MMark status update


Published on February 14, 2018

In November 2017 I announced MMark—a the strict markdown processor for writers. I worked on it actively for some time since then and this post is a little update about the project, which by the way even has its own GitHub organization now.

Getting mature

The initial release I made in November could not even handle blockquotes and lists. The current version closely follows the CommonMark specification and in addition to that supports:

  • parsing of an optional YAML block
  • strikeout using ~~this~~ syntax
  • superscript using ^this^ syntax
  • subscript using ~this~ syntax
  • automatic assignment of ids to headers
  • pipe tables (as on GitHub)

I’m getting interactive feedback from MMark every time I write markdown thanks to a little bit of Emacs Lisp code, which I’ll cover later in the post. As expected, this caused an expansion of the test suite, which currently has 632 tests. Most of them were adapted from the CommonMark spec, but many are custom. I think the library has become quite smooth by now and I finally can recommend it for serious use.


In addition to the built-in functionality, the mmark-ext package provides many useful extensions:

  • Text.MMark.Extension.Comment—turn paragraphs into comments by starting them with a magic prefix
  • Text.MMark.Extension.FontAwesome—insert FontAwesome icons
  • Text.MMark.Extension.Footnotes—insert footnotes
  • Text.MMark.Extension.Kbd—wrap text with kbd tags (MMark does not support arbitrary inline HTML, so for this sort of thing you need a proper extension)
  • Text.MMark.Extension.LinkTarget—specify target attribute of links, so you can make your link e.g. open in new tab
  • Text.MMark.Extension.MathJax—MathJax support
  • Text.MMark.Extension.ObfuscateEmail—email obfuscation
  • Text.MMark.Extension.PunctuationPrettifier—the usual goodies for typographically correct punctuation
  • Text.MMark.Extension.Skylighting—highlight your code blocks like with Pandoc
  • Text.MMark.Extension.TableOfContents—generate and insert tables of contents

If you look closely at the extensions, you’ll find that they do not introduce any new fancy syntax (in fact the extension mechanism cannot affect parsing at all), but rather give additional meaning to the existing markdown syntax/markup structures. I think it’s a nicer approach because it prevents the markup language from growing into a monstrosity like in Pandoc, but your opinion may be different.

The Text.MMark.Extension module describes the extension system in more detail.


I have compared speed and memory consumption of various Haskell markdown libraries by running them on an identical, big-enough markdown document and by rendering it as HTML:

LibraryParsing libraryExecution timeAllocatedMax residency
cmark-0.5.6Custom C code323.4 μs228,4409,608
mmark- ms26,180,27237,792
cheapskate-0.1.1Custom Haskell code10.76 ms44,686,272799,200
markdown-0.1.16Attoparsec14.13 ms69,261,816699,656
pandoc-2.0.5Parsec37.90 ms141,868,8401,471,080

Results are ordered from fastest to slowest.

MMark is the only markdown processor in Haskell without a severe space leak in it. This is mostly because of latest updates in Megaparsec, which were inspired by a suggestion to move from Applicative/Alternative-based combinators to Monad/MonadPlus-based ones. I should thank Vladislav Zavialov for the suggestion. The new combinators are available for everyone in parser-combinators (see Control.Monad.Combinators) go grab them and use for great good.

Command line application

There is a command line app now:

mmark—command line interface to MMark markdown processor

Usage: mmark [-v|--version] [-i|--ifile IFILE] [-o|--ofile OFILE] [-j|--json]
             [-t|--template FILE] [--ext-comment PREFIX] [--ext-font-awesome]
             [--ext-footnotes] [--ext-kbd] [--ext-link-target] [--ext-mathjax]
             [--ext-obfuscate-email CLASS] [--ext-punctuation]
             [--ext-skylighting] [--ext-toc RANGE]
  Command line interface to MMark markdown processor

Available options:
  -h,--help                Show this help text
  -v,--version             Print version of the program
  -i,--ifile IFILE         Read markdown source from this file (otherwise read
                           from stdin)
  -o,--ofile OFILE         Save rendered HTML document to this file (otherwise
                           write to stdout)
  -j,--json                Output parse errors and result in JSON format
  -t,--template FILE       Use the template located at this path
  --ext-comment PREFIX     Remove paragraphs that start with the given prefix
  --ext-font-awesome       Enable support for inserting font awesome icons
  --ext-footnotes          Enable support for footnotes
  --ext-kbd                Enable support for wrapping things in kbd tags
  --ext-link-target        Enable support for specifying link targets
  --ext-mathjax            Enable support for MathJax formulas
  --ext-obfuscate-email CLASS
                           Obfuscate email addresses assigning the specified
  --ext-punctuation        Enable punctuation prettifier
  --ext-skylighting        Enable syntax highlighting of code snippets with
  --ext-toc RANGE          Enable generation of table of contents using the
                           supplied range of headers to include, e.g. "1-6" or

It can also be used for playing with the markdown processor almost interactively because by default input is read from standard input and output is printed to standard output:

$ mmark
So *here* we go!
----------------------- Control-D
<p>So <em>here</em> we go!</p>

Flycheck checker for Emacs users

Finally, Emacs users may find the flycheck-mmark package useful. It defines a Flycheck checker which calls mmark command line app and displays parse errors in the buffer you’re editing. It’s fun and sort of strange to have your markdown checked in this way.

It’s available via MELPA, so you can just M-x package-install RET flycheck-mmark, but make sure that you have mmark on your PATH. Full setup instructions are given in the readme.


This site is fully powered by MMark now. I can’t say that I could actually make an early switch, although I wanted to. I had to wait till MMark becomes more mature because I used a few Pandoc extensions, such as footnotes, so I was in a vendor lock of a sort. However, MMark currently is smooth and powerful enough for me to use it here. It also allowed me to get a more customized markdown processor for my needs because I could easily add several special extensions. For example, I can interpolate my contact info such as email or Twitter account without hardcoding it. Also, previously I used some JavaScript to add Bootstrap classes to tables and images, now it’s done via MMark extensions statically.

So give MMark a try next time you decide to create a static blog or something!