Model Validator (before) does not receive alias in input #696
Description
I recently upgraded pydantic-settings (2.6.1 -> 2.11.0) and now have the following issue. I have a settings class with fields that have a validation_alias using AliasChoices. I want to use the aliases to deprecate an old field name without breaking changes. Therefore, I added model_validator that checks if old field names are supplied and issues a warning.
# file: alias.py
import warnings
from typing import Any
from pydantic import AliasChoices, Field, model_validator
from pydantic_settings import BaseSettings
class Settings(BaseSettings):
NAME: str = Field(
default="",
validation_alias=AliasChoices("NAME", "OLD_NAME"),
)
VALUE: str = Field(
default="",
validation_alias=AliasChoices("VALUE", "OLD_VALUE"),
)
@model_validator(mode="before")
def check_for_deprecated_attributes(cls, data: Any) -> Any:
if isinstance(data, dict):
print("DEBUG: data", data)
old_keys = {k for k in data.keys() if k.startswith("OLD_")}
if old_keys:
warnings.warn(
"The config attributes starting with 'OLD_' are deprecated and will be removed in the"
" future. Please remove the prefix 'OLD_'."
f" The following deprecated attributes were set: {old_keys!r}",
DeprecationWarning,
stacklevel=4,
)
return data
if __name__ == "__main__":
settings = Settings(OLD_NAME="foo")If I run python -Wd alias.py, I get:
DEBUG: data {'NAME': 'foo'}
I.e., the before-validator already receives the "resolved" field name.
But I would expect the following (which works with v2.6.1):
DEBUG: data {'OLD_NAME': 'foo'}
/home/rafael/alias.py:35: DeprecationWarning: The config attributes starting with 'OLD_' are deprecated and will be removed in the future. Please remove the prefix 'OLD_'. The following deprecated attributes were set: {'OLD_NAME'}
settings = Settings(OLD_NAME="foo")
I don't know whether this is a regression or the new behavior is expected.
Thanks for your feedback! 🙂