Provide TypedDict types for json mapping of protobuf

[dilipvamsi]

Hi,
It would be great if json_format definitions are also created while compiling the proto file.

Thanks,
Dilip Vamsi.

[nipunn1313]

Hi - looks like it is in typeshed https://github.com/python/typeshed/blob/master/third_party/2and3/google/protobuf/json_format.pyi

Is there anything more specific you'd like to have in the autogenerated pyi files?
Sorry - I am not familiar with json_format and would appreciate some more details to bring me up to speed.

[nipunn1313]

based on my readthrough, the typing in typeshed covers the json_format. It's not clear to me what autogenerated .pyi definitions would bring to the table. PRs are welcome!

[dilipvamsi]

Hi,
Sorry, what I meant was json mapping https://developers.google.com/protocol-buffers/docs/proto3#json_mapping not json format.
If this is present in pyi file. It is will help in building a dict in accordance with the protobuf.
Using this I need not create an protobuf object and then convert to dict.

[nipunn1313]

What annotation are you looking for? Can you provide some example code which is under-annotated today?

It looks like these json_mapping methods are annotated in typeshed. Is there anything more specific you'd like to have in the autogenerated pyi files? https://github.com/python/typeshed/blob/master/third_party/2and3/google/protobuf/json_format.pyi

[dilipvamsi]

I presenting an example for both proto2 and proto3 implementations.
As converting to json has 2 versions one with preserving the names and other is converting them to the json names.

For proto 2

message Hello {
  required string name = 1;
  required google.protobuf.Timestamp timestamp = 2;
}
message Sample {
  optional string string = 1;
  optional int64 large_int = 2;
  optional Hello hello = 3;
  required int32 small_int = 4;
  repeated int32 numbers = 5; 
}

I want a auto generated typed classes be

class Hello___JSON_PRESERVING(typing.TypedDict):
    name: str
    timestamp: str

class Sample___JSON_PRESERVING(typing.TypedDict):
     string: typing.Optional[str]
     large_int: typing.Optional[str]
     hello: typing.Optional[Hello___JSON_PRESERVING]
     small_int: int
     numbers: typing.Optional[typing.List[int]]

class Hello___JSON(typing.TypedDict):
    name: str
    timestamp: str

class Sample___JSON(typing.TypedDict):
     string: typing.Optional[str]
     largeInt: typing.Optional[str]
     hello: typing.Optional[Hello___JSON]
     smallInt: int
     numbers: typing.Optional[typing.List[int]]

For proto 3

message Hello {
  string name = 1;
  google.protobuf.Timestamp timestamp = 2;
}
message Sample {
  string string = 1;
  int64 large_int = 2;
  Hello hello = 3;
  int32 small_int = 4;
  int32 numbers = 5; 
}

I want a auto generated typed classes be

class Hello___JSON_PRESERVING(typing.TypedDict):
    name: typing.Optional[str]
    timestamp: typing.Optional[str]

class Sample___JSON_PRESERVING(typing.TypedDict):
     string: typing.Optional[str]
     large_int: typing.Optional[str]
     hello: typing.Optional[Hello___JSON_PRESERVING]
     small_int: typing.Optional[int]
     numbers: typing.Optional[typing.List[int]]

class Hello___JSON(typing.TypedDict):
    name: typing.Optional[str]
    timestamp: typing.Optional[str]

class Sample___JSON(typing.TypedDict):
     string: typing.Optional[str]
     largeInt: typing.Optional[str]
     hello: typing.Optional[Hello___JSON]
     smallInt: typing.Optional[int]
     numbers: typing.Optional[typing.List[int]]

[nipunn1313]

Yep - this seems reasonable. I would be amenable to autogenerating these.

They could be generated under the namespace of the original message - making importing easier - as Hello would be generated in both the _pb2.py and the _pb2.pyi

For example

class Hello(typing.TypedDict):
    class __JSONFormatDict(typing.TypedDict):
        [fields]