Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!samsung!munnari.oz.au!manuel!csc.anu.edu.au!bdb112
From: bdb112@csc.anu.edu.au
Newsgroups: comp.lang.idl-pvwave
Subject: Re: How can I integrate? (not so easy?)
Message-ID: <1991Mar26.071841.1@csc.anu.edu.au>
Date: 25 Mar 91 20:18:41 GMT
References: <1991Mar20.233728.1@csc.anu.edu.au>
Sender: news@newshost.anu.edu.au
Organization: Computer Services, Australian National University
Lines: 18

In article <1991Mar20.233728.1@csc.anu.edu.au>, bdb112@csc.anu.edu.au writes:
> How can I efficiently do a running sum of a vector (integrate)?  The explicitly
> coded version (IDL v2)  
> for i=1,n-1  x(i)=x(i) + x(i-1) achieves the desired effect, but takes (VS3100)
> 1 second for a 2000 element array compared to .01 sec for z=x+x, a similar
> number of operations.  This is such an obvious thing to do that I must be
> missing something obvious - I can't find it in the userlib either (just
> deriv).

Two respondents have suggested 
sum_x = total(x(0:i))      ; where i is the maximum idex of the array
When I try this, I get a scalar result, rather like a definite integral, when
what I want is a vector which is a running total, like an indefinite integral.

e.g. x = [1,1,2]   sum_x = [1,2,4]
Of course, the next step would be to allow an optional vector which contained
the abscissae, if not equally spaced, or the spacing if equally spaced, but I
would be happy with the simple result.