Better support for Confluent's kafka AvroSerializer/AvroDeserializer

[robbertvanwaveren]

Is your feature request related to a problem? Please describe.

The confluent AvroSerializer/AvroDeserializer classes require both a schema as well as an object to_dict and from_dict to object function. Using the following code I can successfully use the avro_schema, asdict and parse_obj functions from the generated code from this library to fill those requirements:

    avro_serializer = AvroSerializer(registry_client, msg.avro_schema(), lambda obj, sc: obj.asdict(standardize_factory=standardize_custom_type))
    avro_deserializer = AvroDeserializer(registry_client, msg.avro_schema(), lambda dict, sc: Message.parse_obj(dict))

However its still a bit clunky as the functions do not exactly match. Espescially the current implementation of the to_dict is close but not exactly usable as it does not fill the standardize_factory as above.

Note that successful integration is key to usage of these models in combination with an enterprise setting that relies of a schema-registry and confluents avro encoding header to support schema evolution across different implementation languages.

Describe the solution you'd like

I'd like the generated classes to come with class-level methods matching the provided lambdas, or if that is not feasible at least include the standardize_factory=standardize_custom_type) in the call to asdict from the object's to_dict by default.

Describe alternatives you've considered

Additional context

[marcosschroh]

@robbertvanwaveren

Note that successful integration is key to usage of these models in combination with an enterprise setting that relies of a schema-registry and confluents avro encoding header to support schema evolution across different implementation languages.

I agree but in this case the confluent python ecosystem is not stable at all so I would not go to far in the integration. However, I think we can include standardize_factory: Optional[Callable[..., Any] = None in the to_dict method . PR is welcome.

PS: Documentation and examples would be also nice to have.

[marcosschroh]

Another solution for this could be to have a mixings.Confluent that could add extra functionality for this use case