Emit consider-using-tuple for in comparisons

[cdce8p]

Type of Changes

	Type
✓	✨ New feature

Description

Alternative to #4841
In contrast to #4768, with this PR consider-using-tuple will only be emitted for lists in in comparisons.

As discussed in #4776, this change will not have any impact on performance. CPython does this optimization when converting to bytecode.

# old
for x in [1, 2, 3]:
    ...

# new
for x in (1, 2, 3):
    ...

[coveralls]

Pull Request Test Coverage Report for Build 1152322179

• 9 of 9 (100.0%) changed or added relevant lines in 5 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.003%) to 92.743%

---

Totals
Change from base Build 1152045806:	0.003%
Covered Lines:	13458
Relevant Lines:	14511

---

💛 - Coveralls

[Pierre-Sassoulas]

I thought the point of #4776 was to replace that by use-set-for-membership. Sets are the go to for in membership so if we don't recommend set we should not recommend something else imo. I don't think the problem with hashable being required for set is that huge, use-set-for-membership is salvageable, I'm just delaying it in 2.11 so we can think about it and get more opinions.

[cdce8p]

The reason I opened this one is that IMO we just shouldn't move forward with use-set-for-membership. The minimal performance improvement (if any at all) just isn't worth it. Dev's will overlook that it's a set when they add new items and then wonder why it isn't working. A False-Positive isn't worth it, not for this one.

If we don't recommend sets, there is still some value in using tuples instead. As the docstring for the checker mentions: even only for style consistency.

[Pierre-Sassoulas]

I added more information in #4776 so the discussion can go on.

[cdce8p]

Closing this one, since we are going to move forward with #4841