自定义序列化

Dubbo Python 自定义序列化

在此查看完整示例。

Python 是一种动态语言，其灵活性使得在其他语言中设计通用的序列化层具有挑战性。因此，我们移除了框架级别的序列化层，而是提供接口，允许用户根据需求自行实现（因为用户更了解他们要传输的数据格式）。

序列化通常分为两个部分：序列化和反序列化。我们为这些函数定义了类型，自定义的序列化/反序列化函数必须遵循这些 “格式”。

对于序列化函数，我们规定：

  1. # A function that takes any number of arguments and returns data of type bytes
  2. SerializingFunction = Callable[..., bytes]

对于反序列化函数，我们规定：

  1. # A function that takes an argument of type bytes and returns data of any type
  2. DeserializingFunction = Callable[[bytes], Any]

下面，我将演示如何使用 protobuf 和 json 进行序列化。

protobuf

  1. 有关定义和编译 protobuf 文件的详细说明，请参阅 protobuf tutorial。

  2. 在Client的unary、client_stream、server_stream 和 bi_stream 方法中通过设置request_serializer 和 response_deserializer 设置对应的序列化和反序列化函数。

    1. class GreeterServiceStub:
    2.     def __init__(self, client: dubbo.Client):
    3.         self.unary = client.unary(
    4.             method_name="sayHello",
    5.             request_serializer=greeter_pb2.GreeterRequest.SerializeToString,
    6.             response_deserializer=greeter_pb2.GreeterReply.FromString,
    7.         )
    9.     def say_hello(self, request):
    10.         return self.unary(request)
    12. if __name__ == "__main__":
    13.     reference_config = ReferenceConfig.from_url(
    14.         "tri://127.0.0.1:50051/org.apache.dubbo.samples.data.Greeter"
    15.     )
    16.     dubbo_client = dubbo.Client(reference_config)
    18.     stub = GreeterServiceStub(dubbo_client)
    19.     result = stub.say_hello(greeter_pb2.GreeterRequest(name="Dubbo-python"))
    20.     print(f"Received reply: {result.message}")

  3. 同理在Server中设置对应的序列化和反序列化函数。

    1. class GreeterServiceServicer:
    2.     def say_hello(self, request):
    3.         print(f"Received request: {request}")
    4.         return greeter_pb2.GreeterReply(message=f"Hello, {request.name}")
    6. def build_service_handler():
    7.     # build a method handler
    8.     method_handler = RpcMethodHandler.unary(
    9.         GreeterServiceServicer().say_hello,
    10.         method_name="sayHello",
    11.         request_deserializer=greeter_pb2.GreeterRequest.FromString,
    12.         response_serializer=greeter_pb2.GreeterReply.SerializeToString,
    13.     )
    14.     # build a service handler
    15.     service_handler = RpcServiceHandler(
    16.         service_name="org.apache.dubbo.samples.data.Greeter",
    17.         method_handlers=[method_handler],
    18.     )
    19.     return service_handler
    21. if __name__ == "__main__":
    22.     # build a service handler
    23.     service_handler = build_service_handler()
    24.     service_config = ServiceConfig(
    25.         service_handler=service_handler, host="127.0.0.1", port=50051
    26.     )
    28.     # start the server
    29.     server = dubbo.Server(service_config).start()
    31.     input("Press Enter to stop the server...\n")

Json

我们已经使用 protobuf 实现了单参数的序列化和反序列化。现在，我将演示如何编写一个支持多参数的 Json 序列化和反序列化函数，从而实现多参数方法的远程调用。

  1. 安装orjson

    1. pip install orjson

  2. 定义并设置Client的序列化和反序列化函数

    1. def request_serializer(name: str, age: int) -> bytes:
    2.     return orjson.dumps({"name": name, "age": age})
    4. def response_deserializer(data: bytes) -> str:
    5.     json_dict = orjson.loads(data)
    6.     return json_dict["message"]
    8. class GreeterServiceStub:
    9.     def __init__(self, client: dubbo.Client):
    10.         self.unary = client.unary(
    11.             method_name="unary",
    12.             request_serializer=request_serializer,
    13.             response_deserializer=response_deserializer,
    14.         )
    16.     def say_hello(self, name: str, age: int):
    17.         return self.unary(name, age)
    19. if __name__ == "__main__":
    20.     reference_config = ReferenceConfig.from_url(
    21.         "tri://127.0.0.1:50051/org.apache.dubbo.samples.serialization.json"
    22.     )
    23.     dubbo_client = dubbo.Client(reference_config)
    25.     stub = GreeterServiceStub(dubbo_client)
    26.     result = stub.say_hello("dubbo-python", 18)
    27.     print(result)

  3. 定义并设置Server的序列化和反序列化函数

    1. def request_deserializer(data: bytes) -> Tuple[str, int]:
    2.     json_dict = orjson.loads(data)
    3.     return json_dict["name"], json_dict["age"]
    5. def response_serializer(message: str) -> bytes:
    6.     return orjson.dumps({"message": message})
    8. class GreeterServiceServicer:
    9.     def say_hello(self, request):
    10.         name, age = request
    11.         print(f"Received request: {name}, {age}")
    12.         return f"Hello, {name}, you are {age} years old."
    14. def build_service_handler():
    15.     # build a method handler
    16.     method_handler = RpcMethodHandler.unary(
    17.         GreeterServiceServicer().say_hello,
    18.         method_name="unary",
    19.         request_deserializer=request_deserializer,
    20.         response_serializer=response_serializer,
    21.     )
    22.     # build a service handler
    23.     service_handler = RpcServiceHandler(
    24.         service_name="org.apache.dubbo.samples.serialization.json",
    25.         method_handlers=[method_handler],
    26.     )
    27.     return service_handler
    29. if __name__ == "__main__":
    30.     # build server config
    31.     service_handler = build_service_handler()
    32.     service_config = ServiceConfig(
    33.         service_handler=service_handler, host="127.0.0.1", port=50051
    34.     )
    36.     # start the server
    37.     server = dubbo.Server(service_config).start()
    39.     input("Press Enter to stop the server...\n")

最后修改 November 7, 2024: docs: add python sdk manual (#3056) (26c58b388ff)