node package manager


a strongly specified, highly compatible variant of Markdown


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

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


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 do make dist to produce a standalone JavaScript file js/dist/commonmark.js, suitable for linking into a web page, or fetch the latest from, or bower install commonmark.

To run tests for the JavaScript library:

npm install # if needed to fetch dependencies
make test

To run benchmarks against some other JavaScript converters:

npm install benchmark showdown marked markdown-it
make bench

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

make dingus


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

The following options are currently supported:

  • sourcepos: if true, source position information for block-level elements will be rendered in the data-sourcepos attribute (for HTML) or the sourcepos attribute (for XML).
  • smart: if true, straight quotes will be made curly, -- will be changed to an en dash, --- will be changed to an em dash, and ... will be changed to ellipses.
  • safe: if true, raw HTML will not be passed through to HTML output (it will be replaced by comments), and potentially unsafe URLs in links and images (those beginning with javascript:, vbscript:, file:, and with a few exceptions data:) will be replaced with empty strings.

It is also possible to override the escape and softbreak properties of a renderer. So, to make soft breaks render as hard breaks in HTML:

var writer = new commonmark.HtmlRenderer;
writer.softbreak = "<br />";

To make them render as spaces:

writer.softbreak = " ";

To override escape, pass it a function with two parameters: the first is the string to be escaped, the second is a boolean that is true if the escaped string is to be included in an attribute.

In addition to the HtmlRenderer, there is an XmlRenderer, which 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):

  • type (read-only): a String, one of text, softbreak, linebreak, emph, strong, html_inline, link, image, code, document, paragraph, block_quote, item, list, heading, code_block, html_block, thematic_break.
  • firstChild (read-only): a Node or null.
  • lastChild (read-only): a Node or null.
  • next (read-only): a Node or null.
  • prev (read-only): a Node or null.
  • parent (read-only): a Node or null.
  • sourcepos (read-only): an Array with the following form: [[startline, startcolumn], [endline, endcolumn]].
  • isContainer (read-only): true if the Node can contain other Nodes as children.
  • literal: the literal String content of the node or null.
  • destination: link or image destination (String) or null.
  • title: link or image title (String) or null.
  • info: fenced code block info string (String) or null.
  • level: heading level (Number).
  • listType: a String, either Bullet or Ordered.
  • listTight: true if list is tight.
  • listStart: a Number, the starting number of an ordered list.
  • listDelimiter: a String, either ) or . for an ordered list.
  • onEnter, onExit: Strings, used only for custom_block or custom_inline.

Nodes have the following public methods:

  • appendChild(child): Append a Node child to the end of the Node's children.
  • prependChild(child): Prepend a Node child to the beginning of the Node's children.
  • unlink(): Remove the Node from the tree, severing its links with siblings and parents, and closing up gaps as needed.
  • insertAfter(sibling): Insert a Node sibling after the Node.
  • insertBefore(sibling): Insert a Node sibling before the Node.
  • walker(): Returns a NodeWalker that can be used to iterate through the Node tree rooted in the Node.

The NodeWalker returned by walker() has two methods:

  • next(): Returns an object with properties entering (a boolean, which is true when we enter a Node from a parent or sibling, and false when we reenter it from a child). Returns null when we have finished walking the tree.
  • resumeAt(node, entering): Resets the iterator to resume at the specified node and setting for entering. (Normally this isn't needed unless you do destructive updates to the Node tree.)

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 = {
  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 = {
  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) {
      // remove the empty Emph node 
  } 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 > outputfile.html
commonmark > 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 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

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 1 3.6 3.1 3.9 1 4.8 4.9 4.9 1 11.9 6.8 10.7 1 4.7 12.1 23.0 1 6.2 21.2 19.1 1 5.0 4.8 6.5 1 3.5 3.3 3.5 1 2.1 0.9 3.8 1 5.1 4.9 3.9 1 4.7 4.4 7.4 1 9.5 7.8 17.6 1 0.8 0.5 0.6 1 0.7 0.6 0.9 1 2.3 3.4 2.5 1 7.6 5.3 8.2 1 1.5 1.1 1.6 1 1.8 1.3 1.7 1 2.4 1.5 2.5 1 2.0 3.8 2.7 1 2.2 1.4 5.0 1 2.9 3.7 3.3 1 2.7 2.7 2.2 1 1.4 0.5 0.5 1 2.3 2.0 3.5 1 6.0 2.9 3.3 1 4.6 3.9 6.7

To generate this table,

npm install showdown marked markdown-it benchmark
make bench-detailed


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.