A new approach to fastapi/dataclass interfaces

[msullivan]

We want to be able to support transforming types based on
dataclasses/attrs/pydantic style field descriptors. In order to do
that, we need to be able to consume things like calls to Field.

Our strategy for this is to introduce a new type
InitField[KwargDict] that collects arguments defined by a
KwargDict: TypedDict:

class InitField[KwargDict: BaseTypedDict]:
    def __init__(self, **kwargs: typing.Unpack[KwargDict]) -> None:
        ...

    def _get_kwargs(self) -> KwargDict:
        ...

When InitField or (more likely) a subtype of it is instantiated
inside a class body, we infer a more specific type for it, based on
Literal types for all the arguments passed.

So if we write:

class A:
foo: int = InitField(default=0)

then we would infer the type
InitField[TypedDict('...', {'default': Literal[0]})]
for the initializer, and that would be made available
as the Init field of the Member.

Honestly this is pretty subtle and will probably be controversial, but
maybe less controversial than the Annotated stuff and it will
produce a better result than that, so...

Take a look at test_fastapilike_2.py for the full fastapi + __init__
generation example using this system.

[1st1]

Beautifully complicated, but I love test_fastapilike_2. Feels we're onto something here.

[1st1]

Honestly this is pretty subtle and will probably be controversial, but
maybe less controversial than the Annotated stuff and it will
produce a better result than that, so...

I still want us to push through the Annotated stuff 🤦‍♂️