functools holds tools that take functions and return new ones.

from functools import (
    lru_cache, partial, reduce)

parse_hex = partial(int, base=16)
print(parse_hex("ff"))
print(reduce(lambda a, b: a * b, [2, 3, 4]))

@lru_cache(maxsize=None)
def fib(n):
    if n < 2:
        return n
    return fib(n - 1) + fib(n - 2)

print(fib(80), fib.cache_info().hits)

• partial fixes some arguments now and leaves the rest for later.
• reduce folds an iterable into one value, two items at a time. Give it a start value for empty input. sum, max or a loop is often clearer.
• lru_cache remembers results by argument, so arguments must be hashable. cache is the same without a size limit.

from functools import singledispatch

@singledispatch
def show(x):
    return repr(x)

@show.register
def _(x: list):
    return " | ".join(map(show, x))

print(show([1, "a", [2]]))

singledispatch picks an implementation from the type of the first argument. New types can be registered later, without editing a chain of isinstance checks.

Define parse_binary with partial, so parse_binary("101") is 5. Write product(nums) with reduce, returning 1 for an empty list. Write grid_paths(rows, cols), the number of right-or-down routes across a grid (1 when either side is 1), cached with lru_cache. Write describe(value) with singledispatch: "int 5" for ints, "text 'hi'" for strings and "other" for anything else.

The explanation opens here once every check passes.