Strict typed dict validator

[Wauplin]

This PR adds support for TypedDict validation using the @strict dataclass framework.

Related to huggingface/transformers#40793 and especially the TypedDictAdapter (defined here). Once this PR is merged/released, no need for this logic to be implemented in transformers.

Example:

from typing import Annotated, TypedDict
from huggingface_hub.dataclasses import validate_typed_dict

def positive_int(value: int):
    if not value >= 0:
        raise ValueError(f"Value must be positive, got {value}")

class User(TypedDict):
    name: str
    age: Annotated[int, positive_int]

# Valid data
validate_typed_dict(User, {"name": "John", "age": 30})

# Invalid type for age
validate_typed_dict(User, {"name": "John", "age": "30"})
# huggingface_hub.errors.StrictDataclassFieldValidationError: Validation error for field 'age':
#     TypeError: Field 'age' expected int, got str (value: '30')

# Invalid value for age
validate_typed_dict(User, {"name": "John", "age": -1})
# huggingface_hub.errors.StrictDataclassFieldValidationError: Validation error for field 'age':
#     ValueError: Value must be positive, got -1

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[zucchini-nlp]

Thanks a lot! Visually looks similar to the workaround I had so it should address the issue with processor type-validation

[zucchini-nlp]

nice!

[Wauplin]

Visually looks similar to the workaround I had so it should address the issue with processor type-validation

I definitely took the code from your PR 😄

[gante]

👌

[zucchini-nlp]

can we merge it pls, so I can merge the transformers PR?

[Wauplin]

It'll have to wait for @hanouticelina 's review so probably on Tuesday

[hanouticelina]

Thank you! I left two comments regarding the handling of optional/omittable fields in TypedDict, feel free to correct me if I got that wrong 😄

[Wauplin]

@hanouticelina thanks for your comments. I've addressed them in the 4 latest commits. Now we are handling correctly total=False, Required and NotRequired (on Python 3.11+ or typing_extensions) and optional value have no default (so will fail if None is not passed explicitly).

Can you re-review? 🙏

[hanouticelina]

very nice! thank you

[Wauplin]

pre-released as part of https://pypi.org/project/huggingface-hub/1.0.0rc3/