{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "view-in-github"
   },
   "source": [
    "<a href=\"https://colab.research.google.com/github/apache/beam/blob/master/examples/notebooks/documentation/transforms/python/elementwise/pardo-py.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open in Colab\"/></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "view-the-docs-top"
   },
   "source": [
    "<table align=\"left\"><td><a target=\"_blank\" href=\"https://beam.apache.org/documentation/transforms/python/elementwise/pardo\"><img src=\"https://beam.apache.org/images/logos/full-color/name-bottom/beam-logo-full-color-name-bottom-100.png\" width=\"32\" height=\"32\" />View the docs</a></td></table>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "cellView": "form",
    "id": "_-code"
   },
   "outputs": [],
   "source": [
    "#@title Licensed under the Apache License, Version 2.0 (the \"License\")\n",
    "# Licensed to the Apache Software Foundation (ASF) under one\n",
    "# or more contributor license agreements. See the NOTICE file\n",
    "# distributed with this work for additional information\n",
    "# regarding copyright ownership. The ASF licenses this file\n",
    "# to you under the Apache License, Version 2.0 (the\n",
    "# \"License\"); you may not use this file except in compliance\n",
    "# with the License. You may obtain a copy of the License at\n",
    "#\n",
    "#   http://www.apache.org/licenses/LICENSE-2.0\n",
    "#\n",
    "# Unless required by applicable law or agreed to in writing,\n",
    "# software distributed under the License is distributed on an\n",
    "# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n",
    "# KIND, either express or implied. See the License for the\n",
    "# specific language governing permissions and limitations\n",
    "# under the License."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "pardo"
   },
   "source": [
    "# ParDo\n",
    "\n",
    "<script type=\"text/javascript\">\n",
    "localStorage.setItem('language', 'language-py')\n",
    "</script>\n",
    "\n",
    "<table align=\"left\" style=\"margin-right:1em\">\n",
    "  <td>\n",
    "    <a class=\"button\" target=\"_blank\" href=\"https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.ParDo\"><img src=\"https://beam.apache.org/images/logos/sdks/python.png\" width=\"32px\" height=\"32px\" alt=\"Pydoc\"/> Pydoc</a>\n",
    "  </td>\n",
    "</table>\n",
    "\n",
    "<br/><br/><br/>\n",
    "\n",
    "A transform for generic parallel processing.\n",
    "A `ParDo` transform considers each element in the input `PCollection`,\n",
    "performs some processing function (your user code) on that element,\n",
    "and emits zero or more elements to an output `PCollection`.\n",
    "\n",
    "See more information in the\n",
    "[Beam Programming Guide](https://beam.apache.org/documentation/programming-guide/#pardo)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "setup"
   },
   "source": [
    "## Setup\n",
    "\n",
    "To run a code cell, you can click the **Run cell** button at the top left of the cell,\n",
    "or select it and press **`Shift+Enter`**.\n",
    "Try modifying a code cell and re-running it to see what happens.\n",
    "\n",
    "> To learn more about Colab, see\n",
    "> [Welcome to Colaboratory!](https://colab.sandbox.google.com/notebooks/welcome.ipynb).\n",
    "\n",
    "First, let's install the `apache-beam` module."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "setup-code"
   },
   "outputs": [],
   "source": [
    "!pip install --quiet -U apache-beam"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "examples"
   },
   "source": [
    "## Examples\n",
    "\n",
    "In the following examples, we explore how to create custom `DoFn`s and access\n",
    "the timestamp and windowing information."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "example-1-pardo-with-a-simple-dofn"
   },
   "source": [
    "### Example 1: ParDo with a simple DoFn\n",
    "\n",
    "The following example defines a simple `DoFn` class called `SplitWords`\n",
    "which stores the `delimiter` as an object field.\n",
    "The `process` method is called once per element,\n",
    "and it can yield zero or more output elements."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "example-1-pardo-with-a-simple-dofn-code"
   },
   "outputs": [],
   "source": [
    "import apache_beam as beam\n",
    "\n",
    "class SplitWords(beam.DoFn):\n",
    "  def __init__(self, delimiter=','):\n",
    "    self.delimiter = delimiter\n",
    "\n",
    "  def process(self, text):\n",
    "    for word in text.split(self.delimiter):\n",
    "      yield word\n",
    "\n",
    "with beam.Pipeline() as pipeline:\n",
    "  plants = (\n",
    "      pipeline\n",
    "      | 'Gardening plants' >> beam.Create([\n",
    "          '🍓Strawberry,🥕Carrot,🍆Eggplant',\n",
    "          '🍅Tomato,🥔Potato',\n",
    "      ])\n",
    "      | 'Split words' >> beam.ParDo(SplitWords(','))\n",
    "      | beam.Map(print))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "example-1-pardo-with-a-simple-dofn-2"
   },
   "source": [
    "<table align=\"left\" style=\"margin-right:1em\">\n",
    "  <td>\n",
    "    <a class=\"button\" target=\"_blank\" href=\"https://github.com/apache/beam/blob/master/sdks/python/apache_beam/examples/snippets/transforms/elementwise/pardo.py\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" width=\"32px\" height=\"32px\" alt=\"View source code\"/> View source code</a>\n",
    "  </td>\n",
    "</table>\n",
    "\n",
    "<br/><br/><br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "example-2-pardo-with-timestamp-and-window-information"
   },
   "source": [
    "### Example 2: ParDo with timestamp and window information\n",
    "\n",
    "In this example, we add new parameters to the `process` method to bind parameter values at runtime.\n",
    "\n",
    "* [`beam.DoFn.TimestampParam`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.TimestampParam)\n",
    "  binds the timestamp information as an\n",
    "  [`apache_beam.utils.timestamp.Timestamp`](https://beam.apache.org/releases/pydoc/current/apache_beam.utils.timestamp.html#apache_beam.utils.timestamp.Timestamp)\n",
    "  object.\n",
    "* [`beam.DoFn.WindowParam`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.WindowParam)\n",
    "  binds the window information as the appropriate\n",
    "  [`apache_beam.transforms.window.*Window`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.window.html)\n",
    "  object."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "example-2-pardo-with-timestamp-and-window-information-code"
   },
   "outputs": [],
   "source": [
    "import apache_beam as beam\n",
    "\n",
    "class AnalyzeElement(beam.DoFn):\n",
    "  def process(\n",
    "      self,\n",
    "      elem,\n",
    "      timestamp=beam.DoFn.TimestampParam,\n",
    "      window=beam.DoFn.WindowParam):\n",
    "    yield '\\n'.join([\n",
    "        '# timestamp',\n",
    "        'type(timestamp) -> ' + repr(type(timestamp)),\n",
    "        'timestamp.micros -> ' + repr(timestamp.micros),\n",
    "        'timestamp.to_rfc3339() -> ' + repr(timestamp.to_rfc3339()),\n",
    "        'timestamp.to_utc_datetime() -> ' + repr(timestamp.to_utc_datetime()),\n",
    "        '',\n",
    "        '# window',\n",
    "        'type(window) -> ' + repr(type(window)),\n",
    "        'window.start -> {} ({})'.format(\n",
    "            window.start, window.start.to_utc_datetime()),\n",
    "        'window.end -> {} ({})'.format(\n",
    "            window.end, window.end.to_utc_datetime()),\n",
    "        'window.max_timestamp() -> {} ({})'.format(\n",
    "            window.max_timestamp(), window.max_timestamp().to_utc_datetime()),\n",
    "    ])\n",
    "\n",
    "with beam.Pipeline() as pipeline:\n",
    "  dofn_params = (\n",
    "      pipeline\n",
    "      | 'Create a single test element' >> beam.Create([':)'])\n",
    "      | 'Add timestamp (Spring equinox 2020)' >>\n",
    "      beam.Map(lambda elem: beam.window.TimestampedValue(elem, 1584675660))\n",
    "      |\n",
    "      'Fixed 30sec windows' >> beam.WindowInto(beam.window.FixedWindows(30))\n",
    "      | 'Analyze element' >> beam.ParDo(AnalyzeElement())\n",
    "      | beam.Map(print))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "example-2-pardo-with-timestamp-and-window-information-2"
   },
   "source": [
    "<table align=\"left\" style=\"margin-right:1em\">\n",
    "  <td>\n",
    "    <a class=\"button\" target=\"_blank\" href=\"https://github.com/apache/beam/blob/master/sdks/python/apache_beam/examples/snippets/transforms/elementwise/pardo.py\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" width=\"32px\" height=\"32px\" alt=\"View source code\"/> View source code</a>\n",
    "  </td>\n",
    "</table>\n",
    "\n",
    "<br/><br/><br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "example-3-pardo-with-dofn-methods"
   },
   "source": [
    "### Example 3: ParDo with DoFn methods\n",
    "\n",
    "A [`DoFn`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn)\n",
    "can be customized with a number of methods that can help create more complex behaviors.\n",
    "You can customize what a worker does when it starts and shuts down with `setup` and `teardown`.\n",
    "You can also customize what to do when a\n",
    "[*bundle of elements*](https://beam.apache.org/documentation/runtime/model/#bundling-and-persistence)\n",
    "starts and finishes with `start_bundle` and `finish_bundle`.\n",
    "\n",
    "* [`DoFn.setup()`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.setup):\n",
    "  Called *once per `DoFn` instance* when the `DoFn` instance is initialized.\n",
    "  `setup` need not to be cached, so it could be called more than once per worker.\n",
    "  This is a good place to connect to database instances, open network connections or other resources.\n",
    "\n",
    "* [`DoFn.start_bundle()`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.start_bundle):\n",
    "  Called *once per bundle of elements* before calling `process` on the first element of the bundle.\n",
    "  This is a good place to start keeping track of the bundle elements.\n",
    "\n",
    "* [**`DoFn.process(element, *args, **kwargs)`**](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.process):\n",
    "  Called *once per element*, can *yield zero or more elements*.\n",
    "  Additional `*args` or `**kwargs` can be passed through\n",
    "  [`beam.ParDo()`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.ParDo).\n",
    "  **[required]**\n",
    "\n",
    "* [`DoFn.finish_bundle()`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.finish_bundle):\n",
    "  Called *once per bundle of elements* after calling `process` after the last element of the bundle,\n",
    "  can *yield zero or more elements*. This is a good place to do batch calls on a bundle of elements,\n",
    "  such as running a database query.\n",
    "\n",
    "  For example, you can initialize a batch in `start_bundle`,\n",
    "  add elements to the batch in `process` instead of yielding them,\n",
    "  then running a batch query on those elements on `finish_bundle`, and yielding all the results.\n",
    "\n",
    "  Note that yielded elements from `finish_bundle` must be of the type\n",
    "  [`apache_beam.utils.windowed_value.WindowedValue`](https://github.com/apache/beam/blob/master/sdks/python/apache_beam/utils/windowed_value.py).\n",
    "  You need to provide a timestamp as a unix timestamp, which you can get from the last processed element.\n",
    "  You also need to provide a window, which you can get from the last processed element like in the example below.\n",
    "\n",
    "* [`DoFn.teardown()`](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.DoFn.teardown):\n",
    "  Called *once (as a best effort) per `DoFn` instance* when the `DoFn` instance is shutting down.\n",
    "  This is a good place to close database instances, close network connections or other resources.\n",
    "\n",
    "  Note that `teardown` is called as a *best effort* and is *not guaranteed*.\n",
    "  For example, if the worker crashes, `teardown` might not be called."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "example-3-pardo-with-dofn-methods-code"
   },
   "outputs": [],
   "source": [
    "import apache_beam as beam\n",
    "\n",
    "class DoFnMethods(beam.DoFn):\n",
    "  def __init__(self):\n",
    "    print('__init__')\n",
    "    self.window = beam.window.GlobalWindow()\n",
    "\n",
    "  def setup(self):\n",
    "    print('setup')\n",
    "\n",
    "  def start_bundle(self):\n",
    "    print('start_bundle')\n",
    "\n",
    "  def process(self, element, window=beam.DoFn.WindowParam):\n",
    "    self.window = window\n",
    "    yield '* process: ' + element\n",
    "\n",
    "  def finish_bundle(self):\n",
    "    yield beam.utils.windowed_value.WindowedValue(\n",
    "        value='* finish_bundle: 🌱🌳🌍',\n",
    "        timestamp=0,\n",
    "        windows=[self.window],\n",
    "    )\n",
    "\n",
    "  def teardown(self):\n",
    "    print('teardown')\n",
    "\n",
    "with beam.Pipeline() as pipeline:\n",
    "  results = (\n",
    "      pipeline\n",
    "      | 'Create inputs' >> beam.Create(['🍓', '🥕', '🍆', '🍅', '🥔'])\n",
    "      | 'DoFn methods' >> beam.ParDo(DoFnMethods())\n",
    "      | beam.Map(print))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "example-3-pardo-with-dofn-methods-2"
   },
   "source": [
    "<table align=\"left\" style=\"margin-right:1em\">\n",
    "  <td>\n",
    "    <a class=\"button\" target=\"_blank\" href=\"https://github.com/apache/beam/blob/master/sdks/python/apache_beam/examples/snippets/transforms/elementwise/pardo.py\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" width=\"32px\" height=\"32px\" alt=\"View source code\"/> View source code</a>\n",
    "  </td>\n",
    "</table>\n",
    "\n",
    "<br/><br/><br/>\n",
    "\n",
    "> *Known issues:*\n",
    ">\n",
    "> * [[Issue 19394]](https://github.com/apache/beam/issues/19394)\n",
    ">   `DoFn.teardown()` metrics are lost."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "related-transforms"
   },
   "source": [
    "## Related transforms\n",
    "\n",
    "* [Map](https://beam.apache.org/documentation/transforms/python/elementwise/map) behaves the same, but produces exactly one output for each input.\n",
    "* [FlatMap](https://beam.apache.org/documentation/transforms/python/elementwise/flatmap) behaves the same as `Map`,\n",
    "  but for each input it may produce zero or more outputs.\n",
    "* [Filter](https://beam.apache.org/documentation/transforms/python/elementwise/filter) is useful if the function is just\n",
    "  deciding whether to output an element or not.\n",
    "\n",
    "<table align=\"left\" style=\"margin-right:1em\">\n",
    "  <td>\n",
    "    <a class=\"button\" target=\"_blank\" href=\"https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.core.html#apache_beam.transforms.core.ParDo\"><img src=\"https://beam.apache.org/images/logos/sdks/python.png\" width=\"32px\" height=\"32px\" alt=\"Pydoc\"/> Pydoc</a>\n",
    "  </td>\n",
    "</table>\n",
    "\n",
    "<br/><br/><br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "view-the-docs-bottom"
   },
   "source": [
    "<table align=\"left\"><td><a target=\"_blank\" href=\"https://beam.apache.org/documentation/transforms/python/elementwise/pardo\"><img src=\"https://beam.apache.org/images/logos/full-color/name-bottom/beam-logo-full-color-name-bottom-100.png\" width=\"32\" height=\"32\" />View the docs</a></td></table>"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "ParDo - element-wise transform",
   "toc_visible": true
  },
  "kernelspec": {
   "display_name": "python3",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}