| <!doctype html><html lang=en class=no-js><head><meta charset=utf-8><meta http-equiv=x-ua-compatible content="IE=edge"><meta name=viewport content="width=device-width,initial-scale=1"><title>Beam Programming Guide</title><meta name=description content="Apache Beam is an open source, unified model and set of language-specific SDKs for defining and executing data processing workflows, and also data ingestion and integration flows, supporting Enterprise Integration Patterns (EIPs) and Domain Specific Languages (DSLs). Dataflow pipelines simplify the mechanics of large-scale batch and streaming data processing and can run on a number of runtimes like Apache Flink, Apache Spark, and Google Cloud Dataflow (a cloud service). Beam also brings DSL in different languages, allowing users to easily implement their data integration processes."><link href="https://fonts.googleapis.com/css?family=Roboto:100,300,400,500,700" rel=stylesheet><link rel=preload href=/scss/main.min.408fddfe3e8a45f87a5a8c9a839d77db667c1c534e5e5cd0d957ffc3dd6c14cf.css as=style><link href=/scss/main.min.408fddfe3e8a45f87a5a8c9a839d77db667c1c534e5e5cd0d957ffc3dd6c14cf.css rel=stylesheet integrity><script src=https://code.jquery.com/jquery-2.2.4.min.js></script><style>.body__contained img{max-width:100%}</style><script type=text/javascript src=/js/bootstrap.min.2979f9a6e32fc42c3e7406339ee9fe76b31d1b52059776a02b4a7fa6a4fd280a.js defer></script> |
| <script type=text/javascript src=/js/language-switch-v2.min.121952b7980b920320ab229551857669209945e39b05ba2b433a565385ca44c6.js defer></script> |
| <script type=text/javascript src=/js/fix-menu.min.039174b67107465f2090a493f91e126f7aa797f29420f9edab8a54d9dd4b3d2d.js defer></script> |
| <script type=text/javascript src=/js/section-nav.min.1405fd5e70fab5f6c54037c269b1d137487d8f3d1b3009032525f6db3fbce991.js defer></script> |
| <script type=text/javascript src=/js/page-nav.min.af231204c9c52c5089d53a4c02739eacbb7f939e3be1c6ffcc212e0ac4dbf879.js defer></script> |
| <script type=text/javascript src=/js/expandable-list.min.75a4526624a3b8898fe7fb9e3428c205b581f8b38c7926922467aef17eac69f2.js defer></script> |
| <script type=text/javascript src=/js/copy-to-clipboard.min.364c06423d7e8993fc42bb4abc38c03195bc8386db26d18774ce775d08d5b18d.js defer></script> |
| <script type=text/javascript src=/js/calendar.min.336664054fa0f52b08bbd4e3c59b5cb6d63dcfb2b4d602839746516b0817446b.js defer></script> |
| <script type=text/javascript src=/js/fix-playground-nested-scroll.min.0283f1037cb1b9d5074c6eaf041292b524a8148a7cdb803d5ccd6d1fc4eb3253.js defer></script> |
| <script type=text/javascript src=/js/anchor-content-jump-fix.min.22d3240f81632e4c11179b9d2aaf37a40da9414333c43aa97344e8b21a7df0e4.js defer></script> |
| <link rel=alternate type=application/rss+xml title="Apache Beam" href=/feed.xml><link rel=canonical href=/documentation/programming-guide/ data-proofer-ignore><link rel="shortcut icon" type=image/x-icon href=/images/favicon.ico><link rel=stylesheet href=https://use.fontawesome.com/releases/v5.4.1/css/all.css integrity=sha384-5sAR7xN1Nv6T6+dT2mhtzEpVJvfS3NScPQTrOxhwjIuvcA67KV2R5Jz6kr4abQsz crossorigin=anonymous><link rel=stylesheet href=https://unpkg.com/swiper@8/swiper-bundle.min.css><script async src=https://platform.twitter.com/widgets.js></script> |
| <script>(function(e,t,n,s,o,i,a){e.GoogleAnalyticsObject=o,e[o]=e[o]||function(){(e[o].q=e[o].q||[]).push(arguments)},e[o].l=1*new Date,i=t.createElement(n),a=t.getElementsByTagName(n)[0],i.async=1,i.src=s,a.parentNode.insertBefore(i,a)})(window,document,"script","//www.google-analytics.com/analytics.js","ga"),ga("create","UA-73650088-1","auto"),ga("send","pageview")</script><script>(function(e,t,n,s,o,i){e.hj=e.hj||function(){(e.hj.q=e.hj.q||[]).push(arguments)},e._hjSettings={hjid:2182187,hjsv:6},o=t.getElementsByTagName("head")[0],i=t.createElement("script"),i.async=1,i.src=n+e._hjSettings.hjid+s+e._hjSettings.hjsv,o.appendChild(i)})(window,document,"https://static.hotjar.com/c/hotjar-",".js?sv=")</script></head><body class=body data-spy=scroll data-target=.page-nav data-offset=0><nav class="navigation-bar-mobile header navbar navbar-fixed-top"><div class=navbar-header><a href=/ class=navbar-brand><img alt=Brand style=height:46px;width:43px src=/images/beam_logo_navbar_mobile.png></a> |
| <a class=navbar-link href=/get-started/>Get Started</a> |
| <a class=navbar-link href=/documentation/>Documentation</a> |
| <button type=button class="navbar-toggle menu-open" aria-expanded=false aria-controls=navbar onclick=openMenu()> |
| <span class=sr-only>Toggle navigation</span> |
| <span class=icon-bar></span> |
| <span class=icon-bar></span> |
| <span class=icon-bar></span></button></div><div class="navbar-mask closed"></div><div id=navbar class="navbar-container closed"><button type=button class=navbar-toggle aria-expanded=false aria-controls=navbar id=closeMenu> |
| <span class=sr-only>Toggle navigation</span> |
| <span class=icon-bar></span> |
| <span class=icon-bar></span> |
| <span class=icon-bar></span></button><ul class="nav navbar-nav"><li><div class=searchBar-mobile><script>(function(){var t,n="012923275103528129024:4emlchv9wzi",e=document.createElement("script");e.type="text/javascript",e.async=!0,e.src="https://cse.google.com/cse.js?cx="+n,t=document.getElementsByTagName("script")[0],t.parentNode.insertBefore(e,t)})()</script><gcse:search></gcse:search></div></li><li><a class=navbar-link href=/about>About</a></li><li><a class=navbar-link href=/get-started/>Get Started</a></li><li><span class=navbar-link>Documentation</span><ul><li><a href=/documentation/>General</a></li><li><a href=/documentation/sdks/java/>Languages</a></li><li><a href=/documentation/runners/capability-matrix/>Runners</a></li><li><a href=/documentation/io/connectors/>I/O Connectors</a></li></ul></li><li><a class=navbar-link href=/roadmap/>Roadmap</a></li><li><a class=navbar-link href=/community/>Community</a></li><li><a class=navbar-link href=/contribute/>Contribute</a></li><li><a class=navbar-link href=/blog/>Blog</a></li><li><a class=navbar-link href=/case-studies/>Case Studies</a></li></ul><ul class="nav navbar-nav navbar-right"><li><a href=https://github.com/apache/beam/edit/master/website/www/site/content/en/documentation/programming-guide.md data-proofer-ignore><svg xmlns="http://www.w3.org/2000/svg" width="25" height="24" fill="none" viewBox="0 0 25 24"><path stroke="#ff6d00" stroke-linecap="round" stroke-linejoin="round" stroke-width="2.75" d="M4.543 20h4l10.5-10.5c.53-.53.828-1.25.828-2s-.298-1.47-.828-2-1.25-.828-2-.828-1.47.298-2 .828L4.543 16v4zm9.5-13.5 4 4"/></svg></a></li><li class=dropdown><a href=# class=dropdown-toggle id=apache-dropdown data-toggle=dropdown role=button aria-haspopup=true aria-expanded=false><img src=https://www.apache.org/foundation/press/kit/feather_small.png alt="Apache Logo" style=height:20px> |
| Apache |
| <span class=arrow-icon><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" fill="none" viewBox="0 0 20 20"><circle cx="10" cy="10" r="10" fill="#ff6d00"/><path stroke="#fff" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M8.535 5.28l4.573 4.818-4.573 4.403"/></svg></span></a><ul class="dropdown-menu dropdown-menu-right"><li><a target=_blank href=https://www.apache.org/>ASF Homepage</a></li><li><a target=_blank href=https://www.apache.org/licenses/>License</a></li><li><a target=_blank href=https://www.apache.org/security/>Security</a></li><li><a target=_blank href=https://www.apache.org/foundation/thanks.html>Thanks</a></li><li><a target=_blank href=https://www.apache.org/foundation/sponsorship.html>Sponsorship</a></li><li><a target=_blank href=https://www.apache.org/foundation/policies/conduct>Code of Conduct</a></li></ul></li></ul></div></nav><nav class=navigation-bar-desktop><a href=/ class=navbar-logo><img src=/images/beam_logo_navbar.png alt="Beam Logo"></a><div class=navbar-bar-left><div class=navbar-links><a class=navbar-link href=/about>About</a> |
| <a class=navbar-link href=/get-started/>Get Started</a><li class="dropdown navbar-dropdown navbar-dropdown-documentation"><a href=# class="dropdown-toggle navbar-link" role=button aria-haspopup=true aria-expanded=false>Documentation |
| <span><svg xmlns="http://www.w3.org/2000/svg" width="12" height="11" fill="none" viewBox="0 0 12 11"><path stroke="#ff6d00" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M10.666 4.535 5.847 9.108 1.444 4.535"/></svg></span></a><ul class=dropdown-menu><li><a class=navbar-dropdown-menu-link href=/documentation/>General</a></li><li><a class=navbar-dropdown-menu-link href=/documentation/sdks/java/>Languages</a></li><li><a class=navbar-dropdown-menu-link href=/documentation/runners/capability-matrix/>Runners</a></li><li><a class=navbar-dropdown-menu-link href=/documentation/io/connectors/>I/O Connectors</a></li></ul></li><a class=navbar-link href=/roadmap/>Roadmap</a> |
| <a class=navbar-link href=/community/>Community</a> |
| <a class=navbar-link href=/contribute/>Contribute</a> |
| <a class=navbar-link href=/blog/>Blog</a> |
| <a class=navbar-link href=/case-studies/>Case Studies</a></div><div id=iconsBar><a type=button onclick=showSearch()><svg xmlns="http://www.w3.org/2000/svg" width="25" height="24" fill="none" viewBox="0 0 25 24"><path stroke="#ff6d00" stroke-linecap="round" stroke-linejoin="round" stroke-width="2.75" d="M10.191 17c3.866.0 7-3.134 7-7s-3.134-7-7-7-7 3.134-7 7 3.134 7 7 7zm11 4-6-6"/></svg></a><a target=_blank href=https://github.com/apache/beam/edit/master/website/www/site/content/en/documentation/programming-guide.md data-proofer-ignore><svg xmlns="http://www.w3.org/2000/svg" width="25" height="24" fill="none" viewBox="0 0 25 24"><path stroke="#ff6d00" stroke-linecap="round" stroke-linejoin="round" stroke-width="2.75" d="M4.543 20h4l10.5-10.5c.53-.53.828-1.25.828-2s-.298-1.47-.828-2-1.25-.828-2-.828-1.47.298-2 .828L4.543 16v4zm9.5-13.5 4 4"/></svg></a><li class="dropdown navbar-dropdown navbar-dropdown-apache"><a href=# class=dropdown-toggle role=button aria-haspopup=true aria-expanded=false><img src=https://www.apache.org/foundation/press/kit/feather_small.png alt="Apache Logo" style=height:20px> |
| Apache |
| <span class=arrow-icon><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" fill="none" viewBox="0 0 20 20"><circle cx="10" cy="10" r="10" fill="#ff6d00"/><path stroke="#fff" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M8.535 5.28l4.573 4.818-4.573 4.403"/></svg></span></a><ul class=dropdown-menu><li><a class=navbar-dropdown-menu-link target=_blank href=https://www.apache.org/>ASF Homepage</a></li><li><a class=navbar-dropdown-menu-link target=_blank href=https://www.apache.org/licenses/>License</a></li><li><a class=navbar-dropdown-menu-link target=_blank href=https://www.apache.org/security/>Security</a></li><li><a class=navbar-dropdown-menu-link target=_blank href=https://www.apache.org/foundation/thanks.html>Thanks</a></li><li><a class=navbar-dropdown-menu-link target=_blank href=https://www.apache.org/foundation/sponsorship.html>Sponsorship</a></li><li><a class=navbar-dropdown-menu-link target=_blank href=https://www.apache.org/foundation/policies/conduct>Code of Conduct</a></li></ul></li></div><div class="searchBar disappear"><script>(function(){var t,n="012923275103528129024:4emlchv9wzi",e=document.createElement("script");e.type="text/javascript",e.async=!0,e.src="https://cse.google.com/cse.js?cx="+n,t=document.getElementsByTagName("script")[0],t.parentNode.insertBefore(e,t)})()</script><gcse:search></gcse:search> |
| <a type=button onclick=endSearch()><svg xmlns="http://www.w3.org/2000/svg" width="25" height="25" fill="none" viewBox="0 0 25 25"><path stroke="#ff6d00" stroke-linecap="round" stroke-linejoin="round" stroke-width="2.75" d="M21.122 20.827 4.727 4.432M21.122 4.43 4.727 20.827"/></svg></a></div></div></nav><div class=header-push></div><div class="top-banners swiper"><div class=swiper-wrapper><div class=swiper-slide><a href=https://tour.beam.apache.org><img class=banner-img-desktop src=/images/banners/tour-of-beam/tour-of-beam-desktop.png alt="Start Tour of Beam"> |
| <img class=banner-img-mobile src=/images/banners/tour-of-beam/tour-of-beam-mobile.png alt="Start Tour of Beam"></a></div><div class=swiper-slide><a href=https://beam.apache.org/documentation/ml/overview/><img class=banner-img-desktop src=/images/banners/machine-learning/machine-learning-desktop.jpg alt="Machine Learning"> |
| <img class=banner-img-mobile src=/images/banners/machine-learning/machine-learning-mobile.jpg alt="Machine Learning"></a></div></div><div class=swiper-pagination></div><div class=swiper-button-prev></div><div class=swiper-button-next></div></div><script src=/js/swiper-bundle.min.min.e0e8f81b0b15728d35ff73c07f42ddbb17a108d6f23df4953cb3e60df7ade675.js></script> |
| <script src=/js/sliders/top-banners.min.afa7d0a19acf7a3b28ca369490b3d401a619562a2a4c9612577be2f66a4b9855.js></script> |
| <script>function showSearch(){addPlaceholder();var e,t=document.querySelector(".searchBar");t.classList.remove("disappear"),e=document.querySelector("#iconsBar"),e.classList.add("disappear")}function addPlaceholder(){$("input:text").attr("placeholder","What are you looking for?")}function endSearch(){var e,t=document.querySelector(".searchBar");t.classList.add("disappear"),e=document.querySelector("#iconsBar"),e.classList.remove("disappear")}function blockScroll(){$("body").toggleClass("fixedPosition")}function openMenu(){addPlaceholder(),blockScroll()}</script><div class="clearfix container-main-content"><div class="section-nav closed" data-offset-top=90 data-offset-bottom=500><span class="section-nav-back glyphicon glyphicon-menu-left"></span><nav><ul class=section-nav-list data-section-nav><li><span class=section-nav-list-main-title>Documentation</span></li><li><a href=/documentation>Using the Documentation</a></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Concepts</span><ul class=section-nav-list><li><a href=/documentation/basics/>Basics of the Beam model</a></li><li><a href=/documentation/runtime/model/>How Beam executes a pipeline</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Beam programming guide</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/>Overview</a></li><li><a href=/documentation/programming-guide/#creating-a-pipeline>Pipelines</a></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>PCollections</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#pcollections>Creating a PCollection</a></li><li><a href=/documentation/programming-guide/#pcollection-characteristics>PCollection characteristics</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Transforms</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#applying-transforms>Applying transforms</a></li><li><span class=section-nav-list-title>Core Beam transforms</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#pardo>ParDo</a></li><li><a href=/documentation/programming-guide/#groupbykey>GroupByKey</a></li><li><a href=/documentation/programming-guide/#cogroupbykey>CoGroupByKey</a></li><li><a href=/documentation/programming-guide/#combine>Combine</a></li><li><a href=/documentation/programming-guide/#flatten>Flatten</a></li><li><a href=/documentation/programming-guide/#partition>Partition</a></li></ul></li><li><a href=/documentation/programming-guide/#requirements-for-writing-user-code-for-beam-transforms>Requirements for user code</a></li><li><a href=/documentation/programming-guide/#side-inputs>Side inputs</a></li><li><a href=/documentation/programming-guide/#additional-outputs>Additional outputs</a></li><li><a href=/documentation/programming-guide/#composite-transforms>Composite transforms</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Pipeline I/O</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#pipeline-io>Using I/O transforms</a></li><li><a href=/documentation/io/connectors/>I/O connectors</a></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>I/O connector guides</span><ul class=section-nav-list><li><a href=/documentation/io/built-in/parquet/>Apache Parquet I/O connector</a></li><li><a href=/documentation/io/built-in/hadoop/>Hadoop Input/Output Format IO</a></li><li><a href=/documentation/io/built-in/hcatalog/>HCatalog IO</a></li><li><a href=/documentation/io/built-in/google-bigquery/>Google BigQuery I/O connector</a></li><li><a href=/documentation/io/built-in/snowflake/>Snowflake I/O connector</a></li><li><a href=/documentation/io/built-in/cdap/>CDAP I/O connector</a></li><li><a href=/documentation/io/built-in/sparkreceiver/>Spark Receiver I/O connector</a></li><li><a href=/documentation/io/built-in/singlestore/>SingleStoreDB I/O connector</a></li><li><a href=/documentation/io/built-in/webapis/>Web APIs I/O connector</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Developing new I/O connectors</span><ul class=section-nav-list><li><a href=/documentation/io/developing-io-overview/>Overview: Developing connectors</a></li><li><a href=/documentation/io/developing-io-java/>Developing connectors (Java)</a></li><li><a href=/documentation/io/developing-io-python/>Developing connectors (Python)</a></li><li><a href=/documentation/io/io-standards/>I/O Standards</a></li></ul></li><li><a href=/documentation/io/testing/>Testing I/O transforms</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Schemas</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#what-is-a-schema>What is a schema</a></li><li><a href=/documentation/programming-guide/#schemas-for-pl-types>Schemas for programming language types</a></li><li><a href=/documentation/programming-guide/#schema-definition>Schema definition</a></li><li><a href=/documentation/programming-guide/#logical-types>Logical types</a></li><li><a href=/documentation/programming-guide/#creating-schemas>Creating schemas</a></li><li><a href=/documentation/programming-guide/#using-schemas>Using schemas</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Data encoding and type safety</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#data-encoding-and-type-safety>Data encoding basics</a></li><li><a href=/documentation/programming-guide/#specifying-coders>Specifying coders</a></li><li><a href=/documentation/programming-guide/#default-coders-and-the-coderregistry>Default coders and the CoderRegistry</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Windowing</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#windowing>Windowing basics</a></li><li><a href=/documentation/programming-guide/#provided-windowing-functions>Provided windowing functions</a></li><li><a href=/documentation/programming-guide/#setting-your-pcollections-windowing-function>Setting your PCollection’s windowing function</a></li><li><a href=/documentation/programming-guide/#watermarks-and-late-data>Watermarks and late data</a></li><li><a href=/documentation/programming-guide/#adding-timestamps-to-a-pcollections-elements>Adding timestamps to a PCollection’s elements</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Triggers</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#triggers>Trigger basics</a></li><li><a href=/documentation/programming-guide/#event-time-triggers>Event time triggers and the default trigger</a></li><li><a href=/documentation/programming-guide/#processing-time-triggers>Processing time triggers</a></li><li><a href=/documentation/programming-guide/#data-driven-triggers>Data-driven triggers</a></li><li><a href=/documentation/programming-guide/#setting-a-trigger>Setting a trigger</a></li><li><a href=/documentation/programming-guide/#composite-triggers>Composite triggers</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Metrics</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#metrics>Metrics basics</a></li><li><a href=/documentation/programming-guide/#types-of-metrics>Types of metrics</a></li><li><a href=/documentation/programming-guide/#querying-metrics>Querying metrics</a></li><li><a href=/documentation/programming-guide/#using-metrics>Using metrics in pipeline</a></li><li><a href=/documentation/programming-guide/#export-metrics>Export metrics</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>State and Timers</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#types-of-state>Types of state</a></li><li><a href=/documentation/programming-guide/#deferred-state-reads>Deferred state reads</a></li><li><a href=/documentation/programming-guide/#timers>Timers</a></li><li><a href=/documentation/programming-guide/#garbage-collecting-state>Garbage collecting state</a></li><li><a href=/documentation/programming-guide/#state-timers-examples>State and timers examples</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Splittable DoFns</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#sdf-basics>Basics</a></li><li><a href=/documentation/programming-guide/#sizing-and-progress>Sizing and progress</a></li><li><a href=/documentation/programming-guide/#user-initiated-checkpoint>User-initiated checkpoint</a></li><li><a href=/documentation/programming-guide/#runner-initiated-split>Runner initiated split</a></li><li><a href=/documentation/programming-guide/#watermark-estimation>Watermark estimation</a></li><li><a href=/documentation/programming-guide/#truncating-during-drain>Truncating during drain</a></li><li><a href=/documentation/programming-guide/#bundle-finalization>Bundle finalization</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Multi-language Pipelines</span><ul class=section-nav-list><li><a href=/documentation/programming-guide/#create-x-lang-transforms>Creating cross-language transforms</a></li><li><a href=/documentation/programming-guide/#use-x-lang-transforms>Using cross-language transforms</a></li><li><a href=/documentation/programming-guide/#x-lang-transform-runner-support>Runner Support</a></li></ul></li><li><a href=/documentation/programming-guide/#batched-dofns>Batched DoFns</a></li><li><a href=/documentation/programming-guide/#transform-service>Transform service</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Pipeline development lifecycle</span><ul class=section-nav-list><li><a href=/documentation/pipelines/design-your-pipeline/>Design Your Pipeline</a></li><li><a href=/documentation/pipelines/create-your-pipeline/>Create Your Pipeline</a></li><li><a href=/documentation/pipelines/test-your-pipeline/>Test Your Pipeline</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Common pipeline patterns</span><ul class=section-nav-list><li><a href=/documentation/patterns/overview/>Overview</a></li><li><a href=/documentation/patterns/file-processing/>File processing</a></li><li><a href=/documentation/patterns/side-inputs/>Side inputs</a></li><li><a href=/documentation/patterns/pipeline-options/>Pipeline options</a></li><li><a href=/documentation/patterns/custom-io/>Custom I/O</a></li><li><a href=/documentation/patterns/custom-windows/>Custom windows</a></li><li><a href=/documentation/patterns/bigqueryio/>BigQueryIO</a></li><li><a href=/documentation/patterns/ai-platform/>AI Platform</a></li><li><a href=/documentation/patterns/schema/>Schema</a></li><li><a href=/documentation/patterns/bqml/>BigQuery ML</a></li><li><a href=/documentation/patterns/grouping-elements-for-efficient-external-service-calls/>Grouping elements for efficient external service calls</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>AI/ML pipelines</span><ul class=section-nav-list><li><a href=/documentation/ml/overview/>Get started with AI/ML</a></li><li><a href=/documentation/ml/about-ml/>About Beam ML</a></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Prediction and inference</span><ul class=section-nav-list><li><a href=/documentation/ml/inference-overview/>Overview</a></li><li><a href=/documentation/ml/multi-model-pipelines/>Build a pipeline with multiple models</a></li><li><a href=/documentation/ml/tensorrt-runinference>Build a custom model handler with TensorRT</a></li><li><a href=/documentation/ml/large-language-modeling>Use LLM inference</a></li><li><a href=/documentation/ml/multi-language-inference/>Build a multi-language inference pipeline</a></li><li><a href=/documentation/ml/side-input-updates/>Update your model in production</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Data processing</span><ul class=section-nav-list><li><a href=/documentation/ml/preprocess-data/>Preprocess data</a></li><li><a href=/documentation/ml/data-processing/>Explore your data</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Workflow orchestration</span><ul class=section-nav-list><li><a href=/documentation/ml/orchestration/>Use ML-OPS workflow orchestrators</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Model training</span><ul class=section-nav-list><li><a href=/documentation/ml/per-entity-training>Per-entity training</a></li><li><a href=/documentation/ml/online-clustering/>Online clustering</a></li><li><a href=/documentation/ml/model-evaluation/>ML model evaluation</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Use cases</span><ul class=section-nav-list><li><a href=/documentation/ml/anomaly-detection/>Build an anomaly detection pipeline</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Reference</span><ul class=section-nav-list><li><a href=/documentation/ml/runinference-metrics/>RunInference metrics</a></li><li><a href=/documentation/ml/model-evaluation/>Model validation</a></li></ul></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Runtime systems</span><ul class=section-nav-list><li><a href=/documentation/runtime/environments/>Container environments</a></li><li><a href=/documentation/runtime/resource-hints/>Resource hints</a></li><li><a href=/documentation/runtime/sdk-harness-config/>SDK Harness Configuration</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Transform catalog</span><ul class=section-nav-list><li class=section-nav-item--collapsible><span class=section-nav-list-title>Python</span><ul class=section-nav-list><li><a href=/documentation/transforms/python/overview/>Overview</a></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Element-wise</span><ul class=section-nav-list><li class=section-nav-item--collapsible><span class=section-nav-list-title>Enrichment</span><ul class=section-nav-list><li><a href=/documentation/transforms/python/elementwise/enrichment/>Overview</a></li><li><a href=/documentation/transforms/python/elementwise/enrichment-bigtable/>Bigtable example</a></li><li><a href=/documentation/transforms/python/elementwise/enrichment-vertexai/>Vertex AI Feature Store examples</a></li></ul></li><li><a href=/documentation/transforms/python/elementwise/filter/>Filter</a></li><li><a href=/documentation/transforms/python/elementwise/flatmap/>FlatMap</a></li><li><a href=/documentation/transforms/python/elementwise/keys/>Keys</a></li><li><a href=/documentation/transforms/python/elementwise/kvswap/>KvSwap</a></li><li><a href=/documentation/transforms/python/elementwise/map/>Map</a></li><li><a href=/documentation/transforms/python/elementwise/mltransform/>MLTransform</a></li><li><a href=/documentation/transforms/python/elementwise/pardo/>ParDo</a></li><li><a href=/documentation/transforms/python/elementwise/partition/>Partition</a></li><li><a href=/documentation/transforms/python/elementwise/regex/>Regex</a></li><li><a href=/documentation/transforms/python/elementwise/reify/>Reify</a></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>RunInference</span><ul class=section-nav-list><li><a href=/documentation/transforms/python/elementwise/runinference/>Overview</a></li><li><a href=/documentation/transforms/python/elementwise/runinference-pytorch/>PyTorch examples</a></li><li><a href=/documentation/transforms/python/elementwise/runinference-sklearn/>Sklearn examples</a></li></ul></li><li><a href=/documentation/transforms/python/elementwise/tostring/>ToString</a></li><li><a href=/documentation/transforms/python/elementwise/values/>Values</a></li><li><a href=/documentation/transforms/python/elementwise/withtimestamps/>WithTimestamps</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Aggregation</span><ul class=section-nav-list><li><a href=/documentation/transforms/python/aggregation/approximatequantiles/>ApproximateQuantiles</a></li><li><a href=/documentation/transforms/python/aggregation/approximateunique/>ApproximateUnique</a></li><li><a href=/documentation/transforms/python/aggregation/cogroupbykey/>CoGroupByKey</a></li><li><a href=/documentation/transforms/python/aggregation/combineglobally/>CombineGlobally</a></li><li><a href=/documentation/transforms/python/aggregation/combineperkey/>CombinePerKey</a></li><li><a href=/documentation/transforms/python/aggregation/combinevalues/>CombineValues</a></li><li><a href=/documentation/transforms/python/aggregation/count/>Count</a></li><li><a href=/documentation/transforms/python/aggregation/distinct/>Distinct</a></li><li><a href=/documentation/transforms/python/aggregation/groupby/>GroupBy</a></li><li><a href=/documentation/transforms/python/aggregation/groupbykey/>GroupByKey</a></li><li><a href=/documentation/transforms/python/aggregation/groupintobatches/>GroupIntoBatches</a></li><li><a href=/documentation/transforms/python/aggregation/latest/>Latest</a></li><li><a href=/documentation/transforms/python/aggregation/max/>Max</a></li><li><a href=/documentation/transforms/python/aggregation/mean/>Mean</a></li><li><a href=/documentation/transforms/python/aggregation/min/>Min</a></li><li><a href=/documentation/transforms/python/aggregation/sample/>Sample</a></li><li><a href=/documentation/transforms/python/aggregation/sum/>Sum</a></li><li><a href=/documentation/transforms/python/aggregation/top/>Top</a></li><li><a href=/documentation/transforms/python/aggregation/tolist/>ToList</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Other</span><ul class=section-nav-list><li><a href=/documentation/transforms/python/other/create/>Create</a></li><li><a href=/documentation/transforms/python/other/flatten/>Flatten</a></li><li><a href=/documentation/transforms/python/other/reshuffle/>Reshuffle</a></li><li><a href=/documentation/transforms/python/other/windowinto/>WindowInto</a></li></ul></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Java</span><ul class=section-nav-list><li><a href=/documentation/transforms/java/overview/>Overview</a></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Element-wise</span><ul class=section-nav-list><li><a href=/documentation/transforms/java/elementwise/filter/>Filter</a></li><li><a href=/documentation/transforms/java/elementwise/flatmapelements/>FlatMapElements</a></li><li><a href=/documentation/transforms/java/elementwise/keys/>Keys</a></li><li><a href=/documentation/transforms/java/elementwise/kvswap/>KvSwap</a></li><li><a href=/documentation/transforms/java/elementwise/mapelements/>MapElements</a></li><li><a href=/documentation/transforms/java/elementwise/pardo/>ParDo</a></li><li><a href=/documentation/transforms/java/elementwise/partition/>Partition</a></li><li><a href=/documentation/transforms/java/elementwise/regex/>Regex</a></li><li><a href=/documentation/transforms/java/elementwise/reify/>Reify</a></li><li><a href=/documentation/transforms/java/elementwise/tostring/>ToString</a></li><li><a href=/documentation/transforms/java/elementwise/values/>Values</a></li><li><a href=/documentation/transforms/java/elementwise/withkeys/>WithKeys</a></li><li><a href=/documentation/transforms/java/elementwise/withtimestamps/>WithTimestamps</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Aggregation</span><ul class=section-nav-list><li><a href=/documentation/transforms/java/aggregation/approximatequantiles/>ApproximateQuantiles</a></li><li><a href=/documentation/transforms/java/aggregation/approximateunique/>ApproximateUnique</a></li><li><a href=/documentation/transforms/java/aggregation/cogroupbykey/>CoGroupByKey</a></li><li><a href=/documentation/transforms/java/aggregation/combine/>Combine</a></li><li><a href=/documentation/transforms/java/aggregation/combinewithcontext/>CombineWithContext</a></li><li><a href=/documentation/transforms/java/aggregation/count/>Count</a></li><li><a href=/documentation/transforms/java/aggregation/distinct/>Distinct</a></li><li><a href=/documentation/transforms/java/aggregation/groupbykey/>GroupByKey</a></li><li><a href=/documentation/transforms/java/aggregation/groupintobatches/>GroupIntoBatches</a></li><li><a href=/documentation/transforms/java/aggregation/hllcount/>HllCount</a></li><li><a href=/documentation/transforms/java/aggregation/latest/>Latest</a></li><li><a href=/documentation/transforms/java/aggregation/max/>Max</a></li><li><a href=/documentation/transforms/java/aggregation/mean/>Mean</a></li><li><a href=/documentation/transforms/java/aggregation/min/>Min</a></li><li><a href=/documentation/transforms/java/aggregation/sample/>Sample</a></li><li><a href=/documentation/transforms/java/aggregation/sum/>Sum</a></li><li><a href=/documentation/transforms/java/aggregation/top/>Top</a></li></ul></li><li class=section-nav-item--collapsible><span class=section-nav-list-title>Other</span><ul class=section-nav-list><li><a href=/documentation/transforms/java/other/create/>Create</a></li><li><a href=/documentation/transforms/java/other/flatten/>Flatten</a></li><li><a href=/documentation/transforms/java/other/passert/>PAssert</a></li><li><a href=/documentation/transforms/java/other/view/>View</a></li><li><a href=/documentation/transforms/java/other/window/>Window</a></li></ul></li></ul></li></ul></li><li><a href=/documentation/glossary/>Glossary</a></li><li><a href=https://cwiki.apache.org/confluence/display/BEAM/Apache+Beam>Beam Wiki <img src=/images/external-link-icon.png width=14 height=14 alt="External link."></a></li></ul></nav></div><nav class="page-nav clearfix" data-offset-top=90 data-offset-bottom=500><nav id=TableOfContents><ul><li><a href=#overview>1. Overview</a></li><li><a href=#creating-a-pipeline>2. Creating a pipeline</a><ul><li><a href=#configuring-pipeline-options>2.1. Configuring pipeline options</a><ul><li><a href=#pipeline-options-cli>2.1.1. Setting PipelineOptions from command-line arguments</a></li><li><a href=#creating-custom-options>2.1.2. Creating custom options</a></li></ul></li></ul></li><li><a href=#pcollections>3. PCollections</a><ul><li><a href=#creating-a-pcollection>3.1. Creating a PCollection</a><ul><li><a href=#reading-external-source>3.1.1. Reading from an external source</a></li><li><a href=#creating-pcollection-in-memory>3.1.2. Creating a PCollection from in-memory data</a></li></ul></li><li><a href=#pcollection-characteristics>3.2. PCollection characteristics</a><ul><li><a href=#element-type>3.2.1. Element type</a></li><li><a href=#element-schema>3.2.2. Element schema</a></li><li><a href=#immutability>3.2.3. Immutability</a></li><li><a href=#random-access>3.2.4. Random access</a></li><li><a href=#size-and-boundedness>3.2.5. Size and boundedness</a></li><li><a href=#element-timestamps>3.2.6. Element timestamps</a></li></ul></li></ul></li><li><a href=#transforms>4. Transforms</a><ul><li><a href=#applying-transforms>4.1. Applying transforms</a></li><li><a href=#core-beam-transforms>4.2. Core Beam transforms</a><ul><li><a href=#pardo>4.2.1. ParDo</a></li><li><a href=#groupbykey>4.2.2. GroupByKey</a></li><li><a href=#cogroupbykey>4.2.3. CoGroupByKey</a></li><li><a href=#combine>4.2.4. Combine</a></li><li><a href=#flatten>4.2.5. Flatten</a></li><li><a href=#partition>4.2.6. Partition</a></li></ul></li><li><a href=#requirements-for-writing-user-code-for-beam-transforms>4.3. Requirements for writing user code for Beam transforms</a><ul><li><a href=#user-code-serializability>4.3.1. Serializability</a></li><li><a href=#user-code-thread-compatibility>4.3.2. Thread-compatibility</a></li><li><a href=#user-code-idempotence>4.3.3. Idempotence</a></li></ul></li><li><a href=#side-inputs>4.4. Side inputs</a><ul><li><a href=#side-inputs-pardo>4.4.1. Passing side inputs to ParDo</a></li><li><a href=#side-inputs-windowing>4.4.2. Side inputs and windowing</a></li></ul></li><li><a href=#additional-outputs>4.5. Additional outputs</a><ul><li><a href=#output-tags>4.5.1. Tags for multiple outputs</a></li><li><a href=#multiple-outputs-dofn>4.5.2. Emitting to multiple outputs in your DoFn</a></li><li><a href=#other-dofn-parameters>4.5.3. Accessing additional parameters in your DoFn</a></li></ul></li><li><a href=#composite-transforms>4.6. Composite transforms</a><ul><li><a href=#composite-transform-example>4.6.1. An example composite transform</a></li><li><a href=#composite-transform-creation>4.6.2. Creating a composite transform</a></li><li><a href=#ptransform-style-guide>4.6.3. PTransform Style Guide</a></li></ul></li></ul></li><li><a href=#pipeline-io>5. Pipeline I/O</a><ul><li><a href=#pipeline-io-reading-data>5.1. Reading input data</a></li><li><a href=#pipeline-io-writing-data>5.2. Writing output data</a></li><li><a href=#file-based-data>5.3. File-based input and output data</a><ul><li><a href=#file-based-reading-multiple-locations>5.3.1. Reading from multiple locations</a></li><li><a href=#file-based-writing-multiple-files>5.3.2. Writing to multiple output files</a></li></ul></li><li><a href=#provided-io-transforms>5.4. Beam-provided I/O transforms</a></li></ul></li><li><a href=#schemas>6. Schemas</a><ul><li><a href=#what-is-a-schema>6.1. What is a schema?</a></li><li><a href=#schemas-for-pl-types>6.2. Schemas for programming language types</a></li><li><a href=#schema-definition>6.3. Schema definition</a></li><li><a href=#logical-types>6.4. Logical types</a><ul><li><a href=#defining-a-logical-type>6.4.1. Defining a logical type</a></li><li><a href=#built-in-logical-types>6.4.2. Useful logical types</a></li></ul></li><li><a href=#creating-schemas>6.5. Creating Schemas</a><ul><li><a href=#inferring-schemas>6.5.1. Inferring schemas</a></li></ul></li><li><a href=#using-schemas>6.6. Using Schema Transforms</a><ul><li><a href=#661-field-selection-syntax>6.6.1. Field selection syntax</a></li><li><a href=#662-schema-transforms>6.6.2. Schema transforms</a></li><li><a href=#663-schemas-in-pardo>6.6.3. Schemas in ParDo</a></li></ul></li></ul></li><li><a href=#data-encoding-and-type-safety>7. Data encoding and type safety</a><ul><li><a href=#specifying-coders>7.1. Specifying coders</a></li><li><a href=#default-coders-and-the-coderregistry>7.2. Default coders and the CoderRegistry</a><ul><li><a href=#default-coder-lookup>7.2.1. Looking up a default coder</a></li><li><a href=#setting-default-coder>7.2.2. Setting the default coder for a type</a></li><li><a href=#annotating-custom-type-default-coder>7.2.3. Annotating a custom data type with a default coder</a></li></ul></li></ul></li><li><a href=#windowing>8. Windowing</a><ul><li><a href=#windowing-basics>8.1. Windowing basics</a><ul><li><a href=#windowing-constraints>8.1.1. Windowing constraints</a></li><li><a href=#windowing-bounded-collections>8.1.2. Windowing with bounded PCollections</a></li></ul></li><li><a href=#provided-windowing-functions>8.2. Provided windowing functions</a><ul><li><a href=#fixed-time-windows>8.2.1. Fixed time windows</a></li><li><a href=#sliding-time-windows>8.2.2. Sliding time windows</a></li><li><a href=#session-windows>8.2.3. Session windows</a></li><li><a href=#single-global-window>8.2.4. The single global window</a></li></ul></li><li><a href=#setting-your-pcollections-windowing-function>8.3. Setting your PCollection’s windowing function</a><ul><li><a href=#using-fixed-time-windows>8.3.1. Fixed-time windows</a></li><li><a href=#using-sliding-time-windows>8.3.2. Sliding time windows</a></li><li><a href=#using-session-windows>8.3.3. Session windows</a></li><li><a href=#using-single-global-window>8.3.4. Single global window</a></li></ul></li><li><a href=#watermarks-and-late-data>8.4. Watermarks and late data</a><ul><li><a href=#managing-late-data>8.4.1. Managing late data</a></li></ul></li><li><a href=#adding-timestamps-to-a-pcollections-elements>8.5. Adding timestamps to a PCollection’s elements</a></li></ul></li><li><a href=#triggers>9. Triggers</a><ul><li><a href=#event-time-triggers>9.1. Event time triggers</a><ul><li><a href=#default-trigger>9.1.1. Default trigger</a></li></ul></li><li><a href=#processing-time-triggers>9.2. Processing time triggers</a></li><li><a href=#data-driven-triggers>9.3. Data-driven triggers</a></li><li><a href=#setting-a-trigger>9.4. Setting a trigger</a><ul><li><a href=#window-accumulation-modes>9.4.1. Window accumulation modes</a></li><li><a href=#handling-late-data>9.4.2. Handling late data</a></li></ul></li><li><a href=#composite-triggers>9.5. Composite triggers</a><ul><li><a href=#composite-trigger-types>9.5.1. Composite trigger types</a></li><li><a href=#composite-afterwatermark>9.5.2. Composition with AfterWatermark</a></li><li><a href=#other-composite-triggers>9.5.3. Other composite triggers</a></li></ul></li></ul></li><li><a href=#metrics>10. Metrics</a><ul><li><a href=#101-the-main-concepts-of-beam-metrics>10.1. The main concepts of Beam metrics</a></li><li><a href=#types-of-metrics>10.2. Types of metrics</a></li><li><a href=#querying-metrics>10.3. Querying metrics</a></li><li><a href=#using-metrics>10.4. Using metrics in pipeline</a></li><li><a href=#export-metrics>10.5. Export metrics</a></li></ul></li><li><a href=#state-and-timers>11. State and Timers</a><ul><li><a href=#types-of-state>11.1. Types of state</a><ul><li><a href=#valuestate>ValueState</a></li><li><a href=#combiningstate>CombiningState</a></li><li><a href=#bagstate>BagState</a></li></ul></li><li><a href=#deferred-state-reads>11.2. Deferred state reads</a></li><li><a href=#timers>11.3. Timers</a><ul><li><a href=#event-time-timers>11.3.1. Event-time timers</a></li><li><a href=#processing-time-timers>11.3.2. Processing-time timers</a></li><li><a href=#dynamic-timer-tags>11.3.3. Dynamic timer tags</a></li><li><a href=#timer-output-timestamps>11.3.4. Timer output timestamps</a></li></ul></li><li><a href=#garbage-collecting-state>11.4. Garbage collecting state</a><ul><li></li></ul></li><li><a href=#state-timers-examples>11.5. State and timers examples</a><ul><li><a href=#joining-clicks-and-views>11.5.1. Joining clicks and views</a></li><li><a href=#batching-rpcs>11.5.2. Batching RPCs</a></li></ul></li></ul></li><li><a href=#splittable-dofns>12. Splittable <code>DoFns</code></a><ul><li><a href=#sdf-basics>12.1. SDF basics</a><ul><li><a href=#a-basic-sdf>12.1.1. A basic SDF</a></li></ul></li><li><a href=#sizing-and-progress>12.2. Sizing and progress</a></li><li><a href=#user-initiated-checkpoint>12.3. User-initiated checkpoint</a></li><li><a href=#runner-initiated-split>12.4. Runner-initiated split</a></li><li><a href=#watermark-estimation>12.5. Watermark estimation</a><ul><li><a href=#controlling-the-watermark>12.5.1. Controlling the watermark</a></li></ul></li><li><a href=#truncating-during-drain>12.6. Truncating during drain</a></li><li><a href=#bundle-finalization>12.7. Bundle finalization</a></li></ul></li><li><a href=#multi-language-pipelines>13. Multi-language pipelines</a><ul><li><a href=#create-x-lang-transforms>13.1. Creating cross-language transforms</a><ul><li><a href=#1311-creating-cross-language-java-transforms>13.1.1. Creating cross-language Java transforms</a></li><li><a href=#1312-creating-cross-language-python-transforms>13.1.2. Creating cross-language Python transforms</a></li><li><a href=#1313-creating-cross-language-go-transforms>13.1.3. Creating cross-language Go transforms</a></li><li><a href=#1314-defining-a-urn>13.1.4. Defining a URN</a></li></ul></li><li><a href=#use-x-lang-transforms>13.2. Using cross-language transforms</a><ul><li><a href=#1321-using-cross-language-transforms-in-a-java-pipeline>13.2.1. Using cross-language transforms in a Java pipeline</a></li><li><a href=#1322-using-cross-language-transforms-in-a-python-pipeline>13.2.2. Using cross-language transforms in a Python pipeline</a></li><li><a href=#1323-using-cross-language-transforms-in-a-go-pipeline>13.2.3. Using cross-language transforms in a Go pipeline</a></li><li><a href=#1324-using-cross-language-transforms-in-a-typescript-pipeline>13.2.4. Using cross-language transforms in a Typescript pipeline</a></li></ul></li><li><a href=#x-lang-transform-runner-support>13.3. Runner Support</a></li><li><a href=#x-lang-transform-tips-troubleshooting>13.4 Tips and Troubleshooting</a></li></ul></li><li><a href=#batched-dofns>14 Batched DoFns</a><ul><li><a href=#batched-dofn-basics>14.1 Basics</a></li><li><a href=#batched-dofn-elementwise>14.2 Element-wise Fallback</a></li><li><a href=#batched-dofn-batch-production>14.3 Batch Production vs. Batch Consumption</a></li><li><a href=#batched-dofn-types>14.4 Supported Batch Types</a><ul><li><a href=#numpyhttpsgithubcomapachebeamblobmastersdkspythonapache_beamtypehintsbatchpy><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/typehints/batch.py>numpy</a></a></li><li><a href=#pandashttpsgithubcomapachebeamblobmastersdkspythonapache_beamtypehintspandas_type_compatibilitypy><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/typehints/pandas_type_compatibility.py>pandas</a></a></li><li><a href=#pyarrowhttpsgithubcomapachebeamblobmastersdkspythonapache_beamtypehintsarrow_type_compatibilitypy><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/typehints/arrow_type_compatibility.py>pyarrow</a></a></li><li><a href=#other-types>Other types?</a></li></ul></li><li><a href=#batched-dofn-dynamic-types>14.5 Dynamic Batch Input and Output Types</a></li><li><a href=#batched-dofn-event-time>14.6 Batches and Event-time Semantics</a></li></ul></li><li><a href=#transform-service>15 Transform service</a><ul><li><a href=#transform-service-usage-upgrade>15.1 Using the transform service to upgrade transforms</a></li><li><a href=#transform-service-usage-multi-language>15.2 Using the Transform service for multi-language pipelines</a></li><li><a href=#transform-service-usage-muanual>15.3 Manually starting the transform service</a></li><li><a href=#transform-service-included-transforms>15.4 Portable transforms included in the Transform service</a></li></ul></li></ul></nav></nav><div class="body__contained body__section-nav arrow-list arrow-list--no-mt"><h1 id=apache-beam-programming-guide>Apache Beam Programming Guide</h1><p>The <strong>Beam Programming Guide</strong> is intended for Beam users who want to use the |
| Beam SDKs to create data processing pipelines. It provides guidance for using |
| the Beam SDK classes to build and test your pipeline. The programming guide is |
| not intended as an exhaustive reference, but as a language-agnostic, high-level |
| guide to programmatically building your Beam pipeline. As the programming guide |
| is filled out, the text will include code samples in multiple languages to help |
| illustrate how to implement Beam concepts in your pipelines.</p><p>If you want a brief introduction to Beam’s basic concepts before reading the |
| programming guide, take a look at the |
| <a href=/documentation/basics/>Basics of the Beam model</a> page.</p><nav class=language-switcher><strong>Adapt for:</strong><ul><li data-value=java class=active>Java SDK</li><li data-value=py>Python SDK</li><li data-value=go>Go SDK</li><li data-value=typescript>TypeScript SDK</li><li data-value=yaml>Yaml API</li></ul></nav><p class=language-py>The Python SDK supports Python 3.8, 3.9, 3.10, and 3.11.</p><p class=language-go>The <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam>Go SDK</a> supports Go v1.20+.</p><p class=language-typescript>The Typescript SDK supports Node v16+ and is still experimental.</p><p class=language-yaml>YAML is supported as of Beam 2.52, but is under active development and the most |
| recent SDK is advised.</p><h2 id=overview>1. Overview</h2><p>To use Beam, you need to first create a driver program using the classes in one |
| of the Beam SDKs. Your driver program <em>defines</em> your pipeline, including all of |
| the inputs, transforms, and outputs; it also sets execution options for your |
| pipeline (typically passed in using command-line options). These include the |
| Pipeline Runner, which, in turn, determines what back-end your pipeline will run |
| on.</p><p>The Beam SDKs provide a number of abstractions that simplify the mechanics of |
| large-scale distributed data processing. The same Beam abstractions work with |
| both batch and streaming data sources. When you create your Beam pipeline, you |
| can think about your data processing task in terms of these abstractions. They |
| include:</p><ul><li><p><code>Pipeline</code>: A <code>Pipeline</code> encapsulates your entire data processing task, from |
| start to finish. This includes reading input data, transforming that data, and |
| writing output data. All Beam driver programs must create a <code>Pipeline</code>. When |
| you create the <code>Pipeline</code>, you must also specify the execution options that |
| tell the <code>Pipeline</code> where and how to run.</p></li><li><p><code>PCollection</code>: A <code>PCollection</code> represents a distributed data set that your |
| Beam pipeline operates on. The data set can be <em>bounded</em>, meaning it comes |
| from a fixed source like a file, or <em>unbounded</em>, meaning it comes from a |
| continuously updating source via a subscription or other mechanism. Your |
| pipeline typically creates an initial <code>PCollection</code> by reading data from an |
| external data source, but you can also create a <code>PCollection</code> from in-memory |
| data within your driver program. From there, <code>PCollection</code>s are the inputs and |
| outputs for each step in your pipeline.</p></li><li><p><code>PTransform</code>: A <code>PTransform</code> represents a data processing operation, or a step, |
| in your pipeline. Every <code>PTransform</code> takes one or more <code>PCollection</code> objects as |
| input, performs a processing function that you provide on the elements of that |
| <code>PCollection</code>, and produces zero or more output <code>PCollection</code> objects.</p></li></ul><span class=language-go><ul><li><code>Scope</code>: The Go SDK has an explicit scope variable used to build a <code>Pipeline</code>. |
| A <code>Pipeline</code> can return it’s root scope with the <code>Root()</code> method. The scope |
| variable is passed to <code>PTransform</code> functions to place them in the <code>Pipeline</code> |
| that owns the <code>Scope</code>.</li></ul></span><ul><li>I/O transforms: Beam comes with a number of “IOs” - library <code>PTransform</code>s that |
| read or write data to various external storage systems.</li></ul><span class=language-yaml>Note that in Beam YAML, `PCollection`s are either implicit (e.g. when using |
| `chain`) or referenced by their producing `PTransform`.</span><p>A typical Beam driver program works as follows:</p><ul><li><strong>Create</strong> a <code>Pipeline</code> object and set the pipeline execution options, including |
| the Pipeline Runner.</li><li>Create an initial <code>PCollection</code> for pipeline data, either using the IOs |
| to read data from an external storage system, or using a <code>Create</code> transform to |
| build a <code>PCollection</code> from in-memory data.</li><li><strong>Apply</strong> <code>PTransform</code>s to each <code>PCollection</code>. Transforms can change, filter, |
| group, analyze, or otherwise process the elements in a <code>PCollection</code>. A |
| transform creates a new output <code>PCollection</code> <em>without modifying the input |
| collection</em>. A typical pipeline applies subsequent transforms to each new |
| output <code>PCollection</code> in turn until processing is complete. However, note that |
| a pipeline does not have to be a single straight line of transforms applied |
| one after another: think of <code>PCollection</code>s as variables and <code>PTransform</code>s as |
| functions applied to these variables: the shape of the pipeline can be an |
| arbitrarily complex processing graph.</li><li>Use IOs to write the final, transformed <code>PCollection</code>(s) to an external source.</li><li><strong>Run</strong> the pipeline using the designated Pipeline Runner.</li></ul><p>When you run your Beam driver program, the Pipeline Runner that you designate |
| constructs a <strong>workflow graph</strong> of your pipeline based on the <code>PCollection</code> |
| objects you’ve created and transforms that you’ve applied. That graph is then |
| executed using the appropriate distributed processing back-end, becoming an |
| asynchronous “job” (or equivalent) on that back-end.</p><h2 id=creating-a-pipeline>2. Creating a pipeline</h2><p>The <code>Pipeline</code> abstraction encapsulates all the data and steps in your data |
| processing task. Your Beam driver program typically starts by constructing a |
| <span class=language-java><a href=https://beam.apache.org/releases/javadoc/2.55.1/index.html?org/apache/beam/sdk/Pipeline.html>Pipeline</a></span> |
| <span class=language-py><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/pipeline.py>Pipeline</a></span> |
| <span class=language-go><a href=https://github.com/apache/beam/blob/master/sdks/go/pkg/beam/pipeline.go#L62>Pipeline</a></span> |
| object, and then using that object as the basis for creating the pipeline’s data |
| sets as <code>PCollection</code>s and its operations as <code>Transform</code>s.</p><p>To use Beam, your driver program must first create an instance of the Beam SDK |
| class <code>Pipeline</code> (typically in the <code>main()</code> function). When you create your |
| <code>Pipeline</code>, you’ll also need to set some <strong>configuration options</strong>. You can set |
| your pipeline’s configuration options programmatically, but it’s often easier to |
| set the options ahead of time (or read them from the command line) and pass them |
| to the <code>Pipeline</code> object when you create the object.</p><span class=language-typescript>A Pipeline in the Typescript API is simply a function that will be called |
| with a single `root` object and is passed to a Runner's `run` method.</span><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Start by defining the options for the pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PipelineOptions</span> <span class=n>options</span> <span class=o>=</span> <span class=n>PipelineOptionsFactory</span><span class=o>.</span><span class=na>create</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Then create the pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>Pipeline</span> <span class=n>p</span> <span class=o>=</span> <span class=n>Pipeline</span><span class=o>.</span><span class=na>create</span><span class=o>(</span><span class=n>options</span><span class=o>);</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>import</span> <span class=nn>apache_beam</span> <span class=k>as</span> <span class=nn>beam</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>with</span> <span class=n>beam</span><span class=o>.</span><span class=n>Pipeline</span><span class=p>()</span> <span class=k>as</span> <span class=n>pipeline</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>pass</span> <span class=c1># build your pipeline here</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// beam.Init() is an initialization hook that must be called |
| </span></span></span><span class=line><span class=cl><span class=c1>// near the beginning of main(), before creating a pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>beam</span><span class=p>.</span><span class=nf>Init</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Create the Pipeline object and root scope. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>pipeline</span><span class=p>,</span> <span class=nx>scope</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>NewPipelineWithRoot</span><span class=p>()</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=k>await</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>createRunner</span><span class=p>().</span><span class=nx>run</span><span class=p>(</span><span class=kd>function</span> <span class=nx>pipeline</span><span class=p>(</span><span class=nx>root</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Use root to build a pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>});</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>pipeline</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>options</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span></span></span></code></pre></div></div></div><p>For a more in-depth tutorial on creating basic pipelines |
| in the Python SDK, please read and work through |
| <a href=https://colab.sandbox.google.com/github/apache/beam/blob/master/examples/notebooks/get-started/learn_beam_basics_by_doing.ipynb>this colab notebook</a>.</p><h3 id=configuring-pipeline-options>2.1. Configuring pipeline options</h3><p>Use the pipeline options to configure different aspects of your pipeline, such |
| as the pipeline runner that will execute your pipeline and any runner-specific |
| configuration required by the chosen runner. Your pipeline options will |
| potentially include information such as your project ID or a location for |
| storing files.</p><p class=language-java>When you run the pipeline on a runner of your choice, a copy of the |
| PipelineOptions will be available to your code. For example, if you add a PipelineOptions parameter |
| to a DoFn’s <code>@ProcessElement</code> method, it will be populated by the system.</p><h4 id=pipeline-options-cli>2.1.1. Setting PipelineOptions from command-line arguments</h4><p class="language-java language-py">While you can configure your pipeline by creating a <code>PipelineOptions</code> object and |
| setting the fields directly, the Beam SDKs include a command-line parser that |
| you can use to set fields in <code>PipelineOptions</code> using command-line arguments.</p><p class="language-java language-py">To read options from the command-line, construct your <code>PipelineOptions</code> object |
| as demonstrated in the following example code:</p><p class=language-go>Use Go flags to parse command line arguments to configure your pipeline. Flags must be parsed |
| before <code>beam.Init()</code> is called.</p><p class=language-typescript>Any Javascript object can be used as pipeline options. |
| One can either construct one manually, but it is also common to pass an object |
| created from command line options such as <code>yargs.argv</code>.</p><p class=language-yaml>Pipeline options are simply an optional YAML mapping property that is a sibling to |
| the pipeline definition itself. |
| It will be merged with whatever options are passed on the command line.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PipelineOptions</span> <span class=n>options</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>PipelineOptionsFactory</span><span class=o>.</span><span class=na>fromArgs</span><span class=o>(</span><span class=n>args</span><span class=o>).</span><span class=na>withValidation</span><span class=o>().</span><span class=na>create</span><span class=o>();</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam.options.pipeline_options</span> <span class=kn>import</span> <span class=n>PipelineOptions</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>beam_options</span> <span class=o>=</span> <span class=n>PipelineOptions</span><span class=p>()</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// If beamx or Go flags are used, flags must be parsed first, |
| </span></span></span><span class=line><span class=cl><span class=c1>// before beam.Init() is called. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>flag</span><span class=p>.</span><span class=nf>Parse</span><span class=p>()</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pipeline_options</span> <span class=o>=</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>runner</span><span class=o>:</span> <span class=s2>"default"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>project</span><span class=o>:</span> <span class=s2>"my_project"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl><span class=p>};</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>runner</span> <span class=o>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>createRunner</span><span class=p>(</span><span class=nx>pipeline_options</span><span class=p>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>runnerFromCommandLineOptions</span> <span class=o>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>createRunner</span><span class=p>(</span><span class=nx>yargs</span><span class=p>.</span><span class=nx>argv</span><span class=p>);</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>pipeline</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>options</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>my_pipeline_option</span><span class=p>:</span><span class=w> </span><span class=l>my_value</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span></span></span></code></pre></div></div></div><p>This interprets command-line arguments that follow the format:</p><pre tabindex=0><code>--<option>=<value> |
| </code></pre><span class=language-java><blockquote><p>Appending the method <code>.withValidation</code> will check for required |
| command-line arguments and validate argument values.</p></blockquote></span><p class="language-java language-py">Building your <code>PipelineOptions</code> this way lets you specify any of the options as |
| a command-line argument.</p><p class=language-go>Defining flag variables this way lets you specify any of the options as a command-line argument.</p><blockquote><p><strong>Note:</strong> The <a href=/get-started/wordcount-example>WordCount example pipeline</a> |
| demonstrates how to set pipeline options at runtime by using command-line |
| options.</p></blockquote><h4 id=creating-custom-options>2.1.2. Creating custom options</h4><p>You can add your own custom options in addition to the standard |
| <code>PipelineOptions</code>.<p class=language-java>To add your own options, define an interface with getter and |
| setter methods for each option.</p>The following example shows how to add <code>input</code> and <code>output</code> custom options:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>interface</span> <span class=nc>MyOptions</span> <span class=kd>extends</span> <span class=n>PipelineOptions</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=nf>getInput</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kt>void</span> <span class=nf>setInput</span><span class=o>(</span><span class=n>String</span> <span class=n>input</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=nf>getOutput</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kt>void</span> <span class=nf>setOutput</span><span class=o>(</span><span class=n>String</span> <span class=n>output</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam.options.pipeline_options</span> <span class=kn>import</span> <span class=n>PipelineOptions</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MyOptions</span><span class=p>(</span><span class=n>PipelineOptions</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@classmethod</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>_add_argparse_args</span><span class=p>(</span><span class=bp>cls</span><span class=p>,</span> <span class=n>parser</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>parser</span><span class=o>.</span><span class=n>add_argument</span><span class=p>(</span><span class=s1>'--input'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>parser</span><span class=o>.</span><span class=n>add_argument</span><span class=p>(</span><span class=s1>'--output'</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// Use standard Go flags to define pipeline options. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>var</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>input</span> <span class=p>=</span> <span class=nx>flag</span><span class=p>.</span><span class=nf>String</span><span class=p>(</span><span class=s>"input"</span><span class=p>,</span> <span class=s>""</span><span class=p>,</span> <span class=s>""</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>output</span> <span class=p>=</span> <span class=nx>flag</span><span class=p>.</span><span class=nf>String</span><span class=p>(</span><span class=s>"output"</span><span class=p>,</span> <span class=s>""</span><span class=p>,</span> <span class=s>""</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>options</span> <span class=o>=</span> <span class=nx>yargs</span><span class=p>.</span><span class=nx>argv</span><span class=p>;</span> <span class=c1>// Or an alternative command-line parsing library. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=c1>// Use options.input and options.output during pipeline construction. |
| </span></span></span></code></pre></div></div></div><p>You can also specify a description, which appears when a user passes <code>--help</code> as |
| a command-line argument, and a default value.</p><p class="language-java language-py language-go">You set the description and default value using annotations, as follows:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>interface</span> <span class=nc>MyOptions</span> <span class=kd>extends</span> <span class=n>PipelineOptions</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Description</span><span class=o>(</span><span class=s>"Input for the pipeline"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Default.String</span><span class=o>(</span><span class=s>"gs://my-bucket/input"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=nf>getInput</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kt>void</span> <span class=nf>setInput</span><span class=o>(</span><span class=n>String</span> <span class=n>input</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Description</span><span class=o>(</span><span class=s>"Output for the pipeline"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Default.String</span><span class=o>(</span><span class=s>"gs://my-bucket/output"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=nf>getOutput</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kt>void</span> <span class=nf>setOutput</span><span class=o>(</span><span class=n>String</span> <span class=n>output</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam.options.pipeline_options</span> <span class=kn>import</span> <span class=n>PipelineOptions</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MyOptions</span><span class=p>(</span><span class=n>PipelineOptions</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@classmethod</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>_add_argparse_args</span><span class=p>(</span><span class=bp>cls</span><span class=p>,</span> <span class=n>parser</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>parser</span><span class=o>.</span><span class=n>add_argument</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'--input'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>default</span><span class=o>=</span><span class=s1>'gs://dataflow-samples/shakespeare/kinglear.txt'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>help</span><span class=o>=</span><span class=s1>'The file path for the input text to process.'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>parser</span><span class=o>.</span><span class=n>add_argument</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'--output'</span><span class=p>,</span> <span class=n>required</span><span class=o>=</span><span class=kc>True</span><span class=p>,</span> <span class=n>help</span><span class=o>=</span><span class=s1>'The path prefix for output files.'</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>var</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>input</span> <span class=p>=</span> <span class=nx>flag</span><span class=p>.</span><span class=nf>String</span><span class=p>(</span><span class=s>"input"</span><span class=p>,</span> <span class=s>"gs://my-bucket/input"</span><span class=p>,</span> <span class=s>"Input for the pipeline"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>output</span> <span class=p>=</span> <span class=nx>flag</span><span class=p>.</span><span class=nf>String</span><span class=p>(</span><span class=s>"output"</span><span class=p>,</span> <span class=s>"gs://my-bucket/output"</span><span class=p>,</span> <span class=s>"Output for the pipeline"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span></span></span></code></pre></div></div></div><p class=language-py>For Python, you can also simply parse your custom options with argparse; there |
| is no need to create a separate PipelineOptions subclass.</p><p class=language-java>It’s recommended that you register your interface with <code>PipelineOptionsFactory</code> |
| and then pass the interface when creating the <code>PipelineOptions</code> object. When you |
| register your interface with <code>PipelineOptionsFactory</code>, the <code>--help</code> can find |
| your custom options interface and add it to the output of the <code>--help</code> command. |
| <code>PipelineOptionsFactory</code> will also validate that your custom options are |
| compatible with all other registered options.</p><p class=language-java>The following example code shows how to register your custom options interface |
| with <code>PipelineOptionsFactory</code>:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PipelineOptionsFactory</span><span class=o>.</span><span class=na>register</span><span class=o>(</span><span class=n>MyOptions</span><span class=o>.</span><span class=na>class</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl><span class=n>MyOptions</span> <span class=n>options</span> <span class=o>=</span> <span class=n>PipelineOptionsFactory</span><span class=o>.</span><span class=na>fromArgs</span><span class=o>(</span><span class=n>args</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withValidation</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>as</span><span class=o>(</span><span class=n>MyOptions</span><span class=o>.</span><span class=na>class</span><span class=o>);</span></span></span></code></pre></div></div></div><p>Now your pipeline can accept <code>--input=value</code> and <code>--output=value</code> as command-line arguments.</p><h2 id=pcollections>3. PCollections</h2><p>The <span class=language-java><a href=https://beam.apache.org/releases/javadoc/2.55.1/index.html?org/apache/beam/sdk/values/PCollection.html>PCollection</a></span> |
| <span class=language-py><code>PCollection</code></span> |
| <span class=language-go><a href=https://github.com/apache/beam/blob/master/sdks/go/pkg/beam/pcollection.go#L39>PCollection</a></span> |
| abstraction represents a |
| potentially distributed, multi-element data set. You can think of a |
| <code>PCollection</code> as “pipeline” data; Beam transforms use <code>PCollection</code> objects as |
| inputs and outputs. As such, if you want to work with data in your pipeline, it |
| must be in the form of a <code>PCollection</code>.</p><p>After you’ve created your <code>Pipeline</code>, you’ll need to begin by creating at least |
| one <code>PCollection</code> in some form. The <code>PCollection</code> you create serves as the input |
| for the first operation in your pipeline.</p><span class=language-yaml>In Beam YAML, `PCollection`s are either implicit (e.g. when using `chain`) |
| or referred to by their producing `PTransform`.</span><h3 id=creating-a-pcollection>3.1. Creating a PCollection</h3><p>You create a <code>PCollection</code> by either reading data from an external source using |
| Beam’s <a href=#pipeline-io>Source API</a>, or you can create a <code>PCollection</code> of data |
| stored in an in-memory collection class in your driver program. The former is |
| typically how a production pipeline would ingest data; Beam’s Source APIs |
| contain adapters to help you read from external sources like large cloud-based |
| files, databases, or subscription services. The latter is primarily useful for |
| testing and debugging purposes.</p><h4 id=reading-external-source>3.1.1. Reading from an external source</h4><p>To read from an external source, you use one of the <a href=#pipeline-io>Beam-provided I/O |
| adapters</a>. The adapters vary in their exact usage, but all of them |
| read from some external data source and return a <code>PCollection</code> whose elements |
| represent the data records in that source.</p><p>Each data source adapter has a <code>Read</code> transform; to read, |
| <span class="language-java language-py language-go language-typescript">you must apply that transform to the <code>Pipeline</code> object itself.</span> |
| <span class=language-yaml>place this transform in the <code>source</code> or <code>transforms</code> portion of the pipeline.</span> |
| <span class=language-java><code>TextIO.Read</code></span> |
| <span class=language-py><code>io.TextFileSource</code></span> |
| <span class=language-go><code>textio.Read</code></span> |
| <span class=language-typescript><code>textio.ReadFromText</code></span>, |
| <span class=language-yaml><code>ReadFromText</code></span>, |
| for example, reads from an |
| external text file and returns a <code>PCollection</code> whose elements |
| <span class="language-java language-py language-go language-typescript">are of type <code>String</code> where each <code>String</code></span> |
| represents one line from the text file. Here’s how you |
| would apply <span class=language-java><code>TextIO.Read</code></span> |
| <span class=language-py><code>io.TextFileSource</code></span> |
| <span class=language-go><code>textio.Read</code></span> |
| <span class=language-typescript><code>textio.ReadFromText</code></span> |
| <span class=language-yaml><code>ReadFromText</code></span> |
| to your <code>Pipeline</code> <span class=language-typescript>root</span> to create |
| a <code>PCollection</code>:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>static</span> <span class=kt>void</span> <span class=nf>main</span><span class=o>(</span><span class=n>String</span><span class=o>[]</span> <span class=n>args</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Create the pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PipelineOptions</span> <span class=n>options</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>PipelineOptionsFactory</span><span class=o>.</span><span class=na>fromArgs</span><span class=o>(</span><span class=n>args</span><span class=o>).</span><span class=na>create</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Pipeline</span> <span class=n>p</span> <span class=o>=</span> <span class=n>Pipeline</span><span class=o>.</span><span class=na>create</span><span class=o>(</span><span class=n>options</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Create the PCollection 'lines' by applying a 'Read' transform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>lines</span> <span class=o>=</span> <span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"ReadMyFile"</span><span class=o>,</span> <span class=n>TextIO</span><span class=o>.</span><span class=na>read</span><span class=o>().</span><span class=na>from</span><span class=o>(</span><span class=s>"gs://some/inputData.txt"</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>lines</span> <span class=o>=</span> <span class=n>pipeline</span> <span class=o>|</span> <span class=s1>'ReadMyFile'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>io</span><span class=o>.</span><span class=n>ReadFromText</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'gs://some/inputData.txt'</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// Read the file at the URI 'gs://some/inputData.txt' and return |
| </span></span></span><span class=line><span class=cl><span class=c1>// the lines as a PCollection<string>. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Notice the scope as the first variable when calling |
| </span></span></span><span class=line><span class=cl><span class=c1>// the method as is needed when calling all transforms. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>lines</span> <span class=o>:=</span> <span class=nx>textio</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=s>"gs://some/inputData.txt"</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>async</span> <span class=kd>function</span> <span class=nx>pipeline</span><span class=p>(</span><span class=nx>root</span>: <span class=kt>beam.Root</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Note that textio.ReadFromText is an AsyncPTransform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kr>const</span> <span class=nx>pcoll</span>: <span class=kt>PCollection</span><span class=p><</span><span class=nt>string</span><span class=p>></span> <span class=o>=</span> <span class=k>await</span> <span class=nx>root</span><span class=p>.</span><span class=nx>applyAsync</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>textio</span><span class=p>.</span><span class=nx>ReadFromText</span><span class=p>(</span><span class=s2>"path/to/text_pattern"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>pipeline</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>source</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>ReadFromText</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>path</span><span class=p>:</span><span class=w> </span><span class=l>...</span></span></span></code></pre></div></div></div><p>See the <a href=#pipeline-io>section on I/O</a> to learn more about how to read from the |
| various data sources supported by the Beam SDK.</p><h4 id=creating-pcollection-in-memory>3.1.2. Creating a PCollection from in-memory data</h4><p class=language-java>To create a <code>PCollection</code> from an in-memory Java <code>Collection</code>, you use the |
| Beam-provided <code>Create</code> transform. Much like a data adapter’s <code>Read</code>, you apply |
| <code>Create</code> directly to your <code>Pipeline</code> object itself.</p><p class=language-java>As parameters, <code>Create</code> accepts the Java <code>Collection</code> and a <code>Coder</code> object. The |
| <code>Coder</code> specifies how the elements in the <code>Collection</code> should be |
| <a href=#element-type>encoded</a>.</p><p class=language-py>To create a <code>PCollection</code> from an in-memory <code>list</code>, you use the Beam-provided |
| <code>Create</code> transform. Apply this transform directly to your <code>Pipeline</code> object |
| itself.</p><p class=language-go>To create a <code>PCollection</code> from an in-memory <code>slice</code>, you use the Beam-provided |
| <code>beam.CreateList</code> transform. Pass the pipeline <code>scope</code>, and the <code>slice</code> to this transform.</p><p class=language-typescript>To create a <code>PCollection</code> from an in-memory <code>array</code>, you use the Beam-provided |
| <code>Create</code> transform. Apply this transform directly to your <code>Root</code> object.</p><p class=language-yaml>To create a <code>PCollection</code> from an in-memory <code>array</code>, you use the Beam-provided |
| <code>Create</code> transform. Specify the elements in the pipeline itself.</p><p>The following example code shows how to create a <code>PCollection</code> from an in-memory |
| <span class=language-java><code>List</code></span> |
| <span class=language-py><code>list</code></span> |
| <span class=language-go><code>slice</code></span> |
| <span class="language-typescript language-yaml"><code>array</code></span>:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>static</span> <span class=kt>void</span> <span class=nf>main</span><span class=o>(</span><span class=n>String</span><span class=o>[]</span> <span class=n>args</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Create a Java Collection, in this case a List of Strings. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>final</span> <span class=n>List</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>LINES</span> <span class=o>=</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"To be, or not to be: that is the question: "</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"Whether 'tis nobler in the mind to suffer "</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"The slings and arrows of outrageous fortune, "</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"Or to take arms against a sea of troubles, "</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Create the pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PipelineOptions</span> <span class=n>options</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>PipelineOptionsFactory</span><span class=o>.</span><span class=na>fromArgs</span><span class=o>(</span><span class=n>args</span><span class=o>).</span><span class=na>create</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Pipeline</span> <span class=n>p</span> <span class=o>=</span> <span class=n>Pipeline</span><span class=o>.</span><span class=na>create</span><span class=o>(</span><span class=n>options</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Apply Create, passing the list and the coder, to create the PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Create</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>LINES</span><span class=o>)).</span><span class=na>setCoder</span><span class=o>(</span><span class=n>StringUtf8Coder</span><span class=o>.</span><span class=na>of</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>import</span> <span class=nn>apache_beam</span> <span class=k>as</span> <span class=nn>beam</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>with</span> <span class=n>beam</span><span class=o>.</span><span class=n>Pipeline</span><span class=p>()</span> <span class=k>as</span> <span class=n>pipeline</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>lines</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>pipeline</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Create</span><span class=p>([</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'To be, or not to be: that is the question: '</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"Whether 'tis nobler in the mind to suffer "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'The slings and arrows of outrageous fortune, '</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'Or to take arms against a sea of troubles, '</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>]))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>lines</span> <span class=o>:=</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=s>"To be, or not to be: that is the question: "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"Whether 'tis nobler in the mind to suffer "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"The slings and arrows of outrageous fortune, "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"Or to take arms against a sea of troubles, "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Create the Pipeline object and root scope. |
| </span></span></span><span class=line><span class=cl><span class=c1>// It's conventional to use p as the Pipeline variable and |
| </span></span></span><span class=line><span class=cl><span class=c1>// s as the scope variable. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>p</span><span class=p>,</span> <span class=nx>s</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>NewPipelineWithRoot</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Pass the slice to beam.CreateList, to create the pcollection. |
| </span></span></span><span class=line><span class=cl><span class=c1>// The scope variable s is used to add the CreateList transform |
| </span></span></span><span class=line><span class=cl><span class=c1>// to the pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>linesPCol</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>CreateList</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>lines</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kd>function</span> <span class=nx>pipeline</span><span class=p>(</span><span class=nx>root</span>: <span class=kt>beam.Root</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"To be, or not to be: that is the question: "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"Whether 'tis nobler in the mind to suffer "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"The slings and arrows of outrageous fortune, "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"Or to take arms against a sea of troubles, "</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>])</span> |
| </span></span><span class=line><span class=cl> <span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>pipeline</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>transforms</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Create</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>elements</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=l>A</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=l>B</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=l>...</span></span></span></code></pre></div></div></div><h3 id=pcollection-characteristics>3.2. PCollection characteristics</h3><p>A <code>PCollection</code> is owned by the specific <code>Pipeline</code> object for which it is |
| created; multiple pipelines cannot share a <code>PCollection</code>. |
| <span class=language-java>In some respects, a <code>PCollection</code> functions like |
| a <code>Collection</code> class. However, a <code>PCollection</code> can differ in a few key ways:</span></p><h4 id=element-type>3.2.1. Element type</h4><p>The elements of a <code>PCollection</code> may be of any type, but must all be of the same |
| type. However, to support distributed processing, Beam needs to be able to |
| encode each individual element as a byte string (so elements can be passed |
| around to distributed workers). The Beam SDKs provide a data encoding mechanism |
| that includes built-in encoding for commonly-used types as well as support for |
| specifying custom encodings as needed.</p><h4 id=element-schema>3.2.2. Element schema</h4><p>In many cases, the element type in a <code>PCollection</code> has a structure that can be introspected. |
| Examples are JSON, Protocol Buffer, Avro, and database records. Schemas provide a way to |
| express types as a set of named fields, allowing for more-expressive aggregations.</p><h4 id=immutability>3.2.3. Immutability</h4><p>A <code>PCollection</code> is immutable. Once created, you cannot add, remove, or change |
| individual elements. A Beam Transform might process each element of a |
| <code>PCollection</code> and generate new pipeline data (as a new <code>PCollection</code>), <em>but it |
| does not consume or modify the original input collection</em>.</p><blockquote><p><strong>Note:</strong> Beam SDKs avoid unnecessary copying of elements, so <code>PCollection</code> |
| contents are logically immutable, not physically immutable. Changes to input |
| elements may be visible to other DoFns executing within the same bundle, and may |
| cause correctness issues. |
| As a rule, it’s not safe to modify values provided to a DoFn.</p></blockquote><h4 id=random-access>3.2.4. Random access</h4><p>A <code>PCollection</code> does not support random access to individual elements. Instead, |
| Beam Transforms consider every element in a <code>PCollection</code> individually.</p><h4 id=size-and-boundedness>3.2.5. Size and boundedness</h4><p>A <code>PCollection</code> is a large, immutable “bag” of elements. There is no upper limit |
| on how many elements a <code>PCollection</code> can contain; any given <code>PCollection</code> might |
| fit in memory on a single machine, or it might represent a very large |
| distributed data set backed by a persistent data store.</p><p>A <code>PCollection</code> can be either <strong>bounded</strong> or <strong>unbounded</strong> in size. A |
| <strong>bounded</strong> <code>PCollection</code> represents a data set of a known, fixed size, while an |
| <strong>unbounded</strong> <code>PCollection</code> represents a data set of unlimited size. Whether a |
| <code>PCollection</code> is bounded or unbounded depends on the source of the data set that |
| it represents. Reading from a batch data source, such as a file or a database, |
| creates a bounded <code>PCollection</code>. Reading from a streaming or |
| continuously-updating data source, such as Pub/Sub or Kafka, creates an unbounded |
| <code>PCollection</code> (unless you explicitly tell it not to).</p><p>The bounded (or unbounded) nature of your <code>PCollection</code> affects how Beam |
| processes your data. A bounded <code>PCollection</code> can be processed using a batch job, |
| which might read the entire data set once, and perform processing in a job of |
| finite length. An unbounded <code>PCollection</code> must be processed using a streaming |
| job that runs continuously, as the entire collection can never be available for |
| processing at any one time.</p><p>Beam uses <a href=#windowing>windowing</a> to divide a continuously updating unbounded |
| <code>PCollection</code> into logical windows of finite size. These logical windows are |
| determined by some characteristic associated with a data element, such as a |
| <strong>timestamp</strong>. Aggregation transforms (such as <code>GroupByKey</code> and <code>Combine</code>) work |
| on a per-window basis — as the data set is generated, they process each |
| <code>PCollection</code> as a succession of these finite windows.</p><h4 id=element-timestamps>3.2.6. Element timestamps</h4><p>Each element in a <code>PCollection</code> has an associated intrinsic <strong>timestamp</strong>. The |
| timestamp for each element is initially assigned by the <a href=#pipeline-io>Source</a> |
| that creates the <code>PCollection</code>. Sources that create an unbounded <code>PCollection</code> |
| often assign each new element a timestamp that corresponds to when the element |
| was read or added.</p><blockquote><p><strong>Note</strong>: Sources that create a bounded <code>PCollection</code> for a fixed data set |
| also automatically assign timestamps, but the most common behavior is to |
| assign every element the same timestamp (<code>Long.MIN_VALUE</code>).</p></blockquote><p>Timestamps are useful for a <code>PCollection</code> that contains elements with an |
| inherent notion of time. If your pipeline is reading a stream of events, like |
| Tweets or other social media messages, each element might use the time the event |
| was posted as the element timestamp.</p><p>You can manually assign timestamps to the elements of a <code>PCollection</code> if the |
| source doesn’t do it for you. You’ll want to do this if the elements have an |
| inherent timestamp, but the timestamp is somewhere in the structure of the |
| element itself (such as a “time” field in a server log entry). Beam has |
| <a href=#transforms>Transforms</a> that take a <code>PCollection</code> as input and output an |
| identical <code>PCollection</code> with timestamps attached; see <a href=#adding-timestamps-to-a-pcollections-elements>Adding |
| Timestamps</a> for more information |
| about how to do so.</p><h2 id=transforms>4. Transforms</h2><p>Transforms are the operations in your pipeline, and provide a generic |
| processing framework. You provide processing logic in the form of a function |
| object (colloquially referred to as “user code”), and your user code is applied |
| to each element of an input <code>PCollection</code> (or more than one <code>PCollection</code>). |
| Depending on the pipeline runner and back-end that you choose, many different |
| workers across a cluster may execute instances of your user code in parallel. |
| The user code running on each worker generates the output elements that are |
| ultimately added to the final output <code>PCollection</code> that the transform produces.</p><blockquote><p>Aggregation is an important concept to understand when learning about Beam’s |
| transforms. For an introduction to aggregation, see the Basics of the Beam |
| model <a href=/documentation/basics/#aggregation>Aggregation section</a>.</p></blockquote><p>The Beam SDKs contain a number of different transforms that you can apply to |
| your pipeline’s <code>PCollection</code>s. These include general-purpose core transforms, |
| such as <a href=#pardo>ParDo</a> or <a href=#combine>Combine</a>. There are also pre-written |
| <a href=#composite-transforms>composite transforms</a> included in the SDKs, which |
| combine one or more of the core transforms in a useful processing pattern, such |
| as counting or combining elements in a collection. You can also define your own |
| more complex composite transforms to fit your pipeline’s exact use case.</p><p>For a more in-depth tutorial of applying various transforms |
| in the Python SDK, please read and work through |
| <a href=https://colab.sandbox.google.com/github/liferoad/beam/blob/learn-transforms/examples/notebooks/get-started/learn_beam_transforms_by_doing.ipynb>this colab notebook</a>.</p><h3 id=applying-transforms>4.1. Applying transforms</h3><p>To invoke a transform, you must <strong>apply</strong> it to the input <code>PCollection</code>. Each |
| transform in the Beam SDKs has a generic <code>apply</code> method |
| <span class=language-py>(or pipe operator <code>|</code>)</span>. |
| Invoking multiple Beam transforms is similar to <em>method chaining</em>, but with one |
| slight difference: You apply the transform to the input <code>PCollection</code>, passing |
| the transform itself as an argument, and the operation returns the output |
| <code>PCollection</code>. |
| <span class=language-yaml><code>array</code></span> |
| In YAML, transforms are applied by listing their inputs.</span> |
| This takes the general form:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>[</span><span class=n>Output</span> <span class=n>PCollection</span><span class=o>]</span> <span class=o>=</span> <span class=o>[</span><span class=n>Input</span> <span class=n>PCollection</span><span class=o>].</span><span class=na>apply</span><span class=o>([</span><span class=n>Transform</span><span class=o>])</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=p>[</span><span class=n>Output</span> <span class=n>PCollection</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=n>Input</span> <span class=n>PCollection</span><span class=p>]</span> <span class=o>|</span> <span class=p>[</span><span class=n>Transform</span><span class=p>]</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=p>[</span><span class=nx>Output</span> <span class=nx>PCollection</span><span class=p>]</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=p>[</span><span class=nx>Transform</span><span class=p>],</span> <span class=p>[</span><span class=nx>Input</span> <span class=nx>PCollection</span><span class=p>])</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=p>[</span><span class=nx>Output</span> <span class=nx>PCollection</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=nx>Input</span> <span class=nx>PCollection</span><span class=p>].</span><span class=nx>apply</span><span class=p>([</span><span class=nx>Transform</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=nx>Output</span> <span class=nx>PCollection</span><span class=p>]</span> <span class=o>=</span> <span class=k>await</span> <span class=p>[</span><span class=nx>Input</span> <span class=nx>PCollection</span><span class=p>].</span><span class=nx>applyAsync</span><span class=p>([</span><span class=nx>AsyncTransform</span><span class=p>])</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>pipeline</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>transforms</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransformType</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>MyTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MyTransformType</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span></span></span></code></pre></div></div></div><p class=language-yaml>If a transform has more than one |
| (<a href=https://beam.apache.org/documentation/sdks/yaml-errors/>non-error</a>) output, |
| the various outputs can be identified by explicitly giving the output name.</p><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>pipeline</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>transforms</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransformType</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>MyTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MyTransformType</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransform.output_name</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>MyTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MyTransformType</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransform.another_output_name</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=l>...</span></span></span></code></pre></div></div></div><p><p class=language-yaml>For linear pipelines, this can be further simplified by implicitly determining |
| the inputs based on by the ordering of the transforms by designating and setting |
| the type to <code>chain</code>. For example</p><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>pipeline</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>chain</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>transforms</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>ProducingTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>ReadTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>MyTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MyTransformType</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> </span><span class=l>...</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- <span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>ConsumingTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>WriteTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> </span><span class=l>...</span></span></span></code></pre></div></div></div></p><p class="language-java language-py language-typescript">Because Beam uses a generic <code>apply</code> method for <code>PCollection</code>, you can both chain |
| transforms sequentially and also apply transforms that contain other transforms |
| nested within (called <a href=#composite-transforms>composite transforms</a> in the Beam |
| SDKs).</p><p class=language-go>It’s recommended to create a new variable for each new <code>PCollection</code> to |
| sequentially transform input data. <code>Scope</code>s can be used to create functions |
| that contain other transforms |
| (called <a href=#composite-transforms>composite transforms</a> in the Beam SDKs).</p><p>How you apply your pipeline’s transforms determines the structure of your |
| pipeline. The best way to think of your pipeline is as a directed acyclic graph, |
| where <code>PTransform</code> nodes are subroutines that accept <code>PCollection</code> nodes as |
| inputs and emit <code>PCollection</code> nodes as outputs. |
| <span class="language-java language-py">For example, you can chain together transforms to create a pipeline that successively modifies input data:</span> |
| <span class=language-go>For example, you can successively call transforms on PCollections to modify the input data:</span></p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>[</span><span class=n>Final</span> <span class=n>Output</span> <span class=n>PCollection</span><span class=o>]</span> <span class=o>=</span> <span class=o>[</span><span class=n>Initial</span> <span class=n>Input</span> <span class=n>PCollection</span><span class=o>].</span><span class=na>apply</span><span class=o>([</span><span class=n>First</span> <span class=n>Transform</span><span class=o>])</span> |
| </span></span><span class=line><span class=cl><span class=o>.</span><span class=na>apply</span><span class=o>([</span><span class=n>Second</span> <span class=n>Transform</span><span class=o>])</span> |
| </span></span><span class=line><span class=cl><span class=o>.</span><span class=na>apply</span><span class=o>([</span><span class=n>Third</span> <span class=n>Transform</span><span class=o>])</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=p>[</span><span class=n>Final</span> <span class=n>Output</span> <span class=n>PCollection</span><span class=p>]</span> <span class=o>=</span> <span class=p>([</span><span class=n>Initial</span> <span class=n>Input</span> <span class=n>PCollection</span><span class=p>]</span> <span class=o>|</span> <span class=p>[</span><span class=n>First</span> <span class=n>Transform</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=p>[</span><span class=n>Second</span> <span class=n>Transform</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=p>[</span><span class=n>Third</span> <span class=n>Transform</span><span class=p>])</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=p>[</span><span class=nx>Second</span> <span class=nx>PCollection</span><span class=p>]</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=p>[</span><span class=nx>First</span> <span class=nx>Transform</span><span class=p>],</span> <span class=p>[</span><span class=nx>Initial</span> <span class=nx>Input</span> <span class=nx>PCollection</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=nx>Third</span> <span class=nx>PCollection</span><span class=p>]</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=p>[</span><span class=nx>Second</span> <span class=nx>Transform</span><span class=p>],</span> <span class=p>[</span><span class=nx>Second</span> <span class=nx>PCollection</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=nx>Final</span> <span class=nx>Output</span> <span class=nx>PCollection</span><span class=p>]</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=p>[</span><span class=nx>Third</span> <span class=nx>Transform</span><span class=p>],</span> <span class=p>[</span><span class=nx>Third</span> <span class=nx>PCollection</span><span class=p>])</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=p>[</span><span class=nx>Final</span> <span class=nx>Output</span> <span class=nx>PCollection</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=nx>Initial</span> <span class=nx>Input</span> <span class=nx>PCollection</span><span class=p>].</span><span class=nx>apply</span><span class=p>([</span><span class=nx>First</span> <span class=nx>Transform</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>.</span><span class=nx>apply</span><span class=p>([</span><span class=nx>Second</span> <span class=nx>Transform</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>.</span><span class=nx>apply</span><span class=p>([</span><span class=nx>Third</span> <span class=nx>Transform</span><span class=p>])</span></span></span></code></pre></div></div></div><p>The graph of this pipeline looks like the following:</p><p><img src=/images/design-your-pipeline-linear.svg alt="This linear pipeline starts with one input collection, sequentially appliesthree transforms, and ends with one output collection."></p><p><em>Figure 1: A linear pipeline with three sequential transforms.</em></p><p>However, note that a transform <em>does not consume or otherwise alter</em> the input |
| collection — remember that a <code>PCollection</code> is immutable by definition. This means |
| that you can apply multiple transforms to the same input <code>PCollection</code> to create |
| a branching pipeline, like so:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=n>database</span> <span class=n>table</span> <span class=n>rows</span><span class=o>]</span> <span class=o>=</span> <span class=o>[</span><span class=n>Database</span> <span class=n>Table</span> <span class=n>Reader</span><span class=o>].</span><span class=na>apply</span><span class=o>([</span><span class=n>Read</span> <span class=n>Transform</span><span class=o>])</span> |
| </span></span><span class=line><span class=cl><span class=o>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=sc>'A'</span> <span class=n>names</span><span class=o>]</span> <span class=o>=</span> <span class=o>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=n>database</span> <span class=n>table</span> <span class=n>rows</span><span class=o>].</span><span class=na>apply</span><span class=o>([</span><span class=n>Transform</span> <span class=n>A</span><span class=o>])</span> |
| </span></span><span class=line><span class=cl><span class=o>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=sc>'B'</span> <span class=n>names</span><span class=o>]</span> <span class=o>=</span> <span class=o>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=n>database</span> <span class=n>table</span> <span class=n>rows</span><span class=o>].</span><span class=na>apply</span><span class=o>([</span><span class=n>Transform</span> <span class=n>B</span><span class=o>])</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=p>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=n>database</span> <span class=n>table</span> <span class=n>rows</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=n>Database</span> <span class=n>Table</span> <span class=n>Reader</span><span class=p>]</span> <span class=o>|</span> <span class=p>[</span><span class=n>Read</span> <span class=n>Transform</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=s1>'A'</span> <span class=n>names</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=n>database</span> <span class=n>table</span> <span class=n>rows</span><span class=p>]</span> <span class=o>|</span> <span class=p>[</span><span class=n>Transform</span> <span class=n>A</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=s1>'B'</span> <span class=n>names</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=n>PCollection</span> <span class=n>of</span> <span class=n>database</span> <span class=n>table</span> <span class=n>rows</span><span class=p>]</span> <span class=o>|</span> <span class=p>[</span><span class=n>Transform</span> <span class=n>B</span><span class=p>]</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=p>[</span><span class=nx>PCollection</span> <span class=nx>of</span> <span class=nx>database</span> <span class=nx>table</span> <span class=nx>rows</span><span class=p>]</span> <span class=p>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=p>[</span><span class=nx>Read</span> <span class=nx>Transform</span><span class=p>],</span> <span class=p>[</span><span class=nx>Database</span> <span class=nx>Table</span> <span class=nx>Reader</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=nx>PCollection</span> <span class=nx>of</span> <span class=sc>'A'</span> <span class=nx>names</span><span class=p>]</span> <span class=p>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=p>[</span><span class=nx>Transform</span> <span class=nx>A</span><span class=p>],</span> <span class=p>[</span><span class=nx>PCollection</span> <span class=nx>of</span> <span class=nx>database</span> <span class=nx>table</span> <span class=nx>rows</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=nx>PCollection</span> <span class=nx>of</span> <span class=sc>'B'</span> <span class=nx>names</span><span class=p>]</span> <span class=p>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=p>[</span><span class=nx>Transform</span> <span class=nx>B</span><span class=p>],</span> <span class=p>[</span><span class=nx>PCollection</span> <span class=nx>of</span> <span class=nx>database</span> <span class=nx>table</span> <span class=nx>rows</span><span class=p>])</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=p>[</span><span class=nx>PCollection</span> <span class=k>of</span> <span class=nx>database</span> <span class=nx>table</span> <span class=nx>rows</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=nx>Database</span> <span class=nx>Table</span> <span class=nx>Reader</span><span class=p>].</span><span class=nx>apply</span><span class=p>([</span><span class=nx>Read</span> <span class=nx>Transform</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=nx>PCollection</span> <span class=k>of</span> <span class=s1>'A'</span> <span class=nx>names</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=nx>PCollection</span> <span class=k>of</span> <span class=nx>database</span> <span class=nx>table</span> <span class=nx>rows</span><span class=p>].</span><span class=nx>apply</span><span class=p>([</span><span class=nx>Transform</span> <span class=nx>A</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>[</span><span class=nx>PCollection</span> <span class=k>of</span> <span class=s1>'B'</span> <span class=nx>names</span><span class=p>]</span> <span class=o>=</span> <span class=p>[</span><span class=nx>PCollection</span> <span class=k>of</span> <span class=nx>database</span> <span class=nx>table</span> <span class=nx>rows</span><span class=p>].</span><span class=nx>apply</span><span class=p>([</span><span class=nx>Transform</span> <span class=nx>B</span><span class=p>])</span></span></span></code></pre></div></div></div><p>The graph of this branching pipeline looks like the following:</p><p><img src=/images/design-your-pipeline-multiple-pcollections.svg alt="This pipeline applies two transforms to a single input collection. Eachtransform produces an output collection."></p><p><em>Figure 2: A branching pipeline. Two transforms are applied to a single |
| PCollection of database table rows.</em></p><p>You can also build your own <a href=#composite-transforms>composite transforms</a> that |
| nest multiple transforms inside a single, larger transform. Composite transforms |
| are particularly useful for building a reusable sequence of simple steps that |
| get used in a lot of different places.</p><p class=language-python>The pipe syntax allows one to apply PTransforms to <code>tuple</code>s and <code>dict</code>s of |
| PCollections as well for those transforms accepting multiple inputs (such as |
| <code>Flatten</code> and <code>CoGroupByKey</code>).</p><p class=language-typescript>PTransforms can also be applied to any <code>PValue</code>, which include the Root object, |
| PCollections, arrays of <code>PValue</code>s, and objects with <code>PValue</code> values. |
| One can apply transforms to these composite types by wrapping them with |
| <code>beam.P</code>, e.g. |
| <code>beam.P({left: pcollA, right: pcollB}).apply(transformExpectingTwoPCollections)</code>.</p><p class=language-typescript>PTransforms come in two flavors, synchronous and asynchronous, depending on |
| whether their <em>application</em>* involves asynchronous invocations. |
| An <code>AsyncTransform</code> must be applied with <code>applyAsync</code> and returns a <code>Promise</code> |
| which must be awaited before further pipeline construction.</p><h3 id=core-beam-transforms>4.2. Core Beam transforms</h3><p>Beam provides the following core transforms, each of which represents a different |
| processing paradigm:</p><ul><li><code>ParDo</code></li><li><code>GroupByKey</code></li><li><code>CoGroupByKey</code></li><li><code>Combine</code></li><li><code>Flatten</code></li><li><code>Partition</code></li></ul><p class=language-typescript>The Typescript SDK provides some of the most basic of these transforms |
| as methods on <code>PCollection</code> itself.</p><h4 id=pardo>4.2.1. ParDo</h4><p><code>ParDo</code> is a Beam transform for generic parallel processing. The <code>ParDo</code> |
| processing paradigm is similar to the “Map” phase of a |
| <a href=https://en.wikipedia.org/wiki/MapReduce>Map/Shuffle/Reduce</a>-style |
| algorithm: a <code>ParDo</code> transform considers each element in the input |
| <code>PCollection</code>, performs some processing function (your user code) on that |
| element, and emits zero, one, or multiple elements to an output <code>PCollection</code>.</p><p><code>ParDo</code> is useful for a variety of common data processing operations, including:</p><ul><li><strong>Filtering a data set.</strong> You can use <code>ParDo</code> to consider each element in a |
| <code>PCollection</code> and either output that element to a new collection or discard |
| it.</li><li><strong>Formatting or type-converting each element in a data set.</strong> If your input |
| <code>PCollection</code> contains elements that are of a different type or format than |
| you want, you can use <code>ParDo</code> to perform a conversion on each element and |
| output the result to a new <code>PCollection</code>.</li><li><strong>Extracting parts of each element in a data set.</strong> If you have a |
| <code>PCollection</code> of records with multiple fields, for example, you can use a |
| <code>ParDo</code> to parse out just the fields you want to consider into a new |
| <code>PCollection</code>.</li><li><strong>Performing computations on each element in a data set.</strong> You can use <code>ParDo</code> |
| to perform simple or complex computations on every element, or certain |
| elements, of a <code>PCollection</code> and output the results as a new <code>PCollection</code>.</li></ul><p>In such roles, <code>ParDo</code> is a common intermediate step in a pipeline. You might |
| use it to extract certain fields from a set of raw input records, or convert raw |
| input into a different format; you might also use <code>ParDo</code> to convert processed |
| data into a format suitable for output, like database table rows or printable |
| strings.</p><p class="language-java language-go language-py language-typescript">When you apply a <code>ParDo</code> transform, you’ll need to provide user code in the form |
| of a <code>DoFn</code> object. <code>DoFn</code> is a Beam SDK class that defines a distributed |
| processing function.</p><p class=language-yaml>In Beam YAML, <code>ParDo</code> operations are expressed by the <code>MapToFields</code>, <code>Filter</code>, |
| and <code>Explode</code> transform types. These types can take a UDF in the language of your |
| choice, rather than introducing the notion of a <code>DoFn</code>. |
| See <a href=https://beam.apache.org/documentation/sdks/yaml-udf/>the page on mapping fns</a> for more details.</p><span class="language-java language-py"><blockquote><p>When you create a subclass of <code>DoFn</code>, note that your subclass should adhere to |
| the <a href=#requirements-for-writing-user-code-for-beam-transforms>Requirements for writing user code for Beam transforms</a>.</p></blockquote></span><p class=language-go>All DoFns should be registered using a generic <code>register.DoFnXxY[...]</code> |
| function. This allows the Go SDK to infer an encoding from any inputs/outputs, |
| registers the DoFn for execution on remote runners, and optimizes the runtime |
| execution of the DoFns via reflection.</p><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// ComputeWordLengthFn is a DoFn that computes the word length of string elements. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>ComputeWordLengthFn</span> <span class=kd>struct</span><span class=p>{}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// ProcessElement computes the length of word and emits the result. |
| </span></span></span><span class=line><span class=cl><span class=c1>// When creating structs as a DoFn, the ProcessElement method performs the |
| </span></span></span><span class=line><span class=cl><span class=c1>// work of this step in the pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>ComputeWordLengthFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>ctx</span> <span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>)</span> <span class=kt>int</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 2 inputs and 1 output => DoFn2x1 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Input/output types are included in order in the brackets |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>DoFn2x1</span><span class=p>[</span><span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=kt>string</span><span class=p>,</span> <span class=kt>int</span><span class=p>](</span><span class=o>&</span><span class=nx>ComputeWordLengthFn</span><span class=p>{})</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h5 id=applying-pardo>4.2.1.1. Applying ParDo</h5><p class=language-java>Like all Beam transforms, you apply <code>ParDo</code> by calling the <code>apply</code> method on the |
| input <code>PCollection</code> and passing <code>ParDo</code> as an argument, as shown in the |
| following example code:</p><p class=language-py>Like all Beam transforms, you apply <code>ParDo</code> by calling the <code>beam.ParDo</code> on the |
| input <code>PCollection</code> and passing the <code>DoFn</code> as an argument, as shown in the |
| following example code:</p><p class=language-go><code>beam.ParDo</code> applies the passed in <code>DoFn</code> argument to the input <code>PCollection</code>, |
| as shown in the following example code:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// The input PCollection of Strings. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>words</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// The DoFn to perform on each element in the input PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>static</span> <span class=kd>class</span> <span class=nc>ComputeWordLengthFn</span> <span class=kd>extends</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Integer</span><span class=o>></span> <span class=o>{</span> <span class=o>...</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Apply a ParDo to the PCollection "words" to compute lengths for each word. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>wordLengths</span> <span class=o>=</span> <span class=n>words</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>ParDo</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>ComputeWordLengthFn</span><span class=o>()));</span> <span class=c1>// The DoFn to perform on each element, which |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// we define above. |
| </span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># The input PCollection of Strings.</span> |
| </span></span><span class=line><span class=cl><span class=n>words</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># The DoFn to perform on each element in the input PCollection.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>ComputeWordLengthFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=p>[</span><span class=nb>len</span><span class=p>(</span><span class=n>element</span><span class=p>)]</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Apply a ParDo to the PCollection "words" to compute lengths for each word.</span> |
| </span></span><span class=line><span class=cl><span class=n>word_lengths</span> <span class=o>=</span> <span class=n>words</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>ComputeWordLengthFn</span><span class=p>())</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// ComputeWordLengthFn is the DoFn to perform on each element in the input PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>ComputeWordLengthFn</span> <span class=kd>struct</span><span class=p>{}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// ProcessElement is the method to execute for each element. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>ComputeWordLengthFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// DoFns must be registered with beam. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>RegisterType</span><span class=p>(</span><span class=nx>reflect</span><span class=p>.</span><span class=nf>TypeOf</span><span class=p>((</span><span class=o>*</span><span class=nx>ComputeWordLengthFn</span><span class=p>)(</span><span class=kc>nil</span><span class=p>)))</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 2 inputs and 0 outputs => DoFn2x0 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// 1 input => Emitter1 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Input/output types are included in order in the brackets |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>DoFn2x0</span><span class=p>[</span><span class=kt>string</span><span class=p>,</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>)](</span><span class=o>&</span><span class=nx>ComputeWordLengthFn</span><span class=p>{})</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nx>Emitter1</span><span class=p>[</span><span class=kt>int</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// words is an input PCollection of strings |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>var</span> <span class=nx>words</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nx>wordLengths</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>ComputeWordLengthFn</span><span class=p>{},</span> <span class=nx>words</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=err>#</span> <span class=nx>The</span> <span class=nx>input</span> <span class=nx>PCollection</span> <span class=k>of</span> <span class=nx>Strings</span><span class=p>.</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>words</span> : <span class=kt>PCollection</span><span class=p><</span><span class=nt>string</span><span class=p>></span> <span class=o>=</span> <span class=p>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=err>#</span> <span class=nx>The</span> <span class=nx>DoFn</span> <span class=nx>to</span> <span class=nx>perform</span> <span class=nx>on</span> <span class=nx>each</span> <span class=nx>element</span> <span class=k>in</span> <span class=nx>the</span> <span class=nx>input</span> <span class=nx>PCollection</span><span class=p>.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>function</span> <span class=nx>computeWordLengthFn</span><span class=p>()</span><span class=o>:</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>DoFn</span><span class=p><</span><span class=nt>string</span><span class=err>,</span> <span class=na>number</span><span class=p>></span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>process</span>: <span class=kt>function</span><span class=o>*</span> <span class=p>(</span><span class=nx>element</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=nx>element</span><span class=p>.</span><span class=nx>length</span><span class=p>;</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>};</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>words</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>parDo</span><span class=p>(</span><span class=nx>computeWordLengthFn</span><span class=p>()));</span></span></span></code></pre></div></div></div><p>In the example, our input <code>PCollection</code> contains <span class="language-java language-py"><code>String</code></span> |
| <span class=language-go><code>string</code></span> values. We apply a |
| <code>ParDo</code> transform that specifies a function (<code>ComputeWordLengthFn</code>) to compute |
| the length of each string, and outputs the result to a new <code>PCollection</code> of |
| <span class="language-java language-py"><code>Integer</code></span> |
| <span class=language-go><code>int</code></span> values that stores the length of each word.</p><h5 id=4212-creating-a-dofn>4.2.1.2. Creating a DoFn</h5><p>The <code>DoFn</code> object that you pass to <code>ParDo</code> contains the processing logic that |
| gets applied to the elements in the input collection. When you use Beam, often |
| the most important pieces of code you’ll write are these <code>DoFn</code>s - they’re what |
| define your pipeline’s exact data processing tasks.</p><blockquote><p><strong>Note:</strong> When you create your <code>DoFn</code>, be mindful of the <a href=#requirements-for-writing-user-code-for-beam-transforms>Requirements |
| for writing user code for Beam transforms</a> |
| and ensure that your code follows them.</p></blockquote><p class=language-java>A <code>DoFn</code> processes one element at a time from the input <code>PCollection</code>. When you |
| create a subclass of <code>DoFn</code>, you’ll need to provide type parameters that match |
| the types of the input and output elements. If your <code>DoFn</code> processes incoming |
| <code>String</code> elements and produces <code>Integer</code> elements for the output collection |
| (like our previous example, <code>ComputeWordLengthFn</code>), your class declaration would |
| look like this:</p><p class=language-go>A <code>DoFn</code> processes one element at a time from the input <code>PCollection</code>. When you |
| create a <code>DoFn</code> struct, you’ll need to provide type parameters that match |
| the types of the input and output elements in a ProcessElement method. |
| If your <code>DoFn</code> processes incoming <code>string</code> elements and produces <code>int</code> elements |
| for the output collection (like our previous example, <code>ComputeWordLengthFn</code>), your dofn could |
| look like this:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>static</span> <span class=kd>class</span> <span class=nc>ComputeWordLengthFn</span> <span class=kd>extends</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Integer</span><span class=o>></span> <span class=o>{</span> <span class=o>...</span> <span class=o>}</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// ComputeWordLengthFn is a DoFn that computes the word length of string elements. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>ComputeWordLengthFn</span> <span class=kd>struct</span><span class=p>{}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// ProcessElement computes the length of word and emits the result. |
| </span></span></span><span class=line><span class=cl><span class=c1>// When creating structs as a DoFn, the ProcessElement method performs the |
| </span></span></span><span class=line><span class=cl><span class=c1>// work of this step in the pipeline. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>ComputeWordLengthFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 2 inputs and 0 outputs => DoFn2x0 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// 1 input => Emitter1 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Input/output types are included in order in the brackets |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nf>Function2x0</span><span class=p>(</span><span class=o>&</span><span class=nx>ComputeWordLengthFn</span><span class=p>{})</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nx>Emitter1</span><span class=p>[</span><span class=kt>int</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p class=language-java>Inside your <code>DoFn</code> subclass, you’ll write a method annotated with |
| <code>@ProcessElement</code> where you provide the actual processing logic. You don’t need |
| to manually extract the elements from the input collection; the Beam SDKs handle |
| that for you. Your <code>@ProcessElement</code> method should accept a parameter tagged with |
| <code>@Element</code>, which will be populated with the input element. In order to output |
| elements, the method can also take a parameter of type <code>OutputReceiver</code> which |
| provides a method for emitting elements. The parameter types must match the input |
| and output types of your <code>DoFn</code> or the framework will raise an error. Note: <code>@Element</code> and |
| <code>OutputReceiver</code> were introduced in Beam 2.5.0; if using an earlier release of Beam, a |
| <code>ProcessContext</code> parameter should be used instead.</p><p class="language-py language-typescript">Inside your <code>DoFn</code> subclass, you’ll write a method <code>process</code> where you provide |
| the actual processing logic. You don’t need to manually extract the elements |
| from the input collection; the Beam SDKs handle that for you. Your <code>process</code> method |
| should accept an argument <code>element</code>, which is the input element, and return an |
| iterable with its output values. You can accomplish this by emitting individual |
| elements with <code>yield</code> statements, and use <code>yield from</code> to emit all elements from |
| an iterable, such as a list or a generator. Using <code>return</code> statement |
| with an iterable is also acceptable as long as you don’t mix <code>yield</code> and |
| <code>return</code> statements in the same <code>process</code> method, since that leads to <a href=https://github.com/apache/beam/issues/22969>incorrect behavior</a>.</p><p class=language-go>For your <code>DoFn</code> type, you’ll write a method <code>ProcessElement</code> where you provide |
| the actual processing logic. You don’t need to manually extract the elements |
| from the input collection; the Beam SDKs handle that for you. Your <code>ProcessElement</code> method |
| should accept a parameter <code>element</code>, which is the input element. In order to output elements, |
| the method can also take a function parameter, which can be called to emit elements. |
| The parameter types must match the input and output types of your <code>DoFn</code> |
| or the framework will raise an error.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>static</span> <span class=kd>class</span> <span class=nc>ComputeWordLengthFn</span> <span class=kd>extends</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Integer</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>out</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Use OutputReceiver.output to emit the output element. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>out</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>word</span><span class=o>.</span><span class=na>length</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>ComputeWordLengthFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=p>[</span><span class=nb>len</span><span class=p>(</span><span class=n>element</span><span class=p>)]</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// ComputeWordLengthFn is the DoFn to perform on each element in the input PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>ComputeWordLengthFn</span> <span class=kd>struct</span><span class=p>{}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// ProcessElement is the method to execute for each element. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>ComputeWordLengthFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// DoFns must be registered with beam. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>RegisterType</span><span class=p>(</span><span class=nx>reflect</span><span class=p>.</span><span class=nf>TypeOf</span><span class=p>((</span><span class=o>*</span><span class=nx>ComputeWordLengthFn</span><span class=p>)(</span><span class=kc>nil</span><span class=p>)))</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 2 inputs and 0 outputs => DoFn2x0 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// 1 input => Emitter1 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Input/output types are included in order in the brackets |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>DoFn2x0</span><span class=p>[</span><span class=kt>string</span><span class=p>,</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>)](</span><span class=o>&</span><span class=nx>ComputeWordLengthFn</span><span class=p>{})</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nx>Emitter1</span><span class=p>[</span><span class=kt>int</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kd>function</span> <span class=nx>computeWordLengthFn</span><span class=p>()</span><span class=o>:</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>DoFn</span><span class=p><</span><span class=nt>string</span><span class=err>,</span> <span class=na>number</span><span class=p>></span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>process</span>: <span class=kt>function</span><span class=o>*</span> <span class=p>(</span><span class=nx>element</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=nx>element</span><span class=p>.</span><span class=nx>length</span><span class=p>;</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>};</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p class=language-go>Simple DoFns can also be written as functions.</p><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>ComputeWordLengthFn</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>))</span> <span class=p>{</span> <span class=o>...</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 2 inputs and 0 outputs => DoFn2x0 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// 1 input => Emitter1 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Input/output types are included in order in the brackets |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>DoFn2x0</span><span class=p>[</span><span class=kt>string</span><span class=p>,</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>)](</span><span class=o>&</span><span class=nx>ComputeWordLengthFn</span><span class=p>{})</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nx>Emitter1</span><span class=p>[</span><span class=kt>int</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><span class=language-go><blockquote><p><strong>Note:</strong> Whether using a structural <code>DoFn</code> type or a functional <code>DoFn</code>, they should be registered with |
| beam in an <code>init</code> block. Otherwise they may not execute on distributed runners.</p></blockquote></span><span class=language-java><blockquote><p><strong>Note:</strong> If the elements in your input <code>PCollection</code> are key/value pairs, you |
| can access the key or value by using <code>element.getKey()</code> or |
| <code>element.getValue()</code>, respectively.</p></blockquote></span><span class=language-go><blockquote><p><strong>Note:</strong> If the elements in your input <code>PCollection</code> are key/value pairs, your |
| process element method must have two parameters, for each of the key and value, |
| respectively. Similarly, key/value pairs are also output as separate |
| parameters to a single <code>emitter function</code>.</p></blockquote></span><p>A given <code>DoFn</code> instance generally gets invoked one or more times to process some |
| arbitrary bundle of elements. However, Beam doesn’t guarantee an exact number of |
| invocations; it may be invoked multiple times on a given worker node to account |
| for failures and retries. As such, you can cache information across multiple |
| calls to your processing method, but if you do so, make sure the implementation |
| <strong>does not depend on the number of invocations</strong>.</p><p>In your processing method, you’ll also need to meet some immutability |
| requirements to ensure that Beam and the processing back-end can safely |
| serialize and cache the values in your pipeline. Your method should meet the |
| following requirements:</p><span class=language-java><ul><li>You should not in any way modify an element returned by |
| the <code>@Element</code> annotation or <code>ProcessContext.sideInput()</code> (the incoming |
| elements from the input collection).</li><li>Once you output a value using <code>OutputReceiver.output()</code> you should not modify |
| that value in any way.</li></ul></span><span class="language-py language-typescript"><ul><li>You should not in any way modify the <code>element</code> argument provided to the |
| <code>process</code> method, or any side inputs.</li><li>Once you output a value using <code>yield</code> or <code>return</code>, you should not modify |
| that value in any way.</li></ul></span><span class=language-go><ul><li>You should not in any way modify the parameters provided to the |
| <code>ProcessElement</code> method, or any side inputs.</li><li>Once you output a value using an <code>emitter function</code>, you should not modify |
| that value in any way.</li></ul></span><h5 id=lightweight-dofns>4.2.1.3. Lightweight DoFns and other abstractions</h5><p>If your function is relatively straightforward, you can simplify your use of |
| <code>ParDo</code> by providing a lightweight <code>DoFn</code> in-line, as |
| <span class=language-java>an anonymous inner class instance</span> |
| <span class=language-py>a lambda function</span> |
| <span class=language-go>an anonymous function</span> |
| <span class=language-typescript>a function passed to <code>PCollection.map</code> or <code>PCollection.flatMap</code></span>.</p><p>Here’s the previous example, <code>ParDo</code> with <code>ComputeLengthWordsFn</code>, with the |
| <code>DoFn</code> specified as |
| <span class=language-java>an anonymous inner class instance</span> |
| <span class=language-py>a lambda function</span> |
| <span class=language-go>an anonymous function</span> |
| <span class=language-typescript>a function</span>:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// The input PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>words</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Apply a ParDo with an anonymous DoFn to the PCollection words. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Save the result as the PCollection wordLengths. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>wordLengths</span> <span class=o>=</span> <span class=n>words</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"ComputeWordLengths"</span><span class=o>,</span> <span class=c1>// the transform name |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Integer</span><span class=o>>()</span> <span class=o>{</span> <span class=c1>// a DoFn as an anonymous inner class instance |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>out</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>out</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>word</span><span class=o>.</span><span class=na>length</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># The input PCollection of strings.</span> |
| </span></span><span class=line><span class=cl><span class=n>words</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Apply a lambda function to the PCollection words.</span> |
| </span></span><span class=line><span class=cl><span class=c1># Save the result as the PCollection word_lengths.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>word_lengths</span> <span class=o>=</span> <span class=n>words</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>FlatMap</span><span class=p>(</span><span class=k>lambda</span> <span class=n>word</span><span class=p>:</span> <span class=p>[</span><span class=nb>len</span><span class=p>(</span><span class=n>word</span><span class=p>)])</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>The</span> <span class=nx>Go</span> <span class=nx>SDK</span> <span class=nx>cannot</span> <span class=nx>support</span> <span class=nx>anonymous</span> <span class=nx>functions</span> <span class=nx>outside</span> <span class=nx>of</span> <span class=nx>the</span> <span class=nx>deprecated</span> <span class=nx>Go</span> <span class=nx>Direct</span> <span class=nx>runner</span><span class=p>.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// words is the input PCollection of strings |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>var</span> <span class=nx>words</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nx>lengths</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=kd>func</span> <span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>},</span> <span class=nx>words</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=c1>// The input PCollection of strings. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>words</span> <span class=o>=</span> <span class=p>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>words</span><span class=p>.</span><span class=nx>flatMap</span><span class=p>((</span><span class=nx>word</span><span class=p>)</span> <span class=o>=></span> <span class=p>[</span><span class=nx>word</span><span class=p>.</span><span class=nx>length</span><span class=p>]);</span></span></span></code></pre></div></div></div><p>If your <code>ParDo</code> performs a one-to-one mapping of input elements to output |
| elements–that is, for each input element, it applies a function that produces |
| <em>exactly one</em> output element, <span class=language-go>you can return that |
| element directly.</span><span class="language-java language-py">you can use the higher-level |
| <span class=language-java><code>MapElements</code></span><span class="language-py language-py"><code>Map</code></span> |
| transform.</span><span class=language-java><code>MapElements</code> can accept an anonymous |
| Java 8 lambda function for additional brevity.</span></p><p>Here’s the previous example using <span class=language-java><code>MapElements</code></span> |
| <span class="language-py language-typescript"><code>Map</code></span><span class=language-go>a direct return</span>:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// The input PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>words</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Apply a MapElements with an anonymous lambda function to the PCollection words. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Save the result as the PCollection wordLengths. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>wordLengths</span> <span class=o>=</span> <span class=n>words</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>MapElements</span><span class=o>.</span><span class=na>into</span><span class=o>(</span><span class=n>TypeDescriptors</span><span class=o>.</span><span class=na>integers</span><span class=o>())</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>via</span><span class=o>((</span><span class=n>String</span> <span class=n>word</span><span class=o>)</span> <span class=o>-></span> <span class=n>word</span><span class=o>.</span><span class=na>length</span><span class=o>()));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># The input PCollection of string.</span> |
| </span></span><span class=line><span class=cl><span class=n>words</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Apply a Map with a lambda function to the PCollection words.</span> |
| </span></span><span class=line><span class=cl><span class=c1># Save the result as the PCollection word_lengths.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>word_lengths</span> <span class=o>=</span> <span class=n>words</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=nb>len</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>The</span> <span class=nx>Go</span> <span class=nx>SDK</span> <span class=nx>cannot</span> <span class=nx>support</span> <span class=nx>anonymous</span> <span class=nx>functions</span> <span class=nx>outside</span> <span class=nx>of</span> <span class=nx>the</span> <span class=nx>deprecated</span> <span class=nx>Go</span> <span class=nx>Direct</span> <span class=nx>runner</span><span class=p>.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>wordLengths</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>)</span> <span class=kt>int</span> <span class=p>{</span> <span class=k>return</span> <span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> <span class=nx>register</span><span class=p>.</span><span class=nf>Function1x1</span><span class=p>(</span><span class=nx>wordLengths</span><span class=p>)</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>applyWordLenAnon</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>words</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>wordLengths</span><span class=p>,</span> <span class=nx>words</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=c1>// The input PCollection of string. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>words</span> <span class=o>=</span> <span class=p>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>words</span><span class=p>.</span><span class=nx>map</span><span class=p>((</span><span class=nx>word</span><span class=p>)</span> <span class=o>=></span> <span class=nx>word</span><span class=p>.</span><span class=nx>length</span><span class=p>);</span></span></span></code></pre></div></div></div><span class=language-java><blockquote><p><strong>Note:</strong> You can use Java 8 lambda functions with several other Beam |
| transforms, including <code>Filter</code>, <code>FlatMapElements</code>, and <code>Partition</code>.</p></blockquote></span><span class=language-go><blockquote><p><strong>Note:</strong> Anonymous function DoFns do not work on distributed runners. |
| It’s recommended to use named functions and register them with <code>register.FunctionXxY</code> in |
| an <code>init()</code> block.</p></blockquote></span><h5 id=dofn>4.2.1.4. DoFn lifecycle</h5><p>Here is a sequence diagram that shows the lifecycle of the DoFn during |
| the execution of the ParDo transform. The comments give useful |
| information to pipeline developers such as the constraints that |
| apply to the objects or particular cases such as failover or |
| instance reuse. They also give instantiation use cases. Three key points |
| to note are that:</p><ol><li>Teardown is done on a best effort basis and thus |
| isn’t guaranteed.</li><li>The number of DoFn instances created at runtime is runner-dependent.</li><li>For the Python SDK, the pipeline contents such as DoFn user code, |
| is <a href=https://beam.apache.org/documentation/sdks/python-pipeline-dependencies/#pickling-and-managing-the-main-session>serialized into a bytecode</a>. Therefore, <code>DoFn</code>s should not reference objects that are not serializable, such as locks. To manage a single instance of an object across multiple <code>DoFn</code> instances in the same process, use utilities in the <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.utils.shared.html>shared.py</a> module.</li></ol><p><img src=/images/dofn-sequence-diagram.svg alt="This is a sequence diagram that shows the lifecycle of the DoFn"></p><h4 id=groupbykey>4.2.2. GroupByKey</h4><p><code>GroupByKey</code> is a Beam transform for processing collections of key/value pairs. |
| It’s a parallel reduction operation, analogous to the Shuffle phase of a |
| Map/Shuffle/Reduce-style algorithm. The input to <code>GroupByKey</code> is a collection of |
| key/value pairs that represents a <em>multimap</em>, where the collection contains |
| multiple pairs that have the same key, but different values. Given such a |
| collection, you use <code>GroupByKey</code> to collect all of the values associated with |
| each unique key.</p><p><code>GroupByKey</code> is a good way to aggregate data that has something in common. For |
| example, if you have a collection that stores records of customer orders, you |
| might want to group together all the orders from the same postal code (wherein |
| the “key” of the key/value pair is the postal code field, and the “value” is the |
| remainder of the record).</p><p>Let’s examine the mechanics of <code>GroupByKey</code> with a simple example case, where |
| our data set consists of words from a text file and the line number on which |
| they appear. We want to group together all the line numbers (values) that share |
| the same word (key), letting us see all the places in the text where a |
| particular word appears.</p><p>Our input is a <code>PCollection</code> of key/value pairs where each word is a key, and |
| the value is a line number in the file where the word appears. Here’s a list of |
| the key/value pairs in the input collection:</p><pre tabindex=0><code>cat, 1 |
| dog, 5 |
| and, 1 |
| jump, 3 |
| tree, 2 |
| cat, 5 |
| dog, 2 |
| and, 2 |
| cat, 9 |
| and, 6 |
| ... |
| </code></pre><p><code>GroupByKey</code> gathers up all the values with the same key and outputs a new pair |
| consisting of the unique key and a collection of all of the values that were |
| associated with that key in the input collection. If we apply <code>GroupByKey</code> to |
| our input collection above, the output collection would look like this:</p><pre tabindex=0><code>cat, [1,5,9] |
| dog, [5,2] |
| and, [1,2,6] |
| jump, [3] |
| tree, [2] |
| ... |
| </code></pre><p>Thus, <code>GroupByKey</code> represents a transform from a multimap (multiple keys to |
| individual values) to a uni-map (unique keys to collections of values).</p><p><span class=language-java>Using <code>GroupByKey</code> is straightforward:</span></p><p class="language-py language-typescript language-yaml">While all SDKs have a <code>GroupByKey</code> transform, using <code>GroupBy</code> is |
| generally more natural. |
| The <code>GroupBy</code> transform can be parameterized by the name(s) of properties |
| on which to group the elements of the PCollection, or a function taking |
| the each element as input that maps to a key on which to do grouping.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// The input PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>></span> <span class=n>mapped</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Apply GroupByKey to the PCollection mapped. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Save the result as the PCollection reduced. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Iterable</span><span class=o><</span><span class=n>String</span><span class=o>>>></span> <span class=n>reduced</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>mapped</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>GroupByKey</span><span class=o>.<</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>></span><span class=n>create</span><span class=o>());</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># The input PCollection of (`string`, `int`) tuples.</span> |
| </span></span><span class=line><span class=cl><span class=n>words_and_counts</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>grouped_words</span> <span class=o>=</span> <span class=n>words_and_counts</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>GroupByKey</span><span class=p>()</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// CreateAndSplit creates and returns a PCollection with <K,V> |
| </span></span></span><span class=line><span class=cl><span class=c1>// from an input slice of stringPair (struct with K, V string fields). |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>pairs</span> <span class=o>:=</span> <span class=nf>CreateAndSplit</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>input</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=nx>keyed</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>GroupByKey</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>pairs</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=c1>// A PCollection of elements like |
| </span></span></span><span class=line><span class=cl><span class=c1>// {word: "cat", score: 1}, {word: "dog", score: 5}, {word: "cat", score: 5}, ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kr>const</span> <span class=nx>scores</span> : <span class=kt>PCollection</span><span class=o><</span><span class=p>{</span><span class=nx>word</span>: <span class=kt>string</span><span class=p>,</span> <span class=nx>score</span>: <span class=kt>number</span><span class=p>}</span><span class=o>></span> <span class=o>=</span> <span class=p>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// This will produce a PCollection with elements like |
| </span></span></span><span class=line><span class=cl><span class=c1>// {key: "cat", value: [{ word: "cat", score: 1 }, |
| </span></span></span><span class=line><span class=cl><span class=c1>// { word: "cat", score: 5 }, ...]} |
| </span></span></span><span class=line><span class=cl><span class=c1>// {key: "dog", value: [{ word: "dog", score: 5 }, ...]} |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kr>const</span> <span class=nx>grouped_by_word</span> <span class=o>=</span> <span class=nx>scores</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>groupBy</span><span class=p>(</span><span class=s2>"word"</span><span class=p>));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// This will produce a PCollection with elements like |
| </span></span></span><span class=line><span class=cl><span class=c1>// {key: 3, value: [{ word: "cat", score: 1 }, |
| </span></span></span><span class=line><span class=cl><span class=c1>// { word: "dog", score: 5 }, |
| </span></span></span><span class=line><span class=cl><span class=c1>// { word: "cat", score: 5 }, ...]} |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kr>const</span> <span class=nx>by_word_length</span> <span class=o>=</span> <span class=nx>scores</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>groupBy</span><span class=p>((</span><span class=nx>x</span><span class=p>)</span> <span class=o>=></span> <span class=nx>x</span><span class=p>.</span><span class=nx>word</span><span class=p>.</span><span class=nx>length</span><span class=p>));</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Combine</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>group_by</span><span class=p>:</span><span class=w> </span><span class=l>animal</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>combine</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>weight</span><span class=p>:</span><span class=w> </span><span class=l>group</span></span></span></code></pre></div></div></div><h5 id=groupbykey-and-unbounded-pcollections>4.2.2.1 GroupByKey and unbounded PCollections</h5><p>If you are using unbounded <code>PCollection</code>s, you must use either <a href=#setting-your-pcollections-windowing-function>non-global |
| windowing</a> or an |
| <a href=#triggers>aggregation trigger</a> in order to perform a <code>GroupByKey</code> or |
| <a href=#cogroupbykey>CoGroupByKey</a>. This is because a bounded <code>GroupByKey</code> or |
| <code>CoGroupByKey</code> must wait for all the data with a certain key to be collected, |
| but with unbounded collections, the data is unlimited. Windowing and/or triggers |
| allow grouping to operate on logical, finite bundles of data within the |
| unbounded data streams.</p><p>If you do apply <code>GroupByKey</code> or <code>CoGroupByKey</code> to a group of unbounded |
| <code>PCollection</code>s without setting either a non-global windowing strategy, a trigger |
| strategy, or both for each collection, Beam generates an IllegalStateException |
| error at pipeline construction time.</p><p>When using <code>GroupByKey</code> or <code>CoGroupByKey</code> to group <code>PCollection</code>s that have a |
| <a href=#windowing>windowing strategy</a> applied, all of the <code>PCollection</code>s you want to |
| group <em>must use the same windowing strategy</em> and window sizing. For example, all |
| of the collections you are merging must use (hypothetically) identical 5-minute |
| fixed windows, or 4-minute sliding windows starting every 30 seconds.</p><p>If your pipeline attempts to use <code>GroupByKey</code> or <code>CoGroupByKey</code> to merge |
| <code>PCollection</code>s with incompatible windows, Beam generates an |
| IllegalStateException error at pipeline construction time.</p><h4 id=cogroupbykey>4.2.3. CoGroupByKey</h4><p><code>CoGroupByKey</code> performs a relational join of two or more key/value |
| <code>PCollection</code>s that have the same key type. |
| <a href=/documentation/pipelines/design-your-pipeline/#multiple-sources>Design Your Pipeline</a> |
| shows an example pipeline that uses a join.</p><p>Consider using <code>CoGroupByKey</code> if you have multiple data sets that provide |
| information about related things. For example, let’s say you have two different |
| files with user data: one file has names and email addresses; the other file |
| has names and phone numbers. You can join those two data sets, using the user |
| name as a common key and the other data as the associated values. After the |
| join, you have one data set that contains all of the information (email |
| addresses and phone numbers) associated with each name.</p><p>One can also consider using SqlTransform to perform a join.</p><p>If you are using unbounded <code>PCollection</code>s, you must use either <a href=#setting-your-pcollections-windowing-function>non-global |
| windowing</a> or an |
| <a href=#triggers>aggregation trigger</a> in order to perform a <code>CoGroupByKey</code>. See |
| <a href=#groupbykey-and-unbounded-pcollections>GroupByKey and unbounded PCollections</a> |
| for more details.</p><p class=language-java>In the Beam SDK for Java, <code>CoGroupByKey</code> accepts a tuple of keyed |
| <code>PCollection</code>s (<code>PCollection<KV<K, V>></code>) as input. For type safety, the SDK |
| requires you to pass each <code>PCollection</code> as part of a <code>KeyedPCollectionTuple</code>. |
| You must declare a <code>TupleTag</code> for each input <code>PCollection</code> in the |
| <code>KeyedPCollectionTuple</code> that you want to pass to <code>CoGroupByKey</code>. As output, |
| <code>CoGroupByKey</code> returns a <code>PCollection<KV<K, CoGbkResult>></code>, which groups values |
| from all the input <code>PCollection</code>s by their common keys. Each key (all of type |
| <code>K</code>) will have a different <code>CoGbkResult</code>, which is a map from <code>TupleTag<T></code> to |
| <code>Iterable<T></code>. You can access a specific collection in an <code>CoGbkResult</code> object |
| by using the <code>TupleTag</code> that you supplied with the initial collection.</p><p class=language-py>In the Beam SDK for Python, <code>CoGroupByKey</code> accepts a dictionary of keyed |
| <code>PCollection</code>s as input. As output, <code>CoGroupByKey</code> creates a single output |
| <code>PCollection</code> that contains one key/value tuple for each key in the input |
| <code>PCollection</code>s. Each key’s value is a dictionary that maps each tag to an |
| iterable of the values under they key in the corresponding <code>PCollection</code>.</p><p class=language-go>In the Beam Go SDK, <code>CoGroupByKey</code> accepts an arbitrary number of |
| <code>PCollection</code>s as input. As output, <code>CoGroupByKey</code> creates a single output |
| <code>PCollection</code> that groups each key with value iterator functions for each |
| input <code>PCollection</code>. The iterator functions map to input <code>PCollections</code> in |
| the same order they were provided to the <code>CoGroupByKey</code>.</p><p>The following conceptual examples use two input collections to show the mechanics of |
| <code>CoGroupByKey</code>.</p><p class=language-java>The first set of data has a <code>TupleTag<String></code> called <code>emailsTag</code> and contains names |
| and email addresses. The second set of data has a <code>TupleTag<String></code> called |
| <code>phonesTag</code> and contains names and phone numbers.</p><p class="language-py language-go">The first set of data contains names and email addresses. The second set of |
| data contains names and phone numbers.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>final</span> <span class=n>List</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>></span> <span class=n>emailsList</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"amy"</span><span class=o>,</span> <span class=s>"amy@example.com"</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"carl"</span><span class=o>,</span> <span class=s>"carl@example.com"</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"julia"</span><span class=o>,</span> <span class=s>"julia@example.com"</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"carl"</span><span class=o>,</span> <span class=s>"carl@email.com"</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>final</span> <span class=n>List</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>></span> <span class=n>phonesList</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"amy"</span><span class=o>,</span> <span class=s>"111-222-3333"</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"james"</span><span class=o>,</span> <span class=s>"222-333-4444"</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"amy"</span><span class=o>,</span> <span class=s>"333-444-5555"</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"carl"</span><span class=o>,</span> <span class=s>"444-555-6666"</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>></span> <span class=n>emails</span> <span class=o>=</span> <span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=s>"CreateEmails"</span><span class=o>,</span> <span class=n>Create</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>emailsList</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>></span> <span class=n>phones</span> <span class=o>=</span> <span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=s>"CreatePhones"</span><span class=o>,</span> <span class=n>Create</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>phonesList</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>emails_list</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'amy'</span><span class=p>,</span> <span class=s1>'amy@example.com'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'carl'</span><span class=p>,</span> <span class=s1>'carl@example.com'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'julia'</span><span class=p>,</span> <span class=s1>'julia@example.com'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'carl'</span><span class=p>,</span> <span class=s1>'carl@email.com'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl><span class=p>]</span> |
| </span></span><span class=line><span class=cl><span class=n>phones_list</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'amy'</span><span class=p>,</span> <span class=s1>'111-222-3333'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'james'</span><span class=p>,</span> <span class=s1>'222-333-4444'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'amy'</span><span class=p>,</span> <span class=s1>'333-444-5555'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'carl'</span><span class=p>,</span> <span class=s1>'444-555-6666'</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl><span class=p>]</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>emails</span> <span class=o>=</span> <span class=n>p</span> <span class=o>|</span> <span class=s1>'CreateEmails'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>Create</span><span class=p>(</span><span class=n>emails_list</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=n>phones</span> <span class=o>=</span> <span class=n>p</span> <span class=o>|</span> <span class=s1>'CreatePhones'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>Create</span><span class=p>(</span><span class=n>phones_list</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>stringPair</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>K</span><span class=p>,</span> <span class=nx>V</span> <span class=kt>string</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>splitStringPair</span><span class=p>(</span><span class=nx>e</span> <span class=nx>stringPair</span><span class=p>)</span> <span class=p>(</span><span class=kt>string</span><span class=p>,</span> <span class=kt>string</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>e</span><span class=p>.</span><span class=nx>K</span><span class=p>,</span> <span class=nx>e</span><span class=p>.</span><span class=nx>V</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Register DoFn. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nf>Function1x2</span><span class=p>(</span><span class=nx>splitStringPair</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// CreateAndSplit is a helper function that creates |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>CreateAndSplit</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>input</span> <span class=p>[]</span><span class=nx>stringPair</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>initial</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>CreateList</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>input</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>splitStringPair</span><span class=p>,</span> <span class=nx>initial</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>var</span> <span class=nx>emailSlice</span> <span class=p>=</span> <span class=p>[]</span><span class=nx>stringPair</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"amy"</span><span class=p>,</span> <span class=s>"amy@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"carl"</span><span class=p>,</span> <span class=s>"carl@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"julia"</span><span class=p>,</span> <span class=s>"julia@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"carl"</span><span class=p>,</span> <span class=s>"carl@email.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>var</span> <span class=nx>phoneSlice</span> <span class=p>=</span> <span class=p>[]</span><span class=nx>stringPair</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"amy"</span><span class=p>,</span> <span class=s>"111-222-3333"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"james"</span><span class=p>,</span> <span class=s>"222-333-4444"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"amy"</span><span class=p>,</span> <span class=s>"333-444-5555"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span><span class=s>"carl"</span><span class=p>,</span> <span class=s>"444-555-6666"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=nx>emails</span> <span class=o>:=</span> <span class=nf>CreateAndSplit</span><span class=p>(</span><span class=nx>s</span><span class=p>.</span><span class=nf>Scope</span><span class=p>(</span><span class=s>"CreateEmails"</span><span class=p>),</span> <span class=nx>emailSlice</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=nx>phones</span> <span class=o>:=</span> <span class=nf>CreateAndSplit</span><span class=p>(</span><span class=nx>s</span><span class=p>.</span><span class=nf>Scope</span><span class=p>(</span><span class=s>"CreatePhones"</span><span class=p>),</span> <span class=nx>phoneSlice</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>emails_list</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"amy"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"amy@example.com"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"carl"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"carl@example.com"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"julia"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"julia@example.com"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"carl"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"carl@email.com"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>];</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>phones_list</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"amy"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"111-222-3333"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"james"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"222-333-4444"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"amy"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"333-444-5555"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"carl"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"444-555-6666"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>];</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>emails</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>withName</span><span class=p>(</span><span class=s2>"createEmails"</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>(</span><span class=nx>emails_list</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>phones</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>withName</span><span class=p>(</span><span class=s2>"createPhones"</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>(</span><span class=nx>phones_list</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Create</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>CreateEmails</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>elements</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"amy"</span><span class=nt>, email</span><span class=p>:</span><span class=w> </span><span class=s2>"amy@example.com"</span><span class=w> </span>}<span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"carl"</span><span class=nt>, email</span><span class=p>:</span><span class=w> </span><span class=s2>"carl@example.com"</span><span class=w> </span>}<span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"julia"</span><span class=nt>, email</span><span class=p>:</span><span class=w> </span><span class=s2>"julia@example.com"</span><span class=w> </span>}<span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"carl"</span><span class=nt>, email</span><span class=p>:</span><span class=w> </span><span class=s2>"carl@email.com"</span><span class=w> </span>}<span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Create</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>CreatePhones</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>elements</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"amy"</span><span class=nt>, phone</span><span class=p>:</span><span class=w> </span><span class=s2>"111-222-3333"</span><span class=w> </span>}<span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"james"</span><span class=nt>, phone</span><span class=p>:</span><span class=w> </span><span class=s2>"222-333-4444"</span><span class=w> </span>}<span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"amy"</span><span class=nt>, phone</span><span class=p>:</span><span class=w> </span><span class=s2>"333-444-5555"</span><span class=w> </span>}<span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span>- {<span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=s2>"carl"</span><span class=nt>, phone</span><span class=p>:</span><span class=w> </span><span class=s2>"444-555-6666"</span><span class=w> </span>}</span></span></code></pre></div></div></div><p>After <code>CoGroupByKey</code>, the resulting data contains all data associated with each |
| unique key from any of the input collections.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>final</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>emailsTag</span> <span class=o>=</span> <span class=k>new</span> <span class=n>TupleTag</span><span class=o><>();</span> |
| </span></span><span class=line><span class=cl><span class=kd>final</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>phonesTag</span> <span class=o>=</span> <span class=k>new</span> <span class=n>TupleTag</span><span class=o><>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>final</span> <span class=n>List</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>CoGbkResult</span><span class=o>>></span> <span class=n>expectedResults</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"amy"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>CoGbkResult</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>emailsTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span><span class=s>"amy@example.com"</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>and</span><span class=o>(</span><span class=n>phonesTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span><span class=s>"111-222-3333"</span><span class=o>,</span> <span class=s>"333-444-5555"</span><span class=o>))),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"carl"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>CoGbkResult</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>emailsTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span><span class=s>"carl@email.com"</span><span class=o>,</span> <span class=s>"carl@example.com"</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>and</span><span class=o>(</span><span class=n>phonesTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span><span class=s>"444-555-6666"</span><span class=o>))),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"james"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>CoGbkResult</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>emailsTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>())</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>and</span><span class=o>(</span><span class=n>phonesTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span><span class=s>"222-333-4444"</span><span class=o>))),</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o>.</span><span class=na>of</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"julia"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>CoGbkResult</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>emailsTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span><span class=s>"julia@example.com"</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>and</span><span class=o>(</span><span class=n>phonesTag</span><span class=o>,</span> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>())));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>results</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'amy'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'emails'</span><span class=p>:</span> <span class=p>[</span><span class=s1>'amy@example.com'</span><span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'phones'</span><span class=p>:</span> <span class=p>[</span><span class=s1>'111-222-3333'</span><span class=p>,</span> <span class=s1>'333-444-5555'</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=p>}),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'carl'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'emails'</span><span class=p>:</span> <span class=p>[</span><span class=s1>'carl@email.com'</span><span class=p>,</span> <span class=s1>'carl@example.com'</span><span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'phones'</span><span class=p>:</span> <span class=p>[</span><span class=s1>'444-555-6666'</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=p>}),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'james'</span><span class=p>,</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'emails'</span><span class=p>:</span> <span class=p>[],</span> <span class=s1>'phones'</span><span class=p>:</span> <span class=p>[</span><span class=s1>'222-333-4444'</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=p>}),</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=s1>'julia'</span><span class=p>,</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'emails'</span><span class=p>:</span> <span class=p>[</span><span class=s1>'julia@example.com'</span><span class=p>],</span> <span class=s1>'phones'</span><span class=p>:</span> <span class=p>[]</span> |
| </span></span><span class=line><span class=cl> <span class=p>}),</span> |
| </span></span><span class=line><span class=cl><span class=p>]</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>results</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>CoGroupByKey</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>emails</span><span class=p>,</span> <span class=nx>phones</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nx>contactLines</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>formatCoGBKResults</span><span class=p>,</span> <span class=nx>results</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Synthetic example results of a cogbk. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>results</span> <span class=o>:=</span> <span class=p>[]</span><span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span> <span class=kt>string</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>,</span> <span class=nx>Phones</span> <span class=p>[]</span><span class=kt>string</span> |
| </span></span><span class=line><span class=cl><span class=p>}{</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"amy"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"amy@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"111-222-3333"</span><span class=p>,</span> <span class=s>"333-444-5555"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"carl"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"carl@email.com"</span><span class=p>,</span> <span class=s>"carl@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"444-555-6666"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"james"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"222-333-4444"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"julia"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"julia@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>results</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>name</span><span class=o>:</span> <span class=s2>"amy"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>values</span><span class=o>:</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>emails</span><span class=o>:</span> <span class=p>[{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"amy"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"amy@example.com"</span> <span class=p>}],</span> |
| </span></span><span class=line><span class=cl> <span class=nx>phones</span><span class=o>:</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"amy"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"111-222-3333"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"amy"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"333-444-5555"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>name</span><span class=o>:</span> <span class=s2>"carl"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>values</span><span class=o>:</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>emails</span><span class=o>:</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"carl"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"carl@example.com"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"carl"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"carl@email.com"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=nx>phones</span><span class=o>:</span> <span class=p>[{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"carl"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"444-555-6666"</span> <span class=p>}],</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>name</span><span class=o>:</span> <span class=s2>"james"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>values</span><span class=o>:</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>emails</span><span class=o>:</span> <span class=p>[],</span> |
| </span></span><span class=line><span class=cl> <span class=nx>phones</span><span class=o>:</span> <span class=p>[{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"james"</span><span class=p>,</span> <span class=nx>phone</span><span class=o>:</span> <span class=s2>"222-333-4444"</span> <span class=p>}],</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>name</span><span class=o>:</span> <span class=s2>"julia"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>values</span><span class=o>:</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>emails</span><span class=o>:</span> <span class=p>[{</span> <span class=nx>name</span><span class=o>:</span> <span class=s2>"julia"</span><span class=p>,</span> <span class=nx>email</span><span class=o>:</span> <span class=s2>"julia@example.com"</span> <span class=p>}],</span> |
| </span></span><span class=line><span class=cl> <span class=nx>phones</span><span class=o>:</span> <span class=p>[],</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>];</span></span></span></code></pre></div></div></div><p class="language-java language-py language-typescript">The following code example joins the two <code>PCollection</code>s with <code>CoGroupByKey</code>, |
| followed by a <code>ParDo</code> to consume the result. Then, the code uses tags to look up |
| and format data from each collection.</p><p class=language-go>The following code example joins the two <code>PCollection</code>s with <code>CoGroupByKey</code>, |
| followed by a <code>ParDo</code> to consume the result. The ordering of the <code>DoFn</code> iterator |
| parameters maps to the ordering of the <code>CoGroupByKey</code> inputs.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>CoGbkResult</span><span class=o>>></span> <span class=n>results</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>KeyedPCollectionTuple</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>emailsTag</span><span class=o>,</span> <span class=n>emails</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>and</span><span class=o>(</span><span class=n>phonesTag</span><span class=o>,</span> <span class=n>phones</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>CoGroupByKey</span><span class=o>.</span><span class=na>create</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>contactLines</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>results</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>CoGbkResult</span><span class=o>>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=n>ProcessContext</span> <span class=n>c</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>CoGbkResult</span><span class=o>></span> <span class=n>e</span> <span class=o>=</span> <span class=n>c</span><span class=o>.</span><span class=na>element</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=n>name</span> <span class=o>=</span> <span class=n>e</span><span class=o>.</span><span class=na>getKey</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Iterable</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>emailsIter</span> <span class=o>=</span> <span class=n>e</span><span class=o>.</span><span class=na>getValue</span><span class=o>().</span><span class=na>getAll</span><span class=o>(</span><span class=n>emailsTag</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=n>Iterable</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>phonesIter</span> <span class=o>=</span> <span class=n>e</span><span class=o>.</span><span class=na>getValue</span><span class=o>().</span><span class=na>getAll</span><span class=o>(</span><span class=n>phonesTag</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=n>formattedResult</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>Snippets</span><span class=o>.</span><span class=na>formatCoGbkResults</span><span class=o>(</span><span class=n>name</span><span class=o>,</span> <span class=n>emailsIter</span><span class=o>,</span> <span class=n>phonesIter</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=n>c</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>formattedResult</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># The result PCollection contains one key-value element for each key in the</span> |
| </span></span><span class=line><span class=cl><span class=c1># input PCollections. The key of the pair will be the key from the input and</span> |
| </span></span><span class=line><span class=cl><span class=c1># the value will be a dictionary with two entries: 'emails' - an iterable of</span> |
| </span></span><span class=line><span class=cl><span class=c1># all values for the current key in the emails PCollection and 'phones': an</span> |
| </span></span><span class=line><span class=cl><span class=c1># iterable of all values for the current key in the phones PCollection.</span> |
| </span></span><span class=line><span class=cl><span class=n>results</span> <span class=o>=</span> <span class=p>({</span><span class=s1>'emails'</span><span class=p>:</span> <span class=n>emails</span><span class=p>,</span> <span class=s1>'phones'</span><span class=p>:</span> <span class=n>phones</span><span class=p>}</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CoGroupByKey</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>def</span> <span class=nf>join_info</span><span class=p>(</span><span class=n>name_info</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=n>name</span><span class=p>,</span> <span class=n>info</span><span class=p>)</span> <span class=o>=</span> <span class=n>name_info</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=s1>'</span><span class=si>%s</span><span class=s1>; </span><span class=si>%s</span><span class=s1>; </span><span class=si>%s</span><span class=s1>'</span> <span class=o>%</span>\ |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=n>name</span><span class=p>,</span> <span class=nb>sorted</span><span class=p>(</span><span class=n>info</span><span class=p>[</span><span class=s1>'emails'</span><span class=p>]),</span> <span class=nb>sorted</span><span class=p>(</span><span class=n>info</span><span class=p>[</span><span class=s1>'phones'</span><span class=p>]))</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>contact_lines</span> <span class=o>=</span> <span class=n>results</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=n>join_info</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>formatCoGBKResults</span><span class=p>(</span><span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emailIter</span><span class=p>,</span> <span class=nx>phoneIter</span> <span class=kd>func</span><span class=p>(</span><span class=o>*</span><span class=kt>string</span><span class=p>)</span> <span class=kt>bool</span><span class=p>)</span> <span class=kt>string</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>var</span> <span class=nx>s</span> <span class=kt>string</span> |
| </span></span><span class=line><span class=cl> <span class=kd>var</span> <span class=nx>emails</span><span class=p>,</span> <span class=nx>phones</span> <span class=p>[]</span><span class=kt>string</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nf>emailIter</span><span class=p>(</span><span class=o>&</span><span class=nx>s</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>emails</span> <span class=p>=</span> <span class=nb>append</span><span class=p>(</span><span class=nx>emails</span><span class=p>,</span> <span class=nx>s</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nf>phoneIter</span><span class=p>(</span><span class=o>&</span><span class=nx>s</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>phones</span> <span class=p>=</span> <span class=nb>append</span><span class=p>(</span><span class=nx>phones</span><span class=p>,</span> <span class=nx>s</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Values have no guaranteed order, sort for deterministic output. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>sort</span><span class=p>.</span><span class=nf>Strings</span><span class=p>(</span><span class=nx>emails</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>sort</span><span class=p>.</span><span class=nf>Strings</span><span class=p>(</span><span class=nx>phones</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>fmt</span><span class=p>.</span><span class=nf>Sprintf</span><span class=p>(</span><span class=s>"%s; %s; %s"</span><span class=p>,</span> <span class=nx>key</span><span class=p>,</span> <span class=nf>formatStringIter</span><span class=p>(</span><span class=nx>emails</span><span class=p>),</span> <span class=nf>formatStringIter</span><span class=p>(</span><span class=nx>phones</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nf>Function3x1</span><span class=p>(</span><span class=nx>formatCoGBKResults</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 1 input of type string => Iter1[string] |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>Iter1</span><span class=p>[</span><span class=kt>string</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Synthetic example results of a cogbk. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>results</span> <span class=o>:=</span> <span class=p>[]</span><span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span> <span class=kt>string</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>,</span> <span class=nx>Phones</span> <span class=p>[]</span><span class=kt>string</span> |
| </span></span><span class=line><span class=cl><span class=p>}{</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"amy"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"amy@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"111-222-3333"</span><span class=p>,</span> <span class=s>"333-444-5555"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"carl"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"carl@email.com"</span><span class=p>,</span> <span class=s>"carl@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"444-555-6666"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"james"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"222-333-4444"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Key</span><span class=p>:</span> <span class=s>"julia"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Emails</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=s>"julia@example.com"</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Phones</span><span class=p>:</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{},</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>formatted_results_pcoll</span> <span class=o>=</span> <span class=nx>beam</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>P</span><span class=p>({</span> <span class=nx>emails</span><span class=p>,</span> <span class=nx>phones</span> <span class=p>})</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>coGroupBy</span><span class=p>(</span><span class=s2>"name"</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>map</span><span class=p>(</span><span class=kd>function</span> <span class=nx>formatResults</span><span class=p>({</span> <span class=nx>key</span><span class=p>,</span> <span class=nx>values</span> <span class=p>})</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=kr>const</span> <span class=nx>emails</span> <span class=o>=</span> <span class=nx>values</span><span class=p>.</span><span class=nx>emails</span><span class=p>.</span><span class=nx>map</span><span class=p>((</span><span class=nx>x</span><span class=p>)</span> <span class=o>=></span> <span class=nx>x</span><span class=p>.</span><span class=nx>email</span><span class=p>).</span><span class=nx>sort</span><span class=p>();</span> |
| </span></span><span class=line><span class=cl> <span class=kr>const</span> <span class=nx>phones</span> <span class=o>=</span> <span class=nx>values</span><span class=p>.</span><span class=nx>phones</span><span class=p>.</span><span class=nx>map</span><span class=p>((</span><span class=nx>x</span><span class=p>)</span> <span class=o>=></span> <span class=nx>x</span><span class=p>.</span><span class=nx>phone</span><span class=p>).</span><span class=nx>sort</span><span class=p>();</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=sb>`</span><span class=si>${</span><span class=nx>key</span><span class=si>}</span><span class=sb>; [</span><span class=si>${</span><span class=nx>emails</span><span class=si>}</span><span class=sb>]; [</span><span class=si>${</span><span class=nx>phones</span><span class=si>}</span><span class=sb>]`</span><span class=p>;</span> |
| </span></span><span class=line><span class=cl> <span class=p>});</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MapToFields</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>PrepareEmails</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>CreateEmails</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fields</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>name</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>email</span><span class=p>:</span><span class=w> </span><span class=s2>"[email]"</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>phone</span><span class=p>:</span><span class=w> </span><span class=s2>"[]"</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MapToFields</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>PreparePhones</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>CreatePhones</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fields</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>name</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>email</span><span class=p>:</span><span class=w> </span><span class=s2>"[]"</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>phone</span><span class=p>:</span><span class=w> </span><span class=s2>"[phone]"</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Combine</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>CoGropuBy</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=p>[</span><span class=l>PrepareEmails, PreparePhones]</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>group_by</span><span class=p>:</span><span class=w> </span><span class=p>[</span><span class=l>name]</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>combine</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>email</span><span class=p>:</span><span class=w> </span><span class=l>concat</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>phone</span><span class=p>:</span><span class=w> </span><span class=l>concat</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MapToFields</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>name</span><span class=p>:</span><span class=w> </span><span class=l>FormatResults</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>CoGropuBy</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fields</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>formatted</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=s2>"'%s; %s; %s' % (name, sorted(email), sorted(phone))"</span></span></span></code></pre></div></div></div><p>The formatted data looks like this:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>final</span> <span class=n>List</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>formattedResults</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>Arrays</span><span class=o>.</span><span class=na>asList</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"amy; ['amy@example.com']; ['111-222-3333', '333-444-5555']"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"carl; ['carl@email.com', 'carl@example.com']; ['444-555-6666']"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"james; []; ['222-333-4444']"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"julia; ['julia@example.com']; []"</span><span class=o>);</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>formatted_results</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"amy; ['amy@example.com']; ['111-222-3333', '333-444-5555']"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"carl; ['carl@email.com', 'carl@example.com']; ['444-555-6666']"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"james; []; ['222-333-4444']"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"julia; ['julia@example.com']; []"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl><span class=p>]</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>formattedResults</span> <span class=o>:=</span> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=s>"amy; ['amy@example.com']; ['111-222-3333', '333-444-5555']"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"carl; ['carl@email.com', 'carl@example.com']; ['444-555-6666']"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"james; []; ['222-333-4444']"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s>"julia; ['julia@example.com']; []"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>formatted_results</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"amy; [amy@example.com]; [111-222-3333,333-444-5555]"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"carl; [carl@email.com,carl@example.com]; [444-555-6666]"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"james; []; [222-333-4444]"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"julia; [julia@example.com]; []"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl><span class=p>];</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=s2>"amy; ['amy@example.com']; ['111-222-3333', '333-444-5555']"</span><span class=p>,</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=s2>"carl; ['carl@email.com', 'carl@example.com']; ['444-555-6666']"</span><span class=p>,</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=s2>"james; []; ['222-333-4444']"</span><span class=p>,</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=s2>"julia; ['julia@example.com']; []"</span><span class=p>,</span></span></span></code></pre></div></div></div><h4 id=combine>4.2.4. Combine</h4><p><span class=language-java><a href=https://beam.apache.org/releases/javadoc/2.55.1/index.html?org/apache/beam/sdk/transforms/Combine.html><code>Combine</code></a></span> |
| <span class=language-py><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/transforms/core.py><code>Combine</code></a></span> |
| <span class=language-go><a href=https://github.com/apache/beam/blob/master/sdks/go/pkg/beam/combine.go#L27><code>Combine</code></a></span> |
| <span class=language-typescript><a href=https://github.com/apache/beam/blob/master/sdks/typescript/src/apache_beam/transforms/group_and_combine.ts><code>Combine</code></a></span> |
| is a Beam transform for combining collections of elements or values in your |
| data. <code>Combine</code> has variants that work on entire <code>PCollection</code>s, and some that |
| combine the values for each key in <code>PCollection</code>s of key/value pairs.</p><p>When you apply a <code>Combine</code> transform, you must provide the function that |
| contains the logic for combining the elements or values. The combining function |
| should be commutative and associative, as the function is not necessarily |
| invoked exactly once on all values with a given key. Because the input data |
| (including the value collection) may be distributed across multiple workers, the |
| combining function might be called multiple times to perform partial combining |
| on subsets of the value collection. The Beam SDK also provides some pre-built |
| combine functions for common numeric combination operations such as sum, min, |
| and max.</p><p>Simple combine operations, such as sums, can usually be implemented as a simple |
| function. More complex combination operations might require you to create a |
| <span class="language-java language-py">subclass of</span> <code>CombineFn</code> |
| that has an accumulation type distinct from the input/output type.</p><p>The associativity and commutativity of a <code>CombineFn</code> allows runners to |
| automatically apply some optimizations:</p><ul><li><strong>Combiner lifting</strong>: This is the most significant optimization. Input |
| elements are combined per key and window before they are shuffled, so the |
| volume of data shuffled might be reduced by many orders of magnitude. Another |
| term for this optimization is “mapper-side combine.”</li><li><strong>Incremental combining</strong>: When you have a <code>CombineFn</code> that reduces the data |
| size by a lot, it is useful to combine elements as they emerge from a |
| streaming shuffle. This spreads out the cost of doing combines over the time |
| that your streaming computation might be idle. Incremental combining also |
| reduces the storage of intermediate accumulators.</li></ul><h5 id=simple-combines>4.2.4.1. Simple combinations using simple functions</h5><span class=language-yaml>Beam YAML has the following buit-in CombineFns: count, sum, min, max, |
| mean, any, all, group, and concat. |
| CombineFns from other languages can also be referenced |
| as described in the (full docs on aggregation)[https://beam.apache.org/documentation/sdks/yaml-combine/].</span> |
| The following example code shows a simple combine function. |
| <span class=language-typescript>Combining is done by modifying a grouping transform with the `combining` method. |
| This method takes three parameters: the value to combine (either as a named |
| property of the input elements, or a function of the entire input), |
| the combining operation (either a binary function or a `CombineFn`), |
| and finally a name for the combined value in the output object.</span><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Sum a collection of Integer values. The function SumInts implements the interface SerializableFunction. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>public</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>SumInts</span> <span class=kd>implements</span> <span class=n>SerializableFunction</span><span class=o><</span><span class=n>Iterable</span><span class=o><</span><span class=n>Integer</span><span class=o>>,</span> <span class=n>Integer</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Integer</span> <span class=nf>apply</span><span class=o>(</span><span class=n>Iterable</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>input</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kt>int</span> <span class=n>sum</span> <span class=o>=</span> <span class=n>0</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=o>(</span><span class=kt>int</span> <span class=n>item</span> <span class=o>:</span> <span class=n>input</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>sum</span> <span class=o>+=</span> <span class=n>item</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>sum</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>pc</span> <span class=o>=</span> <span class=p>[</span><span class=mi>1</span><span class=p>,</span> <span class=mi>10</span><span class=p>,</span> <span class=mi>100</span><span class=p>,</span> <span class=mi>1000</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>def</span> <span class=nf>bounded_sum</span><span class=p>(</span><span class=n>values</span><span class=p>,</span> <span class=n>bound</span><span class=o>=</span><span class=mi>500</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>min</span><span class=p>(</span><span class=nb>sum</span><span class=p>(</span><span class=n>values</span><span class=p>),</span> <span class=n>bound</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>small_sum</span> <span class=o>=</span> <span class=n>pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CombineGlobally</span><span class=p>(</span><span class=n>bounded_sum</span><span class=p>)</span> <span class=c1># [500]</span> |
| </span></span><span class=line><span class=cl><span class=n>large_sum</span> <span class=o>=</span> <span class=n>pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CombineGlobally</span><span class=p>(</span><span class=n>bounded_sum</span><span class=p>,</span> <span class=n>bound</span><span class=o>=</span><span class=mi>5000</span><span class=p>)</span> <span class=c1># [1111]</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>sumInts</span><span class=p>(</span><span class=nx>a</span><span class=p>,</span> <span class=nx>v</span> <span class=kt>int</span><span class=p>)</span> <span class=kt>int</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>a</span> <span class=o>+</span> <span class=nx>v</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nf>Function2x1</span><span class=p>(</span><span class=nx>sumInts</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>globallySumInts</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>ints</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Combine</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>sumInts</span><span class=p>,</span> <span class=nx>ints</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>boundedSum</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Bound</span> <span class=kt>int</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>boundedSum</span><span class=p>)</span> <span class=nf>MergeAccumulators</span><span class=p>(</span><span class=nx>a</span><span class=p>,</span> <span class=nx>v</span> <span class=kt>int</span><span class=p>)</span> <span class=kt>int</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>sum</span> <span class=o>:=</span> <span class=nx>a</span> <span class=o>+</span> <span class=nx>v</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Bound</span> <span class=p>></span> <span class=mi>0</span> <span class=o>&&</span> <span class=nx>sum</span> <span class=p>></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Bound</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Bound</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>sum</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nx>Combiner1</span><span class=p>[</span><span class=kt>int</span><span class=p>](</span><span class=o>&</span><span class=nx>boundedSum</span><span class=p>{})</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>globallyBoundedSumInts</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>bound</span> <span class=kt>int</span><span class=p>,</span> <span class=nx>ints</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Combine</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>boundedSum</span><span class=p>{</span><span class=nx>Bound</span><span class=p>:</span> <span class=nx>bound</span><span class=p>},</span> <span class=nx>ints</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span><span class=mi>1</span><span class=p>,</span> <span class=mi>10</span><span class=p>,</span> <span class=mi>100</span><span class=p>,</span> <span class=mi>1000</span><span class=p>]));</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>pcoll</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>groupGlobally</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>combining</span><span class=p>((</span><span class=nx>c</span><span class=p>)</span> <span class=o>=></span> <span class=nx>c</span><span class=p>,</span> <span class=p>(</span><span class=nx>x</span><span class=p>,</span> <span class=nx>y</span><span class=p>)</span> <span class=o>=></span> <span class=nx>x</span> <span class=o>+</span> <span class=nx>y</span><span class=p>,</span> <span class=s2>"sum"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>combining</span><span class=p>((</span><span class=nx>c</span><span class=p>)</span> <span class=o>=></span> <span class=nx>c</span><span class=p>,</span> <span class=p>(</span><span class=nx>x</span><span class=p>,</span> <span class=nx>y</span><span class=p>)</span> <span class=o>=></span> <span class=nx>x</span> <span class=o>*</span> <span class=nx>y</span><span class=p>,</span> <span class=s2>"product"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>expected</span> <span class=o>=</span> <span class=p>{</span> <span class=nx>sum</span>: <span class=kt>1111</span><span class=p>,</span> <span class=nx>product</span>: <span class=kt>1000000</span> <span class=p>}</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Combine</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>group_by</span><span class=p>:</span><span class=w> </span><span class=l>animal</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>combine</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>biggest</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fn</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=s1>'apache_beam.transforms.combiners.TopCombineFn'</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>n</span><span class=p>:</span><span class=w> </span><span class=m>2</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>value</span><span class=p>:</span><span class=w> </span><span class=l>weight</span></span></span></code></pre></div></div></div><p class=language-go>All Combiners should be registered using a generic <code>register.CombinerX[...]</code> |
| function. This allows the Go SDK to infer an encoding from any inputs/outputs, |
| registers the Combiner for execution on remote runners, and optimizes the runtime |
| execution of the Combiner via reflection.</p><p class=language-go>Combiner1 should be used when your accumulator, input, and output are all of the |
| same type. It can be called with <code>register.Combiner1[T](&CustomCombiner{})</code> where <code>T</code> |
| is the type of the input/accumulator/output.</p><p class=language-go>Combiner2 should be used when your accumulator, input, and output are 2 distinct |
| types. It can be called with <code>register.Combiner2[T1, T2](&CustomCombiner{})</code> where |
| <code>T1</code> is the type of the accumulator and <code>T2</code> is the other type.</p><p class=language-go>Combiner3 should be used when your accumulator, input, and output are 3 distinct |
| types. It can be called with <code>register.Combiner3[T1, T2, T3](&CustomCombiner{})</code> |
| where <code>T1</code> is the type of the accumulator, <code>T2</code> is the type of the input, and <code>T3</code> is |
| the type of the output.</p><h5 id=advanced-combines>4.2.4.2. Advanced combinations using CombineFn</h5><p>For more complex combine functions, you can define a |
| <span class="language-java language-py">subclass of</span><code>CombineFn</code>. |
| You should use a <code>CombineFn</code> if the combine function requires a more sophisticated |
| accumulator, must perform additional pre- or post-processing, might change the |
| output type, or takes the key into account.</p><p>A general combining operation consists of five operations. When you create a |
| <span class="language-java language-py">subclass of</span> |
| <code>CombineFn</code>, you must provide five operations by overriding the |
| corresponding methods. Only <code>MergeAccumulators</code> is a required method. The |
| others will have a default interpretation based on the accumulator type. The |
| lifecycle methods are:</p><ol><li><p><strong>Create Accumulator</strong> creates a new “local” accumulator. In the example |
| case, taking a mean average, a local accumulator tracks the running sum of |
| values (the numerator value for our final average division) and the number of |
| values summed so far (the denominator value). It may be called any number of |
| times in a distributed fashion.</p></li><li><p><strong>Add Input</strong> adds an input element to an accumulator, returning the |
| accumulator value. In our example, it would update the sum and increment the |
| count. It may also be invoked in parallel.</p></li><li><p><strong>Merge Accumulators</strong> merges several accumulators into a single accumulator; |
| this is how data in multiple accumulators is combined before the final |
| calculation. In the case of the mean average computation, the accumulators |
| representing each portion of the division are merged together. It may be |
| called again on its outputs any number of times.</p></li><li><p><strong>Extract Output</strong> performs the final computation. In the case of computing a |
| mean average, this means dividing the combined sum of all the values by the |
| number of values summed. It is called once on the final, merged accumulator.</p></li><li><p><strong>Compact</strong> returns a more compact represenation of the accumulator. This is |
| called before an accumulator is sent across the wire, and can be useful in |
| cases where values are buffered or otherwise lazily kept unprocessed when |
| added to the accumulator. Compact should return an equivalent, though |
| possibly modified, accumulator. In most cases, Compact is not necessary. For |
| a real world example of using Compact, see the Python SDK implementation of |
| <a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/transforms/combiners.py#L523>TopCombineFn</a></p></li></ol><p>The following example code shows how to define a <code>CombineFn</code> that computes a |
| mean average:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>AverageFn</span> <span class=kd>extends</span> <span class=n>CombineFn</span><span class=o><</span><span class=n>Integer</span><span class=o>,</span> <span class=n>AverageFn</span><span class=o>.</span><span class=na>Accum</span><span class=o>,</span> <span class=n>Double</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>Accum</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kt>int</span> <span class=n>sum</span> <span class=o>=</span> <span class=n>0</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=kt>int</span> <span class=n>count</span> <span class=o>=</span> <span class=n>0</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Accum</span> <span class=nf>createAccumulator</span><span class=o>()</span> <span class=o>{</span> <span class=k>return</span> <span class=k>new</span> <span class=n>Accum</span><span class=o>();</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Accum</span> <span class=nf>addInput</span><span class=o>(</span><span class=n>Accum</span> <span class=n>accum</span><span class=o>,</span> <span class=n>Integer</span> <span class=n>input</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>accum</span><span class=o>.</span><span class=na>sum</span> <span class=o>+=</span> <span class=n>input</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=n>accum</span><span class=o>.</span><span class=na>count</span><span class=o>++;</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>accum</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Accum</span> <span class=nf>mergeAccumulators</span><span class=o>(</span><span class=n>Iterable</span><span class=o><</span><span class=n>Accum</span><span class=o>></span> <span class=n>accums</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>Accum</span> <span class=n>merged</span> <span class=o>=</span> <span class=n>createAccumulator</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=o>(</span><span class=n>Accum</span> <span class=n>accum</span> <span class=o>:</span> <span class=n>accums</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>merged</span><span class=o>.</span><span class=na>sum</span> <span class=o>+=</span> <span class=n>accum</span><span class=o>.</span><span class=na>sum</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=n>merged</span><span class=o>.</span><span class=na>count</span> <span class=o>+=</span> <span class=n>accum</span><span class=o>.</span><span class=na>count</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>merged</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Double</span> <span class=nf>extractOutput</span><span class=o>(</span><span class=n>Accum</span> <span class=n>accum</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=o>((</span><span class=kt>double</span><span class=o>)</span> <span class=n>accum</span><span class=o>.</span><span class=na>sum</span><span class=o>)</span> <span class=o>/</span> <span class=n>accum</span><span class=o>.</span><span class=na>count</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// No-op |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Accum</span> <span class=nf>compact</span><span class=o>(</span><span class=n>Accum</span> <span class=n>accum</span><span class=o>)</span> <span class=o>{</span> <span class=k>return</span> <span class=n>accum</span><span class=o>;</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>pc</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>AverageFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>CombineFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>create_accumulator</span><span class=p>(</span><span class=bp>self</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=p>(</span><span class=mf>0.0</span><span class=p>,</span> <span class=mi>0</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>add_input</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>sum_count</span><span class=p>,</span> <span class=nb>input</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=nb>sum</span><span class=p>,</span> <span class=n>count</span><span class=p>)</span> <span class=o>=</span> <span class=n>sum_count</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>sum</span> <span class=o>+</span> <span class=nb>input</span><span class=p>,</span> <span class=n>count</span> <span class=o>+</span> <span class=mi>1</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>merge_accumulators</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>accumulators</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>sums</span><span class=p>,</span> <span class=n>counts</span> <span class=o>=</span> <span class=nb>zip</span><span class=p>(</span><span class=o>*</span><span class=n>accumulators</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>sum</span><span class=p>(</span><span class=n>sums</span><span class=p>),</span> <span class=nb>sum</span><span class=p>(</span><span class=n>counts</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>extract_output</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>sum_count</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=nb>sum</span><span class=p>,</span> <span class=n>count</span><span class=p>)</span> <span class=o>=</span> <span class=n>sum_count</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>sum</span> <span class=o>/</span> <span class=n>count</span> <span class=k>if</span> <span class=n>count</span> <span class=k>else</span> <span class=nb>float</span><span class=p>(</span><span class=s1>'NaN'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>compact</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>accumulator</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># No-op</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>accumulator</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>averageFn</span> <span class=kd>struct</span><span class=p>{}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>averageAccum</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Count</span><span class=p>,</span> <span class=nx>Sum</span> <span class=kt>int</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>averageFn</span><span class=p>)</span> <span class=nf>CreateAccumulator</span><span class=p>()</span> <span class=nx>averageAccum</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>averageAccum</span><span class=p>{</span><span class=mi>0</span><span class=p>,</span> <span class=mi>0</span><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>averageFn</span><span class=p>)</span> <span class=nf>AddInput</span><span class=p>(</span><span class=nx>a</span> <span class=nx>averageAccum</span><span class=p>,</span> <span class=nx>v</span> <span class=kt>int</span><span class=p>)</span> <span class=nx>averageAccum</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>averageAccum</span><span class=p>{</span><span class=nx>Count</span><span class=p>:</span> <span class=nx>a</span><span class=p>.</span><span class=nx>Count</span> <span class=o>+</span> <span class=mi>1</span><span class=p>,</span> <span class=nx>Sum</span><span class=p>:</span> <span class=nx>a</span><span class=p>.</span><span class=nx>Sum</span> <span class=o>+</span> <span class=nx>v</span><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>averageFn</span><span class=p>)</span> <span class=nf>MergeAccumulators</span><span class=p>(</span><span class=nx>a</span><span class=p>,</span> <span class=nx>v</span> <span class=nx>averageAccum</span><span class=p>)</span> <span class=nx>averageAccum</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>averageAccum</span><span class=p>{</span><span class=nx>Count</span><span class=p>:</span> <span class=nx>a</span><span class=p>.</span><span class=nx>Count</span> <span class=o>+</span> <span class=nx>v</span><span class=p>.</span><span class=nx>Count</span><span class=p>,</span> <span class=nx>Sum</span><span class=p>:</span> <span class=nx>a</span><span class=p>.</span><span class=nx>Sum</span> <span class=o>+</span> <span class=nx>v</span><span class=p>.</span><span class=nx>Sum</span><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>averageFn</span><span class=p>)</span> <span class=nf>ExtractOutput</span><span class=p>(</span><span class=nx>a</span> <span class=nx>averageAccum</span><span class=p>)</span> <span class=kt>float64</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>a</span><span class=p>.</span><span class=nx>Count</span> <span class=o>==</span> <span class=mi>0</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>math</span><span class=p>.</span><span class=nf>NaN</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>float64</span><span class=p>(</span><span class=nx>a</span><span class=p>.</span><span class=nx>Sum</span><span class=p>)</span> <span class=o>/</span> <span class=nb>float64</span><span class=p>(</span><span class=nx>a</span><span class=p>.</span><span class=nx>Count</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>averageFn</span><span class=p>)</span> <span class=nf>Compact</span><span class=p>(</span><span class=nx>a</span> <span class=nx>averageAccum</span><span class=p>)</span> <span class=nx>averageAccum</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// No-op |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>a</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nx>Combiner3</span><span class=p>[</span><span class=nx>averageAccum</span><span class=p>,</span> <span class=kt>int</span><span class=p>,</span> <span class=kt>float64</span><span class=p>](</span><span class=o>&</span><span class=nx>averageFn</span><span class=p>{})</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>meanCombineFn</span>: <span class=kt>beam.CombineFn</span><span class=p><</span><span class=nt>number</span><span class=err>,</span> <span class=err>[</span><span class=na>number</span><span class=err>,</span> <span class=na>number</span><span class=err>],</span> <span class=na>number</span><span class=p>></span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>createAccumulator</span><span class=o>:</span> <span class=p>()</span> <span class=o>=></span> <span class=p>[</span><span class=mi>0</span><span class=p>,</span> <span class=mi>0</span><span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=nx>addInput</span><span class=o>:</span> <span class=p>([</span><span class=nx>sum</span><span class=p>,</span> <span class=nx>count</span><span class=p>]</span><span class=o>:</span> <span class=p>[</span><span class=kt>number</span><span class=p>,</span> <span class=kt>number</span><span class=p>],</span> <span class=nx>i</span>: <span class=kt>number</span><span class=p>)</span> <span class=o>=></span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=nx>sum</span> <span class=o>+</span> <span class=nx>i</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>count</span> <span class=o>+</span> <span class=mi>1</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=nx>mergeAccumulators</span><span class=o>:</span> <span class=p>(</span><span class=nx>accumulators</span><span class=o>:</span> <span class=p>[</span><span class=kt>number</span><span class=p>,</span> <span class=kt>number</span><span class=p>][])</span> <span class=o>=></span> |
| </span></span><span class=line><span class=cl> <span class=nx>accumulators</span><span class=p>.</span><span class=nx>reduce</span><span class=p>(([</span><span class=nx>sum0</span><span class=p>,</span> <span class=nx>count0</span><span class=p>],</span> <span class=p>[</span><span class=nx>sum1</span><span class=p>,</span> <span class=nx>count1</span><span class=p>])</span> <span class=o>=></span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=nx>sum0</span> <span class=o>+</span> <span class=nx>sum1</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>count0</span> <span class=o>+</span> <span class=nx>count1</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>]),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>extractOutput</span><span class=o>:</span> <span class=p>([</span><span class=nx>sum</span><span class=p>,</span> <span class=nx>count</span><span class=p>]</span><span class=o>:</span> <span class=p>[</span><span class=kt>number</span><span class=p>,</span> <span class=kt>number</span><span class=p>])</span> <span class=o>=></span> <span class=nx>sum</span> <span class=o>/</span> <span class=nx>count</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>};</span></span></span></code></pre></div></div></div><span class=language-go></span><h5 id=combining-pcollection>4.2.4.3. Combining a PCollection into a single value</h5><p>Use the global combine to transform all of the elements in a given <code>PCollection</code> |
| into a single value, represented in your pipeline as a new <code>PCollection</code> |
| containing one element. The following example code shows how to apply the Beam |
| provided sum combine function to produce a single sum value for a <code>PCollection</code> |
| of integers.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Sum.SumIntegerFn() combines the elements in the input PCollection. The resulting PCollection, called sum, |
| </span></span></span><span class=line><span class=cl><span class=c1>// contains one value: the sum of all the elements in the input PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>pc</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>sum</span> <span class=o>=</span> <span class=n>pc</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Combine</span><span class=o>.</span><span class=na>globally</span><span class=o>(</span><span class=k>new</span> <span class=n>Sum</span><span class=o>.</span><span class=na>SumIntegerFn</span><span class=o>()));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># sum combines the elements in the input PCollection.</span> |
| </span></span><span class=line><span class=cl><span class=c1># The resulting PCollection, called result, contains one value: the sum of all</span> |
| </span></span><span class=line><span class=cl><span class=c1># the elements in the input PCollection.</span> |
| </span></span><span class=line><span class=cl><span class=n>pc</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>average</span> <span class=o>=</span> <span class=n>pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CombineGlobally</span><span class=p>(</span><span class=n>AverageFn</span><span class=p>())</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>average</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Combine</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>averageFn</span><span class=p>{},</span> <span class=nx>ints</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span><span class=mi>4</span><span class=p>,</span> <span class=mi>5</span><span class=p>,</span> <span class=mi>6</span><span class=p>]));</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>pcoll</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>groupGlobally</span><span class=p>().</span><span class=nx>combining</span><span class=p>((</span><span class=nx>c</span><span class=p>)</span> <span class=o>=></span> <span class=nx>c</span><span class=p>,</span> <span class=nx>meanCombineFn</span><span class=p>,</span> <span class=s2>"mean"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Combine</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>group_by</span><span class=p>:</span><span class=w> </span><span class=p>[]</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>combine</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>weight</span><span class=p>:</span><span class=w> </span><span class=l>sum</span></span></span></code></pre></div></div></div><h5 id=combine-global-windowing>4.2.4.4. Combine and global windowing</h5><p>If your input <code>PCollection</code> uses the default global windowing, the default |
| behavior is to return a <code>PCollection</code> containing one item. That item’s value |
| comes from the accumulator in the combine function that you specified when |
| applying <code>Combine</code>. For example, the Beam provided sum combine function returns |
| a zero value (the sum of an empty input), while the min combine function returns |
| a maximal or infinite value.</p><p>To have <code>Combine</code> instead return an empty <code>PCollection</code> if the input is empty, |
| specify <code>.withoutDefaults</code> when you apply your <code>Combine</code> transform, as in the |
| following code example:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>pc</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>sum</span> <span class=o>=</span> <span class=n>pc</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Combine</span><span class=o>.</span><span class=na>globally</span><span class=o>(</span><span class=k>new</span> <span class=n>Sum</span><span class=o>.</span><span class=na>SumIntegerFn</span><span class=o>()).</span><span class=na>withoutDefaults</span><span class=o>());</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>pc</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=nb>sum</span> <span class=o>=</span> <span class=n>pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CombineGlobally</span><span class=p>(</span><span class=nb>sum</span><span class=p>)</span><span class=o>.</span><span class=n>without_defaults</span><span class=p>()</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>returnSideOrDefault</span><span class=p>(</span><span class=nx>d</span> <span class=kt>float64</span><span class=p>,</span> <span class=nx>iter</span> <span class=kd>func</span><span class=p>(</span><span class=o>*</span><span class=kt>float64</span><span class=p>)</span> <span class=kt>bool</span><span class=p>)</span> <span class=kt>float64</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>var</span> <span class=nx>c</span> <span class=kt>float64</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nf>iter</span><span class=p>(</span><span class=o>&</span><span class=nx>c</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Side input has a value, so return it. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>c</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Otherwise, return the default |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>d</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> <span class=nx>register</span><span class=p>.</span><span class=nf>Function2x1</span><span class=p>(</span><span class=nx>returnSideOrDefault</span><span class=p>)</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>globallyAverageWithDefault</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>ints</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Setting combine defaults has requires no helper function in the Go SDK. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>average</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Combine</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>averageFn</span><span class=p>{},</span> <span class=nx>ints</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// To add a default value: |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>defaultValue</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Create</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nb>float64</span><span class=p>(</span><span class=mi>0</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>returnSideOrDefault</span><span class=p>,</span> <span class=nx>defaultValue</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>SideInput</span><span class=p>{</span><span class=nx>Input</span><span class=p>:</span> <span class=nx>average</span><span class=p>})</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"alice"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>1.0</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"bob"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>0.99</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"eve"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>0.5</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"eve"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>0.25</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>pcoll</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>groupGlobally</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>combining</span><span class=p>(</span><span class=s2>"accuracy"</span><span class=p>,</span> <span class=nx>combiners</span><span class=p>.</span><span class=nx>mean</span><span class=p>,</span> <span class=s2>"mean"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>combining</span><span class=p>(</span><span class=s2>"accuracy"</span><span class=p>,</span> <span class=nx>combiners</span><span class=p>.</span><span class=nx>max</span><span class=p>,</span> <span class=s2>"max"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>expected</span> <span class=o>=</span> <span class=p>[{</span> <span class=nx>max</span>: <span class=kt>1.0</span><span class=p>,</span> <span class=nx>mean</span>: <span class=kt>0.685</span> <span class=p>}];</span></span></span></code></pre></div></div></div><h5 id=combine-non-global-windowing>4.2.4.5. Combine and non-global windowing</h5><span class="language-java language-py"><p>If your <code>PCollection</code> uses any non-global windowing function, Beam does not |
| provide the default behavior. You must specify one of the following options when |
| applying <code>Combine</code>:</p><ul><li>Specify <code>.withoutDefaults</code>, where windows that are empty in the input |
| <code>PCollection</code> will likewise be empty in the output collection.</li><li>Specify <code>.asSingletonView</code>, in which the output is immediately converted to a |
| <code>PCollectionView</code>, which will provide a default value for each empty window |
| when used as a side input. You’ll generally only need to use this option if |
| the result of your pipeline’s <code>Combine</code> is to be used as a side input later in |
| the pipeline.</li></ul></span><p class=language-go>If your <code>PCollection</code> uses any non-global windowing function, the Beam Go SDK |
| behaves the same way as with global windowing. Windows that are empty in the input |
| <code>PCollection</code> will likewise be empty in the output collection.</p><h5 id=combining-values-in-a-keyed-pcollection>4.2.4.6. Combining values in a keyed PCollection</h5><p>After creating a keyed PCollection (for example, by using a <code>GroupByKey</code> |
| transform), a common pattern is to combine the collection of values associated |
| with each key into a single, merged value. Drawing on the previous example from |
| <code>GroupByKey</code>, a key-grouped <code>PCollection</code> called <code>groupedWords</code> looks like this:</p><pre tabindex=0><code> cat, [1,5,9] |
| dog, [5,2] |
| and, [1,2,6] |
| jump, [3] |
| tree, [2] |
| ... |
| </code></pre><p>In the above <code>PCollection</code>, each element has a string key (for example, “cat”) |
| and an iterable of integers for its value (in the first element, containing [1, |
| 5, 9]). If our pipeline’s next processing step combines the values (rather than |
| considering them individually), you can combine the iterable of integers to |
| create a single, merged value to be paired with each key. This pattern of a |
| <code>GroupByKey</code> followed by merging the collection of values is equivalent to |
| Beam’s Combine PerKey transform. The combine function you supply to Combine |
| PerKey must be an associative reduction function or a |
| <span class="language-java language-py">subclass of</span> <code>CombineFn</code>.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// PCollection is grouped by key and the Double values associated with each key are combined into a Double. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Double</span><span class=o>>></span> <span class=n>salesRecords</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Double</span><span class=o>>></span> <span class=n>totalSalesPerPerson</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>salesRecords</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Combine</span><span class=o>.<</span><span class=n>String</span><span class=o>,</span> <span class=n>Double</span><span class=o>,</span> <span class=n>Double</span><span class=o>></span><span class=n>perKey</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=k>new</span> <span class=n>Sum</span><span class=o>.</span><span class=na>SumDoubleFn</span><span class=o>()));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// The combined value is of a different type than the original collection of values per key. PCollection has |
| </span></span></span><span class=line><span class=cl><span class=c1>// keys of type String and values of type Integer, and the combined value is a Double. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Integer</span><span class=o>>></span> <span class=n>playerAccuracy</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Double</span><span class=o>>></span> <span class=n>avgAccuracyPerPlayer</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>playerAccuracy</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Combine</span><span class=o>.<</span><span class=n>String</span><span class=o>,</span> <span class=n>Integer</span><span class=o>,</span> <span class=n>Double</span><span class=o>></span><span class=n>perKey</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=k>new</span> <span class=n>MeanInts</span><span class=o>())));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># PCollection is grouped by key and the numeric values associated with each key</span> |
| </span></span><span class=line><span class=cl><span class=c1># are averaged into a float.</span> |
| </span></span><span class=line><span class=cl><span class=n>player_accuracies</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>avg_accuracy_per_player</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>player_accuracies</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CombinePerKey</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>combiners</span><span class=o>.</span><span class=n>MeanCombineFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// PCollection is grouped by key and the numeric values associated with each key |
| </span></span></span><span class=line><span class=cl><span class=c1>// are averaged into a float64. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>playerAccuracies</span> <span class=o>:=</span> <span class=o>...</span> <span class=c1>// PCollection<string,int> |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=nx>avgAccuracyPerPlayer</span> <span class=o>:=</span> <span class=nx>stats</span><span class=p>.</span><span class=nf>MeanPerKey</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>playerAccuracies</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// avgAccuracyPerPlayer is a PCollection<string,float64> |
| </span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"alice"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>1.0</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"bob"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>0.99</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"eve"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>0.5</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"eve"</span><span class=p>,</span> <span class=nx>accuracy</span>: <span class=kt>0.25</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>pcoll</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>groupBy</span><span class=p>(</span><span class=s2>"player"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>combining</span><span class=p>(</span><span class=s2>"accuracy"</span><span class=p>,</span> <span class=nx>combiners</span><span class=p>.</span><span class=nx>mean</span><span class=p>,</span> <span class=s2>"mean"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>combining</span><span class=p>(</span><span class=s2>"accuracy"</span><span class=p>,</span> <span class=nx>combiners</span><span class=p>.</span><span class=nx>max</span><span class=p>,</span> <span class=s2>"max"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>expected</span> <span class=o>=</span> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"alice"</span><span class=p>,</span> <span class=nx>mean</span>: <span class=kt>1.0</span><span class=p>,</span> <span class=nx>max</span>: <span class=kt>1.0</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"bob"</span><span class=p>,</span> <span class=nx>mean</span>: <span class=kt>0.99</span><span class=p>,</span> <span class=nx>max</span>: <span class=kt>0.99</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>player</span><span class=o>:</span> <span class=s2>"eve"</span><span class=p>,</span> <span class=nx>mean</span>: <span class=kt>0.375</span><span class=p>,</span> <span class=nx>max</span>: <span class=kt>0.5</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl><span class=p>];</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Combine</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>group_by</span><span class=p>:</span><span class=w> </span><span class=p>[</span><span class=l>animal]</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>combine</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>total_weight</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fn</span><span class=p>:</span><span class=w> </span><span class=l>sum</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>value</span><span class=p>:</span><span class=w> </span><span class=l>weight</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>average_weight</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fn</span><span class=p>:</span><span class=w> </span><span class=l>mean</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>value</span><span class=p>:</span><span class=w> </span><span class=l>weight</span></span></span></code></pre></div></div></div><h4 id=flatten>4.2.5. Flatten</h4><p><span class=language-java><a href=https://beam.apache.org/releases/javadoc/2.55.1/index.html?org/apache/beam/sdk/transforms/Flatten.html><code>Flatten</code></a></span> |
| <span class=language-py><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/transforms/core.py><code>Flatten</code></a></span> |
| <span class=language-go><a href=https://github.com/apache/beam/blob/master/sdks/go/pkg/beam/flatten.go><code>Flatten</code></a></span> |
| <span class=language-typescript><code>Flatten</code></span> |
| is a Beam transform for <code>PCollection</code> objects that store the same data type. |
| <code>Flatten</code> merges multiple <code>PCollection</code> objects into a single logical |
| <code>PCollection</code>.</p><p>The following example shows how to apply a <code>Flatten</code> transform to merge multiple |
| <code>PCollection</code> objects.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Flatten takes a PCollectionList of PCollection objects of a given type. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Returns a single PCollection that contains all of the elements in the PCollection objects in that list. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>pc1</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>pc2</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>pc3</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollectionList</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>collections</span> <span class=o>=</span> <span class=n>PCollectionList</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>pc1</span><span class=o>).</span><span class=na>and</span><span class=o>(</span><span class=n>pc2</span><span class=o>).</span><span class=na>and</span><span class=o>(</span><span class=n>pc3</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>merged</span> <span class=o>=</span> <span class=n>collections</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Flatten</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>pCollections</span><span class=o>());</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># Flatten takes a tuple of PCollection objects.</span> |
| </span></span><span class=line><span class=cl><span class=c1># Returns a single PCollection that contains all of the elements in the PCollection objects in that tuple.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>merged</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=n>pcoll1</span><span class=p>,</span> <span class=n>pcoll2</span><span class=p>,</span> <span class=n>pcoll3</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1># A list of tuples can be "piped" directly into a Flatten transform.</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Flatten</span><span class=p>())</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// Flatten accepts any number of PCollections of the same element type. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Returns a single PCollection that contains all of the elements in input PCollections. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=nx>merged</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Flatten</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>pcol1</span><span class=p>,</span> <span class=nx>pcol2</span><span class=p>,</span> <span class=nx>pcol3</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=c1>// Flatten taken an array of PCollection objects, wrapped in beam.P(...) |
| </span></span></span><span class=line><span class=cl><span class=c1>// Returns a single PCollection that contains a union of all of the elements in all input PCollections. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>fib</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>withName</span><span class=p>(</span><span class=s2>"createFib"</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span><span class=mi>1</span><span class=p>,</span> <span class=mi>1</span><span class=p>,</span> <span class=mi>2</span><span class=p>,</span> <span class=mi>3</span><span class=p>,</span> <span class=mi>5</span><span class=p>,</span> <span class=mi>8</span><span class=p>]))</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pow</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>withName</span><span class=p>(</span><span class=s2>"createPow"</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span><span class=mi>1</span><span class=p>,</span> <span class=mi>2</span><span class=p>,</span> <span class=mi>4</span><span class=p>,</span> <span class=mi>8</span><span class=p>,</span> <span class=mi>16</span><span class=p>,</span> <span class=mi>32</span><span class=p>]))</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span> <span class=o>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>P</span><span class=p>([</span><span class=nx>fib</span><span class=p>,</span> <span class=nx>pow</span><span class=p>]).</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>flatten</span><span class=p>());</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Flatten</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=p>[</span><span class=l>SomeProducingTransform, AnotherProducingTransform]</span></span></span></code></pre></div></div></div><p class=language-yaml>In Beam YAML explicit flattens are not usually needed as one can list |
| multiple inputs for any transform which will be implicitly flattened.</p><h5 id=data-encoding-merged-collections>4.2.5.1. Data encoding in merged collections</h5><p>By default, the coder for the output <code>PCollection</code> is the same as the coder for |
| the first <code>PCollection</code> in the input <code>PCollectionList</code>. However, the input |
| <code>PCollection</code> objects can each use different coders, as long as they all contain |
| the same data type in your chosen language.</p><h5 id=merging-windowed-collections>4.2.5.2. Merging windowed collections</h5><p>When using <code>Flatten</code> to merge <code>PCollection</code> objects that have a windowing |
| strategy applied, all of the <code>PCollection</code> objects you want to merge must use a |
| compatible windowing strategy and window sizing. For example, all the |
| collections you’re merging must all use (hypothetically) identical 5-minute |
| fixed windows or 4-minute sliding windows starting every 30 seconds.</p><p>If your pipeline attempts to use <code>Flatten</code> to merge <code>PCollection</code> objects with |
| incompatible windows, Beam generates an <code>IllegalStateException</code> error when your |
| pipeline is constructed.</p><h4 id=partition>4.2.6. Partition</h4><p><span class=language-java><a href=https://beam.apache.org/releases/javadoc/2.55.1/index.html?org/apache/beam/sdk/transforms/Partition.html><code>Partition</code></a></span> |
| <span class=language-py><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/transforms/core.py><code>Partition</code></a></span> |
| <span class=language-go><a href=https://github.com/apache/beam/blob/master/sdks/go/pkg/beam/partition.go><code>Partition</code></a></span> |
| <span class=language-typescript><code>Partition</code></span> |
| is a Beam transform for <code>PCollection</code> objects that store the same data |
| type. <code>Partition</code> splits a single <code>PCollection</code> into a fixed number of smaller |
| collections.</p><p class=language-typescript>Often in the Typescript SDK the <code>Split</code> transform is more natural to use.</p><p><code>Partition</code> divides the elements of a <code>PCollection</code> according to a partitioning |
| function that you provide. The partitioning function contains the logic that |
| determines how to split up the elements of the input <code>PCollection</code> into each |
| resulting partition <code>PCollection</code>. The number of partitions must be determined |
| at graph construction time. You can, for example, pass the number of partitions |
| as a command-line option at runtime (which will then be used to build your |
| pipeline graph), but you cannot determine the number of partitions in |
| mid-pipeline (based on data calculated after your pipeline graph is constructed, |
| for instance).</p><p>The following example divides a <code>PCollection</code> into percentile groups.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Provide an int value with the desired number of result partitions, and a PartitionFn that represents the |
| </span></span></span><span class=line><span class=cl><span class=c1>// partitioning function. In this example, we define the PartitionFn in-line. Returns a PCollectionList |
| </span></span></span><span class=line><span class=cl><span class=c1>// containing each of the resulting partitions as individual PCollection objects. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>Student</span><span class=o>></span> <span class=n>students</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl><span class=c1>// Split students up into 10 partitions, by percentile: |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollectionList</span><span class=o><</span><span class=n>Student</span><span class=o>></span> <span class=n>studentsByPercentile</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>students</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Partition</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>10</span><span class=o>,</span> <span class=k>new</span> <span class=n>PartitionFn</span><span class=o><</span><span class=n>Student</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>int</span> <span class=nf>partitionFor</span><span class=o>(</span><span class=n>Student</span> <span class=n>student</span><span class=o>,</span> <span class=kt>int</span> <span class=n>numPartitions</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>student</span><span class=o>.</span><span class=na>getPercentile</span><span class=o>()</span> <span class=c1>// 0..99 |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>*</span> <span class=n>numPartitions</span> <span class=o>/</span> <span class=n>100</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}}));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// You can extract each partition from the PCollectionList using the get method, as follows: |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>Student</span><span class=o>></span> <span class=n>fortiethPercentile</span> <span class=o>=</span> <span class=n>studentsByPercentile</span><span class=o>.</span><span class=na>get</span><span class=o>(</span><span class=n>4</span><span class=o>);</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># Provide an int value with the desired number of result partitions, and a partitioning function (partition_fn in this example).</span> |
| </span></span><span class=line><span class=cl><span class=c1># Returns a tuple of PCollection objects containing each of the resulting partitions as individual PCollection objects.</span> |
| </span></span><span class=line><span class=cl><span class=n>students</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>def</span> <span class=nf>partition_fn</span><span class=p>(</span><span class=n>student</span><span class=p>,</span> <span class=n>num_partitions</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>int</span><span class=p>(</span><span class=n>get_percentile</span><span class=p>(</span><span class=n>student</span><span class=p>)</span> <span class=o>*</span> <span class=n>num_partitions</span> <span class=o>/</span> <span class=mi>100</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>by_decile</span> <span class=o>=</span> <span class=n>students</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Partition</span><span class=p>(</span><span class=n>partition_fn</span><span class=p>,</span> <span class=mi>10</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># You can extract each partition from the tuple of PCollection objects as follows:</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>fortieth_percentile</span> <span class=o>=</span> <span class=n>by_decile</span><span class=p>[</span><span class=mi>4</span><span class=p>]</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>decileFn</span><span class=p>(</span><span class=nx>student</span> <span class=nx>Student</span><span class=p>)</span> <span class=kt>int</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>int</span><span class=p>(</span><span class=nb>float64</span><span class=p>(</span><span class=nx>student</span><span class=p>.</span><span class=nx>Percentile</span><span class=p>)</span> <span class=o>/</span> <span class=nb>float64</span><span class=p>(</span><span class=mi>10</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nf>Function1x1</span><span class=p>(</span><span class=nx>decileFn</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Partition returns a slice of PCollections |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>studentsByPercentile</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Partition</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=mi>10</span><span class=p>,</span> <span class=nx>decileFn</span><span class=p>,</span> <span class=nx>students</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=c1>// Each partition can be extracted by indexing into the slice. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>fortiethPercentile</span> <span class=o>:=</span> <span class=nx>studentsByPercentile</span><span class=p>[</span><span class=mi>4</span><span class=p>]</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>deciles</span>: <span class=kt>PCollection</span><span class=p><</span><span class=nt>Student</span><span class=p>>[]</span> <span class=o>=</span> <span class=nx>students</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>partition</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=p>(</span><span class=nx>student</span><span class=p>,</span> <span class=nx>numPartitions</span><span class=p>)</span> <span class=o>=></span> |
| </span></span><span class=line><span class=cl> <span class=nb>Math</span><span class=p>.</span><span class=nx>floor</span><span class=p>((</span><span class=nx>getPercentile</span><span class=p>(</span><span class=nx>student</span><span class=p>)</span> <span class=o>/</span> <span class=mi>100</span><span class=p>)</span> <span class=o>*</span> <span class=nx>numPartitions</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=mi>10</span> |
| </span></span><span class=line><span class=cl> <span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>topDecile</span>: <span class=kt>PCollection</span><span class=p><</span><span class=nt>Student</span><span class=p>></span> <span class=o>=</span> <span class=nx>deciles</span><span class=p>[</span><span class=mi>9</span><span class=p>];</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Partition</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>by</span><span class=p>:</span><span class=w> </span><span class=l>str(percentile // 10)</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>outputs</span><span class=p>:</span><span class=w> </span><span class=p>[</span><span class=s2>"0"</span><span class=p>,</span><span class=w> </span><span class=s2>"1"</span><span class=p>,</span><span class=w> </span><span class=s2>"2"</span><span class=p>,</span><span class=w> </span><span class=s2>"3"</span><span class=p>,</span><span class=w> </span><span class=s2>"4"</span><span class=p>,</span><span class=w> </span><span class=s2>"5"</span><span class=p>,</span><span class=w> </span><span class=s2>"6"</span><span class=p>,</span><span class=w> </span><span class=s2>"7"</span><span class=p>,</span><span class=w> </span><span class=s2>"8"</span><span class=p>,</span><span class=w> </span><span class=s2>"9"</span><span class=p>,</span><span class=w> </span><span class=s2>"10"</span><span class=p>]</span></span></span></code></pre></div></div></div><p class=language-yaml>Note that in Beam YAML, <code>PCollections</code> are partitioned via string rather than integer values.</p><h3 id=requirements-for-writing-user-code-for-beam-transforms>4.3. Requirements for writing user code for Beam transforms</h3><p>When you build user code for a Beam transform, you should keep in mind the |
| distributed nature of execution. For example, there might be many copies of your |
| function running on a lot of different machines in parallel, and those copies |
| function independently, without communicating or sharing state with any of the |
| other copies. Depending on the Pipeline Runner and processing back-end you |
| choose for your pipeline, each copy of your user code function may be retried or |
| run multiple times. As such, you should be cautious about including things like |
| state dependency in your user code.</p><p>In general, your user code must fulfill at least these requirements:</p><ul><li>Your function object must be <strong>serializable</strong>.</li><li>Your function object must be <strong>thread-compatible</strong>, and be aware that <em>the |
| Beam SDKs are not thread-safe</em>.</li></ul><p>In addition, it’s recommended that you make your function object <strong>idempotent</strong>. |
| Non-idempotent functions are supported by Beam, but require additional |
| thought to ensure correctness when there are external side effects.</p><span class="language-java language-py"><blockquote><p><strong>Note:</strong> These requirements apply to subclasses of <code>DoFn</code>(a function object |
| used with the <a href=#pardo>ParDo</a> transform), <code>CombineFn</code> (a function object used |
| with the <a href=#combine>Combine</a> transform), and <code>WindowFn</code> (a function object |
| used with the <a href=#windowing>Window</a> transform).</p></blockquote></span><span class=language-go><blockquote><p><strong>Note:</strong> These requirements apply to <code>DoFn</code>s (a function object |
| used with the <a href=#pardo>ParDo</a> transform), <code>CombineFn</code>s (a function object used |
| with the <a href=#combine>Combine</a> transform), and <code>WindowFn</code>s (a function object |
| used with the <a href=#windowing>Window</a> transform).</p></blockquote></span><h4 id=user-code-serializability>4.3.1. Serializability</h4><p>Any function object you provide to a transform must be <strong>fully serializable</strong>. |
| This is because a copy of the function needs to be serialized and transmitted to |
| a remote worker in your processing cluster. |
| <span class="language-java language-py">The base classes for user code, such |
| as <code>DoFn</code>, <code>CombineFn</code>, and <code>WindowFn</code>, already implement <code>Serializable</code>; |
| however, your subclass must not add any non-serializable members.</span> |
| <span class=language-go>Funcs are serializable as long as |
| they are registered with <code>register.FunctionXxY</code> (for simple functions) or |
| <code>register.DoFnXxY</code> (for structural DoFns), and are not closures. Structural |
| <code>DoFn</code>s will have all exported fields serialized. Unexported fields are unable to |
| be serialized, and will be silently ignored.</span> |
| <span class=language-typescript>The Typescript SDK use <a href=https://github.com/nokia/ts-serialize-closures>ts-serialize-closures</a> |
| to serialize functions (and other objects). |
| This works out of the box for functions that are not closures, and also works |
| for closures as long as the function in question (and any closures it references) |
| are compiled with the |
| <a href=https://github.com/apache/beam/blob/master/sdks/typescript/tsconfig.json><code>ts-closure-transform</code> hooks</a> |
| (e.g. by using <code>ttsc</code> in place of <code>tsc</code>). |
| One can alternatively call |
| <code>requireForSerialization("importableModuleDefiningFunc", {func})</code> |
| to <a href=https://github.com/apache/beam/blob/master/sdks/typescript/src/apache_beam/serialization.ts>register a function directly</a> by name which can be less error-prone. |
| Note that if, as is often the case in Javascript, <code>func</code> returns objects that |
| contain closures, it is not sufficient to register <code>func</code> alone–its return |
| value must be registered if used.</span></p><p>Some other serializability factors you should keep in mind are:</p><ul><li><span class="language-java language-py">Transient</span><span class=language-go>Unexported</span> |
| fields in your function object are <em>not</em> transmitted to worker |
| instances, because they are not automatically serialized.</li><li>Avoid loading a field with a large amount of data before serialization.</li><li>Individual instances of your function object cannot share data.</li><li>Mutating a function object after it gets applied will have no effect.</li></ul><span class=language-java><blockquote><p><strong>Note:</strong> Take care when declaring your function object inline by using an anonymous |
| inner class instance. In a non-static context, your inner class instance will |
| implicitly contain a pointer to the enclosing class and that class’ state. |
| That enclosing class will also be serialized, and thus the same considerations |
| that apply to the function object itself also apply to this outer class.</p></blockquote></span><span class=language-go><blockquote><p><strong>Note:</strong> There’s no way to detect if a function is a closure. Closures will cause |
| runtime errors and pipeline failures. Avoid using anonymous functions when possible.</p></blockquote></span><h4 id=user-code-thread-compatibility>4.3.2. Thread-compatibility</h4><p>Your function object should be thread-compatible. Each instance of your function |
| object is accessed by a single thread at a time on a worker instance, unless you |
| explicitly create your own threads. Note, however, that <strong>the Beam SDKs are not |
| thread-safe</strong>. If you create your own threads in your user code, you must |
| provide your own synchronization. <span class=language-java>Note that static members in your function |
| object are not passed to worker instances and that multiple instances of your |
| function may be accessed from different threads.</span></p><h4 id=user-code-idempotence>4.3.3. Idempotence</h4><p>It’s recommended that you make your function object idempotent–that is, that it |
| can be repeated or retried as often as necessary without causing unintended side |
| effects. Non-idempotent functions are supported, however the Beam model provides |
| no guarantees as to the number of times your user code might be invoked or retried; |
| as such, keeping your function object idempotent keeps your pipeline’s output |
| deterministic, and your transforms’ behavior more predictable and easier to debug.</p><h3 id=side-inputs>4.4. Side inputs</h3><p>In addition to the main input <code>PCollection</code>, you can provide additional inputs |
| to a <code>ParDo</code> transform in the form of side inputs. A side input is an additional |
| input that your <code>DoFn</code> can access each time it processes an element in the input |
| <code>PCollection</code>. When you specify a side input, you create a view of some other |
| data that can be read from within the <code>ParDo</code> transform’s <code>DoFn</code> while processing |
| each element.</p><p>Side inputs are useful if your <code>ParDo</code> needs to inject additional data when |
| processing each element in the input <code>PCollection</code>, but the additional data |
| needs to be determined at runtime (and not hard-coded). Such values might be |
| determined by the input data, or depend on a different branch of your pipeline.</p><p class=language-go>All side input iterables should be registered using a generic <code>register.IterX[...]</code> |
| function. This optimizes runtime execution of the iterable.</p><h4 id=side-inputs-pardo>4.4.1. Passing side inputs to ParDo</h4><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=c1>// Pass side inputs to your ParDo transform by invoking .withSideInputs. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Inside your DoFn, access the side input by using the method DoFn.ProcessContext.sideInput. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The input PCollection to ParDo. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>words</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// A PCollection of word lengths that we'll combine into a single value. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>wordLengths</span> <span class=o>=</span> <span class=o>...;</span> <span class=c1>// Singleton PCollection |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Create a singleton PCollectionView from wordLengths using Combine.globally and View.asSingleton. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>final</span> <span class=n>PCollectionView</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>maxWordLengthCutOffView</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>wordLengths</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Combine</span><span class=o>.</span><span class=na>globally</span><span class=o>(</span><span class=k>new</span> <span class=n>Max</span><span class=o>.</span><span class=na>MaxIntFn</span><span class=o>()).</span><span class=na>asSingletonView</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Apply a ParDo that takes maxWordLengthCutOffView as a side input. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>wordsBelowCutOff</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>words</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>out</span><span class=o>,</span> <span class=n>ProcessContext</span> <span class=n>c</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// In our DoFn, access the side input. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kt>int</span> <span class=n>lengthCutOff</span> <span class=o>=</span> <span class=n>c</span><span class=o>.</span><span class=na>sideInput</span><span class=o>(</span><span class=n>maxWordLengthCutOffView</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(</span><span class=n>word</span><span class=o>.</span><span class=na>length</span><span class=o>()</span> <span class=o><=</span> <span class=n>lengthCutOff</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>out</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>word</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}).</span><span class=na>withSideInputs</span><span class=o>(</span><span class=n>maxWordLengthCutOffView</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>);</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># Side inputs are available as extra arguments in the DoFn's process method or Map / FlatMap's callable.</span> |
| </span></span><span class=line><span class=cl><span class=c1># Optional, positional, and keyword arguments are all supported. Deferred arguments are unwrapped into their</span> |
| </span></span><span class=line><span class=cl><span class=c1># actual values. For example, using pvalue.AsIteor(pcoll) at pipeline construction time results in an iterable</span> |
| </span></span><span class=line><span class=cl><span class=c1># of the actual elements of pcoll being passed into each process invocation. In this example, side inputs are</span> |
| </span></span><span class=line><span class=cl><span class=c1># passed to a FlatMap transform as extra arguments and consumed by filter_using_length.</span> |
| </span></span><span class=line><span class=cl><span class=n>words</span> <span class=o>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Callable takes additional arguments.</span> |
| </span></span><span class=line><span class=cl><span class=k>def</span> <span class=nf>filter_using_length</span><span class=p>(</span><span class=n>word</span><span class=p>,</span> <span class=n>lower_bound</span><span class=p>,</span> <span class=n>upper_bound</span><span class=o>=</span><span class=nb>float</span><span class=p>(</span><span class=s1>'inf'</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>lower_bound</span> <span class=o><=</span> <span class=nb>len</span><span class=p>(</span><span class=n>word</span><span class=p>)</span> <span class=o><=</span> <span class=n>upper_bound</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>word</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Construct a deferred side input.</span> |
| </span></span><span class=line><span class=cl><span class=n>avg_word_len</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>words</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=nb>len</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CombineGlobally</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>combiners</span><span class=o>.</span><span class=n>MeanCombineFn</span><span class=p>()))</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Call with explicit side inputs.</span> |
| </span></span><span class=line><span class=cl><span class=n>small_words</span> <span class=o>=</span> <span class=n>words</span> <span class=o>|</span> <span class=s1>'small'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>FlatMap</span><span class=p>(</span><span class=n>filter_using_length</span><span class=p>,</span> <span class=mi>0</span><span class=p>,</span> <span class=mi>3</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># A single deferred side input.</span> |
| </span></span><span class=line><span class=cl><span class=n>larger_than_average</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>words</span> <span class=o>|</span> <span class=s1>'large'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>FlatMap</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>filter_using_length</span><span class=p>,</span> <span class=n>lower_bound</span><span class=o>=</span><span class=n>pvalue</span><span class=o>.</span><span class=n>AsSingleton</span><span class=p>(</span><span class=n>avg_word_len</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Mix and match.</span> |
| </span></span><span class=line><span class=cl><span class=n>small_but_nontrivial</span> <span class=o>=</span> <span class=n>words</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>FlatMap</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>filter_using_length</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>lower_bound</span><span class=o>=</span><span class=mi>2</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>upper_bound</span><span class=o>=</span><span class=n>pvalue</span><span class=o>.</span><span class=n>AsSingleton</span><span class=p>(</span><span class=n>avg_word_len</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># We can also pass side inputs to a ParDo transform, which will get passed to its process method.</span> |
| </span></span><span class=line><span class=cl><span class=c1># The first two arguments for the process method would be self and element.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>FilterUsingLength</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>lower_bound</span><span class=p>,</span> <span class=n>upper_bound</span><span class=o>=</span><span class=nb>float</span><span class=p>(</span><span class=s1>'inf'</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>lower_bound</span> <span class=o><=</span> <span class=nb>len</span><span class=p>(</span><span class=n>element</span><span class=p>)</span> <span class=o><=</span> <span class=n>upper_bound</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>element</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>small_words</span> <span class=o>=</span> <span class=n>words</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>FilterUsingLength</span><span class=p>(),</span> <span class=mi>0</span><span class=p>,</span> <span class=mi>3</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=o>...</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// Side inputs are provided using `beam.SideInput` in the DoFn's ProcessElement method. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Side inputs can be arbitrary PCollections, which can then be iterated over per element |
| </span></span></span><span class=line><span class=cl><span class=c1>// in a DoFn. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Side input parameters appear after main input elements, and before any output emitters. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>words</span> <span class=p>=</span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// avgWordLength is a PCollection containing a single element, a singleton. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>avgWordLength</span> <span class=o>:=</span> <span class=nx>stats</span><span class=p>.</span><span class=nf>Mean</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>wordLengths</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Side inputs are added as with the beam.SideInput option to beam.ParDo. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>wordsAboveCutOff</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>filterWordsAbove</span><span class=p>,</span> <span class=nx>words</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>SideInput</span><span class=p>{</span><span class=nx>Input</span><span class=p>:</span> <span class=nx>avgWordLength</span><span class=p>})</span> |
| </span></span><span class=line><span class=cl><span class=nx>wordsBelowCutOff</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>filterWordsBelow</span><span class=p>,</span> <span class=nx>words</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>SideInput</span><span class=p>{</span><span class=nx>Input</span><span class=p>:</span> <span class=nx>avgWordLength</span><span class=p>})</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// filterWordsAbove is a DoFn that takes in a word, |
| </span></span></span><span class=line><span class=cl><span class=c1>// and a singleton side input iterator as of a length cut off |
| </span></span></span><span class=line><span class=cl><span class=c1>// and only emits words that are beneath that cut off. |
| </span></span></span><span class=line><span class=cl><span class=c1>// |
| </span></span></span><span class=line><span class=cl><span class=c1>// If the iterator has no elements, an error is returned, aborting processing. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>filterWordsAbove</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>lengthCutOffIter</span> <span class=kd>func</span><span class=p>(</span><span class=o>*</span><span class=kt>float64</span><span class=p>)</span> <span class=kt>bool</span><span class=p>,</span> <span class=nx>emitAboveCutoff</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>var</span> <span class=nx>cutOff</span> <span class=kt>float64</span> |
| </span></span><span class=line><span class=cl> <span class=nx>ok</span> <span class=o>:=</span> <span class=nf>lengthCutOffIter</span><span class=p>(</span><span class=o>&</span><span class=nx>cutOff</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>!</span><span class=nx>ok</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>fmt</span><span class=p>.</span><span class=nf>Errorf</span><span class=p>(</span><span class=s>"no length cutoff provided"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nb>float64</span><span class=p>(</span><span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>))</span> <span class=p>></span> <span class=nx>cutOff</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitAboveCutoff</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// filterWordsBelow is a DoFn that takes in a word, |
| </span></span></span><span class=line><span class=cl><span class=c1>// and a singleton side input of a length cut off |
| </span></span></span><span class=line><span class=cl><span class=c1>// and only emits words that are beneath that cut off. |
| </span></span></span><span class=line><span class=cl><span class=c1>// |
| </span></span></span><span class=line><span class=cl><span class=c1>// If the side input isn't a singleton, a runtime panic will occur. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>filterWordsBelow</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>lengthCutOff</span> <span class=kt>float64</span><span class=p>,</span> <span class=nx>emitBelowCutoff</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nb>float64</span><span class=p>(</span><span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>))</span> <span class=o><=</span> <span class=nx>lengthCutOff</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitBelowCutoff</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nf>Function3x1</span><span class=p>(</span><span class=nx>filterWordsAbove</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nf>Function3x0</span><span class=p>(</span><span class=nx>filterWordsBelow</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 1 input of type string => Emitter1[string] |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>Emitter1</span><span class=p>[</span><span class=kt>string</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 1 input of type float64 => Iter1[float64] |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>Iter1</span><span class=p>[</span><span class=kt>float64</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// The Go SDK doesn't support custom ViewFns. |
| </span></span></span><span class=line><span class=cl><span class=c1>// See https://github.com/apache/beam/issues/18602 for details |
| </span></span></span><span class=line><span class=cl><span class=c1>// on how to contribute them! |
| </span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=c1>// Side inputs are provided by passing an extra context object to |
| </span></span></span><span class=line><span class=cl><span class=c1>// `map`, `flatMap`, or `parDo` transforms. This object will get passed as an |
| </span></span></span><span class=line><span class=cl><span class=c1>// extra argument to the provided function (or `process` method of the `DoFn`). |
| </span></span></span><span class=line><span class=cl><span class=c1>// `SideInputParam` properties (generally created with `pardo.xxxSideInput(...)`) |
| </span></span></span><span class=line><span class=cl><span class=c1>// have a `lookup` method that can be invoked from within the process method. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=c1>// Let words be a PCollection of strings. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kr>const</span> <span class=nx>words</span> : <span class=kt>PCollection</span><span class=p><</span><span class=nt>string</span><span class=p>></span> <span class=o>=</span> <span class=p>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// meanLengthPColl will contain a single number whose value is the |
| </span></span></span><span class=line><span class=cl><span class=c1>// average length of the words |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kr>const</span> <span class=nx>meanLengthPColl</span>: <span class=kt>PCollection</span><span class=p><</span><span class=nt>number</span><span class=p>></span> <span class=o>=</span> <span class=nx>words</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>groupGlobally</span><span class=p><</span><span class=nt>string</span><span class=p>>()</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>combining</span><span class=p>((</span><span class=nx>word</span><span class=p>)</span> <span class=o>=></span> <span class=nx>word</span><span class=p>.</span><span class=nx>length</span><span class=p>,</span> <span class=nx>combiners</span><span class=p>.</span><span class=nx>mean</span><span class=p>,</span> <span class=s2>"mean"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>map</span><span class=p>(({</span> <span class=nx>mean</span> <span class=p>})</span> <span class=o>=></span> <span class=nx>mean</span><span class=p>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Now we use this as a side input to yield only words that are |
| </span></span></span><span class=line><span class=cl><span class=c1>// smaller than average. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kr>const</span> <span class=nx>smallWords</span> <span class=o>=</span> <span class=nx>words</span><span class=p>.</span><span class=nx>flatMap</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// This is the function, taking context as a second argument. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>function</span><span class=o>*</span> <span class=nx>keepSmall</span><span class=p>(</span><span class=nx>word</span><span class=p>,</span> <span class=nx>context</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>(</span><span class=nx>word</span><span class=p>.</span><span class=nx>length</span> <span class=o><</span> <span class=nx>context</span><span class=p>.</span><span class=nx>meanLength</span><span class=p>.</span><span class=nx>lookup</span><span class=p>())</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=nx>word</span><span class=p>;</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// This is the context that will be passed as a second argument. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>{</span> <span class=nx>meanLength</span>: <span class=kt>pardo.singletonSideInput</span><span class=p>(</span><span class=nx>meanLengthPColl</span><span class=p>)</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span></span></span></code></pre></div></div></div><h4 id=side-inputs-windowing>4.4.2. Side inputs and windowing</h4><p>A windowed <code>PCollection</code> may be infinite and thus cannot be compressed into a |
| single value (or single collection class). When you create a <code>PCollectionView</code> |
| of a windowed <code>PCollection</code>, the <code>PCollectionView</code> represents a single entity |
| per window (one singleton per window, one list per window, etc.).</p><p>Beam uses the window(s) for the main input element to look up the appropriate |
| window for the side input element. Beam projects the main input element’s window |
| into the side input’s window set, and then uses the side input from the |
| resulting window. If the main input and side inputs have identical windows, the |
| projection provides the exact corresponding window. However, if the inputs have |
| different windows, Beam uses the projection to choose the most appropriate side |
| input window.</p><p>For example, if the main input is windowed using fixed-time windows of one |
| minute, and the side input is windowed using fixed-time windows of one hour, |
| Beam projects the main input window against the side input window set and |
| selects the side input value from the appropriate hour-long side input window.</p><p>If the main input element exists in more than one window, then <code>processElement</code> |
| gets called multiple times, once for each window. Each call to <code>processElement</code> |
| projects the “current” window for the main input element, and thus might provide |
| a different view of the side input each time.</p><p>If the side input has multiple trigger firings, Beam uses the value from the |
| latest trigger firing. This is particularly useful if you use a side input with |
| a single global window and specify a trigger.</p><h3 id=additional-outputs>4.5. Additional outputs</h3><p class="language-java language-py">While <code>ParDo</code> always produces a main output <code>PCollection</code> (as the return value |
| from <code>apply</code>), you can also have your <code>ParDo</code> produce any number of additional |
| output <code>PCollection</code>s. If you choose to have multiple outputs, your <code>ParDo</code> |
| returns all of the output <code>PCollection</code>s (including the main output) bundled |
| together.</p><p class=language-go>While <code>beam.ParDo</code> always produces an output <code>PCollection</code>, your <code>DoFn</code> can produce any |
| number of additional output <code>PCollections</code>s, or even none at all. |
| If you choose to have multiple outputs, your <code>DoFn</code> needs to be called with the <code>ParDo</code> |
| function that matches the number of outputs. <code>beam.ParDo2</code> for two output <code>PCollection</code>s, |
| <code>beam.ParDo3</code> for three and so on until <code>beam.ParDo7</code>. If you need more, you can |
| use <code>beam.ParDoN</code> which will return a <code>[]beam.PCollection</code>.</p><p class=language-typescript>While <code>ParDo</code> always produces a main output <code>PCollection</code> (as the return value |
| from <code>apply</code>). If you want to have multiple outputs, emit an object with distinct |
| properties in your <code>ParDo</code> operation and follow this operation with a <code>Split</code> |
| to break it into multiple <code>PCollection</code>s.</p><p class=language-yaml>In Beam YAML, one obtains multiple outputs by emitting all outputs to a single |
| <code>PCollection</code>, possibly with an extra field, and then using <code>Partition</code> to |
| split this single <code>PCollection</code> into multiple distinct <code>PCollection</code> |
| outputs.</p><h4 id=output-tags>4.5.1. Tags for multiple outputs</h4><p class=language-typescript>The <code>Split</code> PTransform will take a PCollection of elements of the form |
| <code>{tagA?: A, tagB?: B, ...}</code> and return a object |
| <code>{tagA: PCollection<A>, tagB: PCollection<B>, ...}</code>. |
| The set of expected tags is passed to the operation; how multiple or |
| unknown tags are handled can be specified by passing a non-default |
| <code>SplitOptions</code> instance.</p><p class=language-go>The Go SDK doesn’t use output tags, and instead uses positional ordering for |
| multiple output PCollections.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// To emit elements to multiple output PCollections, create a TupleTag object to identify each collection |
| </span></span></span><span class=line><span class=cl><span class=c1>// that your ParDo produces. For example, if your ParDo produces three output PCollections (the main output |
| </span></span></span><span class=line><span class=cl><span class=c1>// and two additional outputs), you must create three TupleTags. The following example code shows how to |
| </span></span></span><span class=line><span class=cl><span class=c1>// create TupleTags for a ParDo with three output PCollections. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Input PCollection to our ParDo. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>words</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// The ParDo will filter words whose length is below a cutoff and add them to |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// the main output PCollection<String>. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// If a word is above the cutoff, the ParDo will add the word length to an |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// output PCollection<Integer>. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// If a word starts with the string "MARKER", the ParDo will add that word to an |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// output PCollection<String>. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>final</span> <span class=kt>int</span> <span class=n>wordLengthCutOff</span> <span class=o>=</span> <span class=n>10</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Create three TupleTags, one for each output PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Output that contains words below the length cutoff. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>final</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>wordsBelowCutOffTag</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=k>new</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>String</span><span class=o>>(){};</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output that contains word lengths. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>final</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>wordLengthsAboveCutOffTag</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=k>new</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>Integer</span><span class=o>>(){};</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output that contains "MARKER" words. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>final</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>markedWordsTag</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=k>new</span> <span class=n>TupleTag</span><span class=o><</span><span class=n>String</span><span class=o>>(){};</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Passing Output Tags to ParDo: |
| </span></span></span><span class=line><span class=cl><span class=c1>// After you specify the TupleTags for each of your ParDo outputs, pass the tags to your ParDo by invoking |
| </span></span></span><span class=line><span class=cl><span class=c1>// .withOutputTags. You pass the tag for the main output first, and then the tags for any additional outputs |
| </span></span></span><span class=line><span class=cl><span class=c1>// in a TupleTagList. Building on our previous example, we pass the three TupleTags for our three output |
| </span></span></span><span class=line><span class=cl><span class=c1>// PCollections to our ParDo. Note that all of the outputs (including the main output PCollection) are |
| </span></span></span><span class=line><span class=cl><span class=c1>// bundled into the returned PCollectionTuple. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollectionTuple</span> <span class=n>results</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>words</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// DoFn continues here. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>})</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Specify the tag for the main output. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>.</span><span class=na>withOutputTags</span><span class=o>(</span><span class=n>wordsBelowCutOffTag</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Specify the tags for the two additional outputs as a TupleTagList. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>TupleTagList</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>wordLengthsAboveCutOffTag</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>and</span><span class=o>(</span><span class=n>markedWordsTag</span><span class=o>)));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># To emit elements to multiple output PCollections, invoke with_outputs() on the ParDo, and specify the</span> |
| </span></span><span class=line><span class=cl><span class=c1># expected tags for the outputs. with_outputs() returns a DoOutputsTuple object. Tags specified in</span> |
| </span></span><span class=line><span class=cl><span class=c1># with_outputs are attributes on the returned DoOutputsTuple object. The tags give access to the</span> |
| </span></span><span class=line><span class=cl><span class=c1># corresponding output PCollections.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>results</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>words</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>ProcessWords</span><span class=p>(),</span> <span class=n>cutoff_length</span><span class=o>=</span><span class=mi>2</span><span class=p>,</span> <span class=n>marker</span><span class=o>=</span><span class=s1>'x'</span><span class=p>)</span><span class=o>.</span><span class=n>with_outputs</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'above_cutoff_lengths'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'marked strings'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>main</span><span class=o>=</span><span class=s1>'below_cutoff_strings'</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=n>below</span> <span class=o>=</span> <span class=n>results</span><span class=o>.</span><span class=n>below_cutoff_strings</span> |
| </span></span><span class=line><span class=cl><span class=n>above</span> <span class=o>=</span> <span class=n>results</span><span class=o>.</span><span class=n>above_cutoff_lengths</span> |
| </span></span><span class=line><span class=cl><span class=n>marked</span> <span class=o>=</span> <span class=n>results</span><span class=p>[</span><span class=s1>'marked strings'</span><span class=p>]</span> <span class=c1># indexing works as well</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># The result is also iterable, ordered in the same order that the tags were passed to with_outputs(),</span> |
| </span></span><span class=line><span class=cl><span class=c1># the main tag (if specified) first.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>below</span><span class=p>,</span> <span class=n>above</span><span class=p>,</span> <span class=n>marked</span> <span class=o>=</span> <span class=p>(</span><span class=n>words</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>ProcessWords</span><span class=p>(),</span> <span class=n>cutoff_length</span><span class=o>=</span><span class=mi>2</span><span class=p>,</span> <span class=n>marker</span><span class=o>=</span><span class=s1>'x'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=n>with_outputs</span><span class=p>(</span><span class=s1>'above_cutoff_lengths'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'marked strings'</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>main</span><span class=o>=</span><span class=s1>'below_cutoff_strings'</span><span class=p>))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// beam.ParDo3 returns PCollections in the same order as |
| </span></span></span><span class=line><span class=cl><span class=c1>// the emit function parameters in processWords. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>below</span><span class=p>,</span> <span class=nx>above</span><span class=p>,</span> <span class=nx>marked</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo3</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>processWords</span><span class=p>,</span> <span class=nx>words</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// processWordsMixed uses both a standard return and an emitter function. |
| </span></span></span><span class=line><span class=cl><span class=c1>// The standard return produces the first PCollection from beam.ParDo2, |
| </span></span></span><span class=line><span class=cl><span class=c1>// and the emitter produces the second PCollection. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>length</span><span class=p>,</span> <span class=nx>mixedMarked</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo2</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>processWordsMixed</span><span class=p>,</span> <span class=nx>words</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=err>#</span> <span class=nx>Create</span> <span class=nx>three</span> <span class=nx>PCollections</span> <span class=kr>from</span> <span class=nx>a</span> <span class=nx>single</span> <span class=nx>input</span> <span class=nx>PCollection</span><span class=p>.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=p>{</span> <span class=nx>below</span><span class=p>,</span> <span class=nx>above</span><span class=p>,</span> <span class=nx>marked</span> <span class=p>}</span> <span class=o>=</span> <span class=nx>to_split</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>split</span><span class=p>([</span><span class=s2>"below"</span><span class=p>,</span> <span class=s2>"above"</span><span class=p>,</span> <span class=s2>"marked"</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span></span></span></code></pre></div></div></div><h4 id=multiple-outputs-dofn>4.5.2. Emitting to multiple outputs in your DoFn</h4><p class=language-go>Call emitter functions as needed to produce 0 or more elements for its matching |
| <code>PCollection</code>. The same value can be emitted with multiple emitters. |
| As normal, do not mutate values after emitting them from any emitter.</p><p class=language-go>All emitters should be registered using a generic <code>register.EmitterX[...]</code> |
| function. This optimizes runtime execution of the emitter.</p><p class=language-go>DoFns can also return a single element via the standard return. |
| The standard return is always the first PCollection returned from beam.ParDo. |
| Other emitters output to their own PCollections in their defined parameter order.</p><p class=language-yaml><code>MapToFields</code> is always one-to-one. To perform a one-to-many mapping one can |
| first map a field to an iterable type and then follow this transform with an |
| <code>Explode</code> transform that will emit multiple values, one per value of the |
| exploded field.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Inside your ParDo's DoFn, you can emit an element to a specific output PCollection by providing a |
| </span></span></span><span class=line><span class=cl><span class=c1>// MultiOutputReceiver to your process method, and passing in the appropriate TupleTag to obtain an OutputReceiver. |
| </span></span></span><span class=line><span class=cl><span class=c1>// After your ParDo, extract the resulting output PCollections from the returned PCollectionTuple. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Based on the previous example, this shows the DoFn emitting to the main output and two additional outputs. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=n>MultiOutputReceiver</span> <span class=n>out</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(</span><span class=n>word</span><span class=o>.</span><span class=na>length</span><span class=o>()</span> <span class=o><=</span> <span class=n>wordLengthCutOff</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Emit short word to the main output. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// In this example, it is the output with tag wordsBelowCutOffTag. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>out</span><span class=o>.</span><span class=na>get</span><span class=o>(</span><span class=n>wordsBelowCutOffTag</span><span class=o>).</span><span class=na>output</span><span class=o>(</span><span class=n>word</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> <span class=k>else</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Emit long word length to the output with tag wordLengthsAboveCutOffTag. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>out</span><span class=o>.</span><span class=na>get</span><span class=o>(</span><span class=n>wordLengthsAboveCutOffTag</span><span class=o>).</span><span class=na>output</span><span class=o>(</span><span class=n>word</span><span class=o>.</span><span class=na>length</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(</span><span class=n>word</span><span class=o>.</span><span class=na>startsWith</span><span class=o>(</span><span class=s>"MARKER"</span><span class=o>))</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Emit word to the output with tag markedWordsTag. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>out</span><span class=o>.</span><span class=na>get</span><span class=o>(</span><span class=n>markedWordsTag</span><span class=o>).</span><span class=na>output</span><span class=o>(</span><span class=n>word</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># Inside your ParDo's DoFn, you can emit an element to a specific output by wrapping the value and the output tag (str).</span> |
| </span></span><span class=line><span class=cl><span class=c1># using the pvalue.OutputValue wrapper class.</span> |
| </span></span><span class=line><span class=cl><span class=c1># Based on the previous example, this shows the DoFn emitting to the main output and two additional outputs.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>ProcessWords</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>cutoff_length</span><span class=p>,</span> <span class=n>marker</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nb>len</span><span class=p>(</span><span class=n>element</span><span class=p>)</span> <span class=o><=</span> <span class=n>cutoff_length</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Emit this short word to the main output.</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>element</span> |
| </span></span><span class=line><span class=cl> <span class=k>else</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Emit this word's long length to the 'above_cutoff_lengths' output.</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>pvalue</span><span class=o>.</span><span class=n>TaggedOutput</span><span class=p>(</span><span class=s1>'above_cutoff_lengths'</span><span class=p>,</span> <span class=nb>len</span><span class=p>(</span><span class=n>element</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>element</span><span class=o>.</span><span class=n>startswith</span><span class=p>(</span><span class=n>marker</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Emit this word to a different output with the 'marked strings' tag.</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>pvalue</span><span class=o>.</span><span class=n>TaggedOutput</span><span class=p>(</span><span class=s1>'marked strings'</span><span class=p>,</span> <span class=n>element</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Producing multiple outputs is also available in Map and FlatMap.</span> |
| </span></span><span class=line><span class=cl><span class=c1># Here is an example that uses FlatMap and shows that the tags do not need to be specified ahead of time.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>def</span> <span class=nf>even_odd</span><span class=p>(</span><span class=n>x</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>pvalue</span><span class=o>.</span><span class=n>TaggedOutput</span><span class=p>(</span><span class=s1>'odd'</span> <span class=k>if</span> <span class=n>x</span> <span class=o>%</span> <span class=mi>2</span> <span class=k>else</span> <span class=s1>'even'</span><span class=p>,</span> <span class=n>x</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>x</span> <span class=o>%</span> <span class=mi>10</span> <span class=o>==</span> <span class=mi>0</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>x</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>results</span> <span class=o>=</span> <span class=n>numbers</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>FlatMap</span><span class=p>(</span><span class=n>even_odd</span><span class=p>)</span><span class=o>.</span><span class=n>with_outputs</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>evens</span> <span class=o>=</span> <span class=n>results</span><span class=o>.</span><span class=n>even</span> |
| </span></span><span class=line><span class=cl><span class=n>odds</span> <span class=o>=</span> <span class=n>results</span><span class=o>.</span><span class=n>odd</span> |
| </span></span><span class=line><span class=cl><span class=n>tens</span> <span class=o>=</span> <span class=n>results</span><span class=p>[</span><span class=kc>None</span><span class=p>]</span> <span class=c1># the undeclared main output</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// processWords is a DoFn that has 3 output PCollections. The emitter functions |
| </span></span></span><span class=line><span class=cl><span class=c1>// are matched in positional order to the PCollections returned by beam.ParDo3. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>processWords</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitBelowCutoff</span><span class=p>,</span> <span class=nx>emitAboveCutoff</span><span class=p>,</span> <span class=nx>emitMarked</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>const</span> <span class=nx>cutOff</span> <span class=p>=</span> <span class=mi>5</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> <span class=p><</span> <span class=nx>cutOff</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitBelowCutoff</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> <span class=k>else</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitAboveCutoff</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nf>isMarkedWord</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitMarked</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// processWordsMixed demonstrates mixing an emitter, with a standard return. |
| </span></span></span><span class=line><span class=cl><span class=c1>// If a standard return is used, it will always be the first returned PCollection, |
| </span></span></span><span class=line><span class=cl><span class=c1>// followed in positional order by the emitter functions. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>processWordsMixed</span><span class=p>(</span><span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitMarked</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=kt>int</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nf>isMarkedWord</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitMarked</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>len</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nf>Function4x0</span><span class=p>(</span><span class=nx>processWords</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>register</span><span class=p>.</span><span class=nf>Function2x1</span><span class=p>(</span><span class=nx>processWordsMixed</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// 1 input of type string => Emitter1[string] |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>register</span><span class=p>.</span><span class=nx>Emitter1</span><span class=p>[</span><span class=kt>string</span><span class=p>]()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>to_split</span> <span class=o>=</span> <span class=nx>words</span><span class=p>.</span><span class=nx>flatMap</span><span class=p>(</span><span class=kd>function</span><span class=o>*</span> <span class=p>(</span><span class=nx>word</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>(</span><span class=nx>word</span><span class=p>.</span><span class=nx>length</span> <span class=o><</span> <span class=mi>5</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=p>{</span> <span class=nx>below</span>: <span class=kt>word</span> <span class=p>};</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> <span class=k>else</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=p>{</span> <span class=nx>above</span>: <span class=kt>word</span> <span class=p>};</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>(</span><span class=nx>isMarkedWord</span><span class=p>(</span><span class=nx>word</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=p>{</span> <span class=nx>marked</span>: <span class=kt>word</span> <span class=p>};</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>});</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MapToFields</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>SomeProducingTransform</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fields</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>word</span><span class=p>:</span><span class=w> </span><span class=s2>"line.split()"</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span>- <span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>Explode</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>input</span><span class=p>:</span><span class=w> </span><span class=l>MapToFields</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fields</span><span class=p>:</span><span class=w> </span><span class=l>word</span></span></span></code></pre></div></div></div><h4 id=other-dofn-parameters>4.5.3. Accessing additional parameters in your DoFn</h4><p class=language-java>In addition to the element and the <code>OutputReceiver</code>, Beam will populate other parameters to your DoFn’s <code>@ProcessElement</code> method. |
| Any combination of these parameters can be added to your process method in any order.</p><p class=language-py>In addition to the element, Beam will populate other parameters to your DoFn’s <code>process</code> method. |
| Any combination of these parameters can be added to your process method in any order.</p><p class=language-typescript>In addition to the element, Beam will populate other parameters to your DoFn’s <code>process</code> method. |
| These are available by placing accessors in the context argument, just as for side inputs.</p><p class=language-go>In addition to the element, Beam will populate other parameters to your DoFn’s <code>ProcessElement</code> method. |
| Any combination of these parameters can be added to your process method in a standard order.</p><p class=language-go><strong>context.Context:</strong> |
| To support consolidated logging and user defined metrics, a <code>context.Context</code> parameter can be requested. |
| Per Go conventions, if present it’s required to be the first parameter of the <code>DoFn</code> method.</p><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>MyDoFn</span><span class=p>(</span><span class=nx>ctx</span> <span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>)</span> <span class=kt>string</span> <span class=p>{</span> <span class=o>...</span> <span class=p>}</span></span></span></code></pre></div></div></div><p class=language-java><strong>Timestamp:</strong> |
| To access the timestamp of an input element, add a parameter annotated with <code>@Timestamp</code> of type <code>Instant</code>. For example:</p><p class=language-py><strong>Timestamp:</strong> |
| To access the timestamp of an input element, add a keyword parameter default to <code>DoFn.TimestampParam</code>. For example:</p><p class=language-go><strong>Timestamp:</strong> |
| To access the timestamp of an input element, add a <code>beam.EventTime</code> parameter before the element. For example:</p><p class=language-typescript><strong>Timestamp:</strong> |
| To access the window an input element falls into, add a <code>pardo.windowParam()</code> to the context argument.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=nd>@Timestamp</span> <span class=n>Instant</span> <span class=n>timestamp</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>}})</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>import</span> <span class=nn>apache_beam</span> <span class=k>as</span> <span class=nn>beam</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>ProcessRecord</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>timestamp</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimestampParam</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># access timestamp of element.</span> |
| </span></span><span class=line><span class=cl> <span class=k>pass</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>MyDoFn</span><span class=p>(</span><span class=nx>ts</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>)</span> <span class=kt>string</span> <span class=p>{</span> <span class=o>...</span> <span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kd>function</span> <span class=nx>processFn</span><span class=p>(</span><span class=nx>element</span><span class=p>,</span> <span class=nx>context</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>context</span><span class=p>.</span><span class=nx>timestamp</span><span class=p>.</span><span class=nx>lookup</span><span class=p>();</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nx>pcoll</span><span class=p>.</span><span class=nx>map</span><span class=p>(</span><span class=nx>processFn</span><span class=p>,</span> <span class=p>{</span> <span class=nx>timestamp</span>: <span class=kt>pardo.timestampParam</span><span class=p>()</span> <span class=p>});</span></span></span></code></pre></div></div></div><p class=language-java><strong>Window:</strong> |
| To access the window an input element falls into, add a parameter of the type of the window used for the input <code>PCollection</code>. |
| If the parameter is a window type (a subclass of <code>BoundedWindow</code>) that does not match the input <code>PCollection</code>, then an error |
| will be raised. If an element falls in multiple windows (for example, this will happen when using <code>SlidingWindows</code>), then the |
| <code>@ProcessElement</code> method will be invoked multiple time for the element, once for each window. For example, when fixed windows |
| are being used, the window is of type <code>IntervalWindow</code>.</p><p class=language-py><strong>Window:</strong> |
| To access the window an input element falls into, add a keyword parameter default to <code>DoFn.WindowParam</code>. |
| If an element falls in multiple windows (for example, this will happen when using <code>SlidingWindows</code>), then the |
| <code>process</code> method will be invoked multiple time for the element, once for each window.</p><p class=language-go><strong>Window:</strong> |
| To access the window an input element falls into, add a <code>beam.Window</code> parameter before the element. |
| If an element falls in multiple windows (for example, this will happen when using SlidingWindows), |
| then the <code>ProcessElement</code> method will be invoked multiple time for the element, once for each window. |
| Since <code>beam.Window</code> is an interface it’s possible to type assert to the concrete implementation of the window. |
| For example, when fixed windows are being used, the window is of type <code>window.IntervalWindow</code>.</p><p class=language-typescript><strong>Window:</strong> |
| To access the window an input element falls into, add a <code>pardo.windowParam()</code> to the context argument. |
| If an element falls in multiple windows (for example, this will happen when using <code>SlidingWindows</code>), then the |
| function will be invoked multiple time for the element, once for each window.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=n>IntervalWindow</span> <span class=n>window</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>}})</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>import</span> <span class=nn>apache_beam</span> <span class=k>as</span> <span class=nn>beam</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>ProcessRecord</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>window</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>WindowParam</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># access window e.g. window.end.micros</span> |
| </span></span><span class=line><span class=cl> <span class=k>pass</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>MyDoFn</span><span class=p>(</span><span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>)</span> <span class=kt>string</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>iw</span> <span class=o>:=</span> <span class=nx>w</span><span class=p>.(</span><span class=nx>window</span><span class=p>.</span><span class=nx>IntervalWindow</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=nx>pcoll</span><span class=p>.</span><span class=nx>map</span><span class=p>(</span><span class=nx>processWithWindow</span><span class=p>,</span> <span class=p>{</span> <span class=nx>timestamp</span>: <span class=kt>pardo.windowParam</span><span class=p>()</span> <span class=p>});</span></span></span></code></pre></div></div></div><p class=language-java><strong>PaneInfo:</strong> |
| When triggers are used, Beam provides a <code>PaneInfo</code> object that contains information about the current firing. Using <code>PaneInfo</code> |
| you can determine whether this is an early or a late firing, and how many times this window has already fired for this key.</p><p class=language-py><strong>PaneInfo:</strong> |
| When triggers are used, Beam provides a <code>DoFn.PaneInfoParam</code> object that contains information about the current firing. Using <code>DoFn.PaneInfoParam</code> |
| you can determine whether this is an early or a late firing, and how many times this window has already fired for this key. |
| This feature implementation in Python SDK is not fully completed; see more at <a href=https://github.com/apache/beam/issues/18721>Issue 17821</a>.</p><p class=language-go><strong>PaneInfo:</strong> |
| When triggers are used, Beam provides <code>beam.PaneInfo</code> object that contains information about the current firing. Using <code>beam.PaneInfo</code> |
| you can determine whether this is an early or a late firing, and how many times this window has already fired for this key.</p><p class=language-typescript><strong>Window:</strong> |
| To access the window an input element falls into, add a <code>pardo.paneInfoParam()</code> to the context argument. |
| Using <code>beam.PaneInfo</code> you can determine whether this is an early or a late firing, |
| and how many times this window has already fired for this key.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=n>PaneInfo</span> <span class=n>paneInfo</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>}})</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>import</span> <span class=nn>apache_beam</span> <span class=k>as</span> <span class=nn>beam</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>ProcessRecord</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>pane_info</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>PaneInfoParam</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># access pane info, e.g. pane_info.is_first, pane_info.is_last, pane_info.timing</span> |
| </span></span><span class=line><span class=cl> <span class=k>pass</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>extractWordsFn</span><span class=p>(</span><span class=nx>pn</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PaneInfo</span><span class=p>,</span> <span class=nx>line</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>pn</span><span class=p>.</span><span class=nx>Timing</span> <span class=o>==</span> <span class=nx>typex</span><span class=p>.</span><span class=nx>PaneEarly</span> <span class=o>||</span> <span class=nx>pn</span><span class=p>.</span><span class=nx>Timing</span> <span class=o>==</span> <span class=nx>typex</span><span class=p>.</span><span class=nx>PaneOnTime</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... perform operation ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>pn</span><span class=p>.</span><span class=nx>Timing</span> <span class=o>==</span> <span class=nx>typex</span><span class=p>.</span><span class=nx>PaneLate</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... perform operation ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>pn</span><span class=p>.</span><span class=nx>IsFirst</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... perform operation ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>pn</span><span class=p>.</span><span class=nx>IsLast</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... perform operation ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>words</span> <span class=o>:=</span> <span class=nx>strings</span><span class=p>.</span><span class=nf>Split</span><span class=p>(</span><span class=nx>line</span><span class=p>,</span> <span class=s>" "</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>w</span> <span class=o>:=</span> <span class=k>range</span> <span class=nx>words</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitWords</span><span class=p>(</span><span class=nx>w</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=nx>pcoll</span><span class=p>.</span><span class=nx>map</span><span class=p>(</span><span class=nx>processWithPaneInfo</span><span class=p>,</span> <span class=p>{</span> <span class=nx>timestamp</span>: <span class=kt>pardo.paneInfoParam</span><span class=p>()</span> <span class=p>});</span></span></span></code></pre></div></div></div><p class=language-java><strong>PipelineOptions:</strong> |
| The <code>PipelineOptions</code> for the current pipeline can always be accessed in a process method by adding it |
| as a parameter:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>word</span><span class=o>,</span> <span class=n>PipelineOptions</span> <span class=n>options</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>}})</span></span></span></code></pre></div></div></div><p class=language-java><code>@OnTimer</code> methods can also access many of these parameters. Timestamp, Window, key, <code>PipelineOptions</code>, <code>OutputReceiver</code>, and |
| <code>MultiOutputReceiver</code> parameters can all be accessed in an <code>@OnTimer</code> method. In addition, an <code>@OnTimer</code> method can take |
| a parameter of type <code>TimeDomain</code> which tells whether the timer is based on event time or processing time. |
| Timers are explained in more detail in the |
| <a href=/blog/2017/08/28/timely-processing.html>Timely (and Stateful) Processing with Apache Beam</a> blog post.</p><p class=language-py><strong>Timer and State:</strong> |
| In addition to aforementioned parameters, user defined Timer and State parameters can be used in a stateful DoFn. |
| Timers and States are explained in more detail in the |
| <a href=/blog/2017/08/28/timely-processing.html>Timely (and Stateful) Processing with Apache Beam</a> blog post.</p><p class=language-go><strong>Timer and State:</strong> |
| User defined State and Timer parameters can be used in a stateful DoFn. |
| Timers and States are explained in more detail in the |
| <a href=/blog/2017/08/28/timely-processing.html>Timely (and Stateful) Processing with Apache Beam</a> blog post.</p><p class=language-typescript><strong>Timer and State:</strong> |
| This feature isn’t yet implemented in the Typescript SDK, |
| but we welcome <a href=/contribute/>contributions</a>. |
| In the meantime, Typescript pipelines wishing to use state and timers can do so |
| using <a href=#use-x-lang-transforms>cross-language transforms</a>.</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>StatefulDoFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=s2>"""An example stateful DoFn with state and timer"""</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=n>BUFFER_STATE_1</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer1'</span><span class=p>,</span> <span class=n>beam</span><span class=o>.</span><span class=n>BytesCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>BUFFER_STATE_2</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer2'</span><span class=p>,</span> <span class=n>beam</span><span class=o>.</span><span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>WATERMARK_TIMER</span> <span class=o>=</span> <span class=n>TimerSpec</span><span class=p>(</span><span class=s1>'watermark_timer'</span><span class=p>,</span> <span class=n>TimeDomain</span><span class=o>.</span><span class=n>WATERMARK</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>timestamp</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimestampParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>window</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>WindowParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer_1</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>BUFFER_STATE_1</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer_2</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>BUFFER_STATE_2</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>watermark_timer</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimerParam</span><span class=p>(</span><span class=n>WATERMARK_TIMER</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># Do your processing here</span> |
| </span></span><span class=line><span class=cl> <span class=n>key</span><span class=p>,</span> <span class=n>value</span> <span class=o>=</span> <span class=n>element</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Read all the data from buffer1</span> |
| </span></span><span class=line><span class=cl> <span class=n>all_values_in_buffer_1</span> <span class=o>=</span> <span class=p>[</span><span class=n>x</span> <span class=k>for</span> <span class=n>x</span> <span class=ow>in</span> <span class=n>buffer_1</span><span class=o>.</span><span class=n>read</span><span class=p>()]</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>StatefulDoFn</span><span class=o>.</span><span class=n>_is_clear_buffer_1_required</span><span class=p>(</span><span class=n>all_values_in_buffer_1</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># clear the buffer data if required conditions are met.</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer_1</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># add the value to buffer 2</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer_2</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>value</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>StatefulDoFn</span><span class=o>.</span><span class=n>_all_condition_met</span><span class=p>():</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Clear the timer if certain condition met and you don't want to trigger</span> |
| </span></span><span class=line><span class=cl> <span class=c1># the callback method.</span> |
| </span></span><span class=line><span class=cl> <span class=n>watermark_timer</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>element</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@on_timer</span><span class=p>(</span><span class=n>WATERMARK_TIMER</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>on_expiry_1</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>timestamp</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimestampParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>window</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>WindowParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>key</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>KeyParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer_1</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>BUFFER_STATE_1</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer_2</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>BUFFER_STATE_2</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Window and key parameters are really useful especially for debugging issues.</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=s1>'expired1'</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@staticmethod</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>_all_condition_met</span><span class=p>():</span> |
| </span></span><span class=line><span class=cl> <span class=c1># some logic</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>True</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@staticmethod</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>_is_clear_buffer_1_required</span><span class=p>(</span><span class=n>buffer_1_data</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Some business logic</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>True</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// stateAndTimersFn is an example stateful DoFn with state and a timer. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>stateAndTimersFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Buffer1</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Bag</span><span class=p>[</span><span class=kt>string</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Buffer2</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Bag</span><span class=p>[</span><span class=kt>int64</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Watermark</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>EventTime</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>s</span> <span class=o>*</span><span class=nx>stateAndTimersFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>value</span> <span class=kt>int64</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>,</span> <span class=kt>int64</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... handle processing elements here, set a callback timer... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Read all the data from Buffer1 in this window. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>vals</span><span class=p>,</span> <span class=nx>ok</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>s</span><span class=p>.</span><span class=nx>Buffer1</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>ok</span> <span class=o>&&</span> <span class=nx>s</span><span class=p>.</span><span class=nf>shouldClearBuffer</span><span class=p>(</span><span class=nx>vals</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// clear the buffer data if required conditions are met. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span><span class=p>.</span><span class=nx>Buffer1</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Add the value to Buffer2. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span><span class=p>.</span><span class=nx>Buffer2</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>value</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>s</span><span class=p>.</span><span class=nf>allConditionsMet</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Clear the timer if certain condition met and you don't want to trigger |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// the callback method. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span><span class=p>.</span><span class=nx>Watermark</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>tp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nx>key</span><span class=p>,</span> <span class=nx>value</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>s</span> <span class=o>*</span><span class=nx>stateAndTimersFn</span><span class=p>)</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>,</span> <span class=kt>int64</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Window and key parameters are really useful especially for debugging issues. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>s</span><span class=p>.</span><span class=nx>Watermark</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// timer expired, emit a different signal |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nf>emit</span><span class=p>(</span><span class=nx>key</span><span class=p>,</span> <span class=o>-</span><span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>s</span> <span class=o>*</span><span class=nx>stateAndTimersFn</span><span class=p>)</span> <span class=nf>shouldClearBuffer</span><span class=p>([]</span><span class=kt>string</span><span class=p>)</span> <span class=kt>bool</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// some business logic |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=kc>false</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>s</span> <span class=o>*</span><span class=nx>stateAndTimersFn</span><span class=p>)</span> <span class=nf>allConditionsMet</span><span class=p>()</span> <span class=kt>bool</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// other business logic |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=kc>true</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=c1>// Not yet implemented. |
| </span></span></span></code></pre></div></div></div><h3 id=composite-transforms>4.6. Composite transforms</h3><p>Transforms can have a nested structure, where a complex transform performs |
| multiple simpler transforms (such as more than one <code>ParDo</code>, <code>Combine</code>, |
| <code>GroupByKey</code>, or even other composite transforms). These transforms are called |
| composite transforms. Nesting multiple transforms inside a single composite |
| transform can make your code more modular and easier to understand.</p><p>The Beam SDK comes packed with many useful composite transforms. See the API |
| reference pages for a list of transforms:</p><ul><li><a href=https://beam.apache.org/releases/javadoc/2.55.1/index.html?org/apache/beam/sdk/transforms/package-summary.html>Pre-written Beam transforms for Java</a></li><li><a href=https://beam.apache.org/releases/pydoc/2.55.1/apache_beam.transforms.html>Pre-written Beam transforms for Python</a></li><li><a href=https://github.com/apache/beam/tree/master/sdks/go/pkg/beam/transforms>Pre-written Beam transforms for Go</a></li></ul><h4 id=composite-transform-example>4.6.1. An example composite transform</h4><p>The <code>CountWords</code> transform in the <a href=/get-started/wordcount-example/>WordCount example program</a> |
| is an example of a composite transform. <code>CountWords</code> is a <code>PTransform</code> |
| <span class="language-java language-py">subclass</span> that consists |
| of multiple nested transforms.</p><p><span class="language-java language-py">In its <code>expand</code> method, the</span> |
| <span class=language-go>The</span> <code>CountWords</code> transform applies the following |
| transform operations:</p><ol><li>It applies a <code>ParDo</code> on the input <code>PCollection</code> of text lines, producing |
| an output <code>PCollection</code> of individual words.</li><li>It applies the Beam SDK library transform <code>Count</code> on the <code>PCollection</code> of |
| words, producing a <code>PCollection</code> of key/value pairs. Each key represents a |
| word in the text, and each value represents the number of times that word |
| appeared in the original data.</li></ol><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>CountWords</span> <span class=kd>extends</span> <span class=n>PTransform</span><span class=o><</span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Long</span><span class=o>>>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Long</span><span class=o>>></span> <span class=nf>expand</span><span class=o>(</span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>lines</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Convert lines of text into individual words. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>words</span> <span class=o>=</span> <span class=n>lines</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>ExtractWordsFn</span><span class=o>()));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Count the number of times each word occurs. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Long</span><span class=o>>></span> <span class=n>wordCounts</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>words</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Count</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>perElement</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>wordCounts</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># The CountWords Composite Transform inside the WordCount pipeline.</span> |
| </span></span><span class=line><span class=cl><span class=nd>@beam.ptransform_fn</span> |
| </span></span><span class=line><span class=cl><span class=k>def</span> <span class=nf>CountWords</span><span class=p>(</span><span class=n>pcoll</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>pcoll</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Convert lines of text into individual words.</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'ExtractWords'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>ExtractWordsFn</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Count the number of times each word occurs.</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>combiners</span><span class=o>.</span><span class=n>Count</span><span class=o>.</span><span class=n>PerElement</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Format each word and count into a printable string.</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'FormatCounts'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>FormatCountsFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// CountWords is a function that builds a composite PTransform |
| </span></span></span><span class=line><span class=cl><span class=c1>// to count the number of times each word appears. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>CountWords</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>lines</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// A subscope is required for a function to become a composite transform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// We assign it to the original scope variable s to shadow the original |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// for the rest of the CountWords function. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span> <span class=p>=</span> <span class=nx>s</span><span class=p>.</span><span class=nf>Scope</span><span class=p>(</span><span class=s>"CountWords"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Since the same subscope is used for the following transforms, |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// they are in the same composite PTransform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Convert lines of text into individual words. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>words</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>extractWordsFn</span><span class=p>,</span> <span class=nx>lines</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Count the number of times each word occurs. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>wordCounts</span> <span class=o>:=</span> <span class=nx>stats</span><span class=p>.</span><span class=nf>Count</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>words</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Return any PCollections that should be available after |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// the composite transform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>wordCounts</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kd>function</span> <span class=nx>countWords</span><span class=p>(</span><span class=nx>lines</span>: <span class=kt>PCollection</span><span class=p><</span><span class=nt>string</span><span class=p>>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>lines</span> <span class=c1>// |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>.</span><span class=nx>map</span><span class=p>((</span><span class=nx>s</span>: <span class=kt>string</span><span class=p>)</span> <span class=o>=></span> <span class=nx>s</span><span class=p>.</span><span class=nx>toLowerCase</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>flatMap</span><span class=p>(</span><span class=kd>function</span><span class=o>*</span> <span class=nx>splitWords</span><span class=p>(</span><span class=nx>line</span>: <span class=kt>string</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span><span class=o>*</span> <span class=nx>line</span><span class=p>.</span><span class=nx>split</span><span class=p>(</span><span class=sr>/[^a-z]+/</span><span class=p>);</span> |
| </span></span><span class=line><span class=cl> <span class=p>})</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>countPerElement</span><span class=p>());</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>counted</span> <span class=o>=</span> <span class=nx>lines</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>countWords</span><span class=p>);</span></span></span></code></pre></div></div></div><blockquote><p><strong>Note:</strong> Because <code>Count</code> is itself a composite transform, |
| <code>CountWords</code> is also a nested composite transform.</p></blockquote><h4 id=composite-transform-creation>4.6.2. Creating a composite transform</h4><p class=language-typescript>A PTransform in the Typescript SDK is simply a function that accepts and |
| returns <code>PValue</code>s such as <code>PCollection</code>s.</p><p class="language-java language-py">To create your own composite transform, create a subclass of the <code>PTransform</code> |
| class and override the <code>expand</code> method to specify the actual processing logic. |
| You can then use this transform just as you would a built-in transform from the |
| Beam SDK.</p><p class=language-java>For the <code>PTransform</code> class type parameters, you pass the <code>PCollection</code> types |
| that your transform takes as input, and produces as output. To take multiple |
| <code>PCollection</code>s as input, or produce multiple <code>PCollection</code>s as output, use one |
| of the multi-collection types for the relevant type parameter.</p><p class=language-go>To create your own composite <code>PTransform</code> call the <code>Scope</code> method on the current |
| pipeline scope variable. Transforms passed this new sub-<code>Scope</code> will be a part of |
| the same composite <code>PTransform</code>.</p><p class=language-go>To be able to re-use your Composite, build it inside a normal Go function or method. |
| This function is passed a scope and input PCollections, and returns any |
| output PCollections it produces. <strong>Note:</strong> Such functions cannot be passed directly to |
| <code>ParDo</code> functions.</p><p class="language-java language-py">The following code sample shows how to declare a <code>PTransform</code> that accepts a |
| <code>PCollection</code> of <code>String</code>s for input, and outputs a <code>PCollection</code> of <code>Integer</code>s:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>ComputeWordLengths</span> |
| </span></span><span class=line><span class=cl> <span class=kd>extends</span> <span class=n>PTransform</span><span class=o><</span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>ComputeWordLengths</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>PTransform</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>expand</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>pcoll</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Transform logic goes here.</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>pcoll</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=k>lambda</span> <span class=n>x</span><span class=p>:</span> <span class=nb>len</span><span class=p>(</span><span class=n>x</span><span class=p>))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// CountWords is a function that builds a composite PTransform |
| </span></span></span><span class=line><span class=cl><span class=c1>// to count the number of times each word appears. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>CountWords</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>lines</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// A subscope is required for a function to become a composite transform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// We assign it to the original scope variable s to shadow the original |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// for the rest of the CountWords function. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span> <span class=p>=</span> <span class=nx>s</span><span class=p>.</span><span class=nf>Scope</span><span class=p>(</span><span class=s>"CountWords"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Since the same subscope is used for the following transforms, |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// they are in the same composite PTransform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Convert lines of text into individual words. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>words</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>extractWordsFn</span><span class=p>,</span> <span class=nx>lines</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Count the number of times each word occurs. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>wordCounts</span> <span class=o>:=</span> <span class=nx>stats</span><span class=p>.</span><span class=nf>Count</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>words</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Return any PCollections that should be available after |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// the composite transform. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>wordCounts</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p class="language-java language-py">Within your <code>PTransform</code> subclass, you’ll need to override the <code>expand</code> method. |
| The <code>expand</code> method is where you add the processing logic for the <code>PTransform</code>. |
| Your override of <code>expand</code> must accept the appropriate type of input |
| <code>PCollection</code> as a parameter, and specify the output <code>PCollection</code> as the return |
| value.</p><p class="language-java language-py">The following code sample shows how to override <code>expand</code> for the |
| <code>ComputeWordLengths</code> class declared in the previous example:</p><p class=language-go>The following code sample shows how to call the <code>CountWords</code> composite PTransform, |
| adding it to your pipeline:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>ComputeWordLengths</span> |
| </span></span><span class=line><span class=cl> <span class=kd>extends</span> <span class=n>PTransform</span><span class=o><</span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>PCollection</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=nf>expand</span><span class=o>(</span><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// transform logic goes here |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>ComputeWordLengths</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>PTransform</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>expand</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>pcoll</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Transform logic goes here.</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>pcoll</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=k>lambda</span> <span class=n>x</span><span class=p>:</span> <span class=nb>len</span><span class=p>(</span><span class=n>x</span><span class=p>))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>lines</span> <span class=o>:=</span> <span class=o>...</span> <span class=c1>// a PCollection of strings. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=c1>// A Composite PTransform function is called like any other function. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>wordCounts</span> <span class=o>:=</span> <span class=nf>CountWords</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>lines</span><span class=p>)</span> <span class=c1>// returns a PCollection<KV<string,int>> |
| </span></span></span></code></pre></div></div></div><p class="language-java language-py">As long as you override the <code>expand</code> method in your <code>PTransform</code> subclass to |
| accept the appropriate input <code>PCollection</code>(s) and return the corresponding |
| output <code>PCollection</code>(s), you can include as many transforms as you want. These |
| transforms can include core transforms, composite transforms, or the transforms |
| included in the Beam SDK libraries.</p><p class=language-go>Your composite <code>PTransform</code>s can include as many transforms as you want. These |
| transforms can include core transforms, other composite transforms, or the transforms |
| included in the Beam SDK libraries. They can also consume and return as many |
| <code>PCollection</code>s as are necessary.</p><p class="language-java language-py">Your composite transform’s parameters and return value must match the initial |
| input type and final return type for the entire transform, even if the |
| transform’s intermediate data changes type multiple times.</p><p class="language-java language-py"><strong>Note:</strong> The <code>expand</code> method of a <code>PTransform</code> is not meant to be invoked |
| directly by the user of a transform. Instead, you should call the <code>apply</code> method |
| on the <code>PCollection</code> itself, with the transform as an argument. This allows |
| transforms to be nested within the structure of your pipeline.</p><h4 id=ptransform-style-guide>4.6.3. PTransform Style Guide</h4><p>The <a href=/contribute/ptransform-style-guide/>PTransform Style Guide</a> |
| contains additional information not included here, such as style guidelines, |
| logging and testing guidance, and language-specific considerations. The guide |
| is a useful starting point when you want to write new composite PTransforms.</p><h2 id=pipeline-io>5. Pipeline I/O</h2><p>When you create a pipeline, you often need to read data from some external |
| source, such as a file or a database. Likewise, you may |
| want your pipeline to output its result data to an external storage system. |
| Beam provides read and write transforms for a <a href=/documentation/io/built-in/>number of common data storage |
| types</a>. If you want your pipeline |
| to read from or write to a data storage format that isn’t supported by the |
| built-in transforms, you can <a href=/documentation/io/developing-io-overview/>implement your own read and write |
| transforms</a>.</p><h3 id=pipeline-io-reading-data>5.1. Reading input data</h3><p>Read transforms read data from an external source and return a <code>PCollection</code> |
| representation of the data for use by your pipeline. You can use a read |
| transform at any point while constructing your pipeline to create a new |
| <code>PCollection</code>, though it will be most common at the start of your pipeline.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>lines</span> <span class=o>=</span> <span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>TextIO</span><span class=o>.</span><span class=na>read</span><span class=o>().</span><span class=na>from</span><span class=o>(</span><span class=s>"gs://some/inputData.txt"</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>lines</span> <span class=o>=</span> <span class=n>pipeline</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>io</span><span class=o>.</span><span class=n>ReadFromText</span><span class=p>(</span><span class=s1>'gs://some/inputData.txt'</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>lines</span> <span class=o>:=</span> <span class=nx>textio</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=err>'</span><span class=nx>gs</span><span class=p>:</span><span class=c1>//some/inputData.txt') |
| </span></span></span></code></pre></div></div></div><h3 id=pipeline-io-writing-data>5.2. Writing output data</h3><p>Write transforms write the data in a <code>PCollection</code> to an external data source. |
| You will most often use write transforms at the end of your pipeline to output |
| your pipeline’s final results. However, you can use a write transform to output |
| a <code>PCollection</code>’s data at any point in your pipeline.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>output</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>TextIO</span><span class=o>.</span><span class=na>write</span><span class=o>().</span><span class=na>to</span><span class=o>(</span><span class=s>"gs://some/outputData"</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>output</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>io</span><span class=o>.</span><span class=n>WriteToText</span><span class=p>(</span><span class=s1>'gs://some/outputData'</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>textio</span><span class=p>.</span><span class=nf>Write</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=err>'</span><span class=nx>gs</span><span class=p>:</span><span class=c1>//some/inputData.txt', output) |
| </span></span></span></code></pre></div></div></div><h3 id=file-based-data>5.3. File-based input and output data</h3><h4 id=file-based-reading-multiple-locations>5.3.1. Reading from multiple locations</h4><p>Many read transforms support reading from multiple input files matching a glob |
| operator you provide. Note that glob operators are filesystem-specific and obey |
| filesystem-specific consistency models. The following TextIO example uses a glob |
| operator (<code>*</code>) to read all matching input files that have prefix “input-” and the |
| suffix “.csv” in the given location:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=s>"ReadFromText"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>TextIO</span><span class=o>.</span><span class=na>read</span><span class=o>().</span><span class=na>from</span><span class=o>(</span><span class=s>"protocol://my_bucket/path/to/input-*.csv"</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>lines</span> <span class=o>=</span> <span class=n>pipeline</span> <span class=o>|</span> <span class=s1>'ReadFromText'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>io</span><span class=o>.</span><span class=n>ReadFromText</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'path/to/input-*.csv'</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>lines</span> <span class=o>:=</span> <span class=nx>textio</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>scope</span><span class=p>,</span> <span class=s>"path/to/input-*.csv"</span><span class=p>)</span></span></span></code></pre></div></div></div><p>To read data from disparate sources into a single <code>PCollection</code>, read each one |
| independently and then use the <a href=#flatten>Flatten</a> transform to create a single |
| <code>PCollection</code>.</p><h4 id=file-based-writing-multiple-files>5.3.2. Writing to multiple output files</h4><p>For file-based output data, write transforms write to multiple output files by |
| default. When you pass an output file name to a write transform, the file name |
| is used as the prefix for all output files that the write transform produces. |
| You can append a suffix to each output file by specifying a suffix.</p><p>The following write transform example writes multiple output files to a |
| location. Each file has the prefix “numbers”, a numeric tag, and the suffix |
| “.csv”.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>records</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=s>"WriteToText"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>TextIO</span><span class=o>.</span><span class=na>write</span><span class=o>().</span><span class=na>to</span><span class=o>(</span><span class=s>"protocol://my_bucket/path/to/numbers"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withSuffix</span><span class=o>(</span><span class=s>".csv"</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>filtered_words</span> <span class=o>|</span> <span class=s1>'WriteToText'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>io</span><span class=o>.</span><span class=n>WriteToText</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'/path/to/numbers'</span><span class=p>,</span> <span class=n>file_name_suffix</span><span class=o>=</span><span class=s1>'.csv'</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// The Go SDK textio doesn't support sharding on writes yet. |
| </span></span></span><span class=line><span class=cl><span class=c1>// See https://github.com/apache/beam/issues/21031 for ways |
| </span></span></span><span class=line><span class=cl><span class=c1>// to contribute a solution. |
| </span></span></span></code></pre></div></div></div><h3 id=provided-io-transforms>5.4. Beam-provided I/O transforms</h3><p>See the <a href=/documentation/io/built-in/>Beam-provided I/O Transforms</a> |
| page for a list of the currently available I/O transforms.</p><h2 id=schemas>6. Schemas</h2><p>Often, the types of the records being processed have an obvious structure. Common Beam sources produce |
| JSON, Avro, Protocol Buffer, or database row objects; all of these types have well defined structures, |
| structures that can often be determined by examining the type. Even within a SDK pipeline, Simple Java POJOs |
| (or equivalent structures in other languages) are often used as intermediate types, and these also have a |
| clear structure that can be inferred by inspecting the class. By understanding the structure of a pipeline’s |
| records, we can provide much more concise APIs for data processing.</p><h3 id=what-is-a-schema>6.1. What is a schema?</h3><p>Most structured records share some common characteristics:</p><ul><li>They can be subdivided into separate named fields. Fields usually have string names, but sometimes - as in the case of indexed |
| tuples - have numerical indices instead.</li><li>There is a confined list of primitive types that a field can have. These often match primitive types in most programming |
| languages: int, long, string, etc.</li><li>Often a field type can be marked as optional (sometimes referred to as nullable) or required.</li></ul><p>Often records have a nested structure. A nested structure occurs when a field itself has subfields so the |
| type of the field itself has a schema. Fields that are array or map types is also a common feature of these structured |
| records.</p><p>For example, consider the following schema, representing actions in a fictitious e-commerce company:</p><p><strong>Purchase</strong></p><div class=table-container-wrapper><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>userId</td><td>STRING</td></tr><tr><td>itemId</td><td>INT64</td></tr><tr><td>shippingAddress</td><td>ROW(ShippingAddress)</td></tr><tr><td>cost</td><td>INT64</td></tr><tr><td>transactions</td><td>ARRAY[ROW(Transaction)]</td></tr></tbody></table><br></div><p><strong>ShippingAddress</strong></p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>streetAddress</td><td>STRING</td></tr><tr><td>city</td><td>STRING</td></tr><tr><td>state</td><td>nullable STRING</td></tr><tr><td>country</td><td>STRING</td></tr><tr><td>postCode</td><td>STRING</td></tr></tbody></table><br><p><strong>Transaction</strong></p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>bank</td><td>STRING</td></tr><tr><td>purchaseAmount</td><td>DOUBLE</td></tr></tbody></table><br><p>Purchase event records are represented by the above purchase schema. Each purchase event contains a shipping address, which |
| is a nested row containing its own schema. Each purchase also contains an array of credit-card transactions |
| (a list, because a purchase might be split across multiple credit cards); each item in the transaction list is a row |
| with its own schema.</p><p>This provides an abstract description of the types involved, one that is abstracted away from any specific programming |
| language.</p><p>Schemas provide us a type-system for Beam records that is independent of any specific programming-language type. There |
| might be multiple Java classes that all have the same schema (for example a Protocol-Buffer class or a POJO class), |
| and Beam will allow us to seamlessly convert between these types. Schemas also provide a simple way to reason about |
| types across different programming-language APIs.</p><p>A <code>PCollection</code> with a schema does not need to have a <code>Coder</code> specified, as Beam knows how to encode and decode |
| Schema rows; Beam uses a special coder to encode schema types.</p><h3 id=schemas-for-pl-types>6.2. Schemas for programming language types</h3><p>While schemas themselves are language independent, they are designed to embed naturally into the programming languages |
| of the Beam SDK being used. This allows Beam users to continue using native types while reaping the advantage of |
| having Beam understand their element schemas.</p><p class=language-java>In Java you could use the following set of classes to represent the purchase schema. Beam will automatically |
| infer the correct schema based on the members of the class.</p><p class=language-py>In Python you can use the following set of classes to represent the purchase schema. Beam will automatically infer the correct schema based on the members of the class.</p><p class=language-go>In Go, schema encoding is used by default for struct types, with Exported fields becoming part of the schema. |
| Beam will automatically infer the schema based on the fields and field tags of the struct, and their order.</p><p class=language-typescript>In Typescript, JSON objects are used to represent schema’d data. |
| Unfortunately type information in Typescript is not propagated to the runtime layer, |
| so it needs to be manually specified in some places (e.g. when using cross-language pipelines).</p><p class=language-yaml>In Beam YAML, all transforms produce and accept schema’d data which is used to validate the pipeline.</p><p class=language-yaml>In some cases, Beam is unable to figure out the output type of a mapping function. |
| In this case, you can specify it manually using |
| <a href=https://json-schema.org/understanding-json-schema/reference/type>JSON schema syntax</a>.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>JavaBeanSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>Purchase</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getUserId</span><span class=o>();</span> <span class=c1>// Returns the id of the user who made the purchase. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=kt>long</span> <span class=nf>getItemId</span><span class=o>();</span> <span class=c1>// Returns the identifier of the item that was purchased. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=n>ShippingAddress</span> <span class=nf>getShippingAddress</span><span class=o>();</span> <span class=c1>// Returns the shipping address, a nested type. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=kt>long</span> <span class=nf>getCostCents</span><span class=o>();</span> <span class=c1>// Returns the cost of the item. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=n>List</span><span class=o><</span><span class=n>Transaction</span><span class=o>></span> <span class=nf>getTransactions</span><span class=o>();</span> <span class=c1>// Returns the transactions that paid for this purchase (returns a list, since the purchase might be spread out over multiple credit cards). |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=nd>@SchemaCreate</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=nf>Purchase</span><span class=o>(</span><span class=n>String</span> <span class=n>userId</span><span class=o>,</span> <span class=kt>long</span> <span class=n>itemId</span><span class=o>,</span> <span class=n>ShippingAddress</span> <span class=n>shippingAddress</span><span class=o>,</span> <span class=kt>long</span> <span class=n>costCents</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>List</span><span class=o><</span><span class=n>Transaction</span><span class=o>></span> <span class=n>transactions</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>JavaBeanSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>ShippingAddress</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getStreetAddress</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getCity</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Nullable</span> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getState</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getCountry</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getPostCode</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@SchemaCreate</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=nf>ShippingAddress</span><span class=o>(</span><span class=n>String</span> <span class=n>streetAddress</span><span class=o>,</span> <span class=n>String</span> <span class=n>city</span><span class=o>,</span> <span class=nd>@Nullable</span> <span class=n>String</span> <span class=n>state</span><span class=o>,</span> <span class=n>String</span> <span class=n>country</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=n>postCode</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>JavaBeanSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>Transaction</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getBank</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>double</span> <span class=nf>getPurchaseAmount</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@SchemaCreate</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=nf>Transaction</span><span class=o>(</span><span class=n>String</span> <span class=n>bank</span><span class=o>,</span> <span class=kt>double</span> <span class=n>purchaseAmount</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>import</span> <span class=nn>typing</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>Purchase</span><span class=p>(</span><span class=n>typing</span><span class=o>.</span><span class=n>NamedTuple</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>user_id</span><span class=p>:</span> <span class=nb>str</span> <span class=c1># The id of the user who made the purchase.</span> |
| </span></span><span class=line><span class=cl> <span class=n>item_id</span><span class=p>:</span> <span class=nb>int</span> <span class=c1># The identifier of the item that was purchased.</span> |
| </span></span><span class=line><span class=cl> <span class=n>shipping_address</span><span class=p>:</span> <span class=n>ShippingAddress</span> <span class=c1># The shipping address, a nested type.</span> |
| </span></span><span class=line><span class=cl> <span class=n>cost_cents</span><span class=p>:</span> <span class=nb>int</span> <span class=c1># The cost of the item</span> |
| </span></span><span class=line><span class=cl> <span class=n>transactions</span><span class=p>:</span> <span class=n>typing</span><span class=o>.</span><span class=n>Sequence</span><span class=p>[</span><span class=n>Transaction</span><span class=p>]</span> <span class=c1># The transactions that paid for this purchase (a list, since the purchase might be spread out over multiple credit cards).</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>ShippingAddress</span><span class=p>(</span><span class=n>typing</span><span class=o>.</span><span class=n>NamedTuple</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>street_address</span><span class=p>:</span> <span class=nb>str</span> |
| </span></span><span class=line><span class=cl> <span class=n>city</span><span class=p>:</span> <span class=nb>str</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=p>:</span> <span class=n>typing</span><span class=o>.</span><span class=n>Optional</span><span class=p>[</span><span class=nb>str</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=n>country</span><span class=p>:</span> <span class=nb>str</span> |
| </span></span><span class=line><span class=cl> <span class=n>postal_code</span><span class=p>:</span> <span class=nb>str</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>Transaction</span><span class=p>(</span><span class=n>typing</span><span class=o>.</span><span class=n>NamedTuple</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>bank</span><span class=p>:</span> <span class=nb>str</span> |
| </span></span><span class=line><span class=cl> <span class=n>purchase_amount</span><span class=p>:</span> <span class=nb>float</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>Purchase</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ID of the user who made the purchase. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>UserID</span> <span class=kt>string</span> <span class=s>`beam:"userId"`</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Identifier of the item that was purchased. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>ItemID</span> <span class=kt>int64</span> <span class=s>`beam:"itemId"`</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The shipping address, a nested type. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>ShippingAddress</span> <span class=nx>ShippingAddress</span> <span class=s>`beam:"shippingAddress"`</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The cost of the item in cents. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>Cost</span> <span class=kt>int64</span> <span class=s>`beam:"cost"`</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The transactions that paid for this purchase. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// A slice since the purchase might be spread out over multiple |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// credit cards. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>Transactions</span> <span class=p>[]</span><span class=nx>Transaction</span> <span class=s>`beam:"transactions"`</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>ShippingAddress</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>StreetAddress</span> <span class=kt>string</span> <span class=s>`beam:"streetAddress"`</span> |
| </span></span><span class=line><span class=cl> <span class=nx>City</span> <span class=kt>string</span> <span class=s>`beam:"city"`</span> |
| </span></span><span class=line><span class=cl> <span class=nx>State</span> <span class=o>*</span><span class=kt>string</span> <span class=s>`beam:"state"`</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Country</span> <span class=kt>string</span> <span class=s>`beam:"country"`</span> |
| </span></span><span class=line><span class=cl> <span class=nx>PostCode</span> <span class=kt>string</span> <span class=s>`beam:"postCode"`</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>Transaction</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Bank</span> <span class=kt>string</span> <span class=s>`beam:"bank"`</span> |
| </span></span><span class=line><span class=cl> <span class=nx>PurchaseAmount</span> <span class=kt>float64</span> <span class=s>`beam:"purchaseAmount"`</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>intField</span>: <span class=kt>1</span><span class=p>,</span> <span class=nx>stringField</span><span class=o>:</span> <span class=s2>"a"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>intField</span>: <span class=kt>2</span><span class=p>,</span> <span class=nx>stringField</span><span class=o>:</span> <span class=s2>"b"</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>])</span> |
| </span></span><span class=line><span class=cl> <span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Let beam know the type of the elements by providing an exemplar. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>withRowCoder</span><span class=p>({</span> <span class=nx>intField</span>: <span class=kt>0</span><span class=p>,</span> <span class=nx>stringField</span><span class=o>:</span> <span class=s2>""</span> <span class=p>}));</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>MapToFields</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>fields</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>new_field</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>expression</span><span class=p>:</span><span class=w> </span><span class=s2>"hex(weight)"</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>output_type</span><span class=p>:</span><span class=w> </span>{<span class=w> </span><span class=nt>"type": </span><span class=s2>"string"</span><span class=w> </span>}</span></span></code></pre></div></div></div><p class=language-java>Using JavaBean classes as above is one way to map a schema to Java classes. However multiple Java classes might have |
| the same schema, in which case the different Java types can often be used interchangeably. Beam will add implicit |
| conversions between types that have matching schemas. For example, the above |
| <code>Transaction</code> class has the same schema as the following class:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>JavaFieldSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>TransactionPojo</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=n>bank</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>double</span> <span class=n>purchaseAmount</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><p class=language-java>So if we had two <code>PCollection</code>s as follows</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Transaction</span><span class=o>></span> <span class=n>transactionBeans</span> <span class=o>=</span> <span class=n>readTransactionsAsJavaBean</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>TransactionPojos</span><span class=o>></span> <span class=n>transactionPojos</span> <span class=o>=</span> <span class=n>readTransactionsAsPojo</span><span class=o>();</span></span></span></code></pre></div></div></div><p class=language-java>Then these two <code>PCollection</code>s would have the same schema, even though their Java types would be different. This means |
| for example the following two code snippets are valid:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>transactionBeans</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><...>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>TransactionPojo</span> <span class=n>pojo</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><p class=language-java>and</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>transactionPojos</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><...>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>Transaction</span> <span class=n>row</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><p class=language-java>Even though the in both cases the <code>@Element</code> parameter differs from the <code>PCollection</code>’s Java type, since the |
| schemas are the same Beam will automatically make the conversion. The built-in <code>Convert</code> transform can also be used |
| to translate between Java types of equivalent schemas, as detailed below.</p><h3 id=schema-definition>6.3. Schema definition</h3><p>The schema for a <code>PCollection</code> defines elements of that <code>PCollection</code> as an ordered list of named fields. Each field |
| has a name, a type, and possibly a set of user options. The type of a field can be primitive or composite. The following |
| are the primitive types currently supported by Beam:</p><table class=table-wrapper--pr><thead><tr class=header><th>Type</th><th>Description</th></tr></thead><tbody><tr><td>BYTE</td><td>An 8-bit signed value</td></tr><tr><td>INT16</td><td>A 16-bit signed value</td></tr><tr><td>INT32</td><td>A 32-bit signed value</td></tr><tr><td>INT64</td><td>A 64-bit signed value</td></tr><tr><td>DECIMAL</td><td>An arbitrary-precision decimal type</td></tr><tr><td>FLOAT</td><td>A 32-bit IEEE 754 floating point number</td></tr><tr><td>DOUBLE</td><td>A 64-bit IEEE 754 floating point number</td></tr><tr><td>STRING</td><td>A string</td></tr><tr><td>DATETIME</td><td>A timestamp represented as milliseconds since the epoch</td></tr><tr><td>BOOLEAN</td><td>A boolean value</td></tr><tr><td>BYTES</td><td>A raw byte array</td></tr></tbody></table><br><p>A field can also reference a nested schema. In this case, the field will have type ROW, and the nested schema will |
| be an attribute of this field type.</p><p>Three collection types are supported as field types: ARRAY, ITERABLE and MAP:</p><ul><li><strong>ARRAY</strong> This represents a repeated value type, where the repeated elements can have any supported type. Arrays of |
| nested rows are supported, as are arrays of arrays.</li><li><strong>ITERABLE</strong> This is very similar to the array type, it represents a repeated value, but one in which the full list of |
| items is not known until iterated over. This is intended for the case where an iterable might be larger than the |
| available memory, and backed by external storage (for example, this can happen with the iterable returned by a |
| <code>GroupByKey</code>). The repeated elements can have any supported type.</li><li><strong>MAP</strong> This represents an associative map from keys to values. All schema types are supported for both keys and values. |
| Values that contain map types cannot be used as keys in any grouping operation.</li></ul><h3 id=logical-types>6.4. Logical types</h3><p>Users can extend the schema type system to add custom logical types that can be used as a field. A logical type is |
| identified by a unique identifier and an argument. A logical type also specifies an underlying schema type to be used |
| for storage, along with conversions to and from that type. As an example, a logical union can always be represented as |
| a row with nullable fields, where the user ensures that only one of those fields is ever set at a time. However this can |
| be tedious and complex to manage. The OneOf logical type provides a value class that makes it easier to manage the type |
| as a union, while still using a row with nullable fields as its underlying storage. Each logical type also has a |
| unique identifier, so they can be interpreted by other languages as well. More examples of logical types are listed |
| below.</p><h4 id=defining-a-logical-type>6.4.1. Defining a logical type</h4><p>To define a logical type you must specify a Schema type to be used to represent the underlying type as well as a unique |
| identifier for that type. A logical type imposes additional semantics on top a schema type. For example, a logical |
| type to represent nanosecond timestamps is represented as a schema containing an INT64 and an INT32 field. This schema |
| alone does not say anything about how to interpret this type, however the logical type tells you that this represents |
| a nanosecond timestamp, with the INT64 field representing seconds and the INT32 field representing nanoseconds.</p><p>Logical types are also specified by an argument, which allows creating a class of related types. For example, a |
| limited-precision decimal type would have an integer argument indicating how many digits of precision are represented. |
| The argument is represented by a schema type, so can itself be a complex type.</p><p class=language-java>In Java, a logical type is specified as a subclass of the <code>LogicalType</code> class. A custom Java class can be specified to represent the logical type and conversion functions must be supplied to convert back and forth between this Java class and the underlying Schema type representation. For example, the logical type representing nanosecond timestamp might be implemented as follows</p><p class=language-go>In Go, a logical type is specified with a custom implementation of the <code>beam.SchemaProvider</code> interface. |
| For example, the logical type provider representing nanosecond timestamps |
| might be implemented as follows</p><p class=language-typescript>In Typescript, a logical type defined by the <a href=https://github.com/apache/beam/blob/master/sdks/typescript/src/apache_beam/coders/row_coder.ts>LogicalTypeInfo</a> |
| interface which associates a logical type’s URN with its representation |
| and its conversion to and from this representation.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// A Logical type using java.time.Instant to represent the logical type. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>public</span> <span class=kd>class</span> <span class=nc>TimestampNanos</span> <span class=kd>implements</span> <span class=n>LogicalType</span><span class=o><</span><span class=n>Instant</span><span class=o>,</span> <span class=n>Row</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The underlying schema used to represent rows. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>Schema</span> <span class=n>SCHEMA</span> <span class=o>=</span> <span class=n>Schema</span><span class=o>.</span><span class=na>builder</span><span class=o>().</span><span class=na>addInt64Field</span><span class=o>(</span><span class=s>"seconds"</span><span class=o>).</span><span class=na>addInt32Field</span><span class=o>(</span><span class=s>"nanos"</span><span class=o>).</span><span class=na>build</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getIdentifier</span><span class=o>()</span> <span class=o>{</span> <span class=k>return</span> <span class=s>"timestampNanos"</span><span class=o>;</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> <span class=kd>public</span> <span class=n>FieldType</span> <span class=nf>getBaseType</span><span class=o>()</span> <span class=o>{</span> <span class=k>return</span> <span class=n>schema</span><span class=o>;</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Convert the representation type to the underlying Row type. Called by Beam when necessary. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@Override</span> <span class=kd>public</span> <span class=n>Row</span> <span class=nf>toBaseType</span><span class=o>(</span><span class=n>Instant</span> <span class=n>instant</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>Row</span><span class=o>.</span><span class=na>withSchema</span><span class=o>(</span><span class=n>schema</span><span class=o>).</span><span class=na>addValues</span><span class=o>(</span><span class=n>instant</span><span class=o>.</span><span class=na>getEpochSecond</span><span class=o>(),</span> <span class=n>instant</span><span class=o>.</span><span class=na>getNano</span><span class=o>()).</span><span class=na>build</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Convert the underlying Row type to an Instant. Called by Beam when necessary. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@Override</span> <span class=kd>public</span> <span class=n>Instant</span> <span class=nf>toInputType</span><span class=o>(</span><span class=n>Row</span> <span class=n>base</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>Instant</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>row</span><span class=o>.</span><span class=na>getInt64</span><span class=o>(</span><span class=s>"seconds"</span><span class=o>),</span> <span class=n>row</span><span class=o>.</span><span class=na>getInt32</span><span class=o>(</span><span class=s>"nanos"</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// Define a logical provider like so: |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// TimestampNanos is a logical type using time.Time, but |
| </span></span></span><span class=line><span class=cl><span class=c1>// encodes as a schema type. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>TimestampNanos</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Time</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>tn</span> <span class=nx>TimestampNanos</span><span class=p>)</span> <span class=nf>Seconds</span><span class=p>()</span> <span class=kt>int64</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>time</span><span class=p>.</span><span class=nf>Time</span><span class=p>(</span><span class=nx>tn</span><span class=p>).</span><span class=nf>Unix</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>tn</span> <span class=nx>TimestampNanos</span><span class=p>)</span> <span class=nf>Nanos</span><span class=p>()</span> <span class=kt>int32</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nb>int32</span><span class=p>(</span><span class=nx>time</span><span class=p>.</span><span class=nf>Time</span><span class=p>(</span><span class=nx>tn</span><span class=p>).</span><span class=nf>UnixNano</span><span class=p>()</span> <span class=o>%</span> <span class=mi>1000000000</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// tnStorage is the storage schema for TimestampNanos. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>tnStorage</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Seconds</span> <span class=kt>int64</span> <span class=s>`beam:"seconds"`</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Nanos</span> <span class=kt>int32</span> <span class=s>`beam:"nanos"`</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>var</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// reflect.Type of the Value type of TimestampNanos |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>tnType</span> <span class=p>=</span> <span class=nx>reflect</span><span class=p>.</span><span class=nf>TypeOf</span><span class=p>((</span><span class=o>*</span><span class=nx>TimestampNanos</span><span class=p>)(</span><span class=kc>nil</span><span class=p>)).</span><span class=nf>Elem</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=nx>tnStorageType</span> <span class=p>=</span> <span class=nx>reflect</span><span class=p>.</span><span class=nf>TypeOf</span><span class=p>((</span><span class=o>*</span><span class=nx>tnStorage</span><span class=p>)(</span><span class=kc>nil</span><span class=p>)).</span><span class=nf>Elem</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// TimestampNanosProvider implements the beam.SchemaProvider interface. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>TimestampNanosProvider</span> <span class=kd>struct</span><span class=p>{}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// FromLogicalType converts checks if the given type is TimestampNanos, and if so |
| </span></span></span><span class=line><span class=cl><span class=c1>// returns the storage type. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>p</span> <span class=o>*</span><span class=nx>TimestampNanosProvider</span><span class=p>)</span> <span class=nf>FromLogicalType</span><span class=p>(</span><span class=nx>rt</span> <span class=nx>reflect</span><span class=p>.</span><span class=nx>Type</span><span class=p>)</span> <span class=p>(</span><span class=nx>reflect</span><span class=p>.</span><span class=nx>Type</span><span class=p>,</span> <span class=kt>error</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>rt</span> <span class=o>!=</span> <span class=nx>tnType</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span><span class=p>,</span> <span class=nx>fmt</span><span class=p>.</span><span class=nf>Errorf</span><span class=p>(</span><span class=s>"unable to provide schema.LogicalType for type %v, want %v"</span><span class=p>,</span> <span class=nx>rt</span><span class=p>,</span> <span class=nx>tnType</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>tnStorageType</span><span class=p>,</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// BuildEncoder builds a Beam schema encoder for the TimestampNanos type. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>p</span> <span class=o>*</span><span class=nx>TimestampNanosProvider</span><span class=p>)</span> <span class=nf>BuildEncoder</span><span class=p>(</span><span class=nx>rt</span> <span class=nx>reflect</span><span class=p>.</span><span class=nx>Type</span><span class=p>)</span> <span class=p>(</span><span class=kd>func</span><span class=p>(</span><span class=nx>any</span><span class=p>,</span> <span class=nx>io</span><span class=p>.</span><span class=nx>Writer</span><span class=p>)</span> <span class=kt>error</span><span class=p>,</span> <span class=kt>error</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>p</span><span class=p>.</span><span class=nf>FromLogicalType</span><span class=p>(</span><span class=nx>rt</span><span class=p>);</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span><span class=p>,</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>enc</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>coder</span><span class=p>.</span><span class=nf>RowEncoderForStruct</span><span class=p>(</span><span class=nx>tnStorageType</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span><span class=p>,</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kd>func</span><span class=p>(</span><span class=nx>iface</span> <span class=nx>any</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>io</span><span class=p>.</span><span class=nx>Writer</span><span class=p>)</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>v</span> <span class=o>:=</span> <span class=nx>iface</span><span class=p>.(</span><span class=nx>TimestampNanos</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nf>enc</span><span class=p>(</span><span class=nx>tnStorage</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Seconds</span><span class=p>:</span> <span class=nx>v</span><span class=p>.</span><span class=nf>Seconds</span><span class=p>(),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Nanos</span><span class=p>:</span> <span class=nx>v</span><span class=p>.</span><span class=nf>Nanos</span><span class=p>(),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>w</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// BuildDecoder builds a Beam schema decoder for the TimestampNanos type. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>p</span> <span class=o>*</span><span class=nx>TimestampNanosProvider</span><span class=p>)</span> <span class=nf>BuildDecoder</span><span class=p>(</span><span class=nx>rt</span> <span class=nx>reflect</span><span class=p>.</span><span class=nx>Type</span><span class=p>)</span> <span class=p>(</span><span class=kd>func</span><span class=p>(</span><span class=nx>io</span><span class=p>.</span><span class=nx>Reader</span><span class=p>)</span> <span class=p>(</span><span class=nx>any</span><span class=p>,</span> <span class=kt>error</span><span class=p>),</span> <span class=kt>error</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>p</span><span class=p>.</span><span class=nf>FromLogicalType</span><span class=p>(</span><span class=nx>rt</span><span class=p>);</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span><span class=p>,</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>dec</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>coder</span><span class=p>.</span><span class=nf>RowDecoderForStruct</span><span class=p>(</span><span class=nx>tnStorageType</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span><span class=p>,</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kd>func</span><span class=p>(</span><span class=nx>r</span> <span class=nx>io</span><span class=p>.</span><span class=nx>Reader</span><span class=p>)</span> <span class=p>(</span><span class=nx>any</span><span class=p>,</span> <span class=kt>error</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>s</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nf>dec</span><span class=p>(</span><span class=nx>r</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span><span class=p>,</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>tn</span> <span class=o>:=</span> <span class=nx>s</span><span class=p>.(</span><span class=nx>tnStorage</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nf>TimestampNanos</span><span class=p>(</span><span class=nx>time</span><span class=p>.</span><span class=nf>Unix</span><span class=p>(</span><span class=nx>tn</span><span class=p>.</span><span class=nx>Seconds</span><span class=p>,</span> <span class=nb>int64</span><span class=p>(</span><span class=nx>tn</span><span class=p>.</span><span class=nx>Nanos</span><span class=p>))),</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Register it like so: |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=nx>beam</span><span class=p>.</span><span class=nf>RegisterSchemaProvider</span><span class=p>(</span><span class=nx>tnType</span><span class=p>,</span> <span class=o>&</span><span class=nx>TimestampNanosProvider</span><span class=p>{})</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=c1>// Register a logical type: |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=kr>class</span> <span class=nx>Foo</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=kr>constructor</span><span class=p>(</span><span class=kr>public</span> <span class=nx>value</span>: <span class=kt>string</span><span class=p>)</span> <span class=p>{}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=nx>requireForSerialization</span><span class=p>(</span><span class=s2>"apache-beam"</span><span class=p>,</span> <span class=p>{</span> <span class=nx>Foo</span> <span class=p>});</span> |
| </span></span><span class=line><span class=cl><span class=nx>row_coder</span><span class=p>.</span><span class=nx>registerLogicalType</span><span class=p>({</span> |
| </span></span><span class=line><span class=cl> <span class=nx>urn</span><span class=o>:</span> <span class=s2>"beam:logical_type:typescript_foo:v1"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>reprType</span>: <span class=kt>row_coder.RowCoder.inferTypeFromJSON</span><span class=p>(</span><span class=s2>"string"</span><span class=p>,</span> <span class=kc>false</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>toRepr</span><span class=o>:</span> <span class=p>(</span><span class=nx>foo</span><span class=p>)</span> <span class=o>=></span> <span class=nx>foo</span><span class=p>.</span><span class=nx>value</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fromRepr</span><span class=o>:</span> <span class=p>(</span><span class=nx>value</span><span class=p>)</span> <span class=o>=></span> <span class=k>new</span> <span class=nx>Foo</span><span class=p>(</span><span class=nx>value</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl><span class=p>});</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// And use it as follows: |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span><span class=k>new</span> <span class=nx>Foo</span><span class=p>(</span><span class=s2>"a"</span><span class=p>),</span> <span class=k>new</span> <span class=nx>Foo</span><span class=p>(</span><span class=s2>"b"</span><span class=p>)]))</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Use beamLogicalType in the exemplar to indicate its use. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>withRowCoder</span><span class=p>({</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beamLogicalType</span><span class=o>:</span> <span class=s2>"beam:logical_type:typescript_foo:v1"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> <span class=kr>as</span> <span class=kt>any</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>);</span></span></span></code></pre></div></div></div><h4 id=built-in-logical-types>6.4.2. Useful logical types</h4><p class=language-py>Currently the Python SDK provides minimal convenience logical types, |
| other than to handle <code>MicrosInstant</code>.</p><p class=language-go>Currently the Go SDK provides minimal convenience logical types, |
| other than to handle additional integer primitives, and <code>time.Time</code>.</p><h5 id=enumerationtype><strong>EnumerationType</strong></h5><p class=language-py>This convenience builder doesn’t yet exist for the Python SDK.</p><p class=language-go>This convenience builder doesn’t yet exist for the Go SDK.</p><p class=language-java>This logical type allows creating an enumeration type consisting of a set of named constants.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>Schema</span> <span class=n>schema</span> <span class=o>=</span> <span class=n>Schema</span><span class=o>.</span><span class=na>builder</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=err>…</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>addLogicalTypeField</span><span class=o>(</span><span class=s>"color"</span><span class=o>,</span> <span class=n>EnumerationType</span><span class=o>.</span><span class=na>create</span><span class=o>(</span><span class=s>"RED"</span><span class=o>,</span> <span class=s>"GREEN"</span><span class=o>,</span> <span class=s>"BLUE"</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>build</span><span class=o>();</span></span></span></code></pre></div></div></div><p class=language-java>The value of this field is stored in the row as an INT32 type, however the logical type defines a value type that lets |
| you access the enumeration either as a string or a value. For example:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>EnumerationType</span><span class=o>.</span><span class=na>Value</span> <span class=n>enumValue</span> <span class=o>=</span> <span class=n>enumType</span><span class=o>.</span><span class=na>valueOf</span><span class=o>(</span><span class=s>"RED"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl><span class=n>enumValue</span><span class=o>.</span><span class=na>getValue</span><span class=o>();</span> <span class=c1>// Returns 0, the integer value of the constant. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>enumValue</span><span class=o>.</span><span class=na>toString</span><span class=o>();</span> <span class=c1>// Returns "RED", the string value of the constant |
| </span></span></span></code></pre></div></div></div><p class=language-java>Given a row object with an enumeration field, you can also extract the field as the enumeration value.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>EnumerationType</span><span class=o>.</span><span class=na>Value</span> <span class=n>enumValue</span> <span class=o>=</span> <span class=n>row</span><span class=o>.</span><span class=na>getLogicalTypeValue</span><span class=o>(</span><span class=s>"color"</span><span class=o>,</span> <span class=n>EnumerationType</span><span class=o>.</span><span class=na>Value</span><span class=o>.</span><span class=na>class</span><span class=o>);</span></span></span></code></pre></div></div></div><p class=language-java>Automatic schema inference from Java POJOs and JavaBeans automatically converts Java enums to EnumerationType logical |
| types.</p><h5 id=oneoftype><strong>OneOfType</strong></h5><p class=language-py>This convenience builder doesn’t yet exist for the Python SDK.</p><p class=language-go>This convenience builder doesn’t yet exist for the Go SDK.</p><p class=language-java>OneOfType allows creating a disjoint union type over a set of schema fields. For example:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>Schema</span> <span class=n>schema</span> <span class=o>=</span> <span class=n>Schema</span><span class=o>.</span><span class=na>builder</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=err>…</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>addLogicalTypeField</span><span class=o>(</span><span class=s>"oneOfField"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>OneOfType</span><span class=o>.</span><span class=na>create</span><span class=o>(</span><span class=n>Field</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"intField"</span><span class=o>,</span> <span class=n>FieldType</span><span class=o>.</span><span class=na>INT32</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>Field</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"stringField"</span><span class=o>,</span> <span class=n>FieldType</span><span class=o>.</span><span class=na>STRING</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>Field</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"bytesField"</span><span class=o>,</span> <span class=n>FieldType</span><span class=o>.</span><span class=na>BYTES</span><span class=o>)))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>build</span><span class=o>();</span></span></span></code></pre></div></div></div><p class=language-java>The value of this field is stored in the row as another Row type, where all the fields are marked as nullable. The |
| logical type however defines a Value object that contains an enumeration value indicating which field was set and allows |
| getting just that field:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Returns an enumeration indicating all possible case values for the enum. |
| </span></span></span><span class=line><span class=cl><span class=c1>// For the above example, this will be |
| </span></span></span><span class=line><span class=cl><span class=c1>// EnumerationType.create("intField", "stringField", "bytesField"); |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>EnumerationType</span> <span class=n>oneOfEnum</span> <span class=o>=</span> <span class=n>onOfType</span><span class=o>.</span><span class=na>getCaseEnumType</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Creates an instance of the union with the string field set. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>OneOfType</span><span class=o>.</span><span class=na>Value</span> <span class=n>oneOfValue</span> <span class=o>=</span> <span class=n>oneOfType</span><span class=o>.</span><span class=na>createValue</span><span class=o>(</span><span class=s>"stringField"</span><span class=o>,</span> <span class=s>"foobar"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Handle the oneof |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=k>switch</span> <span class=o>(</span><span class=n>oneOfValue</span><span class=o>.</span><span class=na>getCaseEnumType</span><span class=o>().</span><span class=na>toString</span><span class=o>())</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=s>"intField"</span><span class=o>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>processInt</span><span class=o>(</span><span class=n>oneOfValue</span><span class=o>.</span><span class=na>getValue</span><span class=o>(</span><span class=n>Integer</span><span class=o>.</span><span class=na>class</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=s>"stringField"</span><span class=o>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>processString</span><span class=o>(</span><span class=n>oneOfValue</span><span class=o>.</span><span class=na>getValue</span><span class=o>(</span><span class=n>String</span><span class=o>.</span><span class=na>class</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=s>"bytesField"</span><span class=o>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>processBytes</span><span class=o>(</span><span class=n>oneOfValue</span><span class=o>.</span><span class=na>getValue</span><span class=o>(</span><span class=n>bytes</span><span class=o>[].</span><span class=na>class</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><p class=language-java>In the above example we used the field names in the switch statement for clarity, however the enum integer values could |
| also be used.</p><h3 id=creating-schemas>6.5. Creating Schemas</h3><p>In order to take advantage of schemas, your <code>PCollection</code>s must have a schema attached to it. |
| Often, the source itself will attach a schema to the PCollection. |
| For example, when using <code>AvroIO</code> to read Avro files, the source can automatically infer a Beam schema from the Avro schema and attach that to the Beam <code>PCollection</code>. |
| However not all sources produce schemas. |
| In addition, often Beam pipelines have intermediate stages and types, and those also can benefit from the expressiveness of schemas.</p><h4 id=inferring-schemas>6.5.1. Inferring schemas</h4><nav class=language-switcher><strong>Adapt for:</strong><ul><li data-value=java class=active>Java SDK</li><li data-value=py>Python SDK</li><li data-value=go>Go SDK</li><li data-value=typescript>TypeScript SDK</li></ul></nav><p class=language-typescript>Unfortunately, Beam is unable to access Typescript’s type information at runtime. |
| Schemas must be manually declared with <code>beam.withRowCoder</code>. |
| On the other hand, schema-aware operations such as <code>GroupBy</code> can be used |
| without an explicit schema declared.</p><p class=language-java>Beam is able to infer schemas from a variety of common Java types. |
| The <code>@DefaultSchema</code> annotation can be used to tell Beam to infer schemas from a specific type. |
| The annotation takes a <code>SchemaProvider</code> as an argument, and <code>SchemaProvider</code> classes are already built in for common Java types. |
| The <code>SchemaRegistry</code> can also be invoked programmatically for cases where it is not practical to annotate the Java type itself.</p><p class=language-java><strong>Java POJOs</strong></p><p class=language-java>A POJO (Plain Old Java Object) is a Java object that is not bound by any restriction other than the Java Language |
| Specification. A POJO can contain member variables that are primitives, that are other POJOs, or are collections maps or |
| arrays thereof. POJOs do not have to extend prespecified classes or extend any specific interfaces.</p><p class=language-java>If a POJO class is annotated with <code>@DefaultSchema(JavaFieldSchema.class)</code>, Beam will automatically infer a schema for |
| this class. Nested classes are supported as are classes with <code>List</code>, array, and <code>Map</code> fields.</p><p class=language-java>For example, annotating the following class tells Beam to infer a schema from this POJO class and apply it to any |
| <code>PCollection<TransactionPojo></code>.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>JavaFieldSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>TransactionPojo</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>final</span> <span class=n>String</span> <span class=n>bank</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>final</span> <span class=kt>double</span> <span class=n>purchaseAmount</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@SchemaCreate</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=nf>TransactionPojo</span><span class=o>(</span><span class=n>String</span> <span class=n>bank</span><span class=o>,</span> <span class=kt>double</span> <span class=n>purchaseAmount</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>this</span><span class=o>.</span><span class=na>bank</span> <span class=o>=</span> <span class=n>bank</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=k>this</span><span class=o>.</span><span class=na>purchaseAmount</span> <span class=o>=</span> <span class=n>purchaseAmount</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=c1>// Beam will automatically infer the correct schema for this PCollection. No coder is needed as a result. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>TransactionPojo</span><span class=o>></span> <span class=n>pojos</span> <span class=o>=</span> <span class=n>readPojos</span><span class=o>();</span></span></span></code></pre></div></div></div><p class=language-java>The <code>@SchemaCreate</code> annotation tells Beam that this constructor can be used to create instances of TransactionPojo, |
| assuming that constructor parameters have the same names as the field names. <code>@SchemaCreate</code> can also be used to annotate |
| static factory methods on the class, allowing the constructor to remain private. If there is no <code>@SchemaCreate</code> |
| annotation then all the fields must be non-final and the class must have a zero-argument constructor.</p><p class=language-java>There are a couple of other useful annotations that affect how Beam infers schemas. By default the schema field names |
| inferred will match that of the class field names. However <code>@SchemaFieldName</code> can be used to specify a different name to |
| be used for the schema field. <code>@SchemaIgnore</code> can be used to mark specific class fields as excluded from the inferred |
| schema. For example, it’s common to have ephemeral fields in a class that should not be included in a schema |
| (e.g. caching the hash value to prevent expensive recomputation of the hash), and <code>@SchemaIgnore</code> can be used to |
| exclude these fields. Note that ignored fields will not be included in the encoding of these records.</p><p class=language-java>In some cases it is not convenient to annotate the POJO class, for example if the POJO is in a different package that is |
| not owned by the Beam pipeline author. In these cases the schema inference can be triggered programmatically in |
| pipeline’s main function as follows:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>pipeline</span><span class=o>.</span><span class=na>getSchemaRegistry</span><span class=o>().</span><span class=na>registerPOJO</span><span class=o>(</span><span class=n>TransactionPOJO</span><span class=o>.</span><span class=na>class</span><span class=o>);</span></span></span></code></pre></div></div></div><p class=language-java><strong>Java Beans</strong></p><p class=language-java>Java Beans are a de-facto standard for creating reusable property classes in Java. While the full |
| standard has many characteristics, the key ones are that all properties are accessed via getter and setter classes, and |
| the name format for these getters and setters is standardized. A Java Bean class can be annotated with |
| <code>@DefaultSchema(JavaBeanSchema.class)</code> and Beam will automatically infer a schema for this class. For example:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>JavaBeanSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>TransactionBean</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=nf>TransactionBean</span><span class=o>()</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getBank</span><span class=o>()</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>setBank</span><span class=o>(</span><span class=n>String</span> <span class=n>bank</span><span class=o>)</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>double</span> <span class=nf>getPurchaseAmount</span><span class=o>()</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>setPurchaseAmount</span><span class=o>(</span><span class=kt>double</span> <span class=n>purchaseAmount</span><span class=o>)</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=c1>// Beam will automatically infer the correct schema for this PCollection. No coder is needed as a result. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>TransactionBean</span><span class=o>></span> <span class=n>beans</span> <span class=o>=</span> <span class=n>readBeans</span><span class=o>();</span></span></span></code></pre></div></div></div><p class=language-java>The <code>@SchemaCreate</code> annotation can be used to specify a constructor or a static factory method, in which case the |
| setters and zero-argument constructor can be omitted.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>JavaBeanSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>TransactionBean</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@SchemaCreate</span> |
| </span></span><span class=line><span class=cl> <span class=n>Public</span> <span class=nf>TransactionBean</span><span class=o>(</span><span class=n>String</span> <span class=n>bank</span><span class=o>,</span> <span class=kt>double</span> <span class=n>purchaseAmount</span><span class=o>)</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=nf>getBank</span><span class=o>()</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>double</span> <span class=nf>getPurchaseAmount</span><span class=o>()</span> <span class=o>{</span> <span class=err>…</span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><p class=language-java><code>@SchemaFieldName</code> and <code>@SchemaIgnore</code> can be used to alter the schema inferred, just like with POJO classes.</p><p class=language-java><strong>AutoValue</strong></p><p class=language-java>Java value classes are notoriously difficult to generate correctly. There is a lot of boilerplate you must create in |
| order to properly implement a value class. AutoValue is a popular library for easily generating such classes by |
| implementing a simple abstract base class.</p><p class=language-java>Beam can infer a schema from an AutoValue class. For example:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@DefaultSchema</span><span class=o>(</span><span class=n>AutoValueSchema</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=nd>@AutoValue</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>abstract</span> <span class=kd>class</span> <span class=nc>TransactionValue</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>abstract</span> <span class=n>String</span> <span class=nf>getBank</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>abstract</span> <span class=kt>double</span> <span class=nf>getPurchaseAmount</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><p class=language-java>This is all that’s needed to generate a simple AutoValue class, and the above <code>@DefaultSchema</code> annotation tells Beam to |
| infer a schema from it. This also allows AutoValue elements to be used inside of <code>PCollection</code>s.</p><p class=language-java><code>@SchemaFieldName</code> and <code>@SchemaIgnore</code> can be used to alter the schema inferred.</p><p class=language-py>Beam has a few different mechanisms for inferring schemas from Python code.</p><p class=language-py><strong>NamedTuple classes</strong></p><p class=language-py>A <a href=https://docs.python.org/3/library/typing.html#typing.NamedTuple>NamedTuple</a> |
| class is a Python class that wraps a <code>tuple</code>, assigning a name to each element |
| and restricting it to a particular type. Beam will automatically infer the |
| schema for PCollections with <code>NamedTuple</code> output types. For example:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>Transaction</span><span class=p>(</span><span class=n>typing</span><span class=o>.</span><span class=n>NamedTuple</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>bank</span><span class=p>:</span> <span class=nb>str</span> |
| </span></span><span class=line><span class=cl> <span class=n>purchase_amount</span><span class=p>:</span> <span class=nb>float</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>pc</span> <span class=o>=</span> <span class=nb>input</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=k>lambda</span> <span class=o>...</span><span class=p>)</span><span class=o>.</span><span class=n>with_output_types</span><span class=p>(</span><span class=n>Transaction</span><span class=p>)</span></span></span></code></pre></div></div></div><p class=language-py><strong>beam.Row and Select</strong></p><p class=language-py>There are also methods for creating ad-hoc schema declarations. First, you can |
| use a lambda that returns instances of <code>beam.Row</code>:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>input_pc</span> <span class=o>=</span> <span class=o>...</span> <span class=c1># {"bank": ..., "purchase_amount": ...}</span> |
| </span></span><span class=line><span class=cl><span class=n>output_pc</span> <span class=o>=</span> <span class=n>input_pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=k>lambda</span> <span class=n>item</span><span class=p>:</span> <span class=n>beam</span><span class=o>.</span><span class=n>Row</span><span class=p>(</span><span class=n>bank</span><span class=o>=</span><span class=n>item</span><span class=p>[</span><span class=s2>"bank"</span><span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=n>purchase_amount</span><span class=o>=</span><span class=n>item</span><span class=p>[</span><span class=s2>"purchase_amount"</span><span class=p>])</span></span></span></code></pre></div></div></div><p class=language-py>Sometimes it can be more concise to express the same logic with the |
| <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.Select><code>Select</code></a> transform:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>input_pc</span> <span class=o>=</span> <span class=o>...</span> <span class=c1># {"bank": ..., "purchase_amount": ...}</span> |
| </span></span><span class=line><span class=cl><span class=n>output_pc</span> <span class=o>=</span> <span class=n>input_pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Select</span><span class=p>(</span><span class=n>bank</span><span class=o>=</span><span class=k>lambda</span> <span class=n>item</span><span class=p>:</span> <span class=n>item</span><span class=p>[</span><span class=s2>"bank"</span><span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=n>purchase_amount</span><span class=o>=</span><span class=k>lambda</span> <span class=n>item</span><span class=p>:</span> <span class=n>item</span><span class=p>[</span><span class=s2>"purchase_amount"</span><span class=p>])</span></span></span></code></pre></div></div></div><p class=language-py>Note that these declaration don’t include any specific information about the |
| types of the <code>bank</code> and <code>purchase_amount</code> fields, so Beam will attempt to infer |
| type information. If it’s unable to it will fall back to the generic type |
| <code>Any</code>. Sometimes this is not ideal, you can use casts to make sure Beam |
| correctly infers types with <code>beam.Row</code> or with <code>Select</code>:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>input_pc</span> <span class=o>=</span> <span class=o>...</span> <span class=c1># {"bank": ..., "purchase_amount": ...}</span> |
| </span></span><span class=line><span class=cl><span class=n>output_pc</span> <span class=o>=</span> <span class=n>input_pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=k>lambda</span> <span class=n>item</span><span class=p>:</span> <span class=n>beam</span><span class=o>.</span><span class=n>Row</span><span class=p>(</span><span class=n>bank</span><span class=o>=</span><span class=nb>str</span><span class=p>(</span><span class=n>item</span><span class=p>[</span><span class=s2>"bank"</span><span class=p>]),</span> |
| </span></span><span class=line><span class=cl> <span class=n>purchase_amount</span><span class=o>=</span><span class=nb>float</span><span class=p>(</span><span class=n>item</span><span class=p>[</span><span class=s2>"purchase_amount"</span><span class=p>])))</span></span></span></code></pre></div></div></div><p class=language-go>Beam currently only infers schemas for exported fields in Go structs.</p><p class=language-go><strong>Structs</strong></p><p class=language-go>Beam will automatically infer schemas for all Go structs used |
| as PCollection elements, and default to encoding them using |
| schema encoding.</p><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>Transaction</span> <span class=kd>struct</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Bank</span> <span class=kt>string</span> |
| </span></span><span class=line><span class=cl> <span class=nx>PurchaseAmount</span> <span class=kt>float64</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>checksum</span> <span class=p>[]</span><span class=kt>byte</span> <span class=c1>// ignored |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span></span></span></code></pre></div></div></div><p class=language-go>Unexported fields are ignored, and cannot be automatically inferred as part of the schema. |
| Fields of type func, channel, unsafe.Pointer, or uintptr will be ignored by inference. |
| Fields of interface types are ignored, unless a schema provider |
| is registered for them.</p><p class=language-go>By default, schema field names will match the exported struct field names. |
| In the above example, “Bank” and “PurchaseAmount” are the schema field names. |
| A schema field name can be overridden with a struct tag for the field.</p><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>Transaction</span> <span class=kd>struct</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Bank</span> <span class=kt>string</span> <span class=s>`beam:"bank"`</span> |
| </span></span><span class=line><span class=cl> <span class=nx>PurchaseAmount</span> <span class=kt>float64</span> <span class=s>`beam:"purchase_amount"`</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p class=language-go>Overriding schema field names is useful for compatibility cross language transforms, |
| as schema fields may have different requirements or restrictions from Go exported fields.</p><h3 id=using-schemas>6.6. Using Schema Transforms</h3><p>A schema on a <code>PCollection</code> enables a rich variety of relational transforms. The fact that each record is composed of |
| named fields allows for simple and readable aggregations that reference fields by name, similar to the aggregations in |
| a SQL expression.</p><p class=language-go>Beam does not yet support Schema transforms natively in Go. However, it will be implemented with the following behavior.</p><h4 id=661-field-selection-syntax>6.6.1. Field selection syntax</h4><p>The advantage of schemas is that they allow referencing of element fields by name. Beam provides a selection syntax for |
| referencing fields, including nested and repeated fields. This syntax is used by all of the schema transforms when |
| referencing the fields they operate on. The syntax can also be used inside of a DoFn to specify which schema fields to |
| process.</p><p>Addressing fields by name still retains type safety as Beam will check that schemas match at the time the pipeline graph |
| is constructed. If a field is specified that does not exist in the schema, the pipeline will fail to launch. In addition, |
| if a field is specified with a type that does not match the type of that field in the schema, the pipeline will fail to |
| launch.</p><p>The following characters are not allowed in field names: . * [ ] { }</p><h5 id=top-level-fields><strong>Top-level fields</strong></h5><p>In order to select a field at the top level of a schema, the name of the field is specified. For example, to select just |
| the user ids from a <code>PCollection</code> of purchases one would write (using the <code>Select</code> transform)</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span><span class=s>"userId"</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>input_pc</span> <span class=o>=</span> <span class=o>...</span> <span class=c1># {"user_id": ...,"bank": ..., "purchase_amount": ...}</span> |
| </span></span><span class=line><span class=cl><span class=n>output_pc</span> <span class=o>=</span> <span class=n>input_pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Select</span><span class=p>(</span><span class=s2>"user_id"</span><span class=p>)</span></span></span></code></pre></div></div></div><h5 id=nested-fields><strong>Nested fields</strong></h5><p class=language-py>Support for Nested fields hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for Nested fields hasn’t been developed for the Go SDK yet.</p><p class=language-java>Individual nested fields can be specified using the dot operator. For example, to select just the postal code from the |
| shipping address one would write</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span><span class=s>"shippingAddress.postCode"</span><span class=o>));</span></span></span></code></pre></div></div></div><h5 id=wildcards><strong>Wildcards</strong></h5><p class=language-py>Support for wildcards hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for wildcards hasn’t been developed for the Go SDK yet.</p><p class=language-java>The * operator can be specified at any nesting level to represent all fields at that level. For example, to select all |
| shipping-address fields one would write</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span><span class=s>"shippingAddress.*"</span><span class=o>));</span></span></span></code></pre></div></div></div><h5 id=arrays><strong>Arrays</strong></h5><p class=language-java>An array field, where the array element type is a row, can also have subfields of the element type addressed. When |
| selected, the result is an array of the selected subfield type. For example</p><p class=language-py>Support for Array fields hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for Array fields hasn’t been developed for the Go SDK yet.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span><span class=s>"transactions[].bank"</span><span class=o>));</span></span></span></code></pre></div></div></div><p class=language-java>Will result in a row containing an array field with element-type string, containing the list of banks for each |
| transaction.</p><p class=language-java>While the use of [] brackets in the selector is recommended, to make it clear that array elements are being selected, |
| they can be omitted for brevity. In the future, array slicing will be supported, allowing selection of portions of the |
| array.</p><h5 id=maps><strong>Maps</strong></h5><p>A map field, where the value type is a row, can also have subfields of the value type addressed. When selected, the |
| result is a map where the keys are the same as in the original map but the value is the specified type. Similar to |
| arrays, the use of {} curly brackets in the selector is recommended, to make it clear that map value elements are being |
| selected, they can be omitted for brevity. In the future, map key selectors will be supported, allowing selection of |
| specific keys from the map. For example, given the following schema:</p><p><strong>PurchasesByType</strong></p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>purchases</td><td>MAP{STRING, ROW{PURCHASE}</td></tr></tbody></table><br><p>The following</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchasesByType</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span><span class=s>"purchases{}.userId"</span><span class=o>));</span></span></span></code></pre></div></div></div><p class=language-py>Support for Map fields hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for Map fields hasn’t been developed for the Go SDK yet.</p><p>Will result in a row containing a map field with key-type string and value-type string. The selected map will contain |
| all of the keys from the original map, and the values will be the userId contained in the purchase record.</p><p>While the use of {} brackets in the selector is recommended, to make it clear that map value elements are being selected, |
| they can be omitted for brevity. In the future, map slicing will be supported, allowing selection of specific keys from |
| the map.</p><h4 id=662-schema-transforms>6.6.2. Schema transforms</h4><p>Beam provides a collection of transforms that operate natively on schemas. These transforms are very expressive, |
| allowing selections and aggregations in terms of named schema fields. Following are some examples of useful |
| schema transforms.</p><h5 id=selecting-input><strong>Selecting input</strong></h5><p>Often a computation is only interested in a subset of the fields in an input <code>PCollection</code>. The <code>Select</code> transform allows |
| one to easily project out only the fields of interest. The resulting <code>PCollection</code> has a schema containing each selected |
| field as a top-level field. Both top-level and nested fields can be selected. For example, in the Purchase schema, one |
| could select only the userId and streetAddress fields as follows</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"shippingAddress.streetAddress"</span><span class=o>));</span></span></span></code></pre></div></div></div><p class=language-py>Support for Nested fields hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for Nested fields hasn’t been developed for the Go SDK yet.</p><p>The resulting <code>PCollection</code> will have the following schema</p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>userId</td><td>STRING</td></tr><tr><td>streetAddress</td><td>STRING</td></tr></tbody></table><br><p>The same is true for wildcard selections. The following</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"shippingAddress.*"</span><span class=o>));</span></span></span></code></pre></div></div></div><p class=language-py>Support for Wildcards hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for Wildcards hasn’t been developed for the Go SDK yet.</p><p>Will result in the following schema</p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>userId</td><td>STRING</td></tr><tr><td>streetAddress</td><td>STRING</td></tr><tr><td>city</td><td>STRING</td></tr><tr><td>state</td><td>nullable STRING</td></tr><tr><td>country</td><td>STRING</td></tr><tr><td>postCode</td><td>STRING</td></tr></tbody></table><br><p>When selecting fields nested inside of an array, the same rule applies that each selected field appears separately as a |
| top-level field in the resulting row. This means that if multiple fields are selected from the same nested row, each |
| selected field will appear as its own array field. For example</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>fieldNames</span><span class=o>(</span> <span class=s>"transactions.bank"</span><span class=o>,</span> <span class=s>"transactions.purchaseAmount"</span><span class=o>));</span></span></span></code></pre></div></div></div><p class=language-py>Support for nested fields hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for nested fields hasn’t been developed for the Go SDK yet.</p><p class=language-java><p>Will result in the following schema</p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>bank</td><td>ARRAY[STRING]</td></tr><tr><td>purchaseAmount</td><td>ARRAY[DOUBLE]</td></tr></tbody></table><br></p><p>Wildcard selections are equivalent to separately selecting each field.</p><p>Selecting fields nested inside of maps have the same semantics as arrays. If you select multiple fields from a map |
| , then each selected field will be expanded to its own map at the top level. This means that the set of map keys will |
| be copied, once for each selected field.</p><p>Sometimes different nested rows will have fields with the same name. Selecting multiple of these fields would result in |
| a name conflict, as all selected fields are put in the same row schema. When this situation arises, the |
| <code>Select.withFieldNameAs</code> builder method can be used to provide an alternate name for the selected field.</p><p>Another use of the Select transform is to flatten a nested schema into a single flat schema. For example</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Select</span><span class=o>.</span><span class=na>flattenedSchema</span><span class=o>());</span></span></span></code></pre></div></div></div><p class=language-py>Support for nested fields hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for nested fields hasn’t been developed for the Go SDK yet.</p><p class=language-java><p>Will result in the following schema</p><div class=table-container-wrapper><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>userId</td><td>STRING</td></tr><tr><td>itemId</td><td>STRING</td></tr><tr><td>shippingAddress_streetAddress</td><td>STRING</td></tr><tr><td>shippingAddress_city</td><td>nullable STRING</td></tr><tr><td>shippingAddress_state</td><td>STRING</td></tr><tr><td>shippingAddress_country</td><td>STRING</td></tr><tr><td>shippingAddress_postCode</td><td>STRING</td></tr><tr><td>costCents</td><td>INT64</td></tr><tr><td>transactions_bank</td><td>ARRAY[STRING]</td></tr><tr><td>transactions_purchaseAmount</td><td>ARRAY[DOUBLE]</td></tr></tbody></table><br></div></p><h5 id=grouping-aggregations><strong>Grouping aggregations</strong></h5><p class=language-java>The <code>Group</code> transform allows simply grouping data by any number of fields in the input schema, applying aggregations to |
| those groupings, and storing the result of those aggregations in a new schema field. The output of the <code>Group</code> transform |
| has a schema with one field corresponding to each aggregation performed.</p><p class=language-py>The <code>GroupBy</code> transform allows simply grouping data by any number of fields in the input schema, applying aggregations to |
| those groupings, and storing the result of those aggregations in a new schema field. The output of the <code>GroupBy</code> transform |
| has a schema with one field corresponding to each aggregation performed.</p><p class=language-java>The simplest usage of <code>Group</code> specifies no aggregations, in which case all inputs matching the provided set of fields |
| are grouped together into an <code>ITERABLE</code> field. For example</p><p class=language-py>The simplest usage of <code>GroupBy</code> specifies no aggregations, in which case all inputs matching the provided set of fields |
| are grouped together into an <code>ITERABLE</code> field. For example</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Group</span><span class=o>.</span><span class=na>byFieldNames</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"bank"</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>input_pc</span> <span class=o>=</span> <span class=o>...</span> <span class=c1># {"user_id": ...,"bank": ..., "purchase_amount": ...}</span> |
| </span></span><span class=line><span class=cl><span class=n>output_pc</span> <span class=o>=</span> <span class=n>input_pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>GroupBy</span><span class=p>(</span><span class=s1>'user_id'</span><span class=p>,</span><span class=s1>'bank'</span><span class=p>)</span></span></span></code></pre></div></div></div><p class=language-go>Support for schema-aware grouping hasn’t been developed for the Go SDK yet.</p><p class=lanuage-java>The output schema of this is:</p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>key</td><td>ROW{userId:STRING, bank:STRING}</td></tr><tr><td>values</td><td>ITERABLE[ROW[Purchase]]</td></tr></tbody></table><br><p>The key field contains the grouping key and the values field contains a list of all the values that matched that key.</p><p>The names of the key and values fields in the output schema can be controlled using this withKeyField and withValueField |
| builders, as follows:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Group</span><span class=o>.</span><span class=na>byFieldNames</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"shippingAddress.streetAddress"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withKeyField</span><span class=o>(</span><span class=s>"userAndStreet"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withValueField</span><span class=o>(</span><span class=s>"matchingPurchases"</span><span class=o>));</span></span></span></code></pre></div></div></div><p>It is quite common to apply one or more aggregations to the grouped result. Each aggregation can specify one or more fields |
| to aggregate, an aggregation function, and the name of the resulting field in the output schema. For example, the |
| following application computes three aggregations grouped by userId, with all aggregations represented in a single |
| output schema:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Group</span><span class=o>.</span><span class=na>byFieldNames</span><span class=o>(</span><span class=s>"userId"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>aggregateField</span><span class=o>(</span><span class=s>"itemId"</span><span class=o>,</span> <span class=n>Count</span><span class=o>.</span><span class=na>combineFn</span><span class=o>(),</span> <span class=s>"numPurchases"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>aggregateField</span><span class=o>(</span><span class=s>"costCents"</span><span class=o>,</span> <span class=n>Sum</span><span class=o>.</span><span class=na>ofLongs</span><span class=o>(),</span> <span class=s>"totalSpendCents"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>aggregateField</span><span class=o>(</span><span class=s>"costCents"</span><span class=o>,</span> <span class=n>Top</span><span class=o>.<</span><span class=n>Long</span><span class=o>></span><span class=n>largestLongsFn</span><span class=o>(</span><span class=n>10</span><span class=o>),</span> <span class=s>"topPurchases"</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>input_pc</span> <span class=o>=</span> <span class=o>...</span> <span class=c1># {"user_id": ..., "item_Id": ..., "cost_cents": ...}</span> |
| </span></span><span class=line><span class=cl><span class=n>output_pc</span> <span class=o>=</span> <span class=n>input_pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>GroupBy</span><span class=p>(</span><span class=s2>"user_id"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=n>aggregate_field</span><span class=p>(</span><span class=s2>"item_id"</span><span class=p>,</span> <span class=n>CountCombineFn</span><span class=p>,</span> <span class=s2>"num_purchases"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=n>aggregate_field</span><span class=p>(</span><span class=s2>"cost_cents"</span><span class=p>,</span> <span class=nb>sum</span><span class=p>,</span> <span class=s2>"total_spendcents"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=n>aggregate_field</span><span class=p>(</span><span class=s2>"cost_cents"</span><span class=p>,</span> <span class=n>TopCombineFn</span><span class=p>,</span> <span class=s2>"top_purchases"</span><span class=p>)</span></span></span></code></pre></div></div></div><p class=language-go>Support for schema-aware grouping hasn’t been developed for the Go SDK yet.</p><p>The result of this aggregation will have the following schema:</p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>key</td><td>ROW{userId:STRING}</td></tr><tr><td>value</td><td>ROW{numPurchases: INT64, totalSpendCents: INT64, topPurchases: ARRAY[INT64]}</td></tr></tbody></table><br><p>Often <code>Selected.flattenedSchema</code> will be use to flatten the result into a non-nested, flat schema.</p><h5 id=joins><strong>Joins</strong></h5><p>Beam supports equijoins on schema <code>PCollections</code> - namely joins where the join condition depends on the equality of a |
| subset of fields. For example, the following examples uses the Purchases schema to join transactions with the reviews |
| that are likely associated with that transaction (both the user and product match that in the transaction). This is a |
| “natural join” - one in which the same field names are used on both the left-hand and right-hand sides of the join - |
| and is specified with the <code>using</code> keyword:</p><p class=language-py>Support for joins hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for joins hasn’t been developed for the Go SDK yet.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Transaction</span><span class=o>></span> <span class=n>transactions</span> <span class=o>=</span> <span class=n>readTransactions</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Review</span><span class=o>></span> <span class=n>reviews</span> <span class=o>=</span> <span class=n>readReviews</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Row</span><span class=o>></span> <span class=n>joined</span> <span class=o>=</span> <span class=n>transactions</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Join</span><span class=o>.</span><span class=na>innerJoin</span><span class=o>(</span><span class=n>reviews</span><span class=o>).</span><span class=na>using</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"productId"</span><span class=o>));</span></span></span></code></pre></div></div></div><p class=language-java><p>The resulting schema is the following:</p><table class=table-wrapper--pr><thead><tr class=header><th><b>Field Name</b></th><th><b>Field Type</b></th></tr></thead><tbody><tr><td>lhs</td><td>ROW{Transaction}</td></tr><tr><td>rhs</td><td>ROW{Review}</td></tr></tbody></table><br></p><p>Each resulting row contains one Transaction and one Review that matched the join condition.</p><p>If the fields to match in the two schemas have different names, then the on function can be used. For example, if the |
| Review schema named those fields differently than the Transaction schema, then we could write the following:</p><p class=language-py>Support for joins hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for joins hasn’t been developed for the Go SDK yet.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Row</span><span class=o>></span> <span class=n>joined</span> <span class=o>=</span> <span class=n>transactions</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Join</span><span class=o>.</span><span class=na>innerJoin</span><span class=o>(</span><span class=n>reviews</span><span class=o>).</span><span class=na>on</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>FieldsEqual</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>left</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"productId"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>right</span><span class=o>(</span><span class=s>"reviewUserId"</span><span class=o>,</span> <span class=s>"reviewProductId"</span><span class=o>)));</span></span></span></code></pre></div></div></div><p>In addition to inner joins, the Join transform supports full outer joins, left outer joins, and right outer joins.</p><h5 id=complex-joins><strong>Complex joins</strong></h5><p>While most joins tend to be binary joins - joining two inputs together - sometimes you have more than two input |
| streams that all need to be joined on a common key. The <code>CoGroup</code> transform allows joining multiple <code>PCollections</code> |
| together based on equality of schema fields. Each <code>PCollection</code> can be marked as required or optional in the final |
| join record, providing a generalization of outer joins to joins with greater than two input <code>PCollection</code>s. The output |
| can optionally be expanded - providing individual joined records, as in the <code>Join</code> transform. The output can also be |
| processed in unexpanded format - providing the join key along with Iterables of all records from each input that matched |
| that key.</p><p class=language-py>Support for joins hasn’t been developed for the Python SDK yet.</p><p class=language-go>Support for joins hasn’t been developed for the Go SDK yet.</p><h5 id=filtering-events><strong>Filtering events</strong></h5><p>The <code>Filter</code> transform can be configured with a set of predicates, each one based one specified fields. Only records for |
| which all predicates return true will pass the filter. For example the following</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Filter</span><span class=o>.</span><span class=na>create</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>whereFieldName</span><span class=o>(</span><span class=s>"costCents"</span><span class=o>,</span> <span class=n>c</span> <span class=o>-></span> <span class=n>c</span> <span class=o>></span> <span class=n>100</span> <span class=o>*</span> <span class=n>20</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>whereFieldName</span><span class=o>(</span><span class=s>"shippingAddress.country"</span><span class=o>,</span> <span class=n>c</span> <span class=o>-></span> <span class=n>c</span><span class=o>.</span><span class=na>equals</span><span class=o>(</span><span class=s>"de"</span><span class=o>));</span></span></span></code></pre></div></div></div><p>Will produce all purchases made from Germany with a purchase price of greater than twenty cents.</p><h5 id=adding-fields-to-a-schema><strong>Adding fields to a schema</strong></h5><p>The AddFields transform can be used to extend a schema with new fields. Input rows will be extended to the new schema by |
| inserting null values for the new fields, though alternate default values can be specified; if the default null value |
| is used then the new field type will be marked as nullable. Nested subfields can be added using the field selection |
| syntax, including nested fields inside arrays or map values.</p><p>For example, the following application</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>AddFields</span><span class=o>.<</span><span class=n>PurchasePojo</span><span class=o>></span><span class=n>create</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>field</span><span class=o>(</span><span class=s>"timeOfDaySeconds"</span><span class=o>,</span> <span class=n>FieldType</span><span class=o>.</span><span class=na>INT32</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>field</span><span class=o>(</span><span class=s>"shippingAddress.deliveryNotes"</span><span class=o>,</span> <span class=n>FieldType</span><span class=o>.</span><span class=na>STRING</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>field</span><span class=o>(</span><span class=s>"transactions.isFlagged"</span><span class=o>,</span> <span class=n>FieldType</span><span class=o>.</span><span class=na>BOOLEAN</span><span class=o>,</span> <span class=kc>false</span><span class=o>));</span></span></span></code></pre></div></div></div><p>Results in a <code>PCollection</code> with an expanded schema. All of the rows and fields of the input, but also with the specified |
| fields added to the schema. All resulting rows will have null values filled in for the <strong>timeOfDaySeconds</strong> and the |
| <strong>shippingAddress.deliveryNotes</strong> fields, and a false value filled in for the <strong>transactions.isFlagged</strong> field.</p><h5 id=removing-fields-from-a-schema><strong>Removing fields from a schema</strong></h5><p><code>DropFields</code> allows specific fields to be dropped from a schema. Input rows will have their schemas truncated, and any |
| values for dropped fields will be removed from the output. Nested fields can also be dropped using the field selection |
| syntax.</p><p>For example, the following snippet</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>DropFields</span><span class=o>.</span><span class=na>fields</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"shippingAddress.streetAddress"</span><span class=o>));</span></span></span></code></pre></div></div></div><p>Results in a copy of the input with those two fields and their corresponding values removed.</p><h5 id=renaming-schema-fields><strong>Renaming schema fields</strong></h5><p><code>RenameFields</code> allows specific fields in a schema to be renamed. The field values in input rows are left unchanged, only |
| the schema is modified. This transform is often used to prepare records for output to a schema-aware sink, such as an |
| RDBMS, to make sure that the <code>PCollection</code> schema field names match that of the output. It can also be used to rename |
| fields generated by other transforms to make them more usable (similar to SELECT AS in SQL). Nested fields can also be |
| renamed using the field-selection syntax.</p><p>For example, the following snippet</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>RenameFields</span><span class=o>.<</span><span class=n>PurchasePojo</span><span class=o>></span><span class=n>create</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>rename</span><span class=o>(</span><span class=s>"userId"</span><span class=o>,</span> <span class=s>"userIdentifier"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>rename</span><span class=o>(</span><span class=s>"shippingAddress.streetAddress"</span><span class=o>,</span> <span class=s>"shippingAddress.street"</span><span class=o>));</span></span></span></code></pre></div></div></div><p>Results in the same set of unmodified input elements, however the schema on the PCollection has been changed to rename |
| <strong>userId</strong> to <strong>userIdentifier</strong> and <strong>shippingAddress.streetAddress</strong> to <strong>shippingAddress.street</strong>.</p><h5 id=converting-between-types><strong>Converting between types</strong></h5><p>As mentioned, Beam can automatically convert between different Java types, as long as those types have equivalent |
| schemas. One way to do this is by using the <code>Convert</code> transform, as follows.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>PurchaseBean</span><span class=o>></span> <span class=n>purchaseBeans</span> <span class=o>=</span> <span class=n>readPurchasesAsBeans</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>PurchasePojo</span><span class=o>></span> <span class=n>pojoPurchases</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>purchaseBeans</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Convert</span><span class=o>.</span><span class=na>to</span><span class=o>(</span><span class=n>PurchasePojo</span><span class=o>.</span><span class=na>class</span><span class=o>));</span></span></span></code></pre></div></div></div><p>Beam will validate that the inferred schema for <code>PurchasePojo</code> matches that of the input <code>PCollection</code>, and will |
| then cast to a <code>PCollection<PurchasePojo></code>.</p><p>Since the <code>Row</code> class can support any schema, any <code>PCollection</code> with schema can be cast to a <code>PCollection</code> of rows, as |
| follows.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Row</span><span class=o>></span> <span class=n>purchaseRows</span> <span class=o>=</span> <span class=n>purchaseBeans</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Convert</span><span class=o>.</span><span class=na>toRows</span><span class=o>());</span></span></span></code></pre></div></div></div><p>If the source type is a single-field schema, Convert will also convert to the type of the field if asked, effectively |
| unboxing the row. For example, give a schema with a single INT64 field, the following will convert it to a |
| <code>PCollection<Long></code></p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>Long</span><span class=o>></span> <span class=n>longs</span> <span class=o>=</span> <span class=n>rows</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Convert</span><span class=o>.</span><span class=na>to</span><span class=o>(</span><span class=n>TypeDescriptors</span><span class=o>.</span><span class=na>longs</span><span class=o>()));</span></span></span></code></pre></div></div></div><p>In all cases, type checking is done at pipeline graph construction, and if the types do not match the schema then the |
| pipeline will fail to launch.</p><h4 id=663-schemas-in-pardo>6.6.3. Schemas in ParDo</h4><p>A <code>PCollection</code> with a schema can apply a <code>ParDo</code>, just like any other <code>PCollection</code>. However the Beam runner is aware |
| of schemas when applying a <code>ParDo</code>, which enables additional functionality.</p><h5 id=input-conversion><strong>Input conversion</strong></h5><p class=language-go>Beam does not yet support input conversion in Go.</p><p>Since Beam knows the schema of the source <code>PCollection</code>, it can automatically convert the elements to any Java type for |
| which a matching schema is known. For example, using the above-mentioned Transaction schema, say we have the following |
| <code>PCollection</code>:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>PurchasePojo</span><span class=o>></span> <span class=n>purchases</span> <span class=o>=</span> <span class=n>readPurchases</span><span class=o>();</span></span></span></code></pre></div></div></div><p>If there were no schema, then the applied <code>DoFn</code> would have to accept an element of type <code>TransactionPojo</code>. However |
| since there is a schema, you could apply the following DoFn:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>PurchasePojo</span><span class=o>,</span> <span class=n>PurchasePojo</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>PurchaseBean</span> <span class=n>purchase</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><p>Even though the <code>@Element</code> parameter does not match the Java type of the <code>PCollection</code>, since it has a matching schema |
| Beam will automatically convert elements. If the schema does not match, Beam will detect this at graph-construction time |
| and will fail the job with a type error.</p><p>Since every schema can be represented by a Row type, Row can also be used here:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>appy</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>PurchasePojo</span><span class=o>,</span> <span class=n>PurchasePojo</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>Row</span> <span class=n>purchase</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><h5 id=input-selection><strong>Input selection</strong></h5><p>Since the input has a schema, you can also automatically select specific fields to process in the DoFn.</p><p>Given the above purchases <code>PCollection</code>, say you want to process just the userId and the itemId fields. You can do these |
| using the above-described selection expressions, as follows:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>appy</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>PurchasePojo</span><span class=o>,</span> <span class=n>PurchasePojo</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@FieldAccess</span><span class=o>(</span><span class=s>"userId"</span><span class=o>)</span> <span class=n>String</span> <span class=n>userId</span><span class=o>,</span> <span class=nd>@FieldAccess</span><span class=o>(</span><span class=s>"itemId"</span><span class=o>)</span> <span class=kt>long</span> <span class=n>itemId</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><p>You can also select nested fields, as follows.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>purchases</span><span class=o>.</span><span class=na>appy</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>PurchasePojo</span><span class=o>,</span> <span class=n>PurchasePojo</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@FieldAccess</span><span class=o>(</span><span class=s>"shippingAddress.street"</span><span class=o>)</span> <span class=n>String</span> <span class=n>street</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><p>For more information, see the section on field-selection expressions. When selecting subschemas, Beam will |
| automatically convert to any matching schema type, just like when reading the entire row.</p><h2 id=data-encoding-and-type-safety>7. Data encoding and type safety</h2><nav class=language-switcher><strong>Adapt for:</strong><ul><li data-value=java class=active>Java SDK</li><li data-value=py>Python SDK</li><li data-value=go>Go SDK</li><li data-value=typescript>TypeScript SDK</li></ul></nav><p>When Beam runners execute your pipeline, they often need to materialize the |
| intermediate data in your <code>PCollection</code>s, which requires converting elements to |
| and from byte strings. The Beam SDKs use objects called <code>Coder</code>s to describe how |
| the elements of a given <code>PCollection</code> may be encoded and decoded.</p><blockquote><p>Note that coders are unrelated to parsing or formatting data when interacting |
| with external data sources or sinks. Such parsing or formatting should |
| typically be done explicitly, using transforms such as <code>ParDo</code> or |
| <code>MapElements</code>.</p></blockquote><p class=language-java>In the Beam SDK for Java, the type <code>Coder</code> provides the methods required for |
| encoding and decoding data. The SDK for Java provides a number of Coder |
| subclasses that work with a variety of standard Java types, such as Integer, |
| Long, Double, StringUtf8 and more. You can find all of the available Coder |
| subclasses in the <a href=https://github.com/apache/beam/tree/master/sdks/java/core/src/main/java/org/apache/beam/sdk/coders>Coder package</a>.</p><p class=language-py>In the Beam SDK for Python, the type <code>Coder</code> provides the methods required for |
| encoding and decoding data. The SDK for Python provides a number of Coder |
| subclasses that work with a variety of standard Python types, such as primitive |
| types, Tuple, Iterable, StringUtf8 and more. You can find all of the available |
| Coder subclasses in the |
| <a href=https://github.com/apache/beam/tree/master/sdks/python/apache_beam/coders>apache_beam.coders</a> |
| package.</p><p class=language-go>Standard Go types like <code>int</code>, <code>int64</code> <code>float64</code>, <code>[]byte</code>, and <code>string</code> and more are coded using builtin coders. |
| Structs and pointers to structs default using Beam Schema Row encoding. |
| However, users can build and register custom coders with <code>beam.RegisterCoder</code>. |
| You can find available Coder functions in the |
| <a href=https://pkg.go.dev/github.com/apache/beam/sdks/go/pkg/beam/core/graph/coder>coder</a> |
| package.</p><p class=language-typescript>Standard Typescript types like <code>number</code>, <code>UInt8Array</code> and <code>string</code> and more are coded using builtin coders. |
| Json objects and arrays are encoded via a BSON encoding. |
| For these types, coders need not be specified unless interacting with cross-language transforms. |
| Users can build custom coders by extending <code>beam.coders.Coder</code> |
| for use with <code>withCoderInternal</code>, but generally logical types are preferred for this case.</p><blockquote><p>Note that coders do not necessarily have a 1:1 relationship with types. For |
| example, the Integer type can have multiple valid coders, and input and output |
| data can use different Integer coders. A transform might have Integer-typed |
| input data that uses BigEndianIntegerCoder, and Integer-typed output data that |
| uses VarIntCoder.</p></blockquote><h3 id=specifying-coders>7.1. Specifying coders</h3><p>The Beam SDKs require a coder for every <code>PCollection</code> in your pipeline. In most |
| cases, the Beam SDK is able to automatically infer a <code>Coder</code> for a <code>PCollection</code> |
| based on its element type or the transform that produces it, however, in some |
| cases the pipeline author will need to specify a <code>Coder</code> explicitly, or develop |
| a <code>Coder</code> for their custom type.</p><p class=language-java>You can explicitly set the coder for an existing <code>PCollection</code> by using the |
| method <code>PCollection.setCoder</code>. Note that you cannot call <code>setCoder</code> on a |
| <code>PCollection</code> that has been finalized (e.g. by calling <code>.apply</code> on it).</p><p class=language-java>You can get the coder for an existing <code>PCollection</code> by using the method |
| <code>getCoder</code>. This method will fail with an <code>IllegalStateException</code> if a coder has |
| not been set and cannot be inferred for the given <code>PCollection</code>.</p><p>Beam SDKs use a variety of mechanisms when attempting to automatically infer the |
| <code>Coder</code> for a <code>PCollection</code>.</p><p class=language-java>Each pipeline object has a <code>CoderRegistry</code>. The <code>CoderRegistry</code> represents a |
| mapping of Java types to the default coders that the pipeline should use for |
| <code>PCollection</code>s of each type.</p><p class=language-py>The Beam SDK for Python has a <code>CoderRegistry</code> that represents a mapping of |
| Python types to the default coder that should be used for <code>PCollection</code>s of each |
| type.</p><p class=language-go>The Beam SDK for Go allows users to register default coder |
| implementations with <code>beam.RegisterCoder</code>.</p><p class=language-java>By default, the Beam SDK for Java automatically infers the <code>Coder</code> for the |
| elements of a <code>PCollection</code> produced by a <code>PTransform</code> using the type parameter |
| from the transform’s function object, such as <code>DoFn</code>. In the case of <code>ParDo</code>, |
| for example, a <code>DoFn<Integer, String></code> function object accepts an input element |
| of type <code>Integer</code> and produces an output element of type <code>String</code>. In such a |
| case, the SDK for Java will automatically infer the default <code>Coder</code> for the |
| output <code>PCollection<String></code> (in the default pipeline <code>CoderRegistry</code>, this is |
| <code>StringUtf8Coder</code>).</p><p class=language-py>By default, the Beam SDK for Python automatically infers the <code>Coder</code> for the |
| elements of an output <code>PCollection</code> using the typehints from the transform’s |
| function object, such as <code>DoFn</code>. In the case of <code>ParDo</code>, for example a <code>DoFn</code> |
| with the typehints <code>@beam.typehints.with_input_types(int)</code> and |
| <code>@beam.typehints.with_output_types(str)</code> accepts an input element of type int |
| and produces an output element of type str. In such a case, the Beam SDK for |
| Python will automatically infer the default <code>Coder</code> for the output <code>PCollection</code> |
| (in the default pipeline <code>CoderRegistry</code>, this is <code>BytesCoder</code>).</p><p class=language-go>By default, the Beam SDK for Go automatically infers the <code>Coder</code> for the elements of an output <code>PCollection</code> by the output of the transform’s function object, such as a <code>DoFn</code>. |
| In the case of <code>ParDo</code>, for example a <code>DoFn</code> |
| with the parameters of <code>v int, emit func(string)</code> accepts an input element of type <code>int</code> |
| and produces an output element of type <code>string</code>. |
| In such a case, the Beam SDK for Go will automatically infer the default <code>Coder</code> for the output <code>PCollection</code> to be the <code>string_utf8</code> coder.</p><span class=language-java><blockquote><p><strong>Note:</strong> If you create your <code>PCollection</code> from in-memory data by using the |
| <code>Create</code> transform, you cannot rely on coder inference and default coders. |
| <code>Create</code> does not have access to any typing information for its arguments, and |
| may not be able to infer a coder if the argument list contains a value whose |
| exact run-time class doesn’t have a default coder registered.</p></blockquote></span><p class=language-java>When using <code>Create</code>, the simplest way to ensure that you have the correct coder |
| is by invoking <code>withCoder</code> when you apply the <code>Create</code> transform.</p><h3 id=default-coders-and-the-coderregistry>7.2. Default coders and the CoderRegistry</h3><p>Each Pipeline object has a <code>CoderRegistry</code> object, which maps language types to |
| the default coder the pipeline should use for those types. You can use the |
| <code>CoderRegistry</code> yourself to look up the default coder for a given type, or to |
| register a new default coder for a given type.</p><p><code>CoderRegistry</code> contains a default mapping of coders to standard |
| <span class=language-java>Java</span><span class=language-py>Python</span> |
| types for any pipeline you create using the Beam SDK for |
| <span class=language-java>Java</span><span class=language-py>Python</span>. |
| The following table shows the standard mapping:</p><p class=language-java><div class=table-container-wrapper><table class=table-wrapper--pr><thead><tr class=header><th>Java Type</th><th>Default Coder</th></tr></thead><tbody><tr class=odd><td>Double</td><td>DoubleCoder</td></tr><tr class=even><td>Instant</td><td>InstantCoder</td></tr><tr class=odd><td>Integer</td><td>VarIntCoder</td></tr><tr class=even><td>Iterable</td><td>IterableCoder</td></tr><tr class=odd><td>KV</td><td>KvCoder</td></tr><tr class=even><td>List</td><td>ListCoder</td></tr><tr class=odd><td>Map</td><td>MapCoder</td></tr><tr class=even><td>Long</td><td>VarLongCoder</td></tr><tr class=odd><td>String</td><td>StringUtf8Coder</td></tr><tr class=even><td>TableRow</td><td>TableRowJsonCoder</td></tr><tr class=odd><td>Void</td><td>VoidCoder</td></tr><tr class=even><td>byte[ ]</td><td>ByteArrayCoder</td></tr><tr class=odd><td>TimestampedValue</td><td>TimestampedValueCoder</td></tr></tbody></table></div></p><p class=language-py><table class=table-wrapper--pr><thead><tr class=header><th>Python Type</th><th>Default Coder</th></tr></thead><tbody><tr class=odd><td>int</td><td>VarIntCoder</td></tr><tr class=even><td>float</td><td>FloatCoder</td></tr><tr class=odd><td>str</td><td>BytesCoder</td></tr><tr class=even><td>bytes</td><td>StrUtf8Coder</td></tr><tr class=odd><td>Tuple</td><td>TupleCoder</td></tr></tbody></table></p><h4 id=default-coder-lookup>7.2.1. Looking up a default coder</h4><p class=language-java>You can use the method <code>CoderRegistry.getCoder</code> to determine the default |
| Coder for a Java type. You can access the <code>CoderRegistry</code> for a given pipeline |
| by using the method <code>Pipeline.getCoderRegistry</code>. This allows you to determine |
| (or set) the default Coder for a Java type on a per-pipeline basis: i.e. “for |
| this pipeline, verify that Integer values are encoded using |
| <code>BigEndianIntegerCoder</code>.”</p><p class=language-py>You can use the method <code>CoderRegistry.get_coder</code> to determine the default Coder |
| for a Python type. You can use <code>coders.registry</code> to access the <code>CoderRegistry</code>. |
| This allows you to determine (or set) the default Coder for a Python type.</p><p class=language-go>You can use the <code>beam.NewCoder</code> function to determine the default Coder for a Go type.</p><h4 id=setting-default-coder>7.2.2. Setting the default coder for a type</h4><p class="language-java language-py">To set the default Coder for a |
| <span class=language-java>Java</span><span class=language-py>Python</span> |
| type for a particular pipeline, you obtain and modify the pipeline’s |
| <code>CoderRegistry</code>. You use the method |
| <span class=language-java><code>Pipeline.getCoderRegistry</code></span> |
| <span class=language-py><code>coders.registry</code></span> |
| to get the <code>CoderRegistry</code> object, and then use the method |
| <span class=language-java><code>CoderRegistry.registerCoder</code></span> |
| <span class=language-py><code>CoderRegistry.register_coder</code></span> |
| to register a new <code>Coder</code> for the target type.</p><p class=language-go>To set the default Coder for a Go type you use the function <code>beam.RegisterCoder</code> to register a encoder and decoder functions for the target type. |
| However, built in types like <code>int</code>, <code>string</code>, <code>float64</code>, etc cannot have their coders override.</p><p class="language-java language-py">The following example code demonstrates how to set a default Coder, in this case |
| <code>BigEndianIntegerCoder</code>, for |
| <span class=language-java>Integer</span><span class=language-py>int</span> |
| values for a pipeline.</p><p class=language-go>The following example code demonstrates how to set a custom Coder for <code>MyCustomType</code> elements.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PipelineOptions</span> <span class=n>options</span> <span class=o>=</span> <span class=n>PipelineOptionsFactory</span><span class=o>.</span><span class=na>create</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>Pipeline</span> <span class=n>p</span> <span class=o>=</span> <span class=n>Pipeline</span><span class=o>.</span><span class=na>create</span><span class=o>(</span><span class=n>options</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>CoderRegistry</span> <span class=n>cr</span> <span class=o>=</span> <span class=n>p</span><span class=o>.</span><span class=na>getCoderRegistry</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>cr</span><span class=o>.</span><span class=na>registerCoder</span><span class=o>(</span><span class=n>Integer</span><span class=o>.</span><span class=na>class</span><span class=o>,</span> <span class=n>BigEndianIntegerCoder</span><span class=o>.</span><span class=na>class</span><span class=o>);</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>apache_beam</span><span class=o>.</span><span class=n>coders</span><span class=o>.</span><span class=n>registry</span><span class=o>.</span><span class=n>register_coder</span><span class=p>(</span><span class=nb>int</span><span class=p>,</span> <span class=n>BigEndianIntegerCoder</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>MyCustomType</span> <span class=kd>struct</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// See documentation on beam.RegisterCoder for other supported coder forms. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>encode</span><span class=p>(</span><span class=nx>MyCustomType</span><span class=p>)</span> <span class=p>[]</span><span class=kt>byte</span> <span class=p>{</span> <span class=o>...</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>decode</span><span class=p>(</span><span class=nx>b</span> <span class=p>[]</span><span class=kt>byte</span><span class=p>)</span> <span class=nx>MyCustomType</span> <span class=p>{</span> <span class=o>...</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>RegisterCoder</span><span class=p>(</span><span class=nx>reflect</span><span class=p>.</span><span class=nf>TypeOf</span><span class=p>((</span><span class=o>*</span><span class=nx>MyCustomType</span><span class=p>)(</span><span class=kc>nil</span><span class=p>)).</span><span class=nf>Elem</span><span class=p>(),</span> <span class=nx>encode</span><span class=p>,</span> <span class=nx>decode</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h4 id=annotating-custom-type-default-coder>7.2.3. Annotating a custom data type with a default coder</h4><span class=language-java><p>If your pipeline program defines a custom data type, you can use the |
| <code>@DefaultCoder</code> annotation to specify the coder to use with that type. |
| By default, Beam will use <code>SerializableCoder</code> which uses Java serialization, |
| but it has drawbacks:</p><ol><li><p>It is inefficient in encoding size and speed. |
| See this <a href=https://blog.softwaremill.com/the-best-serialization-strategy-for-event-sourcing-9321c299632b>comparison of Java serialization methods.</a></p></li><li><p>It is non-deterministic: it may produce different binary encodings for two |
| equivalent objects.</p><p>For key/value pairs, the correctness of key-based operations |
| (GroupByKey, Combine) and per-key State depends on having a deterministic |
| coder for the key</p></li></ol><p>You can use the <code>@DefaultCoder</code> annotation to set a new default as follows:</p></span><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@DefaultCoder</span><span class=o>(</span><span class=n>AvroCoder</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>MyCustomDataType</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><p class=language-java>If you’ve created a custom coder to match your data type, and you want to use |
| the <code>@DefaultCoder</code> annotation, your coder class must implement a static |
| <code>Coder.of(Class<T>)</code> factory method.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>MyCustomCoder</span> <span class=kd>implements</span> <span class=n>Coder</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>static</span> <span class=n>Coder</span><span class=o><</span><span class=n>T</span><span class=o>></span> <span class=nf>of</span><span class=o>(</span><span class=n>Class</span><span class=o><</span><span class=n>T</span><span class=o>></span> <span class=n>clazz</span><span class=o>)</span> <span class=o>{...}</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nd>@DefaultCoder</span><span class=o>(</span><span class=n>MyCustomCoder</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>MyCustomDataType</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><p class="language-py language-go">The Beam SDK for <span class=language-py>Python</span><span class=language-go>Go</span> |
| does not support annotating data types with a default coder. |
| If you would like to set a default coder, use the method described in the |
| previous section, <em>Setting the default coder for a type</em>.</p><h2 id=windowing>8. Windowing</h2><p>Windowing subdivides a <code>PCollection</code> according to the timestamps of its |
| individual elements. Transforms that aggregate multiple elements, such as |
| <code>GroupByKey</code> and <code>Combine</code>, work implicitly on a per-window basis — they process |
| each <code>PCollection</code> as a succession of multiple, finite windows, though the |
| entire collection itself may be of unbounded size.</p><p>A related concept, called <strong>triggers</strong>, determines when to emit the results of |
| aggregation as unbounded data arrives. You can use triggers to refine the |
| windowing strategy for your <code>PCollection</code>. Triggers allow you to deal with |
| late-arriving data or to provide early results. See the <a href=#triggers>triggers</a> |
| section for more information.</p><h3 id=windowing-basics>8.1. Windowing basics</h3><p>Some Beam transforms, such as <code>GroupByKey</code> and <code>Combine</code>, group multiple |
| elements by a common key. Ordinarily, that grouping operation groups all of the |
| elements that have the same key within the entire data set. With an unbounded |
| data set, it is impossible to collect all of the elements, since new elements |
| are constantly being added and may be infinitely many (e.g. streaming data). If |
| you are working with unbounded <code>PCollection</code>s, windowing is especially useful.</p><p>In the Beam model, any <code>PCollection</code> (including unbounded <code>PCollection</code>s) can be |
| subdivided into logical windows. Each element in a <code>PCollection</code> is assigned to |
| one or more windows according to the <code>PCollection</code>’s windowing function, and |
| each individual window contains a finite number of elements. Grouping transforms |
| then consider each <code>PCollection</code>’s elements on a per-window basis. <code>GroupByKey</code>, |
| for example, implicitly groups the elements of a <code>PCollection</code> by <em>key and |
| window</em>.</p><p><strong>Caution:</strong> Beam’s default windowing behavior is to assign all elements of a |
| <code>PCollection</code> to a single, global window and discard late data, <em>even for |
| unbounded <code>PCollection</code>s</em>. Before you use a grouping transform such as |
| <code>GroupByKey</code> on an unbounded <code>PCollection</code>, you must do at least one of the |
| following:</p><ul><li>Set a non-global windowing function. See <a href=#setting-your-pcollections-windowing-function>Setting your PCollection’s |
| windowing function</a>.</li><li>Set a non-default <a href=#triggers>trigger</a>. This allows the global window to emit |
| results under other conditions, since the default windowing behavior (waiting |
| for all data to arrive) will never occur.</li></ul><p>If you don’t set a non-global windowing function or a non-default trigger for |
| your unbounded <code>PCollection</code> and subsequently use a grouping transform such as |
| <code>GroupByKey</code> or <code>Combine</code>, your pipeline will generate an error upon |
| construction and your job will fail.</p><h4 id=windowing-constraints>8.1.1. Windowing constraints</h4><p>After you set the windowing function for a <code>PCollection</code>, the elements’ windows |
| are used the next time you apply a grouping transform to that <code>PCollection</code>. |
| Window grouping occurs on an as-needed basis. If you set a windowing function |
| using the <code>Window</code> transform, each element is assigned to a window, but the |
| windows are not considered until <code>GroupByKey</code> or <code>Combine</code> aggregates across a |
| window and key. This can have different effects on your pipeline. Consider the |
| example pipeline in the figure below:</p><p><img src=/images/windowing-pipeline-unbounded.svg alt="Diagram of pipeline applying windowing"></p><p><strong>Figure 3:</strong> Pipeline applying windowing</p><p>In the above pipeline, we create an unbounded <code>PCollection</code> by reading a set of |
| key/value pairs using <code>KafkaIO</code>, and then apply a windowing function to that |
| collection using the <code>Window</code> transform. We then apply a <code>ParDo</code> to the |
| collection, and then later group the result of that <code>ParDo</code> using <code>GroupByKey</code>. |
| The windowing function has no effect on the <code>ParDo</code> transform, because the |
| windows are not actually used until they’re needed for the <code>GroupByKey</code>. |
| Subsequent transforms, however, are applied to the result of the <code>GroupByKey</code> – |
| data is grouped by both key and window.</p><h4 id=windowing-bounded-collections>8.1.2. Windowing with bounded PCollections</h4><p>You can use windowing with fixed-size data sets in <strong>bounded</strong> <code>PCollection</code>s. |
| However, note that windowing considers only the implicit timestamps attached to |
| each element of a <code>PCollection</code>, and data sources that create fixed data sets |
| (such as <code>TextIO</code>) assign the same timestamp to every element. This means that |
| all the elements are by default part of a single, global window.</p><p>To use windowing with fixed data sets, you can assign your own timestamps to |
| each element. To assign timestamps to elements, use a <code>ParDo</code> transform with a |
| <code>DoFn</code> that outputs each element with a new timestamp (for example, the |
| <a href=https://beam.apache.org/releases/javadoc/2.55.1/index.html?org/apache/beam/sdk/transforms/WithTimestamps.html>WithTimestamps</a> |
| transform in the Beam SDK for Java).</p><p>To illustrate how windowing with a bounded <code>PCollection</code> can affect how your |
| pipeline processes data, consider the following pipeline:</p><p><img src=/images/unwindowed-pipeline-bounded.svg alt="Diagram of GroupByKey and ParDo without windowing, on a bounded collection"></p><p><strong>Figure 4:</strong> <code>GroupByKey</code> and <code>ParDo</code> without windowing, on a bounded collection.</p><p>In the above pipeline, we create a bounded <code>PCollection</code> by reading lines from a |
| file using <code>TextIO</code>. We then group the collection using <code>GroupByKey</code>, |
| and apply a <code>ParDo</code> transform to the grouped <code>PCollection</code>. In this example, the |
| <code>GroupByKey</code> creates a collection of unique keys, and then <code>ParDo</code> gets applied |
| exactly once per key.</p><p>Note that even if you don’t set a windowing function, there is still a window – |
| all elements in your <code>PCollection</code> are assigned to a single global window.</p><p>Now, consider the same pipeline, but using a windowing function:</p><p><img src=/images/windowing-pipeline-bounded.svg alt="Diagram of GroupByKey and ParDo with windowing, on a bounded collection"></p><p><strong>Figure 5:</strong> <code>GroupByKey</code> and <code>ParDo</code> with windowing, on a bounded collection.</p><p>As before, the pipeline creates a bounded <code>PCollection</code> by reading lines from a |
| file. We then set a <a href=#setting-your-pcollections-windowing-function>windowing function</a> |
| for that <code>PCollection</code>. The <code>GroupByKey</code> transform groups the elements of the |
| <code>PCollection</code> by both key and window, based on the windowing function. The |
| subsequent <code>ParDo</code> transform gets applied multiple times per key, once for each |
| window.</p><h3 id=provided-windowing-functions>8.2. Provided windowing functions</h3><p>You can define different kinds of windows to divide the elements of your |
| <code>PCollection</code>. Beam provides several windowing functions, including:</p><ul><li>Fixed Time Windows</li><li>Sliding Time Windows</li><li>Per-Session Windows</li><li>Single Global Window</li><li>Calendar-based Windows (not supported by the Beam SDK for Python or Go)</li></ul><p>You can also define your own <code>WindowFn</code> if you have a more complex need.</p><p>Note that each element can logically belong to more than one window, depending |
| on the windowing function you use. Sliding time windowing, for example, can |
| create overlapping windows wherein a single element can be assigned to multiple |
| windows. However, each element in a <code>PCollection</code> can only be in one window, so |
| if an element is assigned to multiple windows, the element is conceptually |
| duplicated into each of the windows and each element is identical except for its |
| window.</p><h4 id=fixed-time-windows>8.2.1. Fixed time windows</h4><p>The simplest form of windowing is using <strong>fixed time windows</strong>: given a |
| timestamped <code>PCollection</code> which might be continuously updating, each window |
| might capture (for example) all elements with timestamps that fall into a 30 |
| second interval.</p><p>A fixed time window represents a consistent duration, non overlapping time |
| interval in the data stream. Consider windows with a 30 second duration: all |
| of the elements in your unbounded <code>PCollection</code> with timestamp values from |
| 0:00:00 up to (but not including) 0:00:30 belong to the first window, elements |
| with timestamp values from 0:00:30 up to (but not including) 0:01:00 belong to |
| the second window, and so on.</p><p><img src=/images/fixed-time-windows.png alt="Diagram of fixed time windows, 30s in duration"></p><p><strong>Figure 6:</strong> Fixed time windows, 30s in duration.</p><h4 id=sliding-time-windows>8.2.2. Sliding time windows</h4><p>A <strong>sliding time window</strong> also represents time intervals in the data stream; |
| however, sliding time windows can overlap. For example, each window might |
| capture 60 seconds worth of data, but a new window starts every 30 seconds. |
| The frequency with which sliding windows begin is called the <em>period</em>. |
| Therefore, our example would have a window <em>duration</em> of 60 seconds and a |
| <em>period</em> of 30 seconds.</p><p>Because multiple windows overlap, most elements in a data set will belong to |
| more than one window. This kind of windowing is useful for taking running |
| averages of data; using sliding time windows, you can compute a running average |
| of the past 60 seconds’ worth of data, updated every 30 seconds, in our |
| example.</p><p><img src=/images/sliding-time-windows.png alt="Diagram of sliding time windows, with 1 minute window duration and 30s window period"></p><p><strong>Figure 7:</strong> Sliding time windows, with 1 minute window duration and 30s window |
| period.</p><h4 id=session-windows>8.2.3. Session windows</h4><p>A <strong>session window</strong> function defines windows that contain elements that are |
| within a certain gap duration of another element. Session windowing applies on a |
| per-key basis and is useful for data that is irregularly distributed with |
| respect to time. For example, a data stream representing user mouse activity may |
| have long periods of idle time interspersed with high concentrations of clicks. |
| If data arrives after the minimum specified gap duration time, this initiates |
| the start of a new window.</p><p><img src=/images/session-windows.png alt="Diagram of session windows with a minimum gap duration"></p><p><strong>Figure 8:</strong> Session windows, with a minimum gap duration. Note how each data key |
| has different windows, according to its data distribution.</p><h4 id=single-global-window>8.2.4. The single global window</h4><p>By default, all data in a <code>PCollection</code> is assigned to the single global window, |
| and late data is discarded. If your data set is of a fixed size, you can use the |
| global window default for your <code>PCollection</code>.</p><p>You can use the single global window if you are working with an unbounded data set |
| (e.g. from a streaming data source) but use caution when applying aggregating |
| transforms such as <code>GroupByKey</code> and <code>Combine</code>. The single global window with a |
| default trigger generally requires the entire data set to be available before |
| processing, which is not possible with continuously updating data. To perform |
| aggregations on an unbounded <code>PCollection</code> that uses global windowing, you |
| should specify a non-default trigger for that <code>PCollection</code>.</p><h3 id=setting-your-pcollections-windowing-function>8.3. Setting your PCollection’s windowing function</h3><p>You can set the windowing function for a <code>PCollection</code> by applying the <code>Window</code> |
| transform. When you apply the <code>Window</code> transform, you must provide a <code>WindowFn</code>. |
| The <code>WindowFn</code> determines the windowing function your <code>PCollection</code> will use for |
| subsequent grouping transforms, such as a fixed or sliding time window.</p><p>When you set a windowing function, you may also want to set a trigger for your |
| <code>PCollection</code>. The trigger determines when each individual window is aggregated |
| and emitted, and helps refine how the windowing function performs with respect |
| to late data and computing early results. See the <a href=#triggers>triggers</a> section |
| for more information.</p><p class=language-yaml>In Beam YAML windowing specifications can also be placed directly on any |
| transform rather than requiring an explicit <code>WindowInto</code> transform.</p><h4 id=using-fixed-time-windows>8.3.1. Fixed-time windows</h4><p>The following example code shows how to apply <code>Window</code> to divide a <code>PCollection</code> |
| into fixed windows, each 60 seconds in length:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>items</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>fixedWindowedItems</span> <span class=o>=</span> <span class=n>items</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Window</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>into</span><span class=o>(</span><span class=n>FixedWindows</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>60</span><span class=o>))));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam</span> <span class=kn>import</span> <span class=n>window</span> |
| </span></span><span class=line><span class=cl><span class=n>fixed_windowed_items</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>items</span> <span class=o>|</span> <span class=s1>'window'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>WindowInto</span><span class=p>(</span><span class=n>window</span><span class=o>.</span><span class=n>FixedWindows</span><span class=p>(</span><span class=mi>60</span><span class=p>)))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>fixedWindowedItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewFixedWindows</span><span class=p>(</span><span class=mi>60</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>items</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=nx>pcoll</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>windowInto</span><span class=p>(</span><span class=nx>windowings</span><span class=p>.</span><span class=nx>fixedWindows</span><span class=p>(</span><span class=mi>60</span><span class=p>)))</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>WindowInto</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>windowing</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>fixed</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>size</span><span class=p>:</span><span class=w> </span><span class=l>60s</span></span></span></code></pre></div></div></div><h4 id=using-sliding-time-windows>8.3.2. Sliding time windows</h4><p>The following example code shows how to apply <code>Window</code> to divide a <code>PCollection</code> |
| into sliding time windows. Each window is 30 seconds in length, and a new window |
| begins every five seconds:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>items</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>slidingWindowedItems</span> <span class=o>=</span> <span class=n>items</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Window</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>into</span><span class=o>(</span><span class=n>SlidingWindows</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>30</span><span class=o>)).</span><span class=na>every</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>5</span><span class=o>))));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam</span> <span class=kn>import</span> <span class=n>window</span> |
| </span></span><span class=line><span class=cl><span class=n>sliding_windowed_items</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>items</span> <span class=o>|</span> <span class=s1>'window'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>WindowInto</span><span class=p>(</span><span class=n>window</span><span class=o>.</span><span class=n>SlidingWindows</span><span class=p>(</span><span class=mi>30</span><span class=p>,</span> <span class=mi>5</span><span class=p>)))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>slidingWindowedItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewSlidingWindows</span><span class=p>(</span><span class=mi>5</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>,</span> <span class=mi>30</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>items</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=nx>pcoll</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>windowInto</span><span class=p>(</span><span class=nx>windowings</span><span class=p>.</span><span class=nx>slidingWindows</span><span class=p>(</span><span class=mi>30</span><span class=p>,</span> <span class=mi>5</span><span class=p>)))</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>WindowInto</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>windowing</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>sliding</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>size</span><span class=p>:</span><span class=w> </span><span class=l>5m</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>period</span><span class=p>:</span><span class=w> </span><span class=l>30s</span></span></span></code></pre></div></div></div><h4 id=using-session-windows>8.3.3. Session windows</h4><p>The following example code shows how to apply <code>Window</code> to divide a <code>PCollection</code> |
| into session windows, where each session must be separated by a time gap of at |
| least 10 minutes (600 seconds):</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>items</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>sessionWindowedItems</span> <span class=o>=</span> <span class=n>items</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Window</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>into</span><span class=o>(</span><span class=n>Sessions</span><span class=o>.</span><span class=na>withGapDuration</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>600</span><span class=o>))));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam</span> <span class=kn>import</span> <span class=n>window</span> |
| </span></span><span class=line><span class=cl><span class=n>session_windowed_items</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>items</span> <span class=o>|</span> <span class=s1>'window'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>WindowInto</span><span class=p>(</span><span class=n>window</span><span class=o>.</span><span class=n>Sessions</span><span class=p>(</span><span class=mi>10</span> <span class=o>*</span> <span class=mi>60</span><span class=p>)))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>sessionWindowedItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewSessions</span><span class=p>(</span><span class=mi>600</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>items</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=nx>pcoll</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>windowInto</span><span class=p>(</span><span class=nx>windowings</span><span class=p>.</span><span class=nx>sessions</span><span class=p>(</span><span class=mi>10</span> <span class=o>*</span> <span class=mi>60</span><span class=p>)))</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>WindowInto</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>windowing</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>sessions</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>gap</span><span class=p>:</span><span class=w> </span><span class=l>60s</span></span></span></code></pre></div></div></div><p>Note that the sessions are per-key — each key in the collection will have its |
| own session groupings depending on the data distribution.</p><h4 id=using-single-global-window>8.3.4. Single global window</h4><p>If your <code>PCollection</code> is bounded (the size is fixed), you can assign all the |
| elements to a single global window. The following example code shows how to set |
| a single global window for a <code>PCollection</code>:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>items</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>batchItems</span> <span class=o>=</span> <span class=n>items</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Window</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>into</span><span class=o>(</span><span class=k>new</span> <span class=n>GlobalWindows</span><span class=o>()));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam</span> <span class=kn>import</span> <span class=n>window</span> |
| </span></span><span class=line><span class=cl><span class=n>global_windowed_items</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>items</span> <span class=o>|</span> <span class=s1>'window'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>WindowInto</span><span class=p>(</span><span class=n>window</span><span class=o>.</span><span class=n>GlobalWindows</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>globalWindowedItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewGlobalWindows</span><span class=p>(),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>items</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=nx>pcoll</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>windowInto</span><span class=p>(</span><span class=nx>windowings</span><span class=p>.</span><span class=nx>globalWindows</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>WindowInto</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>windowing</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>global</span></span></span></code></pre></div></div></div><h3 id=watermarks-and-late-data>8.4. Watermarks and late data</h3><p>In any data processing system, there is a certain amount of lag between the time |
| a data event occurs (the “event time”, determined by the timestamp on the data |
| element itself) and the time the actual data element gets processed at any stage |
| in your pipeline (the “processing time”, determined by the clock on the system |
| processing the element). In addition, there are no guarantees that data events |
| will appear in your pipeline in the same order that they were generated.</p><p>For example, let’s say we have a <code>PCollection</code> that’s using fixed-time |
| windowing, with windows that are five minutes long. For each window, Beam must |
| collect all the data with an <em>event time</em> timestamp in the given window range |
| (between 0:00 and 4:59 in the first window, for instance). Data with timestamps |
| outside that range (data from 5:00 or later) belong to a different window.</p><p>However, data isn’t always guaranteed to arrive in a pipeline in time order, or |
| to always arrive at predictable intervals. Beam tracks a <em>watermark</em>, which is |
| the system’s notion of when all data in a certain window can be expected to have |
| arrived in the pipeline. Once the watermark progresses past the end of a window, |
| any further element that arrives with a timestamp in that window is considered |
| <strong>late data</strong>.</p><p>From our example, suppose we have a simple watermark that assumes approximately |
| 30s of lag time between the data timestamps (the event time) and the time the |
| data appears in the pipeline (the processing time), then Beam would close the |
| first window at 5:30. If a data record arrives at 5:34, but with a timestamp |
| that would put it in the 0:00-4:59 window (say, 3:38), then that record is late |
| data.</p><p>Note: For simplicity, we’ve assumed that we’re using a very straightforward |
| watermark that estimates the lag time. In practice, your <code>PCollection</code>’s data |
| source determines the watermark, and watermarks can be more precise or complex.</p><p>Beam’s default windowing configuration tries to determine when all data has |
| arrived (based on the type of data source) and then advances the watermark past |
| the end of the window. This default configuration does <em>not</em> allow late data. |
| <a href=#triggers>Triggers</a> allow you to modify and refine the windowing strategy for |
| a <code>PCollection</code>. You can use triggers to decide when each individual window |
| aggregates and reports its results, including how the window emits late |
| elements.</p><h4 id=managing-late-data>8.4.1. Managing late data</h4><p>You can allow late data by invoking the <code>.withAllowedLateness</code> operation when |
| you set your <code>PCollection</code>’s windowing strategy. The following code example |
| demonstrates a windowing strategy that will allow late data up to two days after |
| the end of a window.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>items</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>fixedWindowedItems</span> <span class=o>=</span> <span class=n>items</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Window</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>into</span><span class=o>(</span><span class=n>FixedWindows</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>1</span><span class=o>)))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withAllowedLateness</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardDays</span><span class=o>(</span><span class=n>2</span><span class=o>)));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl> <span class=n>pc</span> <span class=o>=</span> <span class=p>[</span><span class=n>Initial</span> <span class=n>PCollection</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=n>pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>WindowInto</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>FixedWindows</span><span class=p>(</span><span class=mi>60</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>trigger</span><span class=o>=</span><span class=n>trigger_fn</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>accumulation_mode</span><span class=o>=</span><span class=n>accumulation_mode</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>timestamp_combiner</span><span class=o>=</span><span class=n>timestamp_combiner</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>allowed_lateness</span><span class=o>=</span><span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>2</span><span class=o>*</span><span class=mi>24</span><span class=o>*</span><span class=mi>60</span><span class=o>*</span><span class=mi>60</span><span class=p>))</span> <span class=c1># 2 days</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>windowedItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewFixedWindows</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>),</span> <span class=nx>items</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>AllowedLateness</span><span class=p>(</span><span class=mi>2</span><span class=o>*</span><span class=mi>24</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Hour</span><span class=p>),</span> <span class=c1>// 2 days |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>)</span></span></span></code></pre></div></div></div><p>When you set <code>.withAllowedLateness</code> on a <code>PCollection</code>, that allowed lateness |
| propagates forward to any subsequent <code>PCollection</code> derived from the first |
| <code>PCollection</code> you applied allowed lateness to. If you want to change the allowed |
| lateness later in your pipeline, you must do so explicitly by applying |
| <code>Window.configure().withAllowedLateness()</code>.</p><h3 id=adding-timestamps-to-a-pcollections-elements>8.5. Adding timestamps to a PCollection’s elements</h3><p>An unbounded source provides a timestamp for each element. Depending on your |
| unbounded source, you may need to configure how the timestamp is extracted from |
| the raw data stream.</p><p>However, bounded sources (such as a file from <code>TextIO</code>) do not provide |
| timestamps. If you need timestamps, you must add them to your <code>PCollection</code>’s |
| elements.</p><p>You can assign new timestamps to the elements of a <code>PCollection</code> by applying a |
| <a href=#pardo>ParDo</a> transform that outputs new elements with timestamps that you |
| set.</p><p>An example might be if your pipeline reads log records from an input file, and |
| each log record includes a timestamp field; since your pipeline reads the |
| records in from a file, the file source doesn’t assign timestamps automatically. |
| You can parse the timestamp field from each record and use a <code>ParDo</code> transform |
| with a <code>DoFn</code> to attach the timestamps to each element in your <code>PCollection</code>.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>LogEntry</span><span class=o>></span> <span class=n>unstampedLogs</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>LogEntry</span><span class=o>></span> <span class=n>stampedLogs</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>unstampedLogs</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>LogEntry</span><span class=o>,</span> <span class=n>LogEntry</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>LogEntry</span> <span class=n>element</span><span class=o>,</span> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>LogEntry</span><span class=o>></span> <span class=n>out</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Extract the timestamp from log entry we're currently processing. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>Instant</span> <span class=n>logTimeStamp</span> <span class=o>=</span> <span class=n>extractTimeStampFromLogEntry</span><span class=o>(</span><span class=n>element</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Use OutputReceiver.outputWithTimestamp (rather than |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// OutputReceiver.output) to emit the entry with timestamp attached. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>out</span><span class=o>.</span><span class=na>outputWithTimestamp</span><span class=o>(</span><span class=n>element</span><span class=o>,</span> <span class=n>logTimeStamp</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>AddTimestampDoFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Extract the numeric Unix seconds-since-epoch timestamp to be</span> |
| </span></span><span class=line><span class=cl> <span class=c1># associated with the current log entry.</span> |
| </span></span><span class=line><span class=cl> <span class=n>unix_timestamp</span> <span class=o>=</span> <span class=n>extract_timestamp_from_log_entry</span><span class=p>(</span><span class=n>element</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Wrap and emit the current entry and new timestamp in a</span> |
| </span></span><span class=line><span class=cl> <span class=c1># TimestampedValue.</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>beam</span><span class=o>.</span><span class=n>window</span><span class=o>.</span><span class=n>TimestampedValue</span><span class=p>(</span><span class=n>element</span><span class=p>,</span> <span class=n>unix_timestamp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>timestamped_items</span> <span class=o>=</span> <span class=n>items</span> <span class=o>|</span> <span class=s1>'timestamp'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>AddTimestampDoFn</span><span class=p>())</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// AddTimestampDoFn extracts an event time from a LogEntry. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=nf>AddTimestampDoFn</span><span class=p>(</span><span class=nx>element</span> <span class=nx>LogEntry</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>LogEntry</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>et</span> <span class=o>:=</span> <span class=nf>extractEventTime</span><span class=p>(</span><span class=nx>element</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Defining an emitter with beam.EventTime as the first parameter |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// allows the DoFn to set the event time for the emitted element. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nf>emit</span><span class=p>(</span><span class=nx>mtime</span><span class=p>.</span><span class=nf>FromTime</span><span class=p>(</span><span class=nx>et</span><span class=p>),</span> <span class=nx>element</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Use the DoFn with ParDo as normal. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl><span class=nx>stampedLogs</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>AddTimestampDoFn</span><span class=p>,</span> <span class=nx>unstampedLogs</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-yaml snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-yaml data-lang=yaml><span class=line><span class=cl><span class=nt>type</span><span class=p>:</span><span class=w> </span><span class=l>AssignTimestamps</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w></span><span class=nt>config</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>language</span><span class=p>:</span><span class=w> </span><span class=l>python</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>timestamp</span><span class=p>:</span><span class=w> |
| </span></span></span><span class=line><span class=cl><span class=w> </span><span class=nt>callable</span><span class=p>:</span><span class=w> </span><span class=p>|</span><span class=sd> |
| </span></span></span><span class=line><span class=cl><span class=sd> import datetime |
| </span></span></span><span class=line><span class=cl><span class=sd> |
| </span></span></span><span class=line><span class=cl><span class=sd> def extract_timestamp(x): |
| </span></span></span><span class=line><span class=cl><span class=sd> raw = datetime.datetime.strptime( |
| </span></span></span><span class=line><span class=cl><span class=sd> x.external_timestamp_field, "%Y-%m-%d") |
| </span></span></span><span class=line><span class=cl><span class=sd> return raw.astimezone(datetime.timezone.utc)</span><span class=w> </span></span></span></code></pre></div></div></div><h2 id=triggers>9. Triggers</h2><span class=language-go><blockquote><p><strong>Note:</strong> The Trigger API in the Beam SDK for Go is currently experimental and subject to change.</p></blockquote></span><p>When collecting and grouping data into windows, Beam uses <strong>triggers</strong> to |
| determine when to emit the aggregated results of each window (referred to as a |
| <em>pane</em>). If you use Beam’s default windowing configuration and <a href=#default-trigger>default |
| trigger</a>, Beam outputs the aggregated result when it |
| <a href=#watermarks-and-late-data>estimates all data has arrived</a>, and discards all |
| subsequent data for that window.</p><p>You can set triggers for your <code>PCollection</code>s to change this default behavior. |
| Beam provides a number of pre-built triggers that you can set:</p><ul><li><strong>Event time triggers</strong>. These triggers operate on the event time, as |
| indicated by the timestamp on each data element. Beam’s default trigger is |
| event time-based.</li><li><strong>Processing time triggers</strong>. These triggers operate on the processing time |
| – the time when the data element is processed at any given stage in the |
| pipeline.</li><li><strong>Data-driven triggers</strong>. These triggers operate by examining the data as it |
| arrives in each window, and firing when that data meets a certain property. |
| Currently, data-driven triggers only support firing after a certain number |
| of data elements.</li><li><strong>Composite triggers</strong>. These triggers combine multiple triggers in various |
| ways.</li></ul><p>At a high level, triggers provide two additional capabilities compared to simply |
| outputting at the end of a window:</p><ul><li>Triggers allow Beam to emit early results, before all the data in a given |
| window has arrived. For example, emitting after a certain amount of time |
| elapses, or after a certain number of elements arrives.</li><li>Triggers allow processing of late data by triggering after the event time |
| watermark passes the end of the window.</li></ul><p>These capabilities allow you to control the flow of your data and balance |
| between different factors depending on your use case:</p><ul><li><strong>Completeness:</strong> How important is it to have all of your data before you |
| compute your result?</li><li><strong>Latency:</strong> How long do you want to wait for data? For example, do you wait |
| until you think you have all data? Do you process data as it arrives?</li><li><strong>Cost:</strong> How much compute power/money are you willing to spend to lower the |
| latency?</li></ul><p>For example, a system that requires time-sensitive updates might use a strict |
| time-based trigger that emits a window every <em>N</em> seconds, valuing promptness |
| over data completeness. A system that values data completeness more than the |
| exact timing of results might choose to use Beam’s default trigger, which fires |
| at the end of the window.</p><p>You can also set a trigger for an unbounded <code>PCollection</code> that uses a <a href=#windowing>single |
| global window for its windowing function</a>. This can be useful when |
| you want your pipeline to provide periodic updates on an unbounded data set — |
| for example, a running average of all data provided to the present time, updated |
| every N seconds or every N elements.</p><h3 id=event-time-triggers>9.1. Event time triggers</h3><p>The <code>AfterWatermark</code> trigger operates on <em>event time</em>. The <code>AfterWatermark</code> |
| trigger emits the contents of a window after the |
| <a href=#watermarks-and-late-data>watermark</a> passes the end of the window, based on the |
| timestamps attached to the data elements. The watermark is a global progress |
| metric, and is Beam’s notion of input completeness within your pipeline at any |
| given point. <span class=language-java><code>AfterWatermark.pastEndOfWindow()</code></span> |
| <span class=language-py><code>AfterWatermark</code></span> |
| <span class=language-go><code>trigger.AfterEndOfWindow</code></span> <em>only</em> fires when the |
| watermark passes the end of the window.</p><p>In addition, you can configure triggers that fire if your pipeline receives data |
| before or after the end of the window.</p><p>The following example shows a billing scenario, and uses both early and late |
| firings:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=c1>// Create a bill at the end of the month. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>AfterWatermark</span><span class=o>.</span><span class=na>pastEndOfWindow</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// During the month, get near real-time estimates. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>.</span><span class=na>withEarlyFirings</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>AfterProcessingTime</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>pastFirstElementInPane</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>plusDuration</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>1</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Fire on any late data so the bill can be corrected. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>.</span><span class=na>withLateFirings</span><span class=o>(</span><span class=n>AfterPane</span><span class=o>.</span><span class=na>elementCountAtLeast</span><span class=o>(</span><span class=n>1</span><span class=o>))</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>AfterWatermark</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>early</span><span class=o>=</span><span class=n>AfterProcessingTime</span><span class=p>(</span><span class=n>delay</span><span class=o>=</span><span class=mi>1</span> <span class=o>*</span> <span class=mi>60</span><span class=p>),</span> <span class=n>late</span><span class=o>=</span><span class=n>AfterCount</span><span class=p>(</span><span class=mi>1</span><span class=p>))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>trigger</span> <span class=o>:=</span> <span class=nx>trigger</span><span class=p>.</span><span class=nf>AfterEndOfWindow</span><span class=p>().</span> |
| </span></span><span class=line><span class=cl> <span class=nf>EarlyFiring</span><span class=p>(</span><span class=nx>trigger</span><span class=p>.</span><span class=nf>AfterProcessingTime</span><span class=p>().</span> |
| </span></span><span class=line><span class=cl> <span class=nf>PlusDelay</span><span class=p>(</span><span class=mi>60</span> <span class=o>*</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>)).</span> |
| </span></span><span class=line><span class=cl> <span class=nf>LateFiring</span><span class=p>(</span><span class=nx>trigger</span><span class=p>.</span><span class=nf>Repeat</span><span class=p>(</span><span class=nx>trigger</span><span class=p>.</span><span class=nf>AfterCount</span><span class=p>(</span><span class=mi>1</span><span class=p>)))</span></span></span></code></pre></div></div></div><h4 id=default-trigger>9.1.1. Default trigger</h4><p>The default trigger for a <code>PCollection</code> is based on event time, and emits the |
| results of the window when the Beam’s watermark passes the end of the window, |
| and then fires each time late data arrives.</p><p>However, if you are using both the default windowing configuration and the |
| default trigger, the default trigger emits exactly once, and late data is |
| discarded. This is because the default windowing configuration has an allowed |
| lateness value of 0. See the Handling Late Data section for information about |
| modifying this behavior.</p><h3 id=processing-time-triggers>9.2. Processing time triggers</h3><p>The <code>AfterProcessingTime</code> trigger operates on <em>processing time</em>. For example, |
| the <span class=language-java><code>AfterProcessingTime.pastFirstElementInPane()</code></span> |
| <span class=language-py><code>AfterProcessingTime</code></span> |
| <span class=language-go><code>trigger.AfterProcessingTime()</code></span> trigger emits a window |
| after a certain amount of processing time has passed since data was received. |
| The processing time is determined by the system clock, rather than the data |
| element’s timestamp.</p><p>The <code>AfterProcessingTime</code> trigger is useful for triggering early results from a |
| window, particularly a window with a large time frame such as a single global |
| window.</p><h3 id=data-driven-triggers>9.3. Data-driven triggers</h3><p>Beam provides one data-driven trigger, |
| <span class=language-java><code>AfterPane.elementCountAtLeast()</code></span> |
| <span class=language-py><code>AfterCount</code></span> |
| <span class=language-go><code>trigger.AfterCount()</code></span>. This trigger works on an element |
| count; it fires after the current pane has collected at least <em>N</em> elements. This |
| allows a window to emit early results (before all the data has accumulated), |
| which can be particularly useful if you are using a single global window.</p><p>It is important to note that if, for example, you specify |
| <span class=language-java><code>.elementCountAtLeast(50)</code></span> |
| <span class=language-py>AfterCount(50)</span> |
| <span class=language-go><code>trigger.AfterCount(50)</code></span> and only 32 elements arrive, |
| those 32 elements sit around forever. If the 32 elements are important to you, |
| consider using <a href=#composite-triggers>composite triggers</a> to combine multiple |
| conditions. This allows you to specify multiple firing conditions such as “fire |
| either when I receive 50 elements, or every 1 second”.</p><h3 id=setting-a-trigger>9.4. Setting a trigger</h3><p>When you set a windowing function for a <code>PCollection</code> by using the |
| <span class=language-java><code>Window</code></span><span class=language-py><code>WindowInto</code></span><span class=language-go><code>beam.WindowInto</code></span> |
| transform, you can also specify a trigger.</p><p class=language-java>You set the trigger(s) for a <code>PCollection</code> by invoking the method |
| <code>.triggering()</code> on the result of your <code>Window.into()</code> transform. This code |
| sample sets a time-based trigger for a <code>PCollection</code>, which emits results one |
| minute after the first element in that window has been processed. The last line |
| in the code sample, <code>.discardingFiredPanes()</code>, sets the window’s <strong>accumulation |
| mode</strong>.</p><p class=language-py>You set the trigger(s) for a <code>PCollection</code> by setting the <code>trigger</code> parameter |
| when you use the <code>WindowInto</code> transform. This code sample sets a time-based |
| trigger for a <code>PCollection</code>, which emits results one minute after the first |
| element in that window has been processed. The <code>accumulation_mode</code> parameter |
| sets the window’s <strong>accumulation mode</strong>.</p><p class=language-go>You set the trigger(s) for a <code>PCollection</code> by passing in the <code>beam.Trigger</code> parameter |
| when you use the <code>beam.WindowInto</code> transform. This code sample sets a time-based |
| trigger for a <code>PCollection</code>, which emits results one minute after the first |
| element in that window has been processed. |
| The <code>beam.AccumulationMode</code> parameter sets the window’s <strong>accumulation mode</strong>.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>pc</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>pc</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Window</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>into</span><span class=o>(</span><span class=n>FixedWindows</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>1</span><span class=o>,</span> <span class=n>TimeUnit</span><span class=o>.</span><span class=na>MINUTES</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>triggering</span><span class=o>(</span><span class=n>AfterProcessingTime</span><span class=o>.</span><span class=na>pastFirstElementInPane</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>plusDelayOf</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>1</span><span class=o>)))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>discardingFiredPanes</span><span class=o>());</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl> <span class=n>pcollection</span> <span class=o>|</span> <span class=n>WindowInto</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>FixedWindows</span><span class=p>(</span><span class=mi>1</span> <span class=o>*</span> <span class=mi>60</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>trigger</span><span class=o>=</span><span class=n>AfterProcessingTime</span><span class=p>(</span><span class=mi>1</span> <span class=o>*</span> <span class=mi>60</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>accumulation_mode</span><span class=o>=</span><span class=n>AccumulationMode</span><span class=o>.</span><span class=n>DISCARDING</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>windowedItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewFixedWindows</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>),</span> <span class=nx>pcollection</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>Trigger</span><span class=p>(</span><span class=nx>trigger</span><span class=p>.</span><span class=nf>AfterProcessingTime</span><span class=p>().</span> |
| </span></span><span class=line><span class=cl> <span class=nf>PlusDelay</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>)),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>AllowedLateness</span><span class=p>(</span><span class=mi>30</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>PanesDiscard</span><span class=p>(),</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span></span></span></code></pre></div></div></div><h4 id=window-accumulation-modes>9.4.1. Window accumulation modes</h4><p>When you specify a trigger, you must also set the window’s <strong>accumulation |
| mode</strong>. When a trigger fires, it emits the current contents of the window as a |
| pane. Since a trigger can fire multiple times, the accumulation mode determines |
| whether the system <em>accumulates</em> the window panes as the trigger fires, or |
| <em>discards</em> them.</p><p class=language-java>To set a window to accumulate the panes that are produced when the trigger |
| fires, invoke<code>.accumulatingFiredPanes()</code> when you set the trigger. To set a |
| window to discard fired panes, invoke <code>.discardingFiredPanes()</code>.</p><p class=language-py>To set a window to accumulate the panes that are produced when the trigger |
| fires, set the <code>accumulation_mode</code> parameter to <code>ACCUMULATING</code> when you set the |
| trigger. To set a window to discard fired panes, set <code>accumulation_mode</code> to |
| <code>DISCARDING</code>.</p><p class=language-go>To set a window to accumulate the panes that are produced when the trigger |
| fires, set the <code>beam.AccumulationMode</code> parameter to <code>beam.PanesAccumulate()</code> when you set the |
| trigger. To set a window to discard fired panes, set <code>beam.AccumulationMode</code> to |
| <code>beam.PanesDiscard()</code>.</p><p>Let’s look an example that uses a <code>PCollection</code> with fixed-time windowing and a |
| data-based trigger. This is something you might do if, for example, each window |
| represented a ten-minute running average, but you wanted to display the current |
| value of the average in a UI more frequently than every ten minutes. We’ll |
| assume the following conditions:</p><ul><li>The <code>PCollection</code> uses 10-minute fixed-time windows.</li><li>The <code>PCollection</code> has a repeating trigger that fires every time 3 elements |
| arrive.</li></ul><p>The following diagram shows data events for key X as they arrive in the |
| PCollection and are assigned to windows. To keep the diagram a bit simpler, |
| we’ll assume that the events all arrive in the pipeline in order.</p><p><img src=/images/trigger-accumulation.png alt="Diagram of data events for accumulating mode example"></p><h5 id=accumulating-mode>9.4.1.1. Accumulating mode</h5><p>If our trigger is set to accumulating mode, the trigger emits the following |
| values each time it fires. Keep in mind that the trigger fires every time three |
| elements arrive:</p><pre tabindex=0><code> First trigger firing: [5, 8, 3] |
| Second trigger firing: [5, 8, 3, 15, 19, 23] |
| Third trigger firing: [5, 8, 3, 15, 19, 23, 9, 13, 10] |
| </code></pre><h5 id=discarding-mode>9.4.1.2. Discarding mode</h5><p>If our trigger is set to discarding mode, the trigger emits the following values |
| on each firing:</p><pre tabindex=0><code> First trigger firing: [5, 8, 3] |
| Second trigger firing: [15, 19, 23] |
| Third trigger firing: [9, 13, 10] |
| </code></pre><h4 id=handling-late-data>9.4.2. Handling late data</h4><p>If you want your pipeline to process data that arrives after the watermark |
| passes the end of the window, you can apply an <em>allowed lateness</em> when you set |
| your windowing configuration. This gives your trigger the opportunity to react |
| to the late data. If allowed lateness is set, the default trigger will emit new |
| results immediately whenever late data arrives.</p><p>You set the allowed lateness by using <span class=language-java><code>.withAllowedLateness()</code></span> |
| <span class=language-py><code>allowed_lateness</code></span> |
| <span class=language-go><code>beam.AllowedLateness()</code></span> |
| when you set your windowing function:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>pc</span> <span class=o>=</span> <span class=o>...;</span> |
| </span></span><span class=line><span class=cl> <span class=n>pc</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Window</span><span class=o>.<</span><span class=n>String</span><span class=o>></span><span class=n>into</span><span class=o>(</span><span class=n>FixedWindows</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>1</span><span class=o>,</span> <span class=n>TimeUnit</span><span class=o>.</span><span class=na>MINUTES</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>triggering</span><span class=o>(</span><span class=n>AfterProcessingTime</span><span class=o>.</span><span class=na>pastFirstElementInPane</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>plusDelayOf</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>1</span><span class=o>)))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withAllowedLateness</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>30</span><span class=o>));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl> <span class=n>pc</span> <span class=o>=</span> <span class=p>[</span><span class=n>Initial</span> <span class=n>PCollection</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=n>pc</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>WindowInto</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>FixedWindows</span><span class=p>(</span><span class=mi>60</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>trigger</span><span class=o>=</span><span class=n>AfterProcessingTime</span><span class=p>(</span><span class=mi>60</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>allowed_lateness</span><span class=o>=</span><span class=mi>1800</span><span class=p>)</span> <span class=c1># 30 minutes</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=o>...</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>allowedToBeLateItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewFixedWindows</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>),</span> <span class=nx>pcollection</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>Trigger</span><span class=p>(</span><span class=nx>trigger</span><span class=p>.</span><span class=nf>AfterProcessingTime</span><span class=p>().</span> |
| </span></span><span class=line><span class=cl> <span class=nf>PlusDelay</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>)),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>AllowedLateness</span><span class=p>(</span><span class=mi>30</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span></span></span></code></pre></div></div></div><p>This allowed lateness propagates to all <code>PCollection</code>s derived as a result of |
| applying transforms to the original <code>PCollection</code>. If you want to change the |
| allowed lateness later in your pipeline, you can apply |
| <span class=language-java><code>Window.configure().withAllowedLateness()</code></span> |
| <span class=language-py><code>allowed_lateness</code></span> |
| <span class=language-go><code>beam.AllowedLateness()</code></span> |
| again, explicitly.</p><h3 id=composite-triggers>9.5. Composite triggers</h3><p>You can combine multiple triggers to form <strong>composite triggers</strong>, and can |
| specify a trigger to emit results repeatedly, at most once, or under other |
| custom conditions.</p><h4 id=composite-trigger-types>9.5.1. Composite trigger types</h4><p>Beam includes the following composite triggers:</p><ul><li>You can add additional early firings or late firings to |
| <code>AfterWatermark.pastEndOfWindow</code> via <code>.withEarlyFirings</code> and |
| <code>.withLateFirings</code>.</li><li><code>Repeatedly.forever</code> specifies a trigger that executes forever. Any time the |
| trigger’s conditions are met, it causes a window to emit results and then |
| resets and starts over. It can be useful to combine <code>Repeatedly.forever</code> |
| with <code>.orFinally</code> to specify a condition that causes the repeating trigger |
| to stop.</li><li><code>AfterEach.inOrder</code> combines multiple triggers to fire in a specific |
| sequence. Each time a trigger in the sequence emits a window, the sequence |
| advances to the next trigger.</li><li><code>AfterFirst</code> takes multiple triggers and emits the first time <em>any</em> of its |
| argument triggers is satisfied. This is equivalent to a logical OR operation |
| for multiple triggers.</li><li><code>AfterAll</code> takes multiple triggers and emits when <em>all</em> of its argument |
| triggers are satisfied. This is equivalent to a logical AND operation for |
| multiple triggers.</li><li><code>orFinally</code> can serve as a final condition to cause any trigger to fire one |
| final time and never fire again.</li></ul><h4 id=composite-afterwatermark>9.5.2. Composition with AfterWatermark</h4><p>Some of the most useful composite triggers fire a single time when Beam |
| estimates that all the data has arrived (i.e. when the watermark passes the end |
| of the window) combined with either, or both, of the following:</p><ul><li><p>Speculative firings that precede the watermark passing the end of the window |
| to allow faster processing of partial results.</p></li><li><p>Late firings that happen after the watermark passes the end of the window, |
| to allow for handling late-arriving data</p></li></ul><p>You can express this pattern using <code>AfterWatermark</code>. For example, the following |
| example trigger code fires on the following conditions:</p><ul><li><p>On Beam’s estimate that all the data has arrived (the watermark passes the |
| end of the window)</p></li><li><p>Any time late data arrives, after a ten-minute delay</p></li></ul><p class=language-java><ul><li>After two days, we assume no more data of interest will arrive, and the |
| trigger stops executing</li></ul></p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Window</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>configure</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>triggering</span><span class=o>(</span><span class=n>AfterWatermark</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>pastEndOfWindow</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withLateFirings</span><span class=o>(</span><span class=n>AfterProcessingTime</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>pastFirstElementInPane</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>plusDelayOf</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>10</span><span class=o>))))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withAllowedLateness</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardDays</span><span class=o>(</span><span class=n>2</span><span class=o>)));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>pcollection</span> <span class=o>|</span> <span class=n>WindowInto</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>FixedWindows</span><span class=p>(</span><span class=mi>1</span> <span class=o>*</span> <span class=mi>60</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>trigger</span><span class=o>=</span><span class=n>AfterWatermark</span><span class=p>(</span><span class=n>late</span><span class=o>=</span><span class=n>AfterProcessingTime</span><span class=p>(</span><span class=mi>10</span> <span class=o>*</span> <span class=mi>60</span><span class=p>)),</span> |
| </span></span><span class=line><span class=cl> <span class=n>allowed_lateness</span><span class=o>=</span><span class=mi>10</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>accumulation_mode</span><span class=o>=</span><span class=n>AccumulationMode</span><span class=o>.</span><span class=n>DISCARDING</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>compositeTriggerItems</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>window</span><span class=p>.</span><span class=nf>NewFixedWindows</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>),</span> <span class=nx>pcollection</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>Trigger</span><span class=p>(</span><span class=nx>trigger</span><span class=p>.</span><span class=nf>AfterEndOfWindow</span><span class=p>().</span> |
| </span></span><span class=line><span class=cl> <span class=nf>LateFiring</span><span class=p>(</span><span class=nx>trigger</span><span class=p>.</span><span class=nf>AfterProcessingTime</span><span class=p>().</span> |
| </span></span><span class=line><span class=cl> <span class=nf>PlusDelay</span><span class=p>(</span><span class=mi>10</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>))),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>AllowedLateness</span><span class=p>(</span><span class=mi>2</span><span class=o>*</span><span class=mi>24</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Hour</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span></span></span></code></pre></div></div></div><h4 id=other-composite-triggers>9.5.3. Other composite triggers</h4><p>You can also build other sorts of composite triggers. The following example code |
| shows a simple composite trigger that fires whenever the pane has at least 100 |
| elements, or after a minute.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=n>Repeatedly</span><span class=o>.</span><span class=na>forever</span><span class=o>(</span><span class=n>AfterFirst</span><span class=o>.</span><span class=na>of</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>AfterPane</span><span class=o>.</span><span class=na>elementCountAtLeast</span><span class=o>(</span><span class=n>100</span><span class=o>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>AfterProcessingTime</span><span class=o>.</span><span class=na>pastFirstElementInPane</span><span class=o>().</span><span class=na>plusDelayOf</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>1</span><span class=o>))))</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>pcollection</span> <span class=o>|</span> <span class=n>WindowInto</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>FixedWindows</span><span class=p>(</span><span class=mi>1</span> <span class=o>*</span> <span class=mi>60</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>trigger</span><span class=o>=</span><span class=n>Repeatedly</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>AfterAny</span><span class=p>(</span><span class=n>AfterCount</span><span class=p>(</span><span class=mi>100</span><span class=p>),</span> <span class=n>AfterProcessingTime</span><span class=p>(</span><span class=mi>1</span> <span class=o>*</span> <span class=mi>60</span><span class=p>))),</span> |
| </span></span><span class=line><span class=cl> <span class=n>accumulation_mode</span><span class=o>=</span><span class=n>AccumulationMode</span><span class=o>.</span><span class=n>DISCARDING</span><span class=p>)</span></span></span></code></pre></div></div></div><h2 id=metrics>10. Metrics</h2><p>In the Beam model, metrics provide some insight into the current state of a user pipeline, |
| potentially while the pipeline is running. There could be different reasons for that, for instance:</p><ul><li>Check the number of errors encountered while running a specific step in the pipeline;</li><li>Monitor the number of RPCs made to backend service;</li><li>Retrieve an accurate count of the number of elements that have been processed;</li><li>…and so on.</li></ul><h3 id=101-the-main-concepts-of-beam-metrics>10.1. The main concepts of Beam metrics</h3><ul><li><strong>Named</strong>. Each metric has a name which consists of a namespace and an actual name. The |
| namespace can be used to differentiate between multiple metrics with the same name and also |
| allows querying for all metrics within a specific namespace.</li><li><strong>Scoped</strong>. Each metric is reported against a specific step in the pipeline, indicating what |
| code was running when the metric was incremented.</li><li><strong>Dynamically Created</strong>. Metrics may be created during runtime without pre-declaring them, in |
| much the same way a logger could be created. This makes it easier to produce metrics in utility |
| code and have them usefully reported.</li><li><strong>Degrade Gracefully</strong>. If a runner doesn’t support some part of reporting metrics, the |
| fallback behavior is to drop the metric updates rather than failing the pipeline. If a runner |
| doesn’t support some part of querying metrics, the runner will not return the associated data.</li></ul><p>Reported metrics are implicitly scoped to the transform within the pipeline that reported them. |
| This allows reporting the same metric name in multiple places and identifying the value each |
| transform reported, as well as aggregating the metric across the entire pipeline.</p><blockquote><p><strong>Note:</strong> It is runner-dependent whether metrics are accessible during pipeline execution or only |
| after jobs have completed.</p></blockquote><h3 id=types-of-metrics>10.2. Types of metrics</h3><p>There are three types of metrics that are supported for the moment: <code>Counter</code>, <code>Distribution</code> and |
| <code>Gauge</code>.</p><p class=language-go>In the Beam SDK for Go, a <code>context.Context</code> provided by the framework must be passed to the metric |
| or the metric value will not be recorded. The framework will automatically provide a valid |
| <code>context.Context</code> to <code>ProcessElement</code> and similar methods when it’s the first parameter.</p><p><strong>Counter</strong>: A metric that reports a single long value and can be incremented or decremented.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>Counter</span> <span class=n>counter</span> <span class=o>=</span> <span class=n>Metrics</span><span class=o>.</span><span class=na>counter</span><span class=o>(</span> <span class=s>"namespace"</span><span class=o>,</span> <span class=s>"counter1"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=n>ProcessContext</span> <span class=n>context</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// count the elements |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>counter</span><span class=o>.</span><span class=na>inc</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>var</span> <span class=nx>counter</span> <span class=p>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>NewCounter</span><span class=p>(</span><span class=s>"namespace"</span><span class=p>,</span> <span class=s>"counter1"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>MyDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>ctx</span> <span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=o>...</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// count the elements |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>counter</span><span class=p>.</span><span class=nf>Inc</span><span class=p>(</span><span class=nx>ctx</span><span class=p>,</span> <span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p><strong>Distribution</strong>: A metric that reports information about the distribution of reported values.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>Distribution</span> <span class=n>distribution</span> <span class=o>=</span> <span class=n>Metrics</span><span class=o>.</span><span class=na>distribution</span><span class=o>(</span> <span class=s>"namespace"</span><span class=o>,</span> <span class=s>"distribution1"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=n>ProcessContext</span> <span class=n>context</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>Integer</span> <span class=n>element</span> <span class=o>=</span> <span class=n>context</span><span class=o>.</span><span class=na>element</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// create a distribution (histogram) of the values |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>distribution</span><span class=o>.</span><span class=na>update</span><span class=o>(</span><span class=n>element</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>var</span> <span class=nx>distribution</span> <span class=p>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>NewDistribution</span><span class=p>(</span><span class=s>"namespace"</span><span class=p>,</span> <span class=s>"distribution1"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>MyDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>ctx</span> <span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>v</span> <span class=kt>int64</span><span class=p>,</span> <span class=o>...</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// create a distribution (histogram) of the values |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>distribution</span><span class=p>.</span><span class=nf>Update</span><span class=p>(</span><span class=nx>ctx</span><span class=p>,</span> <span class=nx>v</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p><strong>Gauge</strong>: A metric that reports the latest value out of reported values. Since metrics are |
| collected from many workers the value may not be the absolute last, but one of the latest values.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>Gauge</span> <span class=n>gauge</span> <span class=o>=</span> <span class=n>Metrics</span><span class=o>.</span><span class=na>gauge</span><span class=o>(</span> <span class=s>"namespace"</span><span class=o>,</span> <span class=s>"gauge1"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=n>ProcessContext</span> <span class=n>context</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>Integer</span> <span class=n>element</span> <span class=o>=</span> <span class=n>context</span><span class=o>.</span><span class=na>element</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// create a gauge (latest value received) of the values |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>gauge</span><span class=o>.</span><span class=na>set</span><span class=o>(</span><span class=n>element</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>var</span> <span class=nx>gauge</span> <span class=p>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>NewGauge</span><span class=p>(</span><span class=s>"namespace"</span><span class=p>,</span> <span class=s>"gauge1"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>MyDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>ctx</span> <span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>v</span> <span class=kt>int64</span><span class=p>,</span> <span class=o>...</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// create a gauge (latest value received) of the values |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>gauge</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>ctx</span><span class=p>,</span> <span class=nx>v</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=querying-metrics>10.3. Querying metrics</h3><p class="language-java language-python"><code>PipelineResult</code> has a method <code>metrics()</code> which returns a <code>MetricResults</code> object that allows |
| accessing metrics. The main method available in <code>MetricResults</code> allows querying for all metrics |
| matching a given filter.</p><p class=language-go><code>beam.PipelineResult</code> has a method <code>Metrics()</code> which returns a <code>metrics.Results</code> object that allows |
| accessing metrics. The main method available in <code>metrics.Results</code> allows querying for all metrics |
| matching a given filter. It takes in a predicate with a <code>SingleResult</code> parameter type, which can |
| be used for custom filters.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>interface</span> <span class=nc>PipelineResult</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>MetricResults</span> <span class=nf>metrics</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>abstract</span> <span class=kd>class</span> <span class=nc>MetricResults</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>abstract</span> <span class=n>MetricQueryResults</span> <span class=nf>queryMetrics</span><span class=o>(</span><span class=nd>@Nullable</span> <span class=n>MetricsFilter</span> <span class=n>filter</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>interface</span> <span class=nc>MetricQueryResults</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>Iterable</span><span class=o><</span><span class=n>MetricResult</span><span class=o><</span><span class=n>Long</span><span class=o>>></span> <span class=nf>getCounters</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Iterable</span><span class=o><</span><span class=n>MetricResult</span><span class=o><</span><span class=n>DistributionResult</span><span class=o>>></span> <span class=nf>getDistributions</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Iterable</span><span class=o><</span><span class=n>MetricResult</span><span class=o><</span><span class=n>GaugeResult</span><span class=o>>></span> <span class=nf>getGauges</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>interface</span> <span class=nc>MetricResult</span><span class=o><</span><span class=n>T</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>MetricName</span> <span class=nf>getName</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>String</span> <span class=nf>getStep</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>T</span> <span class=nf>getCommitted</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>T</span> <span class=nf>getAttempted</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>queryMetrics</span><span class=p>(</span><span class=nx>pr</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PipelineResult</span><span class=p>,</span> <span class=nx>ns</span><span class=p>,</span> <span class=nx>n</span> <span class=kt>string</span><span class=p>)</span> <span class=nx>metrics</span><span class=p>.</span><span class=nx>QueryResults</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>pr</span><span class=p>.</span><span class=nf>Metrics</span><span class=p>().</span><span class=nf>Query</span><span class=p>(</span><span class=kd>func</span><span class=p>(</span><span class=nx>r</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>MetricResult</span><span class=p>)</span> <span class=kt>bool</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>r</span><span class=p>.</span><span class=nf>Namespace</span><span class=p>()</span> <span class=o>==</span> <span class=nx>ns</span> <span class=o>&&</span> <span class=nx>r</span><span class=p>.</span><span class=nf>Name</span><span class=p>()</span> <span class=o>==</span> <span class=nx>n</span> |
| </span></span><span class=line><span class=cl> <span class=p>})</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=using-metrics>10.4. Using metrics in pipeline</h3><p>Below, there is a simple example of how to use a <code>Counter</code> metric in a user pipeline.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// creating a pipeline with custom metrics DoFn |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>pipeline</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>apply</span><span class=o>(...)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>MyMetricsDoFn</span><span class=o>()));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>pipelineResult</span> <span class=o>=</span> <span class=n>pipeline</span><span class=o>.</span><span class=na>run</span><span class=o>().</span><span class=na>waitUntilFinish</span><span class=o>(...);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// request the metric called "counter1" in namespace called "namespace" |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>MetricQueryResults</span> <span class=n>metrics</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>pipelineResult</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>metrics</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>queryMetrics</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>MetricsFilter</span><span class=o>.</span><span class=na>builder</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>addNameFilter</span><span class=o>(</span><span class=n>MetricNameFilter</span><span class=o>.</span><span class=na>named</span><span class=o>(</span><span class=s>"namespace"</span><span class=o>,</span> <span class=s>"counter1"</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>build</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// print the metric value - there should be only one line because there is only one metric |
| </span></span></span><span class=line><span class=cl><span class=c1>// called "counter1" in the namespace called "namespace" |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=k>for</span> <span class=o>(</span><span class=n>MetricResult</span><span class=o><</span><span class=n>Long</span><span class=o>></span> <span class=n>counter</span><span class=o>:</span> <span class=n>metrics</span><span class=o>.</span><span class=na>getCounters</span><span class=o>())</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>System</span><span class=o>.</span><span class=na>out</span><span class=o>.</span><span class=na>println</span><span class=o>(</span><span class=n>counter</span><span class=o>.</span><span class=na>getName</span><span class=o>()</span> <span class=o>+</span> <span class=s>":"</span> <span class=o>+</span> <span class=n>counter</span><span class=o>.</span><span class=na>getAttempted</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>MyMetricsDoFn</span> <span class=kd>extends</span> <span class=n>DoFn</span><span class=o><</span><span class=n>Integer</span><span class=o>,</span> <span class=n>Integer</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>private</span> <span class=kd>final</span> <span class=n>Counter</span> <span class=n>counter</span> <span class=o>=</span> <span class=n>Metrics</span><span class=o>.</span><span class=na>counter</span><span class=o>(</span> <span class=s>"namespace"</span><span class=o>,</span> <span class=s>"counter1"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=n>ProcessContext</span> <span class=n>context</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// count the elements |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>counter</span><span class=o>.</span><span class=na>inc</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>context</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>context</span><span class=o>.</span><span class=na>element</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=nf>addMetricDoFnToPipeline</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>input</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>MyMetricsDoFn</span><span class=p>{},</span> <span class=nx>input</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>executePipelineAndGetMetrics</span><span class=p>(</span><span class=nx>ctx</span> <span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>p</span> <span class=o>*</span><span class=nx>beam</span><span class=p>.</span><span class=nx>Pipeline</span><span class=p>)</span> <span class=p>(</span><span class=nx>metrics</span><span class=p>.</span><span class=nx>QueryResults</span><span class=p>,</span> <span class=kt>error</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>pr</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>Run</span><span class=p>(</span><span class=nx>ctx</span><span class=p>,</span> <span class=nx>runner</span><span class=p>,</span> <span class=nx>p</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>metrics</span><span class=p>.</span><span class=nx>QueryResults</span><span class=p>{},</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Request the metric called "counter1" in namespace called "namespace" |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>ms</span> <span class=o>:=</span> <span class=nx>pr</span><span class=p>.</span><span class=nf>Metrics</span><span class=p>().</span><span class=nf>Query</span><span class=p>(</span><span class=kd>func</span><span class=p>(</span><span class=nx>r</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>MetricResult</span><span class=p>)</span> <span class=kt>bool</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>r</span><span class=p>.</span><span class=nf>Namespace</span><span class=p>()</span> <span class=o>==</span> <span class=s>"namespace"</span> <span class=o>&&</span> <span class=nx>r</span><span class=p>.</span><span class=nf>Name</span><span class=p>()</span> <span class=o>==</span> <span class=s>"counter1"</span> |
| </span></span><span class=line><span class=cl> <span class=p>})</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Print the metric value - there should be only one line because there is |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// only one metric called "counter1" in the namespace called "namespace" |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>for</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>c</span> <span class=o>:=</span> <span class=k>range</span> <span class=nx>ms</span><span class=p>.</span><span class=nf>Counters</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fmt</span><span class=p>.</span><span class=nf>Println</span><span class=p>(</span><span class=nx>c</span><span class=p>.</span><span class=nf>Namespace</span><span class=p>(),</span> <span class=s>"-"</span><span class=p>,</span> <span class=nx>c</span><span class=p>.</span><span class=nf>Name</span><span class=p>(),</span> <span class=s>":"</span><span class=p>,</span> <span class=nx>c</span><span class=p>.</span><span class=nx>Committed</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>ms</span><span class=p>,</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>MyMetricsDoFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>counter</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Counter</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>RegisterType</span><span class=p>(</span><span class=nx>reflect</span><span class=p>.</span><span class=nf>TypeOf</span><span class=p>((</span><span class=o>*</span><span class=nx>MyMetricsDoFn</span><span class=p>)(</span><span class=kc>nil</span><span class=p>)))</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>MyMetricsDoFn</span><span class=p>)</span> <span class=nf>Setup</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// While metrics can be defined in package scope or dynamically |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// it's most efficient to include them in the DoFn. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>counter</span> <span class=p>=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>NewCounter</span><span class=p>(</span><span class=s>"namespace"</span><span class=p>,</span> <span class=s>"counter1"</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>MyMetricsDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>ctx</span> <span class=nx>context</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>v</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>V</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>V</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// count the elements |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>counter</span><span class=p>.</span><span class=nf>Inc</span><span class=p>(</span><span class=nx>ctx</span><span class=p>,</span> <span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nx>v</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=export-metrics>10.5. Export metrics</h3><p>Beam metrics can be exported to external sinks. If a metrics sink is set up in the configuration, the runner will push metrics to it at a default 5s period. |
| The configuration is held in the <a href=https://beam.apache.org/releases/javadoc/2.19.0/org/apache/beam/sdk/metrics/MetricsOptions.html>MetricsOptions</a> class. |
| It contains push period configuration and also sink specific options such as type and URL. As for now only the REST HTTP and the Graphite sinks are supported and only |
| Flink and Spark runners support metrics export.</p><p>Also Beam metrics are exported to inner Spark and Flink dashboards to be consulted in their respective UI.</p><h2 id=state-and-timers>11. State and Timers</h2><p>Beam’s windowing and triggering facilities provide a powerful abstraction for grouping and aggregating unbounded input |
| data based on timestamps. However there are aggregation use cases for which developers may require a higher degree of |
| control than provided by windows and triggers. Beam provides an API for manually managing per-key state, allowing for |
| fine-grained control over aggregations.</p><p>Beam’s state API models state per key. To use the state API, you start out with a keyed <code>PCollection</code>, which in Java |
| is modeled as a <code>PCollection<KV<K, V>></code>. A <code>ParDo</code> processing this <code>PCollection</code> can now declare state variables. Inside |
| the <code>ParDo</code> these state variables can be used to write or update state for the current key or to read previous state |
| written for that key. State is always fully scoped only to the current processing key.</p><p>Windowing can still be used together with stateful processing. All state for a key is scoped to the current window. This |
| means that the first time a key is seen for a given window any state reads will return empty, and that a runner can |
| garbage collect state when a window is completed. It’s also often useful to use Beam’s windowed aggregations prior to |
| the stateful operator. For example, using a combiner to preaggregate data, and then storing aggregated data inside of |
| state. Merging windows are not currently supported when using state and timers.</p><p>Sometimes stateful processing is used to implement state-machine style processing inside a <code>DoFn</code>. When doing this, |
| care must be taken to remember that the elements in input PCollection have no guaranteed order and to ensure that the |
| program logic is resilient to this. Unit tests written using the DirectRunner will shuffle the order of element |
| processing, and are recommended to test for correctness.</p><p class=language-java>In Java, DoFn declares states to be accessed by creating final <code>StateSpec</code> member variables representing each state. Each |
| state must be named using the <code>StateId</code> annotation; this name is unique to a ParDo in the graph and has no relation |
| to other nodes in the graph. A <code>DoFn</code> can declare multiple state variables.</p><p class=language-py>In Python, DoFn declares states to be accessed by creating <code>StateSpec</code> class member variables representing each state. Each |
| <code>StateSpec</code> is initialized with a name, this name is unique to a ParDo in the graph and has no relation |
| to other nodes in the graph. A <code>DoFn</code> can declare multiple state variables.</p><p class=language-go>In Go, DoFn declares states to be accessed by creating state struct member variables representing each state. Each |
| state variable is initialized with a key, this key is unique to a ParDo in the graph and has no relation |
| to other nodes in the graph. If no name is supplied, the key defaults to the member variable’s name. |
| A <code>DoFn</code> can declare multiple state variables.</p><span class=language-typescript><blockquote><p><strong>Note:</strong> The Beam SDK for Typescript does not yet support a State and Timer API, |
| but it is possible to use these features from cross-language pipelines (see below).</p></blockquote></span><h3 id=types-of-state>11.1. Types of state</h3><p>Beam provides several types of state:</p><h4 id=valuestate>ValueState</h4><p>A ValueState is a scalar state value. For each key in the input, a ValueState will store a typed value that can be |
| read and modified inside the DoFn’s <code>@ProcessElement</code> or <code>@OnTimer</code> methods. If the type of the ValueState has a coder |
| registered, then Beam will automatically infer the coder for the state value. Otherwise, a coder can be explicitly |
| specified when creating the ValueState. For example, the following ParDo creates a single state variable that |
| accumulates the number of elements seen.</p><p>Note: <code>ValueState</code> is called <code>ReadModifyWriteState</code> in the Python SDK.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>>></span> <span class=n>numElements</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>state</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Read the number element seen so far for this user key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// state.read() returns null if it was never set. The below code allows us to have a default value of 0. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kt>int</span> <span class=n>currentValue</span> <span class=o>=</span> <span class=n>MoreObjects</span><span class=o>.</span><span class=na>firstNonNull</span><span class=o>(</span><span class=n>state</span><span class=o>.</span><span class=na>read</span><span class=o>(),</span> <span class=n>0</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Update the state. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>state</span><span class=o>.</span><span class=na>write</span><span class=o>(</span><span class=n>currentValue</span> <span class=o>+</span> <span class=n>1</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// valueStateFn keeps track of the number of elements seen. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>valueStateFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Val</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=kt>int</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>s</span> <span class=o>*</span><span class=nx>valueStateFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>p</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>book</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Get the value stored in our state |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>val</span><span class=p>,</span> <span class=nx>ok</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>s</span><span class=p>.</span><span class=nx>Val</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>p</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>!</span><span class=nx>ok</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>s</span><span class=p>.</span><span class=nx>Val</span><span class=p>.</span><span class=nf>Write</span><span class=p>(</span><span class=nx>p</span><span class=p>,</span> <span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> <span class=k>else</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>s</span><span class=p>.</span><span class=nx>Val</span><span class=p>.</span><span class=nf>Write</span><span class=p>(</span><span class=nx>p</span><span class=p>,</span> <span class=nx>val</span><span class=o>+</span><span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>val</span> <span class=p>></span> <span class=mi>10000</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Example of clearing and starting again with an empty bag |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span><span class=p>.</span><span class=nx>Val</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>p</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p>Beam also allows explicitly specifying a coder for <code>ValueState</code> values. For example:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>MyType</span><span class=o>>></span> <span class=n>numElements</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>(</span><span class=k>new</span> <span class=n>MyTypeCoder</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>ReadModifyWriteStateDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>STATE_SPEC</span> <span class=o>=</span> <span class=n>ReadModifyWriteStateSpec</span><span class=p>(</span><span class=s1>'num_elements'</span><span class=p>,</span> <span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>state</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>STATE_SPEC</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Read the number element seen so far for this user key.</span> |
| </span></span><span class=line><span class=cl> <span class=n>current_value</span> <span class=o>=</span> <span class=n>state</span><span class=o>.</span><span class=n>read</span><span class=p>()</span> <span class=ow>or</span> <span class=mi>0</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=n>write</span><span class=p>(</span><span class=n>current_value</span><span class=o>+</span><span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'state pardo'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>ReadModifyWriteStateDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>valueStateDoFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Val</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=nx>MyCustomType</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>encode</span><span class=p>(</span><span class=nx>m</span> <span class=nx>MyCustomType</span><span class=p>)</span> <span class=p>[]</span><span class=kt>byte</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>m</span><span class=p>.</span><span class=nf>Bytes</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>decode</span><span class=p>(</span><span class=nx>b</span> <span class=p>[]</span><span class=kt>byte</span><span class=p>)</span> <span class=nx>MyCustomType</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>MyCustomType</span><span class=p>{}.</span><span class=nf>FromBytes</span><span class=p>(</span><span class=nx>b</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>init</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nf>RegisterCoder</span><span class=p>(</span><span class=nx>reflect</span><span class=p>.</span><span class=nf>TypeOf</span><span class=p>((</span><span class=o>*</span><span class=nx>MyCustomType</span><span class=p>)(</span><span class=kc>nil</span><span class=p>)).</span><span class=nf>Elem</span><span class=p>(),</span> <span class=nx>encode</span><span class=p>,</span> <span class=nx>decode</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><div class='language-typescript snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-typescript data-lang=typescript><span class=line><span class=cl><span class=kr>const</span> <span class=nx>pcoll</span> <span class=o>=</span> <span class=nx>root</span><span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>beam</span><span class=p>.</span><span class=nx>create</span><span class=p>([</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>key</span><span class=o>:</span> <span class=s2>"a"</span><span class=p>,</span> <span class=nx>value</span>: <span class=kt>1</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>key</span><span class=o>:</span> <span class=s2>"b"</span><span class=p>,</span> <span class=nx>value</span>: <span class=kt>10</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>{</span> <span class=nx>key</span><span class=o>:</span> <span class=s2>"a"</span><span class=p>,</span> <span class=nx>value</span>: <span class=kt>100</span> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=p>])</span> |
| </span></span><span class=line><span class=cl><span class=p>);</span> |
| </span></span><span class=line><span class=cl><span class=kr>const</span> <span class=nx>result</span>: <span class=kt>PCollection</span><span class=p><</span><span class=nt>number</span><span class=p>></span> <span class=o>=</span> <span class=k>await</span> <span class=nx>pcoll</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>apply</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>withCoderInternal</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=k>new</span> <span class=nx>KVCoder</span><span class=p>(</span><span class=k>new</span> <span class=nx>StrUtf8Coder</span><span class=p>(),</span> <span class=k>new</span> <span class=nx>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>.</span><span class=nx>applyAsync</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>pythonTransform</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Construct a new Transform from source. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=s2>"__constructor__"</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>[</span> |
| </span></span><span class=line><span class=cl> <span class=nx>pythonCallable</span><span class=p>(</span><span class=sb>` |
| </span></span></span><span class=line><span class=cl><span class=sb> # Define a DoFn to be used below. |
| </span></span></span><span class=line><span class=cl><span class=sb> class ReadModifyWriteStateDoFn(beam.DoFn): |
| </span></span></span><span class=line><span class=cl><span class=sb> STATE_SPEC = beam.transforms.userstate.ReadModifyWriteStateSpec( |
| </span></span></span><span class=line><span class=cl><span class=sb> 'num_elements', beam.coders.VarIntCoder()) |
| </span></span></span><span class=line><span class=cl><span class=sb> |
| </span></span></span><span class=line><span class=cl><span class=sb> def process(self, element, state=beam.DoFn.StateParam(STATE_SPEC)): |
| </span></span></span><span class=line><span class=cl><span class=sb> current_value = state.read() or 0 |
| </span></span></span><span class=line><span class=cl><span class=sb> state.write(current_value + 1) |
| </span></span></span><span class=line><span class=cl><span class=sb> yield current_value + 1 |
| </span></span></span><span class=line><span class=cl><span class=sb> |
| </span></span></span><span class=line><span class=cl><span class=sb> class MyPythonTransform(beam.PTransform): |
| </span></span></span><span class=line><span class=cl><span class=sb> def expand(self, pcoll): |
| </span></span></span><span class=line><span class=cl><span class=sb> return pcoll | beam.ParDo(ReadModifyWriteStateDoFn()) |
| </span></span></span><span class=line><span class=cl><span class=sb> `</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Keyword arguments to pass to the transform, if any. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>{},</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output type if it cannot be inferred |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>{</span> <span class=nx>requestedOutputCoders</span><span class=o>:</span> <span class=p>{</span> <span class=nx>output</span>: <span class=kt>new</span> <span class=nx>VarIntCoder</span><span class=p>()</span> <span class=p>}</span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>);</span></span></span></code></pre></div></div></div><h4 id=combiningstate>CombiningState</h4><p><code>CombiningState</code> allows you to create a state object that is updated using a Beam combiner. For example, the previous |
| <code>ValueState</code> example could be rewritten to use <code>CombiningState</code></p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>CombiningState</span><span class=o><</span><span class=n>Integer</span><span class=o>,</span> <span class=kt>int</span><span class=o>[],</span> <span class=n>Integer</span><span class=o>>></span> <span class=n>numElements</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>combining</span><span class=o>(</span><span class=n>Sum</span><span class=o>.</span><span class=na>ofIntegers</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>state</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>1</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>CombiningStateDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>SUM_TOTAL</span> <span class=o>=</span> <span class=n>CombiningValueStateSpec</span><span class=p>(</span><span class=s1>'total'</span><span class=p>,</span> <span class=nb>sum</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>state</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>SUM_TOTAL</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'Combine state pardo'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>CombiningStateDofn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// combiningStateFn keeps track of the number of elements seen. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>combiningStateFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// types are the types of the accumulator, input, and output respectively |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>Val</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Combining</span><span class=p>[</span><span class=kt>int</span><span class=p>,</span> <span class=kt>int</span><span class=p>,</span> <span class=kt>int</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>s</span> <span class=o>*</span><span class=nx>combiningStateFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>p</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>book</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Get the value stored in our state |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>val</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>s</span><span class=p>.</span><span class=nx>Val</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>p</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>s</span><span class=p>.</span><span class=nx>Val</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>p</span><span class=p>,</span> <span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>val</span> <span class=p>></span> <span class=mi>10000</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Example of clearing and starting again with an empty bag |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span><span class=p>.</span><span class=nx>Val</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>p</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>combineState</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>input</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// CombineFn param can be a simple fn like this or a structural CombineFn |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>cFn</span> <span class=o>:=</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeCombiningState</span><span class=p>[</span><span class=kt>int</span><span class=p>,</span> <span class=kt>int</span><span class=p>,</span> <span class=kt>int</span><span class=p>](</span><span class=s>"stateKey"</span><span class=p>,</span> <span class=kd>func</span><span class=p>(</span><span class=nx>a</span><span class=p>,</span> <span class=nx>b</span> <span class=kt>int</span><span class=p>)</span> <span class=kt>int</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>a</span> <span class=o>+</span> <span class=nx>b</span> |
| </span></span><span class=line><span class=cl> <span class=p>})</span> |
| </span></span><span class=line><span class=cl> <span class=nx>combined</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>combiningStateFn</span><span class=p>{</span><span class=nx>Val</span><span class=p>:</span> <span class=nx>cFn</span><span class=p>},</span> <span class=nx>input</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span></code></pre></div></div></div><h4 id=bagstate>BagState</h4><p>A common use case for state is to accumulate multiple elements. <code>BagState</code> allows for accumulating an unordered set |
| of elements. This allows for addition of elements to the collection without requiring the reading of the entire |
| collection first, which is an efficiency gain. In addition, runners that support paged reads can allow individual |
| bags larger than available memory.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>>></span> <span class=n>numElements</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>bag</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>state</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Add the current element to the bag for this key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>state</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>element</span><span class=o>.</span><span class=na>getValue</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(</span><span class=n>shouldFetch</span><span class=o>())</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Occasionally we fetch and process the values. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>Iterable</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>values</span> <span class=o>=</span> <span class=n>state</span><span class=o>.</span><span class=na>read</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>processValues</span><span class=o>(</span><span class=n>values</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> <span class=c1>// Clear the state for this key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>BagStateDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>ALL_ELEMENTS</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer'</span><span class=p>,</span> <span class=n>coders</span><span class=o>.</span><span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element_pair</span><span class=p>,</span> <span class=n>state</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>element_pair</span><span class=p>[</span><span class=mi>1</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>should_fetch</span><span class=p>():</span> |
| </span></span><span class=line><span class=cl> <span class=n>all_elements</span> <span class=o>=</span> <span class=nb>list</span><span class=p>(</span><span class=n>state</span><span class=o>.</span><span class=n>read</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>process_values</span><span class=p>(</span><span class=n>all_elements</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'Bag state pardo'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>BagStateDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// bagStateFn only emits words that haven't been seen |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>bagStateFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Bag</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Bag</span><span class=p>[</span><span class=kt>string</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>s</span> <span class=o>*</span><span class=nx>bagStateFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>p</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>book</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Get all values we've written to this bag state in this window. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>vals</span><span class=p>,</span> <span class=nx>ok</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>s</span><span class=p>.</span><span class=nx>Bag</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>p</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>!</span><span class=nx>ok</span> <span class=o>||</span> <span class=p>!</span><span class=nf>contains</span><span class=p>(</span><span class=nx>vals</span><span class=p>,</span> <span class=nx>word</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emitWords</span><span class=p>(</span><span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>s</span><span class=p>.</span><span class=nx>Bag</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>p</span><span class=p>,</span> <span class=nx>word</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nb>len</span><span class=p>(</span><span class=nx>vals</span><span class=p>)</span> <span class=p>></span> <span class=mi>10000</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Example of clearing and starting again with an empty bag |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>s</span><span class=p>.</span><span class=nx>Bag</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>p</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=deferred-state-reads>11.2. Deferred state reads</h3><p>When a <code>DoFn</code> contains multiple state specifications, reading each one in order can be slow. Calling the <code>read()</code> function |
| on a state can cause the runner to perform a blocking read. Performing multiple blocking reads in sequence adds latency |
| to element processing. If you know that a state will always be read, you can annotate it as @AlwaysFetched, and then the |
| runner can prefetch all of the states necessary. For example:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state1"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>>></span> <span class=n>state1</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state2"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>String</span><span class=o>>></span> <span class=n>state2</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state3"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>>></span> <span class=n>state3</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>bag</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state1"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>state1</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state2"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>state2</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state3"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>state3</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>state1</span><span class=o>.</span><span class=na>read</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>state2</span><span class=o>.</span><span class=na>read</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>state3</span><span class=o>.</span><span class=na>read</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>This</span> <span class=ow>is</span> <span class=ow>not</span> <span class=n>supported</span> <span class=n>yet</span><span class=p>,</span> <span class=n>see</span> <span class=n>https</span><span class=p>:</span><span class=o>//</span><span class=n>github</span><span class=o>.</span><span class=n>com</span><span class=o>/</span><span class=n>apache</span><span class=o>/</span><span class=n>beam</span><span class=o>/</span><span class=n>issues</span><span class=o>/</span><span class=mf>20739.</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>This</span> <span class=nx>is</span> <span class=nx>not</span> <span class=nx>supported</span> <span class=nx>yet</span><span class=p>,</span> <span class=nx>see</span> <span class=nx>https</span><span class=p>:</span><span class=c1>//github.com/apache/beam/issues/22964. |
| </span></span></span></code></pre></div></div></div><p>If however there are code paths in which the states are not fetched, then annotating with @AlwaysFetched will add |
| unnecessary fetching for those paths. In this case, the readLater method allows the runner to know that the state will |
| be read in the future, allowing multiple state reads to be batched together.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state1"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>>></span> <span class=n>state1</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state2"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>String</span><span class=o>>></span> <span class=n>state2</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state3"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>>></span> <span class=n>state3</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>bag</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state1"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>state1</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state2"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>state2</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state3"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>state3</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(</span><span class=cm>/* should read state */</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>state1</span><span class=o>.</span><span class=na>readLater</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>state2</span><span class=o>.</span><span class=na>readLater</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>state3</span><span class=o>.</span><span class=na>readLater</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// The runner can now batch all three states into a single read, reducing latency. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>processState1</span><span class=o>(</span><span class=n>state1</span><span class=o>.</span><span class=na>read</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=n>processState2</span><span class=o>(</span><span class=n>state2</span><span class=o>.</span><span class=na>read</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=n>processState3</span><span class=o>(</span><span class=n>state3</span><span class=o>.</span><span class=na>read</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><h3 id=timers>11.3. Timers</h3><p>Beam provides a per-key timer callback API. This allows for delayed processing of data stored using the state API. |
| Timers can be set to callback at either an event-time or a processing-time timestamp. Every timer is identified with a |
| TimerId. A given timer for a key can only be set for a single timestamp. Calling set on a timer overwrites the previous |
| firing time for that key’s timer.</p><h4 id=event-time-timers>11.3.1. Event-time timers</h4><p>Event-time timers fire when the input watermark for the DoFn passes the time at which the timer is set, meaning that |
| the runner believes that there are no more elements to be processed with timestamps before the timer timestamp. This |
| allows for event-time aggregations.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>>></span> <span class=n>state</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"timer"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>timer</span> <span class=o>=</span> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timer</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>EVENT_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Timestamp</span> <span class=n>Instant</span> <span class=n>elementTs</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>state</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"timer"</span><span class=o>)</span> <span class=n>Timer</span> <span class=n>timer</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Set an event-time timer to the element timestamp. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>timer</span><span class=o>.</span><span class=na>set</span><span class=o>(</span><span class=n>elementTs</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimer</span><span class=o>(</span><span class=s>"timer"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>//Process timer. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>EventTimerDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>ALL_ELEMENTS</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer'</span><span class=p>,</span> <span class=n>coders</span><span class=o>.</span><span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>TIMER</span> <span class=o>=</span> <span class=n>TimerSpec</span><span class=p>(</span><span class=s1>'timer'</span><span class=p>,</span> <span class=n>TimeDomain</span><span class=o>.</span><span class=n>WATERMARK</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element_pair</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>t</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>TimestampParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>TimerParam</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>element_pair</span><span class=p>[</span><span class=mi>1</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Set an event-time timer to the element timestamp.</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>.</span><span class=n>set</span><span class=p>(</span><span class=n>t</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@on_timer</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>expiry_callback</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>buffer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'EventTime timer pardo'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>EventTimerDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>eventTimerDoFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>State</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=kt>int64</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>EventTime</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>eventTimerDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>ts</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>book</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Set an event-time timer to the element timestamp. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Timer</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>ts</span><span class=p>.</span><span class=nf>ToTime</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>eventTimerDoFn</span><span class=p>)</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Timer</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// process callback for this timer |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>AddEventTimeDoFn</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>in</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>eventTimerDoFn</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Timers are given family names so their callbacks can be handled independantly. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>Timer</span><span class=p>:</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>InEventTime</span><span class=p>(</span><span class=s>"processWatermark"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>State</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeValueState</span><span class=p>[</span><span class=kt>int64</span><span class=p>](</span><span class=s>"latest"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>in</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h4 id=processing-time-timers>11.3.2. Processing-time timers</h4><p>Processing-time timers fire when the real wall-clock time passes. This is often used to create larger batches of data |
| before processing. It can also be used to schedule events that should occur at a specific time. Just like with |
| event-time timers, processing-time timers are per key - each key has a separate copy of the timer.</p><p>While processing-time timers can be set to an absolute timestamp, it is very common to set them to an offset relative |
| to the current time. In Java, the <code>Timer.offset</code> and <code>Timer.setRelative</code> methods can be used to accomplish this.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"timer"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>timer</span> <span class=o>=</span> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timer</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>PROCESSING_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@TimerId</span><span class=o>(</span><span class=s>"timer"</span><span class=o>)</span> <span class=n>Timer</span> <span class=n>timer</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Set a timer to go off 30 seconds in the future. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>timer</span><span class=o>.</span><span class=na>offset</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>30</span><span class=o>)).</span><span class=na>setRelative</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimer</span><span class=o>(</span><span class=s>"timer"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>//Process timer. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>ProcessingTimerDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>ALL_ELEMENTS</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer'</span><span class=p>,</span> <span class=n>coders</span><span class=o>.</span><span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>TIMER</span> <span class=o>=</span> <span class=n>TimerSpec</span><span class=p>(</span><span class=s1>'timer'</span><span class=p>,</span> <span class=n>TimeDomain</span><span class=o>.</span><span class=n>REAL_TIME</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element_pair</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>TimerParam</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>element_pair</span><span class=p>[</span><span class=mi>1</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Set a timer to go off 30 seconds in the future.</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>.</span><span class=n>set</span><span class=p>(</span><span class=n>Timestamp</span><span class=o>.</span><span class=n>now</span><span class=p>()</span> <span class=o>+</span> <span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>30</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@on_timer</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>expiry_callback</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>buffer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Process timer.</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'ProcessingTime timer pardo'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>ProcessingTimerDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>processingTimerDoFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>ProcessingTime</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>processingTimerDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>book</span><span class=p>,</span> <span class=nx>word</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Set a timer to go off 30 seconds in the future. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Timer</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>time</span><span class=p>.</span><span class=nf>Now</span><span class=p>().</span><span class=nf>Add</span><span class=p>(</span><span class=mi>30</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>processingTimerDoFn</span><span class=p>)</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Timer</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// process callback for this timer |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>AddProcessingTimeDoFn</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>in</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>processingTimerDoFn</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Timers are given family names so their callbacks can be handled independantly. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>Timer</span><span class=p>:</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>InProcessingTime</span><span class=p>(</span><span class=s>"timer"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>in</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h4 id=dynamic-timer-tags>11.3.3. Dynamic timer tags</h4><p>Beam also supports dynamically setting a timer tag using <code>TimerMap</code> in the Java SDK. This allows for setting multiple different timers |
| in a <code>DoFn</code> and allowing for the timer tags to be dynamically chosen - e.g. based on data in the input elements. A |
| timer with a specific tag can only be set to a single timestamp, so setting the timer again has the effect of |
| overwriting the previous expiration time for the timer with that tag. Each <code>TimerMap</code> is identified with a timer family |
| id, and timers in different timer families are independent.</p><p>In the Python SDK, a dynamic timer tag can be specified while calling <code>set()</code> or <code>clear()</code>. By default, the timer |
| tag is an empty string if not specified.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerFamily</span><span class=o>(</span><span class=s>"actionTimers"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>timer</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timerMap</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>EVENT_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Timestamp</span> <span class=n>Instant</span> <span class=n>elementTs</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerFamily</span><span class=o>(</span><span class=s>"actionTimers"</span><span class=o>)</span> <span class=n>TimerMap</span> <span class=n>timers</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>timers</span><span class=o>.</span><span class=na>set</span><span class=o>(</span><span class=n>element</span><span class=o>.</span><span class=na>getValue</span><span class=o>().</span><span class=na>getActionType</span><span class=o>(),</span> <span class=n>elementTs</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimerFamily</span><span class=o>(</span><span class=s>"actionTimers"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>(</span><span class=nd>@TimerId</span> <span class=n>String</span> <span class=n>timerId</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>LOG</span><span class=o>.</span><span class=na>info</span><span class=o>(</span><span class=s>"Timer fired with id "</span> <span class=o>+</span> <span class=n>timerId</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>TimerDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>ALL_ELEMENTS</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer'</span><span class=p>,</span> <span class=n>coders</span><span class=o>.</span><span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>TIMER</span> <span class=o>=</span> <span class=n>TimerSpec</span><span class=p>(</span><span class=s1>'timer'</span><span class=p>,</span> <span class=n>TimeDomain</span><span class=o>.</span><span class=n>REAL_TIME</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element_pair</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>TimerParam</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>element_pair</span><span class=p>[</span><span class=mi>1</span><span class=p>])</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Set a timer to go off 30 seconds in the future with dynamic timer tag 'first_timer'.</span> |
| </span></span><span class=line><span class=cl> <span class=c1># And set a timer to go off 60 seconds in the future with dynamic timer tag 'second_timer'.</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>.</span><span class=n>set</span><span class=p>(</span><span class=n>Timestamp</span><span class=o>.</span><span class=n>now</span><span class=p>()</span> <span class=o>+</span> <span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>30</span><span class=p>),</span> <span class=n>dynamic_timer_tag</span><span class=o>=</span><span class=s1>'first_timer'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>.</span><span class=n>set</span><span class=p>(</span><span class=n>Timestamp</span><span class=o>.</span><span class=n>now</span><span class=p>()</span> <span class=o>+</span> <span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>60</span><span class=p>),</span> <span class=n>dynamic_timer_tag</span><span class=o>=</span><span class=s1>'second_timer'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Note that a timer can also be explicitly cleared if previously set with a dynamic timer tag:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># timer.clear(dynamic_timer_tag=...)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@on_timer</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>expiry_callback</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>buffer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>),</span> <span class=n>timer_tag</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>DynamicTimerTagParam</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Process timer, the dynamic timer tag associated with expiring timer can be read back with DoFn.DynamicTimerTagParam.</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=p>(</span><span class=n>timer_tag</span><span class=p>,</span> <span class=s1>'fired'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'ProcessingTime timer pardo'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>TimerDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>hasAction</span> <span class=kd>interface</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>Action</span><span class=p>()</span> <span class=kt>string</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>dynamicTagsDoFn</span><span class=p>[</span><span class=nx>V</span> <span class=nx>hasAction</span><span class=p>]</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>EventTime</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>dynamicTagsDoFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>ts</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>value</span> <span class=nx>V</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Set a timer to go off 30 seconds in the future. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Timer</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>ts</span><span class=p>.</span><span class=nf>ToTime</span><span class=p>(),</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>WithTag</span><span class=p>(</span><span class=nx>value</span><span class=p>.</span><span class=nf>Action</span><span class=p>()))</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>dynamicTagsDoFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emitWords</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Timer</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=nx>tag</span> <span class=o>:=</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Tag</span> <span class=c1>// Do something with fired tag |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>_</span> <span class=p>=</span> <span class=nx>tag</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nx>AddDynamicTimerTagsDoFn</span><span class=p>[</span><span class=nx>V</span> <span class=nx>hasAction</span><span class=p>](</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>in</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>dynamicTagsDoFn</span><span class=p>[</span><span class=nx>V</span><span class=p>]{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Timer</span><span class=p>:</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>InEventTime</span><span class=p>(</span><span class=s>"actionTimers"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>in</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h4 id=timer-output-timestamps>11.3.4. Timer output timestamps</h4><p>By default, event-time timers will hold the output watermark of the <code>ParDo</code> to the timestamp of the timer. This means |
| that if a timer is set to 12pm, any windowed aggregations or event-time timers later in the pipeline graph that finish |
| after 12pm will not expire. The timestamp of the timer is also the default output timestamp for the timer callback. This |
| means that any elements output from the onTimer method will have a timestamp equal to the timestamp of the timer firing. |
| For processing-time timers, the default output timestamp and watermark hold is the value of the input watermark at the |
| time the timer was set.</p><p>In some cases, a DoFn needs to output timestamps earlier than the timer expiration time, and therefore also needs to |
| hold its output watermark to those timestamps. For example, consider the following pipeline that temporarily batches |
| records into state, and sets a timer to drain the state. This code may appear correct, but will not work properly.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"elementBag"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>>></span> <span class=n>elementBag</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>bag</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"timerSet"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Boolean</span><span class=o>>></span> <span class=n>timerSet</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>timer</span> <span class=o>=</span> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timer</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>PROCESSING_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"elementBag"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>elementBag</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"timerSet"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Boolean</span><span class=o>></span> <span class=n>timerSet</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=n>Timer</span> <span class=n>timer</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Add the current element to the bag for this key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>elementBag</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>element</span><span class=o>.</span><span class=na>getValue</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(!</span><span class=n>MoreObjects</span><span class=o>.</span><span class=na>firstNonNull</span><span class=o>(</span><span class=n>timerSet</span><span class=o>.</span><span class=na>read</span><span class=o>(),</span> <span class=kc>false</span><span class=o>))</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// If the timer is not current set, then set it to go off in a minute. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>timer</span><span class=o>.</span><span class=na>offset</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>1</span><span class=o>)).</span><span class=na>setRelative</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>timerSet</span><span class=o>.</span><span class=na>write</span><span class=o>(</span><span class=kc>true</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimer</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"elementBag"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>elementBag</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"timerSet"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Boolean</span><span class=o>></span> <span class=n>timerSet</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>output</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=o>(</span><span class=n>ValueT</span> <span class=n>bufferedElement</span> <span class=o>:</span> <span class=n>elementBag</span><span class=o>.</span><span class=na>read</span><span class=o>())</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output each element. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>output</span><span class=o>.</span><span class=na>outputWithTimestamp</span><span class=o>(</span><span class=n>bufferedElement</span><span class=o>,</span> <span class=n>bufferedElement</span><span class=o>.</span><span class=na>timestamp</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=n>elementBag</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Note that the timer has now fired. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>timerSet</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>badTimerOutputTimestampsFn</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>]</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>ElementBag</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Bag</span><span class=p>[</span><span class=nx>V</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=nx>TimerSet</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=kt>bool</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=nx>OutputState</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>ProcessingTime</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>badTimerOutputTimestampsFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>value</span> <span class=nx>V</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Add the current element to the bag for this key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>if</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>ElementBag</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>value</span><span class=p>);</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>set</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>TimerSet</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>!</span><span class=nx>set</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>OutputState</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>time</span><span class=p>.</span><span class=nf>Now</span><span class=p>().</span><span class=nf>Add</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>TimerSet</span><span class=p>.</span><span class=nf>Write</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=kc>true</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>badTimerOutputTimestampsFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>OutputState</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=nx>vs</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>ElementBag</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>v</span> <span class=o>:=</span> <span class=k>range</span> <span class=nx>vs</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output each element |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nf>emit</span><span class=p>(</span><span class=nx>fmt</span><span class=p>.</span><span class=nf>Sprintf</span><span class=p>(</span><span class=s>"%v"</span><span class=p>,</span> <span class=nx>v</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>ElementBag</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Note that the timer has now fired. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>TimerSet</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p>The problem with this code is that the ParDo is buffering elements, however nothing is preventing the watermark |
| from advancing past the timestamp of those elements, so all those elements might be dropped as late data. In order |
| to prevent this from happening, an output timestamp needs to be set on the timer to prevent the watermark from advancing |
| past the timestamp of the minimum element. The following code demonstrates this.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The bag of elements accumulated. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"elementBag"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>>></span> <span class=n>elementBag</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>bag</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The timestamp of the timer set. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"timerTimestamp"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Long</span><span class=o>>></span> <span class=n>timerTimestamp</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The minimum timestamp stored in the bag. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"minTimestampInBag"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>>></span> |
| </span></span><span class=line><span class=cl> <span class=n>minTimestampInBag</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>combining</span><span class=o>(</span><span class=n>Min</span><span class=o>.</span><span class=na>ofLongs</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>timer</span> <span class=o>=</span> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timer</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>PROCESSING_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"elementBag"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>elementBag</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"timerTimestamp"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Long</span><span class=o>></span> <span class=n>timerTimestamp</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"minTimestampInBag"</span><span class=o>)</span> <span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>></span> <span class=n>minTimestamp</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=n>Timer</span> <span class=n>timer</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Add the current element to the bag for this key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>elementBag</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>element</span><span class=o>.</span><span class=na>getValue</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Keep track of the minimum element timestamp currently stored in the bag. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>minTimestamp</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>element</span><span class=o>.</span><span class=na>getValue</span><span class=o>().</span><span class=na>timestamp</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// If the timer is already set, then reset it at the same time but with an updated output timestamp (otherwise |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// we would keep resetting the timer to the future). If there is no timer set, then set one to expire in a minute. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>Long</span> <span class=n>timerTimestampMs</span> <span class=o>=</span> <span class=n>timerTimestamp</span><span class=o>.</span><span class=na>read</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Instant</span> <span class=n>timerToSet</span> <span class=o>=</span> <span class=o>(</span><span class=n>timerTimestamp</span><span class=o>.</span><span class=na>isEmpty</span><span class=o>().</span><span class=na>read</span><span class=o>())</span> |
| </span></span><span class=line><span class=cl> <span class=o>?</span> <span class=n>Instant</span><span class=o>.</span><span class=na>now</span><span class=o>().</span><span class=na>plus</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>1</span><span class=o>))</span> <span class=o>:</span> <span class=k>new</span> <span class=n>Instant</span><span class=o>(</span><span class=n>timerTimestampMs</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Setting the outputTimestamp to the minimum timestamp in the bag holds the watermark to that timestamp until the |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// timer fires. This allows outputting all the elements with their timestamp. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>timer</span><span class=o>.</span><span class=na>withOutputTimestamp</span><span class=o>(</span><span class=n>minTimestamp</span><span class=o>.</span><span class=na>read</span><span class=o>()).</span><span class=na>s</span> <span class=n>et</span><span class=o>(</span><span class=n>timerToSet</span><span class=o>).</span> |
| </span></span><span class=line><span class=cl> <span class=n>timerTimestamp</span><span class=o>.</span><span class=na>write</span><span class=o>(</span><span class=n>timerToSet</span><span class=o>.</span><span class=na>getMillis</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimer</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"elementBag"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>elementBag</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"timerTimestamp"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Long</span><span class=o>></span> <span class=n>timerTimestamp</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>output</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=o>(</span><span class=n>ValueT</span> <span class=n>bufferedElement</span> <span class=o>:</span> <span class=n>elementBag</span><span class=o>.</span><span class=na>read</span><span class=o>())</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output each element. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>output</span><span class=o>.</span><span class=na>outputWithTimestamp</span><span class=o>(</span><span class=n>bufferedElement</span><span class=o>,</span> <span class=n>bufferedElement</span><span class=o>.</span><span class=na>timestamp</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Note that the timer has now fired. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>timerTimestamp</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>Timer</span> <span class=n>output</span> <span class=n>timestamps</span> <span class=ow>is</span> <span class=ow>not</span> <span class=n>yet</span> <span class=n>supported</span> <span class=ow>in</span> <span class=n>Python</span> <span class=n>SDK</span><span class=o>.</span> <span class=n>See</span> <span class=n>https</span><span class=p>:</span><span class=o>//</span><span class=n>github</span><span class=o>.</span><span class=n>com</span><span class=o>/</span><span class=n>apache</span><span class=o>/</span><span class=n>beam</span><span class=o>/</span><span class=n>issues</span><span class=o>/</span><span class=mf>20705.</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>element</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>]</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Timestamp</span> <span class=kt>int64</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Value</span> <span class=nx>V</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>goodTimerOutputTimestampsFn</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>]</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>ElementBag</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Bag</span><span class=p>[</span><span class=nx>element</span><span class=p>[</span><span class=nx>V</span><span class=p>]]</span> <span class=c1>// The bag of elements accumulated. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>TimerTimerstamp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=kt>int64</span><span class=p>]</span> <span class=c1>// The timestamp of the timer set. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>MinTimestampInBag</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Combining</span><span class=p>[</span><span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>]</span> <span class=c1>// The minimum timestamp stored in the bag. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>OutputState</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>ProcessingTime</span> <span class=c1>// The timestamp of the timer. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>goodTimerOutputTimestampsFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>et</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>value</span> <span class=nx>V</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Add the current element to the bag for this key, and preserve the event time. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>if</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>ElementBag</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>element</span><span class=p>[</span><span class=nx>V</span><span class=p>]{</span><span class=nx>Timestamp</span><span class=p>:</span> <span class=nx>et</span><span class=p>.</span><span class=nf>Milliseconds</span><span class=p>(),</span> <span class=nx>Value</span><span class=p>:</span> <span class=nx>value</span><span class=p>});</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Keep track of the minimum element timestamp currently stored in the bag. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>MinTimestampInBag</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>et</span><span class=p>.</span><span class=nf>Milliseconds</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// If the timer is already set, then reset it at the same time but with an updated output timestamp (otherwise |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// we would keep resetting the timer to the future). If there is no timer set, then set one to expire in a minute. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>ts</span><span class=p>,</span> <span class=nx>ok</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>TimerTimerstamp</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=kd>var</span> <span class=nx>tsToSet</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Time</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>ok</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>tsToSet</span> <span class=p>=</span> <span class=nx>time</span><span class=p>.</span><span class=nf>UnixMilli</span><span class=p>(</span><span class=nx>ts</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> <span class=k>else</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>tsToSet</span> <span class=p>=</span> <span class=nx>time</span><span class=p>.</span><span class=nf>Now</span><span class=p>().</span><span class=nf>Add</span><span class=p>(</span><span class=mi>1</span> <span class=o>*</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>minTs</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>MinTimestampInBag</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>outputTs</span> <span class=o>:=</span> <span class=nx>time</span><span class=p>.</span><span class=nf>UnixMilli</span><span class=p>(</span><span class=nx>minTs</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Setting the outputTimestamp to the minimum timestamp in the bag holds the watermark to that timestamp until the |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// timer fires. This allows outputting all the elements with their timestamp. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>OutputState</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>tsToSet</span><span class=p>,</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>WithOutputTimestamp</span><span class=p>(</span><span class=nx>outputTs</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>TimerTimerstamp</span><span class=p>.</span><span class=nf>Write</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>tsToSet</span><span class=p>.</span><span class=nf>UnixMilli</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>goodTimerOutputTimestampsFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=kt>string</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>OutputState</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=nx>vs</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>ElementBag</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>v</span> <span class=o>:=</span> <span class=k>range</span> <span class=nx>vs</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output each element with their timestamp |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nf>emit</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nf>EventTime</span><span class=p>(</span><span class=nx>v</span><span class=p>.</span><span class=nx>Timestamp</span><span class=p>),</span> <span class=nx>fmt</span><span class=p>.</span><span class=nf>Sprintf</span><span class=p>(</span><span class=s>"%v"</span><span class=p>,</span> <span class=nx>v</span><span class=p>.</span><span class=nx>Value</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>ElementBag</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Note that the timer has now fired. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>TimerTimerstamp</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nx>AddTimedOutputBatching</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>](</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>in</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>goodTimerOutputTimestampsFn</span><span class=p>[</span><span class=nx>V</span><span class=p>]{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>ElementBag</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeBagState</span><span class=p>[</span><span class=nx>element</span><span class=p>[</span><span class=nx>V</span><span class=p>]](</span><span class=s>"elementBag"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>TimerTimerstamp</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeValueState</span><span class=p>[</span><span class=kt>int64</span><span class=p>](</span><span class=s>"timerTimestamp"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>MinTimestampInBag</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeCombiningState</span><span class=p>[</span><span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>](</span><span class=s>"minTimestampInBag"</span><span class=p>,</span> <span class=kd>func</span><span class=p>(</span><span class=nx>a</span><span class=p>,</span> <span class=nx>b</span> <span class=kt>int64</span><span class=p>)</span> <span class=kt>int64</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>a</span> <span class=p><</span> <span class=nx>b</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>a</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>b</span> |
| </span></span><span class=line><span class=cl> <span class=p>}),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>OutputState</span><span class=p>:</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>InProcessingTime</span><span class=p>(</span><span class=s>"outputState"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>in</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=garbage-collecting-state>11.4. Garbage collecting state</h3><p>Per-key state needs to be garbage collected, or eventually the increasing size of state may negatively impact |
| performance. There are two common strategies for garbage collecting state.</p><h5 id=using-windows-for-garbage-collection>11.4.1. <strong>Using windows for garbage collection</strong></h5><p>All state and timers for a key is scoped to the window it is in. This means that depending on the timestamp of the |
| input element the ParDo will see different values for the state depending on the window that element falls into. In |
| addition, once the input watermark passes the end of the window, the runner should garbage collect all state for that |
| window. (note: if allowed lateness is set to a positive value for the window, the runner must wait for the watermark to |
| pass the end of the window plus the allowed lateness before garbage collecting state). This can be used as a |
| garbage-collection strategy.</p><p>For example, given the following:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>Window</span><span class=o>.</span><span class=na>into</span><span class=o>(</span><span class=n>CalendarWindows</span><span class=o>.</span><span class=na>days</span><span class=o>(</span><span class=n>1</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withTimeZone</span><span class=o>(</span><span class=n>DateTimeZone</span><span class=o>.</span><span class=na>forID</span><span class=o>(</span><span class=s>"America/Los_Angeles"</span><span class=o>))));</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>>></span> <span class=n>state</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span><span class=nd>@Timestamp</span> <span class=n>Instant</span> <span class=n>ts</span><span class=o>,</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>state</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The state is scoped to a calendar day window. That means that if the input timestamp ts is after |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// midnight PST, then a new copy of the state will be seen for the next day. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>StateDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>ALL_ELEMENTS</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer'</span><span class=p>,</span> <span class=n>coders</span><span class=o>.</span><span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element_pair</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'Windowing'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>WindowInto</span><span class=p>(</span><span class=n>FixedWindows</span><span class=p>(</span><span class=mi>60</span> <span class=o>*</span> <span class=mi>60</span> <span class=o>*</span> <span class=mi>24</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'DoFn'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>StateDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl> <span class=nx>items</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>statefulDoFn</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>S</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeValueState</span><span class=p>[</span><span class=kt>int</span><span class=p>](</span><span class=s>"S"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>elements</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>out</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>WindowInto</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>window</span><span class=p>.</span><span class=nf>NewFixedWindows</span><span class=p>(</span><span class=mi>24</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Hour</span><span class=p>),</span> <span class=nx>items</span><span class=p>)</span></span></span></code></pre></div></div></div><p>This <code>ParDo</code> stores state per day. Once the pipeline is done processing data for a given day, all the state for that |
| day is garbage collected.</p><h5 id=using-timers-for-garbage-collection>11.4.1. <strong>Using timers For garbage collection</strong></h5><p>In some cases, it is difficult to find a windowing strategy that models the desired garbage-collection strategy. For |
| example, a common desire is to garbage collect state for a key once no activity has been seen on the key for some time. |
| This can be done by updating a timer that garbage collects state. For example</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The state for the key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>ValueT</span><span class=o>>></span> <span class=n>state</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// The maximum element timestamp seen so far. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"maxTimestampSeen"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>>></span> |
| </span></span><span class=line><span class=cl> <span class=n>maxTimestamp</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>combining</span><span class=o>(</span><span class=n>Max</span><span class=o>.</span><span class=na>ofLongs</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"gcTimer"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>gcTimer</span> <span class=o>=</span> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timer</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>EVENT_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Timestamp</span> <span class=n>Instant</span> <span class=n>ts</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>state</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"maxTimestampSeen"</span><span class=o>)</span> <span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>></span> <span class=n>maxTimestamp</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"gcTimer"</span><span class=o>)</span> <span class=n>gcTimer</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>updateState</span><span class=o>(</span><span class=n>state</span><span class=o>,</span> <span class=n>element</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=n>maxTimestamp</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>ts</span><span class=o>.</span><span class=na>getMillis</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Set the timer to be one hour after the maximum timestamp seen. This will keep overwriting the same timer, so |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// as long as there is activity on this key the state will stay active. Once the key goes inactive for one hour's |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// worth of event time (as measured by the watermark), then the gc timer will fire. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>Instant</span> <span class=n>expirationTime</span> <span class=o>=</span> <span class=k>new</span> <span class=n>Instant</span><span class=o>(</span><span class=n>maxTimestamp</span><span class=o>.</span><span class=na>read</span><span class=o>()).</span><span class=na>plus</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardHours</span><span class=o>(</span><span class=n>1</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>.</span><span class=na>set</span><span class=o>(</span><span class=n>expirationTime</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimer</span><span class=o>(</span><span class=s>"gcTimer"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>state</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"maxTimestampSeen"</span><span class=o>)</span> <span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>></span> <span class=n>maxTimestamp</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Clear all state for the key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>state</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>maxTimestamp</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>UserDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>ALL_ELEMENTS</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'state'</span><span class=p>,</span> <span class=n>coders</span><span class=o>.</span><span class=n>VarIntCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>MAX_TIMESTAMP</span> <span class=o>=</span> <span class=n>CombiningValueStateSpec</span><span class=p>(</span><span class=s1>'max_timestamp_seen'</span><span class=p>,</span> <span class=nb>max</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>TIMER</span> <span class=o>=</span> <span class=n>TimerSpec</span><span class=p>(</span><span class=s1>'gc-timer'</span><span class=p>,</span> <span class=n>TimeDomain</span><span class=o>.</span><span class=n>WATERMARK</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>t</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>TimestampParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>MAX_TIMESTAMP</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>TimerParam</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>update_state</span><span class=p>(</span><span class=n>state</span><span class=p>,</span> <span class=n>element</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>t</span><span class=o>.</span><span class=n>micros</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># Set the timer to be one hour after the maximum timestamp seen. This will keep overwriting the same timer, so</span> |
| </span></span><span class=line><span class=cl> <span class=c1># as long as there is activity on this key the state will stay active. Once the key goes inactive for one hour's</span> |
| </span></span><span class=line><span class=cl> <span class=c1># worth of event time (as measured by the watermark), then the gc timer will fire.</span> |
| </span></span><span class=line><span class=cl> <span class=n>expiration_time</span> <span class=o>=</span> <span class=n>Timestamp</span><span class=p>(</span><span class=n>micros</span><span class=o>=</span><span class=n>max_timestamp</span><span class=o>.</span><span class=n>read</span><span class=p>())</span> <span class=o>+</span> <span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>60</span><span class=o>*</span><span class=mi>60</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>.</span><span class=n>set</span><span class=p>(</span><span class=n>expiration_time</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@on_timer</span><span class=p>(</span><span class=n>TIMER</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>expiry_callback</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>ALL_ELEMENTS</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp</span> <span class=o>=</span> <span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>MAX_TIMESTAMP</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>state</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'Read per user'</span> <span class=o>>></span> <span class=n>ReadPerUser</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'User DoFn'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>UserDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>timerGarbageCollectionFn</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>]</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>State</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=nx>V</span><span class=p>]</span> <span class=c1>// The state for the key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>MaxTimestampInBag</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Combining</span><span class=p>[</span><span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>]</span> <span class=c1>// The maximum element timestamp seen so far. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>GcTimer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>EventTime</span> <span class=c1>// The timestamp of the timer. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>timerGarbageCollectionFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>et</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>value</span> <span class=nx>V</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>updateState</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>State</span><span class=p>,</span> <span class=nx>key</span><span class=p>,</span> <span class=nx>value</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>MaxTimestampInBag</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>et</span><span class=p>.</span><span class=nf>Milliseconds</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Set the timer to be one hour after the maximum timestamp seen. This will keep overwriting the same timer, so |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// as long as there is activity on this key the state will stay active. Once the key goes inactive for one hour's |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// worth of event time (as measured by the watermark), then the gc timer will fire. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>maxTs</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>MaxTimestampInBag</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>expirationTime</span> <span class=o>:=</span> <span class=nx>time</span><span class=p>.</span><span class=nf>UnixMilli</span><span class=p>(</span><span class=nx>maxTs</span><span class=p>).</span><span class=nf>Add</span><span class=p>(</span><span class=mi>1</span> <span class=o>*</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Hour</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>GcTimer</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>expirationTime</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>timerGarbageCollectionFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>GcTimer</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Clear all the state for the key |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>fn</span><span class=p>.</span><span class=nx>State</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>MaxTimestampInBag</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nx>AddTimerGarbageCollection</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>](</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>in</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>timerGarbageCollectionFn</span><span class=p>[</span><span class=nx>V</span><span class=p>]{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>State</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeValueState</span><span class=p>[</span><span class=nx>V</span><span class=p>](</span><span class=s>"timerTimestamp"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>MaxTimestampInBag</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeCombiningState</span><span class=p>[</span><span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>](</span><span class=s>"maxTimestampInBag"</span><span class=p>,</span> <span class=kd>func</span><span class=p>(</span><span class=nx>a</span><span class=p>,</span> <span class=nx>b</span> <span class=kt>int64</span><span class=p>)</span> <span class=kt>int64</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>a</span> <span class=p>></span> <span class=nx>b</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>a</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>b</span> |
| </span></span><span class=line><span class=cl> <span class=p>}),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>GcTimer</span><span class=p>:</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>InEventTime</span><span class=p>(</span><span class=s>"gcTimer"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>in</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=state-timers-examples>11.5. State and timers examples</h3><p>Following are some example uses of state and timers</p><h4 id=joining-clicks-and-views>11.5.1. Joining clicks and views</h4><p>In this example, the pipeline is processing data from an e-commerce site’s home page. There are two input streams: |
| a stream of views, representing suggested product links displayed to the user on the home page, and a stream of |
| clicks, representing actual user clicks on these links. The goal of the pipeline is to join click events with view |
| events, outputting a new joined event that contains information from both events. Each link has a unique identifier |
| that is present in both the view event and the join event.</p><p>Many view events will never be followed up with clicks. This pipeline will wait one hour for a click, after which it |
| will give up on this join. While every click event should have a view event, some small number of view events may be |
| lost and never make it to the Beam pipeline; the pipeline will similarly wait one hour after seeing a click event, and |
| give up if the view event does not arrive in that time. Input events are not ordered - it is possible to see the click |
| event before the view event. The one hour join timeout should be based on event time, not on processing time.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=c1>// Read the event stream and key it by the link id. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Event</span><span class=o>>></span> <span class=n>eventsPerLinkId</span> <span class=o>=</span> |
| </span></span><span class=line><span class=cl> <span class=n>readEvents</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>WithKeys</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>Event</span><span class=o>::</span><span class=n>getLinkId</span><span class=o>).</span><span class=na>withKeyType</span><span class=o>(</span><span class=n>TypeDescriptors</span><span class=o>.</span><span class=na>strings</span><span class=o>()));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>eventsPerLinkId</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Event</span><span class=o>>,</span> <span class=n>JoinedEvent</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Store the view event. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"view"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>>></span> <span class=n>viewState</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Store the click event. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"click"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>>></span> <span class=n>clickState</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// The maximum element timestamp seen so far. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"maxTimestampSeen"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>>></span> |
| </span></span><span class=line><span class=cl> <span class=n>maxTimestamp</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>combining</span><span class=o>(</span><span class=n>Max</span><span class=o>.</span><span class=na>ofLongs</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Timer that fires when an hour goes by with an incomplete join. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"gcTimer"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>gcTimer</span> <span class=o>=</span> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timer</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>EVENT_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Event</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Timestamp</span> <span class=n>Instant</span> <span class=n>ts</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"view"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>></span> <span class=n>viewState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"click"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>></span> <span class=n>clickState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@AlwaysFetched</span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"maxTimestampSeen"</span><span class=o>)</span> <span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>></span> <span class=n>maxTimestampState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"gcTimer"</span><span class=o>)</span> <span class=n>gcTimer</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>JoinedEvent</span><span class=o>></span> <span class=n>output</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Store the event into the correct state variable. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>Event</span> <span class=n>event</span> <span class=o>=</span> <span class=n>element</span><span class=o>.</span><span class=na>getValue</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>></span> <span class=n>valueState</span> <span class=o>=</span> <span class=n>event</span><span class=o>.</span><span class=na>getType</span><span class=o>().</span><span class=na>equals</span><span class=o>(</span><span class=n>VIEW</span><span class=o>)</span> <span class=o>?</span> <span class=n>viewState</span> <span class=o>:</span> <span class=n>clickState</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=n>valueState</span><span class=o>.</span><span class=na>write</span><span class=o>(</span><span class=n>event</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=n>Event</span> <span class=n>view</span> <span class=o>=</span> <span class=n>viewState</span><span class=o>.</span><span class=na>read</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Event</span> <span class=n>click</span> <span class=o>=</span> <span class=n>clickState</span><span class=o>.</span><span class=na>read</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>(</span><span class=k>if</span> <span class=n>view</span> <span class=o>!=</span> <span class=kc>null</span> <span class=o>&&</span> <span class=n>click</span> <span class=o>!=</span> <span class=kc>null</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// We've seen both a view and a click. Output a joined event and clear state. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>output</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>JoinedEvent</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>view</span><span class=o>,</span> <span class=n>click</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=n>clearState</span><span class=o>(</span><span class=n>viewState</span><span class=o>,</span> <span class=n>clickState</span><span class=o>,</span> <span class=n>maxTimestampState</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> <span class=k>else</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// We've only seen on half of the join. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Set the timer to be one hour after the maximum timestamp seen. This will keep overwriting the same timer, so |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// as long as there is activity on this key the state will stay active. Once the key goes inactive for one hour's |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// worth of event time (as measured by the watermark), then the gc timer will fire. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>maxTimestampState</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>ts</span><span class=o>.</span><span class=na>getMillis</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=n>Instant</span> <span class=n>expirationTime</span> <span class=o>=</span> <span class=k>new</span> <span class=n>Instant</span><span class=o>(</span><span class=n>maxTimestampState</span><span class=o>.</span><span class=na>read</span><span class=o>()).</span><span class=na>plus</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardHours</span><span class=o>(</span><span class=n>1</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=n>gcTimer</span><span class=o>.</span><span class=na>set</span><span class=o>(</span><span class=n>expirationTime</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimer</span><span class=o>(</span><span class=s>"gcTimer"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"view"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>></span> <span class=n>viewState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"click"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>></span> <span class=n>clickState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"maxTimestampSeen"</span><span class=o>)</span> <span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>></span> <span class=n>maxTimestampState</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// An hour has gone by with an incomplete join. Give up and clear the state. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>clearState</span><span class=o>(</span><span class=n>viewState</span><span class=o>,</span> <span class=n>clickState</span><span class=o>,</span> <span class=n>maxTimestampState</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>private</span> <span class=kt>void</span> <span class=nf>clearState</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"view"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>></span> <span class=n>viewState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"click"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Event</span><span class=o>></span> <span class=n>clickState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"maxTimestampSeen"</span><span class=o>)</span> <span class=n>CombiningState</span><span class=o><</span><span class=n>Long</span><span class=o>,</span> <span class=kt>long</span><span class=o>[],</span> <span class=n>Long</span><span class=o>></span> <span class=n>maxTimestampState</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>viewState</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>clickState</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>maxTimestampState</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>JoinDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># stores the view event.</span> |
| </span></span><span class=line><span class=cl> <span class=n>VIEW_STATE_SPEC</span> <span class=o>=</span> <span class=n>ReadModifyWriteStateSpec</span><span class=p>(</span><span class=s1>'view'</span><span class=p>,</span> <span class=n>EventCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=c1># stores the click event.</span> |
| </span></span><span class=line><span class=cl> <span class=n>CLICK_STATE_SPEC</span> <span class=o>=</span> <span class=n>ReadModifyWriteStateSpec</span><span class=p>(</span><span class=s1>'click'</span><span class=p>,</span> <span class=n>EventCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=c1># The maximum element timestamp value seen so far.</span> |
| </span></span><span class=line><span class=cl> <span class=n>MAX_TIMESTAMP</span> <span class=o>=</span> <span class=n>CombiningValueStateSpec</span><span class=p>(</span><span class=s1>'max_timestamp_seen'</span><span class=p>,</span> <span class=nb>max</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Timer that fires when an hour goes by with an incomplete join.</span> |
| </span></span><span class=line><span class=cl> <span class=n>GC_TIMER</span> <span class=o>=</span> <span class=n>TimerSpec</span><span class=p>(</span><span class=s1>'gc'</span><span class=p>,</span> <span class=n>TimeDomain</span><span class=o>.</span><span class=n>WATERMARK</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>view</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>VIEW_STATE_SPEC</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>click</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>CLICK_STATE_SPEC</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp_seen</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>MAX_TIMESTAMP</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>ts</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimestampParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>gc</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimerParam</span><span class=p>(</span><span class=n>GC_TIMER</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>event</span> <span class=o>=</span> <span class=n>element</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>event</span><span class=o>.</span><span class=n>type</span> <span class=o>==</span> <span class=s1>'view'</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>view</span><span class=o>.</span><span class=n>write</span><span class=p>(</span><span class=n>event</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>else</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>click</span><span class=o>.</span><span class=n>write</span><span class=p>(</span><span class=n>event</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=n>previous_view</span> <span class=o>=</span> <span class=n>view</span><span class=o>.</span><span class=n>read</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=n>previous_click</span> <span class=o>=</span> <span class=n>click</span><span class=o>.</span><span class=n>read</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># We've seen both a view and a click. Output a joined event and clear state.</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>previous_view</span> <span class=ow>and</span> <span class=n>previous_click</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=p>(</span><span class=n>previous_view</span><span class=p>,</span> <span class=n>previous_click</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>view</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=n>click</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp_seen</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=k>else</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp_seen</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>ts</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>gc</span><span class=o>.</span><span class=n>set</span><span class=p>(</span><span class=n>max_timestamp_seen</span><span class=o>.</span><span class=n>read</span><span class=p>()</span> <span class=o>+</span> <span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>3600</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@on_timer</span><span class=p>(</span><span class=n>GC_TIMER</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>gc_callback</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>view</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>VIEW_STATE_SPEC</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>click</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>CLICK_STATE_SPEC</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp_seen</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>MAX_TIMESTAMP</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>view</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=n>click</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_timestamp_seen</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>_</span> <span class=o>=</span> <span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=s1>'EventsPerLinkId'</span> <span class=o>>></span> <span class=n>ReadPerLinkEvents</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'Join DoFn'</span> <span class=o>>></span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>JoinDoFn</span><span class=p>()))</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>JoinedEvent</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>View</span><span class=p>,</span> <span class=nx>Click</span> <span class=o>*</span><span class=nx>Event</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>type</span> <span class=nx>joinDoFn</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>View</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=o>*</span><span class=nx>Event</span><span class=p>]</span> <span class=c1>// Store the view event. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>Click</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=o>*</span><span class=nx>Event</span><span class=p>]</span> <span class=c1>// Store the click event. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=nx>MaxTimestampSeen</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Combining</span><span class=p>[</span><span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>]</span> <span class=c1>// The maximum element timestamp seen so far. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>GcTimer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>EventTime</span> <span class=c1>// The timestamp of the timer. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>joinDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>et</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>event</span> <span class=o>*</span><span class=nx>Event</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>JoinedEvent</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>valueState</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>View</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>event</span><span class=p>.</span><span class=nf>isClick</span><span class=p>()</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>valueState</span> <span class=p>=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Click</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>valueState</span><span class=p>.</span><span class=nf>Write</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>event</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>view</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>View</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>click</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Click</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>view</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=o>&&</span> <span class=nx>click</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nx>JoinedEvent</span><span class=p>{</span><span class=nx>View</span><span class=p>:</span> <span class=nx>view</span><span class=p>,</span> <span class=nx>Click</span><span class=p>:</span> <span class=nx>click</span><span class=p>})</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nf>clearState</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>MaxTimestampSeen</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>et</span><span class=p>.</span><span class=nf>Milliseconds</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=nx>expTs</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>MaxTimestampSeen</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>GcTimer</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>time</span><span class=p>.</span><span class=nf>UnixMilli</span><span class=p>(</span><span class=nx>expTs</span><span class=p>).</span><span class=nf>Add</span><span class=p>(</span><span class=mi>1</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Hour</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>joinDoFn</span><span class=p>)</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=kt>string</span><span class=p>))</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>GcTimer</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nf>clearState</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>joinDoFn</span><span class=p>)</span> <span class=nf>clearState</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>View</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>Click</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>MaxTimestampSeen</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nf>AddJoinDoFn</span><span class=p>(</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>in</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>joinDoFn</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>View</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeValueState</span><span class=p>[</span><span class=o>*</span><span class=nx>Event</span><span class=p>](</span><span class=s>"view"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Click</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeValueState</span><span class=p>[</span><span class=o>*</span><span class=nx>Event</span><span class=p>](</span><span class=s>"click"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>MaxTimestampSeen</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeCombiningState</span><span class=p>[</span><span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>,</span> <span class=kt>int64</span><span class=p>](</span><span class=s>"maxTimestampSeen"</span><span class=p>,</span> <span class=kd>func</span><span class=p>(</span><span class=nx>a</span><span class=p>,</span> <span class=nx>b</span> <span class=kt>int64</span><span class=p>)</span> <span class=kt>int64</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>a</span> <span class=p>></span> <span class=nx>b</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>a</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>b</span> |
| </span></span><span class=line><span class=cl> <span class=p>}),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>GcTimer</span><span class=p>:</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>InEventTime</span><span class=p>(</span><span class=s>"gcTimer"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>in</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h4 id=batching-rpcs>11.5.2. Batching RPCs</h4><p>In this example, input elements are being forwarded to an external RPC service. The RPC accepts batch requests - |
| multiple events for the same user can be batched in a single RPC call. Since this RPC service also imposes rate limits, |
| we want to batch ten seconds worth of events together in order to reduce the number of calls.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>></span> <span class=n>perUser</span> <span class=o>=</span> <span class=n>readPerUser</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=n>perUser</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>ParDo</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=k>new</span> <span class=n>DoFn</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>>,</span> <span class=n>OutputT</span><span class=o>>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Store the elements buffered so far. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>>></span> <span class=n>elements</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>bag</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Keep track of whether a timer is currently set or not. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"isTimerSet"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>StateSpec</span><span class=o><</span><span class=n>ValueState</span><span class=o><</span><span class=n>Boolean</span><span class=o>>></span> <span class=n>isTimerSet</span> <span class=o>=</span> <span class=n>StateSpecs</span><span class=o>.</span><span class=na>value</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The processing-time timer user to publish the RPC. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=kd>private</span> <span class=kd>final</span> <span class=n>TimerSpec</span> <span class=n>timer</span> <span class=o>=</span> <span class=n>TimerSpecs</span><span class=o>.</span><span class=na>timer</span><span class=o>(</span><span class=n>TimeDomain</span><span class=o>.</span><span class=na>PROCESSING_TIME</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>process</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>KV</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>ValueT</span><span class=o>></span> <span class=n>element</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>elementsState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"isTimerSet"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Boolean</span><span class=o>></span> <span class=n>isTimerSetState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@TimerId</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=n>Timer</span> <span class=n>timer</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Add the current element to the bag for this key. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>state</span><span class=o>.</span><span class=na>add</span><span class=o>(</span><span class=n>element</span><span class=o>.</span><span class=na>getValue</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(!</span><span class=n>MoreObjects</span><span class=o>.</span><span class=na>firstNonNull</span><span class=o>(</span><span class=n>isTimerSetState</span><span class=o>.</span><span class=na>read</span><span class=o>(),</span> <span class=kc>false</span><span class=o>))</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// If there is no timer currently set, then set one to go off in 10 seconds. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>timer</span><span class=o>.</span><span class=na>offset</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>10</span><span class=o>)).</span><span class=na>setRelative</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>isTimerSetState</span><span class=o>.</span><span class=na>write</span><span class=o>(</span><span class=kc>true</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@OnTimer</span><span class=o>(</span><span class=s>"outputState"</span><span class=o>)</span> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>onTimer</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"state"</span><span class=o>)</span> <span class=n>BagState</span><span class=o><</span><span class=n>ValueT</span><span class=o>></span> <span class=n>elementsState</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@StateId</span><span class=o>(</span><span class=s>"isTimerSet"</span><span class=o>)</span> <span class=n>ValueState</span><span class=o><</span><span class=n>Boolean</span><span class=o>></span> <span class=n>isTimerSetState</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Send an RPC containing the batched elements and clear state. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>sendRPC</span><span class=o>(</span><span class=n>elementsState</span><span class=o>.</span><span class=na>read</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=n>elementsState</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>isTimerSetState</span><span class=o>.</span><span class=na>clear</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}));</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>BufferDoFn</span><span class=p>(</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>BUFFER</span> <span class=o>=</span> <span class=n>BagStateSpec</span><span class=p>(</span><span class=s1>'buffer'</span><span class=p>,</span> <span class=n>EventCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>IS_TIMER_SET</span> <span class=o>=</span> <span class=n>ReadModifyWriteStateSpec</span><span class=p>(</span><span class=s1>'is_timer_set'</span><span class=p>,</span> <span class=n>BooleanCoder</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=n>OUTPUT</span> <span class=o>=</span> <span class=n>TimerSpec</span><span class=p>(</span><span class=s1>'output'</span><span class=p>,</span> <span class=n>TimeDomain</span><span class=o>.</span><span class=n>REAL_TIME</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>BUFFER</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>is_timer_set</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>IS_TIMER_SET</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimerParam</span><span class=p>(</span><span class=n>OUTPUT</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>.</span><span class=n>add</span><span class=p>(</span><span class=n>element</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=ow>not</span> <span class=n>is_timer_set</span><span class=o>.</span><span class=n>read</span><span class=p>():</span> |
| </span></span><span class=line><span class=cl> <span class=n>timer</span><span class=o>.</span><span class=n>set</span><span class=p>(</span><span class=n>Timestamp</span><span class=o>.</span><span class=n>now</span><span class=p>()</span> <span class=o>+</span> <span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>10</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=n>is_timer_set</span><span class=o>.</span><span class=n>write</span><span class=p>(</span><span class=kc>True</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@on_timer</span><span class=p>(</span><span class=n>OUTPUT</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>output_callback</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>BUFFER</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=n>is_timer_set</span><span class=o>=</span><span class=n>DoFn</span><span class=o>.</span><span class=n>StateParam</span><span class=p>(</span><span class=n>IS_TIMER_SET</span><span class=p>)):</span> |
| </span></span><span class=line><span class=cl> <span class=n>send_rpc</span><span class=p>(</span><span class=nb>list</span><span class=p>(</span><span class=n>buffer</span><span class=o>.</span><span class=n>read</span><span class=p>()))</span> |
| </span></span><span class=line><span class=cl> <span class=n>buffer</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=n>is_timer_set</span><span class=o>.</span><span class=n>clear</span><span class=p>()</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>bufferDoFn</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>]</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Elements</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Bag</span><span class=p>[</span><span class=nx>V</span><span class=p>]</span> <span class=c1>// Store the elements buffered so far. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>IsTimerSet</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Value</span><span class=p>[</span><span class=kt>bool</span><span class=p>]</span> <span class=c1>// Keep track of whether a timer is currently set or not. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=nx>OutputElements</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>ProcessingTime</span> <span class=c1>// The processing-time timer user to publish the RPC. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>bufferDoFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>et</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>value</span> <span class=nx>V</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>Elements</span><span class=p>.</span><span class=nf>Add</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=nx>value</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>isSet</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>IsTimerSet</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>!</span><span class=nx>isSet</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>OutputElements</span><span class=p>.</span><span class=nf>Set</span><span class=p>(</span><span class=nx>tp</span><span class=p>,</span> <span class=nx>time</span><span class=p>.</span><span class=nf>Now</span><span class=p>().</span><span class=nf>Add</span><span class=p>(</span><span class=mi>10</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>IsTimerSet</span><span class=p>.</span><span class=nf>Write</span><span class=p>(</span><span class=nx>sp</span><span class=p>,</span> <span class=kc>true</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>bufferDoFn</span><span class=p>[</span><span class=nx>V</span><span class=p>])</span> <span class=nf>OnTimer</span><span class=p>(</span><span class=nx>sp</span> <span class=nx>state</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>tp</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Provider</span><span class=p>,</span> <span class=nx>w</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Window</span><span class=p>,</span> <span class=nx>key</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>timer</span> <span class=nx>timers</span><span class=p>.</span><span class=nx>Context</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>switch</span> <span class=nx>timer</span><span class=p>.</span><span class=nx>Family</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>case</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>OutputElements</span><span class=p>.</span><span class=nx>Family</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=nx>elements</span><span class=p>,</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>_</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>Elements</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nf>sendRpc</span><span class=p>(</span><span class=nx>elements</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>Elements</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>fn</span><span class=p>.</span><span class=nx>IsTimerSet</span><span class=p>.</span><span class=nf>Clear</span><span class=p>(</span><span class=nx>sp</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=nx>AddBufferDoFn</span><span class=p>[</span><span class=nx>V</span> <span class=nx>any</span><span class=p>](</span><span class=nx>s</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>Scope</span><span class=p>,</span> <span class=nx>in</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span><span class=p>)</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>PCollection</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>ParDo</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=o>&</span><span class=nx>bufferDoFn</span><span class=p>[</span><span class=nx>V</span><span class=p>]{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Elements</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeBagState</span><span class=p>[</span><span class=nx>V</span><span class=p>](</span><span class=s>"elements"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>IsTimerSet</span><span class=p>:</span> <span class=nx>state</span><span class=p>.</span><span class=nx>MakeValueState</span><span class=p>[</span><span class=kt>bool</span><span class=p>](</span><span class=s>"isTimerSet"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nx>OutputElements</span><span class=p>:</span> <span class=nx>timers</span><span class=p>.</span><span class=nf>InProcessingTime</span><span class=p>(</span><span class=s>"outputElements"</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> <span class=nx>in</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h2 id=splittable-dofns>12. Splittable <code>DoFns</code></h2><p>A Splittable <code>DoFn</code> (SDF) enables users to create modular components containing I/Os (and some advanced |
| <a href="https://s.apache.org/splittable-do-fn#heading=h.5cep9s8k4fxv">non I/O use cases</a>). Having modular |
| I/O components that can be connected to each other simplify typical patterns that users want. |
| For example, a popular use case is to read filenames from a message queue followed by parsing those |
| files. Traditionally, users were required to either write a single I/O connector that contained the |
| logic for the message queue and the file reader (increased complexity) or choose to reuse a message |
| queue I/O followed by a regular <code>DoFn</code> that read the file (decreased performance). With SDF, |
| we bring the richness of Apache Beam’s I/O APIs to a <code>DoFn</code> enabling modularity while maintaining the |
| performance of traditional I/O connectors.</p><h3 id=sdf-basics>12.1. SDF basics</h3><p>At a high level, an SDF is responsible for processing element and restriction pairs. A |
| restriction represents a subset of work that would have been necessary to have been done when |
| processing the element.</p><p>Executing an SDF follows the following steps:</p><ol><li>Each element is paired with a restriction (e.g. filename is paired with offset range representing the whole file).</li><li>Each element and restriction pair is split (e.g. offset ranges are broken up into smaller pieces).</li><li>The runner redistributes the element and restriction pairs to several workers.</li><li>Element and restriction pairs are processed in parallel (e.g. the file is read). Within this last step, |
| the element and restriction pair can pause its own processing and/or be split into further element and |
| restriction pairs.</li></ol><p><img src=/images/sdf_high_level_overview.svg alt="Diagram of steps that an SDF is composed of"></p><h4 id=a-basic-sdf>12.1.1. A basic SDF</h4><p>A basic SDF is composed of three parts: a restriction, a restriction provider, and a |
| restriction tracker. If you want to control the watermark, especially in a streaming |
| pipeline, two more components are needed: a watermark estimator provider and a watermark estimator.</p><p>The restriction is a user-defined object that is used to represent a subset of |
| work for a given element. For example, we defined <code>OffsetRange</code> as a restriction to represent offset |
| positions in <a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/io/range/OffsetRange.html>Java</a> |
| and <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.io.restriction_trackers.html#apache_beam.io.restriction_trackers.OffsetRange>Python</a>.</p><p>The restriction provider lets SDF authors override default implementations, including the ones for |
| splitting and sizing. In <a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/DoFn.ProcessElement.html>Java</a> |
| and <a href=https://github.com/apache/beam/blob/0f466e6bcd4ac8677c2bd9ecc8e6af3836b7f3b8/sdks/go/pkg/beam/pardo.go#L226>Go</a>, |
| this is the <code>DoFn</code>. <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.RestrictionProvider>Python</a> |
| has a dedicated <code>RestrictionProvider</code> type.</p><p>The restriction tracker is responsible for tracking which subset of the restriction has been |
| completed during processing. For APIs details, read the <a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/RestrictionTracker.html>Java</a> |
| and <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.io.iobase.html#apache_beam.io.iobase.RestrictionTracker>Python</a> |
| reference documentation.</p><p>There are some built-in <code>RestrictionTracker</code> implementations defined in Java:</p><ol><li><a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/OffsetRangeTracker.html>OffsetRangeTracker</a></li><li><a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/GrowableOffsetRangeTracker.html>GrowableOffsetRangeTracker</a></li><li><a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/ByteKeyRangeTracker.html>ByteKeyRangeTracker</a></li></ol><p>The SDF also has a built-in <code>RestrictionTracker</code> implementation in Python:</p><ol><li><a href=https://beam.apache.org/releases/pydoc/current/apache_beam.io.restriction_trackers.html#apache_beam.io.restriction_trackers.OffsetRestrictionTracker>OffsetRangeTracker</a></li></ol><p>Go also has a built-in <code>RestrictionTracker</code> type:</p><ol><li><a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam/io/rtrackers/offsetrange>OffsetRangeTracker</a></li></ol><p>The watermark state is a user-defined object which is used to create a <code>WatermarkEstimator</code> from a |
| <code>WatermarkEstimatorProvider</code>. The simplest watermark state could be a <code>timestamp</code>.</p><p>The watermark estimator provider lets SDF authors define how to initialize the watermark state and |
| create a watermark estimator. In <a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/DoFn.ProcessElement.html>Java</a> and <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam#ParDo>Go</a> |
| this is the <code>DoFn</code>. <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.WatermarkEstimatorProvider>Python</a> |
| has a dedicated <code>WatermarkEstimatorProvider</code> type.</p><p>The watermark estimator tracks the watermark when an element-restriction pair is in progress. |
| For APIs details, read the <a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/WatermarkEstimator.html>Java</a>, <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.io.iobase.html#apache_beam.io.iobase.WatermarkEstimator>Python</a>, and <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam/core/sdf#WatermarkEstimator>Go</a> |
| reference documentation.</p><p>There are some built-in <code>WatermarkEstimator</code> implementations in Java:</p><ol><li><a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/WatermarkEstimators.Manual.html>Manual</a></li><li><a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/WatermarkEstimators.MonotonicallyIncreasing.html>MonotonicallyIncreasing</a></li><li><a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/transforms/splittabledofn/WatermarkEstimators.WallTime.html>WallTime</a></li></ol><p>Along with the default <code>WatermarkEstimatorProvider</code>, there are the same set of built-in |
| <code>WatermarkEstimator</code> implementations in Python:</p><ol><li><a href=https://beam.apache.org/releases/pydoc/current/apache_beam.io.watermark_estimators.html#apache_beam.io.watermark_estimators.ManualWatermarkEstimator>ManualWatermarkEstimator</a></li><li><a href=https://beam.apache.org/releases/pydoc/current/apache_beam.io.watermark_estimators.html#apache_beam.io.watermark_estimators.MonotonicWatermarkEstimator>MonotonicWatermarkEstimator</a></li><li><a href=https://beam.apache.org/releases/pydoc/current/apache_beam.io.watermark_estimators.html#apache_beam.io.watermark_estimators.WalltimeWatermarkEstimator>WalltimeWatermarkEstimator</a></li></ol><p>The following <code>WatermarkEstimator</code> types are implemented in Go:</p><ol><li><a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam/core/sdf#TimestampObservingWatermarkEstimator>TimestampObservingEstimator</a></li><li><a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam/core/sdf#WallTimeWatermarkEstimator>WalltimeWatermarkEstimator</a></li></ol><p>To define an SDF, you must choose whether the SDF is bounded (default) or |
| unbounded and define a way to initialize an initial restriction for an element. The distinction is |
| based on how the amount of work is represented:</p><ul><li>Bounded DoFns are those where the work represented by an element is well-known beforehand and has |
| an end. Examples of bounded elements include a file or group of files.</li><li>Unbounded DoFns are those where the amount of work does not have a specific end or the |
| amount of work is not known beforehand. Examples of unbounded elements include a Kafka or a PubSub |
| topic.</li></ul><p>In Java, you can use <a href=https://beam.apache.org/releases/javadoc/current/index.html?org/apache/beam/sdk/transforms/DoFn.UnboundedPerElement.html>@UnboundedPerElement</a> |
| or <a href=https://beam.apache.org/releases/javadoc/current/index.html?org/apache/beam/sdk/transforms/DoFn.BoundedPerElement.html>@BoundedPerElement</a> |
| to annotate your <code>DoFn</code>. In Python, you can use <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.unbounded_per_element>@unbounded_per_element</a> |
| to annotate the <code>DoFn</code>.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@BoundedPerElement</span> |
| </span></span><span class=line><span class=cl><span class=kd>private</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>FileToWordsFn</span> <span class=kd>extends</span> <span class=n>DoFn</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Integer</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@GetInitialRestriction</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>OffsetRange</span> <span class=nf>getInitialRestriction</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>fileName</span><span class=o>)</span> <span class=kd>throws</span> <span class=n>IOException</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=k>new</span> <span class=n>OffsetRange</span><span class=o>(</span><span class=n>0</span><span class=o>,</span> <span class=k>new</span> <span class=n>File</span><span class=o>(</span><span class=n>fileName</span><span class=o>).</span><span class=na>length</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>String</span> <span class=n>fileName</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>RestrictionTracker</span><span class=o><</span><span class=n>OffsetRange</span><span class=o>,</span> <span class=n>Long</span><span class=o>></span> <span class=n>tracker</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>outputReceiver</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=kd>throws</span> <span class=n>IOException</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>RandomAccessFile</span> <span class=n>file</span> <span class=o>=</span> <span class=k>new</span> <span class=n>RandomAccessFile</span><span class=o>(</span><span class=n>fileName</span><span class=o>,</span> <span class=s>"r"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=n>seekToNextRecordBoundaryInFile</span><span class=o>(</span><span class=n>file</span><span class=o>,</span> <span class=n>tracker</span><span class=o>.</span><span class=na>currentRestriction</span><span class=o>().</span><span class=na>getFrom</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=o>(</span><span class=n>tracker</span><span class=o>.</span><span class=na>tryClaim</span><span class=o>(</span><span class=n>file</span><span class=o>.</span><span class=na>getFilePointer</span><span class=o>()))</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>outputReceiver</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>readNextRecord</span><span class=o>(</span><span class=n>file</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Providing the coder is only necessary if it can not be inferred at runtime. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@GetRestrictionCoder</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Coder</span><span class=o><</span><span class=n>OffsetRange</span><span class=o>></span> <span class=nf>getRestrictionCoder</span><span class=o>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>OffsetRange</span><span class=o>.</span><span class=na>Coder</span><span class=o>.</span><span class=na>of</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>FileToWordsRestrictionProvider</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>transforms</span><span class=o>.</span><span class=n>core</span><span class=o>.</span><span class=n>RestrictionProvider</span> |
| </span></span><span class=line><span class=cl> <span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>initial_restriction</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>file_name</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>OffsetRange</span><span class=p>(</span><span class=mi>0</span><span class=p>,</span> <span class=n>os</span><span class=o>.</span><span class=n>stat</span><span class=p>(</span><span class=n>file_name</span><span class=p>)</span><span class=o>.</span><span class=n>st_size</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>create_tracker</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>restriction</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>beam</span><span class=o>.</span><span class=n>io</span><span class=o>.</span><span class=n>restriction_trackers</span><span class=o>.</span><span class=n>OffsetRestrictionTracker</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>FileToWordsFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>file_name</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Alternatively, we can let FileToWordsFn itself inherit from</span> |
| </span></span><span class=line><span class=cl> <span class=c1># RestrictionProvider, implement the required methods and let</span> |
| </span></span><span class=line><span class=cl> <span class=c1># tracker=beam.DoFn.RestrictionParam() which will use self as</span> |
| </span></span><span class=line><span class=cl> <span class=c1># the provider.</span> |
| </span></span><span class=line><span class=cl> <span class=n>tracker</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>RestrictionParam</span><span class=p>(</span><span class=n>FileToWordsRestrictionProvider</span><span class=p>())):</span> |
| </span></span><span class=line><span class=cl> <span class=k>with</span> <span class=nb>open</span><span class=p>(</span><span class=n>file_name</span><span class=p>)</span> <span class=k>as</span> <span class=n>file_handle</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>file_handle</span><span class=o>.</span><span class=n>seek</span><span class=p>(</span><span class=n>tracker</span><span class=o>.</span><span class=n>current_restriction</span><span class=o>.</span><span class=n>start</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=n>tracker</span><span class=o>.</span><span class=n>try_claim</span><span class=p>(</span><span class=n>file_handle</span><span class=o>.</span><span class=n>tell</span><span class=p>()):</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>read_next_record</span><span class=p>(</span><span class=n>file_handle</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># Providing the coder is only necessary if it can not be inferred at</span> |
| </span></span><span class=line><span class=cl> <span class=c1># runtime.</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>restriction_coder</span><span class=p>(</span><span class=bp>self</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=o>...</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>splittableDoFn</span><span class=p>)</span> <span class=nf>CreateInitialRestriction</span><span class=p>(</span><span class=nx>filename</span> <span class=kt>string</span><span class=p>)</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Start</span><span class=p>:</span> <span class=mi>0</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>End</span><span class=p>:</span> <span class=nf>getFileLength</span><span class=p>(</span><span class=nx>filename</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>splittableDoFn</span><span class=p>)</span> <span class=nf>CreateTracker</span><span class=p>(</span><span class=nx>rest</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>)</span> <span class=o>*</span><span class=nx>sdf</span><span class=p>.</span><span class=nx>LockRTracker</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>sdf</span><span class=p>.</span><span class=nf>NewLockRTracker</span><span class=p>(</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nf>NewTracker</span><span class=p>(</span><span class=nx>rest</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>splittableDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>rt</span> <span class=o>*</span><span class=nx>sdf</span><span class=p>.</span><span class=nx>LockRTracker</span><span class=p>,</span> <span class=nx>filename</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>file</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>os</span><span class=p>.</span><span class=nf>Open</span><span class=p>(</span><span class=nx>filename</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>offset</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nf>seekToNextRecordBoundaryInFile</span><span class=p>(</span><span class=nx>file</span><span class=p>,</span> <span class=nx>rt</span><span class=p>.</span><span class=nf>GetRestriction</span><span class=p>().(</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>).</span><span class=nx>Start</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nx>rt</span><span class=p>.</span><span class=nf>TryClaim</span><span class=p>(</span><span class=nx>offset</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>record</span><span class=p>,</span> <span class=nx>newOffset</span> <span class=o>:=</span> <span class=nf>readNextRecord</span><span class=p>(</span><span class=nx>file</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nx>record</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>offset</span> <span class=p>=</span> <span class=nx>newOffset</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><p>At this point, we have an SDF that supports <a href=#runner-initiated-split>runner-initiated splits</a> |
| enabling dynamic work rebalancing. To increase the rate at which initial parallelization of work occurs |
| or for those runners that do not support runner-initiated splitting, we recommend providing |
| a set of initial splits:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kt>void</span> <span class=nf>splitRestriction</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Restriction</span> <span class=n>OffsetRange</span> <span class=n>restriction</span><span class=o>,</span> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>OffsetRange</span><span class=o>></span> <span class=n>splitReceiver</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kt>long</span> <span class=n>splitSize</span> <span class=o>=</span> <span class=n>64</span> <span class=o>*</span> <span class=o>(</span><span class=n>1</span> <span class=o><<</span> <span class=n>20</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=kt>long</span> <span class=n>i</span> <span class=o>=</span> <span class=n>restriction</span><span class=o>.</span><span class=na>getFrom</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=o>(</span><span class=n>i</span> <span class=o><</span> <span class=n>restriction</span><span class=o>.</span><span class=na>getTo</span><span class=o>()</span> <span class=o>-</span> <span class=n>splitSize</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Compute and output 64 MiB size ranges to process in parallel |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kt>long</span> <span class=n>end</span> <span class=o>=</span> <span class=n>i</span> <span class=o>+</span> <span class=n>splitSize</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=n>splitReceiver</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=k>new</span> <span class=n>OffsetRange</span><span class=o>(</span><span class=n>i</span><span class=o>,</span> <span class=n>end</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=n>i</span> <span class=o>=</span> <span class=n>end</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output the last range |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>splitReceiver</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=k>new</span> <span class=n>OffsetRange</span><span class=o>(</span><span class=n>i</span><span class=o>,</span> <span class=n>restriction</span><span class=o>.</span><span class=na>getTo</span><span class=o>()));</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>FileToWordsRestrictionProvider</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>transforms</span><span class=o>.</span><span class=n>core</span><span class=o>.</span><span class=n>RestrictionProvider</span> |
| </span></span><span class=line><span class=cl> <span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>split</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>file_name</span><span class=p>,</span> <span class=n>restriction</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Compute and output 64 MiB size ranges to process in parallel</span> |
| </span></span><span class=line><span class=cl> <span class=n>split_size</span> <span class=o>=</span> <span class=mi>64</span> <span class=o>*</span> <span class=p>(</span><span class=mi>1</span> <span class=o><<</span> <span class=mi>20</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>i</span> <span class=o>=</span> <span class=n>restriction</span><span class=o>.</span><span class=n>start</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=n>i</span> <span class=o><</span> <span class=n>restriction</span><span class=o>.</span><span class=n>end</span> <span class=o>-</span> <span class=n>split_size</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>OffsetRange</span><span class=p>(</span><span class=n>i</span><span class=p>,</span> <span class=n>i</span> <span class=o>+</span> <span class=n>split_size</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=n>i</span> <span class=o>+=</span> <span class=n>split_size</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>OffsetRange</span><span class=p>(</span><span class=n>i</span><span class=p>,</span> <span class=n>restriction</span><span class=o>.</span><span class=n>end</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>splittableDoFn</span><span class=p>)</span> <span class=nf>SplitRestriction</span><span class=p>(</span><span class=nx>filename</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>rest</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>)</span> <span class=p>(</span><span class=nx>splits</span> <span class=p>[]</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>size</span> <span class=o>:=</span> <span class=mi>64</span> <span class=o>*</span> <span class=p>(</span><span class=mi>1</span> <span class=o><<</span> <span class=mi>20</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>i</span> <span class=o>:=</span> <span class=nx>rest</span><span class=p>.</span><span class=nx>Start</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nx>i</span> <span class=p><</span> <span class=nx>rest</span><span class=p>.</span><span class=nx>End</span> <span class=o>-</span> <span class=nx>size</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Compute and output 64 MiB size ranges to process in parallel |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>end</span> <span class=o>:=</span> <span class=nx>i</span> <span class=o>+</span> <span class=nx>size</span> |
| </span></span><span class=line><span class=cl> <span class=nx>splits</span> <span class=p>=</span> <span class=nb>append</span><span class=p>(</span><span class=nx>splits</span><span class=p>,</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>{</span><span class=nx>i</span><span class=p>,</span> <span class=nx>end</span><span class=p>})</span> |
| </span></span><span class=line><span class=cl> <span class=nx>i</span> <span class=p>=</span> <span class=nx>end</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Output the last range |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>splits</span> <span class=p>=</span> <span class=nb>append</span><span class=p>(</span><span class=nx>splits</span><span class=p>,</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>{</span><span class=nx>i</span><span class=p>,</span> <span class=nx>rest</span><span class=p>.</span><span class=nx>End</span><span class=p>})</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>splits</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=sizing-and-progress>12.2. Sizing and progress</h3><p>Sizing and progress are used during execution of an SDF to inform runners so that they may |
| perform intelligent decisions about which restrictions to split and how to parallelize work.</p><p>Before processing an element and restriction, an initial size may be used by a runner to choose |
| how and who processes the restrictions attempting to improve initial balancing and parallelization |
| of work. During the processing of an element and restriction, sizing and progress are used to choose |
| which restrictions to split and who should process them.</p><p>By default, we use the restriction tracker’s estimate for work remaining falling back to assuming |
| that all restrictions have an equal cost. To override the default, SDF authors can provide the |
| appropriate method within the restriction provider. SDF authors need to be aware that the |
| sizing method will be invoked concurrently during bundle processing due to runner initiated splitting |
| and progress estimation.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@GetSize</span> |
| </span></span><span class=line><span class=cl><span class=kt>double</span> <span class=nf>getSize</span><span class=o>(</span><span class=nd>@Element</span> <span class=n>String</span> <span class=n>fileName</span><span class=o>,</span> <span class=nd>@Restriction</span> <span class=n>OffsetRange</span> <span class=n>restriction</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=o>(</span><span class=n>fileName</span><span class=o>.</span><span class=na>contains</span><span class=o>(</span><span class=s>"expensiveRecords"</span><span class=o>)</span> <span class=o>?</span> <span class=n>2</span> <span class=o>:</span> <span class=n>1</span><span class=o>)</span> <span class=o>*</span> <span class=n>restriction</span><span class=o>.</span><span class=na>getTo</span><span class=o>()</span> |
| </span></span><span class=line><span class=cl> <span class=o>-</span> <span class=n>restriction</span><span class=o>.</span><span class=na>getFrom</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># The RestrictionProvider is responsible for calculating the size of given</span> |
| </span></span><span class=line><span class=cl><span class=c1># restriction.</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MyRestrictionProvider</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>transforms</span><span class=o>.</span><span class=n>core</span><span class=o>.</span><span class=n>RestrictionProvider</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>restriction_size</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>file_name</span><span class=p>,</span> <span class=n>restriction</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>weight</span> <span class=o>=</span> <span class=mi>2</span> <span class=k>if</span> <span class=s2>"expensiveRecords"</span> <span class=ow>in</span> <span class=n>file_name</span> <span class=k>else</span> <span class=mi>1</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>restriction</span><span class=o>.</span><span class=n>size</span><span class=p>()</span> <span class=o>*</span> <span class=n>weight</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>splittableDoFn</span><span class=p>)</span> <span class=nf>RestrictionSize</span><span class=p>(</span><span class=nx>filename</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>rest</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>)</span> <span class=kt>float64</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>weight</span> <span class=o>:=</span> <span class=nb>float64</span><span class=p>(</span><span class=mi>1</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>strings</span><span class=p>.</span><span class=nf>Contains</span><span class=p>(</span><span class=nx>filename</span><span class=p>,</span> <span class=err>“</span><span class=nx>expensiveRecords</span><span class=err>”</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>weight</span> <span class=p>=</span> <span class=mi>2</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>weight</span> <span class=o>*</span> <span class=p>(</span><span class=nx>rest</span><span class=p>.</span><span class=nx>End</span> <span class=o>-</span> <span class=nx>rest</span><span class=p>.</span><span class=nx>Start</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=user-initiated-checkpoint>12.3. User-initiated checkpoint</h3><p>Some I/Os cannot produce all of the data necessary to complete a restriction within the lifetime of a |
| single bundle. This typically happens with unbounded restrictions, but can also happen with bounded |
| restrictions. For example, there could be more data that needs to be ingested but is not available yet. |
| Another cause of this scenario is the source system throttling your data.</p><p>Your SDF can signal to you that you are not done processing the current restriction. This |
| signal can suggest a time to resume at. While the runner tries to honor the resume time, this is not |
| guaranteed. This allows execution to continue on a restriction that has available work improving |
| resource utilization.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=n>ProcessContinuation</span> <span class=nf>processElement</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>RestrictionTracker</span><span class=o><</span><span class=n>OffsetRange</span><span class=o>,</span> <span class=n>Long</span><span class=o>></span> <span class=n>tracker</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>RecordPosition</span><span class=o>></span> <span class=n>outputReceiver</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kt>long</span> <span class=n>currentPosition</span> <span class=o>=</span> <span class=n>tracker</span><span class=o>.</span><span class=na>currentRestriction</span><span class=o>().</span><span class=na>getFrom</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=n>Service</span> <span class=n>service</span> <span class=o>=</span> <span class=n>initializeService</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=k>try</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=o>(</span><span class=kc>true</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>List</span><span class=o><</span><span class=n>RecordPosition</span><span class=o>></span> <span class=n>records</span> <span class=o>=</span> <span class=n>service</span><span class=o>.</span><span class=na>readNextRecords</span><span class=o>(</span><span class=n>currentPosition</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(</span><span class=n>records</span><span class=o>.</span><span class=na>isEmpty</span><span class=o>())</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Return a short delay if there is no data to process at the moment. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=n>ProcessContinuation</span><span class=o>.</span><span class=na>resume</span><span class=o>().</span><span class=na>withResumeDelay</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>10</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=o>(</span><span class=n>RecordPosition</span> <span class=n>record</span> <span class=o>:</span> <span class=n>records</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(!</span><span class=n>tracker</span><span class=o>.</span><span class=na>tryClaim</span><span class=o>(</span><span class=n>record</span><span class=o>.</span><span class=na>getPosition</span><span class=o>()))</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>ProcessContinuation</span><span class=o>.</span><span class=na>stop</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=n>currentPosition</span> <span class=o>=</span> <span class=n>record</span><span class=o>.</span><span class=na>getPosition</span><span class=o>()</span> <span class=o>+</span> <span class=n>1</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=n>outputReceiver</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>record</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> <span class=k>catch</span> <span class=o>(</span><span class=n>ThrottlingException</span> <span class=n>exception</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Return a longer delay in case we are being throttled. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=n>ProcessContinuation</span><span class=o>.</span><span class=na>resume</span><span class=o>().</span><span class=na>withResumeDelay</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardSeconds</span><span class=o>(</span><span class=n>60</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>MySplittableDoFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>restriction_tracker</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>RestrictionParam</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>MyRestrictionProvider</span><span class=p>())):</span> |
| </span></span><span class=line><span class=cl> <span class=n>current_position</span> <span class=o>=</span> <span class=n>restriction_tracker</span><span class=o>.</span><span class=n>current_restriction</span><span class=o>.</span><span class=n>start</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=kc>True</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Pull records from an external service.</span> |
| </span></span><span class=line><span class=cl> <span class=k>try</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>records</span> <span class=o>=</span> <span class=n>external_service</span><span class=o>.</span><span class=n>fetch</span><span class=p>(</span><span class=n>current_position</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>records</span><span class=o>.</span><span class=n>empty</span><span class=p>():</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Set a shorter delay in case we are being throttled.</span> |
| </span></span><span class=line><span class=cl> <span class=n>restriction_tracker</span><span class=o>.</span><span class=n>defer_remainder</span><span class=p>(</span><span class=n>timestamp</span><span class=o>.</span><span class=n>Duration</span><span class=p>(</span><span class=n>second</span><span class=o>=</span><span class=mi>10</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=n>record</span> <span class=ow>in</span> <span class=n>records</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=n>restriction_tracker</span><span class=o>.</span><span class=n>try_claim</span><span class=p>(</span><span class=n>record</span><span class=o>.</span><span class=n>position</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>current_position</span> <span class=o>=</span> <span class=n>record</span><span class=o>.</span><span class=n>position</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>record</span> |
| </span></span><span class=line><span class=cl> <span class=k>else</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> |
| </span></span><span class=line><span class=cl> <span class=k>except</span> <span class=ne>TimeoutError</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Set a longer delay in case we are being throttled.</span> |
| </span></span><span class=line><span class=cl> <span class=n>restriction_tracker</span><span class=o>.</span><span class=n>defer_remainder</span><span class=p>(</span><span class=n>timestamp</span><span class=o>.</span><span class=n>Duration</span><span class=p>(</span><span class=n>seconds</span><span class=o>=</span><span class=mi>60</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>checkpointingSplittableDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>rt</span> <span class=o>*</span><span class=nx>sdf</span><span class=p>.</span><span class=nx>LockRTracker</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=nx>Record</span><span class=p>))</span> <span class=p>(</span><span class=nx>sdf</span><span class=p>.</span><span class=nx>ProcessContinuation</span><span class=p>,</span> <span class=kt>error</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>position</span> <span class=o>:=</span> <span class=nx>rt</span><span class=p>.</span><span class=nf>GetRestriction</span><span class=p>().(</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>).</span><span class=nx>Start</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>records</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>ExternalService</span><span class=p>.</span><span class=nf>readNextRecords</span><span class=p>(</span><span class=nx>position</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>==</span> <span class=nx>fn</span><span class=p>.</span><span class=nx>ExternalService</span><span class=p>.</span><span class=nx>ThrottlingErr</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Resume at a later time to avoid throttling. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>sdf</span><span class=p>.</span><span class=nf>ResumeProcessingIn</span><span class=p>(</span><span class=mi>60</span> <span class=o>*</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>),</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>sdf</span><span class=p>.</span><span class=nf>StopProcessing</span><span class=p>(),</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nb>len</span><span class=p>(</span><span class=nx>records</span><span class=p>)</span> <span class=o>==</span> <span class=mi>0</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Wait for data to be available. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>sdf</span><span class=p>.</span><span class=nf>ResumeProcessingIn</span><span class=p>(</span><span class=mi>10</span> <span class=o>*</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Second</span><span class=p>),</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nx>_</span><span class=p>,</span> <span class=nx>record</span> <span class=o>:=</span> <span class=k>range</span> <span class=nx>records</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=p>!</span><span class=nx>rt</span><span class=p>.</span><span class=nf>TryClaim</span><span class=p>(</span><span class=nx>position</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Records have been claimed, finish processing. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>sdf</span><span class=p>.</span><span class=nf>StopProcessing</span><span class=p>(),</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>position</span> <span class=o>+=</span> <span class=mi>1</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nx>record</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=runner-initiated-split>12.4. Runner-initiated split</h3><p>A runner at any time may attempt to split a restriction while it is being processed. This allows the |
| runner to either pause processing of the restriction so that other work may be done (common for |
| unbounded restrictions to limit the amount of output and/or improve latency) or split the restriction |
| into two pieces, increasing the available parallelism within the system. Different runners (e.g., |
| Dataflow, Flink, Spark) have different strategies to issue splits under batch and streaming |
| execution.</p><p>Author an SDF with this in mind since the end of the restriction may change. When writing the |
| processing loop, use the result from trying to claim a piece of the restriction instead of assuming |
| you can process until the end.</p><p>One incorrect example could be:</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kt>void</span> <span class=nf>badTryClaimLoop</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>String</span> <span class=n>fileName</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>RestrictionTracker</span><span class=o><</span><span class=n>OffsetRange</span><span class=o>,</span> <span class=n>Long</span><span class=o>></span> <span class=n>tracker</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>OutputReceiver</span><span class=o><</span><span class=n>Integer</span><span class=o>></span> <span class=n>outputReceiver</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=kd>throws</span> <span class=n>IOException</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>RandomAccessFile</span> <span class=n>file</span> <span class=o>=</span> <span class=k>new</span> <span class=n>RandomAccessFile</span><span class=o>(</span><span class=n>fileName</span><span class=o>,</span> <span class=s>"r"</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=n>seekToNextRecordBoundaryInFile</span><span class=o>(</span><span class=n>file</span><span class=o>,</span> <span class=n>tracker</span><span class=o>.</span><span class=na>currentRestriction</span><span class=o>().</span><span class=na>getFrom</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// The restriction tracker can be modified by another thread in parallel |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// so storing state locally is ill advised. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kt>long</span> <span class=n>end</span> <span class=o>=</span> <span class=n>tracker</span><span class=o>.</span><span class=na>currentRestriction</span><span class=o>().</span><span class=na>getTo</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=o>(</span><span class=n>file</span><span class=o>.</span><span class=na>getFilePointer</span><span class=o>()</span> <span class=o><</span> <span class=n>end</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Only after successfully claiming should we produce any output and/or |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// perform side effects. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=n>tracker</span><span class=o>.</span><span class=na>tryClaim</span><span class=o>(</span><span class=n>file</span><span class=o>.</span><span class=na>getFilePointer</span><span class=o>());</span> |
| </span></span><span class=line><span class=cl> <span class=n>outputReceiver</span><span class=o>.</span><span class=na>output</span><span class=o>(</span><span class=n>readNextRecord</span><span class=o>(</span><span class=n>file</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>BadTryClaimLoop</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>file_name</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>tracker</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>RestrictionParam</span><span class=p>(</span><span class=n>FileToWordsRestrictionProvider</span><span class=p>())):</span> |
| </span></span><span class=line><span class=cl> <span class=k>with</span> <span class=nb>open</span><span class=p>(</span><span class=n>file_name</span><span class=p>)</span> <span class=k>as</span> <span class=n>file_handle</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>file_handle</span><span class=o>.</span><span class=n>seek</span><span class=p>(</span><span class=n>tracker</span><span class=o>.</span><span class=n>current_restriction</span><span class=o>.</span><span class=n>start</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=c1># The restriction tracker can be modified by another thread in parallel</span> |
| </span></span><span class=line><span class=cl> <span class=c1># so storing state locally is ill advised.</span> |
| </span></span><span class=line><span class=cl> <span class=n>end</span> <span class=o>=</span> <span class=n>tracker</span><span class=o>.</span><span class=n>current_restriction</span><span class=o>.</span><span class=n>end</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl> <span class=k>while</span> <span class=n>file_handle</span><span class=o>.</span><span class=n>tell</span><span class=p>()</span> <span class=o><</span> <span class=n>end</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Only after successfully claiming should we produce any output and/or</span> |
| </span></span><span class=line><span class=cl> <span class=c1># perform side effects.</span> |
| </span></span><span class=line><span class=cl> <span class=n>tracker</span><span class=o>.</span><span class=n>try_claim</span><span class=p>(</span><span class=n>file_handle</span><span class=o>.</span><span class=n>tell</span><span class=p>())</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>read_next_record</span><span class=p>(</span><span class=n>file_handle</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>badTryClaimLoop</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>rt</span> <span class=o>*</span><span class=nx>sdf</span><span class=p>.</span><span class=nx>LockRTracker</span><span class=p>,</span> <span class=nx>filename</span> <span class=kt>string</span><span class=p>,</span> <span class=nx>emit</span> <span class=kd>func</span><span class=p>(</span><span class=kt>int</span><span class=p>))</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>file</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nx>os</span><span class=p>.</span><span class=nf>Open</span><span class=p>(</span><span class=nx>filename</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=nx>offset</span><span class=p>,</span> <span class=nx>err</span> <span class=o>:=</span> <span class=nf>seekToNextRecordBoundaryInFile</span><span class=p>(</span><span class=nx>file</span><span class=p>,</span> <span class=nx>rt</span><span class=p>.</span><span class=nf>GetRestriction</span><span class=p>().(</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>).</span><span class=nx>Start</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=nx>err</span> <span class=o>!=</span> <span class=kc>nil</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>err</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// The restriction tracker can be modified by another thread in parallel |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// so storing state locally is ill advised. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>end</span> <span class=p>=</span> <span class=nx>rt</span><span class=p>.</span><span class=nf>GetRestriction</span><span class=p>().(</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>).</span><span class=nx>End</span> |
| </span></span><span class=line><span class=cl> <span class=k>for</span> <span class=nx>offset</span> <span class=p><</span> <span class=nx>end</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Only after successfully claiming should we produce any output and/or |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// perform side effects. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>rt</span><span class=p>.</span><span class=nf>TryClaim</span><span class=p>(</span><span class=nx>offset</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>record</span><span class=p>,</span> <span class=nx>newOffset</span> <span class=o>:=</span> <span class=nf>readNextRecord</span><span class=p>(</span><span class=nx>file</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nf>emit</span><span class=p>(</span><span class=nx>record</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=nx>offset</span> <span class=p>=</span> <span class=nx>newOffset</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=watermark-estimation>12.5. Watermark estimation</h3><p>The default watermark estimator does not produce a watermark estimate. Therefore, the output watermark |
| is solely computed by the minimum of upstream watermarks.</p><p>An SDF can advance the output watermark by specifying a lower bound for all future output |
| that this element and restriction pair will produce. The runner computes the minimum output watermark |
| by taking the minimum over all upstream watermarks and the minimum reported by each element and |
| restriction pair. The reported watermark must monotonically increase for each element and restriction |
| pair across bundle boundaries. When an element and restriction pair stops processing its watermark, |
| it is no longer considered part of the above calculation.</p><p>Tips:</p><ul><li>If you author an SDF that outputs records with timestamps, you should expose ways to allow users of |
| this SDF to configure which watermark estimator to use.</li><li>Any data produced before the watermark may be considered late. See |
| <a href=#watermarks-and-late-data>watermarks and late data</a> for more details.</li></ul><h4 id=controlling-the-watermark>12.5.1. Controlling the watermark</h4><p>There are two general types of watermark estimators: timestamp observing and external clock observing. |
| Timestamp observing watermark estimators use the output timestamp of each record to compute the watermark |
| estimate while external clock observing watermark estimators control the watermark by using a clock that |
| is not associated to any individual output, such as the local clock of the machine or a clock exposed |
| through an external service.</p><p>The watermark estimator provider lets you override the default watermark estimation logic and use an existing |
| watermark estimator implementation. You can also provide your own watermark estimator implementation.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl> <span class=c1>// (Optional) Define a custom watermark state type to save information between bundle |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// processing rounds. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>MyCustomWatermarkState</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=nf>MyCustomWatermarkState</span><span class=o>(</span><span class=n>String</span> <span class=n>element</span><span class=o>,</span> <span class=n>OffsetRange</span> <span class=n>restriction</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Store data necessary for future watermark computations |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// (Optional) Choose which coder to use to encode the watermark estimator state. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nd>@GetWatermarkEstimatorStateCoder</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Coder</span><span class=o><</span><span class=n>MyCustomWatermarkState</span><span class=o>></span> <span class=nf>getWatermarkEstimatorStateCoder</span><span class=o>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>AvroCoder</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>MyCustomWatermarkState</span><span class=o>.</span><span class=na>class</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Define a WatermarkEstimator |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>MyCustomWatermarkEstimator</span> |
| </span></span><span class=line><span class=cl> <span class=kd>implements</span> <span class=n>TimestampObservingWatermarkEstimator</span><span class=o><</span><span class=n>MyCustomWatermarkState</span><span class=o>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=nf>MyCustomWatermarkEstimator</span><span class=o>(</span><span class=n>MyCustomWatermarkState</span> <span class=n>type</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Initialize watermark estimator state |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>observeTimestamp</span><span class=o>(</span><span class=n>Instant</span> <span class=n>timestamp</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Will be invoked on each output from the SDF |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Instant</span> <span class=nf>currentWatermark</span><span class=o>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Return a monotonically increasing value |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=n>currentWatermark</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>MyCustomWatermarkState</span> <span class=nf>getState</span><span class=o>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Return state to resume future watermark estimation after a checkpoint/split |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=kc>null</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// Then, update the DoFn to generate the initial watermark estimator state for all new element |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// and restriction pairs and to create a new instance given watermark estimator state. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=nd>@GetInitialWatermarkEstimatorState</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>MyCustomWatermarkState</span> <span class=nf>getInitialWatermarkEstimatorState</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>String</span> <span class=n>element</span><span class=o>,</span> <span class=nd>@Restriction</span> <span class=n>OffsetRange</span> <span class=n>restriction</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Compute and return the initial watermark estimator state for each element and |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// restriction. All subsequent processing of an element and restriction will be restored |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// from the existing state. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=k>new</span> <span class=n>MyCustomWatermarkState</span><span class=o>(</span><span class=n>element</span><span class=o>,</span> <span class=n>restriction</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@NewWatermarkEstimator</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>WatermarkEstimator</span><span class=o><</span><span class=n>MyCustomWatermarkState</span><span class=o>></span> <span class=nf>newWatermarkEstimator</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@WatermarkEstimatorState</span> <span class=n>MyCustomWatermarkState</span> <span class=n>oldState</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=k>new</span> <span class=n>MyCustomWatermarkEstimator</span><span class=o>(</span><span class=n>oldState</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># (Optional) Define a custom watermark state type to save information between</span> |
| </span></span><span class=line><span class=cl><span class=c1># bundle processing rounds.</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MyCustomerWatermarkEstimatorState</span><span class=p>(</span><span class=nb>object</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=fm>__init__</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>restriction</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Store data necessary for future watermark computations</span> |
| </span></span><span class=line><span class=cl> <span class=k>pass</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Define a WatermarkEstimator</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MyCustomWatermarkEstimator</span><span class=p>(</span><span class=n>WatermarkEstimator</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=fm>__init__</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>estimator_state</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=bp>self</span><span class=o>.</span><span class=n>state</span> <span class=o>=</span> <span class=n>estimator_state</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>observe_timestamp</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>timestamp</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Will be invoked on each output from the SDF</span> |
| </span></span><span class=line><span class=cl> <span class=k>pass</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>current_watermark</span><span class=p>(</span><span class=bp>self</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Return a monotonically increasing value</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>current_watermark</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>get_estimator_state</span><span class=p>(</span><span class=bp>self</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Return state to resume future watermark estimation after a</span> |
| </span></span><span class=line><span class=cl> <span class=c1># checkpoint/split</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=bp>self</span><span class=o>.</span><span class=n>state</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Then, a WatermarkEstimatorProvider needs to be created for this</span> |
| </span></span><span class=line><span class=cl><span class=c1># WatermarkEstimator</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MyWatermarkEstimatorProvider</span><span class=p>(</span><span class=n>WatermarkEstimatorProvider</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>initial_estimator_state</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>restriction</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>MyCustomerWatermarkEstimatorState</span><span class=p>(</span><span class=n>element</span><span class=p>,</span> <span class=n>restriction</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>create_watermark_estimator</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>estimator_state</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>MyCustomWatermarkEstimator</span><span class=p>(</span><span class=n>estimator_state</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Finally, define the SDF using your estimator.</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MySplittableDoFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>element</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>restriction_tracker</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>RestrictionParam</span><span class=p>(</span><span class=n>MyRestrictionProvider</span><span class=p>()),</span> |
| </span></span><span class=line><span class=cl> <span class=n>watermark_estimator</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>WatermarkEstimatorParam</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>MyWatermarkEstimatorProvider</span><span class=p>())):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># The current watermark can be inspected.</span> |
| </span></span><span class=line><span class=cl> <span class=n>watermark_estimator</span><span class=o>.</span><span class=n>current_watermark</span><span class=p>()</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// WatermarkState is a custom type.` |
| </span></span></span><span class=line><span class=cl><span class=c1>// |
| </span></span></span><span class=line><span class=cl><span class=c1>// It is optional to write your own state type when making a custom estimator. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>WatermarkState</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Watermark</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Time</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// CustomWatermarkEstimator is a custom watermark estimator. |
| </span></span></span><span class=line><span class=cl><span class=c1>// You may use any type here, including some of Beam's built in watermark estimator types, |
| </span></span></span><span class=line><span class=cl><span class=c1>// e.g. sdf.WallTimeWatermarkEstimator, sdf.TimestampObservingWatermarkEstimator, and sdf.ManualWatermarkEstimator |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>type</span> <span class=nx>CustomWatermarkEstimator</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>state</span> <span class=nx>WatermarkState</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// CurrentWatermark returns the current watermark and is invoked on DoFn splits and self-checkpoints. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Watermark estimators must implement CurrentWatermark() time.Time |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>e</span> <span class=o>*</span><span class=nx>CustomWatermarkEstimator</span><span class=p>)</span> <span class=nf>CurrentWatermark</span><span class=p>()</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Time</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>e</span><span class=p>.</span><span class=nx>state</span><span class=p>.</span><span class=nx>Watermark</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// ObserveTimestamp is called on the output timestamps of all |
| </span></span></span><span class=line><span class=cl><span class=c1>// emitted elements to update the watermark. It is optional |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>e</span> <span class=o>*</span><span class=nx>CustomWatermarkEstimator</span><span class=p>)</span> <span class=nf>ObserveTimestamp</span><span class=p>(</span><span class=nx>ts</span> <span class=nx>time</span><span class=p>.</span><span class=nx>Time</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>e</span><span class=p>.</span><span class=nx>state</span><span class=p>.</span><span class=nx>Watermark</span> <span class=p>=</span> <span class=nx>ts</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// InitialWatermarkEstimatorState defines an initial state used to initialize the watermark |
| </span></span></span><span class=line><span class=cl><span class=c1>// estimator. It is optional. If this is not defined, WatermarkEstimatorState may not be |
| </span></span></span><span class=line><span class=cl><span class=c1>// defined and CreateWatermarkEstimator must not take in parameters. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>weDoFn</span><span class=p>)</span> <span class=nf>InitialWatermarkEstimatorState</span><span class=p>(</span><span class=nx>et</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>EventTime</span><span class=p>,</span> <span class=nx>rest</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>,</span> <span class=nx>element</span> <span class=kt>string</span><span class=p>)</span> <span class=nx>WatermarkState</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Return some watermark state |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=nx>WatermarkState</span><span class=p>{</span><span class=nx>Watermark</span><span class=p>:</span> <span class=nx>time</span><span class=p>.</span><span class=nf>Now</span><span class=p>()}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// CreateWatermarkEstimator creates the watermark estimator used by this Splittable DoFn. |
| </span></span></span><span class=line><span class=cl><span class=c1>// Must take in a state parameter if InitialWatermarkEstimatorState is defined, otherwise takes no parameters. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>weDoFn</span><span class=p>)</span> <span class=nf>CreateWatermarkEstimator</span><span class=p>(</span><span class=nx>initialState</span> <span class=nx>WatermarkState</span><span class=p>)</span> <span class=o>*</span><span class=nx>CustomWatermarkEstimator</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=o>&</span><span class=nx>CustomWatermarkEstimator</span><span class=p>{</span><span class=nx>state</span><span class=p>:</span> <span class=nx>initialState</span><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// WatermarkEstimatorState returns the state used to resume future watermark estimation |
| </span></span></span><span class=line><span class=cl><span class=c1>// after a checkpoint/split. It is required if InitialWatermarkEstimatorState is defined, |
| </span></span></span><span class=line><span class=cl><span class=c1>// otherwise it must not be defined. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>weDoFn</span><span class=p>)</span> <span class=nf>WatermarkEstimatorState</span><span class=p>(</span><span class=nx>e</span> <span class=o>*</span><span class=nx>CustomWatermarkEstimator</span><span class=p>)</span> <span class=nx>WatermarkState</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>e</span><span class=p>.</span><span class=nx>state</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// ProcessElement is the method to execute for each element. |
| </span></span></span><span class=line><span class=cl><span class=c1>// It can optionally take in a watermark estimator. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>weDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>e</span> <span class=o>*</span><span class=nx>CustomWatermarkEstimator</span><span class=p>,</span> <span class=nx>element</span> <span class=kt>string</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>e</span><span class=p>.</span><span class=nx>state</span><span class=p>.</span><span class=nx>Watermark</span> <span class=p>=</span> <span class=nx>time</span><span class=p>.</span><span class=nf>Now</span><span class=p>()</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=truncating-during-drain>12.6. Truncating during drain</h3><p>Runners which support draining pipelines need the ability to drain SDFs; otherwise, the |
| pipeline may never stop. By default, bounded restrictions process the remainder of the restriction while |
| unbounded restrictions finish processing at the next SDF-initiated checkpoint or runner-initiated split. |
| You are able to override this default behavior by defining the appropriate method on the restriction |
| provider.</p><p class=language-go>Note: Once the pipeline drain starts and truncate restriction transform is triggered, the <code>sdf.ProcessContinuation</code> |
| will not be rescheduled.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@TruncateRestriction</span> |
| </span></span><span class=line><span class=cl><span class=nd>@Nullable</span> |
| </span></span><span class=line><span class=cl><span class=n>TruncateResult</span><span class=o><</span><span class=n>OffsetRange</span><span class=o>></span> <span class=nf>truncateRestriction</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@Element</span> <span class=n>String</span> <span class=n>fileName</span><span class=o>,</span> <span class=nd>@Restriction</span> <span class=n>OffsetRange</span> <span class=n>restriction</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=o>(</span><span class=n>fileName</span><span class=o>.</span><span class=na>contains</span><span class=o>(</span><span class=s>"optional"</span><span class=o>))</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// Skip optional files |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=k>return</span> <span class=kc>null</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>TruncateResult</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=n>restriction</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>MyRestrictionProvider</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>transforms</span><span class=o>.</span><span class=n>core</span><span class=o>.</span><span class=n>RestrictionProvider</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>truncate</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>file_name</span><span class=p>,</span> <span class=n>restriction</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>if</span> <span class=s2>"optional"</span> <span class=ow>in</span> <span class=n>file_name</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Skip optional files</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>None</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>restriction</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=c1>// TruncateRestriction is a transform that is triggered when pipeline starts to drain. It helps to finish a |
| </span></span></span><span class=line><span class=cl><span class=c1>// pipeline quicker by truncating the restriction. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>splittableDoFn</span><span class=p>)</span> <span class=nf>TruncateRestriction</span><span class=p>(</span><span class=nx>rt</span> <span class=o>*</span><span class=nx>sdf</span><span class=p>.</span><span class=nx>LockRTracker</span><span class=p>,</span> <span class=nx>element</span> <span class=kt>string</span><span class=p>)</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>start</span> <span class=o>:=</span> <span class=nx>rt</span><span class=p>.</span><span class=nf>GetRestriction</span><span class=p>().(</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>).</span><span class=nx>Start</span> |
| </span></span><span class=line><span class=cl> <span class=nx>prevEnd</span> <span class=o>:=</span> <span class=nx>rt</span><span class=p>.</span><span class=nf>GetRestriction</span><span class=p>().(</span><span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>).</span><span class=nx>End</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// truncate the restriction by half. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>newEnd</span> <span class=o>:=</span> <span class=nx>prevEnd</span> <span class=o>/</span> <span class=mi>2</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=nx>offsetrange</span><span class=p>.</span><span class=nx>Restriction</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Start</span><span class=p>:</span> <span class=nx>start</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>End</span><span class=p>:</span> <span class=nx>newEnd</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h3 id=bundle-finalization>12.7. Bundle finalization</h3><p>Bundle finalization enables a <code>DoFn</code> to perform side effects by registering a callback. |
| The callback is invoked once the runner has acknowledged that it has durably persisted the output. |
| For example, a message queue might need to acknowledge messages that it has ingested into the pipeline. |
| Bundle finalization is not limited to SDFs but is called out here since this is the primary |
| use case.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@ProcessElement</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kt>void</span> <span class=nf>processElement</span><span class=o>(</span><span class=n>ProcessContext</span> <span class=n>c</span><span class=o>,</span> <span class=n>BundleFinalizer</span> <span class=n>bundleFinalizer</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... produce output ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=n>bundleFinalizer</span><span class=o>.</span><span class=na>afterBundleCommit</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>Instant</span><span class=o>.</span><span class=na>now</span><span class=o>().</span><span class=na>plus</span><span class=o>(</span><span class=n>Duration</span><span class=o>.</span><span class=na>standardMinutes</span><span class=o>(</span><span class=n>5</span><span class=o>)),</span> |
| </span></span><span class=line><span class=cl> <span class=o>()</span> <span class=o>-></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... perform a side effect ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=o>});</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>MySplittableDoFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>,</span> <span class=n>bundle_finalizer</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>BundleFinalizerParam</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># ... produce output ...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># Register callback function for this bundle that performs the side</span> |
| </span></span><span class=line><span class=cl> <span class=c1># effect.</span> |
| </span></span><span class=line><span class=cl> <span class=n>bundle_finalizer</span><span class=o>.</span><span class=n>register</span><span class=p>(</span><span class=n>my_callback_func</span><span class=p>)</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>func</span> <span class=p>(</span><span class=nx>fn</span> <span class=o>*</span><span class=nx>splittableDoFn</span><span class=p>)</span> <span class=nf>ProcessElement</span><span class=p>(</span><span class=nx>bf</span> <span class=nx>beam</span><span class=p>.</span><span class=nx>BundleFinalization</span><span class=p>,</span> <span class=nx>rt</span> <span class=o>*</span><span class=nx>sdf</span><span class=p>.</span><span class=nx>LockRTracker</span><span class=p>,</span> <span class=nx>element</span> <span class=kt>string</span><span class=p>)</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... produce output ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=nx>bf</span><span class=p>.</span><span class=nf>RegisterCallback</span><span class=p>(</span><span class=mi>5</span><span class=o>*</span><span class=nx>time</span><span class=p>.</span><span class=nx>Minute</span><span class=p>,</span> <span class=kd>func</span><span class=p>()</span> <span class=kt>error</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=c1>// ... perform a side effect ... |
| </span></span></span><span class=line><span class=cl><span class=c1></span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=kc>nil</span> |
| </span></span><span class=line><span class=cl> <span class=p>})</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span></span></span></code></pre></div></div></div><h2 id=multi-language-pipelines>13. Multi-language pipelines</h2><p>This section provides comprehensive documentation of multi-language pipelines. To get started creating a multi-language pipeline, see:</p><ul><li><a href=/documentation/sdks/python-multi-language-pipelines>Python multi-language pipelines quickstart</a></li><li><a href=/documentation/sdks/java-multi-language-pipelines>Java multi-language pipelines quickstart</a></li></ul><p>Beam lets you combine transforms written in any supported SDK language (currently, Java and Python) and use them in one multi-language pipeline. This capability makes it easy to provide new functionality simultaneously in different Apache Beam SDKs through a single cross-language transform. For example, the <a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/io/kafka.py>Apache Kafka connector</a> and <a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/transforms/sql.py>SQL transform</a> from the Java SDK can be used in Python pipelines.</p><p>Pipelines that use transforms from more than one SDK-language are known as <em>multi-language pipelines</em>.</p><p class=language-yaml>Beam YAML is built entirely on top of cross-language transforms. |
| In addition to the built in transforms, you can author your own transforms |
| (using the full expressivity of the Beam API) and surface them via a concept |
| called <a href=https://beam.apache.org/documentation/sdks/yaml/#providers>providers</a>.</p><h3 id=create-x-lang-transforms>13.1. Creating cross-language transforms</h3><p>To make transforms written in one language available to pipelines written in another language, Beam uses an <em>expansion service</em>, which creates and injects the appropriate language-specific pipeline fragments into the pipeline.</p><p>In the following example, a Beam Python pipeline starts up a local Java expansion service to create and inject the appropriate Java pipeline fragments for executing the Java Kafka cross-language transform into the Python pipeline. The SDK then downloads and stages the necessary Java dependencies needed to execute these transforms.</p><p><img src=/images/multi-language-pipelines-diagram.svg alt="Diagram of multi-language pipeline execution flow."></p><p>At runtime, the Beam runner will execute both Python and Java transforms to run the pipeline.</p><p>In this section, we will use <a href=https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/io/kafka/KafkaIO.Read.html>KafkaIO.Read</a> to illustrate how to create a cross-language transform for Java and a test example for Python.</p><h4 id=1311-creating-cross-language-java-transforms>13.1.1. Creating cross-language Java transforms</h4><p>There are two ways to make Java transforms available to other SDKs.</p><ul><li>Option 1: In some cases, you can use existing Java transforms from other SDKs without writing any additional Java code.</li><li>Option 2: You can use arbitrary Java transforms from other SDKs by adding a few Java classes.</li></ul><h5 id=13111-using-existing-java-transforms-without-writing-more-java-code>13.1.1.1 Using existing Java transforms without writing more Java code</h5><p>Starting with Beam 2.34.0, Python SDK users can use some Java transforms without writing additional Java code. This can be useful in many cases. For example:</p><ul><li>A developer not familiar with Java may need to use an existing Java transform from a Python pipeline.</li><li>A developer may need to make an existing Java transform available to a Python pipeline without writing/releasing more Java code.</li></ul><blockquote><p><strong>Note:</strong> This feature is currently only available when using Java transforms from a Python pipeline.</p></blockquote><p>To be eligible for direct usage, the API of the Java transform has to meet the following requirements:</p><ol><li>The Java transform can be constructed using an available public constructor or a public static method (a constructor method) in the same Java class.</li><li>The Java transform can be configured using one or more builder methods. Each builder method should be public and should return an instance of the Java transform.</li></ol><p>Here’s an example Java class that can be directly used from the Python API.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kd>public</span> <span class=kd>class</span> <span class=nc>JavaDataGenerator</span> <span class=kd>extends</span> <span class=n>PTransform</span><span class=o><</span><span class=n>PBegin</span><span class=o>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span> <span class=o>.</span> <span class=o>.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// The following method satisfies requirement 1. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=c1>// Note that you could use a class constructor instead of a static method. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=kd>static</span> <span class=n>JavaDataGenerator</span> <span class=nf>create</span><span class=o>(</span><span class=n>Integer</span> <span class=n>size</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=k>new</span> <span class=n>JavaDataGenerator</span><span class=o>(</span><span class=n>size</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>JavaDataGeneratorConfig</span> <span class=kd>implements</span> <span class=n>Serializable</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=n>prefix</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>long</span> <span class=n>length</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>String</span> <span class=n>suffix</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span> <span class=o>.</span> <span class=o>.</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1>// The following method conforms to requirement 2. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=kd>public</span> <span class=n>JavaDataGenerator</span> <span class=nf>withJavaDataGeneratorConfig</span><span class=o>(</span><span class=n>JavaDataGeneratorConfig</span> <span class=n>dataConfig</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=k>new</span> <span class=n>JavaDataGenerator</span><span class=o>(</span><span class=k>this</span><span class=o>.</span><span class=na>size</span><span class=o>,</span> <span class=n>javaDataGeneratorConfig</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=o>.</span> <span class=o>.</span> <span class=o>.</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span></code></pre></div><p>For a complete example, see <a href=https://github.com/apache/beam/blob/master/examples/multi-language/src/main/java/org/apache/beam/examples/multilanguage/JavaDataGenerator.java>JavaDataGenerator</a>.</p><p>To use a Java class that conforms to the above requirements from a Python SDK pipeline, follow these steps:</p><ol><li>Create a <em>yaml</em> allowlist that describes the Java transform classes and methods that will be directly accessed from Python.</li><li>Start an expansion service, using the <code>javaClassLookupAllowlistFile</code> option to pass the path to the allowlist.</li><li>Use the Python <a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/transforms/external.py>JavaExternalTransform</a> API to directly access Java transforms defined in the allowlist from the Python side.</li></ol><p>Starting with Beam 2.36.0, steps 1 and 2 can be skipped, as described in the corresponding sections below.</p><p><strong>Step 1</strong></p><p>To use an eligible Java transform from Python, define a <em>yaml</em> allowlist. This allowlist lists the class names, |
| constructor methods, and builder methods that are directly available to be used from the Python side.</p><p>Starting with Beam 2.35.0, you have the option to pass <code>*</code> to the <code>javaClassLookupAllowlistFile</code> option instead of defining an actual allowlist. The <code>*</code> specifies that all supported transforms in the classpath of the expansion service can be accessed through the API. We encourage using an actual allowlist for production, because allowing clients to access arbitrary Java classes can pose a security risk.</p><div class=snippet><div class="notebook-skip code-snippet without_switcher"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><pre tabindex=0><code>version: v1 |
| allowedClasses: |
| - className: my.beam.transforms.JavaDataGenerator |
| allowedConstructorMethods: |
| - create |
| allowedBuilderMethods: |
| - withJavaDataGeneratorConfig</code></pre></div></div><p><strong>Step 2</strong></p><p>Provide the allowlist as an argument when starting up the Java expansion service. For example, you can start the expansion service |
| as a local Java process using the following command:</p><div class=snippet><div class="notebook-skip code-snippet without_switcher"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><pre tabindex=0><code>java -jar <jar file> <port> --javaClassLookupAllowlistFile=<path to the allowlist file></code></pre></div></div><p>Starting with Beam 2.36.0, the <code>JavaExternalTransform</code> API will automatically start up an expansion service with a given <code>jar</code> file dependency if an expansion service address was not provided.</p><p><strong>Step 3</strong></p><p>You can use the Java class directly from your Python pipeline using a stub transform created from the <code>JavaExternalTransform</code> API. This API allows you to construct the transform using the Java class name and allows you to invoke builder methods to configure the class.</p><p>Constructor and method parameter types are mapped between Python and Java using a Beam schema. The schema is auto-generated using the object types |
| provided on the Python side. If the Java class constructor method or builder method accepts any complex object types, make sure that the Beam schema |
| for these objects is registered and available for the Java expansion service. If a schema has not been registered, the Java expansion service will |
| try to register a schema using <a href=/documentation/programming-guide/#creating-schemas>JavaFieldSchema</a>. In Python, arbitrary objects |
| can be represented using <code>NamedTuple</code>s, which will be represented as Beam rows in the schema. Here is a Python stub transform that represents the above |
| mentioned Java transform:</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>JavaDataGeneratorConfig</span> <span class=o>=</span> <span class=n>typing</span><span class=o>.</span><span class=n>NamedTuple</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl><span class=s1>'JavaDataGeneratorConfig'</span><span class=p>,</span> <span class=p>[(</span><span class=s1>'prefix'</span><span class=p>,</span> <span class=nb>str</span><span class=p>),</span> <span class=p>(</span><span class=s1>'length'</span><span class=p>,</span> <span class=nb>int</span><span class=p>),</span> <span class=p>(</span><span class=s1>'suffix'</span><span class=p>,</span> <span class=nb>str</span><span class=p>)])</span> |
| </span></span><span class=line><span class=cl><span class=n>data_config</span> <span class=o>=</span> <span class=n>JavaDataGeneratorConfig</span><span class=p>(</span><span class=n>prefix</span><span class=o>=</span><span class=s1>'start'</span><span class=p>,</span> <span class=n>length</span><span class=o>=</span><span class=mi>20</span><span class=p>,</span> <span class=n>suffix</span><span class=o>=</span><span class=s1>'end'</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>java_transform</span> <span class=o>=</span> <span class=n>JavaExternalTransform</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl><span class=s1>'my.beam.transforms.JavaDataGenerator'</span><span class=p>,</span> <span class=n>expansion_service</span><span class=o>=</span><span class=s1>'localhost:<port>'</span><span class=p>)</span><span class=o>.</span><span class=n>create</span><span class=p>(</span><span class=n>numpy</span><span class=o>.</span><span class=n>int32</span><span class=p>(</span><span class=mi>100</span><span class=p>))</span><span class=o>.</span><span class=n>withJavaDataGeneratorConfig</span><span class=p>(</span><span class=n>data_config</span><span class=p>)</span> |
| </span></span></code></pre></div><p>You can use this transform in a Python pipeline along with other Python transforms. For a complete example, see <a href=https://github.com/apache/beam/blob/master/examples/multi-language/python/javadatagenerator.py>javadatagenerator.py</a>.</p><h5 id=13112-using-the-api-to-make-existing-java-transforms-available-to-other-sdks>13.1.1.2 Using the API to make existing Java transforms available to other SDKs</h5><p>To make your Beam Java SDK transform portable across SDK languages, you must implement two interfaces: <a href=https://github.com/apache/beam/blob/master/sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/ExternalTransformBuilder.java>ExternalTransformBuilder</a> and <a href=https://github.com/apache/beam/blob/master/sdks/java/core/src/main/java/org/apache/beam/sdk/expansion/ExternalTransformRegistrar.java>ExternalTransformRegistrar</a>. The <code>ExternalTransformBuilder</code> interface constructs the cross-language transform using configuration values passed in from the pipeline, and the <code>ExternalTransformRegistrar</code> interface registers the cross-language transform for use with the expansion service.</p><p><strong>Implementing the interfaces</strong></p><ol><li><p>Define a Builder class for your transform that implements the <code>ExternalTransformBuilder</code> interface and overrides the <code>buildExternal</code> method that will be used to build your transform object. Initial configuration values for your transform should be defined in the <code>buildExternal</code> method. In most cases, it’s convenient to make the Java transform builder class implement <code>ExternalTransformBuilder</code>.</p><blockquote><p><strong>Note:</strong> <code>ExternalTransformBuilder</code> requires you to define a configuration object (a simple POJO) to capture a set of parameters sent by external SDKs to initiate the Java transform. Usually these parameters directly map to constructor parameters of the Java transform.</p></blockquote><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@AutoValue.Builder</span> |
| </span></span><span class=line><span class=cl><span class=kd>abstract</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>Builder</span><span class=o><</span><span class=n>K</span><span class=o>,</span> <span class=n>V</span><span class=o>></span> |
| </span></span><span class=line><span class=cl> <span class=kd>implements</span> <span class=n>ExternalTransformBuilder</span><span class=o><</span><span class=n>External</span><span class=o>.</span><span class=na>Configuration</span><span class=o>,</span> <span class=n>PBegin</span><span class=o>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>K</span><span class=o>,</span> <span class=n>V</span><span class=o>>>></span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>abstract</span> <span class=n>Builder</span><span class=o><</span><span class=n>K</span><span class=o>,</span> <span class=n>V</span><span class=o>></span> <span class=nf>setConsumerConfig</span><span class=o>(</span><span class=n>Map</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Object</span><span class=o>></span> <span class=n>config</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>abstract</span> <span class=n>Builder</span><span class=o><</span><span class=n>K</span><span class=o>,</span> <span class=n>V</span><span class=o>></span> <span class=nf>setTopics</span><span class=o>(</span><span class=n>List</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>topics</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=cm>/** Remaining property declarations omitted for clarity. */</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>abstract</span> <span class=n>Read</span><span class=o><</span><span class=n>K</span><span class=o>,</span> <span class=n>V</span><span class=o>></span> <span class=nf>build</span><span class=o>();</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>PTransform</span><span class=o><</span><span class=n>PBegin</span><span class=o>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>KV</span><span class=o><</span><span class=n>K</span><span class=o>,</span> <span class=n>V</span><span class=o>>>></span> <span class=nf>buildExternal</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>External</span><span class=o>.</span><span class=na>Configuration</span> <span class=n>config</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=n>setTopics</span><span class=o>(</span><span class=n>ImmutableList</span><span class=o>.</span><span class=na>copyOf</span><span class=o>(</span><span class=n>config</span><span class=o>.</span><span class=na>topics</span><span class=o>));</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=cm>/** Remaining property defaults omitted for clarity. */</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span></code></pre></div><p>For complete examples, see <a href=https://github.com/apache/beam/blob/master/examples/multi-language/src/main/java/org/apache/beam/examples/multilanguage/JavaCountBuilder.java>JavaCountBuilder</a> and <a href=https://github.com/apache/beam/blob/master/examples/multi-language/src/main/java/org/apache/beam/examples/multilanguage/JavaPrefixBuilder.java>JavaPrefixBuilder</a>.</p><p>Note that the <code>buildExternal</code> method can perform additional operations before setting properties received from external SDKs in the transform. For example, <code>buildExternal</code> can validate properties available in the configuration object before setting them in the transform.</p></li><li><p>Register the transform as an external cross-language transform by defining a class that implements <code>ExternalTransformRegistrar</code>. You must annotate your class with the <code>AutoService</code> annotation to ensure that your transform is registered and instantiated properly by the expansion service.</p></li><li><p>In your registrar class, define a Uniform Resource Name (URN) for your transform. The URN must be a unique string that identifies your transform with the expansion service.</p></li><li><p>From within your registrar class, define a configuration class for the parameters used during the initialization of your transform by the external SDK.</p><p>The following example from the KafkaIO transform shows how to implement steps two through four:</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=nd>@AutoService</span><span class=o>(</span><span class=n>ExternalTransformRegistrar</span><span class=o>.</span><span class=na>class</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl><span class=kd>public</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>External</span> <span class=kd>implements</span> <span class=n>ExternalTransformRegistrar</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>static</span> <span class=kd>final</span> <span class=n>String</span> <span class=n>URN</span> <span class=o>=</span> <span class=s>"beam:external:java:kafka:read:v1"</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@Override</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=n>Map</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>Class</span><span class=o><?</span> <span class=kd>extends</span> <span class=n>ExternalTransformBuilder</span><span class=o><?,</span> <span class=o>?,</span> <span class=o>?>>></span> <span class=n>knownBuilders</span><span class=o>()</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>ImmutableMap</span><span class=o>.</span><span class=na>of</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>URN</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=o>(</span><span class=n>Class</span><span class=o><?</span> <span class=kd>extends</span> <span class=n>ExternalTransformBuilder</span><span class=o><?,</span> <span class=o>?,</span> <span class=o>?>>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>(</span><span class=n>Class</span><span class=o><?>)</span> <span class=n>AutoValue_KafkaIO_Read</span><span class=o>.</span><span class=na>Builder</span><span class=o>.</span><span class=na>class</span><span class=o>);</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=cm>/** Parameters class to expose the Read transform to an external SDK. */</span> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kd>static</span> <span class=kd>class</span> <span class=nc>Configuration</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=kd>private</span> <span class=n>Map</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>></span> <span class=n>consumerConfig</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=kd>private</span> <span class=n>List</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>topics</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>setConsumerConfig</span><span class=o>(</span><span class=n>Map</span><span class=o><</span><span class=n>String</span><span class=o>,</span> <span class=n>String</span><span class=o>></span> <span class=n>consumerConfig</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>this</span><span class=o>.</span><span class=na>consumerConfig</span> <span class=o>=</span> <span class=n>consumerConfig</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=kd>public</span> <span class=kt>void</span> <span class=nf>setTopics</span><span class=o>(</span><span class=n>List</span><span class=o><</span><span class=n>String</span><span class=o>></span> <span class=n>topics</span><span class=o>)</span> <span class=o>{</span> |
| </span></span><span class=line><span class=cl> <span class=k>this</span><span class=o>.</span><span class=na>topics</span> <span class=o>=</span> <span class=n>topics</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=cm>/** Remaining properties omitted for clarity. */</span> |
| </span></span><span class=line><span class=cl> <span class=o>}</span> |
| </span></span><span class=line><span class=cl><span class=o>}</span> |
| </span></span></code></pre></div><p>For additional examples, see <a href=https://github.com/apache/beam/blob/master/examples/multi-language/src/main/java/org/apache/beam/examples/multilanguage/JavaCountRegistrar.java>JavaCountRegistrar</a> and <a href=https://github.com/apache/beam/blob/master/examples/multi-language/src/main/java/org/apache/beam/examples/multilanguage/JavaPrefixRegistrar.java>JavaPrefixRegistrar</a>.</p></li></ol><p>After you have implemented the <code>ExternalTransformBuilder</code> and <code>ExternalTransformRegistrar</code> interfaces, your transform can be registered and created successfully by the default Java expansion service.</p><p><strong>Starting the expansion service</strong></p><p>You can use an expansion service with multiple transforms in the same pipeline. The Beam Java SDK provides a default expansion service for Java transforms. You can also write your own expansion service, but that’s generally not needed, so it’s not covered in this section.</p><p>Perform the following to start up a Java expansion service directly:</p><div class=snippet><div class="notebook-skip code-snippet without_switcher"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><pre tabindex=0><code># Build a JAR with both your transform and the expansion service |
| |
| # Start the expansion service at the specified port. |
| $ jar -jar /path/to/expansion_service.jar <PORT_NUMBER></code></pre></div></div><p>The expansion service is now ready to serve transforms on the specified port.</p><p>When creating SDK-specific wrappers for your transform, you may be able to use SDK-provided utilities to start up an expansion service. For example, the Python SDK provides the utilities <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.JavaJarExpansionService><code>JavaJarExpansionService</code></a> and <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.BeamJarExpansionService><code>BeamJarExpansionService</code></a> for starting up a Java expansion service using a JAR file.</p><p><strong>Including dependencies</strong></p><p>If your transform requires external libraries, you can include them by adding them to the classpath of the expansion service. After they are included in the classpath, they will be staged when your transform is expanded by the expansion service.</p><p><strong>Writing SDK-specific wrappers</strong></p><p>Your cross-language Java transform can be called through the lower-level <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.ExternalTransform><code>ExternalTransform</code></a> class in a multi-language pipeline (as described in the next section); however, if possible, you should write an SDK-specific wrapper in the language of the pipeline (such as Python) to access the transform instead. This higher-level abstraction will make it easier for pipeline authors to use your transform.</p><p>To create an SDK wrapper for use in a Python pipeline, do the following:</p><ol><li><p>Create a Python module for your cross-language transform(s).</p></li><li><p>In the module, use one of the <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.PayloadBuilder><code>PayloadBuilder</code></a> classes to build the payload for the initial cross-language transform expansion request.</p><p>The parameter names and types of the payload should map to parameter names and types of the configuration POJO provided to the Java <code>ExternalTransformBuilder</code>. Parameter types are mapped across SDKs using a <a href=https://github.com/apache/beam/blob/master/model/pipeline/src/main/proto/org/apache/beam/model/pipeline/v1/schema.proto>Beam schema</a>. Parameter names are mapped by simply converting Python underscore-separated variable names to camel-case (Java standard).</p><p>In the following example, <a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/io/kafka.py>kafka.py</a> uses <code>NamedTupleBasedPayloadBuilder</code> to build the payload. The parameters map to the Java <a href=https://github.com/apache/beam/blob/master/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/KafkaIO.java>KafkaIO.External.Configuration</a> config object defined in the previous section.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>ReadFromKafkaSchema</span><span class=p>(</span><span class=n>typing</span><span class=o>.</span><span class=n>NamedTuple</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=n>consumer_config</span><span class=p>:</span> <span class=n>typing</span><span class=o>.</span><span class=n>Mapping</span><span class=p>[</span><span class=nb>str</span><span class=p>,</span> <span class=nb>str</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=n>topics</span><span class=p>:</span> <span class=n>typing</span><span class=o>.</span><span class=n>List</span><span class=p>[</span><span class=nb>str</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Other properties omitted for clarity.</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>payload</span> <span class=o>=</span> <span class=n>NamedTupleBasedPayloadBuilder</span><span class=p>(</span><span class=n>ReadFromKafkaSchema</span><span class=p>(</span><span class=o>...</span><span class=p>))</span> |
| </span></span></code></pre></div></li><li><p>Start an expansion service, unless one is specified by the pipeline creator. The Beam Python SDK provides the utilities <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.JavaJarExpansionService><code>JavaJarExpansionService</code></a> and <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.BeamJarExpansionService><code>BeamJarExpansionService</code></a> for starting up an expansion service using a JAR file. <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.JavaJarExpansionService><code>JavaJarExpansionService</code></a> can be used to start up an expansion service using the path (a local path or a URL) to a given JAR file. <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.BeamJarExpansionService><code>BeamJarExpansionService</code></a> can be used to start an expansion service from a JAR released with Beam.</p><p>For transforms released with Beam, do the following:</p><ol><li><p>Add a Gradle target to Beam that can be used to build a shaded expansion service JAR for the target Java transform. This target should produce a Beam JAR that contains all dependencies needed for expanding the Java transform, and the JAR should be released with Beam. You might be able to use an existing Gradle target that offers an aggregated version of an expansion service JAR (for example, for all GCP IO).</p></li><li><p>In your Python module, instantiate <code>BeamJarExpansionService</code> with the Gradle target.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>expansion_service</span> <span class=o>=</span> <span class=n>BeamJarExpansionService</span><span class=p>(</span><span class=s1>'sdks:java:io:expansion-service:shadowJar'</span><span class=p>)</span> |
| </span></span></code></pre></div></li></ol></li><li><p>Add a Python wrapper transform class that extends <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.ExternalTransform><code>ExternalTransform</code></a>. Pass the payload and expansion service defined above as parameters to the constructor of the <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.ExternalTransform><code>ExternalTransform</code></a> parent class.</p></li></ol><h4 id=1312-creating-cross-language-python-transforms>13.1.2. Creating cross-language Python transforms</h4><p>Any Python transforms defined in the scope of the expansion service should be accessible by specifying their fully qualified names. For example, you could use Python’s <code>ReadFromText</code> transform in a Java pipeline with its fully qualified name <code>apache_beam.io.ReadFromText</code>:</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=s>"Read"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>PythonExternalTransform</span><span class=o>.<</span><span class=n>PBegin</span><span class=o>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>>></span><span class=n>from</span><span class=o>(</span><span class=s>"apache_beam.io.ReadFromText"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withKwarg</span><span class=o>(</span><span class=s>"file_pattern"</span><span class=o>,</span> <span class=n>options</span><span class=o>.</span><span class=na>getInputFile</span><span class=o>())</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withKwarg</span><span class=o>(</span><span class=s>"validate"</span><span class=o>,</span> <span class=kc>false</span><span class=o>))</span> |
| </span></span></code></pre></div><p><code>PythonExternalTransform</code> has other useful methods such as <code>withExtraPackages</code> for staging PyPI package dependencies and <code>withOutputCoder</code> for setting an output coder. If your transform exists in an external package, make sure to specify that package using <code>withExtraPackages</code>, for example:</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>p</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=s>"Read"</span><span class=o>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>PythonExternalTransform</span><span class=o>.<</span><span class=n>PBegin</span><span class=o>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>String</span><span class=o>>></span><span class=n>from</span><span class=o>(</span><span class=s>"my_python_package.BeamReadPTransform"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withExtraPackages</span><span class=o>(</span><span class=n>ImmutableList</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"my_python_package"</span><span class=o>)))</span> |
| </span></span></code></pre></div><p>Alternatively, you may want to create a Python module that registers an existing Python transform as a cross-language transform for use with the Python expansion service and calls into that existing transform to perform its intended operation. A registered URN can be used later in an expansion request for indicating an expansion target.</p><p><strong>Defining the Python module</strong></p><ol><li><p>Define a Uniform Resource Name (URN) for your transform. The URN must be a unique string that identifies your transform with the expansion service.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>TEST_COMPK_URN</span> <span class=o>=</span> <span class=s2>"beam:transforms:xlang:test:compk"</span> |
| </span></span></code></pre></div></li><li><p>For an existing Python transform, create a new class to register the URN with the Python expansion service.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=nd>@ptransform.PTransform.register_urn</span><span class=p>(</span><span class=n>TEST_COMPK_URN</span><span class=p>,</span> <span class=kc>None</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>CombinePerKeyTransform</span><span class=p>(</span><span class=n>ptransform</span><span class=o>.</span><span class=n>PTransform</span><span class=p>):</span> |
| </span></span></code></pre></div></li><li><p>From within the class, define an expand method that takes an input PCollection, runs the Python transform, and then returns the output PCollection.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>def</span> <span class=nf>expand</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>pcoll</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>pcoll</span> \ |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>CombinePerKey</span><span class=p>(</span><span class=nb>sum</span><span class=p>)</span><span class=o>.</span><span class=n>with_output_types</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>typing</span><span class=o>.</span><span class=n>Tuple</span><span class=p>[</span><span class=n>unicode</span><span class=p>,</span> <span class=nb>int</span><span class=p>])</span> |
| </span></span></code></pre></div></li><li><p>As with other Python transforms, define a <code>to_runner_api_parameter</code> method that returns the URN.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>def</span> <span class=nf>to_runner_api_parameter</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>unused_context</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>TEST_COMPK_URN</span><span class=p>,</span> <span class=kc>None</span> |
| </span></span></code></pre></div></li><li><p>Define a static <code>from_runner_api_parameter</code> method that returns an instantiation of the cross-language Python transform.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=nd>@staticmethod</span> |
| </span></span><span class=line><span class=cl><span class=k>def</span> <span class=nf>from_runner_api_parameter</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>unused_ptransform</span><span class=p>,</span> <span class=n>unused_parameter</span><span class=p>,</span> <span class=n>unused_context</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>CombinePerKeyTransform</span><span class=p>()</span> |
| </span></span></code></pre></div></li></ol><p><strong>Starting the expansion service</strong></p><p>An expansion service can be used with multiple transforms in the same pipeline. The Beam Python SDK provides a default expansion service for you to use with your Python transforms. You are free to write your own expansion service, but that is generally not needed, so it is not covered in this section.</p><p>Perform the following steps to start up the default Python expansion service directly:</p><ol><li><p>Create a virtual environment and <a href=/get-started/quickstart-py/>install the Apache Beam SDK</a>.</p></li><li><p>Start the Python SDK’s expansion service with a specified port.</p><div class=snippet><div class="notebook-skip code-snippet without_switcher"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><pre tabindex=0><code>$ export PORT_FOR_EXPANSION_SERVICE=12345 |
| </code></pre></div></div></li><li><p>Import any modules that contain transforms to be made available using the expansion service.</p><div class=snippet><div class="notebook-skip code-snippet without_switcher"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><pre tabindex=0><code>$ python -m apache_beam.runners.portability.expansion_service_test -p $PORT_FOR_EXPANSION_SERVICE --pickle_library=cloudpickle |
| </code></pre></div></div></li><li><p>This expansion service is now ready to serve up transforms on the address `localhost:$PORT_FOR_EXPANSION_SERVICE</p></li></ol><h4 id=1313-creating-cross-language-go-transforms>13.1.3. Creating cross-language Go transforms</h4><p>Go currently does not support creating cross-language transforms, only using cross-language |
| transforms from other languages; see more at <a href=https://github.com/apache/beam/issues/21767>Issue 21767</a>.</p><h4 id=1314-defining-a-urn>13.1.4. Defining a URN</h4><p>Developing a cross-language transform involves defining a URN for registering the transform with an expansion service. In this section |
| we provide a convention for defining such URNs. Following this convention is optional but it will ensure that your transform |
| will not run into conflicts when registering in an expansion service along with transforms developed by other developers.</p><h5 id=13141-schema>13.1.4.1. Schema</h5><p>A URN should consist of the following components:</p><ul><li><strong>ns-id</strong>: A namespace identifier. Default recommendation is <code>beam:transform</code>.</li><li><strong>org-identifier</strong>: Identifies the organization where the transform was defined. Transforms defined in Apache Beam use <code>org.apache.beam</code> for this.</li><li><strong>functionality-identifier</strong>: Identifies the functionality of the cross-language transform.</li><li><strong>version</strong>: a version number for the transform.</li></ul><p>We provide the schema from the URN convention in <a href=https://en.wikipedia.org/wiki/Augmented_Backus%E2%80%93Naur_form>augmented Backus–Naur</a> form. |
| Keywords in upper case are from the <a href=https://datatracker.ietf.org/doc/html/rfc8141>URN spec</a>.</p><div class=snippet><div class="notebook-skip code-snippet without_switcher"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><pre tabindex=0><code>transform-urn = ns-id “:” org-identifier “:” functionality-identifier “:” version |
| ns-id = (“beam” / NID) “:” “transform” |
| id-char = ALPHA / DIGIT / "-" / "." / "_" / "~" ; A subset of characters allowed in a URN |
| org-identifier = 1*id-char |
| functionality-identifier = 1*id-char |
| version = “v” 1*(DIGIT / “.”) ; For example, ‘v1.2’</code></pre></div></div><h5 id=13142-examples>13.1.4.2. Examples</h5><p>Below we’ve given some example transform classes and corresponding URNs to be used.</p><ul><li>A transform offered with Apache Beam that writes Parquet files.<ul><li><code>beam:transform:org.apache.beam:parquet_write:v1</code></li></ul></li><li>A transform offered with Apache Beam that reads from Kafka with metadata.<ul><li><code>beam:transform:org.apache.beam:kafka_read_with_metadata:v1</code></li></ul></li><li>A transform developed by organization abc.org that reads from data store MyDatastore.<ul><li><code>beam:transform:org.abc:mydatastore_read:v1</code></li></ul></li></ul><h3 id=use-x-lang-transforms>13.2. Using cross-language transforms</h3><p>Depending on the SDK language of the pipeline, you can use a high-level SDK-wrapper class, or a low-level transform class to access a cross-language transform.</p><h4 id=1321-using-cross-language-transforms-in-a-java-pipeline>13.2.1. Using cross-language transforms in a Java pipeline</h4><p>Users have three options to use cross-language transforms in a Java pipeline. At the highest level of abstraction, some popular Python transforms are accessible through dedicated Java wrapper transforms. For example, the Java SDK has the <code>DataframeTransform</code> class, which uses the Python SDK’s <code>DataframeTransform</code>, and it has the <code>RunInference</code> class, which uses the Python SDK’s <code>RunInference</code>, and so on. When an SDK-specific wrapper transform is not available for a target Python transform, you can use the lower-level <a href=https://github.com/apache/beam/blob/master/sdks/java/extensions/python/src/main/java/org/apache/beam/sdk/extensions/python/PythonExternalTransform.java>PythonExternalTransform</a> class instead by specifying the fully qualified name of the Python transform. If you want to try external transforms from SDKs other than Python (including Java SDK itself), you can also use the lowest-level <a href=https://github.com/apache/beam/blob/master/runners/core-construction-java/src/main/java/org/apache/beam/runners/core/construction/External.java>External</a> class.</p><p><strong>Using an SDK wrapper</strong></p><p>To use a cross-language transform through an SDK wrapper, import the module for the SDK wrapper and call it from your pipeline, as shown in the example:</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=kn>import</span> <span class=nn>org.apache.beam.sdk.extensions.python.transforms.DataframeTransform</span><span class=o>;</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>input</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span><span class=n>DataframeTransform</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"lambda df: df.groupby('a').sum()"</span><span class=o>).</span><span class=na>withIndexes</span><span class=o>())</span> |
| </span></span></code></pre></div><p><strong>Using the PythonExternalTransform class</strong></p><p>When an SDK-specific wrapper is not available, you can access the Python cross-language transform through the <code>PythonExternalTransform</code> class by specifying the fully qualified name and the constructor arguments of the target Python transform.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>input</span><span class=o>.</span><span class=na>apply</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>PythonExternalTransform</span><span class=o>.<</span><span class=n>PCollection</span><span class=o><</span><span class=n>Row</span><span class=o>>,</span> <span class=n>PCollection</span><span class=o><</span><span class=n>Row</span><span class=o>>></span><span class=n>from</span><span class=o>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"apache_beam.dataframe.transforms.DataframeTransform"</span><span class=o>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withKwarg</span><span class=o>(</span><span class=s>"func"</span><span class=o>,</span> <span class=n>PythonCallableSource</span><span class=o>.</span><span class=na>of</span><span class=o>(</span><span class=s>"lambda df: df.groupby('a').sum()"</span><span class=o>))</span> |
| </span></span><span class=line><span class=cl> <span class=o>.</span><span class=na>withKwarg</span><span class=o>(</span><span class=s>"include_indexes"</span><span class=o>,</span> <span class=kc>true</span><span class=o>))</span> |
| </span></span></code></pre></div><p><strong>Using the External class</strong></p><ol><li><p>Make sure you have any runtime environment dependencies (like the JRE) installed on your local machine (either directly on the local machine or available through a container). See the expansion service section for more details.</p><blockquote><p><strong>Note:</strong> When including Python transforms from within a Java pipeline, all Python dependencies have to be included in the SDK harness container.</p></blockquote></li><li><p>Start up the expansion service for the SDK that is in the language of the transform you’re trying to consume, if not available.</p><p>Make sure the transform you are trying to use is available and can be used by the expansion service.</p></li><li><p>Include <a href=https://github.com/apache/beam/blob/master/runners/core-construction-java/src/main/java/org/apache/beam/runners/core/construction/External.java>External.of(…)</a> when instantiating your pipeline. Reference the URN, payload, and expansion service. For examples, see the <a href=https://github.com/apache/beam/blob/master/runners/core-construction-java/src/test/java/org/apache/beam/runners/core/construction/ValidateRunnerXlangTest.java>cross-language transform test suite</a>.</p></li><li><p>After the job has been submitted to the Beam runner, shutdown the expansion service by terminating the expansion service process.</p></li></ol><h4 id=1322-using-cross-language-transforms-in-a-python-pipeline>13.2.2. Using cross-language transforms in a Python pipeline</h4><p>If a Python-specific wrapper for a cross-language transform is available, use that. Otherwise, you have to use the lower-level <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.ExternalTransform><code>ExternalTransform</code></a> class to access the transform.</p><p><strong>Using an SDK wrapper</strong></p><p>To use a cross-language transform through an SDK wrapper, import the module for the SDK wrapper and call it from your pipeline, as shown in the example:</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam.io.kafka</span> <span class=kn>import</span> <span class=n>ReadFromKafka</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=n>kafka_records</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>pipeline</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=s1>'ReadFromKafka'</span> <span class=o>>></span> <span class=n>ReadFromKafka</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>consumer_config</span><span class=o>=</span><span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'bootstrap.servers'</span><span class=p>:</span> <span class=bp>self</span><span class=o>.</span><span class=n>bootstrap_servers</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=s1>'auto.offset.reset'</span><span class=p>:</span> <span class=s1>'earliest'</span> |
| </span></span><span class=line><span class=cl> <span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=n>topics</span><span class=o>=</span><span class=p>[</span><span class=bp>self</span><span class=o>.</span><span class=n>topic</span><span class=p>],</span> |
| </span></span><span class=line><span class=cl> <span class=n>max_num_records</span><span class=o>=</span><span class=n>max_num_records</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>expansion_service</span><span class=o>=<</span><span class=n>Address</span> <span class=n>of</span> <span class=n>expansion</span> <span class=n>service</span><span class=o>></span><span class=p>))</span> |
| </span></span></code></pre></div><p><strong>Using the ExternalTransform class</strong></p><p>When an SDK-specific wrapper isn’t available, you will have to access the cross-language transform through the <code>ExternalTransform</code> class.</p><ol><li><p>Make sure you have any runtime environment dependencies (like the JRE) installed on your local machine. See the expansion service section for more details.</p></li><li><p>Start up the expansion service for the SDK that is in the language of the transform you’re trying to consume, if not available. |
| Python provides several classes for automatically starting expansion java services such as |
| <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.JavaJarExpansionService>JavaJarExpansionService</a> |
| and <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.BeamJarExpansionService>BeamJarExpansionService</a> |
| which can be passed directly as an expansion service to <code>beam.ExternalTransform</code>. |
| Make sure the transform you’re trying to use is available and can be used by the expansion service.</p><p>For Java, make sure the builder and registrar for the transform are available in the classpath of the expansion service.</p></li><li><p>Include <code>ExternalTransform</code> when instantiating your pipeline. Reference the URN, payload, and expansion service. |
| You can use one of the available <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.PayloadBuilder><code>PayloadBuilder</code></a> classes to build the payload for <code>ExternalTransform</code>.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>with</span> <span class=n>pipeline</span> <span class=k>as</span> <span class=n>p</span><span class=p>:</span> |
| </span></span><span class=line><span class=cl> <span class=n>res</span> <span class=o>=</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>p</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Create</span><span class=p>([</span><span class=s1>'a'</span><span class=p>,</span> <span class=s1>'b'</span><span class=p>])</span><span class=o>.</span><span class=n>with_output_types</span><span class=p>(</span><span class=n>unicode</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ExternalTransform</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=n>TEST_PREFIX_URN</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>ImplicitSchemaPayloadBuilder</span><span class=p>({</span><span class=s1>'data'</span><span class=p>:</span> <span class=s1>'0'</span><span class=p>}),</span> |
| </span></span><span class=line><span class=cl> <span class=o><</span><span class=n>expansion</span> <span class=n>service</span><span class=o>></span><span class=p>))</span> |
| </span></span><span class=line><span class=cl> <span class=n>assert_that</span><span class=p>(</span><span class=n>res</span><span class=p>,</span> <span class=n>equal_to</span><span class=p>([</span><span class=s1>'0a'</span><span class=p>,</span> <span class=s1>'0b'</span><span class=p>]))</span> |
| </span></span></code></pre></div><p>For additional examples, see <a href=https://github.com/apache/beam/blob/master/examples/multi-language/python/addprefix.py>addprefix.py</a> and <a href=https://github.com/apache/beam/blob/master/examples/multi-language/python/javacount.py>javacount.py</a>.</p></li><li><p>After the job has been submitted to the Beam runner, shut down any manually started expansion services by terminating the expansion service process.</p></li></ol><p><strong>Using the JavaExternalTransform class</strong></p><p>Python has the ability to invoke Java-defined transforms via <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.external.html#apache_beam.transforms.external.JavaExternalTransform>proxy objects</a> |
| as if they were Python transforms. |
| These are invoked as follows</p><pre><code>```py |
| MyJavaTransform = beam.JavaExternalTransform('fully.qualified.ClassName', classpath=[jars]) |
| |
| with pipeline as p: |
| res = ( |
| p |
| | beam.Create(['a', 'b']).with_output_types(unicode) |
| | MyJavaTransform(javaConstructorArg, ...).builderMethod(...) |
| assert_that(res, equal_to(['0a', '0b'])) |
| ``` |
| </code></pre><p>Python’s <code>getattr</code> method can be used if the method names in java are reserved |
| Python keywords such as <code>from</code>.</p><p>As with other external transforms, either a pre-started expansion service can |
| be provided, or jar files that include the transform, its dependencies, and |
| Beam’s expansion service in which case an expansion service will be auto-started.</p><h4 id=1323-using-cross-language-transforms-in-a-go-pipeline>13.2.3. Using cross-language transforms in a Go pipeline</h4><p>If a Go-specific wrapper for a cross-language is available, use that. Otherwise, you have to use the |
| lower-level <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam#CrossLanguage>CrossLanguage</a> |
| function to access the transform.</p><p><strong>Expansion Services</strong></p><p>The Go SDK supports automatically starting Java expansion services if an expansion address is not provided, although this is slower than |
| providing a persistent expansion service. Many wrapped Java transforms manage perform this automatically; if you wish to do this manually, use the <code>xlangx</code> package’s |
| <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2@v2.40.0/go/pkg/beam/core/runtime/xlangx#UseAutomatedJavaExpansionService>UseAutomatedJavaExpansionService()</a> function. In order to use Python cross-language transforms, you must manually start any necessary expansion |
| services on your local machine and ensure they are accessible to your code during pipeline construction.</p><p><strong>Using an SDK wrapper</strong></p><p>To use a cross-language transform through an SDK wrapper, import the package for the SDK wrapper |
| and call it from your pipeline as shown in the example:</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kn>import</span> <span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=s>"github.com/apache/beam/sdks/v2/go/pkg/beam/io/xlang/kafkaio"</span> |
| </span></span><span class=line><span class=cl><span class=p>)</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1>// Kafka Read using previously defined values. |
| </span></span></span><span class=line><span class=cl><span class=c1></span><span class=nx>kafkaRecords</span> <span class=o>:=</span> <span class=nx>kafkaio</span><span class=p>.</span><span class=nf>Read</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=nx>s</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=nx>expansionAddr</span><span class=p>,</span> <span class=c1>// Address of expansion service. |
| </span></span></span><span class=line><span class=cl><span class=c1></span> <span class=nx>bootstrapAddr</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>[]</span><span class=kt>string</span><span class=p>{</span><span class=nx>topicName</span><span class=p>},</span> |
| </span></span><span class=line><span class=cl> <span class=nx>kafkaio</span><span class=p>.</span><span class=nf>MaxNumRecords</span><span class=p>(</span><span class=nx>numRecords</span><span class=p>),</span> |
| </span></span><span class=line><span class=cl> <span class=nx>kafkaio</span><span class=p>.</span><span class=nf>ConsumerConfigs</span><span class=p>(</span><span class=kd>map</span><span class=p>[</span><span class=kt>string</span><span class=p>]</span><span class=kt>string</span><span class=p>{</span><span class=s>"auto.offset.reset"</span><span class=p>:</span> <span class=s>"earliest"</span><span class=p>}))</span> |
| </span></span></code></pre></div><p><strong>Using the CrossLanguage function</strong></p><p>When an SDK-specific wrapper isn’t available, you will have to access the cross-language transform through the <code>beam.CrossLanguage</code> function.</p><ol><li><p>Make sure you have the appropriate expansion service running. See the expansion service section for details.</p></li><li><p>Make sure the transform you’re trying to use is available and can be used by the expansion service. |
| Refer to <a href=#create-x-lang-transforms>Creating cross-language transforms</a> for details.</p></li><li><p>Use the <code>beam.CrossLanguage</code> function in your pipeline as appropriate. Reference the URN, payload, |
| expansion service address, and define inputs and outputs. You can use the |
| <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam#CrossLanguagePayload>beam.CrossLanguagePayload</a> |
| function as a helper for encoding a payload. You can use the |
| <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam#UnnamedInput>beam.UnnamedInput</a> and |
| <a href=https://pkg.go.dev/github.com/apache/beam/sdks/v2/go/pkg/beam#UnnamedOutput>beam.UnnamedOutput</a> |
| functions as shortcuts for single, unnamed inputs/outputs or define a map for named ones.</p><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=kd>type</span> <span class=nx>prefixPayload</span> <span class=kd>struct</span> <span class=p>{</span> |
| </span></span><span class=line><span class=cl> <span class=nx>Data</span> <span class=kt>string</span> <span class=s>`beam:"data"`</span> |
| </span></span><span class=line><span class=cl><span class=p>}</span> |
| </span></span><span class=line><span class=cl><span class=nx>urn</span> <span class=o>:=</span> <span class=s>"beam:transforms:xlang:test:prefix"</span> |
| </span></span><span class=line><span class=cl><span class=nx>payload</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>CrossLanguagePayload</span><span class=p>(</span><span class=nx>prefixPayload</span><span class=p>{</span><span class=nx>Data</span><span class=p>:</span> <span class=nx>prefix</span><span class=p>})</span> |
| </span></span><span class=line><span class=cl><span class=nx>expansionAddr</span> <span class=o>:=</span> <span class=s>"localhost:8097"</span> |
| </span></span><span class=line><span class=cl><span class=nx>outT</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>UnnamedOutput</span><span class=p>(</span><span class=nx>typex</span><span class=p>.</span><span class=nf>New</span><span class=p>(</span><span class=nx>reflectx</span><span class=p>.</span><span class=nx>String</span><span class=p>))</span> |
| </span></span><span class=line><span class=cl><span class=nx>res</span> <span class=o>:=</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>CrossLanguage</span><span class=p>(</span><span class=nx>s</span><span class=p>,</span> <span class=nx>urn</span><span class=p>,</span> <span class=nx>payload</span><span class=p>,</span> <span class=nx>expansionAddr</span><span class=p>,</span> <span class=nx>beam</span><span class=p>.</span><span class=nf>UnnamedInput</span><span class=p>(</span><span class=nx>inputPCol</span><span class=p>),</span> <span class=nx>outT</span><span class=p>)</span> |
| </span></span></code></pre></div></li><li><p>After the job has been submitted to the Beam runner, shutdown the expansion service by |
| terminating the expansion service process.</p></li></ol><h4 id=1324-using-cross-language-transforms-in-a-typescript-pipeline>13.2.4. Using cross-language transforms in a Typescript pipeline</h4><p>Using a Typescript wrapper for a cross-language pipeline is similar to using any |
| other transform, provided the dependencies (e.g. a recent Python interpreter or |
| a Java JRE) is available. For example, most of the Typescript IOs are simply |
| wrappers around Beam transforms from other languages.</p><p class=paragraph-wrap>If a wrapper is not already available, one can use it explicitly using |
| <a href=https://github.com/apache/beam/blob/master/sdks/typescript/src/apache_beam/transforms/external.ts target=_blank rel="noopener noreferrer">apache_beam.transforms.external.rawExternalTransform</a>. |
| which takes a `urn` (a string identifying the transform), |
| a `payload` (a binary or json object parameterizing the transform), |
| and a `expansionService` which can either be an address of a pre-started service |
| or a callable returning an auto-started expansion service object.</p><p>For example, one could write</p><pre tabindex=0><code>pcoll.applyAsync( |
| rawExternalTransform( |
| "beam:registered:urn", |
| {arg: value}, |
| "localhost:expansion_service_port" |
| ) |
| ); |
| </code></pre><p>Note that <code>pcoll</code> must have a cross-language compatible coder coder such as <code>SchemaCoder</code>. |
| This can be ensured with the <a href=https://github.com/apache/beam/blob/master/sdks/typescript/src/apache_beam/transforms/internal.ts>withCoderInternal</a> |
| or <a href=https://github.com/apache/beam/blob/master/sdks/typescript/src/apache_beam/transforms/internal.ts>withRowCoder</a> |
| transforms, e.g.</p><pre tabindex=0><code> |
| const result = pcoll.apply( |
| beam.withRowCoder({ intFieldName: 0, stringFieldName: "" }) |
| ); |
| |
| </code></pre><p>Coder can also be specified on the output if it cannot be inferred, e.g.</p><p>In addition, there are several utilities such as <a href=https://github.com/apache/beam/blob/master/sdks/typescript/src/apache_beam/transforms/python.ts>pythonTransform</a> |
| that make it easier to invoke transforms from specific languages:</p><pre tabindex=0><code> |
| const result: PCollection<number> = await pcoll |
| .apply( |
| beam.withName("UpdateCoder1", beam.withRowCoder({ a: 0, b: 0 })) |
| ) |
| .applyAsync( |
| pythonTransform( |
| // Fully qualified name |
| "apache_beam.transforms.Map", |
| // Positional arguments |
| [pythonCallable("lambda x: x.a + x.b")], |
| // Keyword arguments |
| {}, |
| // Output type if it cannot be inferred |
| { requestedOutputCoders: { output: new VarIntCoder() } } |
| ) |
| ); |
| |
| </code></pre><p>Cross-language transforms can also be defined in line, which can be useful |
| for accessing features or libraries not available in the calling SDK</p><pre tabindex=0><code> |
| const result: PCollection<string> = await pcoll |
| .apply(withCoderInternal(new StrUtf8Coder())) |
| .applyAsync( |
| pythonTransform( |
| // Define an arbitrary transform from a callable. |
| "__callable__", |
| [ |
| pythonCallable(` |
| def apply(pcoll, prefix, postfix): |
| return pcoll | beam.Map(lambda s: prefix + s + postfix) |
| `), |
| ], |
| // Keyword arguments to pass above, if any. |
| { prefix: "x", postfix: "y" }, |
| // Output type if it cannot be inferred |
| { requestedOutputCoders: { output: new StrUtf8Coder() } } |
| ) |
| ); |
| |
| </code></pre><h3 id=x-lang-transform-runner-support>13.3. Runner Support</h3><p>Currently, portable runners such as Flink, Spark, and the direct runner can be used with multi-language pipelines.</p><p>Dataflow supports multi-language pipelines through the Dataflow Runner v2 backend architecture.</p><h3 id=x-lang-transform-tips-troubleshooting>13.4 Tips and Troubleshooting</h3><p>For additional tips and troubleshooting information, see <a href=https://cwiki.apache.org/confluence/display/BEAM/Multi-language+Pipelines+Tips>here</a>.</p><h2 id=batched-dofns>14 Batched DoFns</h2><nav class=language-switcher><strong>Adapt for:</strong><ul><li data-value=java class=active>Java SDK</li><li data-value=py>Python SDK</li><li data-value=go>Go SDK</li><li data-value=typescript>TypeScript SDK</li></ul></nav><p class="language-go language-java language-typescript">Batched DoFns are currently a Python-only feature.</p><p class=language-py>Batched DoFns enable users to create modular, composable components that |
| operate on batches of multiple logical elements. These DoFns can leverage |
| vectorized Python libraries, like numpy, scipy, and pandas, which operate on |
| batches of data for efficiency.</p><h3 id=batched-dofn-basics>14.1 Basics</h3><p class="language-go language-java language-typescript">Batched DoFns are currently a Python-only feature.</p><p class=language-py>A trivial Batched DoFn might look like this:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>MultiplyByTwo</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Type</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process_batch</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>batch</span><span class=p>:</span> <span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>)</span> <span class=o>-></span> <span class=n>Iterator</span><span class=p>[</span><span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>]:</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>batch</span> <span class=o>*</span> <span class=mi>2</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># Declare what the element-wise output type is</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>infer_output_type</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>input_element_type</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>input_element_type</span></span></span></code></pre></div></div></div><p class=language-py>This DoFn can be used in a Beam pipeline that otherwise operates on individual |
| elements. Beam will implicitly buffer elements and create numpy arrays on the |
| input side, and on the output side it will explode the numpy arrays back into |
| individual elements:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Create</span><span class=p>([</span><span class=mi>1</span><span class=p>,</span> <span class=mi>2</span><span class=p>,</span> <span class=mi>3</span><span class=p>,</span> <span class=mi>4</span><span class=p>])</span><span class=o>.</span><span class=n>with_output_types</span><span class=p>(</span><span class=n>np</span><span class=o>.</span><span class=n>int64</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>MultiplyByTwo</span><span class=p>())</span> <span class=c1># Implicit buffering and batch creation</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Map</span><span class=p>(</span><span class=k>lambda</span> <span class=n>x</span><span class=p>:</span> <span class=n>x</span><span class=o>/</span><span class=mi>3</span><span class=p>))</span> <span class=c1># Implicit batch explosion</span></span></span></code></pre></div></div></div><p class=language-py>Note that we use |
| <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.ptransform.html#apache_beam.transforms.ptransform.PTransform.with_output_types><code>PTransform.with_output_types</code></a> to |
| set the <em>element-wise</em> typehint for the output of <code>beam.Create</code>. Then, when |
| <code>MultiplyByTwo</code> is applied to this <code>PCollection</code>, Beam recognizes that |
| <code>np.ndarray</code> is an acceptable batch type to use in conjunction with <code>np.int64</code> |
| elements. We will use numpy typehints like these throughout this guide, but |
| Beam supports typehints from other libraries as well, see <a href=#batched-dofn-types>Supported Batch |
| Types</a>.</p><p class=language-py>In the previous case, Beam will implicitly create and explode batches at the |
| input and output boundaries. However, if Batched DoFns with equivalent types are |
| chained together, this batch creation and explosion will be elided. The batches |
| will be passed straight through! This makes it much simpler to efficiently |
| compose transforms that operate on batches.</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=p>(</span><span class=n>p</span> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>Create</span><span class=p>([</span><span class=mi>1</span><span class=p>,</span> <span class=mi>2</span><span class=p>,</span> <span class=mi>3</span><span class=p>,</span> <span class=mi>4</span><span class=p>])</span><span class=o>.</span><span class=n>with_output_types</span><span class=p>(</span><span class=n>np</span><span class=o>.</span><span class=n>int64</span><span class=p>)</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>MultiplyByTwo</span><span class=p>())</span> <span class=c1># Implicit buffering and batch creation</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>MultiplyByTwo</span><span class=p>())</span> <span class=c1># Batches passed through</span> |
| </span></span><span class=line><span class=cl> <span class=o>|</span> <span class=n>beam</span><span class=o>.</span><span class=n>ParDo</span><span class=p>(</span><span class=n>MultiplyByTwo</span><span class=p>()))</span></span></span></code></pre></div></div></div><h3 id=batched-dofn-elementwise>14.2 Element-wise Fallback</h3><p class="language-go language-java language-typescript">Batched DoFns are currently a Python-only feature.</p><p class=language-py>For some DoFns you may be able to provide both a batched and an element-wise |
| implementation of your desired logic. You can do this by simply defining both |
| <code>process</code> and <code>process_batch</code>:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>MultiplyByTwo</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>element</span><span class=p>:</span> <span class=n>np</span><span class=o>.</span><span class=n>int64</span><span class=p>)</span> <span class=o>-></span> <span class=n>Iterator</span><span class=p>[</span><span class=n>np</span><span class=o>.</span><span class=n>int64</span><span class=p>]:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Multiply an individual int64 by 2</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>element</span> <span class=o>*</span> <span class=mi>2</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process_batch</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>batch</span><span class=p>:</span> <span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>)</span> <span class=o>-></span> <span class=n>Iterator</span><span class=p>[</span><span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>]:</span> |
| </span></span><span class=line><span class=cl> <span class=c1># Multiply a _batch_ of int64s by 2</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>batch</span> <span class=o>*</span> <span class=mi>2</span></span></span></code></pre></div></div></div><p class=language-py>When executing this DoFn, Beam will select the best implementation to use given |
| the context. Generally, if the inputs to a DoFn are already batched Beam will |
| use the batched implementation; otherwise it will use the element-wise |
| implementation defined in the <code>process</code> method.</p><p class=language-py>Note that, in this case, there is no need to define <code>infer_output_type</code>. This is |
| because Beam can get the output type from the typehint on <code>process</code>.</p><h3 id=batched-dofn-batch-production>14.3 Batch Production vs. Batch Consumption</h3><p class="language-go language-java language-typescript">Batched DoFns are currently a Python-only feature.</p><p class=language-py>By convention, Beam assumes that the <code>process_batch</code> method, which consumes |
| batched inputs, will also produce batched outputs. Similarly, Beam assumes the |
| <code>process</code> method will produce individual elements. This can be overridden with |
| the <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.yields_elements><code>@beam.DoFn.yields_elements</code></a> and |
| <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.yields_batches><code>@beam.DoFn.yields_batches</code></a> decorators. For example:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># Consumes elements, produces batches</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>ReadFromFile</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=nd>@beam.DoFn.yields_batches</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>path</span><span class=p>:</span> <span class=nb>str</span><span class=p>)</span> <span class=o>-></span> <span class=n>Iterator</span><span class=p>[</span><span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>]:</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>array</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=c1># Declare what the element-wise output type is</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>infer_output_type</span><span class=p>(</span><span class=bp>self</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>np</span><span class=o>.</span><span class=n>int64</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=c1># Consumes batches, produces elements</span> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>WriteToFile</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=nd>@beam.DoFn.yields_elements</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process_batch</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>batch</span><span class=p>:</span> <span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>)</span> <span class=o>-></span> <span class=n>Iterator</span><span class=p>[</span><span class=nb>str</span><span class=p>]:</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>output_path</span></span></span></code></pre></div></div></div><h3 id=batched-dofn-types>14.4 Supported Batch Types</h3><p class="language-go language-java language-typescript">Batched DoFns are currently a Python-only feature.</p><p class=language-py>We’ve used numpy types in the Batched DoFn implementations in this guide – |
| <code>np.int64 </code> as the element typehint and <code>np.ndarray</code> as the corresponding |
| batch typehint – but Beam supports typehints from other libraries as well.</p><h4 id=numpyhttpsgithubcomapachebeamblobmastersdkspythonapache_beamtypehintsbatchpy><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/typehints/batch.py>numpy</a></h4><table><thead><tr><th>Element Typehint</th><th>Batch Typehint</th></tr></thead><tbody><tr><td>Numeric types (<code>int</code>, <code>np.int32</code>, <code>bool</code>, …)</td><td>np.ndarray (or NumpyArray)</td></tr></tbody></table><h4 id=pandashttpsgithubcomapachebeamblobmastersdkspythonapache_beamtypehintspandas_type_compatibilitypy><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/typehints/pandas_type_compatibility.py>pandas</a></h4><table><thead><tr><th>Element Typehint</th><th>Batch Typehint</th></tr></thead><tbody><tr><td>Numeric types (<code>int</code>, <code>np.int32</code>, <code>bool</code>, …)</td><td><code>pd.Series</code></td></tr><tr><td><code>bytes</code></td><td></td></tr><tr><td><code>Any</code></td><td></td></tr><tr><td><a href=#schemas>Beam Schema Types</a></td><td><code>pd.DataFrame</code></td></tr></tbody></table><h4 id=pyarrowhttpsgithubcomapachebeamblobmastersdkspythonapache_beamtypehintsarrow_type_compatibilitypy><a href=https://github.com/apache/beam/blob/master/sdks/python/apache_beam/typehints/arrow_type_compatibility.py>pyarrow</a></h4><table><thead><tr><th>Element Typehint</th><th>Batch Typehint</th></tr></thead><tbody><tr><td>Numeric types (<code>int</code>, <code>np.int32</code>, <code>bool</code>, …)</td><td><code>pd.Series</code></td></tr><tr><td><code>Any</code></td><td></td></tr><tr><td><code>List</code></td><td></td></tr><tr><td><code>Mapping</code></td><td></td></tr><tr><td><a href=#schemas>Beam Schema Types</a></td><td><code>pa.Table</code></td></tr></tbody></table><h4 id=other-types>Other types?</h4><p>If there are other batch types you would like to use with Batched DoFns, please |
| <a href=https://github.com/apache/beam/issues/new/choose>file an issue</a>.</p><h3 id=batched-dofn-dynamic-types>14.5 Dynamic Batch Input and Output Types</h3><p class="language-go language-java language-typescript">Batched DoFns are currently a Python-only feature.</p><p class=language-py>For some Batched DoFns, it may not be sufficient to declare batch types |
| statically, with typehints on <code>process</code> and/or <code>process_batch</code>. You may need to |
| declare these types dynamically. You can do this by overriding the |
| <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.get_input_batch_type><code>get_input_batch_type</code></a> |
| and |
| <a href=https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.get_output_batch_type><code>get_output_batch_type</code></a> |
| methods on your DoFn:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=c1># Utilize Beam's parameterized NumpyArray typehint</span> |
| </span></span><span class=line><span class=cl><span class=kn>from</span> <span class=nn>apache_beam.typehints.batch</span> <span class=kn>import</span> <span class=n>NumpyArray</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl><span class=k>class</span> <span class=nc>MultipyByTwo</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=c1># No typehints needed</span> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process_batch</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>batch</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>yield</span> <span class=n>batch</span> <span class=o>*</span> <span class=mi>2</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>get_input_batch_type</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>input_element_type</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>NumpyArray</span><span class=p>[</span><span class=n>input_element_type</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>get_output_batch_type</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>input_element_type</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>NumpyArray</span><span class=p>[</span><span class=n>input_element_type</span><span class=p>]</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>infer_output_type</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>input_element_type</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>input_element_type</span></span></span></code></pre></div></div></div><h3 id=batched-dofn-event-time>14.6 Batches and Event-time Semantics</h3><p class="language-go language-java language-typescript">Batched DoFns are currently a Python-only feature.</p><p class=language-py>Currently, batches must have a single set of timing information (event time, |
| windows, etc…) that applies to every logical element in the batch. There is |
| currently no mechanism to create batches that span multiple timestamps. However, |
| it is possible to retrieve this timing information in Batched DoFn |
| implementations. This information can be accessed by using the conventional |
| <code>DoFn.*Param</code> attributes:</p><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=k>class</span> <span class=nc>RetrieveTimingDoFn</span><span class=p>(</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>process_batch</span><span class=p>(</span> |
| </span></span><span class=line><span class=cl> <span class=bp>self</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>batch</span><span class=p>:</span> <span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>timestamp</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>TimestampParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=n>pane_info</span><span class=o>=</span><span class=n>beam</span><span class=o>.</span><span class=n>DoFn</span><span class=o>.</span><span class=n>PaneInfoParam</span><span class=p>,</span> |
| </span></span><span class=line><span class=cl> <span class=p>)</span> <span class=o>-></span> <span class=n>Iterator</span><span class=p>[</span><span class=n>np</span><span class=o>.</span><span class=n>ndarray</span><span class=p>]:</span> |
| </span></span><span class=line><span class=cl> <span class=o>...</span> |
| </span></span><span class=line><span class=cl> |
| </span></span><span class=line><span class=cl> <span class=k>def</span> <span class=nf>infer_output_type</span><span class=p>(</span><span class=bp>self</span><span class=p>,</span> <span class=n>input_type</span><span class=p>):</span> |
| </span></span><span class=line><span class=cl> <span class=k>return</span> <span class=n>input_type</span></span></span></code></pre></div></div></div><h2 id=transform-service>15 Transform service</h2><p>The Apache Beam SDK versions 2.49.0 and later include a <a href=https://docs.docker.com/compose/>Docker Compose</a> |
| service named <em>Transform service</em>.</p><p>The following diagram illustrates the basic architecture of the Transform service.</p><p><img src=/images/transform_service.png alt="Diagram of the Transform service architecture"></p><p>To use the Transform service, Docker must be available on the machine that starts the service.</p><p>The Transform service has several primary use cases.</p><h3 id=transform-service-usage-upgrade>15.1 Using the transform service to upgrade transforms</h3><p>Transform service can be used to upgrade (or downgrade) the Beam SDK versions of supported individual transforms used by Beam pipelines without changing the Beam version of the pipelines. |
| This feature is currently only available for Beam Java SDK 2.53.0 and later. Currently, the following transforms are available for upgrading:</p><ul><li>BigQuery read transform (URN: <em>beam:transform:org.apache.beam:bigquery_read:v1</em>)</li><li>BigQuery write transform (URN: <em>beam:transform:org.apache.beam:bigquery_write:v1</em>)</li><li>Kafka read transform (URN: <em>beam:transform:org.apache.beam:kafka_read_with_metadata:v2</em>)</li><li>Kafka write transform (URN: <em>beam:transform:org.apache.beam:kafka_write:v2</em>)</li></ul><p>To use this feature, you can simply execute a Java pipeline with additional pipeline options that specify the URNs of the transforms you would like to upgrade and the Beam version you would like to upgrade the transforms to. All transforms in the pipeline with matching URNs will be upgraded.</p><p>For example, to upgrade the BigQuery read transform for a pipeline run using Beam <code>2.53.0</code> to a future Beam version <code>2.xy.z</code>, you can specify the following additional pipelines options.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=o>--</span><span class=n>transformsToOverride</span><span class=o>=</span><span class=n>beam</span><span class=o>:</span><span class=n>transform</span><span class=o>:</span><span class=n>org</span><span class=o>.</span><span class=na>apache</span><span class=o>.</span><span class=na>beam</span><span class=o>:</span><span class=n>bigquery_read</span><span class=o>:</span><span class=n>v1</span> <span class=o>--</span><span class=n>transformServiceBeamVersion</span><span class=o>=</span><span class=n>2</span><span class=o>.</span><span class=na>xy</span><span class=o>.</span><span class=na>z</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>This</span> <span class=n>feature</span> <span class=ow>is</span> <span class=n>currently</span> <span class=ow>not</span> <span class=n>available</span> <span class=k>for</span> <span class=n>Python</span> <span class=n>SDK</span><span class=o>.</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>This</span> <span class=nx>feature</span> <span class=nx>is</span> <span class=nx>currently</span> <span class=nx>not</span> <span class=nx>available</span> <span class=k>for</span> <span class=nx>Go</span> <span class=nx>SDK</span><span class=p>.</span></span></span></code></pre></div></div></div><p>Note that the framework will automatically download the relevant Docker containers and startup the transform service for you.</p><p>Please see <a href=https://cwiki.apache.org/confluence/display/BEAM/Transform+Service#TransformService-Upgradetransformswithoutupgradingthepipeline>here</a> for a full example that uses this feature to upgrade BigQuery read and write transforms.</p><h3 id=transform-service-usage-multi-language>15.2 Using the Transform service for multi-language pipelines</h3><p>Transform service implements the Beam expansion API. This allows Beam multi-language pipelines to use the transform service when expanding transforms available within the transform service. |
| The main advantage here is that multi-language pipelines will be able to operate without installing support for additional language runtimes. For example, Beam Python pipelines that use Java transforms such as |
| <code>KafkaIO</code> can operate without installing Java locally during job submission as long as Docker is available in the system.</p><p>In some cases, Apache Beam SDKs can automatically start the Transform service.</p><ul><li><p>The Java <a href=https://github.com/apache/beam/blob/master/sdks/java/extensions/python/src/main/java/org/apache/beam/sdk/extensions/python/PythonExternalTransform.java><code>PythonExternalTransform</code> API</a> automatically |
| starts the Transform service when a Python runtime isn’t available locally, but Docker is.</p></li><li><p>The Apache Beam Python multi-language wrappers might automatically start the Transform service when you’re using Java transforms, a Java language runtime isn’t available locally, and Docker is available locally.</p></li></ul><p>Beam users also have the option to <a href=/documentation/programming-guide/#transform-service-usage-muanual>manually start</a> a transform service and use that as the expansion service used by multi-language pipelines.</p><h3 id=transform-service-usage-muanual>15.3 Manually starting the transform service</h3><p>A Beam Transform service instance can be manually started by using utilities provided with Apache Beam SDKs.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>java</span> <span class=o>-</span><span class=n>jar</span> <span class=n>beam</span><span class=o>-</span><span class=n>sdks</span><span class=o>-</span><span class=n>java</span><span class=o>-</span><span class=n>transform</span><span class=o>-</span><span class=n>service</span><span class=o>-</span><span class=n>app</span><span class=o>-<</span><span class=n>Beam</span> <span class=n>version</span> <span class=k>for</span> <span class=n>the</span> <span class=n>jar</span><span class=o>>.</span><span class=na>jar</span> <span class=o>--</span><span class=n>port</span> <span class=o><</span><span class=n>port</span><span class=o>></span> <span class=o>--</span><span class=n>beam_version</span> <span class=o><</span><span class=n>Beam</span> <span class=n>version</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>project_name</span> <span class=o><</span><span class=n>a</span> <span class=n>unique</span> <span class=n>ID</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>command</span> <span class=n>up</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>python</span> <span class=o>-</span><span class=n>m</span> <span class=n>apache_beam</span><span class=o>.</span><span class=n>utils</span><span class=o>.</span><span class=n>transform_service_launcher</span> <span class=o>--</span><span class=n>port</span> <span class=o><</span><span class=n>port</span><span class=o>></span> <span class=o>--</span><span class=n>beam_version</span> <span class=o><</span><span class=n>Beam</span> <span class=n>version</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>project_name</span> <span class=o><</span><span class=n>a</span> <span class=n>unique</span> <span class=n>ID</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>command</span> <span class=n>up</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>This</span> <span class=nx>feature</span> <span class=nx>is</span> <span class=nx>currently</span> <span class=nx>in</span> <span class=nx>development</span><span class=p>.</span></span></span></code></pre></div></div></div><p>To stop the transform service, use the following commands.</p><div class='language-java snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-java data-lang=java><span class=line><span class=cl><span class=n>java</span> <span class=o>-</span><span class=n>jar</span> <span class=n>beam</span><span class=o>-</span><span class=n>sdks</span><span class=o>-</span><span class=n>java</span><span class=o>-</span><span class=n>transform</span><span class=o>-</span><span class=n>service</span><span class=o>-</span><span class=n>app</span><span class=o>-<</span><span class=n>Beam</span> <span class=n>version</span> <span class=k>for</span> <span class=n>the</span> <span class=n>jar</span><span class=o>>.</span><span class=na>jar</span> <span class=o>--</span><span class=n>port</span> <span class=o><</span><span class=n>port</span><span class=o>></span> <span class=o>--</span><span class=n>beam_version</span> <span class=o><</span><span class=n>Beam</span> <span class=n>version</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>project_name</span> <span class=o><</span><span class=n>a</span> <span class=n>unique</span> <span class=n>ID</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>command</span> <span class=n>down</span></span></span></code></pre></div></div></div><div class='language-py snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-py data-lang=py><span class=line><span class=cl><span class=n>python</span> <span class=o>-</span><span class=n>m</span> <span class=n>apache_beam</span><span class=o>.</span><span class=n>utils</span><span class=o>.</span><span class=n>transform_service_launcher</span> <span class=o>--</span><span class=n>port</span> <span class=o><</span><span class=n>port</span><span class=o>></span> <span class=o>--</span><span class=n>beam_version</span> <span class=o><</span><span class=n>Beam</span> <span class=n>version</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>project_name</span> <span class=o><</span><span class=n>a</span> <span class=n>unique</span> <span class=n>ID</span> <span class=k>for</span> <span class=n>the</span> <span class=n>transform</span> <span class=n>service</span><span class=o>></span> <span class=o>--</span><span class=n>command</span> <span class=n>down</span></span></span></code></pre></div></div></div><div class='language-go snippet'><div class="notebook-skip code-snippet"><a class=copy type=button data-bs-toggle=tooltip data-bs-placement=bottom title="Copy to clipboard"><img src=/images/copy-icon.svg></a><div class=highlight><pre tabindex=0 class=chroma><code class=language-go data-lang=go><span class=line><span class=cl><span class=nx>This</span> <span class=nx>feature</span> <span class=nx>is</span> <span class=nx>currently</span> <span class=nx>in</span> <span class=nx>development</span><span class=p>.</span></span></span></code></pre></div></div></div><h3 id=transform-service-included-transforms>15.4 Portable transforms included in the Transform service</h3><p>Beam Transform service includes a number of transforms implemented in the Apache Beam Java and Python SDKs.</p><p>Currently, the following transforms are included in the Transform service:</p><ul><li><p>Java transforms: Google Cloud I/O connectors, the Kafka I/O connector, and the JDBC I/O connector</p></li><li><p>Python transforms: all portable transforms implemented within the Apache Beam Python SDK, such as |
| <a href=/documentation/transforms/python/elementwise/runinference/>RunInference</a> and |
| <a href=/documentation/dsls/dataframes/overview/>DataFrame</a> transforms.</p></li></ul><p>For a more comprehensive list of available transforms, see the |
| <a href=https://cwiki.apache.org/confluence/display/BEAM/Transform+Service#TransformService-TransformsincludedintheTransformservice>Transform service</a> developer guide.</p><div class=feedback><p class=update>Last updated on 2024/04/23</p><h3>Have you found everything you were looking for?</h3><p class=description>Was it all useful and clear? Is there anything that you would like to change? Let us know!</p><button class=load-button><a href="https://docs.google.com/forms/d/e/1FAIpQLSfID7abne3GE6k6RdJIyZhPz2Gef7UkpggUEhTIDjjplHuxSA/viewform?usp=header_link" target=_blank>SEND FEEDBACK</a></button></div></div></div><footer class=footer><div class=footer__contained><div class=footer__cols><div class="footer__cols__col footer__cols__col__logos"><div class=footer__cols__col__logo><img src=/images/beam_logo_circle.svg class=footer__logo alt="Beam logo"></div><div class=footer__cols__col__logo><img src=/images/apache_logo_circle.svg class=footer__logo alt="Apache logo"></div></div><div class=footer-wrapper><div class=wrapper-grid><div class=footer__cols__col><div class=footer__cols__col__title>Start</div><div class=footer__cols__col__link><a href=/get-started/beam-overview/>Overview</a></div><div class=footer__cols__col__link><a href=/get-started/quickstart-java/>Quickstart (Java)</a></div><div class=footer__cols__col__link><a href=/get-started/quickstart-py/>Quickstart (Python)</a></div><div class=footer__cols__col__link><a href=/get-started/quickstart-go/>Quickstart (Go)</a></div><div class=footer__cols__col__link><a href=/get-started/downloads/>Downloads</a></div></div><div class=footer__cols__col><div class=footer__cols__col__title>Docs</div><div class=footer__cols__col__link><a href=/documentation/programming-guide/>Concepts</a></div><div class=footer__cols__col__link><a href=/documentation/pipelines/design-your-pipeline/>Pipelines</a></div><div class=footer__cols__col__link><a href=/documentation/runners/capability-matrix/>Runners</a></div></div><div class=footer__cols__col><div class=footer__cols__col__title>Community</div><div class=footer__cols__col__link><a href=/contribute/>Contribute</a></div><div class=footer__cols__col__link><a href=https://projects.apache.org/committee.html?beam target=_blank>Team<img src=/images/external-link-icon.png width=14 height=14 alt="External link."></a></div><div class=footer__cols__col__link><a href=/community/presentation-materials/>Media</a></div><div class=footer__cols__col__link><a href=/community/in-person/>Events/Meetups</a></div><div class=footer__cols__col__link><a href=/community/contact-us/>Contact Us</a></div></div><div class=footer__cols__col><div class=footer__cols__col__title>Resources</div><div class=footer__cols__col__link><a href=/blog/>Blog</a></div><div class=footer__cols__col__link><a href=https://github.com/apache/beam>GitHub</a></div></div></div><div class=footer__bottom>© |
| <a href=https://www.apache.org>The Apache Software Foundation</a> |
| | <a href=/privacy_policy>Privacy Policy</a> |
| | <a href=/feed.xml>RSS Feed</a><br><br>Apache Beam, Apache, Beam, the Beam logo, and the Apache feather logo are either registered trademarks or trademarks of The Apache Software Foundation. All other products or name brands are trademarks of their respective holders, including The Apache Software Foundation.</div></div><div class="footer__cols__col footer__cols__col__logos"><div class=footer__cols__col--group><div class=footer__cols__col__logo><a href=https://github.com/apache/beam><img src=/images/logos/social-icons/github-logo-150.png class=footer__logo alt="Github logo"></a></div><div class=footer__cols__col__logo><a href=https://www.linkedin.com/company/apache-beam/><img src=/images/logos/social-icons/linkedin-logo-150.png class=footer__logo alt="Linkedin logo"></a></div></div><div class=footer__cols__col--group><div class=footer__cols__col__logo><a href=https://twitter.com/apachebeam><img src=/images/logos/social-icons/twitter-logo-150.png class=footer__logo alt="Twitter logo"></a></div><div class=footer__cols__col__logo><a href=https://www.youtube.com/channel/UChNnb_YO_7B0HlW6FhAXZZQ><img src=/images/logos/social-icons/youtube-logo-150.png class=footer__logo alt="Youtube logo"></a></div></div></div></div></div></footer></body></html> |
