djb2.py

gistfile1.py

#!/usr/bin/env
# -*- coding: utf-8 -*-

from collections import Counter
from contextlib import contextmanager

def djb2(uuid):
    """ See for details: http://www.cse.yorku.ca/~oz/hash.html """

    _hash = 5381

    for i in xrange(0, len(uuid)):
       _hash = ((_hash << 5) + _hash) + ord(uuid[i])

    return _hash

def stats(iterable):
    """ Returns mean, median and std.dev of iterable"""
    try:
        import numpy as np
    except ImportError:
        return (0,0,0)

    data = list(iterable)
    return [func(data) for func in [np.mean, np.median, np.std]]

@contextmanager
def benchmark():
    import time
    start = time.time()
    yield
    print "Total time: {0}".format(time.time() - start)

if __name__ == "__main__":
    import uuid
    from zlib import crc32
    from hashlib import md5, sha256

    result = {}
    hash_to_long = lambda x: int(x.hexdigest(), base=16)
    hashes = [
                (djb2, lambda x: x),
                (crc32, lambda x: x),
                (md5, hash_to_long),
                (sha256, hash_to_long),
             ]
    for func, convert in hashes:
        print func.__name__
        with benchmark():
            histogram = Counter(convert(func(uuid.uuid1().get_hex())) % 16 for _ in xrange(100000))
        print histogram
        print stats(histogram.values())
        print

This shows that djb2 is not so uniform =(