Bug#497448: ITP: python-griddata -- Python function to interpolate irregularly spaced data to a grid
On Tue, Sep 02, 2008 at 07:01:08PM +0200, Michael Hanke wrote:
> On Tue, Sep 02, 2008 at 04:31:55PM +0100, Chris Walker wrote:
> > Michael Hanke <michael.hanke@gmail.com> writes:
> >
> > > * URL : http://code.google.com/p/griddata-python/
> > > * License : MITish
> >
> > That URL claims the licence is "GNU General Public License v2"
> Yeah, but not all of it ... but read on.
>
> > > This module provides a single function, 'griddata', that fits a surface
> > > to nonuniformly spaced data points. It behaves basically like its equivalent
> > > in Matlab.
> >
> >
> > How does this relate to the algorithm mentioned in
> > http://lwn.net/Articles/292979/ where it says:
> >
> > "Jeffrey Whitaker has added support for gridding irregularly spaced
> > data using the Matlab (TM) equivalent griddata function. This is a
> > long-standing feature request for matplotlib and a major
> > enhancement. matplotlib now ships with Robert Kern's delaunay
> > triangularization code (BSD license), which supports the default
> > griddata implementation, but there are some known corner cases where
> > this routine fails. As such, Jeff has provided a python wrapper to
> > the NCAR natgrid routines, whose licensing terms are a bit murkier,
> > for those who need bullet proof gridding routines. If the NCAR
> > toolkit is installed, griddata will detect it and use it. See
> > http://matplotlib.sf.net/matplotlib.mlab.html#-griddata for details.
> > Thanks Robert and Jeff."
> That seems to be exactly the same -- great! The griddata package would
> have contained NCAR, but under this circumstances it is not really
> necessary, IMHO.
>
Further examination of
http://matplotlib.sf.net/matplotlib.mlab.html#-griddata suggests that
NCAR is not actually provided with matplotlib - so if you can surmount
the murkey licence problems (which presumably exist in your package
anyway) it would be useful to have in Debian (mind you if you can
surmount those, perhaps it can be in matplotlib proper).
If you do package it, then making sure it is available to matplotlib
would be nice.
Chris
Reply to: