Nuclear Planning Manual


    1.0.1 • Public • Published

    Parse Google Docs

    This is a parser for journo-developers. It's for this workflow:

    1. Share a Google Doc in your newsroom
    2. Collaborate on the story
    3. Publish that story through your code
    4. Repeat as necessary

    Google Docs is fantastic for editing but icky for publishing. This library lets you decide how to convert your Google Doc to HTML (or any other format).


    google-docs-markup is a parser that outputs a JSON Array. You can render its output with any programming language. To use the parser, though, you'll need NodeJS, at least version 5.11.

    Using google-docs-markup

    First, install it:

    npm init -y # if you haven't already
    npm install --save-dev google-docs-markup

    Then pass it HTML:

    const gdm = require('google-docs-markup')
    const html = '' // get a document's HTML somehow
    const blocks = gdm.parse_google_docs_html(html)

    Now you have "Blocks"!

    Downloading from the Google Drive API

    We skipped the hard part: grabbing the HTML from the Google Drive API. Beware: you must use the API, because "published" Google Docs use different HTML conventions. The API uses OAuth, so it's confusing. We suggest you use google-docs-markup offline and store the resulting documents in your project's repository. Here's how we do it:

    Install google-docs-console-download:

    npm install --save-dev google-docs-console-download

    Create a function to download a file:

    const gdcd = require('google-docs-console-download')(null)
    function downloadGoogleDocAsBlocks(docId, callback) {, (err, html) => {
        if (err) return callback(err)
        const blocks = gdm.parse_google_docs_html(html)
        callback(null, blocks)

    And if your project uses multiple files, maybe you'll want a config file. Perhaps write "config/google-docs.json" with these contents:

      { "slug": "index", "googleId": "1qLoJYmUEJvpQdP4Xplp6I5JBsMpRY9RZTnak2gPhiEQ" },
      { "slug": "page1", "googleId": "1qLoJYmUEJvpQdP4Xplp6I5JBsMpRY9RZTnak2gPhiEQ" }

    Write a script that downloads them all:

    const docs_config = require('./config/google-docs')
    const fs = require('fs')
    function downloadAndSaveAll(docs, dirname, callback) {
      const todo = docs.slice()
      (function tick() {
        const doc = todo.shift()
        if (!doc) return callback()
        downloadGoogleDocAsBlocks(doc.googleId, (err, blocks) => {
          if (err) return callback(err)
          fs.writeFile(`${dirname}/${doc.slug}.json`, JSON.stringify(blocks), callback)
    downloadAndSaveAll(docs_config, '.', (err) => {
      if (err) console.error(err)

    Run that script manually every time editors modify the Google Docs.

    Now you have JSON markup in ./index.json and ./page1.json. If you're loading it in Node, it's easy:

    const index_blocks = require('./index')
    const page1_blocks = require('./page1')

    "Blocks" Output Format

    google-docs-markup output looks like this:

    const blocks = [
      { type: 'h1', texts: [ { type: '', text: 'The title' } ] },
      { type: 'p', texts: [
        { text: 'This sentence contains ' },
        { text: 'italics', italic: true },
        { text: '' },
        { text: 'bold italics', italic: true, bold: true },
        { text: ' and ' },
        { text: 'underlines', underline: true },
        { text: '.' }
      ] },
      { type: 'page-break' },
      { type: 'hr' },
      { type: 'ul', blocks: [
        { type: 'li', texts: [ { text: 'List item' } ] }
      ] },
      { type: 'ol', blocks: [
        { type: 'li', texts: [ { text: 'Numbered list item' } ] }
      ] }


    Think of this tree as a super-duper-simple HTML document. A document is made of "blocks". Some blocks have "block" children. Other blocks have "text" children. We never mix the two.

    Blocks can have these types, derived from HTML:

    • h1, h2, h3, h4: headers. Contain texts.
    • p: paragraph. Contains texts.
    • ul, ol: lists. Contain li blocks.
    • li: list item. Contains texts.
    • hr: horizontal rule. Contains nothing.
    • page-break: page break. Contains nothing. Not HTML.

    There is no recursive nesting. You can't build a list of lists. You can't put a header in a list or a list in a paragraph.

    There are no attributes. Blocks don't have HTML classes or any other formatting information. We leave that to texts.

    There are no empty blocks. If there's no text, there's nothing.


    Texts always have a text attribute. They can also have:

    • italic: true if the text is italicized in the Google Doc
    • bold: true if the text is bolded in the Google Doc
    • underline: true if the text is underlined in the Google Doc

    There are no empty texts.

    Tips and tricks

    Use underline to build a templating engine

    Here's the nifty thing: you probably aren't going to output underlined text. Who even does that these days? We underline code.

    For example, in our Google Doc, we underline something like render(...). Then when we render our blocks and texts, we replace any text with underlined:true with eval(text).

    Beware smart quotes: Google Docs usually turns render("foo") into render(“foo”), which is invalid in most programming languages. It does the same thing with single quotes. Try to keep those characters out of your underlined code, or your newsroom practices will probably break it:

    • If you're rendering with eval() in JavaScript, write Strings \like this`` (backticks are an alternative to single or double quotes).
    • If you're rendering with eval() in Ruby, write Strings %{like this} (that escape sequence is an alternative to single or double quotes).
    • If you're using a programming language with fewer features (like Python or a compiled language), don't use eval(): parse the underlying code using a mini-language you design yourself (one that doesn't use require single or double quotes).

    Include instructions below a page break

    We maintain an example Google Doc that shows off all features. It doubles as a page of instructions.

    Copy/paste that page into a fresh Google Doc. Write your story above the instructions; then write any notes (a scratchpad), then leave the instructions. Add a page break after the story and before the notes.

    The instructions will help journalists with syntax. And you can easily drop everything after the first page break. For instance:

    const html = ...;
    const blocks = gdm.parse_google_docs_html(html);
    blocks.splice(blocks.find(b => b.type === 'page-break'))

    Adding features and fixing bugs

    1. Run node_modules/.bin/mocha -w and verify all tests pass
    2. Write a new test in the test/ directory and verify that mocha reports failure
    3. Edit code in index.js until tests pass
    4. git commit -a with a descriptive message
    5. Repeat from step 2 until this library has all the features you want and no bugs you know of
    6. git push


    After you've developed, you probably want other people to use your changes.

    So publish to npm:

    1. Edit package.json and set a new semantic version number.
    2. git commit package.json
    3. git push
    4. npm publish

    Then make your other projects depend on this new version:

    npm install --save google-docs-markup


    npm i google-docs-markup

    DownloadsWeekly Downloads






    Last publish


    • adamhooper