Discriminative Union

[julien-blanchon]

I'm using a lot the following discriminative union pattern:

from typing import Annotated, Literal
from pydantic import Field
from simple_parsing import ArgumentParser

@dataclass
class DatasetImage:
    type: Literal["image"]
    path: str
    resolution: tuple[int, int]


@dataclass
class DatasetVideo:
    type: Literal["video"]
    path: str
    resolution: tuple[int, int]
    fps: int


@dataclass
class Config:
    dataset: Annotated[DatasetImage | DatasetVideo, Field(discriminator="type")]


parser = ArgumentParser()

if __name__ == "__main__":
    parser.add_arguments(Config, "example")
    args = parser.parse_args()
    print(args)

It's working with cyclopts (but still need pydantic for the discriminator field).

With simple-parsing I get this

❯ python main.py --help 
usage: main.py [-h] --dataset Annotated

options:
  -h, --help           show this help message and exit

Config ['example']:
  Config(dataset: Annotated[__main__.DatasetImage | __main__.DatasetVideo, FieldInfo(annotation=NoneType, required=True, discriminator='type')])

  --dataset Annotated

[lebrice]

Hey there @julien-blanchon , thanks for posting!

Hmm, I'm not sure I understand, what would you like the behaviour to be in this case? If you want to dispatch the type of dataclass based on a field, you can use the subgroups feature. Have you already tried that?

Otherwise, I'm all for more pydantic compatibility stuff, I just don't have time (or a clear idea of what needs to be added) for this.

Let me know what you think! :)