Earmark is a pure-Elixir Markdown converter. It is intended to be used as a library (just call Earmark.as_html), but can also be used as a command-line tool (run mix escript.build first). Output generation is pluggable.



Earmark—A Pure Elixir Markdown Processor

Build Status Hex.pm Hex.pm Hex.pm

Table Of Contents


{ :earmark, "> x.y.z" }




  {:ok, html_doc, []}                   = Earmark.as_html(markdown)
  {:ok, html_doc, deprecation_messages} = Earmark.as_html(markdown)
  {:error, html_doc, error_messages}    = Earmark.as_html(markdown)


  html_doc = Earmark.as_html!(markdown, options)

All messages are printed to stderr.


Options can be passed into as_html or as_html! according to the documentation.

  html_doc = Earmark.as_html!(markdown)

  html_doc = Earmark.as_html!(markdown, options)

Formats the error_messages returned by as_html and adds the filename to each. Then prints them to stderr and just returns the html_doc

NEW and EXPERIMENTAL: Earmark.as_ast

Although well tested the way the exposed AST will look in future versions may change, a stable API is expected for Earmark v1.6, when the rendered HTML shall be derived from the ast too.

More details can be found in the function's description below.

Command line

$ mix escript.build
$ ./earmark file.md

Some options defined in the Earmark.Options struct can be specified as command line switches.


$ ./earmark --help

to find out more, but here is a short example

$ ./earmark --smartypants false --code-class-prefix "a- b-" file.md

will call

Earmark.as_html!( ..., %Earmark.Options{smartypants: false, code_class_prefix: "a- b-"})


Standard Gruber markdown.


Github Flavored Markdown

GFM is supported by default, however as GFM is a moving target and all GFM extension do not make sense in a general context, Earmark does not support all of it, here is a list of what is supported:

  • Strike Through

      iex(1)> Earmark.as_html! ["~~hello~~"]
  • Syntax Highlighting

The generated code blocks have a corresponding class attribute:

  iex(2)> Earmark.as_html! ["```elixir", "   [] |> Enum.into(%{})", "```"]
  "<pre><code class=\"elixir\">   [] |&gt; Enum.into(%{})</code></pre>\n"

which can be customized with the code_class_prefix option

  iex(3)> Earmark.as_html! ["```elixir", "   [] |> Enum.into(%{})", "```"] , %Earmark.Options{code_class_prefix: "lang-"}
  "<pre><code class=\"elixir lang-elixir\">   [] |&gt; Enum.into(%{})</code></pre>\n"
  • Tables

Are supported as long as they are preceded by an empty line.

    State | Abbrev | Capital
    ----: | :----: | -------
    Texas | TX     | Austin
    Maine | ME     | Augusta

Tables may have leading and trailing vertical bars on each line

    | State | Abbrev | Capital |
    | ----: | :----: | ------- |
    | Texas | TX     | Austin  |
    | Maine | ME     | Augusta |

Tables need not have headers, in which case all column alignments default to left.

    | Texas | TX     | Austin  |
    | Maine | ME     | Augusta |

Currently we assume there are always spaces around interior vertical unless there are exterior bars.

However in order to be more GFM compatible the gfm_tables: true option can be used to interpret only interior vertical bars as a table if a seperation line is given, therefor

     Elixir  | awesome

is a table (iff gfm_tables: true) while

     Elixir  | awesome

never is.

Adding HTML attributes with the IAL extension

To block elements

HTML attributes can be added to any block-level element. We use the Kramdown syntax: add the line {: attrs } following the block.

attrs can be one or more of:

  • .className
  • #id
  • name=value, name="value", or name='value'

For example:

    # Warning
    {: .red}

    Do not turn off the engine
    if you are at altitude.
    {: .boxed #warning spellcheck="true"}

To links or images

It is possible to add IAL attributes to generated links or images in the following format.

  iex(4)> markdown = "[link](url) {: .classy}"
  ...(4)> Earmark.as_html(markdown)
  { :ok, "<p><a href=\"url\" class=\"classy\">link</a></p>\n", []}

For both cases, malformed attributes are ignored and warnings are issued.

  iex(5)> [ "Some text", "{:hello}" ] |> Enum.join("\n") |> Earmark.as_html()
  {:error, "<p>Some text</p>\n", [{:warning, 2,"Illegal attributes [\"hello\"] ignored in IAL"}]}

It is possible to escape the IAL in both forms if necessary

  iex(6)> markdown = "[link](url)\\{: .classy}"
  ...(6)> Earmark.as_html(markdown)
  {:ok, "<p><a href=\"url\">link</a>{: .classy}</p>\n", []}

This of course is not necessary in code blocks or text lines containing an IAL-like string, as in the following example

  iex(7)> markdown = "hello {:world}"
  ...(7)> Earmark.as_html!(markdown)
  "<p>hello {:world}</p>\n"


  • Block-level HTML is correctly handled only if each HTML tag appears on its own line. So


    will work. However. the following won't

  • John Gruber's tests contain an ambiguity when it comes to lines that might be the start of a list inside paragraphs.

    One test says that

      This is the text
      * of a paragraph
      that I wrote

    is a single paragraph. The "*" is not significant. However, another test has

      *   A list item
          * an another

    and expects this to be a nested list. But, in reality, the second could just be the continuation of a paragraph.

    I've chosen always to use the second interpretation—a line that looks like a list item will always be a list item.

  • Rendering of block and inline elements.

    Block or void HTML elements that are at the absolute beginning of a line end the preceding paragraph.


      <hr />


      <hr />


       <hr />

    will be transformed into

       <hr /></p>


Syntax Highlighting

All backquoted or fenced code blocks with a language string are rendered with the given language as a class attribute of the code tag.

For example:

  iex(8)> [
  ...(8)>    "```elixir",
  ...(8)>    " @tag :hello",
  ...(8)>    "```"
  ...(8)> ] |> Earmark.as_html!()
  "<pre><code class=\"elixir\"> @tag :hello</code></pre>\n"

will be rendered as shown in the doctest above.

If you want to integrate with a syntax highlighter with different conventions you can add more classes by specifying prefixes that will be put before the language string.

Prism.js for example needs a class language-elixir. In order to achieve that goal you can add language- as a code_class_prefix to Earmark.Options.

In the following example we want more than one additional class, so we add more prefixes.

  Earmark.as_html!(..., %Earmark.Options{code_class_prefix: "lang- language-"})

which is rendering

   <pre><code class="elixir lang-elixir language-elixir">...

As for all other options code_class_prefix can be passed into the earmark executable as follows:

  earmark --code-class-prefix "language- lang-" ...


By default, that is if the timeout option is not set Earmark uses parallel mapping as implemented in Earmark.pmap/2, which uses Task.await with its default timeout of 5000ms.

In rare cases that might not be enough.

By indicating a longer timeout option in milliseconds Earmark will use parallel mapping as implemented in Earmark.pmap/3, which will pass timeout to Task.await.

In both cases one can override the mapper function with either the mapper option (used if and only if timeout is nil) or the mapper_with_timeout function (used otherwise).

For the escript only the timeout command line argument can be used.


Please be aware that Markdown is not a secure format. It produces HTML from Markdown and HTML. It is your job to sanitize and or filter the output of Earmark.as_html if you cannot trust the input and are to serve the produced HTML on the Web.



Given a markdown document (as either a list of lines or a string containing newlines), returns a tuple containing either {:ok, html_doc, error_messages}, or {:error, html_doc, error_messages} Where html_doc is an HTML representation of the markdown document and error_messages is a list of tuples with the following elements

  • severity e.g. :error, :warning or :deprecation
  • line number in input where the error occurred
  • description of the error

options can be an %Earmark.Options{} structure, or can be passed in as a Keyword argument (with legal keys for %Earmark.Options

  • renderer: ModuleName

    The module used to render the final document. Defaults to Earmark.HtmlRenderer

  • gfm: boolean

    True by default. Turns on the supported Github Flavored Markdown extensions

  • breaks: boolean

    Only applicable if gfm is enabled. Makes all line breaks significant (so every line in the input is a new line in the output.

  • code_class_prefix: binary

    Code blocks will be rendered with prefixed class names, which might be necessary for usage with 3rd party libraries.

      Earmark.as_html("```elixir\nCode\n```", code_class_prefix: "my_prefix_")
      {:ok, "<pre><code class=\"elixir my_prefix_elixir\">Code\```</code></pre>\n", []}
  • smartypants: boolean

    Turns on smartypants processing, so quotes become curly, two or four hyphens become en and em dashes, and so on. True by default.

    So, to format the document in original and disable smartypants, you'd call

      alias Earmark.Options
      Earmark.as_html(original, %Options{smartypants: false})
  • pure_links: boolean

    Pure links of the form ~r{\bhttps?://\S+\b} are rendered as links from now on. However, by setting the pure_links option to false this can be disabled and pre 1.4 behavior can be used.


EXPERIMENTAL, but well tested, just expect API changes in the 1.4 branch

  iex(9)> markdown = "My `code` is **best**"
  ...(9)> {:ok, ast, []} = Earmark.as_ast(markdown)
  ...(9)> ast
  [{"p", [], ["My ", {"code", [{"class", "inline"}], ["code"]}, " is ", {"strong", [], ["best"]}]}] 

Options are passes like to as_html, some do not have an effect though (e.g. smartypants) as formatting and escaping is not done for the AST.

  iex(10)> markdown = "```elixir\nIO.puts 42\n```"
  ...(10)> {:ok, ast, []} = Earmark.as_ast(markdown, code_class_prefix: "lang-")
  ...(10)> ast
  [{"pre", [], [{"code", [{"class", "elixir lang-elixir"}], ["IO.puts 42"]}]}]


The AST is exposed in the spirit of Floki's there might be some subtle WS differences and we chose to always have tripples, even for comments. We also do return a list for a single node

  Floki.parse("<!-- comment -->")           
  {:comment, " comment "}

  Earmark.as_ast("<!-- comment -->")
  {:ok, [{:comment, [], [" comment "]}], []}

Therefore as_ast is of the following type

  @typep tag  :: String.t | :comment
  @typep att  :: {String.t, String.t}
  @typep atts :: list(att)
  @typep node :: String.t | {tag, atts, ast}

  @type  ast  :: list(node)


EXPERIMENTAL But well tested, just expect API changes in the 1.4 branch Takes an ast, and optional options (I love this pun), which can be a map or keyword list of which the following keys will be used:

  • smartypants: boolean

  • initial_indent: number

  • indent: number

    iex(1)> transform({"p", [], [{"em", [], "help"}, "me"]})
    "<p>\n  <em>\n    help\n  </em>\n  me\n</p>\n"

Right now only transformation to HTML is supported.

The transform is also agnostic to any annotation map that is added to the AST.

Only the :meta key is reserved and by passing annotation maps with a :meta key into the AST the result might become altered or an exception might be raised, otherwise...

  iex(2)> transform({"p", [], [{"em", [], ["help"], %{inner: true}}], %{level: 1}})
  "<p>\n  <em>\n    help\n  </em>\n</p>\n"


Pull Requests are happily accepted.

Please be aware of one caveat when correcting/improving README.md.

The README.md is generated by the mix task readme from README.template and docstrings by means of %moduledoc or %functiondoc directives.

Please identify the origin of the generated text you want to correct and then apply your changes there.

Then issue the mix task readme, this is important to have a correctly updated README.md after the merge of your PR.

Thank you all who have already helped with Earmark, your names are duely noted in CHANGELOG.md.


Copyright © 2014,5,6,7,8 Dave Thomas, The Pragmatic Programmers @/+pragdave, dave@pragprog.com


Same as Elixir, which is Apache License v2.0. Please refer to LICENSE for details.

SPDX-License-Identifier: Apache-2.0