Numerical Python question

[Alex Martelli]

David Mertz wrote:
   ...
> But in general, an average is 'sum(highs)/len(highs)'.  The version of
> 'sum()' in Numeric will work a lot faster though... although it makes no
> difference for 10 prices.  If you start worrying about a million prices,
> you might see a significant boost using Numeric.

No kidding: Numeric.sum is an *order of magnitude faster* for this:

[alex at lancelot pop]$ timeit.py -s'import Numeric' 
-s'x=Numeric.array(map(float,range(1000000)))' 'Numeric.sum(x)'
10 loops, best of 3: 2.11e+04 usec per loop

[alex at lancelot pop]$ timeit.py -s'import Numeric' 
-s'x=Numeric.array(map(float,range(1000000)))' 'sum(x)'
10 loops, best of 3: 3.05e+05 usec per loop

the difference between 21 milliseconds with Numeric.sum, and 300 with
the built-in sum, can be *very* significant if you sum millions of floats in
your bottlenecks.

I think the chapter on Numeric in "Python in a Nutshell" is a good intro
(thanks to lots of crucial input from Eric Jones and Paul DuBois!), and
you can read it for free with the usual trick -- get an O'Reilly safari
subscription and use the free first two weeks to read the parts that
interest you, then cancel the subscription so you won't have to pay.


Alex