I am implementing the Cartesian product of multiple sets in python using recursion.

Here is my implementation:

def car_two_sets(a, b):
    result = []
    for x in a:
        for y in b:
            result.append(str(x)   str(y))
    return result


def car_multiple_sets(lists):
    if len(lists) == 2:
        return car_two_sets(lists[0], lists[1])
    else:
        return car_multiple_sets([car_two_sets(lists[0], lists[1])]   lists[2:])


a = [1, 2]
b = [3, 4]
c = [6, 7, 8]
lists = [a, b, c]
print(car_multiple_sets(lists))

The code works correctly, but for larger number of sets, it is slow. Any ideas on how to improve this implementation? I thought of memoization, but could not find any repetitive calculations to cache.

I do not want to use itertools functions.

CodePudding user response：

Benchmark with three times more lists:

 221 us   223 us   223 us  h
 225 us   227 us   227 us  k3
 228 us   229 us   229 us  k2
 267 us   267 us   267 us  k
 340 us   341 us   342 us  g
1177 us  1185 us  1194 us  car_multiple_sets
3057 us  3082 us  3084 us  f

Code (Try it online!):

from timeit import repeat
from random import shuffle
from bisect import insort
from itertools import product, starmap
from operator import concat

def car_two_sets(a, b):
    result = []
    for x in a:
        for y in b:
            result.append(str(x)   str(y))
    return result


def car_multiple_sets(lists):
    if len(lists) == 2:
        return car_two_sets(lists[0], lists[1])
    else:
        return car_multiple_sets([car_two_sets(lists[0], lists[1])]   lists[2:])

def f(lists):
    return [''.join(map(str,a)) for a in product(*lists)]

def g(lists):
    return [''.join(a) for a in product(*[map(str,a)for a in lists])]

def h(lists):
    return list(map(''.join, product(*[map(str,a)for a in lists])))

def k(lists):
    result = ['']
    for lst in lists:
        lst = [*map(str, lst)]
        result = [S   s for S in result for s in lst]
    return result

def k2(lists):
    result = ['']
    for lst in lists:
        result = list(starmap(concat, product(result, map(str, lst))))
    return result

def k3(lists):
    result = ['']
    for lst in lists:
        result = starmap(concat, product(result, map(str, lst)))
    return list(result)

funcs = [car_multiple_sets, f, g, h, k, k2, k3]

a = [1, 2]
b = [3, 4]
c = [6, 7, 8]
lists = [a, b, c]

for func in funcs:
  print(func(lists), func.__name__)

times = {func: [] for func in funcs}
lists *= 3
for _ in range(50):
  shuffle(funcs)
  for func in funcs:
    t = min(repeat(lambda: func(lists), number=1))
    insort(times[func], t)
for func in sorted(funcs, key=times.get):
    print(*('M us ' % (t * 1e6) for t in times[func][:3]), func.__name__)

(f and g are from a currently deleted answer, the k functions are from me)

CodePudding user response：

A few comments:

• If you think about it, what car_multiple_sets is doing is iterating on its parameter lists. You're doing that using recursion, but iterating on a list can also be done with a for-loop. And it so happens that recursion is somewhat slow and memory-inefficient in python, so for-loops are preferable.

• You don't need to convert to str to group the ints together. You can use tuples. That's precisely what they're for. Replace str(x) str(y) with (x,y) to get a pair of two integers instead of a string.

def car_two_sets(a, b):
    if all(isinstance(x, tuple) for x in a):
        return [(*x, y) for x in a for y in b]
    else:
        return [(x, y) for x in a for y in b]

def car_multiple_sets(lists):
    if len(lists) == 0:
        return [()]
    elif len(lists) == 1:
        return [(x,) for x in lists[0]]
    else:
        result = car_two_sets(lists[0], lists[1])
        for l in lists[2:]:
            result = car_two_sets(result, l)
        return result

print( car_multiple_sets((range(3), 'abc', range(2))) )
# [(0, 'a', 0), (0, 'a', 1), (0, 'b', 0), (0, 'b', 1), (0, 'c', 0), (0, 'c', 1),
#  (1, 'a', 0), (1, 'a', 1), (1, 'b', 0), (1, 'b', 1), (1, 'c', 0), (1, 'c', 1),
#  (2, 'a', 0), (2, 'a', 1), (2, 'b', 0), (2, 'b', 1), (2, 'c', 0), (2, 'c', 1)]