Name 'typing.io.IO' is not defined

[FichteFoll]

The Python 3.5 docs list a typing.io something ("wrapper namespace" according to docs, class according to interpreter), which you can access with IO to get an IO type, or for example TextIO. Using this in an interpreter works fine, but mypy complains about this.

>>> typing.io.IO[str]
typing.IO[str]
>>> typing.io.TextIO
typing.TextIO[str]

prettify.py:40: error: Name 'typing.io.IO' is not defined

I understand (now) that typing.IO is also possible, but the docs are very vague about that being a possibility, which is why I didn't even try it. Docs aside, since this works in the interpreter I think it should also work in mypy, which is why I'm reporting it.

[gvanrossum]

Right, it's meant as a namespace, it seems mypy doesn't understand this
yet. Bug.

[gvanrossum]

We added this to PEP 484 and typing.py in a relatively late stage, as a preferred way to separate support for specific modules from support for fundamentals (including containers). I guess mypy never implemented it. Maybe you can also submit a doc bug for Python 3.5?

[gvanrossum]

It's a passive-aggressive way of telling people that using typing.io.IO is deprecated. :-)

Seriously, we should add it, but I'd also like to deprecate it, at least in the docs.

[rohitkg98]

How do we go around adding type aliases inside the io module.
I tried doing something like this:

class io:

    __all__ = ['IO', 'TextIO', 'BinaryIO']

    IO = Union[IO]
    TextIO = Union[TextIO]
    BinaryIO = Union[BinaryIO]

but this introduces cyclic definition error in mypy

Also, the Union is used as workaround for this

[JelleZijlstra]

We should probably just give up on typing.io. It's never worked with mypy and I haven't seen a lot of demand for it.

[gvanrossum]

We should probably just give up on typing.io. It's never worked with mypy and I haven't seen a lot of demand for it.

Okay, let's close this then. Can someone file an issue with cpython to deprecate typing.io?

[rohitkg98]

Created the issue to deprecate: https://bugs.python.org/issue42001