Introduction of a generic schema used to type load and loads

[Vince-LD]

Hello,

This week, I've started to use marshmallow and marshmallow_dataclass at work for the first time and it's absolutely great. However, I was wondering if there was a reason not to use a generic schema to keep track of the wrapped dataclass when using class_schema factory.

The basic idea is pretty simple:

# have a custom schema class that is used only for type checking
class DataclassSchema(Generic[_T]):
    if TYPE_CHECKING:
        # use some overloading here to chose between the two result types
        def load(...) -> Union[_T, List[_T]]: ...

# update `class_schema` signature
def class_schema(clazz: Typ[_T], ...) -> Type[DataclassSchema[_T]]: ...

Then we can create a dataclass and the corresponding schema:

@dataclass
class Test:
    a : int
    b : str

# The types can actually be inferred by Pyright
schema: DataclassSchema[Test] = class_schema(Test)()
data: Test = schema.load({"a": 0, "b": "hellow world :)")

I have created a first implementation of this proposal in this fork. I am not familiar with mypy (I use Pyright through Pylance) and really advanced typing so I don't know if this is the best way to achieve my goal. If this subject interests you I'd be glad to contribute :)

Cheers,
Vincent

Edit: Sorry for the tag....

[ddorian]

Like https://github.com/daniil-berg/marshmallow-generic ?

[Vince-LD]

Yes, I'm proposing something similar. Actually, my first implementation works really well in real conditions (used it at work this week). I'm going to try another implentation by using marshmallow-generic but it would still require to modify the signature of marshmallow_dataclass.class_schema to use this generic Schema and a dataclass TypeVar instead of type and Schema.