[MarcoGorelli]

Summary

The idea is:

• If we have a subscript expression, like x = my_list[0] where we know that my_list is a list[int], then x gets inferred as int. That's what

  pyrefly/pyrefly/lib/alt/expr.rs

  Lines 945 to 951 in 8003f9b

  	self.subscript_infer(
  	&base,
  	&x.slice,
  	x.range(),
  	type_form_context.unwrap_or(TypeFormContext::TypeExpression),
  	errors,
  	)

  does
• If however we have a subscript assignment like x = NotRequired[int], then NotRequired is a _SpecialForm which is typed to return object for __getitem__, so we need to infer x as object rather than reporting it as an invalid annotation
• For cases in class definitions, like class MyClass(TypedDict, total=False), however, we want to preserve the current behaviour and raise on something like extra_items=NotRequired[int], which is why in fn expr_class_keyword we need to pass down Some(TypeFormContext::ClassKeyword) so that if type_form_context.is_none() in expr_infer_without_shape_comment doesn't evaluate to true

Note that x: NotRequired[int] still gets checked as-before, as this change only covers the if type_form_context.is_none() case, and in the x: NotRequired[int] case, there is a type form context

Fixes #5429

• I am a human, and I fully understand the code I am submitting.

I used Claude to generate the code, but the description here is fully written by me

Test Plan

[meta-codesync]

This pull request has been imported. If you are a Meta employee, you can view this in D128312510. (Because this pull request was imported automatically, there will not be any future comments.)

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅