[ty] Infer type of self as typing.Self in method body

[Glyphack]

Summary

Part of astral-sh/ty#159

Add support for adding a synthetic typing.Self type for self arguments in methods.
typing.Self is assigned as the type if there are no annotations.
This PR only adds the functionality when the symbol lookup is happening.

#18007 is handling the case when a call is happening.

Surfaced issues

After running tests and mypy primer some tests started failing because self was not Any anymore:

MDtests

• possibly unbound and unresolved attribute diagnostic when storing attributes ty#664
• Using super() in generic classes will fail if the self parameter type is set ty#697

Too many cycle iterations

• Too-many fix point iterations in growing union of literals ty#660

• too many cycle iterations in cyclic dependent attributes ty#692

• incrementing a value inside a tuple: solved by [ty] Infer type of self as typing.Self in method body #18473 (comment)

class ChunkedBytesIO:
    def __init__(self: Self, contents: list[bytes]) -> None:
        self.contents = contents
        self.pos = (0, 0)

    def read(self: Self):
        chunk, cursor = self.pos
        self.pos = (chunk + 1, cursor)

Test Plan

• Updated md tests.
• Added new TODOs because some tests started failing once the self was not unknown anymore.

[dcreager]

Hi @Glyphack, saw that you opened this up! Let me know if you want to find some time to chat through the options.

Edit: Tomorrow, that is, it's the end of the day for me 😄

[codspeed-hq]

CodSpeed Performance Report

Merging #18473 will degrade performances by 16.08%

_{Comparing Glyphack:typing-self-function-scope (a4d6a2a) with main (058fc37)¹}

Summary

❌ 6 regressions
✅ 42 untouched
⏩ 4 skipped²

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Mode	Benchmark	BASE	HEAD	Change
❌	Instrumentation	anyio	819.6 ms	915.2 ms	-10.44%
❌	WallTime	large[sympy]	38.4 s	41.9 s	-8.26%
❌	WallTime	small[altair]	2.3 s	2.4 s	-4.26%
❌	WallTime	small[freqtrade]	4.2 s	4.8 s	-12.64%
❌	WallTime	small[pydantic]	2.4 s	2.5 s	-7.4%
❌	WallTime	small[tanjun]	1.6 s	1.9 s	-16.08%

Footnotes

1. No successful run was found on david/type-of-self-in-methods-integration (ec9faa3) during the generation of this report, so main (058fc37) was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

2. 4 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[MichaReiser]

I "forked" this PR in #20812 and updated the salsa version. There are 4 new projects that now fail with too many iteration errors https://micha-typing-self-function-s.ecosystem-663.pages.dev/timing. I only checked setuptools, and the error looks legit. The type becomes massive! I think this should now be unblocked (once we figure out how to fix the non-convergent cases).

The perf impact is still pretty "bad" but I think that's expected given that we now check much more code https://codspeed.io/astral-sh/ruff/branches/micha%2Ftyping-self-function-scope

[sharkdp]

I will squash-merge this into #20922 for the same reasons that applied for the previous type-of-self PR. I'd like to clean up the looong history a bit (I will transfer comments that are still relevant), and I'd like to iterate faster on the ecosystem results, which are more convenient to use when it's not a contributor PR. @Glyphack thank you!