Question

In the command line, I can do:

python -m timeit 'a = 1'

According to the docs:

If -n is not given, a suitable number of loops is calculated by trying successive powers
of 10 until the total time is at least 0.2 seconds.

This works great, but how can I get the same behavior when using timeit in my program? If I leave out the number argument to the timeit.timeit call, it will simply default to 1000000.

Answer 1

The docs don’t make it obvious that there exists such functionality, I’ll assume there isn’t. Fortunately, defining a wrapper that does it for you is isn’t too hard:

import timeit

def auto_timeit(stmt='pass', setup='pass'):
    n = 1
    t = timeit.timeit(stmt, setup, number=n)

    while t < 0.2:
        n *= 10
        t = timeit.timeit(stmt, setup, number=n)

    return t / n # normalise to time-per-run