A benchmark of two approaches to constructing an array. Bytearray vs array+join. Limitation: one should know the length of the array. Within pypy-1.9 the bytearray approach is one order of magnitude faster. In CPython-2.7, bytearray aprox. 35% faster.

array_append_len_known.py

import timeit

DATA = 'abcdef'

def bytearr():
    bytesn = 1024 * 1024 * 5
    s = bytearray(bytesn)
    i = 0 
    while i < bytesn:
        s[i] = DATA[i%len(DATA)]
        i += 1
    return str(s)

def strs():
    bytesn = 1024 * 1024 * 5
    s = []
    i = 0 
    while i < bytesn:
        s.append(DATA[i%len(DATA)])
        i += 1
    ss = "".join(s)
    return ss

if __name__ == "__main__":
    print timeit.timeit(bytearr,number=10)
    print timeit.timeit(strs,number=10)
    
    
#(pypy-1.9)➜ ✗ python array_append_len_known.py
#0.774618959427
#9.52767682076

#(cPyhton-2.7)➜ ✗ python array_append_len_known.py
#16.6827712059
#21.0312628746

on 2.7 GHz Intel Core i7, 8G RAM, SSD disk.