A markdown parser built for speed


A full-featured markdown parser and compiler, written in JavaScript. Built for speed.

npm install marked --save

Minimal usage:

console.log(marked('I am using __markdown__.'));
// Outputs: <p>I am using <i>markdown</i>.</p> 

Example using all options:

// Set default options except highlight which has no default 
  gfm: true,
  highlightfunction (codelangcallback) {
    pygmentize({ lang: lang, format: 'html' }, code, function (errresult) {
      callback(err, result.toString());
  tables: true,
  breaks: false,
  pedantic: false,
  sanitize: true,
  smartLists: true,
  smartypants: false,
  langPrefix: 'lang-'
console.log(marked('I am using __markdown__.'));

Type: String

String of markdown source to be compiled.

Type: Object

Hash of options. Can also be set using the marked.setOptions method as seen above.

Type: Function

Function called when the markdownString has been fully parsed. If the options argument is omitted, this can be used as the second argument as follows:

marked(markdownString, function (errcontent) {
  if (err) throw err;
  // Outputs parsed html 

Type: Boolean Default: true

Enable GitHub flavored markdown.

Type: Function

A function to highlight code blocks. The function takes three arguments: code, lang, and callback. The above example uses async highlighting with node-pygementize-bundled, and here is a synchronous example using highlight.js which doesn't require the callback argument:

  highlightfunction (langcode) {
    return hljs.highlightAuto(lang, code).value;


Type: String

The section of code to pass to the highlighter.


Type: String

The programming language specified in the code block.


Type: String

The callback function to call when using an async highlighter.

Type: Boolean Default: true

Enable GFM tables. This option requires the gfm option to be true.

Type: Boolean Default: false

Enable GFM line breaks. This option requires the gfm option to be true.

Type: Boolean Default: false

Conform to obscure parts of as much as possible. Don't fix any of the original markdown bugs or poor behavior.

Type: Boolean Default: true

Sanitize the output. Ignore any HTML that has been input.

Type: Boolean Default: true

Use smarter list behavior than the original markdown. May eventually be default with the old behavior moved into pedantic.

Type: Boolean Default: false

Use "smart" typograhic punctuation for things like quotes and dashes.

Type: String Default: lang-

Set the prefix for code block classes.

You also have direct access to the lexer and parser if you so desire.

var tokens = marked.lexer(text, options);
var lexer = new marked.Lexer(options);
var tokens = lexer.lex(text);
$ marked -o hello.html
hello world
$ cat hello.html
<p>hello world</p>

node v0.4.x

$ node test --bench
marked completed in 12071ms.
showdown (reuse converter) completed in 27387ms.
showdown (new converter) completed in 75617ms.
markdown-js completed in 70069ms.

node v0.6.x

$ node test --bench
marked completed in 6448ms.
marked (gfm) completed in 7357ms.
marked (pedantic) completed in 6092ms.
discount completed in 7314ms.
showdown (reuse converter) completed in 16018ms.
showdown (new converter) completed in 18234ms.
markdown-js completed in 24270ms.

Marked is now faster than Discount, which is written in C.

For those feeling skeptical: These benchmarks run the entire markdown test suite 1000 times. The test suite tests every feature. It doesn't cater to specific aspects.

node v0.8.x

$ node test --bench
marked completed in 3411ms.
marked (gfm) completed in 3727ms.
marked (pedantic) completed in 3201ms.
robotskirt completed in 808ms.
showdown (reuse converter) completed in 11954ms.
showdown (new converter) completed in 17774ms.
markdown-js completed in 17191ms.

The point of marked was to create a markdown compiler where it was possible to frequently parse huge chunks of markdown without having to worry about caching the compiled output somehow...or blocking for an unnecesarily long time.

marked is very concise and still implements all markdown features. It is also now fully compatible with the client-side.

marked more or less passes the official markdown test suite in its entirety. This is important because a surprising number of markdown compilers cannot pass more than a few tests. It was very difficult to get marked as compliant as it is. It could have cut corners in several areas for the sake of performance, but did not in order to be exactly what you expect in terms of a markdown rendering. In fact, this is why marked could be considered at a disadvantage in the benchmarks above.

Along with implementing every markdown feature, marked also implements GFM features.

$ node
> require('marked').lexer('> i am using marked.')
[ { type: 'blockquote_start' },
  { type: 'paragraph',
    text: 'i am using marked.' },
  { type: 'blockquote_end' },
  links: {} ]

Copyright (c) 2011-2013, Christopher Jeffrey. (MIT License)

See LICENSE for more info.