Order of alias given in AliasChoices yield different results

[andlogreg]

Versions:

pydantic                       2.9.2
pydantic_core                  2.23.4
pydantic-settings              [current main branch so that this fix is present: https://github.com/pydantic/pydantic-settings/pull/400]

In the continuation of #406 I've found that the order of the alias given in AliasChoices change the behaviour, which I believe, shouldn't be the case.

Given the following example:

# test_config.py
from pydantic import Field
from pydantic_settings import BaseSettings, SettingsConfigDict
from pydantic import Field, AliasChoices, BaseModel, AliasGenerator, ConfigDict
from typing import List

validation_alias_a = lambda x: AliasChoices(x, x.replace('_','-'))
validation_alias_b = lambda x: AliasChoices(x.replace('_','-'), x)

my_validation_alias = validation_alias_a

class SubModel(BaseModel):
    sub_list: List[str] = Field(
        default_factory=list, description="A list argument",
    )
    model_config = ConfigDict(
        alias_generator=AliasGenerator(validation_alias=my_validation_alias)
    )

class MySettings(BaseSettings):
    sub_model: SubModel = Field(
        SubModel(),
    )

    model_config = SettingsConfigDict(
        cli_parse_args=True,
        env_nested_delimiter="__",
        alias_generator=AliasGenerator(validation_alias=my_validation_alias)
    )

settings = MySettings()

print(settings)

The following works as expected:

# python test_config.py --sub-model.sub-list=a,b,c
sub_model=SubModel(sub_list=['a', 'b', 'c'])

But if I change the validation_alias:

(...)
my_validation_alias = validation_alias_b
(...)

Then the following fails

# python test_config.py --sub-model.sub-list=a,b,c
Traceback (most recent call last):
  File "./test_config.py", line 30, in <module>
    settings = MySettings()
  File "/usr/local/lib/python3.10/site-packages/pydantic_settings/main.py", line 155, in __init__
    super().__init__(
  File "/usr/local/lib/python3.10/site-packages/pydantic/main.py", line 212, in __init__
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
pydantic_core._pydantic_core.ValidationError: 1 validation error for MySettings
sub_model.sub-list
  Input should be a valid list [type=list_type, input_value='["a","b","c"]', input_type=str]
    For further information visit https://errors.pydantic.dev/2.9/v/list_type

---

In fact, if both of the alias do not include the field name itself (the x), for example:

(...)
validation_alias_a = lambda x: AliasChoices(x.replace('_','__'), x.replace('_','-'))
validation_alias_b = lambda x: AliasChoices(x.replace('_','-'), x.replace('_','__'))
(...)

then the error is always thrown.

---

Not sure if it's in pydantic or pydantic-settings. But I guess this is a bug?

Note this is not a deal breaker since with the workaround of including x as the first alias choice mitigates it.

[hramezani]

Thanks @andlogreg for reporting this bug.

I've created #411 as a fix. Could you please confirm?

[andlogreg]

Thanks @andlogreg for reporting this bug.

I've created #411 as a fix. Could you please confirm?

Thanks @hramezani I can confirm the example above works.

[hramezani]

Thanks @andlogreg for confirming!