Skip to main content

A Doxygen filter for Python

Project description

A more Pythonic version of doxypy, a Doxygen filter for Python.


For now Doxygen has limited support for Python. It recognizes Python comments, but otherwise treats the language as being more or less like Java. It doesn’t understand basic Python syntax constructs like docstrings, keyword arguments, generators, nested functions, decorators, or lambda expressions. It likewise doesn’t understand conventional constructs like doctests or ZOPE-style interfaces. It does however support inline filters that can be used to make input source code a little more like what it’s expecting.

The excellent doxypy makes it possible to embed Doxygen commands in Python docstrings, and have those docstrings converted to Doxygen-recognized comments on the fly per Doxygen’s regular input filtering process. It however does not address any of the other previously mentioned areas of difficulty.

This project started off as a fork of doxypy but quickly became quite distinct. It shares little (if any) of the same code at this point (but maintains the original license just in case). It is meant to support all the same command line options as doxypy, but handle additional Python syntax beyond docstrings.

Additional Syntax Supported

Python can have functions and classes within both functions and classes. Doxygen best understands this concept via its notion of namespaces. This filter thus can supply Doxygen tags marking namespaces on every function and class. This addresses the issue of Doxygen merging inner functions’ documentation with the documentation of the parent.

Python class members whose names begin with a double-underscore are mangled and kept private by the language. Doxygen does not understand this natively yet, so this filter additionally provides Doxygen tags to label such variables as private.

Python frequently embeds doctests within docstrings. This filter makes it trivial to mark off such sections of the docstring so they get displayed as code.

ZOPE-style interfaces overload class definitions to be interface definitions, use embedded variable assignments to identify attributes, and use specific function calls to indicate interface adherence. Furthermore, they frequently don’t have any code beyond their docstrings, so naively removing docstrings would result in broken Python. This filter has basic understanding of these interfaces and treats them accordingly, supplying Doxygen tags as appropriate.

Fundamentally Python docstrings are meant for humans and not machines, and ought not to have special mark-up beyond conventional structured text. This filter heuristically examines Python docstrings, and ones like the sample for complex in PEP 257 or that generally follow the stricter Google Python Style Guide will get appropriate Doxygen tags automatically added.

How It Works

This project takes a radically different approach than doxypy. Rather than use regular expressions tied to a state machine to figure out syntax, Python’s own Abstract Syntax Tree module is used to extract items of interest. If the autobrief option is enabled, docstrings are parsed via a set of regular expressions and a producer / consumer pair of coroutines.


This filter will correctly process code like the following working (albeit contrived) example:

def myfunction(arg1, arg2, kwarg='whatever.'):
    Does nothing more than demonstrate syntax.

    This is an example of how a Pythonic human-readable docstring can
    get parsed by doxypypy and marked up with Doxygen commands as a
    regular input filter to Doxygen.

        arg1:   A positional argument.
        arg2:   Another positional argument.

        kwarg:  A keyword argument.

        A string holding the result.

        ZeroDivisionError, AssertionError, & ValueError.

        >>> myfunction(2, 3)
        '5 - 0, whatever.'
        >>> myfunction(5, 0, 'oops.')
        Traceback (most recent call last):
        ZeroDivisionError: integer division or modulo by zero
        >>> myfunction(4, 1, 'got it.')
        '5 - 4, got it.'
        >>> myfunction(23.5, 23, 'oh well.')
        Traceback (most recent call last):
        >>> myfunction(5, 50, 'too big.')
        Traceback (most recent call last):
    assert isinstance(arg1, int)
    if arg2 > 23:
        raise ValueError
    return '{0} - {1}, {2}'.format(arg1 + arg2, arg1 / arg2, kwarg)

There are a few points to note:

1. No special tags are used. Best practice human-readable section headers are enough.

2. Some flexibility is allowed. Most common names for sections are accepted, and items and descriptions may be separated by either colons or dashes.

3. The brief must be the first item and be no longer than one line.

4. Everything thrown into an examples section will be treated as code, so it’s the perfect place for doctests.

Additional more comprehensive examples can be found in the test area.

Installing doxypypy

One can use either pip or easy_install for installation. Running either:

pip install doxypypy


easy_install doxypypy

with administrator privileges should do the trick.

Many Linux distributions have packages for doxypypy, so if you are using Linux you may find it more convenient to use aptitude, apt, apt-get, yum, dnf, etc. as appropriate for your system to install the version tested by the distribution maintainer. It will often be available as separate packages for both Python 3 and Python 2.

Previewing doxypypy Output

After successful installation, doxypypy can be run from the command line to preview the filtered results with:

doxypypy -a -c

Typically you’ll want to redirect output to a file for viewing in a text editor:

doxypypy -a -c >

Invoking doxypypy from Doxygen

To make Doxygen run your Python code through doxypypy, set the FILTER_PATTERNS tag in your Doxyfile as follows:

FILTER_PATTERNS        = *.py=py_filter

py_filter must be available in your path as a shell script (or Windows batch file). If you wish to run py_filter in a particular directory you can include the full or relative path.

For Unix-like operating systems, py_filter should like something like this:

doxypypy -a -c $1

In Windows, the batch file should be named py_filter.bat, and need only contain the one line:

doxypypy -a -c %1

Running Doxygen as usual should now run all Python code through doxypypy. Be sure to carefully browse the Doxygen output the first time to make sure that Doxygen properly found and executed doxypypy.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

doxypypy- (45.9 kB view hashes)

Uploaded source

Built Distribution

doxypypy- (86.6 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page