[aio types] Fix some grpc.aio python types

[Tasssadar]

With these, it is actually possible to have typed client stubs where the return type is correctly inferred.

It's only for the non-streaming calls, because there is RequestIterableType for the streaming ones (but it's just Any with extra steps and would require much more work).

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: Tasssadar / name: Vojtěch Boček (47531bf)

[XuanWang-Amos]

Hi, thanks for the PR, the change looks fine to me except for one concern about changing Awaitable to Generator since it's not backward compatible and might break some peoples integration. Can you share more context on the issue you're having now and how this PR can help address it?

[Tasssadar]

Hello, thanks for taking a look.

The Awaitable type for __await__ is wrong. As per mypy docs and PEP492, it's supposed to return a Generator, and it simply does not work (typecheck nor infer return type) with Awaitable.

Besides, if you look at the actual implementation of __await__, that's what the method is returning, the Awaitable was never correct: https://github.com/grpc/grpc/pull/32475/files#diff-986f9aff481adca09f2b4e49c2d5daa8971a6c5e113bdd96002681e208c0aa1fR268-R293

Now, as for what I'm trying to achieve: mypy-protobuf generates type stubs like this for gRPC services:

class FooServiceStub:
    def __init__(self, channel: grpc.Channel) -> None: ...

    GetFoo: grpc.UnaryUnaryMultiCallable[FooRequest,FooResponse]

It generates non-aio stubs, true, but if you put import grpc.aio as grpc on top, plus the changes in this PR, it actually typechecks and infers argument and return value types correctly.

[XuanWang-Amos]

Hi, sorry for late reply and thanks for the additional context.

We agree with the proposed change, but we would like to verify that it does not break any mypy or pytype checks before merging it. We will update this thread once we had a chance to verify that.

In the meantime, please let us know if the change has already been tested.

[Tasssadar]

Hi, thank you. I'm already using this in some projects, but they are not open-source so I can't provide any examples unfortunately.

[XuanWang-Amos]

Sorry about the delay. I just verified with both mypy and pytype, good news is we're now sure this change is good to merge, but bad news is we're not checking aio files in our pytype workflow, we might need fix that later.

I added a small comment, can you help address that and fix merge conflict?

Thanks again for the contribution!

[Tasssadar]

Hello, sorry for the long turnaround, but I added the Generic and rebased the PR.

[XuanWang-Amos]

Looks like some bazel tests are failing, you can check them locally using:
bazel test //src/python/grpcio_tests/tests/unit:all.
For sanity (format) related fixes, you can run:
./tools/distrib/sanitize.sh (Or ./tools/distrib/yapf_code.sh and ./tools/distrib/isort_code.sh etc manually).