[Question] Is it still possible to pass None through environment variables? about pydantic-settings HOT 8 CLOSED

@kschwab I think it's better to create new issue with your question or maybe I should reopen this one again?

from pydantic-settings.

Agree @kschwab. Let's define it as str

from pydantic-settings.

Thanks @eremeevfd for the issue.

It worked in Pydantic 1.* and I've read that we can use ENV_VAR=None|none|null to parse as None in python.

Where did you read it?

You already provided a value(None as a string) for env variable, So, pydantic-settings will pick that.
If you don't provide any value for the env, it will use the default value which is None in you case.

from pydantic-settings.

@hramezani Okay, I'm sorry, I think I've mixed topics of boolean parsing (true/True/false/False) with None parsing and somehow thought that null/None from envvar should be parse to None type also 😆
Everything works as expected and worked this way!

from pydantic-settings.

@hramezani I was going to submit a similar question, but as inverse of the case presented above:

import os

from pydantic_settings import BaseSettings

class Settings(BaseSettings):
    ENV_VAR: str | None = 'Default string'

if __name__ == '__main__':

    os.environ['ENV_VAR'] = 'None' # or 'null'
    print(type(Settings().ENV_VAR) == type(None))
    '''
    False
    '''

The use case I was thinking of was Optional[int] = 2 where it's very clear that 'None' (str) should be coerced into None (NoneType). For the above Optional[str] case, adding a env_none_str spec in the SettingsConfigDict for environment variables seems like a good solution.

If env_none_str=None (the default), no coercion will take place. If env_none_str="None", environment variables with "None" string values will be coerced into None NoneTypes.

Thoughts?

from pydantic-settings.

@kschwab I would suggest adding a new config flag named e.g. parse_env_none or something like this.
The default value should be False and it means string'None' should be considered as string.(to don't introduce breaking change).
By setting the flag to True, string 'None' should be parse as None

from pydantic-settings.

@hramezani thanks for the feedback.

My rationale behind using a str flag instead of a bool flag was when considering compliance with e.g. a JSON spec null value. It may be desirable that users want "null" (str) to be coerced into None instead of only "None" (str) to None. i.e., the None str parse value is user defined. Defining the flag itself as None would be the default so as to not introduce breaking changes.

I actually don't have a strong preference either way. A bool flag is simpler and likely the dominant case. Let me know how you would like to proceed. I'll take a stab at initial PR for your feedback.

from pydantic-settings.

@eremeevfd

As an update, once PR #206 merges, the below will now work:

import os

from pydantic_settings import BaseSettings

class Settings(BaseSettings):
    ENV_VAR: str | None = 'Default string'

if __name__ == '__main__':

    os.environ['ENV_VAR'] = 'None'
    print(type(Settings().ENV_VAR) == type(None))
    '''
    False
    '''

    print(type(Settings(_env_parse_none_str='None').ENV_VAR) == type(None))
    '''
    True
    '''

    os.environ['ENV_VAR'] = 'null'
    print(type(Settings(_env_parse_none_str='null').ENV_VAR) == type(None))
    '''
    True
    '''

from pydantic-settings.