348 lines (260 sloc) 12.8 KB


Docco is a quick-and-dirty documentation generator, written in Literate CoffeeScript. It produces an HTML document that displays your comments intermingled with your code. All prose is passed through Markdown, and code is passed through Highlight.js syntax highlighting. This page is the result of running Docco against its own source file.

  1. Install Docco with npm: sudo npm install -g docco

  2. Run it against your code: docco src/*.coffee

There is no "Step 3". This will generate an HTML page for each of the named source files, with a menu linking to the other pages, saving the whole mess into a docs folder (configurable).

The Docco source is available on GitHub, and is released under the MIT license.

Docco can be used to process code written in any programming language. If it doesn't handle your favorite yet, feel free to add it to the list. Finally, the "literate" style of any language is also supported — just tack an .md extension on the end:,, and so on.

Partners in Crime:

  • If Node.js doesn't run on your platform, or you'd prefer a more convenient package, get Ryan Tomayko's Rocco, the Ruby port that's available as a gem.

  • If you're writing shell scripts, try Shocco, a port for the POSIX shell, also by Mr. Tomayko.

  • If Python is more your speed, take a look at Nick Fitzgerald's Pycco.

  • For Clojure fans, Fogus's Marginalia is a bit of a departure from "quick-and-dirty", but it'll get the job done.

  • There's a Go port called Gocco, written by Nikhil Marathe.

  • For all you PHP buffs out there, Fredi Bach's sourceMakeup (we'll let the faux pas with respect to our naming scheme slide), should do the trick nicely.

  • Lua enthusiasts can get their fix with Robert Gieseke's Locco.

  • And if you happen to be a .NET aficionado, check out Don Wilson's Nocco.

  • Going further afield from the quick-and-dirty, Groc is a CoffeeScript fork of Docco that adds a searchable table of contents, and aims to gracefully handle large projects with complex hierarchies of code.

Note that not all ports will support all Docco features ... yet.

Main Documentation Generation Functions

Generate the documentation for our configured source file by copying over static assets, reading all the source files in, splitting them up into prose+code sections, highlighting each file in the appropriate language, and printing them out in an HTML template.

document = (options = {}, callback) ->
  config = configure options

  fs.mkdirs config.output, ->

    callback or= (error) -> throw error if error
    copyAsset  = (file, callback) ->
      return callback() unless fs.existsSync file
      fs.copy file, path.join(config.output, path.basename(file)), callback
    complete   = ->
      copyAsset config.css, (error) ->
        return callback error if error
        return copyAsset config.public, callback if fs.existsSync config.public

    files = config.sources.slice()

    nextFile = ->
      source = files.shift()
      fs.readFile source, (error, buffer) ->
        return callback error if error

        code = buffer.toString()
        sections = parse source, code, config
        format source, sections, config
        write source, sections, config
        if files.length then nextFile() else complete()


Given a string of source code, parse out each block of prose and the code that follows it — by detecting which is which, line by line — and then create an individual section for it. Each section is an object with docsText and codeText properties, and eventually docsHtml and codeHtml as well.

parse = (source, code, config = {}) ->
  lines    = code.split '\n'
  sections = []
  lang     = getLanguage source, config
  hasCode  = docsText = codeText = ''

  save = ->
    sections.push {docsText, codeText}
    hasCode = docsText = codeText = ''

Our quick-and-dirty implementation of the literate programming style. Simply invert the prose and code relationship on a per-line basis, and then continue as normal below.

  if lang.literate
    isText = maybeCode = yes
    for line, i in lines
      lines[i] = if maybeCode and match = /^([ ]{4}|[ ]{0,3}\t)/.exec line
        isText = no
      else if maybeCode = /^\s*$/.test line
        if isText then lang.symbol else ''
        isText = yes
        lang.symbol + ' ' + line

  for line in lines
    if line.match(lang.commentMatcher) and not line.match(lang.commentFilter)
      save() if hasCode
      docsText += (line = line.replace(lang.commentMatcher, '')) + '\n'
      save() if /^(---+|===+)$/.test line
      hasCode = yes
      codeText += line + '\n'


To format and highlight the now-parsed sections of code, we use Highlight.js over stdio, and run the text of their corresponding comments through Markdown, using Marked.

format = (source, sections, config) ->
  language = getLanguage source, config

Pass any user defined options to Marked if specified via command line option

  markedOptions =
    smartypants: true

  if config.marked
    markedOptions = config.marked

  marked.setOptions markedOptions

Tell Marked how to highlight code blocks within comments, treating that code as either the language specified in the code block or the language of the file if not specified.

  marked.setOptions {
    highlight: (code, lang) ->
      lang or=

      if highlightjs.getLanguage(lang)
        highlightjs.highlight(lang, code).value
        console.warn "docco: couldn't highlight code block with unknown language '#{lang}' in #{source}"

  for section, i in sections
    code = highlightjs.highlight(, section.codeText).value
    code = code.replace(/\s+$/, '')
    section.codeHtml = "<div class='highlight'><pre>#{code}</pre></div>"
    section.docsHtml = marked(section.docsText)

Once all of the code has finished highlighting, we can write the resulting documentation file by passing the completed HTML sections into the template, and rendering it to the specified output path.

write = (source, sections, config) ->

  destination = (file) ->
    path.join(config.output, path.basename(file, path.extname(file)) + '.html')

The title of the file is either the first heading in the prose, or the name of the source file.

  firstSection = _.find sections, (section) ->
    section.docsText.length > 0
  first = marked.lexer(firstSection.docsText)[0] if firstSection
  hasTitle = first and first.type is 'heading' and first.depth is 1
  title = if hasTitle then first.text else path.basename source

  html = config.template {sources: config.sources, css: path.basename(config.css),
    title, hasTitle, sections, path, destination,}

  console.log "docco: #{source} -> #{destination source}"
  fs.writeFileSync destination(source), html


Default configuration options. All of these may be extended by user-specified options.

defaults =
  layout:     'parallel'
  output:     'docs'
  template:   null
  css:        null
  extension:  null
  languages:  {}
  marked:     null

Configure this particular run of Docco. We might use a passed-in external template, or one of the built-in layouts. We only attempt to process source files for languages for which we have definitions.

configure = (options) ->
  config = _.extend {}, defaults, _.pick(options, _.keys(defaults)...)

  config.languages = buildMatchers config.languages

The user is able to override the layout file used with the --template parameter. In this case, it is also neccessary to explicitly specify a stylesheet file. These custom templates are compiled exactly like the predefined ones, but the public folder is only copied for the latter.

  if options.template
    unless options.css
      console.warn "docco: no stylesheet file specified"
    config.layout = null
    dir = config.layout = path.join __dirname, 'resources', config.layout
    config.public       = path.join dir, 'public' if fs.existsSync path.join dir, 'public'
    config.template     = path.join dir, 'docco.jst'
    config.css          = options.css or path.join dir, 'docco.css'
  config.template = _.template fs.readFileSync(config.template).toString()

  if options.marked
    config.marked = JSON.parse fs.readFileSync(options.marked)

  config.sources = options.args.filter((source) ->
    lang = getLanguage source, config
    console.warn "docco: skipped unknown type (#{path.basename source})" unless lang


Helpers & Initial Setup

Require our external dependencies.

_           = require 'underscore'
fs          = require 'fs-extra'
path        = require 'path'
marked      = require 'marked'
commander   = require 'commander'
highlightjs = require 'highlight.js'

Languages are stored in JSON in the file resources/languages.json. Each item maps the file extension to the name of the language and the symbol that indicates a line comment. To add support for a new programming language to Docco, just add it to the file.

languages = JSON.parse fs.readFileSync(path.join(__dirname, 'resources', 'languages.json'))

Build out the appropriate matchers and delimiters for each language.

buildMatchers = (languages) ->
  for ext, l of languages

Does the line begin with a comment?

    l.commentMatcher = ///^\s*#{l.symbol}\s?///

Ignore hashbangs and interpolations...

    l.commentFilter = /(^#![/]|^\s*#\{)/
languages = buildMatchers languages

A function to get the current language we're documenting, based on the file extension. Detect and tag "literate" variants.

getLanguage = (source, config) ->
  ext  = config.extension or path.extname(source) or path.basename(source)
  lang = config.languages?[ext] or languages[ext]
  if lang and is 'markdown'
    codeExt = path.extname(path.basename(source, ext))
    if codeExt and codeLang = languages[codeExt]
      lang = _.extend {}, codeLang, {literate: yes}

Keep it DRY. Extract the docco version from package.json

version = JSON.parse(fs.readFileSync(path.join(__dirname, 'package.json'))).version

Command Line Interface

Finally, let's define the interface to run Docco from the command line. Parse options using Commander.

run = (args = process.argv) ->
  c = defaults
    .usage('[options] files')
    .option('-L, --languages [file]', 'use a custom languages.json', _.compose JSON.parse, fs.readFileSync)
    .option('-l, --layout [name]',    'choose a layout (parallel, linear or classic)', c.layout)
    .option('-o, --output [path]',    'output to a given folder', c.output)
    .option('-c, --css [file]',       'use a custom css file', c.css)
    .option('-t, --template [file]',  'use a custom .jst template', c.template)
    .option('-e, --extension [ext]',  'assume a file extension for all inputs', c.extension)
    .option('-m, --marked [file]',    'use custom marked options', c.marked)
    .name = "docco"
  if commander.args.length
    document commander
    console.log commander.helpInformation()

Public API

Docco = module.exports = {run, document, parse, format, version}