TypeAdapter should accept Annotated[T, ...]

[rijenkii]

Initial Checks

• I confirm that I'm using Pydantic V2

Description

See https://discuss.python.org/t/is-annotated-compatible-with-type-t/43898, microsoft/pyright#7091, https://github.com/python/typing/pull/1618/files.

Basically, I use TypeAdapter(Annotated[...]).validate_python in my custom validators quite ofter, but now after pydantic began to actually follow the PEPs, following examples no longer pass the type checker.

I have submitted this issue because Jelle Zijlstra explicitly said that this is a problem with pydantic types: https://discuss.python.org/t/is-annotated-compatible-with-type-t/43898/24

Example Code

from typing import Annotated, Any
import shortuuid
from uuid import UUID
from pydantic import PlainValidator, Field, TypeAdapter

def _shortuuid_validator(value: Any) -> UUID:
    value = TypeAdapter(
        Annotated[str, Field(min_length=22, max_length=22)]
    ).validate_python(value)
    return shortuuid.decode(value)


type ShortUUID = Annotated[UUID, PlainValidator(_shortuuid_validator)]

from typing import Annotated, Any
from pydantic import BeforeValidator, Field, TypeAdapter

class ArtifactObject(BaseModel):
    name: Annotated[str, Field(pattern=r"^[a-zA-Z0-9_\-]+$")]
    path: Annotated[str, Field(pattern=r"^[a-zA-Z0-9_\-/\\]+$")]


type ArtifactLongStr = Annotated[
    ArtifactObject,
    BeforeValidator(
        lambda x: (lambda y: {"name": y[0], "path": y[1]})(
            TypeAdapter(
                Annotated[str, Field(pattern=r"^[a-zA-Z0-9_\-]+:[a-zA-Z0-9_\-/\\]+$")],
            )
            .validate_python(x)
            .split(":"),
        ),
    ),
]

type ArtifactShortStr = Annotated[
    ArtifactObject,
    BeforeValidator(
        lambda x: (lambda y: {"name": y, "path": y})(
            TypeAdapter(
                Annotated[str, Field(pattern=r"^[a-zA-Z0-9_\-]+$")],
            ).validate_python(x),
        ),
    ),
]

from typing import Annotated, Any, Literal
from pydantic import BeforeValidator, Field, TypeAdapter, AliasPath, Json, Base64Bytes


class _PushBody(BaseModel):
    ref: str
    after: str
    clone_url: Annotated[
        str, Field(validation_alias=AliasPath("repository", "clone_url"))
    ]


class PushPayload(BaseModel):
    event: Annotated[
        Literal["push"], Field(validation_alias=AliasPath("headers", "x-github-event"))
    ]
    body: Annotated[
        Json[_PushBody], BeforeValidator(TypeAdapter(Base64Bytes).validate_python)
    ]

Python, Pydantic & OS Version

> python -c "import pydantic.version; print(pydantic.version.version_info())"
             pydantic version: 2.5.2
        pydantic-core version: 2.14.5
          pydantic-core build: profile=release pgo=true
                 install path: /home/rijenkii/.local/lib/python3.12/site-packages/pydantic
               python version: 3.12.1 (main, Dec 18 2023, 00:00:00) [GCC 13.2.1 20231205 (Red Hat 13.2.1-6)]
                     platform: Linux-6.6.14-200.fc39.x86_64-x86_64-with-glibc2.38
             related packages: email-validator-2.1.0.post1 mypy-1.7.1 pyright-1.1.339 typing_extensions-4.9.0 fastapi-0.109.0

Pylance v2024.1.104 (pre-release) (Pyright 1.1.348).

[sydney-runkle]

@rijenkii,

Thanks for reporting this here as well. @adriangb recently responded on the discussion linked above 👍.

[adriangb]

It is a "problem" with pydantic types that as far as I know does not have a real solution until Python makes changes to it's typing system to offer one. And it wasn't a problem until (necessary, correct) breaking changes were made to Python typing that were outside of Pydantic's control.

In the meantime, we relaxed our typing to Any so you won't get a type error as of the next release #8923. But you also won't get any useful typing.

The only thing you can do now is upvote discussion related to TypeForm.

Also, and unrelated, I highly recommend you don't create TypeAdapter instances dynamically. Create them as global constants and reference them from within your lambda functions.

[sydney-runkle]

Thanks for the thoughtful reply @adriangb. Makes sense!!