biais.org

import time
 
def timemeth(func):
    "Internal class method timing decorator"
    def _inner(self, *args, **kw):
        start = time.time()
        res = func(self, *args, **kw)
        self.timed = time.time() - start
        return res
    return _inner
 
class TimedClass:
    def __init__(self):
        self.timed = 0
 
class Sample(TimedClass):
    @timemeth
    def learn(self):
        for i in range(1000000):
            i * i
 
if __name__ == "__main__": 
    s = Sample()
    print "Learning sample..."
    s.learn()
    print "Learning done ! (%.2f secs)" % s.timed

$ python timedclass.py
Learning sample...
Learning done ! (2.01 secs)

Comments

1. Tuesday 30 January 2007 at 12:55, by Justin

I am of the opinion that this is not how decorators should be used.

Benchmarking of an implementation is clearly a different aspect and should not be included into the implementation itself. Something like

def timeit(func, n, args, kwargs):
..a = []
..for i in range(n):
....start = time.time()
....func(*args, **kwargs)
....a.append(time.time() - start)
..return sum(a) / len(a)

seperates benchmarking from implementation.

The only point where this makes sense, is in an environment where you cannot easily "fake" the input (e.G. web development, where it is hard to emulate surfing behaviour).

2. Tuesday 30 January 2007 at 13:03, by maxime

I agree, and I did not use it to benchmark or profile my applications. I work on an application that take about 60 seconds to load (because of network checks, agent distribution, ...). I use this decorator to print time informations about elapsed parts of the loading. like:

network check 1 : OK (done in 1.24 secs)
network check 2 : OK (done in 4.11 secs)
computing things : OK (done in 0.25 secs)

For profiling I use this decorator: www.biais.org/blog/index.... and for benchmarking I'm using a third I will publish soon.

3. Tuesday 30 January 2007 at 13:54, by Justin

Okay, that seems like one of the cases I meant.

What you can do in those cases is, to put the "timing logic" out of the "business logic" via not using the decorator syntax, but instead something like
>>> s = Sample()
>>> s.learn = timemeth(s.learn)

4. Tuesday 30 January 2007 at 14:21, by maxime

Sure, I may use this:

import time

def timemeth(func):
    def _inner(*args, **kw):
        start = time.time()
        res = func(*args, **kw)
        timed = time.time() - start
        return res, timed
    return _inner

class Sample:
    def learn(self):
        for i in range(1000000):
            i * i

s = Sample()
learn = timemeth(s.learn)
dummy, t = learn()

print "Learning done ! (%.2f secs)" % t

Write your comment

Name:

Email (optional):

Web site(optional):

Comment:

Captcha anti-spam :
What is the color of the white horse of Napoleon ?

HTML code will be displayed as text

Remember me