Skip to content

SuperFences

Overview

SuperFences provides a number of features:

  1. Allowing the nesting of fences under blockquotes, lists, or other block elements (see Limitations for more info).
  2. Ability to specify custom fences to provide features like flowcharts, sequence diagrams, or other custom blocks.
  3. Allow disabling of indented code blocks in favor of only using the fenced variant (off by default).
  4. Experimental feature that preserves tabs within a code block instead of converting them to spaces which is Python Markdown's default behavior.

Reminder

Remember to read the Usage Notes for information that may be relevant when using this extension!

The SuperFences extension can be included in Python Markdown by using the following:

import markdown
md = markdown.Markdown(extensions=['pymdownx.superfences'])

Nested Fence Format

  1. Start and end fence boundaries are specified with either 3 or more backticks or tildes.

  2. Start and end fence boundaries must both use matching symbols (backticks or tildes) and must be of the same number of symbols. If start is 3 backticks, the fence will end with 3 backticks.

  3. Start and end fence boundaries must be aligned to the same indentation level.

  4. Content between fences must be indented at least the same amount as the start and end boundaries. Empty lines are exempted.

  5. If you are using a fenced block inside a blockquote, at the very least, the first line of the fenced block needs to have the appropriate number of > characters signifying the quote depth.

    > ```
      a fenced block
      ```
    
  6. Too many blank lines will cause a blockquote to terminate, so remember to use > markers accordingly if not marking every line.

    > ```
      a fenced block
    
    > with blank lines
      ```
    
  7. If using a fenced block as the first line of a list, you will have to leave the first line blank, but remember that the list marker must be immediately followed by at least one space. To avoid accidentally deleting the space and to make your intentions clear, you might want to also add an explicit unicode space ( ) as shown here:

    -  
        ```
        a fenced block
        ```
    
    Definition
    :  
        ```
        a fenced block
        ```
    
  8. Fenced blocks should be separated from other blocks by an empty line.

    Paragraph.
    
    ```
    a fenced block
    ```
    
    Another paragraph.
    

Injecting Classes, IDs, and Attributes

You can use the brace format to specify classes and IDs where IDs use the form #id and classes use the form .class. Arbitrary attributes in the form key="value" can be inserted as well if the attr_list extension is enabled, but when using Pygments, only key="value" attributes that start with the data- prefix will be recognized, all others will be treated as options for Pygments and will be rejected if not valid.

Injecting Classes
```python {.extra-class #id linenums="1"}
import hello_world
```
<div id="id" class="extra-class highlight"><table class="highlighttable"><tr><td class="linenos"><div class="linenodiv"><pre><span></span><span class="normal">1</span></pre></div></td><td class="code"><div><pre><span></span><code><span class="kn">import</span> <span class="nn">hello_world</span>
</code></pre></div></td></tr></table></div>

Alternatively, if a language is not provided, the first class is assumed to specify the language.

Injecting Classes
```{.python .extra-class #id linenums="1"}
import hello_world
```
<div id="id" class="extra-class highlight"><table class="highlighttable"><tr><td class="linenos"><div class="linenodiv"><pre><span></span><span class="normal">1</span></pre></div></td><td class="code"><div><pre><span></span><code><span class="kn">import</span> <span class="nn">hello_world</span>
</code></pre></div></td></tr></table></div>

When generating additional classes on a JavaScript style code block (non-Pygments code blocks), classes are injected in the code block.

Non-Pygments Injecting Classes
```python {.extra-class #id linenums="1"}
import hello_world
```
<pre class="highlight"><code id="id" class="language-python extra-class" linenums="1">import hello_world</code></pre>

When using Pygments, all extra classes and attributes will be attached to the Pygments' wrapper <div>.

When using a built in custom formatter, all classes and IDs are injected on to the first element <div> or <pre>. This preserves previous behavior, but you can write your own and inject them in the way that suites your needs.

New 9.0

Added support for #id in Pygments brace headers. Also added support for arbitrary data- attributes if the attr_list extension is enabled.

Preserve Tabs

Python Markdown has an approach where it normalizes whitespace. This means \r\n is converted to \n and \t is converted to spaces. In 99% of Markdown, this is never really an issue, but with code blocks it can be. Tabs can sometimes be very useful for aligning certain kinds of data, especially when dealing with characters of varying width.

Tabs in Content
```
============================================================
T   Tp  Sp  D   Dp  S   D7  T
------------------------------------------------------------
A   F#m Bm  E   C#m D   E7  A
A#  Gm  Cm  F   Dm  D#  F7  A#
B♭  Gm  Cm  F   Dm  E♭m F7  B♭
```
============================================================
T	Tp	Sp	D	Dp	S	D7	T
------------------------------------------------------------
A	F#m	Bm	E	C#m	D	E7	A
A#	Gm	Cm	F	Dm	D#	F7	A#
B♭	Gm	Cm	F	Dm	E♭m	F7	B♭

If you have a scenario where preserving tabs is a requirement, you can use SuperFences preserve_tabs option to prevent converting tabs to spaces inside fenced code blocks. This only applies to fenced code blocks. Indented code blocks and inline code blocks will still be treated with Python Markdown's default behavior.

This feature is experimental and actually applies the fences before whitespace normalization and bypasses the normalization for the code content.

Code Highlighting

Assuming Pygments is installed, code highlighting will be handled by Pygments by default. If Pygments is not installed, or disabled, code blocks will be created using HTML5 style tags for a JavaScript syntax highlighter: <pre class="highlight"><code class="language-mylanguage"></code></pre>. If you disable highlight_code, specified languages will be ignored, and the content will be wrapped in a simple pre and code tags with no classes.

Highlighting can be further controlled via the pymdownx.highlight extension. You must include pymdownx.highlight in the extensions list in order to be able to configure it.

When using fenced code blocks, you can specify a specific syntax language to highlight with by specifying the language name directly after the opening tokens (either ``` or ~~~). Whether using Pygments or some other JavaScript highlighter, the syntax is the same, but please consult your highlighter's documentation for recognized language syntax specifiers.

Highlighting
```python
import foo.bar
```
import foo.bar

Showing Line Numbers

Line numbers are provided via Pygments and can either be shown per code block or globally for all. To show globally via pymdownx.highlight, you must set linenums to True in the respective extension.

To set line numbers per code block, you can specify a special setting directly after the opening tokens (and language if present). Simply specify the starting line number with option linenums="1". The setting is followed by the equal sign and the value must be quoted. Valid line numbers are n > 0. If linenums is enabled globally, this will just control the starting line shown in the block.

Line Numbers
```python {linenums="1"}
import foo.bar
```
import foo.bar

And, if we wanted to start with a different starting line number, we would just specify something other than 1.

Custom Line Number Start
```python {linenums="2"}
import foo.bar
```
import foo.bar

Pygments also has a few additional options in regards to line numbers. One is "line step" which, if set to a number n > 1, will print only every nth line number. The other option is a setting that can mark line numbers as "special" with a span and class special. If the special parameter is set to a number n > 0, every nth line number is given the CSS class special.

So to set showing only every other line number, we could do the following. Line options are separated by a space, and "line step" is always the second option, so you must specify line start before line step.

Nth Line
```python {linenums="1 2"}
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz
```
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz

To set every other line as special, you must set the third linenums option (specify line start and step before it). Special must be a value of n > 0. Additionally, you can set this globally with linenums_special in the Highlight extension.

Special Lines
``` {linenums="1 1 2"}
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz
```
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz

For JavaScript libraries, a class of linenums is written to the block. This may or may not be leveraged by your chosen highlighter. It is uncertain at this time whether line number support for JavaScript highlighters will be enhanced beyond this.w what to inject.

Highlighting Lines

Via Pygments, certain lines can be specified for highlighting. This is done by specifying a special setting directly after the opening tokens (and language if present). The setting is named hl_lines and the value should be the targeted line numbers separated by spaces.

Highlight Lines
```python {hl_lines="1 3"}
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz
```
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz

Line numbers are always referenced starting at 1 ignoring what the line number is labeled as when showing line numbers.

Highlight Lines with Line Numbers
```python {hl_lines="1 3" linenums="2"}
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz
```
"""Some file."""
import foo.bar
import boo.baz
import foo.bar.baz

If you'd like to do a range of lines, you can use the notation x-y where x is the starting line and y is the ending line. You can do multiple ranges and even mix them with non ranges.

Highlight Ranges
```python {hl_lines="1-2 5 7-8"}
import foo
import boo.baz
import foo.bar.baz

class Foo:
   def __init__(self):
       self.foo = None
       self.bar = None
       self.baz = None
```
import foo
import boo.baz
import foo.bar.baz

class Foo:
   def __init__(self):
       self.foo = None
       self.bar = None
       self.baz = None

Code Block Title Headers

New 9.0

Title headers are new in version 9.0.

When Pygments is enabled, a header with a title can be applied with the title option. This essentially controls the Pygments' filename option under the hood. It made more sense to use the term title as people can really set any arbitrary title, not just filenames.

Pygments will simply output the HTML below. The user is responsible for provided CSS to style the header. Pygments uses the class name of filename as that is the feature being used under the hood.

<div class="highlight">
<span class="filename">Some Title</span>
<pre><code></code></pre>
</div>

If using line numbers and the linenums_style set to table (the default), the title will be inserted in <th> element at the start of the table set to span both the line number column and the line column.

<div class="highlight">
<table class="highlighttable">
<tr>
  <th colspan="2" class="filename"><span class="filename">My title</span></th>
</tr>
<tr>
  <td class="linenos"><div class="linenodiv"><pre><span></span><span class="normal">1</span></pre></div></td>
  <td class="code"><div class="highlight"><pre><code></code></pre></div></td>
</tr>
</table>
</div>
Adding Titles
```python {title="My Cool Header"}
import foo.bar
import boo.baz
import foo.bar.baz
```
My Cool Header
import foo.bar
import boo.baz
import foo.bar.baz

If the auto_title option is enabled in the Highlight extension, the title will be auto populated with the name of the lexer used to highlight the code (unless title is manually specified). In the example below, no title is specified, but the title is extracted from the Python lexer.

Automatic Titles
```python
import foo.bar
import boo.baz
import foo.bar.baz
```
Python
import foo.bar
import boo.baz
import foo.bar.baz

There may be some cases where a Lexer returns a result that is undesired. For example, when a user specifies the pycon lexer, the title that is returned is quite verbose.

Undesired Automatic Title
```pycon
>>> 3 + 3
6
```
Python Console Session
>>> 3 + 3
6

In a case like above, it may be desired to simply use the title Python. We can configure the Highlight extension to override any auto returned title. Simply create mapping via the auto_title_map specifying the title you wish to override as the key, and the desired title as the value.

extension_configs = {
    "pymdownx.highlight": {
        "auto_title": True,
        "auto_title_map": {
            "Python Console Session": "Python"
        }
    }
}

Now whenever we use pycon, we get "Python" instead of "Python Console Session".

Python
>>> 3 + 3
6

Pygments Line Anchors and Spans

New 9.0

The various line wrapping options are new to version 9.0.

Pygments offers a couple of options that will wrap lines, line numbers even create anchor links for line numbers. SuperFences, when using Pygments, exposes these options under similar names.

In this example, we will wrap each line of code in a span. Pygments will create an ID for each span using the prefix that we provide. We simply set the global config option line_spans and specify the desired "prefix" (_codeline), and then every line will be wrapped in a span with the ID prefix-x-y where prefix is the user specified prefix, x is a unique number for the code block, and y is the line number. After that, it is up to the user to do as they with to target the ID with either JavaScript and/or CSS.

extension_configs = {
    "pymdownx.highlight": {
        'linenums_style': 'inline',
        'line_spans': '__codeline'
    }
}
```python {linenums="1 1" }
import foo
```
<div class="highlight">
<pre>
<span></span>
<code>
<span id="__codeline-0-1"><span class="linenos">1</span><span class="kn">import</span> <span class="nn">foo</span>
</span>
</code>
</pre>
</div>

We can also wrap line numbers with a link and inject anchors so you can click line numbers and be taken to said line. To do this, anchor_linenums must be enabled and then a prefix should be provided via line_anchors, just like line_spans, line_anchors will produce an ID in the form prefix-x-y where x is a unique number for the code block and y is the line number. If you wish to not have the line numbers clickable, and just have the anchors inserted, you can omit enabling anchor_linenums.

extension_configs = {
    "pymdownx.highlight": {
        'linenums_style': 'inline',
        'line_spans': '__codeline',
        'line_anchors': '__codelineno',
        'anchor_linenums': True
    }
}
```python {linenums="1 1" }
import foo
```
<div class="highlight">
<pre>
<span></span>
<code>
<span id="__codeline-0-1"><a id="__codelineno-0-1" name="__codelineno-0-1"></a><a href="#__codelineno-0-1"><span class="linenos">1</span></a><span class="kn">import</span> <span class="nn">foo</span>
</span>
</code>
</pre>
</div>

Custom Fences

SuperFences allows defining custom fences for special purposes. For instance, we could create special fences for diagrams that we could later run Mermaid on.

Flow Chart
```diagram
graph TD
    A[Hard] -->|Text| B(Round)
    B --> C{Decision}
    C -->|One| D[Result 1]
    C -->|Two| E[Result 2]
```
graph TD
    A[Hard] -->|Text| B(Round)
    B --> C{Decision}
    C -->|One| D[Result 1]
    C -->|Two| E[Result 2]

This would allow us access to all the diagrams Mermaid offers. Some Mermaid diagrams are less practical to use as they don't scale well, but many work great.

In the above example, we have set up a custom fence called diagram which allows us to process special charts with Mermaid. In this case, our special fence preserves the content and sends them to a an element so that Mermaid can then find them and convert them when the document is loaded. To learn more see UML Diagram Example.

Custom fences are created via the custom_fences option. custom_fences takes an array of dictionaries where each dictionary defines a custom fence. The dictionaries require the following keys:

Keys Description
name The language name that is specified when using the fence in Markdown. If given *, it will override the base fence logic, the default for all fence names not handled by other custom fences.
class The class name assigned to the HTML element when converting from Markdown to HTML.
format A function that formats the HTML output. The function should return a string as HTML.
validator An optional parameter that is used to provide a function to validate custom fence parameters.

Logging

When a custom fence fails, the error will be swallowed up and the error will be handled gracefully. If logging is desired, custom logging should be added into the custom function. SuperFences will not provide any.

New in 7.0

Starting in 7.0, you can override the base fence logic (the syntax highlighter) by specifying the custom fence with a name of *. This means that if a fence does not match any other custom fences, the default, fallback fence would be handled by your custom * fence. This can be useful for tailoring a fence output with custom parameters for a specific, favorite JavaScript highlighter.

Formatters

SuperFences provides two format functions by default, but you can always write your own:

Format Function Description
superfences.fence_code_format Places the HTML escaped content of the fence under a <pre><code> block.
superfences.fence_div_format Places the HTML escaped content of the fence under a <div> block.

In general, formatters take five parameters: the source found between the fences, the specified language, the class name originally defined via the class option in the custom_fence entry, custom options, additional classes defined in brace style headers, an optional ID, any attributes defined in the brace style header, and the Markdown object (in case you want access to meta data etc.).

def custom_formatter(source, language, css_class, options, md, classes=None, id_value='', attrs=None, **kwargs):
    return string

or

def custom_formatter(source, language, css_class, options, md, **kwargs):
    return string

All formatters should return a string as HTML.

YAML Configuration Format

If you are attempting to configure these options in a YAML based configuration (like in MkDocs), please see the FAQ to see how to specify function references in YAML.

New 7.0

The addition of the parameters classes and id_value is new in 7.0. If injecting additional classes or ids via brace headers, only then will classes and id_value be passed in to preserve backwards compatibility with old custom formatters. Users, moving forward, should at the very least update their formatters with **kwargs to future proof their custom formatters in case additional parameters are added in the future.

Changes 8.0

Formatters now take the keyword parameter attrs.

Validators

The validator is used to provide a function that allows the validation of inputs. Inputs are then sorted to either attrs or options. While a formatter can treat attrs and options however they like, the intention is that key value pairs assigned to attrs are assigned directly to the fenced block's element, while options are used to conditionally control logic within the formatter. If no validator is defined, the default is used which assigns all inputs to attrs.

SuperFences will only pass attrs to a formatter if an attribute style header is used for a fenced block (```lang {attr="value"}) and the attr_list extension is enabled. Attribute are not supported in the form (```lang attr=value) and will cause the parsing of the fenced block to abort.

Custom options can be used as keys with quoted values (key="value"), or as keys with no value (key). If a key is given with no value, the value will be the key value. SuperFences will parse the options in the fence and then pass them to the validator. If the validator returns true, the options will be accepted and later passed to the formatter. attrs will only be passed to the formatter if the attr_list extension is enabled. attrs will also be ignored during Pygments highlighting.

Assuming a validator fails, the next validator/formatter defined will be tried.

For a contrived example: if we wanted to define a custom fence named test that accepts an option opt that can only be assigned a value of A, we could define the validator below. Notice that if we get an input that matches opt, but doesn't validate with the appropriate value, we abort handling the fenced block for this validator/formatter by returning False. All other inputs are assigned as attrs which will be passed to the formatter if the attr_list extension is enabled.

def custom_validator(language, inputs, options, attrs, md):
    """Custom validator."""

    okay = True
    for k, v in inputs.items():
        if k == 'opt':
            if v != "A":
                okay = False
                break
            else:
                options[k] = v
        else:
            attrs[k] = v

    return okay

Assuming validation passed, the formatter would be given the option and any attributes (though the formatter below doesn't use the attributes).

def custom_format(source, language, class_name, options, md, **kwargs):
    """Custom format."""

    return '<div class_name="%s %s", data-option="%s">%s</div>' % (language, class_name, options['opt'], html_escape(source))

This would allow us to use the following custom fence:

```test {opt="A"}
test
```

YAML Configuration Format

If you are attempting to configure these options in a YAML based configuration (like in MkDocs), please see the FAQ to see how to specify function references in YAML.

Changes 8.0

  • validator now accepts the following variables:
    • inputs: with all the parsed options/attributes (validator should not modify this structure).
    • options: a dictionary to which all valid options should be assigned to.
    • attrs: a dictionary to which all valid attributes should be assigned to.
    • md: the Markdown object.
  • If the attr_list extension is enabled and the brace style header is used, any key/value pairs that were assigned as attributes by the validator will be passed to the formatter's attrs parameter.
  • Options in the form of key= (which have no value) will are no longer be allowed. A key with no value will assume the value to be the key name. This brings consistency as options are now parsed with attr_list.
  • If a validator fails, the next validator/formatter pair will be tired.

Exception Handling

In general, if an exception occurs in either a validator or a formatter, SuperFences will gracefully ignore the validator or formatter. In the case of validators, this usually means it will try the next validator in the line. In the case of a formatter, the fenced content will just be abandoned as we are already past the point of "validating" and will not go back to validate the use of a different formatter.

Some users may want such failures to not silently go by. SuperFences exposes a special exception called SuperFencesException which, if raised, will not gracefully be handled. If SuperFencesException is raised, this will likely bubble all the way up and cause Markdown parsing to halt. SuperFencesException can be used to raise other exceptions if desired.

Here we have a contrived example. When no options are provided in this validator, we will try and access a non-existent key in the validator which will throw a KeyError. In this case, we capture the KeyError and raise it with SuperFencesException. This will display the SuperFencesException and the original KeyError when the failure occurs:

def custom_validator_except(language, inputs, options, attrs, md):
    """Custom validator."""

    okay = True
    try:
        for k in inputs.keys():
            if k != 'opt':
                okay = False
                break
        if okay:
            if inputs['opt'] != "A":
                okay = False
            else:
                options['opt'] = inputs['opt']
    except KeyError as e:
        raise SuperFencesException from e

    return okay

New 9.2

SuperFencesException added in 9.2

UML Diagram Example

Support

This example is mainly used to illustrate how custom fences work. While Mermaid is used in the example, it is the bare minimum needed to get diagrams working.

Please reference Mermaid's documents for more information on configuring features. PyMdown Extensions does not offer direct support for issues you may have in using Mermaid, feel free to use their issue tracker to report problems with their library.

We do provide some advanced notes on Mermaid if you are interested in learning how we use and configure it in our own documentation.

This example illustrates how you can use the custom_fences option to do UML diagrams with Mermaid. The settings below show us creating a new custom fence called mermaid. The special fence is set under the custom_fences option.

The mermaid fences will pass the content through the superfences.fence_div_format format function which will wrap the content in <div> blocks and attach the class mermaid to the <div> block.

extension_configs = {
    "pymdownx.superfences": {
        "custom_fences": [
            {
                'name': 'mermaid',
                'class': 'mermaid',
                'format': pymdownx.superfences.fence_div_format
            }
        ]
    }
}

This would allow a user to create fenced blocks with the mermaid specifier:

```mermaid
...content here
```

The format function we used in this example only escapes the content to be included in HTML. We will rely on the Mermaid JavaScript library to render our flowcharts/diagrams in the browser. Mermaid, on load, will automatically find the <div> elements with the mermaid class and render them. We can include Mermaid via its CDN.

<script src="https://unpkg.com/mermaid@9.4.0/dist/mermaid.min.js"></script>

Limitations

This extension suffers from some of the same quirks that the original fenced block extension suffers from. Normally Python Markdown does not parse content inside HTML tags unless they are marked with the attribute markdown='1'. But since this is run as a preprocessor, it is not aware of the HTML blocks.

SuperFences is made to work with the default extensions out of the box. It will probably not work with other extensions such as Grid Tables, since that extension allows for characters to obscure the blocks like the blockquote syntax does (though this has been designed to work with blockquotes). Ideally fenced blocks need to be handled by a block parser, but there is much work to be done on Python Markdown's internal block handlers before this is possible.

SuperFences works best when following the guidelines. If the guidelines are not followed, odd results may be encountered.

For the reasons above, the nested fences feature really is just a workaround. But for a lot of people, this functionality is more than sufficient.

Options

Option Type Default Description
css_class string '' Class name is applied to the wrapper element of the code. If configured, this setting will override the css_class option of Highlight. If nothing is configured here or in Highlight, the class highlight will be used.
disable_indented_code_blocks bool False Disables Python Markdown's indented code block parsing. This is nice if you only ever use fenced blocks.
custom_fences [dictionary] [] Custom fences.
preserve_tabs bool False Experimental feature that preserves tabs in fenced code blocks.