Add deserialize context to __post_deserialize__

[kdkavanagh-jt]

I have a yaml file containing some custom tags like:

sub_config: !include /path/to/my/config

I've updated my yaml.SafeLoader to properly handle those tags, (resolving them to a dictionary in this case) before handing the data off to mashumaro. Those resolvers are able to set an attribute on data["sub_config"] to indicate that it was loaded from an !include tag. I can then detect this in __pre_deserialize__ by inspecting the dict, but I dont currently have a way to set some hidden attr on the resolved object, since I dont have access to that data: dict in __post_deserialize__.

I'd like to carry that metadata thru to the dataclass such that when mashumaro serializes the type back to a yaml file, I can intercept data: dict in __post_deserialize__ and reattach that metadata for my yaml representer to convert back to !include ...

In other words, something like this would do the trick:

@dataclass
class ConfigItem(DataClassYAMLMixin):

  def __post_deserialize(cls, obj: T, data: dict):
           for k,v in data.items():
            if hasattr(v, YAML_INCLUDE_FROM_SOURCE):
               # Copy the attr over from the dict into the final object
               setattr(getattr(obj, k), YAML_INCLUDE_FROM_SOURCE, getattr(v, YAML_INCLUDE_FROM_SOURCE))

A workaround is to do something like this, but this is undesirable as it introduces a new "proper" attribute on the dataclass, making inheritance messy, etc:

@dataclass
class ConfigItem(DataClassYAMLMixin):
    __include_keys__: list[tuple[str, pathlib.Path]]

    @classmethod
    def __pre_deserialize__(cls, data: dict) -> dict: # pylint: disable=arguments-renamed
        inc_keys = data.setdefault("__include_keys__", [])
        for k,v in data.items():
            if hasattr(v, YAML_INCLUDE_FROM_SOURCE):
                inc_keys.append((k, getattr(v, YAML_INCLUDE_FROM_SOURCE)))
        return data

Is there a better way to do this already?