Add executable kwarg to trio.run_process type hints

[dj-foxxy]

The executable kwarg is missing from the type hints of trio.run_process. It works at run time but Pyright is complaining.

[A5rocks]

The executable kwarg is missing from the type hints of trio.run_process. It works at run time but Pyright is complaining.

Absolutely, and it looks like we're missing other kwargs in the signature too. (and same for trio.open_process) Namely:

• executable (of course)
• group (3.9+, unix)
• extra_groups (3.9+, unix)
• user (3.9+, unix)
• umask (3.9+, unix)
• pipesize (3.10+, linux but is ignored so all platforms? or maybe unix)
• process_group (3.11+, unix)

Feel free to make a PR -- you'll have to add some if sys.version_info checks and update the cheat sheet comment on top of the overloads.

---

If you no longer want Latta AI to attempt solving issues on your repository, you can block this account.

Done!

[jakkdl]

That if TYPE_CHECKING block is gonna become an absolute monstrosity - it's already a ~third of the file with 400 lines. This might warrant a separate file (_subprocess_type_hints.py?) and maybe also a dedicated test that compares the type-checker inferred signature with that of subprocess.run (although given the false alarm rate on test_static_tool_sees_[...] I'm not terribly excited about that)

[A5rocks]

maybe also a dedicated test that compares the type-checker inferred signature with that of subprocess.run (although given the false alarm rate on test_static_tool_sees_[...] I'm not terribly excited about that)

I don't think this is possible because typeshed structures its overloads differently to ours (eg no platform specific ones)

[dj-foxxy]

Happy to give this a go. Can I create script than generates it? I feel like doing this by hand is too much work and maintenance nightmare when they add more kwargs.

[A5rocks]

Can I create script than generates it? I feel like doing this by hand is too much work and maintenance nightmare when they add more kwargs.

If you want to! I don't think it's necessary: we'll just change the 2 overloads for things on Unix into 6 overloads across 3 different branches. But also 6 overloads still is somewhat large, plus open_process. (If you do make a script, don't bother with making it work completely on its own IMO)

[jakkdl]

I think maintaining a script across platforms etc is going to be way more work than adding a kwarg at most once a year with a new python release. Lotsa stuff needs updating with new releases anyway.
But for generating the initial set, an LLM and/or a quick script might well be worth it.