feat(requirements): Support flake8 v5

[DmytroLitvinov]

Closes #45

[jonyscathe]

I am guessing that this is what is stuffing up those tests: PyCQA/flake8#1609

Not sure the best way around that... I don't think it causes any issues, but does mean a few tests aren't quite valid anymore.

Also, given how infrequently this project is updated, perhaps the flake8 upper bound should just be completely removed?

[mschwager]

Yes, those TestFlake8Extension tests rely a bit on the implementation details of flake8. Not ideal, but it was the only way I could test some of the non-standard optimizations done in dlint.

The changelog states (https://flake8.pycqa.org/en/latest/release-notes/5.0.0.html):

Properly differentiate between explicitly ignored / selected and default ignored / selected options (See also #284, #1576, #1609).

Is this just a matter of updating the defaults in create_options in tests/test_extension.py?

[jonyscathe]

@mschwager changing the defaults in create_options can almost do it.

Currently the ignore and extend_ignore tests fail because all linter_classes come up as ignored given that they are not explicitly selected (only implicitly selected with the extended_default_select default of ["DUO"]).
Those two tests can pass if the default value for the "select" kwarg is changed to ["DUO"].
And if that change is made then the only failed test is test_flake8_extension_get_linter_classes_extended_default_select

And I'm not sure that the extended_default_select option can be tested using the current test methods. The extended_default_select setting doesn't appear to have any affect on the engine.decision_for() value for any given code. If a code (or "DUO" for all DUOXXX codes) is explicitly selected with "select" or "extend_select" and its code is not in the ignore or extend_ignore list then its decision_for() value is style_guide.Decision.Selected. If it isn't explicitly selected with "select" or "extend_select" then its decision_for() value will be style_guide.Decision.Ignored no matter what the value of extended_default_select is.

I'm not entirely sure how important the extended_default_select test is for dlint. Nor how important the behaviour of get_linter_classes() is with different option args so I'm not sure what the best way forward is for that test.

[mschwager]

Thanks for digging into this.

Now I remember where I got the initial create_options code: https://github.com/PyCQA/flake8/blob/main/tests/unit/test_decision_engine.py

Copying the latest create_options changes from that file to Dlint made the tests pass for me. In other words, select, extend_select, ignore, and extend_ignore must default to None, not []. If you add those changes to this PR I suspect CI will succeed and we'll be good to go 🎉

That change shouldn't affect any functionality in Dlint. Since None vs. [] is an implementation detail of CLI arguments passed to flake8, the Dlint optimizations in get_linter_classes shouldn't be affected, and it will do the correct thing in flake8<5.0.0 and flake8>=5.0.0.

[jonyscathe]

Ah yep, awesome. Sounds good to me.
@DmytroLitvinov are you able to make those changes from https://github.com/PyCQA/flake8/blob/main/tests/unit/test_decision_engine.py into tests/test_extension.py since this is a PR from your fork? Or if not I guess I can make a PR into your fork (or possibly just a new PR might be simpler given this has minimal changes)

[DmytroLitvinov]

Good morning everyone 👋
I have pushed changes. Could some of the maintainers run CI?

[mschwager]

It passed! Thanks again for the PR, I will cut a new release of Dlint now 👍

[mschwager]

It's up: https://pypi.org/project/dlint/0.13.0/