list-comprehension and map question (simple)

runsun pan runsun at gmail.com
Wed Mar 30 18:32:57 CEST 2005


Thx, Robert. I did some tests:

>>> xs = range(10000)
>>> ys = range(0,20000,2)
>>> def m1(x, count=100):
... 	for c in range(count):
... 	   y = map(float, x)
... 	return y
>>> def L1(x, count=100):
... 	for c in range(count):
... 	   y = [float(z) for z in x]
... 	return y
>>> def m2(xs, yx, count=100):
... 	for c in range(count):
... 		y = map(lambda x,y: x+y, xs, ys)
... 	return y
>>> def L2(xs, yx, count=100):
... 	for c in range(count):
... 		y =[x+y for x,y in itertools.izip(xs, ys)]
... 	return y

>>> profile.run('m1(xs)')
         105 function calls in 0.991 CPU seconds

   Ordered by: standard name

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      100    0.484    0.005    0.484    0.005 :0(map)
        1    0.000    0.000    0.000    0.000 :0(range)
        1    0.485    0.485    0.485    0.485 :0(setprofile)
        1    0.022    0.022    0.506    0.506 <interactive input>:1(m1)
        1    0.000    0.000    0.506    0.506 <string>:1(?)
        1    0.000    0.000    0.991    0.991 profile:0(m1(xs))
        0    0.000             0.000          profile:0(profiler)

>>> profile.run('L1(xs)')
         5 function calls in 0.724 CPU seconds

   Ordered by: standard name

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.000    0.000    0.000    0.000 :0(range)
        1    0.000    0.000    0.000    0.000 :0(setprofile)
        1    0.724    0.724    0.724    0.724 <interactive input>:1(L1)
        1    0.000    0.000    0.724    0.724 <string>:1(?)
        1    0.000    0.000    0.724    0.724 profile:0(L1(xs))
        0    0.000             0.000          profile:0(profiler)

>>> profile.run('m2(xs, ys)')
         1000105 function calls in 6.171 CPU seconds

   Ordered by: standard name

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      100    3.336    0.033    6.149    0.061 :0(map)
        1    0.000    0.000    0.000    0.000 :0(range)
        1    0.000    0.000    0.000    0.000 :0(setprofile)
        1    0.022    0.022    6.171    6.171 <interactive input>:1(m2)
  1000000    2.813    0.000    2.813    0.000 <interactive
input>:3(<lambda>)
        1    0.000    0.000    6.171    6.171 <string>:1(?)
        1    0.000    0.000    6.171    6.171 profile:0(m2(xs, ys))
        0    0.000             0.000          profile:0(profiler)

>>> profile.run('L2(xs, ys)')
         5 function calls in 0.418 CPU seconds

   Ordered by: standard name

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.000    0.000    0.000    0.000 :0(range)
        1    0.000    0.000    0.000    0.000 :0(setprofile)
        1    0.418    0.418    0.418    0.418 <interactive input>:1(L2)
        1    0.000    0.000    0.418    0.418 <string>:1(?)
        1    0.000    0.000    0.418    0.418 profile:0(L2(xs, ys))
        0    0.000             0.000          profile:0(profiler)

First of all, map(float, some_list_of_numbers) didn't seems to be
faster than [float(x) for x in some_list_of_numbers]. Maybe I did the
test in a wrong way ?

Secondly, [x+y for x,y in itertools.izip(xs, ys)] did go much faster
than map(lambda x,y: x+y, xs, ys). The latter is not only the slowest
one, but with an amazingly slow speed of 15 times slowdown.




More information about the Python-list mailing list