Allow serialization of dataclass objects from and to dict

[christianwbrock]

Allow serialization of dataclass objects from and to dict

Currently I know only two workarounds.
After explaining them I will ask my questions at the end ;-)

The first workarounds involves using json.loads and json.dumps:

def model_to_dict(model):
    json_serializer = JsonSerializer()
    json_string = json_serializer.render(model)
    return json.loads(json_string)

def dict_to_model(a_dict):
    json_string = json.dumps(a_dict)
    json_parser = JsonParser()
    return json_parser.parse(json_string, generated.Model)

This works but requires the extra step of creating a json string.

The second workaround exploits the fact that xsdata uses dicts internally.

For serialization this is kind of supported and documented:

def model_to_dict(model):
    result = {}

    def dict_dumper(obj, *_args, **_kwargs):
        result.update(obj)

    json_serializer = JsonSerializer(dump_factory=dict_dumper)
    json_serializer.write(None, model)

    return result

For parsing it exploits undocumented internals:

def dict_to_model(a_dict):
    class FakeSourceHavingReadAttribute:
        def read(self):
            pass

    def dict_loader(_src):
        return a_dict

    json_parser = JsonParser(load_factory=dict_loader)
    model = json_parser.parse(FakeSourceHavingReadAttribute(), generated.Model)

    return model

Two questions

1. Is there a better way?
2. Should we modify JsonParser to get rid of the FakeSourceHavingReadAttribute hack?

[tefra]

The parser exists but it's not documented, it needs some refactoring to follow the parser interface, the serializer is also planned.

They are on top of my priorities after I finish refactoring all the docstrings/documentation.

[tefra]

I surfaced the dict parser as the dict decoder and extracted the dict encoder from the json serializer