typing.Annotated should be considered a dangerous default

[lattwood]

Current problem

If a programmer accidentally uses typing.Annotated as a default value (not as a default type), pylint doesn't complain.

Ex, you're converting some FastAPI code from plain Depends to Annotated and make a typo, like below:

Original:

def some_method(
    session: AsyncSession = Depends(get_session_read),
) -> None:
    ...

Brainfart:

def some_method(
    session= typing.Annotated[AsyncSession, Depends(get_session_read)],
) -> None:
    ...

What you're actually supposed to do:

def some_method(
    session: typing.Annotated[AsyncSession, Depends(get_session_read)],
) -> None:
    ...

Desired solution

PyLint should make me feel bad about myself for making this typo. /s

[Pierre-Sassoulas]

Thank you for opening the issue. I'm not sure we can easily create a checker for this as it's possible to use typing classes as default value for function argument voluntarily (meta-programming or advanced use cases?).

I.e. something like this:

def some_meta_method(
    typing_type: Type[typing.Annotated] = typing.Annotated[AsyncSession, Depends(get_session_read)],
) -> None:
    ...

How would you deal with that ? Raise only if there's no defined type for typing_type ?