Cost of raising exceptions.

[Les Schaffer]

On Tue, 29 May 2001 17:29:24 +0200, "Alex Martelli" <aleaxit at yahoo.com>
wrote:

>So: if we're having an exception about one time in 20, the
>cost of the exceptions-in-loop approach is about 20%. 

i had a similar question when i was doing lots of stats computations on
generic types.

i was curious how expensive it was to try/except vs. checking types.

so here's my little play script:

#!/usr/bin/python -O

# test whether exceptions or type testing take more time.
import time

x = 'NA'
xp = 1.000
y = 3.1415926

NMAX = 100000

def timeIt( func ):
    start = time.time()
    name = func()
    end = time.time()
    print '%s took %5.4f seconds'%(name, end - start)

    
def typeCheck():
    for i in range(NMAX):
        if i < CUTOFF:
            X = x
        else:
            X = xp
            
        if type(X) == type(y):
            z = X*y
        else:
            pass

    return 'typeCheck'

def tryExcept():
    for i in range(NMAX):
        if i < CUTOFF:
            X = x
        else:
            X = xp
            
        try:
            z = X*y
        except TypeError, e:
            #print e
            pass

    return 'tryExcept'

for FRACTION_THAT_ARE_NA in [ 0, .2, .4, .6, .8, 1]:
    print 'For fraction that are NA = %f:'%FRACTION_THAT_ARE_NA
    CUTOFF = FRACTION_THAT_ARE_NA * NMAX
    timeIt( typeCheck )
    timeIt( tryExcept )
    print

====== results ======
For fraction that are NA = 0.000000:
typeCheck took 0.4220 seconds
tryExcept took 0.2180 seconds

For fraction that are NA = 0.200000:
typeCheck took 0.4690 seconds
tryExcept took 0.6090 seconds

For fraction that are NA = 0.400000:
typeCheck took 0.4690 seconds
tryExcept took 0.9850 seconds

For fraction that are NA = 0.600000:
typeCheck took 0.4530 seconds
tryExcept took 1.3750 seconds

For fraction that are NA = 0.800000:
typeCheck took 0.4220 seconds
tryExcept took 1.7500 seconds

For fraction that are NA = 1.000000:
typeCheck took 0.3750 seconds
tryExcept took 2.0000 seconds