@tiangolo this is green now. I believe it is technically a breaking change.

3 months and no review :(

Dang, I just checked out your PR and tried it for my test. It doesn't fix it because I think this code block still needs to strip out ctx and input from pydantic v2s ValidationError.errors() call:

            except ValidationError as exc:
                errors: List[ErrorDetails] = [
                    {**err, "loc": loc + err.get("loc", ())}  # type: ignore [typeddict-unknown-key]
                    for err in exc.errors(include_url=False)
                ]
                return None, errors

perhaps

for err in exc.errors(include_url=False, include_input=False, include_context=False)

See here in pydantic docs: https://docs.pydantic.dev/latest/api/pydantic_core/#pydantic_core.ValidationError.errors

Hi @tamird 👋

We'll be reviewing this PR soon. 🤓

Thank you for your interest and time in helping us with your contribution 🙇

@tamird thanks for the interest. There's no need to ping or contact individual maintainers, we already have this PR in the project and we'll review it eventually.

If you would like to help accelerate things, there are many ways to contribute, for example helping others with questions: https://fastapi.tiangolo.com/help-fastapi/#help-others-with-questions-in-github

But in VSCode I can still see that elements of errors are of type Any:

@tamird, could you please take a look?

I do not see the token "Any" in your screenshot.

I do not see the token "Any" in your screenshot.

You can see that errors is just Sequence instead of Sequence[ErrorDetails]. If I access the element as errors[0] it will be of type Any:

I do not see the token "Any" in your screenshot.

You can see that errors is just Sequence instead of Sequence[ErrorDetails]. If I access the element as errors[0] it will be of type Any:

I see. Rather than post a screenshot, perhaps you could add a type annotation in the relevant place? I'm not sure how your local environment is set up, and I'm not sure exactly what code you're looking at. A test change would be a robust way to ensure that the type annotations are working as intended.

I'm not sure how your local environment is set up, and I'm not sure exactly what code you're looking at.

I looked at the code from this comment.

And I use fresh virtual environment with the latest versions of packages installed using pip install -r requirements.txt.

I don't think this is a problem of my environment set up. If I create my own function with the same return type annotation, it works well

And if I try the same with:

from fastapi.exceptions import ValidationException

exc = ValidationException([])

b = exc.errors()[0]

It gives my Any

Full code example that you or anybody else can try:

from collections.abc import Sequence

from fastapi.exceptions import ValidationException
from pydantic_core import ErrorDetails


# =============================================
def my_errors() -> Sequence["ErrorDetails"]:
    pass

a = my_errors()[0]  # ErrorDetails


# =============================================
exc = ValidationException([])

b = exc.errors()[0]  # Any

Rather than post a screenshot, perhaps you could add a type annotation in the relevant place?

A test change would be a robust way to ensure that the type annotations are working as intended.

Sorry, I didn't get the above messages. Could you please clarify what you mean?

@YuriiMotov I have added type annotations in the relevant tests and confirmed locally that changing the annotations to Sequence[str] does cause mypy to complain (see updated commit message).

This still doesn't show me type hints in VSCode for the following code example with Pydantic V2:

from fastapi.exceptions import ValidationException

exc = ValidationException([])

b = exc.errors()[0]  # VSCode shows `(variable) b: Any`

At the same time it does show type correctly with Pydantic V1.

I think we should either make it working with Pydantic V2 or explain what is the purpose of this PR (what it gives us)

This still doesn't show me type hints in VSCode for the following code example with Pydantic V2:

from fastapi.exceptions import ValidationException

exc = ValidationException([])

b = exc.errors()[0]  # VSCode shows `(variable) b: Any`

At the same time it does show type correctly with Pydantic V1.

I think we should either make it working with Pydantic V2 or explain what is the purpose of this PR (what it gives us)

I'd prefer not to try to debug your VSCode setup. If you run mypy over the tests which have the type annotation showing that the type is Sequence[ErrorDetails] (e.g. tests/test_filter_pydantic_sub_model_pv2.py), does it pass?

I'd prefer not to try to debug your VSCode setup. If you run mypy over the tests which have the type annotation showing that the type is Sequence[ErrorDetails] (e.g. tests/test_filter_pydantic_sub_model_pv2.py), does it pass?

I think it's not a problem of my VSCode setup. It works perfectly fine in other cases.
Does it work in your IDE?
Could anybody else check it?

It looks like the if TYPE_CHECKING guards are interfering with the type checker, at least in the IDE.

Let me play with it some more.

This pull request has a merge conflict that needs to be resolved.