MDEx
Fast and Extensible Markdown for Elixir.
Features
- Support formats:
- Markdown (CommonMark)
- HTML
- JSON
- XML
- Floki-like Document AST
- Req-like Pipeline API
- Compliant with the CommonMark spec
- Additional features:
- GitHub Flavored Markdown
- Discord and GitLab features
- Wiki-style links
- Emoji shortcodes
- Syntax highlighting for code blocks
- HTML sanitization
- ~MD sigil for Markdown, HTML, JSON, and XML
Foundation
The library is built on top of:
- comrak - a fast Rust port of GitHub's CommonMark parser
- ammonia for HTML sanitization
- autumnus for syntax highlighting
Installation
Add :mdex dependency:
def deps do
[
{:mdex, "~> 0.7"}
]
end
Usage
Mix.install([{:mdex, "~> 0.7"}])
iex> MDEx.to_html!("# Hello")
"<h1>Hello</h1>"
iex> MDEx.to_html!("# Hello :smile:", extension: [shortcodes: true])
"<h1>Hello π</h1>"
Plugins
- mdex_mermaid - Render Mermaid diagrams in code blocks
Req-like Pipeline
MDEx.Pipe provides a high-level API to manipulate a Markdown document and build plugins that can be attached to a pipeline:
document = ~s"""
# Super Diagram
```mermaid
graph TD
A[Enter Chart Definition] --> B(Preview)
B --> C{decide}
C --> D[Keep]
C --> E[Edit Definition]
E --> B
D --> F[Save Image and Code]
F --> B
```
"""
MDEx.new()
|> MDExMermaid.attach(mermaid_version: "11")
|> MDEx.to_html(document: document)
~MD Sigil
Convert and generate AST (MDEx.Document), Markdown (CommonMark), HTML, JSON, and XML formats.
iex> import MDEx.Sigil
iex> ~MD|# Hello from `~MD` sigil|
%MDEx.Document{
nodes: [
%MDEx.Heading{
nodes: [
%MDEx.Text{literal: "Hello from "},
%MDEx.Code{num_backticks: 1, literal: "~MD"},
%MDEx.Text{literal: " sigil"}
],
level: 1,
setext: false
}
]
}
iex> import MDEx.Sigil
iex> ~MD|`~MD` also converts to HTML format|HTML
"<p><code>~MD</code> also converts to HTML format</p>"
iex> import MDEx.Sigil
iex> ~MD|and to XML as well|XML
"<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<!DOCTYPE document SYSTEM \"CommonMark.dtd\">\n<document xmlns=\"http://commonmark.org/xml/1.0\">\n <paragraph>\n <text xml:space=\"preserve\">and to XML as well</text>\n </paragraph>\n</document>"
~MD also accepts an assigns map to pass variables to the document:
iex> import MDEx.Sigil
iex> assigns = %{lang: "Elixir"}
iex> ~MD|Running <%= @lang %>|HTML
"<p>Running Elixir</p>"
See more info at https://hexdocs.pm/mdex/MDEx.Sigil.html
Code Block Decorators
Code block decorators allow you to customize the appearance and behavior of individual code blocks by adding special attributes to the info string (the part after the opening backticks).
Prerequisites
To use code block decorators, you must enable both :render options:
render: [
github_pre_lang: true,
full_info_string: true
]
Available Decorators
| Decorator | Description | Supported Formatters | Example |
|---|---|---|---|
theme |
Override the syntax highlighting theme | All | theme=github_dark |
pre_class |
Add custom CSS classes to <pre> element |
All | pre_class="my-class" |
highlight_lines |
Highlight single and/or range of lines | All | highlight_lines="1,3-5" |
highlight_lines_style |
Custom inline styles for highlighted lines | HTML inline only | highlight_lines_style="background: yellow" |
highlight_lines_class |
Custom CSS class for highlighted lines | All | highlight_lines_class="emphasis" |
include_highlights |
Add syntax token names as data attributes | All | include_highlights |
Examples
Following examples assume render: [github_pre_lang: true, full_info_string: true] is set.
Override Theme
Change the syntax highlighting theme for a specific code block:
```elixir theme=github_dark
def hello do
"Hello, world!"
end
```
Output: <pre class="athl" style="color: #c9d1d9; background-color: #0d1117;">...
Add Custom CSS Classes
Add your own CSS classes to the <pre> element:
```javascript pre_class="code-example interactive"
console.log("Hello!");
```
Output: <pre class="athl code-example interactive">...
Highlight Specific Lines
Highlight individual lines or ranges (inclusive):
# Highlight lines 1, 4, 5, and 6
```python highlight_lines="1,4-6"
import math
def calculate(x):
result = x * 2
# return calculated result
return math.sqrt(x)
```
With :html_inline formatter, lines get styles from the theme's highlight color, for eg:
<span style="background-color: #dae9f9;" data-line="1">...
With :html_linked formatter, the class highlighted is added to the highlighted lines, for eg:
<span class="line highlighted" data-line="1">...
Custom Highlight Styling
Use either highlight_lines_style or highlight_lines_class to customize the appearance of highlighted lines:
```ruby highlight_lines="2" highlight_lines_style="background: #ffeb3b; font-weight: bold;"
class User
def initialize(name)
@name = name
end
end
```
Include Syntax Token Information
Add syntax token names in data-highlight attributes, useful for debugging or custom styling:
```rust include_highlights
let x: i32 = 42;
```
Output: <span data-highlight="keyword">let</span>
Combine Multiple Decorators
Use multiple decorators together:
```typescript theme=github_light pre_class="example" highlight_lines="2-3" include_highlights
interface User {
name: string; // highlighted
email: string; // highlighted
age?: number;
}
```
Important Notes
-
Formatter Support: Not all decorators work with all formatters:
highlight_lines_styleonly works with:html_inlineformatterthemeonly works with:html_inlineformatter
-
CSS Classes: The
athlclass is always added to<pre>elements when using syntax highlighting -
Line Numbers: The
data-lineattribute is added to each line for reference -
Order: It's expected the first word of the code fence info string to be the language name, followed by decorators.
- For example,
elixir theme=github_darkis valid, buttheme=github_dark elixiris not.
- For example,
-
Performance: Decorators are processed at render time, so using many decorators may impact performance
Safety
For security reasons, every piece of raw HTML is omitted from the output by default:
iex> MDEx.to_html!("<h1>Hello</h1>")
"<!-- raw HTML omitted -->"
That's not very useful for most cases, but you have a few options:
Escape
The most basic is render raw HTML but escape it:
iex> MDEx.to_html!("<h1>Hello</h1>", render: [escape: true])
"<h1>Hello</h1>"
Sanitize
But if the input is provided by external sources, it might be a good idea to sanitize it:
iex> MDEx.to_html!("<a href=https://elixir-lang.org>Elixir</a>", render: [unsafe: true], sanitize: MDEx.default_sanitize_options())
"<p><a href=\"https://elixir-lang.org\" rel=\"noopener noreferrer\">Elixir</a></p>"
Note that you must pass the unsafe: true option to first generate the raw HTML in order to sanitize it.
It does clean HTML with a conservative set of defaults that works for most cases, but you can overwrite those rules for further customization.
For example, let's modify the link rel attribute
to add "nofollow" into the rel attribute:
iex> MDEx.to_html!("<a href=https://someexternallink.com>External</a>", render: [unsafe: true], sanitize: [link_rel: "nofollow noopener noreferrer"])
"<p><a href=\"https://someexternallink.com\" rel=\"nofollow noopener noreferrer\">External</a></p>"
In this case the default rule set is still applied but the link_rel rule is overwritten.
Unsafe
If those rules are too strict and you really trust the input, or you really need to render raw HTML, then you can just render it directly without escaping nor sanitizing:
iex> MDEx.to_html!("<script>alert('hello')</script>", render: [unsafe: true])
"<script>alert('hello')</script>"
Parsing
Converts Markdown to an AST data structure that can be inspected and manipulated to change the content of the document programmatically.
The data structure format is inspired on Floki (with :attributes_as_maps = true) so we can keep similar APIs and keep the same mental model when
working with these documents, either Markdown or HTML, where each node is represented as a struct holding the node name as the struct name and its attributes and children, for eg:
%MDEx.Heading{
level: 1
nodes: [...],
}
The parent node that represents the root of the document is the MDEx.Document struct, where you can find more more information about the AST and what operations are available.
The complete list of nodes is listed in the documentation, section Document Nodes.
Formatting
Formatting is the process of converting from one format to another, for example from AST or Markdown to HTML. Formatting to XML and to Markdown is also supported.
You can use MDEx.parse_document/2 to generate an AST or any of the to_* functions
to convert to Markdown (CommonMark), HTML, JSON, or XML.
Examples
GitHub Flavored Markdown with emojis
MDEx.to_html!(~S"""
# GitHub Flavored Markdown :rocket:
- [x] Task A
- [x] Task B
- [ ] Task C
| Feature | Status |
| ------- | ------ |
| Fast | :white_check_mark: |
| GFM | :white_check_mark: |
Check out the spec at https://github.github.com/gfm/
""",
extension: [
strikethrough: true,
tagfilter: true,
table: true,
autolink: true,
tasklist: true,
footnotes: true,
shortcodes: true,
],
parse: [
smart: true,
relaxed_tasklist_matching: true,
relaxed_autolinks: true
],
render: [
github_pre_lang: true,
unsafe: true,
]) |> IO.puts()
"""
<p>GitHub Flavored Markdown π</p>
<ul>
<li><input type="checkbox" checked="" disabled="" /> Task A</li>
<li><input type="checkbox" checked="" disabled="" /> Task B</li>
<li><input type="checkbox" disabled="" /> Task C</li>
</ul>
<table>
<thead>
<tr>
<th>Feature</th>
<th>Status</th>
</tr>
</thead>
<tbody>
<tr>
<td>Fast</td>
<td>β
</td>
</tr>
<tr>
<td>GFM</td>
<td>β
</td>
</tr>
</tbody>
</table>
<p>Check out the spec at <a href="https://github.github.com/gfm/">https://github.github.com/gfm/</a></p>
"""
Code Syntax Highlighting
MDEx.to_html!(~S"""
```elixir
String.upcase("elixir")
```
""",
syntax_highlight: [
formatter: {:html_inline, theme: "catppuccin_latte"}
]
) |> IO.puts()
"""
<pre class=\"autumn highlight\" style=\"background-color: #282C34; color: #ABB2BF;\">
<code class=\"language-elixir\" translate=\"no\">
<span class=\"namespace\" style=\"color: #61AFEF;\">String</span><span class=\"operator\" style=\"color: #C678DD;\">.</span><span class=\"function\" style=\"color: #61AFEF;\">upcase</span><span class=\"\" style=\"color: #ABB2BF;\">(</span><span class=\"string\" style=\"color: #98C379;\">"elixir"</span><span class=\"\" style=\"color: #ABB2BF;\">)</span>
</code>
</pre>
"""
Pre-compilation
Pre-compiled binaries are available for the following targets, so you don't need to have Rust installed to compile and use this library:
aarch64-apple-darwinaarch64-unknown-linux-gnuaarch64-unknown-linux-muslarm-unknown-linux-gnueabihfriscv64gc-unknown-linux-gnux86_64-apple-darwinx86_64-pc-windows-gnux86_64-pc-windows-msvcx86_64-unknown-freebsdx86_64-unknown-linux-gnux86_64-unknown-linux-musl
Note: The pre-compiled binaries for Linux are compiled using Ubuntu 22 on libc 2.35, which requires minimum Ubuntu 22, Debian Bookworm or a system with a compatible libc version. For older Linux systems, you'll need to compile manually.
Compile manually
But in case you need or want to compile it yourself, you can do the following:
-
Install a C compiler or build packages
It depends on your OS, for example in Ubuntu you can install the build-essential package.
- Run:
export MDEX_BUILD=1
mix deps.get
mix compile
Legacy CPUs
Modern CPU features are enabled by default but if your environment has an older CPU,
you can use legacy artifacts by adding the following configuration to your config.exs:
config :mdex, use_legacy_artifacts: true
Used By
Are you using MDEx and want to list your project here? Please send a PR!
Motivation
MDEx was born out of the necessity of parsing CommonMark files, to parse hundreds of files quickly, and to be easily extensible by consumers of the library.
- earmark is extensible but can't parse all kinds of documents and is slow to convert hundreds of markdowns.
- md is very extensible but the doc says "If one needs to perfectly parse the common markdown, Md is probably not the correct choice" and CommonMark was a requirement to parse many existing files.
- markdown is not precompiled and has not received updates in a while.
- cmark is a fast CommonMark parser but it requires compiling the C library, is hard to extend, and was archived on Apr 2024.
Comparison
| Feature | MDEx | Earmark | md | cmark |
|---|---|---|---|---|
| Active | β | β | β | β |
| Pure Elixir | β | β | β | β |
| Extensible | β | β | β | β |
| Syntax Highlighting | β | β | β | β |
| Code Block Decorators | β | β | β | β |
| AST | β | β | β | β |
| AST to Markdown | β | β οΈΒ² | β | β |
| To HTML | β | β | β | β |
| To JSON | β | β | β | β |
| To XML | β | β | β | β |
| To Manpage | β | β | β | β |
| To LaTeX | β | β | β | β |
| Emoji | β | β | β | β |
| GFMΒ³ | β | β | β | β |
| GitLabβ΄ | β οΈΒΉ | β | β | β |
| Discordβ΅ | β οΈΒΉ | β | β | β |
- Partial support
- Possible with earmark_reversal
- GitHub Flavored Markdown
- GitLab Flavored Markdown
- Discord Flavored Markdown
Benchmark
A simple script is available to compare existing libs:
Name ips average deviation median 99th %
cmark 7.17 K 0.139 ms Β±4.20% 0.138 ms 0.165 ms
mdex 2.71 K 0.37 ms Β±7.95% 0.36 ms 0.45 ms
md 0.196 K 5.11 ms Β±2.51% 5.08 ms 5.55 ms
earmark 0.0372 K 26.91 ms Β±2.09% 26.77 ms 30.25 ms
Comparison:
cmark 7.17 K
mdex 2.71 K - 2.65x slower +0.23 ms
md 0.196 K - 36.69x slower +4.98 ms
earmark 0.0372 K - 193.04x slower +26.77 ms
The most performance gain is using the ~MD sigil to compile the Markdown instead of parsing it at runtime,
prefer using it when possible:
Comparison:
mdex_sigil_MD 176948.46 K
cmark 31.47 K - 5622.76x slower +31.77 ΞΌs
mdex_to_html/1 7.32 K - 24184.36x slower +136.67 ΞΌs
md 2.05 K - 86176.93x slower +487.01 ΞΌs
earmark 0.21 K - 855844.67x slower +4836.68 ΞΌs
To finish, a friendly reminder that all libs have their own strengths and trade-offs so use the one that better suits your needs.
Acknowledgements
- comrak crate for all the heavy work on parsing Markdown and rendering HTML
- Floki for the AST
- Req for the pipeline API
- Logo based on markdown-mark