Investigate false negatives for ``unsupported-binary-operation``

[DudeNr33]

Following the discussion here:

Originally posted by @DanielNoord in #6353 (comment)

unsupported-binary-operation was previously disabled due to too many false positives.
Maybe we can investigate if we can re-enable it?

[omarandlorraine]

Have we got a regression test file for (at least some of the) false positives?

[DanielNoord]

I don't think so. If you want to look into this you could (for example) use some of the projects we have in our primer and see what happens with:
pylint --disable=all --enable=unsupported-binary-operation after having re-enabled the checker. That might show whether there are too many false positives still.

[jacobtylerwalls]

Actually, I think this issue is wrong -- the check is enabled currently, see _detect_unsupported_alternative_union_syntax and #4256

[jacobtylerwalls]

Clearly a cleanup is needed to prevent confusion and remove the todos.

[DanielNoord]

I think the message is partially emitted. There is visit_binop method which seems to do something, but there is also a _visit_binop and a _visit_augassign both of which call _check_binop_errors. Those two methods are decorated with @only_required_for_messages("unsupported-binary-operation") but never called.

We probably need to check if we should indeed look at AugAssign and whether there is something in _check_binop_errors that _detect_unsupported_alternative_union_syntax does not do.

[jacobtylerwalls]

_check_binop_errors doesn't exist anymore :D

I think we can just remove the TODOs, there are tests for += in unsupported_binary_operation.py

[DanielNoord]

https://github.com/PyCQA/pylint/blob/0a4204fd7555cfedd43f43017c94d24ef48244a5/pylint/checkers/typecheck.py#L1921-L1929

🤔

[jacobtylerwalls]

oh my word. the things that happen when you investigate an issue by deleting the routine you're talking about. ☕

[jacobtylerwalls]

The code there looks very similar to visit_unary_op() which should also be investigated and cleaned up.