How can I lazily generate and process combinations of results from a generator?

Question

I have a generator, and I'd like to perform a nested loop on it in such a way that the inner loop will start from where the outer loop stands at the moment. For example, I have a generator that produces the list [1,2,3], and my loop should produce: (1,2),(1,3),(2,3). The code I came up with is the following:

from itertools import tee

def my_gen():
    my_list = [1, 2, 3]
    for x in my_list:
        yield  x

first_it = my_gen()
while True:
    try:
        a = next(first_it)
        first_it, second_it = tee(first_it)
        for b in second_it:
            print(a,b)
    except StopIteration:
        break

This code is cumbersome, not efficient and does not look very pythonic to me. Please notice that I cannot use combinations_with_replacement because I need an inner loop for processing a specific value from the outer loop.

Any suggestions for a more elegant and pythonic code?

Possible duplicate of [Python: return unique only cartesian product of a list containing lists?](https://stackoverflow.com/questions/19946195/python-return-unique-only-cartesian-product-of-a-list-containing-lists) — Saif Ahmad, Dec 26 '17 at 10:51
These are combinations - without replacement; not a cartesian product. — Karl Knechtel, Jun 25 '22 at 00:43

user2390182 · Accepted Answer · 2017-12-26T11:15:26.180

3

The repeated cloning and exhausting of only one of the resulting iterators is not very efficient. As per the itertools.tee docs:

In general, if one iterator uses most or all of the data before another iterator starts, it is faster to use list() instead of tee().

from itertools import islice

my_list = [1, 2, 3]
# or, more generally
# my_list = list(my_gen())

for i, a in enumerate(my_list):
    for b in islice(my_list, i+1, None):
        print((a, b))
(1, 2)
(1, 3)
(2, 3)

edited Dec 26 '17 at 11:15

answered Dec 26 '17 at 11:05

user2390182

72,016
6
67
89

This is a great answer, but I'd rather stay with a generators, and not work with lists. – Gino Dec 26 '17 at 11:46
1

@Gino Generators are usually used to avoid loading a lot of things into memory. Using a tee basically has a behind-the-scenes list that is filled while iterating one, and emptied while iterating the other. Using a tee and iterating over one until the end is the same as loading the whole generator into memory, so making a list doesn't hurt. – Artyer Dec 26 '17 at 11:53

score 0 · Answer 2 · answered Dec 26 '17 at 11:10

The result (1,2),(1,3),(2,3) isn't the cartesian product. Do you mean to obtain all the combinations without repetition?

For combinations without repetition

Use itertools.combinations function:

from itertools import combinations

print(list(combinations([1,2,3], 2)))

For actual cartesian product

Using list comprehension you can reduce your code into a more pythonic and elegant way:

my_list = [1, 2, 3]

c_prod = [(i, j) for i in iter(my_list) for j in iter(my_list)]

print(c_prod)

Another choice is to use the itertools.product function:

import itertools

for i in itertools.product(my_list, my_list):
    print(i)

How can I lazily generate and process combinations of results from a generator?

2 Answers2