From bug-request at octave dot org Thu Jan  5 05:54:06 2006
Subject: Re: interp1.m (octave-forge) *linear extrapolation
From: Olaf Till <olaf dot till at uni-jena dot de>
To: bug at octave dot org
Date: Thu, 5 Jan 2006 03:57:11 -0600

Sorry, but there is a further issue:

in the same lines of code cited below, line 134

idx = floor(t);

seems not to allow extrapolation (I get 'invalid row index 0' from
octave when it evaluates line 139.

Adding (in line 135)

idx = max (idx, 1); idx = min (idx, nx);

seems to solve this problem for me.

But a remark: for my data, then, '*linear' seems to last almost twice
as long as 'linear'.

Olaf

On Thu, Jan 05, 2006 at 09:02:41AM +0000, olaf wrote:
> This is a bug in a component of octave-forge, please forward if this
> is not the correct mailing list. If someone wants to respond --- I am
> not subscribed to bug at octave dot org dot 
> 
> Function: interp1.m
> Author: Paul Kienzle
> Version: Date of last change documented in the file: 2002-01-23
> 
> Description:
> In method '*linear' the results are different from method 'linear'
> (despite of even spacing of x) when the difference between neighboured
> x's is not equal to 1.
> 
> Suggested fix:
> 
> in lines 131 to 139:
> 
>   elseif strcmp(method, '*linear')
>     ## find the interval containing the test point
>     t = (xi - minx)/dx + 1;
>     idx = floor(t);
> 
>     ## use the endpoints of the interval to define a line
>     dy = [y(2:ny,:) - y(1:ny-1,:); zeros(1,nc)];
>     s = (t - idx)./dx;
>     yi(range,:) = s(:,ones(1,nc)).*dy(idx,:) + y(idx,:); 
> 
> there is a doubled division by `dx' (in line 133 and in line 138).
> Changing line 138
> 
>     s = (t - idx)./dx;
> 
> to
> 
>     s = t - idx;
> 
> makes method '*linear' return the same results as 'linear' even with
> non integer x's for me.
> 
> Regards, Olaf



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------