Thanks. That indeed did work (after downloading python-dev package). I just didn't know that 'install' was the argument that I was supposed to pass to it =)

After installing I tried griddata again. My input data was originally a list. The new griddata didn't like this and so I simply used a=array(thelist) to convert to an array and tried again. It took quite a bit longer than the old griddata, but the resulting output now looks correct! Better a slow and correct answer than a fast and garbage one.

Thanks again Jeff (and thanks for the new griddata if you are the one that made it)! 


On Wed, Dec 21, 2011 at 5:55 AM, Jeff Whitaker <> wrote:
On 12/21/11 12:31 AM, Brad Malone wrote:
Hi, I'm still working on my interpolating from an irregularly space grid and then running pcolormesh on the resulting output. With some of the newer data I've been plotting I've noticed that my plots are complete garbage. I realized that this was actually because of the output from griddata rather than some problem with pcolormesh/pcolor/etc (basically I get huge negative values like -80000 from the interpolation when all of my data points lie within [0,20]) . 

Googling I found out that the default griddata has some problems, and that there is a better, more robust version available through natgrid.  I downloaded the natgrid-0.2.1 package from here

My question now is, how do I install this and give it a shot? I'm running on Ubuntu (or Xubuntu rather). The README doesn't seem to have any directions.

python install should do it.  matplotlib will automatically use it if it's installed.


Also, let's say that this new griddata doesn't work for me, is there something else I could try? The interpolation problems are strange, because I can break my data into 3 segments (I read 3 files to obtain the data so this is the natural way to do it) and I can plot and interpolate correctly any segment individually. It's only when I do all 3 segments together that the interpolation begins to fail. 

Any ideas?

Thanks for the continued help!


Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity.

Matplotlib-users mailing list