[Python] Generated type annotations do not pass mypy strict

[alexprengere]

Hi! Here is a request to improve pyi Python type annotations generated.
Using the latest protoc version, and assuming this proto file:

syntax = "proto3";

message X {
  string s = 1;
}

message M {
  repeated X xs = 1;
}

After generating the code and pyi file:

$ protoc repro.proto --python_out=. --pyi_out=.

Here is the generated pyi file:

from google.protobuf.internal import containers as _containers
from google.protobuf import descriptor as _descriptor
from google.protobuf import message as _message
from collections.abc import Iterable as _Iterable, Mapping as _Mapping
from typing import ClassVar as _ClassVar, Optional as _Optional, Union as _Union

DESCRIPTOR: _descriptor.FileDescriptor

class X(_message.Message):
    __slots__ = ("s",)
    S_FIELD_NUMBER: _ClassVar[int]
    s: str
    def __init__(self, s: _Optional[str] = ...) -> None: ...

class M(_message.Message):
    __slots__ = ("xs",)
    XS_FIELD_NUMBER: _ClassVar[int]
    xs: _containers.RepeatedCompositeFieldContainer[X]
    def __init__(self, xs: _Optional[_Iterable[_Union[X, _Mapping]]] = ...) -> None: ...

mypy (the standard Python type checker) raises a type-arg issue in strict mode:

# python -m mypy repro_pb2.pyi  --strict, needs types-protobuf
repro_pb2.pyi:19: error: Missing type arguments for generic type "Mapping"  [type-arg]

In this example, the _Mapping is indeed lacking type arguments, and could be typed as Mapping[str, str].
AFAICT this is the only issue I encountered of such type. Fixing it could help the adoption of static typing of projects relying on generated protobuf code.

[JasonLunn]

We would accept a pull request to add typing as Mapping[Any, Any]

[alexprengere]

The pull requested has been opened, I signed the CLA.
I tested it myself locally on the example above, and it fixes the issue.
I am not super familiar with the codebase here, so this was helped by an AI agent (some projects prefer contributions to disclose this).