I’m looking for a way to generate a single Markdown (.md) file that includes all the file names, function definitions, and docstrings from my Python files. Ideally, this should traverse a directory recursively while respecting the .gitignore rules.
Does anyone know of a package that accomplishes this?
There are A LOT of these. I have had good experience with https://pypi.org/project/git2md
Says “Support for .gitignore, .globalignore and .mdignore for local projects: Automatically excludes files/directories specified in .gitignore, .globalignore or .mdignore.”
If you don’t find what you like, I’ve also found it easy to ask LLMs to just write a tool customized to my use case, usually when I need to somehow condense the code base to make it fit in the context.
Sphinx’s autodoc tool does gathering documentation and will traverse the whole package with the right config. It can read markdown with myst-parser. It provides HTML, TeX etc as output formats out of the box, however if you need markdown as output you many need an extension.
Just spent the last two weeks (no exaggeration) putting documentation together. Ensuring references resolve correctly is the time consuming bit.
extlinks, nitpick_ignore, autodoc_type_aliases, intersphinx_mapping, multiple extensions, and building inventories using sphobjinv, Table of contents using sphinx-external-toc-strict links to both ref and docs.
The table of contents supports both
.mdand.rstfiles. With the caveat that the initial file must beindex.rst.Having to relearn all this crap to do the same thing (e.g. mkdocs)? Would rather stick a fork in my eye.
Here is a sample TOC config file wreck _toc.yml from package wreck. Shows
globs:,file:,url:links.This one includes both
.rstand.mdstrict TOC _toc.ymlOutput is a user manual web site or a pdf.
markdown and restructuredText are input files, not the end goal.
Would seriously love someone to rip down Sphinxdoc and put this baby to bed.
In my history of using docstrings, I’ve put maybe 20-40minutes of docs into a 200k+ codebase. It’s super easy.
In my history of using Sphinx autodoc, I’ve spent about 5-10 hours fumbling with Makefiles, configs, and other nightmares and it still has the worst developer experience after about 8 years of using it.
For the
Makefileandconf.py, it’s a Ctrl+P job.tox.inimore Ctrl+P. Crap that goes intopyproject.tomlCtrl+P.ETA ~20 mins including bathroom breaks and chatting with friends that randomly arrive at the cafe.
Then comes dealing with fixing up the URLs and references. That is the only time consuming bit.
Have posted links to packages i’ve written, can just go there and five finger discount the configuration files.
Howto autodoc, i’m not licensed to sell without a medical license. But there is a shot for that. ;-)
Will checkout your repos. Thanks for comment. It’s been a breeze to write python and the docstrings are pretty great. I’ve had periods where everything just worked for Sphinx autodoc and then others where it’s just broken for months. Thanks again
If/when run into issues with Sphinx, 1st the project should be accessible online then message me with the URL and ask for help. I’m battle hardened; will conquer whatever Sphinx autodoc demons assails you.
This goes for everyone else as well. If have the Sphinx blues, don’t whine about it, just ask for help.
Especially if your project is awe inspiring. Online read lots of complaining about writing documentation, but see it as just as important as the code or business model. Your documentation should not suck. Don’t half ass it.
Also i’m really curious how autodoc stays broken (for months). Take that to mean, there are things that need to be learned. Not knowing those things Sphinx can be frustrating. For example, dealing with Type aliases and table of contents.
