[ty] Eagerly evaluate `types.UnionType` elements as type expressions #21531

sharkdp · 2025-11-19T20:58:31Z

Summary

Eagerly evaluate the elements of a PEP 604 union in value position (e.g. IntOrStr = int | str) as type expressions and store the result (the corresponding Type::Union if all elements are valid type expressions, or the first encountered InvalidTypeExpressionError) on the UnionTypeInstance, such that the Type::Union(…) does not need to be recomputed every time the implicit type alias is used in a type annotation.

This might lead to performance improvements for large unions, but is also necessary for correctness, because the elements of the union might refer to type variables that need to be looked up in the scope of the type alias, not at the usage site.

Test Plan

New Markdown tests

sharkdp · 2025-11-19T21:00:03Z

crates/ty_python_semantic/src/types.rs

-                        builder = builder.add(if inferred_as.type_expression() {
-                            *element
-                        } else {
-                            element.in_type_expression(db, scope_id, typevar_binding_context)?


The previous version was also only able to surface the first error. I'm not sure if this is a problem, or something that we can live with.

sharkdp · 2025-11-19T21:03:54Z

crates/ty_python_semantic/src/types.rs

+                    UnionTypeInstance::PEP604(instance) => {
+                        // Cloning here is cheap if the result is a `Type` (which is `Copy`). It's more
+                        // expensive if there are errors.
+                        instance.union_type(db).clone()


@carljm In the end, this didn't require a new salsa query. Instead, we're now eagerly building that Type::Union in case there were no errors, and interning the result. And then we only need to copy the result here when we see a use of this implicit type alias in a type expression.

This means that we do some extra work if someone defines IntOrStr = int | str and then never uses it as a type. But that feels okay to me?

Yeah, that seems reasonable.

What's potentially more problematic is the fact that when we have a LargeUnion = C1 | C2 | … | Cn, we would build $n - 1$ union types eagerly, instead of just one. But if that turns out to be a problem, we can turn the union_type field into a salsa query.

crates/ty_python_semantic/src/types/narrow.rs

astral-sh-bot · 2025-11-19T21:08:43Z

Diagnostic diff on typing conformance tests

No changes detected when running ty on typing conformance tests ✅

astral-sh-bot · 2025-11-20T12:54:17Z

`mypy_primer` results

Changes were detected when running on open source projects

scikit-build-core (https://github.com/scikit-build/scikit-build-core)
+ src/scikit_build_core/build/wheel.py:98:20: error[no-matching-overload] No overload of bound method `__init__` matches arguments
- Found 43 diagnostics
+ Found 44 diagnostics

No memory usage changes detected ✅

crates/ty_python_semantic/resources/mdtest/narrow/isinstance.md

sharkdp · 2025-11-20T14:20:49Z

crates/ty_python_semantic/resources/mdtest/implicit_type_aliases.md

    reveal_type(callable_or_int)  # revealed: ((str, /) -> bytes) | int
    # TODO should be Unknown | int
-    reveal_type(type_var_or_int)  # revealed: T@_ | int
+    reveal_type(type_var_or_int)  # revealed: typing.TypeVar | int


We still don't support generic implicit type aliases, but this is a step in the right direction. T@_ is just wrong.

sharkdp · 2025-11-20T14:21:49Z

crates/ty_python_semantic/src/types/narrow.rs

-        // TODO: add support for PEP 604 union types on the right hand side of `isinstance`
-        // and `issubclass`, for example `isinstance(x, str | (int | float))`.


This was recently added

AlexWaygood

Nice!

crates/ty_python_semantic/resources/mdtest/narrow/isinstance.md

crates/ty_python_semantic/src/types.rs

crates/ty_python_semantic/resources/mdtest/narrow/isinstance.md

sharkdp · 2025-11-20T15:12:38Z

Looks like there are minor performance improvements

sharkdp added the ty Multi-file analysis & type inference label Nov 19, 2025

sharkdp commented Nov 19, 2025

View reviewed changes

crates/ty_python_semantic/src/types/narrow.rs Outdated Show resolved Hide resolved

sharkdp force-pushed the david/eager-in-type-expression branch from 41258a6 to d058e94 Compare November 19, 2025 21:06

carljm mentioned this pull request Nov 20, 2025

[ty] support PEP 613 type aliases #21394

Merged

[ty] Eagerly evaluate UnionType elements as type expressions

de8fda4

sharkdp force-pushed the david/eager-in-type-expression branch from d058e94 to 41ecb33 Compare November 20, 2025 12:50

sharkdp force-pushed the david/eager-in-type-expression branch from 41ecb33 to 3dc7c48 Compare November 20, 2025 14:14

AlexWaygood reviewed Nov 20, 2025

View reviewed changes

crates/ty_python_semantic/resources/mdtest/narrow/isinstance.md Outdated Show resolved Hide resolved

Even simpler representation

08a1f4e

sharkdp force-pushed the david/eager-in-type-expression branch from 3dc7c48 to 08a1f4e Compare November 20, 2025 14:19

sharkdp marked this pull request as ready for review November 20, 2025 14:20

sharkdp requested a review from dcreager as a code owner November 20, 2025 14:20

sharkdp commented Nov 20, 2025

View reviewed changes

AlexWaygood approved these changes Nov 20, 2025

View reviewed changes

crates/ty_python_semantic/resources/mdtest/narrow/isinstance.md Show resolved Hide resolved

crates/ty_python_semantic/src/types.rs Show resolved Hide resolved

crates/ty_python_semantic/resources/mdtest/narrow/isinstance.md Show resolved Hide resolved

Review findings

ab92379

sharkdp merged commit 0761ea4 into main Nov 20, 2025
41 checks passed

sharkdp deleted the david/eager-in-type-expression branch November 20, 2025 16:28

		// TODO: add support for PEP 604 union types on the right hand side of `isinstance`
		// and `issubclass`, for example `isinstance(x, str \| (int \| float))`.

[ty] Eagerly evaluate types.UnionType elements as type expressions #21531

[ty] Eagerly evaluate types.UnionType elements as type expressions #21531

Uh oh!

Conversation

sharkdp commented Nov 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test Plan

Uh oh!

sharkdp Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

sharkdp Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

carljm Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

sharkdp Nov 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

astral-sh-bot bot commented Nov 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Diagnostic diff on typing conformance tests

Uh oh!

astral-sh-bot bot commented Nov 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

mypy_primer results

Uh oh!

Uh oh!

sharkdp Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

sharkdp Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

AlexWaygood left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sharkdp commented Nov 20, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

[ty] Eagerly evaluate `types.UnionType` elements as type expressions #21531

[ty] Eagerly evaluate `types.UnionType` elements as type expressions #21531

sharkdp commented Nov 19, 2025 •

edited

Loading

sharkdp Nov 20, 2025 •

edited

Loading

astral-sh-bot bot commented Nov 19, 2025 •

edited

Loading

astral-sh-bot bot commented Nov 20, 2025 •

edited

Loading

`mypy_primer` results