Programming :: Numpy 'filter' Type Function?

I'm a n00b to both numpy and python. I was wondering what the most intelligent way to write the trapezoid rule function was:

Code:
>>> import numpy as np
>>> a=np.array([1,2,3,4,5])
>>> def mytrapz(vec):
total=0
for i in xrange(0,len(vec)-1):
total+=(vec[i]+vec[i+1])/2.0
return total

>>> mytrapz(a)12.0 This is exactly what I would do in C and I'm sure the for loop is not the fastest way (as is the case for my interpreted langs). Yes, I know there is already numpy.mytrapz... I'm interested in the most efficient way to loop over arrays in such a way that each iteration accesses more than one cell.