Nested message produces different proto schema than the one from compiled .proto file

[nodarai]

The message built with proto-plus package produces slightly different proto schema. The difference is that ._default_package is added to the nested field's type_name.
It causes an error when trying to use that schema to write to GCP BigQuery's Storage API: 400 Invalid proto schema: BqMessage.proto: UserSchema.team: "._default_package.UserSchema.Team" is not defined

Environment details

• Programming language: Python
• OS: Manjaro with 5.19 kernel
• Language runtime version: 3.8.12
• Package version: 1.22.0

Steps to reproduce

1. Create the user.proto file

syntax = "proto2";

message UserSchema {
  required string username = 1;
  required string email = 2;

  message Team { required string name = 1; };

  required Team team = 3;
}

2. Compile it

protoc --python_out=. user.proto

3. Run the code

import asyncio

import proto
from google.cloud.bigquery_storage_v1beta2.types.protobuf import ProtoSchema
from google.protobuf.descriptor_pb2 import DescriptorProto

from . import user_pb2


class UserSchema(proto.Message):
    class Team(proto.Message):
        name = proto.Field(proto.STRING, number=1)

    username = proto.Field(proto.STRING, number=1)
    email = proto.Field(proto.STRING, number=2)
    team = proto.Field(Team, number=3)


async def main():
    proto_descriptor = DescriptorProto()
    UserSchema.pb().DESCRIPTOR.CopyToProto(proto_descriptor)
    proto_schema = ProtoSchema(proto_descriptor=proto_descriptor)

proto_descriptor2 = DescriptorProto()
user_pb2.UserSchema.DESCRIPTOR.CopyToProto(proto_descriptor2)
proto_schema2 = ProtoSchema(proto_descriptor=proto_descriptor2)

    print(proto_schema)
    print(proto_schema2)


if __name__ == "__main__":
    asyncio.run(main())

4. Compare the output's team field's type names

proto_descriptor {
  name: "UserSchema"
  field {
    name: "username"
    number: 1
    label: LABEL_OPTIONAL
    type: TYPE_STRING
  }
  field {
    name: "email"
    number: 2
    label: LABEL_OPTIONAL
    type: TYPE_STRING
  }
  field {
    name: "team"
    number: 3
    label: LABEL_OPTIONAL
    type: TYPE_MESSAGE
    type_name: "._default_package.UserSchema.Team"
  }
  nested_type {
    name: "Team"
    field {
      name: "name"
      number: 1
      label: LABEL_OPTIONAL
      type: TYPE_STRING
    }
    options {
    }
  }
  options {
  }
}
proto_descriptor {
  name: "UserSchema"
  field {
    name: "username"
    number: 1
    label: LABEL_REQUIRED
    type: TYPE_STRING
  }
  field {
    name: "email"
    number: 2
    label: LABEL_REQUIRED
    type: TYPE_STRING
  }
  field {
    name: "team"
    number: 3
    label: LABEL_REQUIRED
    type: TYPE_MESSAGE
    type_name: ".UserSchema.Team"
  }
  nested_type {
    name: "Team"
    field {
      name: "name"
      number: 1
      label: LABEL_REQUIRED
      type: TYPE_STRING
    }
  }
}

5. type_name: "._default_package.UserSchema.Team" from proto.Message class and type_name: ".UserSchema.Team" from compiled .proto file

[nodarai]

Just found this:
https://github.com/googleapis/proto-plus-python/blob/main/proto/_package_info.py#L40

    # TODO: Revert to empty string as a package value after protobuf fix.
    # When package is empty, upb based protobuf fails with an
    # "TypeError: Couldn't build proto file into descriptor pool: invalid name: empty part ()' means"
    # during an attempt to add to descriptor pool.

Which is probably the reason for the above error.

[nodarai]

Is the above mentioned issue being worked on?
Or is it at least in the roadmap?

[nodarai]

The cause issue was fixed in protobuf version 4.21.6, so I have created a PR #341 to apply the fix to this package.
Meanwhile, before the PR is merged and released, to make things work with protobuf=4.21.6, adding the next line to the above code will fix the issue:

__protobuf__ = proto.module(package="")