Discriminator and recursive fields

[George Sakkis]

I'm trying to define a recursive type with a discriminator as follows:

from __future__ import annotations

from dataclasses import dataclass, field
from typing import Annotated

import apischema


@dataclass
class FileApi:
    name: str


@dataclass
class DirectoryApi:
    name: str
    contents: list[FileApi | DirectoryApi] = field(default_factory=list)


NodeApi = Annotated[
    DirectoryApi | FileApi ,
    apischema.discriminator("type", {"file": FileApi, "directory": DirectoryApi})
]

print(apischema.deserialize(NodeApi, {"type": "file", "name": "foo.txt"}))
print(apischema.deserialize(NodeApi, {"type": "directory", "name": "images"}))

The first print succeeds (FileApi(name='foo.txt')) but the second fails with
apischema.validation.errors.ValidationError: ValidationError: [{'loc': ['type'], 'err': 'unexpected property'}]

What's the right way to make this work?

[Joseph Perez]

Hi, thank you for opening this issue.
I've started to investigate, and I've found that if you remove the fields contents: list[FileApi | DirectoryApi] , it works. So this is definitely a bug, and I will investigate further.

[Joseph Perez]

I've isolated the bug and fixed it. I will push the PR soon.