how to measure execution time of functions (automatically) in Python

Question

I need to have a base class which I will use to inherit other classes which I would like to measure execution time of its functions.

So intead of having something like this:

class Worker():
    def doSomething(self):
        start = time.time()
        ... do something
        elapsed = (time.time() - start)
        print "doSomething() took ", elapsed, " time to finish"

#outputs: doSomething() took XX time to finish

I would like to have something like this:

class Worker(BaseClass):
    def doSomething(self):
        ... do something

#outputs the same: doSomething() took XX time to finish

So the BaseClass needs to dealing with measuring time

I'd suggest a class decorator instead of implementing the timing in BaseClass. — MattH
– MattH, Commented Feb 11, 2010 at 14:48
Why not just use a profiler like cProfiler whenever you want to know the timings? — Justin Peel
– Justin Peel, Commented Feb 11, 2010 at 14:49
@Justin cProfiler is good but what I needed was to know about timings each time to report them. — sarp
– sarp, Commented Feb 11, 2010 at 16:01

Geoff Reedy · Accepted Answer · 2010-02-11 15:02:49Z

65

One way to do this would be with a decorator (PEP for decorators) (first of a series of tutorial articles on decorators). Here's an example that does what you want.

from functools import wraps
from time import time

def timed(f):
  @wraps(f)
  def wrapper(*args, **kwds):
    start = time()
    result = f(*args, **kwds)
    elapsed = time() - start
    print "%s took %d time to finish" % (f.__name__, elapsed)
    return result
  return wrapper

This is an example of its use

@timed
def somefunction(countto):
  for i in xrange(countto):
    pass
  return "Done"

To show how it works I called the function from the python prompt:

>>> timedec.somefunction(10000000)
somefunction took 0 time to finish
'Done'
>>> timedec.somefunction(100000000)
somefunction took 2 time to finish
'Done'
>>> timedec.somefunction(1000000000)
somefunction took 22 time to finish
'Done'

answered Feb 11, 2010 at 15:02

Geoff Reedy

36.3k3 gold badges60 silver badges79 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

André Staltz Over a year ago

Great example of clean code to show how decorators are one of the best joys in Python

Geoff Reedy Over a year ago

@ronnieaka I'ts just the module name. I wrote the definitions of timed and somefunction in timedec.py and imported it from the interpreter to test them.

Cullen SUN Over a year ago

Good stuff. Thank you. One thing to share with other people. It makes more sense to measure the execution time in milliseconds. e.g. Geoff's answer worked for me, but I only saw the following. abc_function took 0 time to finish. This change will convert the time to milliseconds print "%s took %d milliSeconds to finish" % (f.__name__, elapsed*1000)

p.marino · Accepted Answer · 2010-02-11 14:53:27Z

10

Have you checked the "profile" module?

I.e. are you sure you need to implement your own custom framework instead of using the default profiling mechanism for the language?

You could also google for "python hotshot" for a similar solution.

answered Feb 11, 2010 at 14:53

p.marino

6,2823 gold badges27 silver badges37 bronze badges

Comments

jathanism · Accepted Answer · 2010-02-11 15:07:09Z

6

There is also timeit, which is part of the standard library, and is really easy to use. Remember: don't reinvent the wheel!

answered Feb 11, 2010 at 15:07

jathanism

33.9k9 gold badges72 silver badges89 bronze badges

2 Comments

Leonidas Tsampros Over a year ago

Indeed very simple and easy to use!

sarp Over a year ago

timeit seems promising! I will combine it with Geoff's solution.

Collectives™ on Stack Overflow

how to measure execution time of functions (automatically) in Python

3 Answers 3

3 Comments

Comments

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

3 Comments

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related