Markdown parser and renderer for Hack

Installs: 9 830

Dependents: 3

Suggesters: 0

Security: 0

Stars: 28

Watchers: 10

Forks: 9

Open Issues: 7


v1.6.5 2020-12-21 22:50 UTC

This package is auto-updated.

Last update: 2021-10-12 20:07:09 UTC


FBMarkdown is an extensible parser and renderer for GitHub Flavored Markdown, written in Hack.

It is used to render the Hack and HHVM documentation.

Why another Markdown library?

For docs.hhvm.com, we wanted:

  • GitHub Flavored Markdown for familiarity
  • support for custom extensions

Originally, the Ruby GFM pipeline was the best fit; over time, we started to want to:

  • make it easier and faster to contribute to docs.hhvm.com
  • remove the Ruby dependencies to make it easy to render markdown in other Hack projects
  • produce and mutate an AST before rendering
  • support multiple renders

FBMarkdown exists to address all of these goals.


Installing FBMarkdown

hhvm composer.phar require facebook/fbmarkdown

Using FBMarkdown

use namespace Facebook\Markdown;

function render(string $markdown): string {
  $ast = Markdown\parse(new Markdown\ParserContext(), $markdown);

  $html = (new Markdown\HTMLRenderer(
    new Markdown\RenderContext()

  return $html;

FBMarkdown currently supports three types of Markdown sources, with plans to expand: trusted, sponsored, and user-generated content.

  • Trusted content mode: Embedded HTML is enabled, and all URI schemes are enabled and will be parsed as links. In addition, all images are processed normally.

  • Sponsored mode: HTML rendering is enabled, but limited to allowed tags only (defined in TagFilterExtension, based on the GFM spec). Additionally, URIs are limited to the {http, https, irc, and mailto} schemes, and rel="nofollow ugc" is added to all links.

  • User-generated content: All HTML is disabled, as are links and images regardless of schemes. If links are re-enabled, rel="nofollow ugc" will be added to all links.

To make changes to these default settings:

  • You may alter the keyset of allowed URI schemes by calling the Parser function setAllowedURISchemes().
  • You may enable embedded HTML by calling the Parser function enableHTML_UNSAFE(). N.B.: For complete compatibility with GitHub Flavored Markdown, support for embedded HTML must be enabled.
  • You may disable image filtering by calling the Renderer function disableImageFiltering().
  • You may add rel="nofollow ugc" to all links by calling the Renderer function addNoFollowUGCAllLinks().

If you are re-using contexts to render multiple independent snippets, you will need to call ->resetFileData() on the context.

How FBMarkdown works


  1. The classes in the Facebook\Markdown\UnparsedBlocks namespace convert markdown text to a tree of nodes representing the block structure of the document, however the content of the blocks is unparsed.
  2. The contents of the blocks ('inlines') are parsed using the classes in the Facebook\Markdown\Inlines namespace.
  3. Finally, the classes of the Facebook\Markdown\Blocks namespace are used to represent the fully parsed AST - blocks and Inlines.


The AST is recursively walked, emitting output for each note - e.g. the HTML renderer produces strings.

Extending FBMarkdown

There are 2 main ways to extend FBMarkdown: extending the parser, and transforming the AST.

Extending The Parser


Extend Facebook\Markdown\Inlines\Inline or a subclass, and pass your classname to $render_ctx->getInlineContext()->prependInlineTypes(...).

There are then several approaches to rendering:

  • instantiate your subclass, and add support for it to a custom renderer
  • instantiate your subclass, and make it implement the Facebook\Markdown\RenderableAsHTML interface
  • if it could be replaced with several existing inlines, return a Facebook\Markdown\Inlines\InlineSequence, then you won't need to extend the renderer.


You will need to implement the Facebook\Markdown\UnparsedBlocks\BlockProducer interface, and pass your classname to $render_ctx->getBlockContext()->prependBlockTypes(...).

There are then several approaches to rendering:

  • create a subclass of Block, and add support for it to a custom renderer
  • create a subclass of Block, and make it implement the Facebook\Markdown\RenderableAsHTML interface
  • if it could be replaced with several existing blocks, return a Facebook\Markdown\Blocks\BlockSequence
  • if it could be replaced with a paragraph of inlines, return a Facebook\Markdown\Blocks\InlineSequenceBlock

Transforming The AST

Extend Facebook\Markdown\RenderFilter, and pass it to $render_ctx->appendFilters(...).


The Hack and HHVM documentation uses most of these approaches; see:


FBMarkdown is MIT-licensed.

FBMarkdown may contain third-party software; see third_party_notices.txt for details.