Re: [OctDev] interp2.m

[Paul K. <pki...@us...>]

On Jun 12, 2005, at 10:00 PM, Tom Holroyd wrote:

>> The bug is in the documentation for not saying that the function is 
>> restricted to meshgrids :-)
>
> Ah, yeah, I was afraid of that.  :-)
>
> It actually interpolates each point correctly, so one workaround is to 
> just call it repeatedly for each point.  That's a little bit slow, but 
> if it only has to be done once it's almost tolerable.  I might get 
> around to it fixing it eventually but it's a bit low on my list at the 
> moment.

On second thought, it is not so bad.  Rather than using (i,j) indexing, 
which grabs regular submatrices, use (i+(j-1)*m), which grabs 
individual elements.  See attached.

Note that unlike interp2, calling interp2fi with vectors xi,yi will not 
automatically form a meshgrid.

Note also that x,y are still constrained to meshgrids.  Working around 
that will require griddata.

- Paul