It's not a lot of code maybe it makes sense to build smth for us with slots and all?

@Poolitzer that's what I was thinking as well, maybe open an issue in Ruff, cause it will get implemented pretty quickly if accepted

It's not a lot of code maybe it makes sense to build smth for us with slots and all?

Have already forked :D https://github.com/Bibo-Joshi/sort-all

that's what I was thinking as well, maybe open an issue in Ruff, cause it will get implemented pretty quickly if accepted

also a nice idea! would you mind doing that? Things to mention IMO are

• the sort-all package for reference
• comment-handling
• extensibility for general iterables of strings with user defined names

If you like the idea in general, I'd like to go ahead with the package that we have so far :)

@Bibo-Joshi @clot27 points to astral-sh/ruff#1198 in which it's mentioned that isort can do this already. I'll have to check

isort does support sorting __all__ via the sort_reexports option even though this option is not documted in the options list (only here). One major bummer however is that this doesn't seem to honor the black settings, in particular it uses single quotes (' instead of ") and it collapses multiline definitions (see also PyCQA/isort#2193). One would have to apply black again and the pre-commit step would always fails.
I would therefore not consider the isort implementation as useful for us ATM.

PyCQA/isort#2193 also mentions https://pypi.org/project/asort/. Even though that package is even less maintained than sort-all, it worked well on first try and even kept comments. I'll have another look at it in time

Welp, asort completely deletes the content of __all__ when run against our code base and also messes up several unrelated things. so not an option …