You are right, I didn't see the isort skip until you mentioned it. I did some tests and it appears ruff only respects this line if it is a comment, and not in the docstring:

This doesn't work

"""
isort:skip_file
"""

nor this

"""
# isort:skip_file
"""

but this works

"""
...
"""
# isort:skip_file

But as you said, there is no compelling reason to skip sorting imports in these files

I think I'll create a stopgap PR to skip running ruff's isort rules on *_pb2.pyi files, since that preserves the status quo and will stop pre-commit-ci pushing unwelcome "autofixes" to every PR that's filed at typeshed. But let's keep this open since, as I say, there doesn't appear (to me) to be a compelling reason not to sort imports in these files.

I think I'll create a stopgap PR to skip running ruff's isort rules on *_pb2.pyi files, since that preserves the status quo and will stop pre-commit-ci pushing unwelcome "autofixes" to every PR that's filed at typeshed. But let's keep this open since, as I say, there doesn't appear (to me) to be a compelling reason not to sort imports in these files.

Done in #10940. This is now a PR to enable running ruff's isort rules on *_pb2.pyi files!

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

On the one hand, all these files have isort:skip_file in their module docstrings. It seems obvious why ruff isn't respecting that directive in the same way isort used to!

I can also see why it'd be ignored from a docstring, could be worth raising the deviation to be documented in https://docs.astral.sh/ruff/linter/#action-comments