This package provides Haskell bindings for libcmark, the reference parser for CommonMark, a fully specified variant of Markdown. It includes sources for libcmark and does not require prior installation of the C library.
cmark provides the following advantages over existing Markdown libraries for Haskell:
Speed: cmark can render a Markdown version of War and Peace in the blink of an eye. Conversion speed is on par with the sundown library, though we were unable to benchmark precisely, because sundown raised a malloc error when compiled into our benchmark suite. Relative to other Haskell Markdown libraries: cmark was 82 times faster than cheapskate, 59 times faster than markdown, 105 times faster than pandoc, and 3 times faster than discount.
Robustness: cmark can handle whatever is thrown at it, without the exponential blowups in parsing time that sometimes afflict other libraries. (The input
bench/full-sample.md, for example, causes both pandoc and markdown to grind to a halt.) libcmark has been extensively fuzz-tested.
Accuracy: cmark passes the CommonMark spec’s suite of over 600 conformance tests.
Standardization: Since there is a spec and a comprehensive suite of tests, we can have a high degree of confidence that any two CommonMark implementations will behave the same. Thus, for example, one could use this library for server-side rendering and commonmark.js for client-side previewing.
Multiple renderers. Output in HTML, groff man, LaTeX, CommonMark, and a custom XML format is supported. And it is easy to write new renderers to support other formats.
Ease of installation: cmark is portable and has minimal dependencies.
cmark does not provide Haskell versions of the whole libcmark
API, which is built around mutable
cmark_node objects. Instead, it
provides functions for converting CommonMark to HTML (and other
formats), and a function for converting CommonMark to a
tree that can be processed further using Haskell.
A note on security: This library does not attempt to sanitize
HTML output. We recommend using xss-sanitize to filter the output,
optSafe to filter out all raw HTML and potentially
A note on stability: There is a good chance the API will change significantly after this early release.