fix Python interpreter gets error, using [int] with a function call while pyrefly doesn't get error #2590

[asukaminato0721]

Summary

Fixes #2590

blocked subscripting generic functions/callables so func[int] now emits UnsupportedOperation, and added a regression test.

Test Plan

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

mitmproxy (https://github.com/mitmproxy/mitmproxy)
- ERROR test/mitmproxy/proxy/layers/http/test_http3.py:89:41-66: Expected a type form, got instance of `(cls: type[bytes] = ...) -> _Placeholder[bytes] | bytes` [not-a-type]
- ERROR test/mitmproxy/proxy/layers/http/test_http3.py:91:35-67: `|` is not supported between `(cls: type[bytes] = ...) -> _Placeholder[bytes] | bytes` and `None` [unsupported-operation]
+ ERROR test/mitmproxy/proxy/layers/http/test_http3.py:89:41-66: `Placeholder` is not subscriptable [unsupported-operation]
+ ERROR test/mitmproxy/proxy/layers/http/test_http3.py:91:35-60: `Placeholder` is not subscriptable [unsupported-operation]

[github-actions]

Primer Diff Classification

✅ 1 improvement(s) | 1 project(s) total

1 improvement(s) across mitmproxy.

Project	Verdict	Changes	Error Kinds	Root Cause
mitmproxy	✅ Improvement	+2, -2	not-a-type, unsupported-operation	expr_untype()

Detailed analysis

✅ Improvement (1)

mitmproxy (+2, -2)

This is an improvement. The new errors correctly catch a real bug - attempting to subscript generic functions like tutils.Placeholder[bytes] when tutils.Placeholder is being treated as a function rather than a class. Looking at the code, lines 89 and 91 show tutils.Placeholder[bytes] being used in type annotations, but the old errors suggested pyrefly was incorrectly inferring Placeholder as a function type rather than recognizing it as a generic class. The new unsupported-operation errors correctly identify that if something is inferred as a function, it cannot be subscripted. The removed errors were false positives - tutils.Placeholder[bytes] is valid syntax when Placeholder is a generic class, and the union with None is also valid. The PR change improved type inference to properly distinguish between generic classes (subscriptable) and generic functions (not subscriptable).
Attribution: The change to expr_untype() in pyrefly/lib/alt/expr.rs added a check that only allows subscripting for TypeAlias foralls, blocking subscripting of generic functions/callables. This correctly identifies that tutils.Placeholder (a generic class) should be subscriptable, while generic functions should not be.

---

Was this helpful? React with 👍 or 👎

_{Classification by primer-classifier (1 LLM)}

[Copilot]

Pull request overview

Fixes a mismatch between CPython runtime behavior and pyrefly when subscripting generic functions/callables (e.g., func[int](...)), making pyrefly report an error and adding a regression test for issue #2590.

Changes:

• Block subscripting Type::Forall values unless the forall body is a TypeAlias, otherwise emit UnsupportedOperation (“X is not subscriptable”).
• Add a regression test asserting func[int](100) produces an error.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
pyrefly/lib/alt/expr.rs	Changes subscript inference to reject subscripting generic functions/callables (non-type-alias Forall).
pyrefly/lib/test/calls.rs	Adds a regression testcase covering func[int](100) erroring as “not subscriptable”.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

matches!(forall.body, ...) moves forall.body, so *forall / forall.body.name() in the branches will fail to compile due to a partial move. Use matches!(&forall.body, ...) (or pattern-match on &forall.body) and only move *forall when specializing.

[meta-codesync]

@yangdanny97 has imported this pull request. If you are a Meta employee, you can view this in D95062809.

[stroxler]

Review automatically exported from Phabricator review in Meta.