[SciPy-user] organizing data

[Gael Varoquaux]

On Wed, Nov 21, 2007 at 08:45:22AM +0100, Gael Varoquaux wrote:
> If I understand properly, you have non-regular gridded data that you want
> to interpolate on a regular grid (though only through a cut) to plot as
> images. If I get it right, that does not sound an incredibly easy
> problem. There are algorithmes to do this, but I don't think there are
> wrapped in scipy. VTK but do this, but AFAIK, only in 3D.

I am stupid. Of course there are things in scipy !

First you can look at a Delaunay natural neighbor interpolation
(http://www.scipy.org/Cookbook/Matplotlib/Gridding_irregularly_spaced_data).
Unfortunately I think the algorithm in Scipy can only do this for 2D
data. Robert Kern will chime in, when he wakes up.

Second, you can use radial basis function for interpolation, see
http://www.scipy.org/Cookbook/RadialBasisFunctions
Unfortunately, you will have to rebuild scipy for this.


To everybody, is it possible to change the sandbox so that when scipy is
built, all the packages in the sandbox are built, but the build doesn't
complain if it fails for these (or you need a special switch to have it
complain). There are some really nice packages in the sandbox, but most
people cannot build them because they don't know how to build. IMHO it is
better to have the binary packages of scipy shipping with the fraction of
the sandbox that builds, and might not even pass tests, than to have
these packages simply not available for users who cannot build.

Cheers,

Gaël