LISTS: Extract every other element

[Charles G Waldman]

Skip Montanaro writes:
 > 
 >     Randall> I want to take a large list:
 >     Randall>   [ 1,2,3,4,5,6,7,... ]
 > 
 >     Randall> and build a list with every other element:
 >     Randall>   [ 1,3,5,7,... ]
 > 
 >     Randall> Is there a faster way than looping over indices?:
 > 
 > I'm pretty sure you can do this with NumPy.  I'm not a NumPy user though, so
 > I'll have to defer to the rocket scientists... ;-)
 > 

(Disclaimer: I may work at a government lab, but I'm not a rocket
scientist)

You could do this:

>>> from Numeric import *
>>> a = array(mylist, 'O')
>>> b = take(a,range(0,len(a),2)

For the case where "mylist" is a list of numbers, this works fine:
>>> mylist = [3,1,4,1,5,9,2,6,5,3,6]
>>> a = array(mylist, 'O')
>>> b = take(a,range(0,len(a),2)
>>> b
array([3 , 4 , 5 , 2 , 5 , 6 ],'O')


It *should* work for general lists, using Numeric arrays of type 'O'
(Python Object).

However, in the course of writing and testing this I found a bug in
Numeric, where it mis-handles initialization of arrays of type 'O'
(Python Object).

>>> from Numeric import array
>>> l = ['abc','def','ghi']
>>> array(l,'O')
array([[abc , abc , abc ],
       [def , def , def ],
       [ghi , ghi , ghi ]],'O')

Oops!  That should have been

array([abc, def, ghi], 'O')


So, in fact, you *could* use Numeric to do this, if it weren't
broken.  Sigh.  I'm taking this thread over to the Matrix-SIG.