How to make Pydantic's non-strict, coercive mode apply to integer literals?

Question

I'm validating inputs to a function using Pydantic's @validate_call as follows:

from typing import Literal
from pydantic import validate_call

@validate_call
def foo(a: Literal[0, 90, 180, 270]) -> None:
    print(a, type(a))

I want Pydantic to perform its default type coercion like it does with the int type:

foo(90)    # Works as expected
foo('90')  # Doesn't work, but I want it to

If I use the annotation a: int, it will coerce strings like '180', but then I have to manually validate which integers are given.

How do I make Pydantic perform type coercion on Literals?

Note: I'll accept a solution that requires a to be a string type instead of an integer, as long as it still allows both integer and string input.

---

Bad Solutions

• I don't want to add every literal case. Literal[0, 90, 180, 270, '0', '90', '180', '270'] is bad because it doesn't allow the strings '-0' or '180.0'.

• I could do Annotated[int, Field(ge=0, le=0)] | Annotated[int, Field(ge=90, le=90)] | ..., but that's stupidly verbose.

• I don't want to define some separate function or model. At that point, it's easier to just accept a: int and validate the particular value inside the method.

Answer 1

You can combine the BeforeValidator and the Literal like this:

from typing import Annotated, Literal
from pydantic import validate_call, BeforeValidator, ValidationError

# First try has the following validator:
# BeforeValidator(int)

@validate_call
def foo(a: Annotated[Literal[0, 90, 180, 270], BeforeValidator(float)]) -> None:
    print(a, type(a))


if __name__ == "__main__":
    foo("90")
    foo("180.0")
    foo("180.0")
    try:
        foo(0.1)
    except ValidationError as err:
        print(err)
    try:
        foo("70")
    except ValidationError as err:
        print(err)
    try:
        foo("can't convert to int")
    except ValueError as err:
        print(err)

The BeforeValidator function will be called before checks and thus the literal validation will be done against an integer.

Edit: better manage string with decimal number.