| <!doctype html><html lang=en-US dir=ltr class="docs-wrapper plugin-docs plugin-id-default docs-version-0.10 docs-doc-page docs-doc-id-specification/xlang_serialization_spec" data-has-hydrated=false><head><meta charset=UTF-8><meta name=generator content="Docusaurus v3.10.1"><title data-rh=true>Fury Xlang Serialization Format | Apache Fory™</title><meta data-rh=true name=viewport content="width=device-width, initial-scale=1.0"/><meta data-rh=true property=og:url content=https://fory.apache.org/docs/0.10/specification/xlang_serialization_spec /><meta data-rh=true property=og:locale content=en_US /><meta data-rh=true property=og:locale:alternate content=zh_CN /><meta data-rh=true name=docusaurus_locale content=en-US /><meta data-rh=true name=docsearch:language content=en-US /><meta data-rh=true http-equiv=Content-Security-Policy content="frame-src 'self' https://ghbtns.com/;"/><meta data-rh=true property=og:image content=https://fory.apache.org/img/logo.png /><meta data-rh=true property=og:image:width content=1500 /><meta data-rh=true property=og:image:height content=1500 /><meta data-rh=true name=twitter:card content=summary_large_image /><meta data-rh=true name=twitter:image content=https://fory.apache.org/img/logo.png /><meta data-rh=true name=docusaurus_version content=0.10 /><meta data-rh=true name=docusaurus_tag content=docs-default-0.10 /><meta data-rh=true name=docsearch:version content=0.10 /><meta data-rh=true name=docsearch:docusaurus_tag content=docs-default-0.10 /><meta data-rh=true property=og:title content="Fury Xlang Serialization Format | Apache Fory™"/><meta data-rh=true name=description content="Cross-language Serialization Specification"/><meta data-rh=true property=og:description content="Cross-language Serialization Specification"/><link data-rh=true rel=icon href=/img/favicon.ico /><link data-rh=true rel=canonical href=https://fory.apache.org/docs/0.10/specification/xlang_serialization_spec /><link data-rh=true rel=alternate href=https://fory.apache.org/docs/0.10/specification/xlang_serialization_spec hreflang=en-US /><link data-rh=true rel=alternate href=https://fory.apache.org/zh-CN/docs/0.10/specification/xlang_serialization_spec hreflang=zh-CN /><link data-rh=true rel=alternate href=https://fory.apache.org/docs/0.10/specification/xlang_serialization_spec hreflang=x-default /><script data-rh=true type=application/ld+json>{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","item":"https://fory.apache.org/docs/0.10/specification/xlang_serialization_spec","name":"Fury Xlang Serialization Format","position":1}]}</script><link rel=alternate type=application/rss+xml href=/blog/rss.xml title="Apache Fory™ RSS Feed"><link rel=alternate type=application/atom+xml href=/blog/atom.xml title="Apache Fory™ Atom Feed"><script type=text/javascript>"fury.apache.org"===window.location.host&&(window.location.href="https://fory.apache.org"),function(){for(var o=window.location.pathname,a=["guide","introduction","start"],n=0;n<a.length;n++){var t=a[n],i=RegExp("^((?:/[a-z]{2}-[A-Z]{2})?/docs(?:/(?:next|[0-9]+.[0-9]+))?)/docs/"+t+"(/.*)?$"),c=o.match(i);if(c){var r=c[1]+"/"+t+(c[2]||"");window.location.replace(r+window.location.search+window.location.hash);return}}}()</script><link rel=stylesheet href=/assets/css/styles.3a1edaa1.css /><script src=/assets/js/runtime~main.605b11b1.js defer></script><script src=/assets/js/main.283fea4b.js defer></script></head><body><svg style="display: none;"><defs> |
| <symbol id=theme-svg-external-link viewBox="0 0 24 24"><path fill=currentColor d="M21 13v10h-21v-19h12v2h-10v15h17v-8h2zm3-12h-10.988l4.035 4-6.977 7.07 2.828 2.828 6.977-7.07 4.125 4.172v-11z"/></symbol> |
| </defs></svg> |
| <script>!function(){var t=function(){try{return new URLSearchParams(window.location.search).get("docusaurus-theme")}catch(t){}}()||function(){try{return window.localStorage.getItem("theme")}catch(t){}}();document.documentElement.setAttribute("data-theme",t||"light"),document.documentElement.setAttribute("data-theme-choice",t||"light")}(),function(){try{for(var[t,e]of new URLSearchParams(window.location.search).entries())if(t.startsWith("docusaurus-data-")){var a=t.replace("docusaurus-data-","data-");document.documentElement.setAttribute(a,e)}}catch(t){}}()</script><div id=__docusaurus><div role=region aria-label="Skip to main content"><a class=skipToContent_fXgn href=#__docusaurus_skipToContent_fallback>Skip to main content</a></div><nav aria-label=Main class="theme-layout-navbar navbar navbar--fixed-top"><div class=navbar__inner><div class="theme-layout-navbar-left navbar__items"><button aria-label="Toggle navigation bar" aria-expanded=false class="navbar__toggle clean-btn" type=button><svg width=30 height=30 viewBox="0 0 30 30" aria-hidden=true><path stroke=currentColor stroke-linecap=round stroke-miterlimit=10 stroke-width=2 d="M4 7h22M4 15h22M4 23h22"/></svg></button><a class=navbar__brand href=/><div class=navbar__logo><img src=/img/fory-logo-light.png alt="Apache Fory™ Logo" class="themedComponent_mlkZ themedComponent--light_NVdE"/><img src=/img/fory-logo-dark.png alt="Apache Fory™ Logo" class="themedComponent_mlkZ themedComponent--dark_xIcU"/></div><b class="navbar__title text--truncate"></b></a><a class="navbar__item navbar__link" href=/docs/0.10/introduction/overview>Docs</a><a aria-current=page class="navbar__item navbar__link navbar__link--active" href=/docs/0.10/specification/xlang_serialization_spec>Specification</a><a class="navbar__item navbar__link" href=/docs/0.10/community/>Community</a><a class="navbar__item navbar__link" href=/user>Users</a><a class="navbar__item navbar__link" href=/download>Download</a><a class="navbar__item navbar__link" href=/blog>Blog</a></div><div class="theme-layout-navbar-right navbar__items navbar__items--right"><div class="navbar__item dropdown dropdown--hoverable dropdown--right"><a href=# aria-haspopup=true aria-expanded=false role=button class=navbar__link>ASF</a><ul class=dropdown__menu><li><a href=https://www.apache.org/ target=_blank rel="noopener noreferrer" class=dropdown__link>Foundation</a><li><a href=https://www.apache.org/licenses/ target=_blank rel="noopener noreferrer" class=dropdown__link>License</a><li><a href=https://www.apache.org/events/current-event.html target=_blank rel="noopener noreferrer" class=dropdown__link>Events</a><li><a href=https://privacy.apache.org/policies/privacy-policy-public.html target=_blank rel="noopener noreferrer" class=dropdown__link>Privacy</a><li><a href=https://www.apache.org/security/ target=_blank rel="noopener noreferrer" class=dropdown__link>Security</a><li><a href=https://www.apache.org/foundation/sponsorship.html target=_blank rel="noopener noreferrer" class=dropdown__link>Sponsorship</a><li><a href=https://www.apache.org/foundation/thanks.html target=_blank rel="noopener noreferrer" class=dropdown__link>Thanks</a><li><a href=https://www.apache.org/foundation/policies/conduct.html target=_blank rel="noopener noreferrer" class=dropdown__link>Code of Conduct</a></ul></div><div class="navbar__item dropdown dropdown--hoverable dropdown--right"><a class=navbar__link aria-haspopup=true aria-expanded=false role=button href=/docs/0.10/specification/xlang_serialization_spec>0.10</a><ul class=dropdown__menu><li><a class=dropdown__link href=/docs/next/specification/xlang_serialization_spec>dev</a><li><a class=dropdown__link href=/docs/specification/xlang_serialization_spec>1.3.0</a><li><a class=dropdown__link href=/docs/1.2.0/specification/xlang_serialization_spec>1.2.0</a><li><a class=dropdown__link href=/docs/1.1.0/specification/xlang_serialization_spec>1.1.0</a><li><a class=dropdown__link href=/docs/1.0.0/specification/xlang_serialization_spec>1.0.0</a><li><a class=dropdown__link href=/docs/0.17/specification/xlang_serialization_spec>0.17</a><li><a class=dropdown__link href=/docs/0.16/specification/xlang_serialization_spec>0.16</a><li><a class=dropdown__link href=/docs/0.15/specification/xlang_serialization_spec>0.15</a><li><a class=dropdown__link href=/docs/0.14/specification/xlang_serialization_spec>0.14</a><li><a class=dropdown__link href=/docs/0.13/specification/xlang_serialization_spec>0.13</a><li><a class=dropdown__link href=/docs/0.12/specification/xlang_serialization_spec>0.12</a><li><a class=dropdown__link href=/docs/0.11/specification/xlang_serialization_spec>0.11</a><li><a aria-current=page class="dropdown__link dropdown__link--active" href=/docs/0.10/specification/xlang_serialization_spec>0.10</a></ul></div><a href=https://github.com/apache/fory target=_blank rel="noopener noreferrer" class="navbar__item navbar__link header-github-link" aria-label="GitHub repository"></a><div class="navbar__item dropdown dropdown--hoverable dropdown--right"><a href=# aria-haspopup=true aria-expanded=false role=button class=navbar__link><svg viewBox="0 0 24 24" width=20 height=20 aria-hidden=true class=iconLanguage_nlXk><path fill=currentColor d="M12.87 15.07l-2.54-2.51.03-.03c1.74-1.94 2.98-4.17 3.71-6.53H17V4h-7V2H8v2H1v1.99h11.17C11.5 7.92 10.44 9.75 9 11.35 8.07 10.32 7.3 9.19 6.69 8h-2c.73 1.63 1.73 3.17 2.98 4.56l-5.09 5.02L4 19l5-5 3.11 3.11.76-2.04zM18.5 10h-2L12 22h2l1.12-3h4.75L21 22h2l-4.5-12zm-2.62 7l1.62-4.33L19.12 17h-3.24z"/></svg>English</a><ul class=dropdown__menu><li><a href=/docs/0.10/specification/xlang_serialization_spec target=_self rel="noopener noreferrer" class="dropdown__link dropdown__link--active" lang=en-US>English</a><li><a href=/zh-CN/docs/0.10/specification/xlang_serialization_spec target=_self rel="noopener noreferrer" class=dropdown__link lang=zh-CN>简体中文</a></ul></div><div class="toggle_vylO colorModeToggle_DEke"><button class="clean-btn toggleButton_gllP toggleButtonDisabled_aARS" type=button disabled title="system mode" aria-label="Switch between dark and light mode (currently system mode)"><svg viewBox="0 0 24 24" width=24 height=24 aria-hidden=true class="toggleIcon_g3eP lightToggleIcon_pyhR"><path fill=currentColor d="M12,9c1.65,0,3,1.35,3,3s-1.35,3-3,3s-3-1.35-3-3S10.35,9,12,9 M12,7c-2.76,0-5,2.24-5,5s2.24,5,5,5s5-2.24,5-5 S14.76,7,12,7L12,7z M2,13l2,0c0.55,0,1-0.45,1-1s-0.45-1-1-1l-2,0c-0.55,0-1,0.45-1,1S1.45,13,2,13z M20,13l2,0c0.55,0,1-0.45,1-1 s-0.45-1-1-1l-2,0c-0.55,0-1,0.45-1,1S19.45,13,20,13z M11,2v2c0,0.55,0.45,1,1,1s1-0.45,1-1V2c0-0.55-0.45-1-1-1S11,1.45,11,2z M11,20v2c0,0.55,0.45,1,1,1s1-0.45,1-1v-2c0-0.55-0.45-1-1-1C11.45,19,11,19.45,11,20z M5.99,4.58c-0.39-0.39-1.03-0.39-1.41,0 c-0.39,0.39-0.39,1.03,0,1.41l1.06,1.06c0.39,0.39,1.03,0.39,1.41,0s0.39-1.03,0-1.41L5.99,4.58z M18.36,16.95 c-0.39-0.39-1.03-0.39-1.41,0c-0.39,0.39-0.39,1.03,0,1.41l1.06,1.06c0.39,0.39,1.03,0.39,1.41,0c0.39-0.39,0.39-1.03,0-1.41 L18.36,16.95z M19.42,5.99c0.39-0.39,0.39-1.03,0-1.41c-0.39-0.39-1.03-0.39-1.41,0l-1.06,1.06c-0.39,0.39-0.39,1.03,0,1.41 s1.03,0.39,1.41,0L19.42,5.99z M7.05,18.36c0.39-0.39,0.39-1.03,0-1.41c-0.39-0.39-1.03-0.39-1.41,0l-1.06,1.06 c-0.39,0.39-0.39,1.03,0,1.41s1.03,0.39,1.41,0L7.05,18.36z"/></svg><svg viewBox="0 0 24 24" width=24 height=24 aria-hidden=true class="toggleIcon_g3eP darkToggleIcon_wfgR"><path fill=currentColor d="M9.37,5.51C9.19,6.15,9.1,6.82,9.1,7.5c0,4.08,3.32,7.4,7.4,7.4c0.68,0,1.35-0.09,1.99-0.27C17.45,17.19,14.93,19,12,19 c-3.86,0-7-3.14-7-7C5,9.07,6.81,6.55,9.37,5.51z M12,3c-4.97,0-9,4.03-9,9s4.03,9,9,9s9-4.03,9-9c0-0.46-0.04-0.92-0.1-1.36 c-0.98,1.37-2.58,2.26-4.4,2.26c-2.98,0-5.4-2.42-5.4-5.4c0-1.81,0.89-3.42,2.26-4.4C12.92,3.04,12.46,3,12,3L12,3z"/></svg><svg viewBox="0 0 24 24" width=24 height=24 aria-hidden=true class="toggleIcon_g3eP systemToggleIcon_QzmC"><path fill=currentColor d="m12 21c4.971 0 9-4.029 9-9s-4.029-9-9-9-9 4.029-9 9 4.029 9 9 9zm4.95-13.95c1.313 1.313 2.05 3.093 2.05 4.95s-0.738 3.637-2.05 4.95c-1.313 1.313-3.093 2.05-4.95 2.05v-14c1.857 0 3.637 0.737 4.95 2.05z"/></svg></button></div><div class=navbarSearchContainer_Bca1><div class=navbar__search><span aria-label="expand searchbar" role=button class=search-icon tabindex=0></span><input id=search_input_react type=search placeholder=Loading... aria-label=Search class="navbar__search-input search-bar" disabled/></div></div></div></div><div role=presentation class=navbar-sidebar__backdrop></div></nav><div id=__docusaurus_skipToContent_fallback class="theme-layout-main main-wrapper mainWrapper_z2l0"><div class=docsWrapper_hBAB><button aria-label="Scroll back to top" class="clean-btn theme-back-to-top-button backToTopButton_sjWU" type=button></button><div class=docRoot_UBD9><aside class="theme-doc-sidebar-container docSidebarContainer_YfHR"><div class=sidebarViewport_aRkj><div class=sidebar_njMd><nav aria-label="Docs sidebar" class="menu thin-scrollbar menu_SIkG"><ul class="theme-doc-sidebar-menu menu__list"><li class="theme-doc-sidebar-item-link theme-doc-sidebar-item-link-level-1 menu__list-item"><a class="menu__link menu__link--active" aria-current=page href=/docs/0.10/specification/xlang_serialization_spec><span title="Fury Xlang Serialization Format" class=linkLabel_WmDU>Fury Xlang Serialization Format</span></a><li class="theme-doc-sidebar-item-link theme-doc-sidebar-item-link-level-1 menu__list-item"><a class=menu__link href=/docs/0.10/specification/java_serialization_spec><span title="Fury Java Serialization Format" class=linkLabel_WmDU>Fury Java Serialization Format</span></a><li class="theme-doc-sidebar-item-link theme-doc-sidebar-item-link-level-1 menu__list-item"><a class=menu__link href=/docs/0.10/specification/row_format_spec><span title="Fury Row Format" class=linkLabel_WmDU>Fury Row Format</span></a></ul></nav></div></div></aside><main class=docMainContainer_TBSr><div class="container padding-top--md padding-bottom--lg"><div class=row><div class="col docItemCol_VOVn"><div class="theme-doc-version-banner alert alert--warning margin-bottom--md" role=alert><div>This is documentation for <!-- -->Apache Fory™<!-- --> <b>0.10</b>, which is no longer actively maintained.</div><div class=margin-top--md>For up-to-date documentation, see the <b><a href=/docs/specification/xlang_serialization_spec>latest version</a></b> (<!-- -->1.3.0<!-- -->).</div></div><div class=docItemContainer_Djhp><article><nav class="theme-doc-breadcrumbs breadcrumbsContainer_Z_bl" aria-label=Breadcrumbs><ul class=breadcrumbs><li class=breadcrumbs__item><a aria-label="Home page" class=breadcrumbs__link href=/><svg viewBox="0 0 24 24" class=breadcrumbHomeIcon_YNFT><path d="M10 19v-5h4v5c0 .55.45 1 1 1h3c.55 0 1-.45 1-1v-7h1.7c.46 0 .68-.57.33-.87L12.67 3.6c-.38-.34-.96-.34-1.34 0l-8.36 7.53c-.34.3-.13.87.33.87H5v7c0 .55.45 1 1 1h3c.55 0 1-.45 1-1z" fill=currentColor /></svg></a><li class="breadcrumbs__item breadcrumbs__item--active"><span class=breadcrumbs__link>Fury Xlang Serialization Format</span></ul></nav><span class="theme-doc-version-badge badge badge--secondary">Version: 0.10</span><div class="tocCollapsible_ETCw theme-doc-toc-mobile tocMobile_ITEo"><button type=button class="clean-btn tocCollapsibleButton_TO0P">On this page</button></div><div class="theme-doc-markdown markdown"><header><h1>Fury Xlang Serialization Format</h1></header><h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=cross-language-serialization-specification>Cross-language Serialization Specification<a href=#cross-language-serialization-specification class=hash-link aria-label="Direct link to Cross-language Serialization Specification" title="Direct link to Cross-language Serialization Specification" translate=no></a></h2> |
| <blockquote> |
| <p>Format Version History:</p> |
| <ul> |
| <li class="">Version 0.1 - serialization spec formalized</li> |
| </ul> |
| </blockquote> |
| <p>Fury xlang serialization is an automatic object serialization framework that supports reference and polymorphism. |
| Fury will convert an object from/to fury xlang serialization binary format. |
| Fury has two core concepts for xlang serialization:</p> |
| <ul> |
| <li class=""><strong>Fury xlang binary format</strong></li> |
| <li class=""><strong>Framework implemented in different languages to convert object to/from Fury xlang binary format</strong></li> |
| </ul> |
| <p>The serialization format is a dynamic binary format. The dynamics and reference/polymorphism support make Fury flexible, |
| much more easy to use, but |
| also introduce more complexities compared to static serialization frameworks. So the format will be more complex.</p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=type-systems>Type Systems<a href=#type-systems class=hash-link aria-label="Direct link to Type Systems" title="Direct link to Type Systems" translate=no></a></h2> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=data-types>Data Types<a href=#data-types class=hash-link aria-label="Direct link to Data Types" title="Direct link to Data Types" translate=no></a></h3> |
| <ul> |
| <li class="">bool: a boolean value (true or false).</li> |
| <li class="">int8: a 8-bit signed integer.</li> |
| <li class="">int16: a 16-bit signed integer.</li> |
| <li class="">int32: a 32-bit signed integer.</li> |
| <li class="">var_int32: a 32-bit signed integer which use fury var_int32 encoding.</li> |
| <li class="">int64: a 64-bit signed integer.</li> |
| <li class="">var_int64: a 64-bit signed integer which use fury PVL encoding.</li> |
| <li class="">sli_int64: a 64-bit signed integer which use fury SLI encoding.</li> |
| <li class="">float16: a 16-bit floating point number.</li> |
| <li class="">float32: a 32-bit floating point number.</li> |
| <li class="">float64: a 64-bit floating point number including NaN and Infinity.</li> |
| <li class="">string: a text string encoded using Latin1/UTF16/UTF-8 encoding.</li> |
| <li class="">enum: a data type consisting of a set of named values. Rust enum with non-predefined field values are not supported as |
| an enum.</li> |
| <li class="">named_enum: an enum whose value will be serialized as the registered name.</li> |
| <li class="">struct: a morphic(final) type serialized by Fury Struct serializer. i.e. it doesn't have subclasses. Suppose we're |
| deserializing <code>List<SomeClass></code>, we can save dynamic serializer dispatch since <code>SomeClass</code> is morphic(final).</li> |
| <li class="">compatible_struct: a morphic(final) type serialized by Fury compatible Struct serializer.</li> |
| <li class="">named_struct: a <code>struct</code> whose type mapping will be encoded as a name.</li> |
| <li class="">named_compatible_struct: a <code>compatible_struct</code> whose type mapping will be encoded as a name.</li> |
| <li class="">ext: a type which will be serialized by a customized serializer.</li> |
| <li class="">named_ext: an <code>ext</code> type whose type mapping will be encoded as a name.</li> |
| <li class="">list: a sequence of objects.</li> |
| <li class="">set: an unordered set of unique elements.</li> |
| <li class="">map: a map of key-value pairs. Mutable types such as <code>list/map/set/array/tensor/arrow</code> are not allowed as key of map.</li> |
| <li class="">duration: an absolute length of time, independent of any calendar/timezone, as a count of nanoseconds.</li> |
| <li class="">timestamp: a point in time, independent of any calendar/timezone, as a count of nanoseconds. The count is relative |
| to an epoch at UTC midnight on January 1, 1970.</li> |
| <li class="">local_date: a naive date without timezone. The count is days relative to an epoch at UTC midnight on Jan 1, 1970.</li> |
| <li class="">decimal: exact decimal value represented as an integer value in two's complement.</li> |
| <li class="">binary: an variable-length array of bytes.</li> |
| <li class="">array: only allow 1d numeric components. Other arrays will be taken as List. The implementation should support the |
| interoperability between array and list.<!-- --> |
| <ul> |
| <li class="">bool_array: one dimensional bool array.</li> |
| <li class="">int8_array: one dimensional int8 array.</li> |
| <li class="">int16_array: one dimensional int16 array.</li> |
| <li class="">int32_array: one dimensional int32 array.</li> |
| <li class="">int64_array: one dimensional int64 array.</li> |
| <li class="">float16_array: one dimensional half_float_16 array.</li> |
| <li class="">float32_array: one dimensional float32 array.</li> |
| <li class="">float64_array: one dimensional float64 array.</li> |
| </ul> |
| </li> |
| <li class="">tensor: multidimensional array which every sub-array have same size and type.</li> |
| <li class="">arrow record batch: an arrow <a href=https://arrow.apache.org/docs/cpp/tables.html#record-batches target=_blank rel="noopener noreferrer" class="">record batch</a> object.</li> |
| <li class="">arrow table: an arrow <a href=https://arrow.apache.org/docs/cpp/tables.html#tables target=_blank rel="noopener noreferrer" class="">table</a> object.</li> |
| </ul> |
| <p>Note:</p> |
| <ul> |
| <li class="">Unsigned int/long are not added here, since not every language support those types.</li> |
| </ul> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=polymorphisms>Polymorphisms<a href=#polymorphisms class=hash-link aria-label="Direct link to Polymorphisms" title="Direct link to Polymorphisms" translate=no></a></h3> |
| <p>For polymorphism, if one non-final class is registered, and only one subclass is registered, then we can take all |
| elements in List/Map have same type, thus reduce runtime check cost.</p> |
| <p>Collection/Array polymorphism are not fully supported, since some languages such as golang have only one collection |
| type. If users want to get exactly the type he passed, he must pass that type when deserializing or annotate that type |
| to the field of struct.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=type-disambiguation>Type disambiguation<a href=#type-disambiguation class=hash-link aria-label="Direct link to Type disambiguation" title="Direct link to Type disambiguation" translate=no></a></h3> |
| <p>Due to differences between type systems of languages, those types can't be mapped one-to-one between languages. When |
| deserializing, Fury use the target data structure type and the data type in the data jointly to determine how to |
| deserialize and populate the target data structure. For example:</p> |
| <div class="language-java codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-java codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token keyword" style=color:#00009f>class</span><span class="token plain"> </span><span class="token class-name">Foo</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>{</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>int</span><span class="token punctuation" style=color:#393A34>[</span><span class="token punctuation" style=color:#393A34>]</span><span class="token plain"> intArray</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token class-name">Object</span><span class="token punctuation" style=color:#393A34>[</span><span class="token punctuation" style=color:#393A34>]</span><span class="token plain"> objects</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token class-name">List</span><span class="token generics punctuation" style=color:#393A34><</span><span class="token generics class-name">Object</span><span class="token generics punctuation" style=color:#393A34>></span><span class="token plain"> objectList</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token punctuation" style=color:#393A34>}</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain" style=display:inline-block></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token keyword" style=color:#00009f>class</span><span class="token plain"> </span><span class="token class-name">Foo2</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>{</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>int</span><span class="token punctuation" style=color:#393A34>[</span><span class="token punctuation" style=color:#393A34>]</span><span class="token plain"> intArray</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token class-name">List</span><span class="token generics punctuation" style=color:#393A34><</span><span class="token generics class-name">Object</span><span class="token generics punctuation" style=color:#393A34>></span><span class="token plain"> objects</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token class-name">List</span><span class="token generics punctuation" style=color:#393A34><</span><span class="token generics class-name">Object</span><span class="token generics punctuation" style=color:#393A34>></span><span class="token plain"> objectList</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token punctuation" style=color:#393A34>}</span><br/></div></code></pre></div></div> |
| <p><code>intArray</code> has an <code>int32_array</code> type. But both <code>objects</code> and <code>objectList</code> fields in the serialize data have <code>list</code> data |
| type. When deserializing, the implementation will create an <code>Object</code> array for <code>objects</code>, but create a <code>ArrayList</code> |
| for <code>objectList</code> to populate its elements. And the serialized data of <code>Foo</code> can be deserialized into <code>Foo2</code> too.</p> |
| <p>Users can also provide meta hints for fields of a type, or the type whole. Here is an example in java which use |
| annotation to provide such information.</p> |
| <div class="language-java codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-java codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token annotation punctuation" style=color:#393A34>@FuryObject</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">fieldsNullable </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token boolean" style=color:#36acaa>false</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> trackingRef </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token boolean" style=color:#36acaa>false</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token keyword" style=color:#00009f>class</span><span class="token plain"> </span><span class="token class-name">Foo</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>{</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token annotation punctuation" style=color:#393A34>@FuryField</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">trackingRef </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token boolean" style=color:#36acaa>false</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>int</span><span class="token punctuation" style=color:#393A34>[</span><span class="token punctuation" style=color:#393A34>]</span><span class="token plain"> intArray</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token annotation punctuation" style=color:#393A34>@FuryField</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">polymorphic </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token boolean" style=color:#36acaa>true</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token class-name">Object</span><span class="token plain"> object</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token annotation punctuation" style=color:#393A34>@FuryField</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">tagId </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token number" style=color:#36acaa>1</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> nullable </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token boolean" style=color:#36acaa>true</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token class-name">List</span><span class="token generics punctuation" style=color:#393A34><</span><span class="token generics class-name">Object</span><span class="token generics punctuation" style=color:#393A34>></span><span class="token plain"> objectList</span><span class="token punctuation" style=color:#393A34>;</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token punctuation" style=color:#393A34>}</span><br/></div></code></pre></div></div> |
| <p>Such information can be provided in other languages too:</p> |
| <ul> |
| <li class="">cpp: use macro and template.</li> |
| <li class="">golang: use struct tag.</li> |
| <li class="">python: use typehint.</li> |
| <li class="">rust: use macro.</li> |
| </ul> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=type-id>Type ID<a href=#type-id class=hash-link aria-label="Direct link to Type ID" title="Direct link to Type ID" translate=no></a></h3> |
| <p>All internal data types are expressed using an ID in range <code>0~64</code>. Users can use <code>0~4096</code> for representing their |
| types.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=type-mapping>Type mapping<a href=#type-mapping class=hash-link aria-label="Direct link to Type mapping" title="Direct link to Type mapping" translate=no></a></h3> |
| <p>See <a class="" href=/docs/0.10/guide/xlang_type_mapping>Type mapping</a></p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=spec-overview>Spec overview<a href=#spec-overview class=hash-link aria-label="Direct link to Spec overview" title="Direct link to Spec overview" translate=no></a></h2> |
| <p>Here is the overall format:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| fury header | object ref meta | object type meta | object value data |</span><br/></div></code></pre></div></div> |
| <p>The data are serialized using little endian byte order overall. If bytes swap is costly for some object, |
| Fury will write the byte order for that object into the data instead of converting it to little endian.</p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=fury-header>Fury header<a href=#fury-header class=hash-link aria-label="Direct link to Fury header" title="Direct link to Fury header" translate=no></a></h2> |
| <p>Fury header consists starts one byte:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| 2 bytes | 4 bits | 1 bit | 1 bit | 1 bit | 1 bit | 1 byte | optional 4 bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+--------------+---------------+-------+-------+--------+-------+------------+------------------------------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| magic number | reserved bits | oob | xlang | endian | null | language | unsigned int for meta start offset |</span><br/></div></code></pre></div></div> |
| <ul> |
| <li class="">magic number: used to identify fury serialization protocol, current version use <code>0x62d4</code>.</li> |
| <li class="">null flag: 1 when object is null, 0 otherwise. If an object is null, other bits won't be set.</li> |
| <li class="">endian flag: 1 when data is encoded by little endian, 0 for big endian.</li> |
| <li class="">xlang flag: 1 when serialization uses xlang format, 0 when serialization uses Fury java format.</li> |
| <li class="">oob flag: 1 when passed <code>BufferCallback</code> is not null, 0 otherwise.</li> |
| <li class="">language: the language when serializing objects, such as JAVA, PYTHON, GO, etc. Fury can use this flag to determine whether spend more time on serialization to make the deserialization faster for dynamic languages.</li> |
| </ul> |
| <p>If meta share mode is enabled, an uncompressed unsigned int is appended to indicate the start offset of metadata.</p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=reference-meta>Reference Meta<a href=#reference-meta class=hash-link aria-label="Direct link to Reference Meta" title="Direct link to Reference Meta" translate=no></a></h2> |
| <p>Reference tracking handles whether the object is null, and whether to track reference for the object by writing |
| corresponding flags and maintaining internal state.</p> |
| <p>Reference flags:</p> |
| <table><thead><tr><th>Flag<th>Byte Value<th>Description<tbody><tr><td>NULL FLAG<td><code>-3</code><td>This flag indicates the object is a null value. We don't use another byte to indicate REF, so that we can save one byte.<tr><td>REF FLAG<td><code>-2</code><td>This flag indicates the object is already serialized previously, and fury will write a ref id with unsigned varint format instead of serialize it again<tr><td>NOT_NULL VALUE FLAG<td><code>-1</code><td>This flag indicates the object is a non-null value and fury doesn't track ref for this type of object.<tr><td>REF VALUE FLAG<td><code>0</code><td>This flag indicates the object is referencable and the first time to serialize.</table> |
| <p>When reference tracking is disabled globally or for specific types, or for certain types within a particular |
| context(e.g., a field of a type), only the <code>NULL</code> and <code>NOT_NULL VALUE</code> flags will be used for reference meta.</p> |
| <p>For languages which doesn't support reference such as rust, reference tracking must be disabled for correct |
| deserialization by fury rust implementation.</p> |
| <p>For languages whose object values are not null by default:</p> |
| <ul> |
| <li class="">In rust, Fury takes <code>Option:None</code> as a null value</li> |
| <li class="">In c++, Fury takes <code>std::nullopt</code> as a null value</li> |
| <li class="">In golang, Fury takes <code>null interface/pointer</code> as a null value</li> |
| </ul> |
| <p>If one want to deserialize in languages like <code>Java/Python/JavaScript</code>, he should mark the type with all fields |
| not-null by default, or using schema-evolution mode to carry the not-null fields info in the data.</p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=type-meta>Type Meta<a href=#type-meta class=hash-link aria-label="Direct link to Type Meta" title="Direct link to Type Meta" translate=no></a></h2> |
| <p>For every type to be serialized, it have a type id to indicate its type.</p> |
| <ul> |
| <li class="">basic types: the type id</li> |
| <li class="">enum:<!-- --> |
| <ul> |
| <li class=""><code>Type.ENUM</code> + registered id</li> |
| <li class=""><code>Type.NAMED_ENUM</code> + registered namespace+typename</li> |
| </ul> |
| </li> |
| <li class="">list: <code>Type.List</code></li> |
| <li class="">set: <code>Type.SET</code></li> |
| <li class="">map: <code>Type.MAP</code></li> |
| <li class="">ext:<!-- --> |
| <ul> |
| <li class=""><code>Type.EXT</code> + registered id</li> |
| <li class=""><code>Type.NAMED_EXT</code> + registered namespace+typename</li> |
| </ul> |
| </li> |
| <li class="">struct:<!-- --> |
| <ul> |
| <li class=""><code>Type.STRUCT</code> + struct meta</li> |
| <li class=""><code>Type.NAMED_STRUCT</code> + struct meta</li> |
| </ul> |
| </li> |
| </ul> |
| <p>Every type must be registered with an ID or name first. The registration can be used for security check and type |
| identification.</p> |
| <p>Struct is a special type, depending whether schema compatibility is enabled, Fury will write struct meta |
| differently.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=struct-schema-consistent>Struct Schema consistent<a href=#struct-schema-consistent class=hash-link aria-label="Direct link to Struct Schema consistent" title="Direct link to Struct Schema consistent" translate=no></a></h3> |
| <ul> |
| <li class="">If schema consistent mode is enabled globally when creating fury, type meta will be written as a fury unsigned varint |
| of <code>type_id</code>. Schema evolution related meta will be ignored.</li> |
| <li class="">If schema evolution mode is enabled globally when creating fury, and current class is configured to use schema |
| consistent mode like <code>struct</code> vs <code>table</code> in flatbuffers:<!-- --> |
| <ul> |
| <li class="">Type meta will be add to <code>captured_type_defs</code>: <code>captured_type_defs[type def stub] = map size</code> ahead when |
| registering type.</li> |
| <li class="">Get index of the meta in <code>captured_type_defs</code>, write that index as <code>| unsigned varint: index |</code>.</li> |
| </ul> |
| </li> |
| </ul> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=struct-schema-evolution>Struct Schema evolution<a href=#struct-schema-evolution class=hash-link aria-label="Direct link to Struct Schema evolution" title="Direct link to Struct Schema evolution" translate=no></a></h3> |
| <p>If schema evolution mode is enabled globally when creating fury, and enabled for current type, type meta will be written |
| using one of the following mode. Which mode to use is configured when creating fury.</p> |
| <ul> |
| <li class=""> |
| <p>Normal mode(meta share not enabled):</p> |
| <ul> |
| <li class="">If type meta hasn't been written before, add <code>type def</code> |
| to <code>captured_type_defs</code>: <code>captured_type_defs[type def] = map size</code>.</li> |
| <li class="">Get index of the meta in <code>captured_type_defs</code>, write that index as <code>| unsigned varint: index |</code>.</li> |
| <li class="">After finished the serialization of the object graph, fury will start to write <code>captured_type_defs</code>:<!-- --> |
| <ul> |
| <li class=""> |
| <p>Firstly, set current to <code>meta start offset</code> of fury header</p> |
| </li> |
| <li class=""> |
| <p>Then write <code>captured_type_defs</code> one by one:</p> |
| <div class="language-python codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-python codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_var_uint32</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">len</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">writting_type_defs</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"> </span><span class="token operator" style=color:#393A34>-</span><span class="token plain"> </span><span class="token builtin">len</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">schema_consistent_type_def_stubs</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token keyword" style=color:#00009f>for</span><span class="token plain"> type_meta </span><span class="token keyword" style=color:#00009f>in</span><span class="token plain"> writting_type_defs</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>if</span><span class="token plain"> </span><span class="token keyword" style=color:#00009f>not</span><span class="token plain"> type_meta</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">is_stub</span><span class="token punctuation" style=color:#393A34>(</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> type_meta</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_type_def</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain">writing_type_defs </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> copy</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">schema_consistent_type_def_stubs</span><span class="token punctuation" style=color:#393A34>)</span><br/></div></code></pre></div></div> |
| </li> |
| </ul> |
| </li> |
| </ul> |
| </li> |
| <li class=""> |
| <p>Meta share mode: the writing steps are same as the normal mode, but <code>captured_type_defs</code> will be shared across |
| multiple serializations of different objects. For example, suppose we have a batch to serialize:</p> |
| <div class="language-python codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-python codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">captured_type_defs </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>{</span><span class="token punctuation" style=color:#393A34>}</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain">stream </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token comment" style=color:#999988;font-style:italic># add `Type1` to `captured_type_defs` and write `Type1`</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain">fury</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">serialize</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">stream</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>[</span><span class="token plain">Type1</span><span class="token punctuation" style=color:#393A34>(</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>]</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token comment" style=color:#999988;font-style:italic># add `Type2` to `captured_type_defs` and write `Type2`, `Type1` is written before.</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain">fury</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">serialize</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">stream</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>[</span><span class="token plain">Type1</span><span class="token punctuation" style=color:#393A34>(</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> Type2</span><span class="token punctuation" style=color:#393A34>(</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>]</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token comment" style=color:#999988;font-style:italic># `Type1` and `Type2` are written before, no need to write meta.</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain">fury</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">serialize</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">stream</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>[</span><span class="token plain">Type1</span><span class="token punctuation" style=color:#393A34>(</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> Type2</span><span class="token punctuation" style=color:#393A34>(</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>]</span><span class="token punctuation" style=color:#393A34>)</span><br/></div></code></pre></div></div> |
| </li> |
| <li class=""> |
| <p>Streaming mode(streaming mode doesn't support meta share):</p> |
| <ul> |
| <li class=""> |
| <p>If type meta hasn't been written before, the data will be written as:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| unsigned varint: 0b11111111 | type def |</span><br/></div></code></pre></div></div> |
| </li> |
| <li class=""> |
| <p>If type meta has been written before, the data will be written as:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| unsigned varint: written index << 1 |</span><br/></div></code></pre></div></div> |
| <p><code>written index</code> is the id in <code>captured_type_defs</code>.</p> |
| </li> |
| <li class=""> |
| <p>With this mode, <code>meta start offset</code> can be omitted.</p> |
| </li> |
| </ul> |
| </li> |
| </ul> |
| <blockquote> |
| <p>The normal mode and meta share mode will forbid streaming writing since it needs to look back for update the start |
| offset after the whole object graph writing and meta collecting is finished. Only in this way we can ensure |
| deserialization failure in meta share mode doesn't lost shared meta.</p> |
| </blockquote> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=type-def>Type Def<a href=#type-def class=hash-link aria-label="Direct link to Type Def" title="Direct link to Type Def" translate=no></a></h4> |
| <p>Here we mainly describe the meta layout for schema evolution mode:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| 8 bytes header | variable bytes | variable bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+----------------------+--------------------+-------------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| global binary header | meta header | fields meta |</span><br/></div></code></pre></div></div> |
| <p>For languages which support inheritance, if parent class and subclass has fields with same name, using field in |
| subclass.</p> |
| <h5 class="anchor anchorTargetStickyNavbar_Vzrq" id=global-binary-header>Global binary header<a href=#global-binary-header class=hash-link aria-label="Direct link to Global binary header" title="Direct link to Global binary header" translate=no></a></h5> |
| <p><code>50 bits hash + 1bit compress flag + write fields meta + 12 bits meta size</code>. Right is the lower bits.</p> |
| <ul> |
| <li class="">lower 12 bits are used to encode meta size. If meta size <code>>= 0b1111_1111_1111</code>, then write |
| <code>meta_ size - 0b1111_1111_1111</code> next.</li> |
| <li class="">13rd bit is used to indicate whether to write fields meta. When this class is schema-consistent or use registered |
| serializer, fields meta will be skipped. Class Meta will be used for share namespace + type name only.</li> |
| <li class="">14rd bit is used to indicate whether meta is compressed.</li> |
| <li class="">Other 50 bits is used to store the unique hash of <code>flags + all layers class meta</code>.</li> |
| </ul> |
| <h5 class="anchor anchorTargetStickyNavbar_Vzrq" id=meta-header>Meta header<a href=#meta-header class=hash-link aria-label="Direct link to Meta header" title="Direct link to Meta header" translate=no></a></h5> |
| <p>Meta header is a 8 bits number value.</p> |
| <ul> |
| <li class="">Lowest 5 digits <code>0b00000~0b11110</code> are used to record num fields. <code>0b11111</code> is preserved to indicate that Fury need to |
| read more bytes for length using Fury unsigned int encoding. Note that num_fields is the number of compatible fields. |
| Users can use tag id to mark some fields as compatible fields in schema consistent context. In such cases, schema |
| consistent fields will be serialized first, then compatible fields will be serialized next. At deserialization, |
| Fury will use fields info of those fields which aren't annotated by tag id for deserializing schema consistent |
| fields, then use fields info in meta for deserializing compatible fields.</li> |
| <li class="">The 6th bit: 0 for registered by id, 1 for registered by name.</li> |
| <li class="">Remaining 2 bits are reserved for future extension.</li> |
| </ul> |
| <h5 class="anchor anchorTargetStickyNavbar_Vzrq" id=fields-meta>Fields meta<a href=#fields-meta class=hash-link aria-label="Direct link to Fields meta" title="Direct link to Fields meta" translate=no></a></h5> |
| <p>Format:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| field info: variable bytes | variable bytes | ... |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+---------------------------------+-----------------+-----+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| header + type info + field name | next field info | ... |</span><br/></div></code></pre></div></div> |
| <h6 class="anchor anchorTargetStickyNavbar_Vzrq" id=field-header>Field Header<a href=#field-header class=hash-link aria-label="Direct link to Field Header" title="Direct link to Field Header" translate=no></a></h6> |
| <p>Field Header is 8 bits, annotation can be used to provide more specific info. If annotation not exists, fury will infer |
| those info automatically.</p> |
| <p>The format for field header is:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">2 bits field name encoding + 4 bits size + nullability flag + ref tracking flag</span><br/></div></code></pre></div></div> |
| <p>Detailed spec:</p> |
| <ul> |
| <li class="">2 bits field name encoding:<!-- --> |
| <ul> |
| <li class="">encoding: <code>UTF8/ALL_TO_LOWER_SPECIAL/LOWER_UPPER_DIGIT_SPECIAL/TAG_ID</code></li> |
| <li class="">If tag id is used, field name will be written by an unsigned varint tag id, and 2 bits encoding will be <code>11</code>.</li> |
| </ul> |
| </li> |
| <li class="">size of field name:<!-- --> |
| <ul> |
| <li class="">The <code>4 bits size: 0~14</code> will be used to indicate length <code>1~15</code>, the value <code>15</code> indicates to read more bytes, |
| the encoding will encode <code>size - 15</code> as a varint next.</li> |
| <li class="">If encoding is <code>TAG_ID</code>, then num_bytes of field name will be used to store tag id.</li> |
| </ul> |
| </li> |
| <li class="">ref tracking: when set to 1, ref tracking will be enabled for this field.</li> |
| <li class="">nullability: when set to 1, this field can be null.</li> |
| </ul> |
| <h6 class="anchor anchorTargetStickyNavbar_Vzrq" id=field-type-info>Field Type Info<a href=#field-type-info class=hash-link aria-label="Direct link to Field Type Info" title="Direct link to Field Type Info" translate=no></a></h6> |
| <p>Field type info is written as unsigned int8. Detailed id spec is:</p> |
| <ul> |
| <li class="">For struct registered by id, it will be <code>Type.STRUCT</code>.</li> |
| <li class="">For struct registered by name, it will be <code>Type.NAMED_STRUCT</code>.</li> |
| <li class="">For enum registered by id, it will be <code>Type.ENUM</code>.</li> |
| <li class="">For enum registered by name, it will be <code>Type.NAMED_ENUM</code>.</li> |
| <li class="">For ext type registered by id, it will be <code>Type.EXT</code>.</li> |
| <li class="">For ext type registered by name, it will be <code>Type.NAMED_EXT</code>.</li> |
| <li class="">For list/set type, it will be written as <code>Type.LIST/SET</code>, then write element type recursively.</li> |
| <li class="">For 1D primitive array type, it will be written as <code>Type.XXX_ARRAY</code>.</li> |
| <li class="">For multi-dimensional primitive array type with same size on each dim, it will be written as <code>Type.TENSOR</code>.</li> |
| <li class="">For other array type, it will be written as <code>Type.LIST</code>, then write element type recursively.</li> |
| <li class="">For map type, it will be written as <code>Type.MAP</code>, then write key and value type recursively.</li> |
| <li class="">For other types supported by fury directly, it will be fury type id for that type.</li> |
| <li class="">For other types not determined at compile time, write <code>Type.UNKNOWN</code> instead. For such types, actual type |
| will be written when serializing such field values.</li> |
| </ul> |
| <p>Polymorphism spec:</p> |
| <ul> |
| <li class=""><code>struct/named_struct/ext/named_ext</code> are taken as polymorphic, the meta for those types are written separately |
| instead of inlining here to reduce meta space cost if object of this type is serialized in current object graph |
| multiple times, and the field value may be null too.</li> |
| <li class=""><code>enum</code> is taken as morphic, if deserialization doesn't have this field, or the type is not enum, enum value |
| will be skipped.</li> |
| <li class=""><code>list/map/set</code> are taken as morphic, when serializing values of those type, the concrete types won't be written |
| again.</li> |
| <li class="">Other types that fury supported are taken as morphic too.</li> |
| </ul> |
| <p>List/Set/Map nested type spec:</p> |
| <ul> |
| <li class=""><code>list</code>: <code>| list type id | nested type id << 2 + nullability flag + ref tracking flag | ... multi-layer type info |</code></li> |
| <li class=""><code>set</code>: <code>| set type id | nested type id << 2 + nullability flag + ref tracking flag | ... multi-layer type info |</code></li> |
| <li class=""><code>map</code>: <code>| set type id | key type info | value type info |</code> |
| <ul> |
| <li class="">Key type format: <code>| nested type id << 2 + nullability flag + ref tracking flag | ... multi-layer type info |</code></li> |
| <li class="">Value type format: <code>| nested type id << 2 + nullability flag + ref tracking flag | ... multi-layer type info |</code></li> |
| </ul> |
| </li> |
| </ul> |
| <h6 class="anchor anchorTargetStickyNavbar_Vzrq" id=field-name>Field Name<a href=#field-name class=hash-link aria-label="Direct link to Field Name" title="Direct link to Field Name" translate=no></a></h6> |
| <p>If tag id is set, tag id will be used instead. Otherwise meta string of field name will be written instead.</p> |
| <h6 class="anchor anchorTargetStickyNavbar_Vzrq" id=field-order>Field order<a href=#field-order class=hash-link aria-label="Direct link to Field order" title="Direct link to Field order" translate=no></a></h6> |
| <p>Field order are left as implementation details, which is not exposed to specification, the deserialization need to |
| resort fields based on Fury fields sort algorithms. In this way, fury can compute statistics for field names or types and |
| using a more compact encoding.</p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=extended-type-meta-with-inheritance-support>Extended Type Meta with Inheritance support<a href=#extended-type-meta-with-inheritance-support class=hash-link aria-label="Direct link to Extended Type Meta with Inheritance support" title="Direct link to Extended Type Meta with Inheritance support" translate=no></a></h2> |
| <p>If one want to support inheritance for struct, one can implement following spec.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=schema-consistent>Schema consistent<a href=#schema-consistent class=hash-link aria-label="Direct link to Schema consistent" title="Direct link to Schema consistent" translate=no></a></h3> |
| <p>Fields are serialized from parent type to leaf type. Fields are sorted using fury struct fields sort algorithms.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=schema-evolution>Schema Evolution<a href=#schema-evolution class=hash-link aria-label="Direct link to Schema Evolution" title="Direct link to Schema Evolution" translate=no></a></h3> |
| <p>Meta layout for schema evolution mode:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| 8 bytes header | variable bytes | variable bytes | variable bytes | variable bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+----------------------+----------------+----------------+--------------------+--------------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| global binary header | meta header | fields meta | parent meta header | parent fields meta |</span><br/></div></code></pre></div></div> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=meta-header-1>Meta header<a href=#meta-header-1 class=hash-link aria-label="Direct link to Meta header" title="Direct link to Meta header" translate=no></a></h4> |
| <p>Meta header is a 64 bits number value encoded in little endian order.</p> |
| <ul> |
| <li class="">Lowest 4 digits <code>0b0000~0b1110</code> are used to record num classes. <code>0b1111</code> is preserved to indicate that Fury need to |
| read more bytes for length using Fury unsigned int encoding. If current type doesn't has parent type, or parent |
| type doesn't have fields to serialize, or we're in a context which serialize fields of current type |
| only, num classes will be 1.</li> |
| <li class="">The 5th bit is used to indicate whether this type needs schema evolution.</li> |
| <li class="">Other 56 bits are used to store the unique hash of <code>flags + all layers type meta</code>.</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=single-layer-type-meta>Single layer type meta<a href=#single-layer-type-meta class=hash-link aria-label="Direct link to Single layer type meta" title="Direct link to Single layer type meta" translate=no></a></h4> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| unsigned varint | var uint | field info: variable bytes | variable bytes | ... |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------------+----------+-------------------------------+-----------------+-----+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| num_fields | type id | header + type id + field name | next field info | ... |</span><br/></div></code></pre></div></div> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=other-layers-type-meta>Other layers type meta<a href=#other-layers-type-meta class=hash-link aria-label="Direct link to Other layers type meta" title="Direct link to Other layers type meta" translate=no></a></h4> |
| <p>Same encoding algorithm as the previous layer.</p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=meta-string>Meta String<a href=#meta-string class=hash-link aria-label="Direct link to Meta String" title="Direct link to Meta String" translate=no></a></h2> |
| <p>Meta string is mainly used to encode meta strings such as field names.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=encoding-algorithms>Encoding Algorithms<a href=#encoding-algorithms class=hash-link aria-label="Direct link to Encoding Algorithms" title="Direct link to Encoding Algorithms" translate=no></a></h3> |
| <p>String binary encoding algorithm:</p> |
| <table><thead><tr><th>Algorithm<th>Pattern<th>Description<tbody><tr><td>LOWER_SPECIAL<td><code>a-z._$|</code><td>every char is written using 5 bits, <code>a-z</code>: <code>0b00000~0b11001</code>, <code>._$|</code>: <code>0b11010~0b11101</code>, prepend one bit at the start to indicate whether strip last char since last byte may have 7 redundant bits(1 indicates strip last char)<tr><td>LOWER_UPPER_DIGIT_SPECIAL<td><code>a-zA-Z0~9._</code><td>every char is written using 6 bits, <code>a-z</code>: <code>0b00000~0b11001</code>, <code>A-Z</code>: <code>0b11010~0b110011</code>, <code>0~9</code>: <code>0b110100~0b111101</code>, <code>._</code>: <code>0b111110~0b111111</code>, prepend one bit at the start to indicate whether strip last char since last byte may have 7 redundant bits(1 indicates strip last char)<tr><td>UTF-8<td>any chars<td>UTF-8 encoding</table> |
| <p>Encoding flags:</p> |
| <table><thead><tr><th>Encoding Flag<th>Pattern<th>Encoding Algorithm<tbody><tr><td>LOWER_SPECIAL<td>every char is in <code>a-z._|</code><td><code>LOWER_SPECIAL</code><tr><td>FIRST_TO_LOWER_SPECIAL<td>every char is in <code>a-z._</code> except first char is upper case<td>replace first upper case char to lower case, then use <code>LOWER_SPECIAL</code><tr><td>ALL_TO_LOWER_SPECIAL<td>every char is in <code>a-zA-Z._</code><td>replace every upper case char by <code>|</code> + <code>lower case</code>, then use <code>LOWER_SPECIAL</code>, use this encoding if it's smaller than Encoding <code>LOWER_UPPER_DIGIT_SPECIAL</code><tr><td>LOWER_UPPER_DIGIT_SPECIAL<td>every char is in <code>a-zA-Z._</code><td>use <code>LOWER_UPPER_DIGIT_SPECIAL</code> encoding if it's smaller than Encoding <code>FIRST_TO_LOWER_SPECIAL</code><tr><td>UTF8<td>any utf-8 char<td>use <code>UTF-8</code> encoding<tr><td>Compression<td>any utf-8 char<td>lossless compression</table> |
| <p>Notes:</p> |
| <ul> |
| <li class="">Depending on cases, one can choose encoding <code>flags + data</code> jointly, uses 3 bits of first byte for flags and other |
| bytes |
| for data.</li> |
| </ul> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=value-format>Value Format<a href=#value-format class=hash-link aria-label="Direct link to Value Format" title="Direct link to Value Format" translate=no></a></h2> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=basic-types>Basic types<a href=#basic-types class=hash-link aria-label="Direct link to Basic types" title="Direct link to Basic types" translate=no></a></h3> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=bool>bool<a href=#bool class=hash-link aria-label="Direct link to bool" title="Direct link to bool" translate=no></a></h4> |
| <ul> |
| <li class="">size: 1 byte</li> |
| <li class="">format: 0 for <code>false</code>, 1 for <code>true</code></li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=int8>int8<a href=#int8 class=hash-link aria-label="Direct link to int8" title="Direct link to int8" translate=no></a></h4> |
| <ul> |
| <li class="">size: 1 byte</li> |
| <li class="">format: write as pure byte.</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=int16>int16<a href=#int16 class=hash-link aria-label="Direct link to int16" title="Direct link to int16" translate=no></a></h4> |
| <ul> |
| <li class="">size: 2 byte</li> |
| <li class="">byte order: raw bytes of little endian order</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=unsigned-int32>unsigned int32<a href=#unsigned-int32 class=hash-link aria-label="Direct link to unsigned int32" title="Direct link to unsigned int32" translate=no></a></h4> |
| <ul> |
| <li class="">size: 4 byte</li> |
| <li class="">byte order: raw bytes of little endian order</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=unsigned-varint32>unsigned varint32<a href=#unsigned-varint32 class=hash-link aria-label="Direct link to unsigned varint32" title="Direct link to unsigned varint32" translate=no></a></h4> |
| <ul> |
| <li class="">size: 1~5 byte</li> |
| <li class="">Format: The most significant bit (MSB) in every byte indicates whether to have the next byte. If first bit is set |
| i.e. <code>b & 0x80 == 0x80</code>, then |
| the next byte should be read until the first bit of the next byte is unset.</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=signed-int32>signed int32<a href=#signed-int32 class=hash-link aria-label="Direct link to signed int32" title="Direct link to signed int32" translate=no></a></h4> |
| <ul> |
| <li class="">size: 4 byte</li> |
| <li class="">byte order: raw bytes of little endian order</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=signed-varint32>signed varint32<a href=#signed-varint32 class=hash-link aria-label="Direct link to signed varint32" title="Direct link to signed varint32" translate=no></a></h4> |
| <ul> |
| <li class="">size: 1~5 byte</li> |
| <li class="">Format: First convert the number into positive unsigned int by <code>(v << 1) ^ (v >> 31)</code> ZigZag algorithm, then encode |
| it as an unsigned varint.</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=unsigned-int64>unsigned int64<a href=#unsigned-int64 class=hash-link aria-label="Direct link to unsigned int64" title="Direct link to unsigned int64" translate=no></a></h4> |
| <ul> |
| <li class="">size: 8 byte</li> |
| <li class="">byte order: raw bytes of little endian order</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=unsigned-varint64>unsigned varint64<a href=#unsigned-varint64 class=hash-link aria-label="Direct link to unsigned varint64" title="Direct link to unsigned varint64" translate=no></a></h4> |
| <ul> |
| <li class="">size: 1~9 byte</li> |
| <li class="">Fury SLI(Small long as int) Encoding:<!-- --> |
| <ul> |
| <li class="">If long is in <code>[0, 2147483647]</code>, encode as 4 bytes int: <code>| little-endian: ((int) value) << 1 |</code></li> |
| <li class="">Otherwise write as 9 bytes: <code>| 0b1 | little-endian 8 bytes long |</code></li> |
| </ul> |
| </li> |
| <li class="">Fury PVL(Progressive Variable-length Long) Encoding:<!-- --> |
| <ul> |
| <li class="">positive long format: first bit in every byte indicates whether to have the next byte. If first bit is set |
| i.e. <code>b & 0x80 == 0x80</code>, then the next byte should be read until the first bit is unset.</li> |
| </ul> |
| </li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=signed-int64>signed int64<a href=#signed-int64 class=hash-link aria-label="Direct link to signed int64" title="Direct link to signed int64" translate=no></a></h4> |
| <ul> |
| <li class="">size: 8 byte</li> |
| <li class="">byte order: raw bytes of little endian order</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=signed-varint64>signed varint64<a href=#signed-varint64 class=hash-link aria-label="Direct link to signed varint64" title="Direct link to signed varint64" translate=no></a></h4> |
| <ul> |
| <li class="">size: 1~9 byte</li> |
| <li class="">Fury SLI(Small long as int) Encoding:<!-- --> |
| <ul> |
| <li class="">If long is in <code>[-1073741824, 1073741823]</code>, encode as 4 bytes int: <code>| little-endian: ((int) value) << 1 |</code></li> |
| <li class="">Otherwise write as 9 bytes: <code>| 0b1 | little-endian 8 bytes long |</code></li> |
| </ul> |
| </li> |
| <li class="">Fury PVL(Progressive Variable-length Long) Encoding:<!-- --> |
| <ul> |
| <li class="">First convert the number into positive unsigned long by <code>(v << 1) ^ (v >> 63)</code> ZigZag algorithm to reduce cost of |
| small negative numbers, then encoding it as an unsigned long.</li> |
| </ul> |
| </li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=float32>float32<a href=#float32 class=hash-link aria-label="Direct link to float32" title="Direct link to float32" translate=no></a></h4> |
| <ul> |
| <li class="">size: 4 byte</li> |
| <li class="">format: encode the specified floating-point value according to the IEEE 754 floating-point "single format" bit layout, |
| preserving Not-a-Number (NaN) values, then write as binary by little endian order.</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=float64>float64<a href=#float64 class=hash-link aria-label="Direct link to float64" title="Direct link to float64" translate=no></a></h4> |
| <ul> |
| <li class="">size: 8 byte</li> |
| <li class="">format: encode the specified floating-point value according to the IEEE 754 floating-point "double format" bit layout, |
| preserving Not-a-Number (NaN) values. then write as binary by little endian order.</li> |
| </ul> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=string>string<a href=#string class=hash-link aria-label="Direct link to string" title="Direct link to string" translate=no></a></h3> |
| <p>Format:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| unsigned varint64: size << 2 `bitor` 2 bits encoding flags | binary data |</span><br/></div></code></pre></div></div> |
| <ul> |
| <li class=""><code>size + encoding</code> will be concat as a long and encoded as an unsigned varint64. The little 2 bits is used for |
| encoding: |
| 0 for <code>latin1(ISO-8859-1)</code>, 1 for <code>utf-16</code>, 2 for <code>utf-8</code>.</li> |
| <li class="">encoded string binary data based on encoding: <code>latin/utf-16/utf-8</code>.</li> |
| </ul> |
| <p>Which encoding to choose:</p> |
| <ul> |
| <li class="">For JDK8: fury detect <code>latin</code> at runtime, if string is <code>latin</code> string, then use <code>latin</code> encoding, otherwise |
| use <code>utf-16</code>.</li> |
| <li class="">For JDK9+: fury use <code>coder</code> in <code>String</code> object for encoding, <code>latin</code>/<code>utf-16</code> will be used for encoding.</li> |
| <li class="">If the string is encoded by <code>utf-8</code>, then fury will use <code>utf-8</code> to decode the data. Cross-language string |
| serialization of fury uses <code>utf-8</code> by default.</li> |
| </ul> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=list>list<a href=#list class=hash-link aria-label="Direct link to list" title="Direct link to list" translate=no></a></h3> |
| <p>Format:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| unsigned varint64: length | 1 byte elements header | elements data |</span><br/></div></code></pre></div></div> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=elements-header>elements header<a href=#elements-header class=hash-link aria-label="Direct link to elements header" title="Direct link to elements header" translate=no></a></h4> |
| <p>In most cases, all elements are same type and not null, elements header will encode those homogeneous |
| information to avoid the cost of writing it for every element. Specifically, there are four kinds of information |
| which will be encoded by elements header, each use one bit:</p> |
| <ul> |
| <li class="">If track elements ref, use the first bit <code>0b1</code> of the header to flag it.</li> |
| <li class="">If the elements have null, use the second bit <code>0b10</code> of the header to flag it. If ref tracking is enabled for this |
| element type, this flag is invalid.</li> |
| <li class="">If the element types are not the declared type, use the 3rd bit <code>0b100</code> of the header to flag it.</li> |
| <li class="">If the element types are different, use the 4rd bit <code>0b1000</code> header to flag it.</li> |
| </ul> |
| <p>By default, all bits are unset, which means all elements won't track ref, all elements are same type, not null and |
| the actual element is the declared type in the custom type field.</p> |
| <p>The implementation can generate different deserialization code based read header, and look up the generated code from |
| a linear map/list.</p> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=elements-data>elements data<a href=#elements-data class=hash-link aria-label="Direct link to elements data" title="Direct link to elements data" translate=no></a></h4> |
| <p>Based on the elements header, the serialization of elements data may skip <code>ref flag</code>/<code>null flag</code>/<code>element type info</code>.</p> |
| <div class="language-python codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-python codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">fury </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token builtin">buffer</span><span class="token plain"> </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain">elems </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> </span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token keyword" style=color:#00009f>if</span><span class="token plain"> element_type_is_same</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>if</span><span class="token plain"> </span><span class="token keyword" style=color:#00009f>not</span><span class="token plain"> is_declared_type</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> fury</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_type</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem_type</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> elem_serializer </span><span class="token operator" style=color:#393A34>=</span><span class="token plain"> get_serializer</span><span class="token punctuation" style=color:#393A34>(</span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>.</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>if</span><span class="token plain"> track_ref</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>for</span><span class="token plain"> elem </span><span class="token keyword" style=color:#00009f>in</span><span class="token plain"> elems</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>if</span><span class="token plain"> </span><span class="token keyword" style=color:#00009f>not</span><span class="token plain"> ref_resolver</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_ref_or_null</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem</span><span class="token punctuation" style=color:#393A34>)</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> elem_serializer</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>elif</span><span class="token plain"> has_null</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>for</span><span class="token plain"> elem </span><span class="token keyword" style=color:#00009f>in</span><span class="token plain"> elems</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>if</span><span class="token plain"> elem </span><span class="token keyword" style=color:#00009f>is</span><span class="token plain"> </span><span class="token boolean" style=color:#36acaa>None</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_byte</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">null_flag</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>else</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_byte</span><span class="token punctuation" style=color:#393A34>(</span><span class="token plain">not_null_flag</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> elem_serializer</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>else</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>for</span><span class="token plain"> elem </span><span class="token keyword" style=color:#00009f>in</span><span class="token plain"> elems</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> elem_serializer</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"></span><span class="token keyword" style=color:#00009f>else</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>if</span><span class="token plain"> track_ref</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>for</span><span class="token plain"> elem </span><span class="token keyword" style=color:#00009f>in</span><span class="token plain"> elems</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> fury</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_ref</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>elif</span><span class="token plain"> has_null</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>for</span><span class="token plain"> elem </span><span class="token keyword" style=color:#00009f>in</span><span class="token plain"> elems</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> fury</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_nullable</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem</span><span class="token punctuation" style=color:#393A34>)</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>else</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> </span><span class="token keyword" style=color:#00009f>for</span><span class="token plain"> elem </span><span class="token keyword" style=color:#00009f>in</span><span class="token plain"> elems</span><span class="token punctuation" style=color:#393A34>:</span><span class="token plain"></span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> fury</span><span class="token punctuation" style=color:#393A34>.</span><span class="token plain">write_value</span><span class="token punctuation" style=color:#393A34>(</span><span class="token builtin">buffer</span><span class="token punctuation" style=color:#393A34>,</span><span class="token plain"> elem</span><span class="token punctuation" style=color:#393A34>)</span><br/></div></code></pre></div></div> |
| <p><a href=https://github.com/apache/fury/blob/20a1a78b17a75a123a6f5b7094c06ff77defc0fe/java/fury-core/src/main/java/org/apache/fury/serializer/collection/AbstractCollectionSerializer.java#L302 target=_blank rel="noopener noreferrer" class=""><code>CollectionSerializer#writeElements</code></a> |
| can be taken as an example.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=array>array<a href=#array class=hash-link aria-label="Direct link to array" title="Direct link to array" translate=no></a></h3> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=primitive-array>primitive array<a href=#primitive-array class=hash-link aria-label="Direct link to primitive array" title="Direct link to primitive array" translate=no></a></h4> |
| <p>Primitive array are taken as a binary buffer, serialization will just write the length of array size as an unsigned int, |
| then copy the whole buffer into the stream.</p> |
| <p>Such serialization won't compress the array. If users want to compress primitive array, users need to register custom |
| serializers for such types or mark it as list type.</p> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=tensor>Tensor<a href=#tensor class=hash-link aria-label="Direct link to Tensor" title="Direct link to Tensor" translate=no></a></h4> |
| <p>Tensor is a special primitive multi-dimensional array which all dimensions have same size and type. The serialization |
| format is:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| num_dims(unsigned varint) | shape[0](unsigned varint) | shape[...] | shape[N] | element type | data |</span><br/></div></code></pre></div></div> |
| <p>The data is continuous to reduce copy and may zero-copy in some cases.</p> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=object-array>object array<a href=#object-array class=hash-link aria-label="Direct link to object array" title="Direct link to object array" translate=no></a></h4> |
| <p>Object array is serialized using the list format. Object component type will be taken as list element |
| generic type.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=map>map<a href=#map class=hash-link aria-label="Direct link to map" title="Direct link to map" translate=no></a></h3> |
| <p>Map uses a chunk by chunk based Format:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| length(unsigned varint) | key value chunk data | ... | key value chunk data |</span><br/></div></code></pre></div></div> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=map-key-value-chunk-data>map key-value chunk data<a href=#map-key-value-chunk-data class=hash-link aria-label="Direct link to map key-value chunk data" title="Direct link to map key-value chunk data" translate=no></a></h4> |
| <p>Map iteration is too expensive, Fury won't compute the header like for list since it introduce |
| <a href=https://github.com/apache/fury/issues/925 target=_blank rel="noopener noreferrer" class="">considerable overhead</a>. |
| Users can use <code>MapFieldInfo</code> annotation to provide the header in advance. Otherwise Fury will use first key-value pair |
| to predict header optimistically, and update the chunk header if the prediction failed at some pair.</p> |
| <p>Fury will serialize the map chunk by chunk, every chunk has 255 pairs at most.</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| 1 byte | 1 byte | variable bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+----------------+----------------+-----------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| KV header | chunk size: N | N*2 objects |</span><br/></div></code></pre></div></div> |
| <p>KV header:</p> |
| <ul> |
| <li class="">If track key ref, use the first bit <code>0b1</code> of the header to flag it.</li> |
| <li class="">If the key has null, use the second bit <code>0b10</code> of the header to flag it. If ref tracking is enabled for this |
| key type, this flag is invalid.</li> |
| <li class="">If the actual key type of map is not the declared key type, use the 3rd bit <code>0b100</code> of the header to flag it.</li> |
| <li class="">If track value ref, use the 4th bit <code>0b1000</code> of the header to flag it.</li> |
| <li class="">If the value has null, use the 5th bit <code>0b10000</code> of the header to flag it. If ref tracking is enabled for this |
| value type, this flag is invalid.</li> |
| <li class="">If the value type of map is not the declared value type, use the 6rd bit <code>0b100000</code> of the header to flag it.</li> |
| <li class="">If key or value is null, that key and value will be written as a separate chunk, and chunk size writing will be |
| skipped too.</li> |
| </ul> |
| <p>If streaming write is enabled, which means Fury can't update written <code>chunk size</code>. In such cases, map key-value data |
| format will be:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| 1 byte | variable bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+----------------+-----------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| KV header | N*2 objects |</span><br/></div></code></pre></div></div> |
| <p><code>KV header</code> will be a header marked by <code>MapFieldInfo</code> in java. For languages such as golang, this can be computed in |
| advance for non-interface types most times. The implementation can generate different deserialization code based read |
| header, and look up the generated code from a linear map/list.</p> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=why-serialize-chunk-by-chunk>Why serialize chunk by chunk?<a href=#why-serialize-chunk-by-chunk class=hash-link aria-label="Direct link to Why serialize chunk by chunk?" title="Direct link to Why serialize chunk by chunk?" translate=no></a></h4> |
| <p>When fury will use first key-value pair to predict header optimistically, it can't know how many pairs have same |
| meta(tracking kef ref, key has null and so on). If we don't write chunk by chunk with max chunk size, we must write at |
| least <code>X</code> bytes to take up a place for later to update the number which has same elements, <code>X</code> is the num_bytes for |
| encoding varint encoding of map size.</p> |
| <p>And most map size are smaller than 255, if all pairs have same data, the chunk will be 1. This is common in golang/rust, |
| which object are not reference by default.</p> |
| <p>Also, if only one or two keys have different meta, we can make it into a different chunk, so that most pairs can share |
| meta.</p> |
| <p>The implementation can accumulate read count with map size to decide whether to read more chunks.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=enum>enum<a href=#enum class=hash-link aria-label="Direct link to enum" title="Direct link to enum" translate=no></a></h3> |
| <p>Enums are serialized as an unsigned var int. If the order of enum values change, the deserialized enum value may not be |
| the value users expect. In such cases, users must register enum serializer by make it write enum value as an enumerated |
| string with unique hash disabled.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=decimal>decimal<a href=#decimal class=hash-link aria-label="Direct link to decimal" title="Direct link to decimal" translate=no></a></h3> |
| <p>Not supported for now.</p> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=struct>struct<a href=#struct class=hash-link aria-label="Direct link to struct" title="Direct link to struct" translate=no></a></h3> |
| <p>Struct means object of <code>class/pojo/struct/bean/record</code> type. |
| Struct will be serialized by writing its fields data in fury order.</p> |
| <p>Depending on schema compatibility, structs will have different formats.</p> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=field-order-1>field order<a href=#field-order-1 class=hash-link aria-label="Direct link to field order" title="Direct link to field order" translate=no></a></h4> |
| <p>Field will be ordered as following, every group of fields will have its own order:</p> |
| <ul> |
| <li class="">primitive fields:<!-- --> |
| <ul> |
| <li class="">larger size type first, smaller later, variable size type last.</li> |
| <li class="">when same size, sort by type id</li> |
| <li class="">when same size and type id, sort by snake case field name</li> |
| <li class="">types: bool/int8/int16/int32/varint32/int64/varint64/sliint64/float16/float32/float64</li> |
| </ul> |
| </li> |
| <li class="">nullable primitive fields: same order as primitive fields</li> |
| <li class="">morphic fields: same type together, then sorted by field name lexicographically using snake case style.</li> |
| <li class="">unknown fields: same sort algorithms as morphic fields</li> |
| <li class="">list fields: same sort algorithms as morphic fields</li> |
| <li class="">set fields: same sort algorithms as morphic fields</li> |
| <li class="">map fields: same sort algorithms as morphic fields</li> |
| </ul> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=field-order-2>Field order<a href=#field-order-2 class=hash-link aria-label="Direct link to Field order" title="Direct link to Field order" translate=no></a></h4> |
| <p>Fields in a struct are sorted in a ascending order by:</p> |
| <ul> |
| <li class="">primitive fields first: bool/int8/int16/int32/varint32/int64/varint64/sliint64/float16/float32/float64, sorted by |
| type id.</li> |
| <li class="">nullable primitive fields</li> |
| <li class="">morphic types except <code>list/set/map</code></li> |
| <li class="">unknown types</li> |
| <li class="">list types</li> |
| <li class="">set types</li> |
| <li class="">map types</li> |
| </ul> |
| <p>If two fields have same type, then sort by snake_case styled field name.</p> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=schema-consistent-1>schema consistent<a href=#schema-consistent-1 class=hash-link aria-label="Direct link to schema consistent" title="Direct link to schema consistent" translate=no></a></h4> |
| <p>Object will be written as:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">| 4 byte | variable bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+---------------+------------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| type hash | field values |</span><br/></div></code></pre></div></div> |
| <p>Type hash is used to check the type schema consistency across languages. Type hash will be the first 32 bits of 56 bits |
| value of the type meta.</p> |
| <p>Object fields will be serialized one by one using following format:</p> |
| <div class="language-text codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-text codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">not null primitive field value:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| var bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+----------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| value data |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+----------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">nullable primitive field value:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| one byte | var bytes |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+---------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| null flag | field value |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+---------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">field value of final type with ref tracking:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| var bytes | var objects |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+-------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| ref meta | value data |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+-------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">field value of final type without ref tracking:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| one byte | var objects |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+-------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| null flag | field value |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+-------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">field value of non-final type with ref tracking:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| one byte | var bytes | var objects |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+-------------+-------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| ref meta | type meta | value data |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+-------------+-------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">field value of non-final type without ref tracking:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| one byte | var bytes | var objects |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+------------+------------+</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">| null flag | type meta | value data |</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">+-----------+------------+------------+</span><br/></div></code></pre></div></div> |
| <h4 class="anchor anchorTargetStickyNavbar_Vzrq" id=schema-evolution-1>Schema evolution<a href=#schema-evolution-1 class=hash-link aria-label="Direct link to Schema evolution" title="Direct link to Schema evolution" translate=no></a></h4> |
| <p>Schema evolution have similar format as schema consistent mode for object except:</p> |
| <ul> |
| <li class="">For the object type, <code>schema consistent</code> mode will write type by id only, but <code>schema evolution</code> mode will |
| write type consisting of field names, types and other meta too, see <a href=#type-meta class="">Type meta</a>.</li> |
| <li class="">Type meta of <code>final custom type</code> needs to be written too, because peers may not have this type defined.</li> |
| </ul> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=type>Type<a href=#type class=hash-link aria-label="Direct link to Type" title="Direct link to Type" translate=no></a></h3> |
| <p>Type will be serialized using type meta format.</p> |
| <h2 class="anchor anchorTargetStickyNavbar_Vzrq" id=implementation-guidelines>Implementation guidelines<a href=#implementation-guidelines class=hash-link aria-label="Direct link to Implementation guidelines" title="Direct link to Implementation guidelines" translate=no></a></h2> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=how-to-reduce-memory-readwrite-code>How to reduce memory read/write code<a href=#how-to-reduce-memory-readwrite-code class=hash-link aria-label="Direct link to How to reduce memory read/write code" title="Direct link to How to reduce memory read/write code" translate=no></a></h3> |
| <ul> |
| <li class="">Try to merge multiple bytes into an int/long write before writing to reduce memory IO and bound check cost.</li> |
| <li class="">Read multiple bytes as an int/long, then split into multiple bytes to reduce memory IO and bound check cost.</li> |
| <li class="">Try to use one varint/long to write flags and length together to save one byte cost and reduce memory io.</li> |
| <li class="">Condition branches are less expensive compared to memory IO cost unless there are too many branches.</li> |
| </ul> |
| <h3 class="anchor anchorTargetStickyNavbar_Vzrq" id=fast-deserialization-for-static-languages-without-runtime-codegen-support>Fast deserialization for static languages without runtime codegen support<a href=#fast-deserialization-for-static-languages-without-runtime-codegen-support class=hash-link aria-label="Direct link to Fast deserialization for static languages without runtime codegen support" title="Direct link to Fast deserialization for static languages without runtime codegen support" translate=no></a></h3> |
| <p>For type evolution, the serializer will encode the type meta into the serialized data. The deserializer will compare |
| this meta with class meta in the current process, and use the diff to determine how to deserialize the data.</p> |
| <p>For java/javascript/python, we can use the diff to generate serializer code at runtime and load it as class/function for |
| deserialization. In this way, the type evolution will be as fast as type consist mode.</p> |
| <p>For C++/Rust, we can't generate the serializer code at runtime. So we need to generate the code at compile-time using |
| meta programming. But at that time, we don't know the type schema in other processes, so we can't generate the |
| serializer code for such inconsistent types. We may need to generate the code which has a loop and compare field name |
| one by one to decide whether to deserialize and assign the field or skip the field value.</p> |
| <p>One fast way is that we can optimize the string comparison into <code>jump</code> instructions:</p> |
| <ul> |
| <li class="">Assume the current type has <code>n</code> fields, and the peer type has <code>n1</code> fields.</li> |
| <li class="">Generate an auto growing <code>field id</code> from <code>0</code> for every sorted field in the current type at the compile time.</li> |
| <li class="">Compare the received type meta with current type, generate same id if the field name is same, otherwise generate an |
| auto growing id starting from <code>n</code>, cache this meta at runtime.</li> |
| <li class="">Iterate the fields of received type meta, use a <code>switch</code> to compare the <code>field id</code> to deserialize data |
| and <code>assign/skip</code> field value. <strong>Continuous</strong> field id will be optimized into <code>jump</code> in <code>switch</code> block, so it will |
| very fast.</li> |
| </ul> |
| <p>Here is an example, suppose process A has a class <code>Foo</code> with version 1 defined as <code>Foo1</code>, process B has a class <code>Foo</code> |
| with version 2 defined as <code>Foo2</code>:</p> |
| <div class="language-c++ codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-c++ codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">// class Foo with version 1</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">class Foo1 {</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> int32_t v1; // id 0</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> std::string v2; // id 1</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">};</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">// class Foo with version 2</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">class Foo2 {</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> // id 0, but will have id 2 in process A</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> bool v0;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> // id 1, but will have id 0 in process A</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> int32_t v1;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> // id 2, but will have id 3 in process A</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> int64_t long_value;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> // id 3, but will have id 1 in process A</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> std::string v2;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> // id 4, but will have id 4 in process A</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> std::vector<std::string> list;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">};</span><br/></div></code></pre></div></div> |
| <p>When process A received serialized <code>Foo2</code> from process B, here is how it deserialize the data:</p> |
| <div class="language-c++ codeBlockContainer_Ckt0 theme-code-block" style=--prism-color:#393A34;--prism-background-color:#f6f8fa><div class=codeBlockContent_QJqH><pre tabindex=0 class="prism-code language-c++ codeBlock_bY9V thin-scrollbar" style=color:#393A34;background-color:#f6f8fa><code class=codeBlockLines_e6Vv><div class=token-line style=color:#393A34><span class="token plain">Foo1 foo1 = ...;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">const std::vector<fury::FieldInfo> &field_infos = type_meta.field_infos;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">for (const auto &field_info : field_infos) {</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> switch (field_info.field_id) {</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> case 0:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> foo1.v1 = buffer.read_varint32();</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> break;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> case 1:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> foo1.v2 = fury.read_string();</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> break;</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> default:</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> fury.skip_data(field_info);</span><br/></div><div class=token-line style=color:#393A34><span class="token plain"> }</span><br/></div><div class=token-line style=color:#393A34><span class="token plain">}</span><br/></div></code></pre></div></div></div><footer class="theme-doc-footer docusaurus-mt-lg"><div class="row margin-top--sm theme-doc-footer-edit-meta-row"><div class="col noPrint_WFHX"><a href=https://github.com/apache/fory-site/tree/main/docs/specification/xlang_serialization_spec.md target=_blank rel="noopener noreferrer" class=theme-edit-this-page><svg fill=currentColor height=20 width=20 viewBox="0 0 40 40" class=iconEdit_Z9Sw aria-hidden=true><g><path d="m34.5 11.7l-3 3.1-6.3-6.3 3.1-3q0.5-0.5 1.2-0.5t1.1 0.5l3.9 3.9q0.5 0.4 0.5 1.1t-0.5 1.2z m-29.5 17.1l18.4-18.5 6.3 6.3-18.4 18.4h-6.3v-6.2z"/></g></svg>Edit this page</a></div><div class="col lastUpdated_JAkA"></div></div></footer></article><nav class="docusaurus-mt-lg pagination-nav" aria-label="Docs pages"><a class="pagination-nav__link pagination-nav__link--next" href=/docs/0.10/specification/java_serialization_spec><div class=pagination-nav__sublabel>Next</div><div class=pagination-nav__label>Fury Java Serialization Format</div></a></nav></div></div><div class="col col--3"><div class="tableOfContents_bqdL thin-scrollbar theme-doc-toc-desktop"><ul class="table-of-contents table-of-contents__left-border"><li><a href=#cross-language-serialization-specification class="table-of-contents__link toc-highlight">Cross-language Serialization Specification</a><li><a href=#type-systems class="table-of-contents__link toc-highlight">Type Systems</a><ul><li><a href=#data-types class="table-of-contents__link toc-highlight">Data Types</a><li><a href=#polymorphisms class="table-of-contents__link toc-highlight">Polymorphisms</a><li><a href=#type-disambiguation class="table-of-contents__link toc-highlight">Type disambiguation</a><li><a href=#type-id class="table-of-contents__link toc-highlight">Type ID</a><li><a href=#type-mapping class="table-of-contents__link toc-highlight">Type mapping</a></ul><li><a href=#spec-overview class="table-of-contents__link toc-highlight">Spec overview</a><li><a href=#fury-header class="table-of-contents__link toc-highlight">Fury header</a><li><a href=#reference-meta class="table-of-contents__link toc-highlight">Reference Meta</a><li><a href=#type-meta class="table-of-contents__link toc-highlight">Type Meta</a><ul><li><a href=#struct-schema-consistent class="table-of-contents__link toc-highlight">Struct Schema consistent</a><li><a href=#struct-schema-evolution class="table-of-contents__link toc-highlight">Struct Schema evolution</a></ul><li><a href=#extended-type-meta-with-inheritance-support class="table-of-contents__link toc-highlight">Extended Type Meta with Inheritance support</a><ul><li><a href=#schema-consistent class="table-of-contents__link toc-highlight">Schema consistent</a><li><a href=#schema-evolution class="table-of-contents__link toc-highlight">Schema Evolution</a></ul><li><a href=#meta-string class="table-of-contents__link toc-highlight">Meta String</a><ul><li><a href=#encoding-algorithms class="table-of-contents__link toc-highlight">Encoding Algorithms</a></ul><li><a href=#value-format class="table-of-contents__link toc-highlight">Value Format</a><ul><li><a href=#basic-types class="table-of-contents__link toc-highlight">Basic types</a><li><a href=#string class="table-of-contents__link toc-highlight">string</a><li><a href=#list class="table-of-contents__link toc-highlight">list</a><li><a href=#array class="table-of-contents__link toc-highlight">array</a><li><a href=#map class="table-of-contents__link toc-highlight">map</a><li><a href=#enum class="table-of-contents__link toc-highlight">enum</a><li><a href=#decimal class="table-of-contents__link toc-highlight">decimal</a><li><a href=#struct class="table-of-contents__link toc-highlight">struct</a><li><a href=#type class="table-of-contents__link toc-highlight">Type</a></ul><li><a href=#implementation-guidelines class="table-of-contents__link toc-highlight">Implementation guidelines</a><ul><li><a href=#how-to-reduce-memory-readwrite-code class="table-of-contents__link toc-highlight">How to reduce memory read/write code</a><li><a href=#fast-deserialization-for-static-languages-without-runtime-codegen-support class="table-of-contents__link toc-highlight">Fast deserialization for static languages without runtime codegen support</a></ul></ul></div></div></div></div></main></div></div></div><footer class="theme-layout-footer footer footer--dark"><div class="container container-fluid"><div class="row footer__links"><div class="theme-layout-footer-column col footer__col"><div class=footer__title>Community</div><ul class="footer__items clean-list"><li class=footer__item><a href=https://lists.apache.org/list.html?dev@fory.apache.org target=_blank rel="noopener noreferrer" class=footer__link-item>Mailing list<svg width=13.5 height=13.5 aria-label="(opens in new tab)" class=iconExternalLink_nPIU><use href=#theme-svg-external-link /></svg></a><li class=footer__item><a href=https://join.slack.com/t/fory-project/shared_invite/zt-1u8soj4qc-ieYEu7ciHOqA2mo47llS8A target=_blank rel="noopener noreferrer" class=footer__link-item>Slack<svg width=13.5 height=13.5 aria-label="(opens in new tab)" class=iconExternalLink_nPIU><use href=#theme-svg-external-link /></svg></a><li class=footer__item><a href=https://twitter.com/ApacheFory target=_blank rel="noopener noreferrer" class=footer__link-item>Twitter<svg width=13.5 height=13.5 aria-label="(opens in new tab)" class=iconExternalLink_nPIU><use href=#theme-svg-external-link /></svg></a></ul></div><div class="theme-layout-footer-column col footer__col"><div class=footer__title>Docs</div><ul class="footer__items clean-list"><li class=footer__item><a class=footer__link-item href=/docs/start/install>Install</a><li class=footer__item><a class=footer__link-item href=/docs/start/usage>Usage</a><li class=footer__item><a class=footer__link-item href=/docs/introduction/benchmark>Benchmark</a></ul></div><div class="theme-layout-footer-column col footer__col"><div class=footer__title>Repositories</div><ul class="footer__items clean-list"><li class=footer__item><a href=https://github.com/apache/fory target=_blank rel="noopener noreferrer" class=footer__link-item>Apache Fory™<svg width=13.5 height=13.5 aria-label="(opens in new tab)" class=iconExternalLink_nPIU><use href=#theme-svg-external-link /></svg></a><li class=footer__item><a href=https://github.com/apache/fory-site target=_blank rel="noopener noreferrer" class=footer__link-item>Website<svg width=13.5 height=13.5 aria-label="(opens in new tab)" class=iconExternalLink_nPIU><use href=#theme-svg-external-link /></svg></a></ul></div></div><div class="footer__bottom text--center"><div class=margin-bottom--sm><a href=https://apache.org/ rel="noopener noreferrer" class=footerLogoLink_BH7S><img src=/img/asf_logo.svg alt="ASF Logo" class="footer__logo themedComponent_mlkZ themedComponent--light_NVdE" width=200 /><img src=/img/asf_logo.svg alt="ASF Logo" class="footer__logo themedComponent_mlkZ themedComponent--dark_xIcU" width=200 /></a></div><div class=footer__copyright><div> |
| <p> |
| Copyright © 2026 The Apache Software Foundation, Licensed under the Apache License, Version 2.0. <br/> |
| Apache Fory, Fory, Apache, the Apache Logo and the Apache Fory logo are either registered trademarks or trademarks of the Apache Software Foundation in the United States and/or other countries. |
| </p> |
| </div></div></div></div></footer></div></body> |