[bug?] B018 useless-expression does not work

[smackesey]

None of the following trigger B018 useless-expression for me, when I would expect them all to:

"foo" + "bar"  # should trigger B018

"foo"  # should trigger B018

object().__class__  # should trigger B018

I might be missing something about what this rule is supposed to do though.

[charliermarsh]

I think this is consistent with flake8-bugbear although I'm not sure why their rule is limited to those cases.

[charliermarsh]

I'm guessing they got rid of strings to avoid false positives with docstrings (https://github.com/PyCQA/flake8-bugbear/pull/209/files).

I'm guessing they don't detect function calls since it's not really possible (with current information) to know whether a function call is effect-ful or pure.

We could definitely detect binary operations though.

[charliermarsh]

In general I think we can improve this even if it deviates a bit from bugbear.

[smackesey]

I was actually unable to get this rule to trigger at all-- do you have an example of something that triggers it? Like, 1 + 5 doesn't trigger it either.

I'm guessing they don't detect function calls since it's not really possible (with current information) to know whether a function call is effect-ful or pure.

Yeah, the expected behavior is not to detect a straight function call, but when it's followed by attribute access (as in my example) then I would think it should be detected. FWIW that's pylint's behavior.

[charliermarsh]

Prior to my open PR, the rule was only enforced in function and class bodies. I think this was consistent with bugbear but has since changed, so it’s reflected in the PR.

[smackesey]

Ah nice, somehow my eyes glossed over the PR notification. Thx for addressing.

[charliermarsh]

All good, that PR still doesn't include the object().__class__ case which I may add.