Intended Audience
- Developers
License
- OSI Approved :: BSD License
Programming Language
- Python :: 3.3
Operating System
- OS Independent
Topic
- Documentation
Use Python 3 annotations in sphinx-enabled docstrings
If you’re on Python 3 and writing sphinx-enabled docstrings, you might feel like doing needless work when typing :type arg: or :rtype: directives. After all, why not use annotations for this?
Sure, :param str arg: description is not a lot of work, but when you want to document your argument as a specific class for which you have a :class: link, then you need to use :type: and it’s cumbersome. By using this sphinx extension, you can turn this:
def f(a):
"""Do something.
:param a: description for a
:type a: :class:`ClassForA`
:rtype: str
"""
into:
def f(a: ClassForA) -> str:
"""Do something.
:param a: description for a
"""
Installation
First, you need Python 3.3+ and a Sphinx documentation (with autodoc enabled).
You can install sphinx-autodoc-annotation with:
$ pip install sphinx-autodoc-annotation
Then, you need to enable it in your conf.py file:
extensions = [
'sphinx.ext.autodoc',
'sphinx_autodoc_annotation',
]
You’re done!
Usage
All you need to do to use this extension is to properly annotate your functions and methods with expected types for your arguments and return value. :type: and :rtype: directives will automatically be added to your docstring.
These directives behave like if you added them manually, that is, your argument is not going to show up only with :type arg: you need :param arg: to be there (with a description of what it does) for your type to show up.
When there are no annotations, argument types are deduced from default values. If your default value is a bool, str, int or float, the argument is going to be considered of that type. That feature is there mainly because f(flag: bool = False) feels a bit redundant.
In all cases, :type: and :rtype: directives in the docstring will always have precedence over annotations and defaults.