Fastapi: Is it possible to define a value within a response model that contains at least 2 decimal digits [QUESTION]

[meteoDaniel]

Description

Is it possible to define a value within a response model that contains at least 2 decimal digits.?

Additional context

The response model I am talking about looks like this:

class WindData(BaseModel):
    speed: Union[List[float], List[None]] = Schema(
        None,
        title="Wind Speed [m/s]")
    dir: Union[List[float], List[None]] = Schema(
        None,
        title="Wind Direction [°]")
    gust: Union[List[float], List[None]] = Schema(
        None,
        title="Wind Speed [m/s]")

And the values should have maximal 2 decimal digits. I hope you can help me

Thanks a lot !

[dmontagu]

This could probably be done by using a validator that rounds inputs to the nearest 0.01? If you need to preserve the original value then things would get more complicated.

[sm-Fifteen]

If actual decimal values matter, you should probably be using Decimal instead of float. As for precision, If you want to require a certain amount of decimals so that 12 != 12.00 without doing that sort of validation on the incoming strings by hand, Decimal has a same_quantum test that validates that both numbers have the same exponent value.

import Decimal from decimal

Decimal("2.2").same_quantum(Decimal("1.00")) # False
Decimal("2.20").same_quantum(Decimal("1.00")) # True
Decimal("2.200").same_quantum(Decimal("1.00")) # False

You would still need to write a custom validator for this, but you don't risk accidentally adding significant figures to the input data.

If you just want to limit allowed precision but not forbid values with a different level of precision, you can instead quantize your decimal values to ensure they're rounded at the precision level you want.

[tiangolo]

Thanks @dmontagu and @sm-Fifteen ! :bowing_man: :cake:

Does that solve your question @meteoDaniel ?

[sm-Fifteen]

I should probably mention that Pydantic has had max_digits and decimal_places restrictions for its condecimal type since very early versions (though those barely get a mention in the current documentation), which can be used together to have a constraint similar to NUMERIC(precision, scale) in SQL. There's no way to require a minimum or exact number of significant figures, though, at least not without a custom validator.

[dmontagu]

Yeah, there is also another challenge with decimal that pydantic json-encodes it to a float rather than a string (which can come with some float rounding issues if dumped and re-encoded).

[tiangolo]

Yep, good points.

@meteoDaniel if that solved your question, may we close this issue?

[github-actions[bot]]

Assuming the original issue was solved, it will be automatically closed now. But feel free to add more comments or create new issues.