commonmark.js

Build Status NPM version

CommonMark is a rationalized version of Markdown syntax, with a spec and BSD-licensed reference implementations in C and JavaScript.

For more information, see http://commonmark.org.

This repository contains the JavaScript reference implementation. It provides a library with functions for parsing CommonMark documents to an abstract syntax tree (AST), manipulating the AST, and rendering the document to HTML or to an XML representation of the AST.

To play with this library without installing it, see the live dingus at http://try.commonmark.org/.

Installing

You can install the library using npm:

npm install commonmark

This package includes the commonmark library and a command-line executable, commonmark.

For client-side use, you can use one of the single-file distributions provided in the dist/ subdirectory of the node installation (node_modules/commonmark/dist/). Use either commonmark.js (readable source) or commonmark.min.js (minimized source).

Alternatively, bower install commonmark will install the needed distribution files in bower_components/commonmark/dist.

You can also use the version hosted by unpkg: for example, https://unpkg.com/commonmark@0.29.3/dist/commonmark.js for the unminimized version 0.29.3.

Building

Make sure to fetch dependencies with:

npm install

To run tests for the JavaScript library:

npm test

(Running the tests will also rebuild distribution files in dist/.)

To run benchmarks against some other JavaScript converters:

make bench

To start an interactive dingus that you can use to try out the library:

make dingus

Usage

Instead of converting Markdown directly to HTML, as most converters do, commonmark.js parses Markdown to an AST (abstract syntax tree), and then renders this AST as HTML. This opens up the possibility of manipulating the AST between parsing and rendering. For example, one could transform emphasis into ALL CAPS.

Here's a basic usage example:

var reader = new commonmark.Parser();
var writer = new commonmark.HtmlRenderer();
var parsed = reader.parse("Hello *world*"); // parsed is a 'Node' tree
// transform parsed if you like...
var result = writer.render(parsed); // result is a String

The constructors for Parser and HtmlRenderer take an optional options parameter:

var reader = new commonmark.Parser({smart: true});
var writer = new commonmark.HtmlRenderer({sourcepos: true});

Parser currently supports the following:

Both HtmlRenderer and XmlRenderer (see below) support these options:

For example, to make soft breaks render as hard breaks in HTML:

var writer = new commonmark.HtmlRenderer({softbreak: "<br />"});

To make them render as spaces:

var writer = new commonmark.HtmlRenderer({softbreak: " "});

XmlRenderer serves as an alternative to HtmlRenderer and will produce an XML representation of the AST:

var writer = new commonmark.XmlRenderer({sourcepos: true});

The parser returns a Node. The following public properties are defined (those marked "read-only" have only a getter, not a setter):

Nodes have the following public methods:

The NodeWalker returned by walker() has two methods:

Here is an example of the use of a NodeWalker to iterate through the tree, making transformations. This simple example converts the contents of all text nodes to ALL CAPS:

var walker = parsed.walker();
var event, node;

while ((event = walker.next())) {
  node = event.node;
  if (event.entering && node.type === 'text') {
    node.literal = node.literal.toUpperCase();
  }
}

This more complex example converts emphasis to ALL CAPS:

var walker = parsed.walker();
var event, node;
var inEmph = false;

while ((event = walker.next())) {
  node = event.node;
  if (node.type === 'emph') {
    if (event.entering) {
      inEmph = true;
    } else {
      inEmph = false;
      // add Emph node's children as siblings
      while (node.firstChild) {
        node.insertBefore(node.firstChild);
      }
      // remove the empty Emph node
      node.unlink()
    }
  } else if (inEmph && node.type === 'text') {
      node.literal = node.literal.toUpperCase();
  }
}

Exercises for the reader: write a transform to

  1. De-linkify a document, transforming links to regular text.
  2. Remove all raw HTML (html_inline and html_block nodes).
  3. Run fenced code blocks marked with a language name through a syntax highlighting library, replacing them with an HtmlBlock containing the highlighted code.
  4. Print warnings to the console for images without image descriptions or titles.

Command line

The command line executable parses CommonMark input from the specified files, or from stdin if no files are specified, and renders the result to stdout as HTML. If multiple input files are specified, their contents are concatenated before parsing, with newlines between them.

commonmark inputfile.md > outputfile.html
commonmark intro.md chapter1.md chapter2.md > book.html

Use commonmark --help to get a summary of options.

A note on security

The library does not attempt to sanitize link attributes or raw HTML. If you use this library in applications that accept untrusted user input, you should either enable the safe option (see above) or run the output through an HTML sanitizer to protect against XSS attacks.

Performance

Performance is excellent, roughly on par with marked. On a benchmark converting an 11 MB Markdown file built by concatenating the Markdown sources of all localizations of the first edition of Pro Git by Scott Chacon, the command-line tool, commonmark is just a bit slower than the C program discount, roughly ten times faster than PHP Markdown, a hundred times faster than Python Markdown, and more than a thousand times faster than Markdown.pl.

Here are some focused benchmarks of four JavaScript libraries (using versions available on 24 Jan 2015). They test performance on different kinds of Markdown texts. (Most of these samples are taken from the markdown-it repository.) Results show a ratio of ops/second (higher is better) against showdown (which is usually the slowest implementation). Versions: showdown 1.3.0, marked 0.3.5, commonmark.js 0.22.1, markdown-it 5.0.2, node 5.3.0. Hardware: 1.6GHz Intel Core i5, Mac OSX.

| Sample |showdown |commonmark|marked |markdown-it| |--------------------------|---------:|---------:|---------:|----------:| |README.md | 1| 3.6| 3.1| 3.9| |block-bq-flat.md | 1| 4.8| 4.9| 4.9| |block-bq-nested.md | 1| 11.9| 6.8| 10.7| |block-code.md | 1| 4.7| 12.1| 23.0| |block-fences.md | 1| 6.2| 21.2| 19.1| |block-heading.md | 1| 5.0| 4.8| 6.5| |block-hr.md | 1| 3.5| 3.3| 3.5| |block-html.md | 1| 2.1| 0.9| 3.8| |block-lheading.md | 1| 5.1| 4.9| 3.9| |block-list-flat.md | 1| 4.7| 4.4| 7.4| |block-list-nested.md | 1| 9.5| 7.8| 17.6| |block-ref-flat.md | 1| 0.8| 0.5| 0.6| |block-ref-nested.md | 1| 0.7| 0.6| 0.9| |inline-autolink.md | 1| 2.3| 3.4| 2.5| |inline-backticks.md | 1| 7.6| 5.3| 8.2| |inline-em-flat.md | 1| 1.5| 1.1| 1.6| |inline-em-nested.md | 1| 1.8| 1.3| 1.7| |inline-em-worst.md | 1| 2.4| 1.5| 2.5| |inline-entity.md | 1| 2.0| 3.8| 2.7| |inline-escape.md | 1| 2.2| 1.4| 5.0| |inline-html.md | 1| 2.9| 3.7| 3.3| |inline-links-flat.md | 1| 2.7| 2.7| 2.2| |inline-links-nested.md | 1| 1.4| 0.5| 0.5| |inline-newlines.md | 1| 2.3| 2.0| 3.5| |lorem1.md | 1| 6.0| 2.9| 3.3| |rawtabs.md | 1| 4.6| 3.9| 6.7|

To generate this table:

make bench-detailed

Authors

John MacFarlane wrote the first version of the JavaScript implementation. The block parsing algorithm was worked out together with David Greenspan. Kārlis Gaņģis helped work out a better parsing algorithm for links and emphasis, eliminating several worst-case performance issues. Vitaly Puzrin has offered much good advice about optimization and other issues.