[ty] Bind bare generic property receivers in protocol checks

[charliermarsh]

Summary

#25332 checks protocol property getter types covariantly, but it resolves a property accessor to its return type before binding an explicitly generic receiver. For a getter like as def value(self: T) -> T, the unresolved T can make an incompatible property appear to satisfy the protocol.

This binds a bare, unbounded accessor-local receiver type variable to the concrete implementing instance before comparing the getter result:

from typing import Protocol, TypeVar

T = TypeVar("T")

class HasValue(Protocol):
    @property
    def value(self: T) -> T: ...

class Good:
    @property
    def value(self) -> "Good":
        return self

class Bad:
    @property
    def value(self) -> int:
        return 1

good: HasValue = Good()
bad: HasValue = Bad()  # invalid

This is intentionally limited to a single non-overloaded getter whose receiver annotation is exactly one unbounded, unconstrained type variable bound by that accessor. Bounded or constrained type variables, nested receiver annotations such as Box[T], overloaded getters, generic setters, and protocol-to-protocol generic receiver relations remain unsupported and are rejected conservatively, so valid programs using those forms may still be rejected. Broader support should reuse a receiver-aware callable abstraction rather than adding a second generic solver inside protocol checking; the lazy receiver binding discussed in astral-sh/ty#1172 is the likely shared foundation.

[astral-sh-bot]

Typing conformance results

The percentage of diagnostics emitted that were expected errors increased from 94.44% to 94.52%. The percentage of expected errors that received a diagnostic decreased from 90.21% to 90.03%. The number of fully passing files held steady at 95/134.

Summary

How are test cases classified?

Each test case represents one expected error annotation or a group of annotations sharing a tag. Counts are per test case, not per diagnostic — multiple diagnostics on the same line count as one. Required annotations (E) are true positives when ty flags the expected location and false negatives when it does not. Optional annotations (E?) are true positives when flagged but true negatives (not false negatives) when not. Tagged annotations (E[tag]) require ty to flag exactly one of the tagged lines; tagged multi-annotations (E[tag+]) allow any number up to the tag count. Flagging unexpected locations counts as a false positive.

Metric	Old	New	Diff	Outcome
True Positives	968	966	-2	⏬ (❌)
False Positives	57	56	-1	⏬ (✅)
False Negatives	105	107	+2	⏫ (❌)
Total Diagnostics	1073	1070	-3	⏬
Precision	94.44%	94.52%	+0.08%	⏫ (✅)
Recall	90.21%	90.03%	-0.19%	⏬ (❌)
Passing Files	95/134	95/134	+0

Test file breakdown

1 file altered

File	True Positives	False Positives	False Negatives	Status
protocols_generic.py	7 (-2) ❌	0 (-1) ✅	2 (+2) ❌	📉 Regressing
Total (all files)	966 (-2) ❌	56 (-1) ✅	107 (+2) ❌	95/134

False positives removed (1)

1 diagnostic

Test case	Diff
protocols_generic.py:144	-error[invalid-assignment] Object of type `ConcreteHasProperty1` is not assignable to `HasPropertyProto`

True positives removed (2)

2 diagnostics

Test case	Diff
protocols_generic.py:145	-error[invalid-assignment] Object of type `ConcreteHasProperty2` is not assignable to `HasPropertyProto`
protocols_generic.py:147	-error[invalid-assignment] Object of type `ConcreteHasProperty4` is not assignable to `HasPropertyProto`

[astral-sh-bot]

Memory usage report

Memory usage unchanged ✅

[astral-sh-bot]

ecosystem-analyzer results

No diagnostic changes detected ✅

Flaky changes detected. This PR summary excludes flaky changes; see the HTML report for details.

Full report with detailed diff (timing results)

[codspeed-hq]

Merging this PR will degrade performance by 18.18%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 1 improved benchmark
❌ 2 regressed benchmarks
✅ 72 untouched benchmarks
⏩ 64 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	ty_micro[many_protocol_members_mismatch]	364.3 ms	801.5 ms	-54.55%
❌	WallTime	colour_science	43.3 s	45.9 s	-5.69%
⚡	Simulation	DateType	258 ms	201.9 ms	+27.78%

Tip

Investigate this regression with the CodSpeed MCP and your agent.

---

_{Comparing charlie/protocol-property-generic-self (5399dc5) with main (f82a36b)²}

Footnotes

1. 64 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. ↩

2. No successful run was found on charlie/protocol-property-check-2 (8f86bcf) during the generation of this report, so main (f82a36b) was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩