I am learning about timing analysis in Python. I used time module before it was pointed out that timeit module would be better for timing analysis.
My main focus on timing analysis is for conducting many test cases on 2 variations of the same function. In this example code I used 2 variations of function for reversing a number.
Can the timing analysis be done in a better way? Is the method scalable for many test cases?
def reverse_num(num):
'''returns the reverse of an integer '''
if num < 0:
return - int(str(-num)[::-1])
else:
return int(str(num)[::-1])
def reverse_num2(num):
rev = 0
while num > 0:
rev *= 10
rev += num % 10
num /= 10
return rev
if __name__ == "__main__":
from timeit import Timer
def test(f):
for i in xrange(1000):
f(i)
print Timer(lambda: test(reverse_num)).timeit(number = 100)
print Timer(lambda: test(reverse_num2)).timeit(number = 100)
I am posting sample test runs for my usage of timeit module for timing for the following code. I wrote it for this question.
def mysort(words):
mylist1 = sorted([i for i in words if i[:1] == "s"])
mylist2 = sorted([i for i in words if i[:1] != "s"])
list = mylist1 + mylist2
return list
def mysort3(words):
ans = []
p = ans.append
q = words.remove
words.sort()
for i in words[:]:
if i[0] == 's':
p(i)
q(i)
return ans + words
def mysort4(words):
ans1 = []
ans2 = []
p = ans1.append
q = ans2.append
for i in words:
if i[0] == 's':
p(i)
else:
q(i)
ans1.sort()
ans2.sort()
return ans1 + ans2
def mysort6(words):
return ( sorted([i for i in words if i[:1] == "s"]) +
sorted([i for i in words if i[:1] != "s"])
)
if __name__ == "__main__":
from timeit import Timer
def test(f):
f(['a','b','c','abcd','s','se', 'ee', 'as'])
print Timer(lambda: test(mysort)).timeit(number = 10000)
print Timer(lambda: test(mysort3)).timeit(number = 10000)
print Timer(lambda: test(mysort4)).timeit(number = 10000)
print Timer(lambda: test(mysort6)).timeit(number = 10000)
The timing results are
>>> ================================ RESTART ================================ >>> 0.0643831414457 0.0445699517515 0.0446611241927 0.0616633662243 >>> ================================ RESTART ================================ >>> 0.0625827896417 0.045101588386 0.0443568108447 0.0607123363607 >>> ================================ RESTART ================================ >>> 0.0647336488305 0.0596154305912 0.0445711673841 0.0614101094434 >>> ================================ RESTART ================================ >>> 0.0689924148581 0.0502542495475 0.0443466805735 0.0903267660811 >>> ================================ RESTART ================================ >>> 0.0695374234506 0.0579001730656 0.0443790974415 0.0835670386907 >>> ================================ RESTART ================================ >>> 0.0675612101379 0.05079925814 0.044170413854 0.0681050030978
As you can see that the results vary each time. I understand that the times are small but that is the whole purpose of number = 10000 in timeit to measure consistently by doing timing many times and finding average or the best.
In most runs the 2nd function takes more time than 3rd function but sometimes it doesn't. The 4th function takes more time than 1st function in some cases and sometimes less.
I think it is correct according to usage but why these variable results? Is this the proper way of doing timing or not? I am really confused about this.
After some discussion in the chat room and some more thinking I came up with for testing functions that have integer inputs. Additional nested functions can be defined for more tests and the repetitions can be made more. It may take some time but it gives good results as far as I can see. Simple function calls need to be added for variations of the same function.
def testing():
from timeit import Timer
import random
def tests(f, times):
def test1(f):
f(random.randint(1, 1000))
def test2(f):
f(random.randint(100000, 1000000))
print(f.__name__)
print(Timer(lambda: test1(f)).timeit(number = times))
print(Timer(lambda: test2(f)).timeit(number = times//10))
print()
tests(reverse_num, 10000)
tests(reverse_num2, 10000)
Please note that I started using Python 3 while I was still discussing the question. I won't be using anything specific to Python 3 so all codes can be run on Python 2. Anyone wanting to run new codes will just need to change the
print statements. Nothing else.
