Model exact narrowing with type(x) checks #20703
Conversation
test-data/unit/check-narrowing.test
Outdated
|
|
||
| T_value = TypeVar("T_value", A, B) | ||
|
|
||
| def f2(self, obj: T_value) -> T_value: |
There was a problem hiding this comment.
What is the difference between f1 and f2? Both are generic in a type variable with values.
There was a problem hiding this comment.
Nvm, I see you just fixed it.
There was a problem hiding this comment.
Oh sorry, I fixed that already :-)
More generally, the behaviour here isn't yet perfect.
I opened this PR to fix the testTypeEqualsCheckUsingIsNonOverlappingChild-xfail case, discovered we had an open issue it affected sterliakov/mypy-issues#259 , so adding the test case for it.
The test case I added highlights two preexisting issues. Ideally a) the narrowing works in the unconstrained case, b) we issue an error in both branches in the bad constrained case
| [builtins fixtures/isinstance.pyi] | ||
|
|
||
| [case testTypeEqualsCheckUsingIsNonOverlappingChild-xfail] | ||
| [case testTypeEqualsCheckUsingIsNonOverlappingChild] |
There was a problem hiding this comment.
This is the test case I was targeting with this PR
This comment has been minimized.
This comment has been minimized.
|
static-frame primer has a wrong type annotation and they already have a type-ignore for it |
|
Diff from mypy_primer, showing the effect of this PR on open source code: static-frame (https://github.com/static-frame/static-frame)
+ static_frame/core/archive_npy.py:232: error: Statement is unreachable [unreachable]
|
2 participants
I have some changes planned to improve final support, which will pair nicely with this
Fixes biggest thing in #10302 (original issue in there is not an issue; fixes this comment #10302 (comment) )