Rule proposal: in-empty-collection

[naslundx]

Summary

After discussions in #15732 and #15729 an idea was proposed to add a rule for checking unnecessary membership test in empty collections.

This would reinstate the old behavior in rule PLR6201 to explicitly check for things like if 1 in [] or if 'a' not in ''.

Should ruff include it?

Example implementation is here: #16480

[MichaReiser]

Unrelated on whether we decide to add this rule. I think I'd prefer naming it in-empty-sequence to match python's semantics OR exclude str from the rule

Strings are immutable sequences of Unicode code points. String literals are written in a variety of ways:
source

[MichaReiser]

I think having this as a rule does make sense. What would you suggest as proposed fix?

[MichaReiser]

@dylwil3 you asked in #16480 to open this issue. Do you have any concerns with the rule or should we accept this request and land the PR?

[dylwil3]

I'm personally happy with the rule, was just checking boxes - sorry for the bureaucratic red tape! @naslundx would you like to sync with main, respond to the remaining comments on the PR and we can get this landed?

[Avasam]

On the name semantics, in is the operator for __contains__ * (see message below for caveat), and typing.Container is the Protocol for def __contains__. So in-empty-container ?
That covers strings, lists, tuples, dicts, sets, deque, and I'm sure many other stdlib containers I'm forgetting.
Ok, these are all collections... tbh I'm having a hard time finding a stdlib container that isn't a collection XD

Btw sequences are collections with a few extra methods. So the originally proposed name in-empty-collection still works whilst including strings imo.

[Avasam]

"" in iter(()) and "" in iter("") (ie: Iterator[Never]) seem to be in the spirit of this rule, but not a Container in the Protocol sense. But in also works for Iterables and implementers of __getitem__() https://docs.python.org/3/reference/expressions.html#in

Actually, any empty <whatever this rule checks for> passed to iter I guess.

[naslundx]

Not a problem at all @dylwil3 , sorry for dropping this for a while.

Agree with @Avasam that collection seems like the more proper name considering it doesn't just apply to sequences.

I will polish up the PR.

[dylwil3]

Resolved in #16480