Let's suppose we want raise x to a number y.

You'd write this as:

def raise_power(x, y):
    return x**y

What if your y value can assume a finite set of values?

Let's suppose y can be one of [3,4,5] and let's say you don't want offer end user the possibility to use such function since it is very computationally intensive. In fact you would check if provided y assumes a valid value and rewrite your function as:

def raise(x, y):
    if y in (3,4,5):
        return x**y
    raise NumberNotInRangeException("You should provide a valid exponent")

Messy? Let's use the abstract form and specialize it to all three cases: let's implement them partially.

from functors import partial
raise_to_three = partial(raise, y=3)
raise_to_four = partial(raise, y=4)
raise_to_five = partial(raise, y=5)

What happens here? We fixed the y params and we defined three different functions.

No need to use the abstract function defined above (you could make it private) but you could use partial applied functions to deal with raising a number to a fixed value.