Is Pyrefly supposed to check Pytest fixture types?

[gpollo]

I have this simple dumb test.

import pytest

@pytest.fixture
def my_fixture() -> str:
    return "hello"

def test_my_test(my_fixture: int) -> None:
    print(my_fixture)

In Python, this is 100% valid typing, but not with the Pytest dependency injection mecanism. The fixture my_fixture should be annotated as str.

I know Pyrefly has built-in support for Pytest, but does that include checking fixture type errors? In my case, it is not raising any error on this code.

It's not clear from the documentation, so I don't want to raise an issue right away. Thanks!

[sueun-dev]

ran your snippet through pyrefly 1.2.0, mypy 1.19.1 and pyright: all three stay silent, while a plain str-into-int control errors in all three. pytest resolves fixtures by name at runtime, so there's no call site to compare the annotation against. #3776 tracks adding that inference.

side note, without a pyrefly.toml the basic preset gave me 0 errors even on the control, so pyrefly init first.

[gpollo]

Yeah, I know other tools don't check that (I've tried them too). As I said, the typing is "correct", but not when it's used with Pytest.

I'm just searching for a way to check these types, because I work in a codebase where people either don't type their fixtures or just put the wrong types lol

Anyway thanks, the ticket seems to be tracking that feature at least.

[yangdanny97]

I don’t think there's special type checking for pytest yet, iirc the pytest support we currently have is mostly around code navigation, but we are not opposed to adding this support