是否可以使用布尔返回函数作为类型提示？

Question

Below are two examples of functions having some type hints:

def sum_dem_digits(digits:str) -> int:
    return sum(int(d) for d in digits)

def do_stuff(x:int):
    lizzard = ["eyes", "nose", "body", "tail"]
    return lizzard[x]

Is there some way to set an arbitrary boolean-returning function to be the type hint?

import string

def numeric_string(chs:str) -> bool:
    """
        BAD:  "ANACONDA, RATTLESNAKE, SIDEWINDER"
        BAD:  "WHAT I HAD FOR LUNCH"
        GOOD: "1234"
        GOOD: "99"
    """
    return all([ch in string.digits for ch in chs])

def sum_dem_digits(digits:numeric_string) -> int:
    return sum(int(d) for d in digits)

#######################################################################

def do_stuff_helper(x:int):
    return (0 <= x) and (x <= 3)

def do_stuff(x:do_stuff_helper):
    lizzard = ["eyes", "nose", "body", "tail"]
    return lizzard[x]

Answer 1

The answer isn't quite a straightforward yes or no.

Arbitrary functions are not valid static types. They are technically legal annotations, because you can use any object as an annotation and Python won't care, but anything that does type analysis with type hints will see this as an error.

More importantly, why are you doing this? If you were hoping that tools like mypy would be able to check whether inputs pass your function, then that's hopeless. mypy is a static tool, and running your function would be a runtime thing. mypy can't see runtime argument values, or figure out what running do_stuff_helper on those values would do.

If you want to build your own tool that performs runtime constraint checking based on constraint functions supplied as type annotations, then sure, you can do that. Ideally, you'd use typing.Annotated to supply your constraint functions in a way that doesn't interfere with type hints:

class Constraint:
    def __init__(self, constraint_function):
        self.constraint_function = constraint_function

@some_checker_decorator_you_write
def do_stuff(x: Annotated[int, Constraint(do_stuff_helper)]):
    ...