Restrict PYI034 for in-place operations to enclosing class

[charliermarsh]

Summary

In .py and .pyi files, we now only flag cases in which the return type is the enclosing class, like:

class A:
    def __iadd__(self) -> A:
        return self

As opposed to:

class A:
    def __iadd__(self) -> int:
        return self

Closes #24462.

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

ℹ️ ecosystem check detected linter changes. (+0 -2 violations, +0 -0 fixes in 1 projects; 55 projects unchanged)

python/typeshed (+0 -2 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --no-fix --output-format concise --no-preview --select E,F,FA,I,PYI,RUF,UP,W

- stdlib/typing_extensions.pyi:265:9: PYI034 `__ior__` methods in classes like `_TypedDict` usually return `self` at runtime
- stubs/boltons/boltons/dictutils.pyi:96:9: PYI034 `__ior__` methods in classes like `FrozenDict` usually return `self` at runtime

Changes by rule (1 rules affected)

code	total	+ violation	- violation	+ fix	- fix
PYI034	2	0	2	0	0

Linter (preview)

ℹ️ ecosystem check detected linter changes. (+0 -2 violations, +0 -0 fixes in 1 projects; 55 projects unchanged)

python/typeshed (+0 -2 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --no-fix --output-format concise --preview --select E,F,FA,I,PYI,RUF,UP,W

- stdlib/typing_extensions.pyi:265:9: PYI034 `__ior__` methods in classes like `_TypedDict` usually return `self` at runtime
- stubs/boltons/boltons/dictutils.pyi:96:9: PYI034 `__ior__` methods in classes like `FrozenDict` usually return `self` at runtime

Changes by rule (1 rules affected)

code	total	+ violation	- violation	+ fix	- fix
PYI034	2	0	2	0	0

[MichaReiser]

Do you think the special case is worth it? To me, defining __iadd__(self, other: A) -> int seems a pretty high signal that using Self here is not what the user wants.

[charliermarsh]

Hmm, not sure. I was following @AlexWaygood's guidance in #24462 (comment) where he said it did make sense to enforce this in stub files, though I also agree that (without that guidance) I would've made this change for both .py and .pyi files.

[AlexWaygood]

The least confusing thing for Ruff might be just to make the rule consistent between .py and .pyi files. If the rule only supported .pyi files, I think there probably would not be a strong motivation for any change. But if Ruff makes this change for both .py files and .pyi files, maybe it makes sense for us to make the same change upstream in flake8-pyi, so that Ruff and flake8-pyi continue to be consistent about how they enforce the rule.

[charliermarsh]

(I updated the PR to match that logic.)