[Python] How does `pytest.mark.parametrize` infer a Literal type for a function argument?

I'm using pytest.mark.parametrize in my test function, and I noticed something interesting when hovering over the input parameter in VSCode. The inferred type is:

input: Literal['bar', 'baz', 'BAR'] | None


Here is my test code:

@pytest.mark.parametrize(
"input,status_code,expected",
[
("bar", status.HTTP_200_OK, "bar"),
("baz", status.HTTP_200_OK, "baz"),
(None, status.HTTP_422_UNPROCESSABLE_ENTITY, None),
("BAR", status.HTTP_200_OK, "BAR"),
],
)
def test_path(app, input, status_code, expected):
@app.get("/items/{item_id}")
def read_item(item_id: Annotated[str, Path()]):
assert item_id == expected

res = read_item({"pathParameters": {"item_id": input}}, {})
assert res["statusCode"] == status_code


It seems like pytest.mark.parametrize is influencing the inferred type of input, making it a Literal type based on the provided test cases.

My questions:

1. How does pytest.mark.parametrize affect the type inference of function parameters?
2. How to add type hints for function arguments by only using decorators in Python

I couldn't find documentation explicitly mentioning this behavior, so I'd appreciate any insights or references.



Continue reading...