Testing 'x in s' performance for various python sequence types

testseq.py

import timeit

seq_types = ('set', 'list', 'tuple')
intervals = (10, 100, 1000)

def test(t, i):
    r = t(range(i))
    for j in range(i):
        if i in r:
            pass

for t in seq_types:
    for i in intervals:
        duration = timeit.timeit(
            'test({t}, {i})'.format(t=t, i=i),
            setup='from __main__ import test', number=10000
        )
        print '{i}-long {t} took {d}'.format(
            i=i, t=t, d=duration
        )

My results:

10-long set took 0.0126779079437
100-long set took 0.0721209049225
1000-long set took 0.667522907257
10-long list took 0.0193102359772
100-long list took 1.023209095
1000-long list took 98.0308930874
10-long tuple took 0.0185379981995
100-long tuple took 1.02909088135
1000-long tuple took 97.4343318939

Sets win by a mile. Tuples are fractionally faster at this than lists.

The time required for each type rises super-linearly with the length of the sequence.