A comparison of two methods to canonicalize (to byte strings) a list of 2-tuples containing byte strings and unicode strings.

conversion-benchmark.py

#!/usr/bin/env python

from __future__ import unicode_literals, print_function

"""Compare two algorithms that canonicalize a list of tuples.

Python 2.7.0:

Good headers:
List creation and iteration: 2.98474693298
List iteration and substitution: 3.3568974336
11.0861445122% penalty

Mixed headers:
List creation and iteration: 7.56983908017
List iteration and substitution: 3.39938569069
55.0930256947% improvement

Bad headers:
List creation and iteration: 10.04327027
List iteration and substitution: 3.36879269282
66.4572136142% improvement


Python 3.1.3:

Good headers:
List creation and iteration: 3.96421766281
List iteration and substitution: 4.36530995369
9.18817438242% penalty

Mixed headers:
List creation and iteration: 11.3370667299
List iteration and substitution: 4.5164826711
60.1618057058% improvement

Bad headers:
List creation and iteration: 16.6681366762
List iteration and substitution: 4.34285736084
73.9451538872% improvement


"""


import sys

# Python 3.x compatibility.
if sys.version_info >= (3, 0):
    unicode = str

else:
    range = xrange



headers_good = [
        (b'Content-Type', b"text/plain"),
        (b'Content-Length', b'27'),
        (b'Content-MD5', b"1234567890123456789012")
    ]

headers_mixed = [
        (b'Content-Type', "text/plain"),
        (b'Content-Length', b'27'),
        ('Content-MD5', "1234567890123456789012")
    ]

headers_bad = [
        ('Content-Type', "text/plain"),
        ('Content-Length', '27'),
        ('Content-MD5', "1234567890123456789012")
    ]


def foo(headers):
    headers_ = list()
    for name, value in headers:
        if not isinstance(name, unicode) and not isinstance(value, unicode):
            continue
        
        if isinstance(name, unicode):
            name = name.encode('iso-8859-1')
        
        if isinstance(value, unicode):
            value = value.encode('iso-8859-1')
        
        headers_.append((name, value))
    return headers_


def bar(headers):
    for i in range(len(headers)):
        name, value = headers[i]
        
        if not isinstance(name, unicode) and not isinstance(value, unicode):
            continue
        
        if isinstance(name, unicode):
            name = name.encode('iso-8859-1')
        
        if isinstance(value, unicode):
            value = value.encode('iso-8859-1')
        
        headers[i] = (name, value)
    
    return headers


def avgresult(c, count=3.0):
    return sum([c() for i in range(3)]) / count


def result(t1, t2):
    print(abs(1.0 - (t1 if t2 > t1 else t2) / (t2 if t2 > t1 else t1)) * 100, '%', " improvement" if 1.0 - t1 / t2 < 0 else " penalty", sep='')


if __name__ == '__main__':
    from timeit import Timer
    
    print("Running each test three times and averaging the result;\nplease wait, this could take some time.\n\nGood headers:")
    
    t1 = avgresult(lambda: Timer("foo(headers_good)", "from __main__ import foo, headers_good").timeit())
    print("List creation and iteration:", t1)
    
    t2 = avgresult(lambda: Timer("bar(headers_good)", "from __main__ import bar, headers_good").timeit())
    print("List iteration and substitution:", t2)
    
    result(t1, t2)
    
    print("\nMixed headers:")
    
    t1 = avgresult(lambda: Timer("foo(headers_mixed)", "from __main__ import foo, headers_mixed").timeit())
    print("List creation and iteration:", t1)
    
    t2 = avgresult(lambda: Timer("bar(headers_mixed)", "from __main__ import bar, headers_mixed").timeit())
    print("List iteration and substitution:", t2)
    
    result(t1, t2)
    
    print("\nBad headers:")
    
    t1 = avgresult(lambda: Timer("foo(headers_bad)", "from __main__ import foo, headers_bad").timeit())
    print("List creation and iteration:", t1)
    
    t2 = avgresult(lambda: Timer("bar(headers_bad)", "from __main__ import bar, headers_bad").timeit())
    print("List iteration and substitution:", t2)
    
    result(t1, t2)