why_groupby_is_slow.py

'''
Created on 2009-12-15

@author: lreqc
'''

import timeit

SETUP = """\
import random

alphabet = ['a', 'b', 'c', ' ']
rng = random.Random()

charlist = [ rng.choice(alphabet) for _ in xrange(1000) ]
"""

GROUPBY_LAMBDA = """
import itertools

def list_split(iterable):
    return [''.join(slowo) for jest, slowo in itertools.groupby(iterable, lambda x: bool(x.strip())) if jest]    
"""

GROUPBY_PLAIN = """
import itertools

def list_split(iterable):
    return [''.join(word) for ommit, word in itertools.groupby(iterable, str.isspace) if not ommit]    
"""

FORLOOP = """
def list_split(lista):
    slowa = []
    aktualne_slowo = ''
    for litera in lista:
        if litera.isspace(): 
            if aktualne_slowo: 
                slowa.append(aktualne_slowo) 
                aktualne_slowo = ''
        else:
            aktualne_slowo += litera
 
    if aktualne_slowo:
        slowa.append(aktualne_slowo)
 
    return slowa
"""

if __name__ == '__main__':
    print "Groupby with lambda (average time from 10 runs, 500 calls per):"
    print sum(timeit.repeat("list_split(charlist)", setup = SETUP + GROUPBY_LAMBDA, number = 500, repeat = 10)) / 10
    print "Groupby with builtin:"
    print sum(timeit.repeat("list_split(charlist)", setup = SETUP + GROUPBY_PLAIN, number = 500, repeat = 10)) / 10
    print "For loop with builtins and lists:"
    print sum(timeit.repeat("list_split(charlist)", setup = SETUP + FORLOOP, number = 500, repeat = 10)) / 10