Create a sphinx extension to customize your docs

With Sphinx is easy to generate documentation of your python project, as long as you don’t require some custom code. This is a tutorial of how to create a quick & dirty Sphinx extension to personalize the docs of your project.

In our open-source project, Devicehub, we have some Schemas that configures our API (like a User schema), and we want to generate documentation for those. The documentation takes variables from the schemas and prints them nicely. This is the code, this is where our extension is called in a .rst file, and this is the outcome.

In this tutorial we use our extension as an example of a simple Sphinx custom extension. We assume you already know the basics of Sphinx and reStructured text.

A minimalist Sphinx extension requires two things:

  1. A class that extends from docutils.parsers.rst.Directive that has our code.
  2. A function called setup that manually adds our class to Sphinx.

Both things can be in our project’s sphinx’s If you check our code, you see our class and our setup function. Read this Sphinx page that explains this class and setup function and then come back here :-). The following sections explore parts not covered by the sphinx’s tutorial.

Extension class skeleton and arguments

Our extension has only one parameter (called too “argument”): a string with the path where our Schema is, and we want our extension to print some variables of the schema and its subschemas.

class DhlistDirective(Directive):
    has_content = False
    option_spec = {'module': directives.unchanged}

The first portion of our extension class is about configurations of our directive, passed-in as class variables (here you have the full list of configurations of the directive). Leave has_content as Falseoption_spec is a dictionary of the parameters of your directive and their type. Sphinx coerces the type for each argument, and the safest one to use is directives.unchanged, which returns the argument as a string, just as the user writes it. In our case we define a parameter called module, and the user has to write it like this in a .rst file:

.. dhlist::
    :module: ereuse_devicehub.resources.device.schemas

And our directive gets the value “ereuse_devicehub.resources.device.schemas” in self.options['module']:

    def run(self):
        my_argument = self.options['module']

run is the main method of our extension. It is executed every time the user writes ..dhlist:: in a .rst file. In our case we use the string of self.options['module'] as a path to import the Schema is referencing at.

Docutil’s elements

The objective of our extension is to return something to Sphinx so that it can translate it to HTML or to a PDF file. This something is a list of Docutil’s element, and they define a link, a bullet list, a section of a document, etc. Here you have the reference docs. In fact, when reading a rst file, we can think that Sphinx parses the rst syntax into Docutil’s elements, and then tranforms them to HTML, PDF…

So, we have that a section element in rst is the same as the Docutil’s section element.

Element hierarchy

These elements have a strong hierarchy that you always have to have in mind when using them, otherwise it fails:

| document  [may begin with a title, subtitle, decoration, docinfo]  |
|                             +--------------------------------------+
|                             | sections  [each begins with a title] |
| [body elements:]                                      | (sections) |
|         | - literal | - lists  |       | - hyperlink  +------------+
|         |   blocks  | - tables |       |   targets    |
| para-   | - doctest | - block  | foot- | - sub. defs  |
| graphs  |   blocks  |   quotes | notes | - comments   |
| [text]+ | [text]    | (body elements)  | [text]       |
| (inline +-----------+------------------+--------------+
| markup) |

In our extension we return a list of section elements, as we create full blocks of content. If we do something like this in our .rst file:

The following schema represents all the device types and their

.. dhlist::
    :module: ereuse_devicehub.resources.device.schemas

Note that the header Schema is a subsection (HTML’s <h2>), our module is inserted inside Schema’s subsection, so our module ends up being a subsubsection (HTML’s <h3>), which means that our module plays nice with the rest of content.

In the following sections we learn how to use docutil’s elements by showing the ones that appear in our project.

Base: section, title, link

The following example of run returns a list of elements (which is what Sphinx expects). There is only one element, a section.

import docutils.nodes as n

def run(self):
    section = n.section(ids=n.make_id('the-html-anchor'))
    section += n.title('Text', 'Text') # Just add the text twice
    section += n.any_other_element_that_can_be_in_a_section()
    sections = [section]
    return sections<br>

The section can have an id, which is the anchor used in HTML/PDF for linking. In this case we add a title (we need to add its value twice for some strange reason), and any other element that we want inside the section (note the section += ...)

The following code shows ho to generate the hyperlink that takes you to the section (this generates the <a> HTML tag, for example):

ref = n.reference(text='a link!')
ref['refuri'] = '#the-html-anchor'
p = n.paragraph()
p += ref
# We can add the paragraph to a section
section += p

Lists (bullet points)

l = n.bullet_list('')
l += n.list_item('', n.paragraph('Text'))
sublist = n.bullet_list('')
sublist += n.list_item('', n.paragraph('Sublist!'))
l += sublist
# We can add to a section:
section += l


# Create a field list
fields = n.field_list()
# Create a field
field_name = n.field_name(text='Some text')
field_body = n.field_body('', 
    n.paragraph('Some text'),
    n.paragraph('Some more text in another paragraph.')
field = n.field('', field_name, field_body)
# Add the field to the field list
fields += field
# We can add fields to an existing section, for example:
section += fields

Parse rst strings to docutil’s elements

In our project we end up reading string variables with rst syntax. The following code allows us to transform those strings into docutil’s elements that we can add, for example, in our sections (check the source code here):

from docutils.statemachine import StringList, string2lines
import inspect

def parse(self, text) -> n.container:
    """Parses text possibly containing ReST stuff and adds it in
     a node."""
    p = n.container('')
    self.state.nested_parse(StringList(string2lines(inspect.cleandoc(text))), 0, p)
    return p

Further info

The following tutorials can help you dive deeper into Sphinx extension development:

Happy coding 🙂

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.