ndreynolds / earmark

Markdown parser for Elixir

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Earmark—A Pure Elixir Markdown Processor

Build Status Hex.pm Hex.pm Hex.pm

Table Of Contents

Dependency

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

Usage

API

Earmark.as_html

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

Earmark.as_html!

  html_doc = Earmark.as_html!(markdown, options)

Any error messages are printed to stderr.

Options:

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

Command line

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

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

Use

$ ./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-"})

Supports

Standard Gruber markdown.

Extensions

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~~"]
      "<p><del>hello</del></p>\n"
    
  • 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 bars. It isn't clear what the expectation 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"

Limitations

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

      <div>
      <div>
      hello
      </div>
      </div>
    

    will work. However. the following won't

      <div>
      hello</div>
    
  • 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.

    Thusly

      mypara
      <hr>
    

    Becomes

      <p>mypara</p>
      <hr>
    

    While

      mypara
       <hr>
    

    will be transformed into

      <p>mypara
       <hr></p>
    

Integration

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-" ...

Timeouts

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.

Security

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.

Details

Given a markdown document (as either a list of lines or a string containing newlines), returns a tuple containing either {:ok, html_doc}, 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 or :warning
  • line number in input where the error occurred
  • description of the error

The options are a %Earmark.Options{} structure:

  • 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.

  • 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})

Plugins

Plugins are modules that implement a render function. Right now that is as_html.

Plugin API

Plugin Registration

When invoking Earmark.as_html(some_md, options) we can register plugins inside the plugins map, where each plugin is a value pointed to by the prefix triggering it.

Prefixes are appended to "$$" and lines starting by that string will be rendered by the registered plugin.

%Earmark.Options{plugins: %{"" => CommentPlugin}} would trigger the CommentPlugin for each block of lines prefixed by $$, while %Earmark.Options{plugins: %{"cp" => CommentPlugin}} would do the same for blocks of lines prefixed by $$cp.

Please see the documentation of Plugin.define for a convenience function that helps creating the necessary Earmark.Options structs for the usage of plugins.

Plugin Invocation

as_html (or other render functions in the future) is invoked with a list of pairs containing the text and line number of the lines in the block. As an example, if our plugin was registered with the default prefix of "" and the markdown to be converted was:

  # Plugin output ahead
  $$ line one
  $$
  $$ line two

as_html would be invoked as follows:

  as_html([{"line one", 2}, {"", 3}, {"line two", 4})

Plugin Output

Earmark's render function will invoke the plugin's render function as explained above. It can then integrate the return value of the function into the generated rendering output if it complies to the following criteria.

  1. It returns a string
  2. It returns a list of strings
  3. It returns a pair of lists containing a list of strings and a list of error/warning tuples. Where the tuples are of the form {:error | :warning, line_number, descriptive_text}

A complete example

  iex> defmodule MyPlug do
  ...>   def as_html(lines) do
  ...>     # to demonstrate the three possible return values
  ...>     case render(lines) do
  ...>       {[line], []} -> line
  ...>       {lines, []} -> lines
  ...>       tuple       -> tuple
  ...>     end
  ...>   end
  ...>
  ...>   defp render(lines) do
  ...>     Enum.map(lines, &render_line/1) |> Enum.split_with(&ok?/1)
  ...>   end
  ...>
  ...>   defp render_line({"", _}), do: "<hr/>"
  ...>   defp render_line({"line one", _}), do: "<p>first line</p>\n"
  ...>   defp render_line({line, lnb}), do: {:error, lnb, line}
  ...>
  ...>   defp ok?({_, _, _}), do: false
  ...>   defp ok?(_), do: true
  ...> end
  ...>
  ...> lines = [
  ...>   "# Plugin Ahead",
  ...>   "$$ line one",
  ...>   "$$",
  ...>   "$$ line two",
  ...> ]
  ...> Earmark.as_html(lines, Earmark.Plugin.define(MyPlug))
  {:error, "<h1>Plugin Ahead</h1>\n<p>first line</p>\n<hr/>", [{ :error, 4, "line two"}]}

Plugins, reusing Earmark

As long as you avoid endless recursion there is absolutely no problem to call Earmark.as_html in your plugin, consider the following example in which the plugin will parse markdown and render html verbatim (which is stupid, that is what Earmark already does for you, but just to demonstrate the possibilities):

  iex> defmodule Again do
  ...>   def as_html(lines) do
  ...>     text_lines = Enum.map(lines, fn {str, _} -> str end)
  ...>     {_, html, errors} = Earmark.as_html(text_lines)
  ...>     { Enum.join([html | text_lines]), errors }
  ...>   end
  ...> end
  ...>  lines = [
  ...>    "$$a * one",
  ...>    "$$a * two",
  ...>  ]
  ...>  Earmark.as_html(lines, Earmark.Plugin.define({Again, "a"}))
  {:ok, "<ul>\n<li>one\n</li>\n<li>two\n</li>\n</ul>\n* one* two", []}

Contributing

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.

Author

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

LICENSE

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

SPDX-License-Identifier: Apache-2.0

About

Markdown parser for Elixir

License:Other


Languages

Language:Elixir 84.8%Language:HTML 14.1%Language:Erlang 1.1%