[ty] Use annotated parameters as type context #20635
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ibraheemdev
requested review from
AlexWaygood,
carljm,
dcreager and
sharkdp
as code owners
Contributor
Diagnostic diff on typing conformance testsChanges were detected when running ty on typing conformance tests--- old-output.txt 2025-10-03 20:48:40.319496259 +0000
+++ new-output.txt 2025-10-03 20:48:43.574517278 +0000
@@ -1,6 +1,6 @@
WARN ty is pre-release software and not ready for production use. Expect to encounter bugs, missing features, and fatal errors.
fatal[panic] Panicked at /home/runner/.cargo/git/checkouts/salsa-e6f3bb7c2a062968/29ab321/src/function/execute.rs:217:25 when checking `/home/runner/work/ruff/ruff/typing/conformance/tests/aliases_type_statement.py`: `PEP695TypeAliasType < 'db >::value_type_(Id(cc17)): execute: too many cycle iterations`
-fatal[panic] Panicked at /home/runner/.cargo/git/checkouts/salsa-e6f3bb7c2a062968/29ab321/src/function/execute.rs:217:25 when checking `/home/runner/work/ruff/ruff/typing/conformance/tests/aliases_typealiastype.py`: `infer_definition_types(Id(15c32)): execute: too many cycle iterations`
+fatal[panic] Panicked at /home/runner/.cargo/git/checkouts/salsa-e6f3bb7c2a062968/29ab321/src/function/execute.rs:217:25 when checking `/home/runner/work/ruff/ruff/typing/conformance/tests/aliases_typealiastype.py`: `infer_definition_types(Id(16432)): execute: too many cycle iterations`
_directives_deprecated_library.py:15:31: error[invalid-return-type] Function always implicitly returns `None`, which is not assignable to return type `int`
_directives_deprecated_library.py:30:26: error[invalid-return-type] Function always implicitly returns `None`, which is not assignable to return type `str`
_directives_deprecated_library.py:36:41: error[invalid-return-type] Function always implicitly returns `None`, which is not assignable to return type `Self@__add__` |
Contributor
|
ibraheemdev
force-pushed
the
ibraheem/argument-inference
branch
4 times, most recently
from
a90814a to
49661f2
Compare
Member
Author
|
It looks like a lot of these errors are duplicated diagnostics now that we infer argument expressions multiple times for callable unions/overloads, e.g., if flag:
def f(_: int): ...
else:
def f(_: int): ...
if flag:
x = 0
# warning[possibly-unresolved-reference] Name `x` used when possibly not defined
# warning[possibly-unresolved-reference] Name `x` used when possibly not defined
f(x)(In this example, multi-inference is not strictly necessary because the signatures are identical, but that's a separate performance concern). I'm not sure there's a great way around this other than adding checks around all diagnostics that aren't type-context specific? |
ibraheemdev
commented
Sep 29, 2025
Contributor
|
| Lint rule | Added | Removed | Changed |
|---|---|---|---|
missing-typed-dict-key |
162 | 0 | 0 |
unresolved-attribute |
2 | 109 | 0 |
type-assertion-failure |
30 | 3 | 0 |
invalid-argument-type |
4 | 0 | 9 |
invalid-context-manager |
0 | 13 | 0 |
non-subscriptable |
9 | 0 | 0 |
unsupported-operator |
2 | 2 | 2 |
possibly-missing-attribute |
3 | 0 | 1 |
redundant-cast |
3 | 1 | 0 |
invalid-assignment |
3 | 0 | 0 |
invalid-return-type |
2 | 0 | 0 |
possibly-missing-implicit-call |
1 | 0 | 0 |
unused-ignore-comment |
0 | 1 | 0 |
| Total | 221 | 129 | 12 |
Contributor
Instead of adding checks for every non-type-context-specific context, can we use state on |
ibraheemdev
force-pushed
the
ibraheem/argument-inference
branch
from
b68d4ca to
f9a4ae1
Compare
CodSpeed Instrumentation Performance ReportMerging #20635 will degrade performances by 17.12%Comparing Summary
Benchmarks breakdown
|
MichaReiser
reviewed
Oct 1, 2025
ibraheemdev
force-pushed
the
ibraheem/argument-inference
branch
2 times, most recently
from
d8b1f25 to
f5e0cb1
Compare
CodSpeed WallTime Performance ReportMerging #20635 will degrade performances by 21.13%Comparing Summary
Benchmarks breakdown
|
Contributor
|
Member
Author
|
I've decided to remove the idea of "lints that may be affected by bidirectional type context", as this seemed quite. Instead, the new implementation does the following for each argument:
This does mean we get worse diagnostics for I also went through the ecosystem report, and most diagnostic changes seem to be correct, or the result of an unrelated known limitation (namely |
ibraheemdev
force-pushed
the
ibraheem/argument-inference
branch
2 times, most recently
from
e8b8f6d to
ecd6491
Compare
ibraheemdev
commented
Oct 3, 2025
| match call_expression_tcx { | ||
| // A type variable is not a useful type-context for expression inference, and applying it | ||
| // to the return type can lead to confusing unions in nested generic calls. | ||
| Type::TypeVar(_) => {} |
Member
Author
There was a problem hiding this comment.
This is mostly for better diagnostics. Without it, reveal_type(identity(1)) is revealed as T@reveal_type | int. It also doesn't quite work for parameters annotated as T | None — we still might union the inferred type with T@f in that case.
ibraheemdev
force-pushed
the
ibraheem/argument-inference
branch
from
ecd6491 to
af00ebe
Compare
dcreager
approved these changes
Oct 3, 2025
|
|
||
| fn get_argument_node( | ||
| node: ast::AnyNodeRef<'_>, | ||
| argument_index: Option<usize>, |
Member
There was a problem hiding this comment.
nit: I think you can update this helper to take in argument_index: usize, and use argument_index? up above when you call the helper. Then you only need to match on node below.
Member
Author
There was a problem hiding this comment.
We also use get_argument_node in the diagnostic code with an Option<usize>, so I'm not sure the change is worth it.
Comment on lines
+4201
to
+4199
| let val_ty = | ||
| BoundTypeVarInstance::synthetic(db, "T", TypeVarVariance::Invariant); |
Member
There was a problem hiding this comment.
Were there tests that were failing with the old signature? I can see that the new signature means that there's now a return value, which will have the same type as the value being checked, but I don't see any tests that rely on that behavior.
Member
Author
There was a problem hiding this comment.
The issue that showed up in the ecosystem report was that the f(1) in assert_type(f(1), int) was inferred as int | Any, because the signature of assert_type was (Any, ..) -> ... I added a test.
ibraheemdev
force-pushed
the
ibraheem/argument-inference
branch
from
af00ebe to
6227fca
Compare
Member
Author
|
Deduplicating the parameter types helps the performance regression somewhat, but outside of that I think it's unavoidable. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Use the type annotation of function parameters as bidirectional type context when inferring the argument expression. For example, the following example now type-checks:
Part of astral-sh/ty#168.