Re: [Tutor] A faster x in S
I used the s.intersection(t) function in the set type as it was the most appropriate. The performance was phenomenal. Thank-you! Dinesh - Original Message - From: bob gailer To: Dinesh B Vadhia Cc: tutor@python.org Sent: Tuesday, January 15, 2008 2:03 PM Subject: Re: [Tutor] A faster x in S Dinesh B Vadhia wrote: For some significant data pre-processing we have to perform the following simple process: Is the integer x in a list of 13K sorted integers. That's it except this has to be done 100m times with different x's (multiple times). Yep, a real pain! I've put the 13K integers in a list S and am using the is 'x in S' function. I was wondering if there is anything faster? I agree with Kent. l = range(13000) s=set(l) d=dict(enumerate(l)) import time def f(lookupVal, times, values): .. st=time.time() .. for i in range(times): .. z = lookupVal in values .. return time.time()-st f(6499,1000,l) 0.3126376037598 f(6499,100,s) 0.3123623962402 So set is 1000 times faster than list! f(6499,100,d) 0.31300020217895508 And dict is (as expected) about the same as set. So 100,000,000 lookups should take about 30 seconds. Not bad, eh? Let's explore another angle. What range are the integers in (min and max)? Bob ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
[Tutor] A faster x in S
For some significant data pre-processing we have to perform the following simple process: Is the integer x in a list of 13K sorted integers. That's it except this has to be done 100m times with different x's (multiple times). Yep, a real pain! I've put the 13K integers in a list S and am using the is 'x in S' function. I was wondering if there is anything faster? Dinesh ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
Re: [Tutor] A faster x in S
Dinesh B Vadhia wrote: For some significant data pre-processing we have to perform the following simple process: Is the integer x in a list of 13K sorted integers. That's it except this has to be done 100m times with different x's (multiple times). Yep, a real pain! I've put the 13K integers in a list S and am using the is 'x in S' function. I was wondering if there is anything faster? Yes. Put the integers in a set and test for membership there. If for some reason the integers have to be in a list, and the list is sorted, use the bisect module to do a binary search, rather than the linear search used by 'x in S'. http://docs.python.org/lib/module-bisect.html Kent ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
Re: [Tutor] A faster x in S
Dinesh B Vadhia wrote: For some significant data pre-processing we have to perform the following simple process: Is the integer x in a list of 13K sorted integers. That's it except this has to be done 100m times with different x's (multiple times). Yep, a real pain! I've put the 13K integers in a list S and am using the is 'x in S' function. I was wondering if there is anything faster? I agree with Kent. l = range(13000) s=set(l) d=dict(enumerate(l)) import time def f(lookupVal, times, values): ... st=time.time() ... for i in range(times): ... z = lookupVal in values ... return time.time()-st f(6499,1000,l) 0.3126376037598 f(6499,100,s) 0.3123623962402 So set is 1000 times faster than list! f(6499,100,d) 0.31300020217895508 And dict is (as expected) about the same as set. So 100,000,000 lookups should take about 30 seconds. Not bad, eh? Let's explore another angle. What range are the integers in (min and max)? Bob ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
[Tutor] A faster x in S
Kent and Bob, Are you thinking of the first problem in Bentley's Programming Pearls? The original poster's questions sounds like it could be in that domain. http://netlib.bell-labs.com/cm/cs/pearls/cto.html So I agree: the next questions we probably should ask the original poster: * Why are you trying to search for a number in those sorted integers? * Is there anything characteristic about those sorted integers that might be peculiar or useful? Do the numbers have streaks? Are the integers large or small? ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor