Type aliases disable automatic JSON parsing

[GeorchW]

When declaring a type alias, any JSON inside an env var will stop being parsed. Consider the following code (run with uv run reproduction.py):

# /// script
# requires-python = ">=3.12"
# dependencies = [
#     "pydantic-settings",
# ]
# ///
import os

from pydantic_settings import BaseSettings

type MyType = list[int]

print(MyType, type(MyType))  # prints "MyType <class 'typing.TypeAliasType'>"


class Settings(BaseSettings):
    MY_VAR: MyType


os.environ["MY_VAR"] = "[1, 2, 3]"
settings = Settings()  # type: ignore
print(settings.MY_VAR, type(settings.MY_VAR))

This fails with an error message:

Traceback (most recent call last):
  File "reproduction.py", line 21, in <module>
    settings = Settings()  # type: ignore
               ^^^^^^^^^^
  File ".../site-packages/pydantic_settings/main.py", line 188, in __init__
    super().__init__(
  File ".../site-packages/pydantic/main.py", line 253, in __init__
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
pydantic_core._pydantic_core.ValidationError: 1 validation error for Settings
MY_VAR
  Input should be a valid list [type=list_type, input_value='[1, 2, 3]', input_type=str]
    For further information visit https://errors.pydantic.dev/2.11/v/list_type

It works fine if the type is not an alias. When replacing like this, there's no error:

-type MyType = list[int]
+MyType = list[int]

Instead, the output looks as expected:

list[int] <class 'types.GenericAlias'>  # output of print(MyType, type(MyType))
[1, 2, 3] <class 'list'>                # output of print(settings.MY_VAR, type(settings.MY_VAR))

This just took me a bit to figure out. I'd expect a type alias not to change any behavior of the parsing.

[Viicos]

I believe already fixed by #644. Works on main.

[GeorchW]

Great! Is there a time frame for the release?