gistfile1.py

import random
import timeit
from copy import deepcopy

def inter_reduce(sets):
    return reduce(set.intersection, sets)

def inter_for_deepcopy(sets):
    sets = iter(sets)
    result = deepcopy(next(sets))
    for s in sets:
        result.intersection_update(s)
    return result

def inter_for_copy(sets):
    sets = iter(sets)
    result = next(sets).copy()
    for s in sets:
        result.intersection_update(s)
    return result

def inter_for_union(sets):
    sets = iter(sets)
    result = set().union(next(sets))
    for s in sets:
        result.intersection_update(s)
    return result

def inter_for(sets):
    sets = iter(sets)
    result = next(sets)
    for s in sets:
        result.intersection_update(s)
    return result

random.seed(42)
sets = [set(random.randrange(1000) for i in range(4000))
        for j in range(100)]
print len(sets[0])
print "inter_reduce: %s" % timeit.timeit("inter_reduce(sets)",
                    "from __main__ import inter_reduce, sets", number=1000)
print len(sets[0])
print "inter_for_deepcopy: %s" % timeit.timeit("inter_for_deepcopy(sets)",
                    "from __main__ import inter_for_deepcopy, sets", number=1000)
print len(sets[0])
print "inter_for_union: %s" % timeit.timeit("inter_for_union(sets)",
                    "from __main__ import inter_for_union, sets", number=1000)
print len(sets[0])
print "inter_for_copy: %s" % timeit.timeit("inter_for_copy(sets)",
                    "from __main__ import inter_for_copy, sets", number=1000)
print len(sets[0])
print "inter_for: %s" % timeit.timeit("inter_for(sets)",
                    "from __main__ import inter_for, sets", number=1000)
print len(sets[0])

Yeah I suppose that's true, since none of the elements of the set are modified (so no need to copy them all). Added this to the snippet, when running it was almost identical to the union() version give or take a couple hundredths of a second.