Fix Code Quality CI: checkout action and complex function

[glatterf42]

This PR intends to fix a few things about the Code Quality CI:

• It fixes a typo such that the Code Quality check will consider the changes here, not those already on main
• It reduces the complexity of the cool_tech() function to below 11, which is our accepted limit here (@adrivinca, this may serve as a template for the future, if you're interested: I have simply extracted one if/else clause and moved that to its own function, following the idea that every function should have one clearly defined task)
• I've also deleted the pre-commit cache in hopes of it now picking up the latest version of genno, which should fix the mypy errors we've been seeing in the nightly tests

Housekeeping:

• Mirror the 'fix' from Adjust type hints for xarray 2025.3 khaeru/genno#166 for a mypy error.
• Adjust import of the Annotation class from sdmx1 to avoid warnings added in v2.22.

How to review

• Read the diff and note that the CI checks all pass.

PR checklist

• Continuous integration checks all ✅
• Add or expand tests; coverage checks both ✅
• [ ] Add, expand, or update documentation. Just CI.
• [ ] Update doc/whatsnew. Just CI.

[glatterf42]

I have also deleted the cache entry called setup-uv-1-x86_64-unknown-linux-gnu-3.9-..., as this is used by the setup-uv action in the code quality run, but it doesn't seem to have any effect.
Also, even though the linked workflow file uses needs.check.outputs.ref for the code quality check now, the checkout action still resolves to a0d5420, the latest commit on main.

I'll have to do some organizing for the Community Meeting now and can only come back to this later, I'm afraid.

[codecov]

Codecov Report

Attention: Patch coverage is 90.00000% with 5 lines in your changes missing coverage. Please review.

Project coverage is 77.4%. Comparing base (a0d5420) to head (2540694).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
message_ix_models/model/material/cli.py	0.0%	2 Missing ⚠️
message_ix_models/tools/wb.py	33.3%	2 Missing ⚠️
message_ix_models/report/operator.py	0.0%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #320     +/-   ##
=======================================
- Coverage   78.3%   77.4%   -0.9%     
=======================================
  Files        217     217             
  Lines      16913   16925     +12     
=======================================
- Hits       13254   13116    -138     
- Misses      3659    3809    +150     

Files with missing lines	Coverage Δ
message_ix_models/model/bare.py	100.0% <100.0%> (ø)
message_ix_models/model/disutility.py	100.0% <100.0%> (ø)
message_ix_models/model/structure.py	100.0% <100.0%> (ø)
message_ix_models/model/transport/build.py	94.9% <100.0%> (ø)
...essage_ix_models/model/water/data/water_for_ppl.py	94.7% <100.0%> (+<0.1%)	⬆️
message_ix_models/project/edits/__init__.py	94.9% <100.0%> (+<0.1%)	⬆️
message_ix_models/project/ssp/structure.py	100.0% <100.0%> (ø)
message_ix_models/tests/model/test_disutility.py	100.0% <100.0%> (ø)
message_ix_models/tests/model/test_macro.py	100.0% <100.0%> (ø)
message_ix_models/tests/model/test_structure.py	100.0% <100.0%> (ø)
... and 7 more

... and 7 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[khaeru]

even though the linked workflow file uses needs.check.outputs.ref for the code quality check now, the checkout action still resolves to a0d5420, the latest commit on main.

I think that is expected. The web display on GitHub itself seems to be erroneous; the pull_request_target trigger always uses the version of the workflow file on the target branch, in this case main.

In order to debug, you can do what it says here:

message-ix-models/.github/workflows/pytest.yaml

Lines 4 to 9 in a0d5420

	# To debug the workflow, uncomment this entry AND comment pull_request_target
	# pull_request:
	# branches: [ main ]
	pull_request_target:
	branches: [ main, "migrate**" ]
	types: [ labeled, opened, reopened, synchronize ]

in a commit with a message like "TEMPORARY Run 'pytest' workflow from PR branch"; preserve a link to the successful run; and then rebase the branch to drop the temporary commit.

[glatterf42]

Thanks, I had forgotten that was necessary.

[adrivinca]

Thanks, looks good to me. should we write a test for the new function?

[glatterf42]

We could, but as long as cool_tech() is tested, the new function is already tested, too! That's the beauty of it: it's simply part of the larger function, so when cool_tech() runs, it also runs.
I'm trying to get the Code Quality check to pass, then we can worry about adding a test, but it's not strictly necessary here. Though it would be good practice.

[glatterf42]

Even adding the main branch of genno to the additional requirement for mypy does not seem to resolve the errors. I don't really understand what's going on there as all these type hints are very abstract. I could dive in, but I think I should be doing other things instead. So I would opt for applying type: ignore if we have tests for both places where mypy raises errors and these pass.

Unfortunately, there is now another error originating from the latest release of sdmx1, and I don't really have time to look into that, either :/

[khaeru]

Unfortunately, there is now another error originating from the latest release of sdmx1, and I don't really have time to look into that, either :/

That's something I caused, so I'll step in with a commit to fix it.

[khaeru]

I'll step in with a commit to fix it.

I believe I've addressed the issues related to genno and sdmx1, so back to you @glatterf42.

[adrivinca]

At the end there were more changes elsewhere than in the water module, that looks fine for me. But have to admit I did not dig much in all the scripts.
You and Paul know what you are doing :)

[glatterf42]

These tests will fail again, at least in part, because they're using the code stored on main. Since they were passing before, I'll merge without waiting for them.