Demo of why sum(list_of_lists, []) is a bad idea (from a performance perspective)

flatten_time.py

"""
Demonstration of how slow sum can be for flattening lists-of-lists

Output on my machine:
0.019118324998999014 itertools.chain.from_iterable
0.03325132900135941 comprehension
10.947899631002656 sum

Using sum on a list results in a loop inside a loop (a new list is made for
each new sub-list) so the more sub lists there are, the bigger the timing
difference will be.
"""
import itertools
import random
import timeit


list_of_lists = [
    [random.randint(1, 1000) for _ in range(3)]
    for _ in range(7_000)
]


def flatten_chain(iterables):
    return list(itertools.chain.from_iterable(iterables))


def flatten_comp(iterables):
    return [
        item
        for iterable in iterables
        for item in iterable
    ]


def flatten_sum(iterables):
    return sum(iterables, [])


def time(function):
    return max(timeit.repeat(
        f"{function}(list_of_lists)",
        number=50,
        repeat=3,
        globals=globals(),
    ))


print(time("flatten_chain"), "itertools.chain.from_iterable")
print(time("flatten_comp"), "comprehension")
print(time("flatten_sum"), "sum")