Enable flake8-type-checking rule

[CoolCat467]

This PR enables the ruff flake8-type-checking rule (TCH).
From flake8-type-checking's description, the main benefit of this would be avoiding the overhead of type hint imports at runtime.

[codecov]

Codecov Report

Merging #2813 (f98500d) into master (e6e18bd) will decrease coverage by 0.01%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2813      +/-   ##
==========================================
- Coverage   99.18%   99.18%   -0.01%     
==========================================
  Files         115      115              
  Lines       17611    17551      -60     
  Branches     3149     3149              
==========================================
- Hits        17468    17408      -60     
  Misses         99       99              
  Partials       44       44              

Files	Coverage Δ
trio/_channel.py	100.00% <ø> (ø)
trio/_core/_asyncgens.py	100.00% <ø> (ø)
trio/_core/_ki.py	100.00% <ø> (ø)
trio/_core/_multierror.py	100.00% <ø> (ø)
trio/_core/_parking_lot.py	100.00% <ø> (ø)
trio/_core/_run.py	100.00% <100.00%> (ø)
trio/_core/_tests/test_asyncgen.py	99.08% <100.00%> (-0.01%)	⬇️
trio/_core/_tests/test_guest_mode.py	100.00% <ø> (ø)
trio/_core/_tests/test_instrumentation.py	100.00% <100.00%> (ø)
trio/_core/_tests/test_io.py	100.00% <ø> (ø)
... and 38 more

[TeamSpen210]

I would argue against this, type hints are useful to have at runtime. This will break Sphinx for instance.

[jakkdl]

I would argue against this, type hints are useful to have at runtime. This will break Sphinx for instance.

is Sphinx == ReadTheDocs here? There are (only) 4 errors, or is the signatures in the documentation broken?

Digging around a bit, it seems you can get Sphinx to run with TYPE_CHECKING = True?
https://github.com/sphinx-doc/sphinx/blob/bb74aec2b6aa1179868d83134013450c9ff9d4d6/sphinx/ext/autodoc/importer.py#L107

[CoolCat467]

pre-commit.ci autofix

[jakkdl]

RTD build is failing, which I'm pretty sure it's not on master, so something is being messed up somewhere.

[CoolCat467]

RTD build is failing, which I'm pretty sure it's not on master, so something is being messed up somewhere.

Errors I'm seeing in the most recent run RTD had:

trio/__init__.py:docstring of trio.run_process:1: WARNING: py:class reference target not found: StrOrBytesPath
trio/__init__.py:docstring of trio.run_process:1: WARNING: py:class reference target not found: StrOrBytesPath
trio/lowlevel.py:docstring of trio.lowlevel.Task.iter_await_frames:1: WARNING: py:class reference target not found: FrameType
trio/lowlevel.py:docstring of trio.lowlevel.start_guest_run:1: WARNING: py:class reference target not found: Outcome

Not completely sure what this means, something about something missing from docstrings? I'm sure if I looked into it deeper I could figure it out but if anyone has more information I would be glad for the advise. Otherwise I'll do some research later and we can get this taken care of.

[jakkdl]

similar problems required aliases or similar in docs/source/conf.py, or shuffling around imports. You can run the build process locally with make html standing in the docs/ directory, which should make it much quicker to try to playing around with the code and figure out what is erroring where and why.
You could e.g. git blame the lines in it to dig up the PRs that added them, but I tried to add decent comments since Sphinx is kind of magic to me.

[TeamSpen210]

Sphinx isn't able to find those names when it's trying to evaluate the annotations - possibly because those functions were imported into a different class or something. Probably need to move the imports out of TYPE_CHECKING. It could also be caused by us importing trio in conf.py to get __version__ - maybe change that to read the file itself, so the module isn't loaded?

[jakkdl]

Probably need to move the imports out of TYPE_CHECKING.

this shouldn't be the case if 4764ba2 is working as expected.

[CoolCat467]

google/pytype#1066 seems to be having a similar issue with StrOrBytesPath

[jakkdl]

I kind of expected the RTD build to fail, but apparently it runs on 3.9+, whoop whoop.

google/pytype#1066 seems to be having a similar issue with StrOrBytesPath

I didn't manage to figure out any way of getting StrOrBytesPath working, but it does not seem related to the above afaict.

[CoolCat467]

I kind of expected the RTD build to fail, but apparently it runs on 3.9+, whoop whoop.

google/pytype#1066 seems to be having a similar issue with StrOrBytesPath

I didn't manage to figure out any way of getting StrOrBytesPath working, but it does not seem related to the above afaict.

I thought I would mention it, one of the only internet search results with anything potentially related. Maybe I was just looking in the wrong places, but it seems like sphinx doesn't have very much information about fixing errors.

Thank you so much for helping with this, I was definitely struggling a bit!

[TeamSpen210]

For StrOrBytesPath, try adding it to autodoc_type_aliases, or we could use autodoc_process_signature() to expand it. That function gets passed the entire function signature as a string, so you can edit it however you like.

[jakkdl]

For StrOrBytesPath, try adding it to autodoc_type_aliases, or we could use autodoc_process_signature() to expand it. That function gets passed the entire function signature as a string, so you can edit it however you like.

I tried autodoc_type_aliases without any success

[jakkdl]

teamwork! 🚀