Slight mistake in docs "Multivariate data interpolation on a regular grid" (original) (raw)

Hello,

I think I spotted a slight mistake in the docs here.

The CartesianGridInterpolator from Johanness Buchner is told to be an “almost drop-in” replacement for RegularGridInterpolator for N-D images. I have some code written with RGI, so I tried to move to the proposed implementation.

self.limits = np.array([[min(x), max(x)] for x in points]) (line 1 of init)

does not mirror the behavior of RGI. The correct implementation of RGI’s behavior, is, I think, something like this :

self.limits = np.array([[x[0], x[-1]] for x in points])

My correction might be wrong, I’m by no means an experienced dev or an interpolation expert, but I am pretty sure that the current proposed implementation is wrong and can mislead the user.

BTW, thank you to you all for the excellent toolkit that is scipy

ev-br March 19, 2025, 12:26pm 2

Hi,

It’s great that you’re taking time to improve the example, much appreciated!

Now, the example is definitely a quick untested demo, it very well might be faulty. Could you explain a bit more though, what is wrong with the original version, and how yours fixes the problem? Here or in a pull request to the scipy repository.

Cheers,

Evgeni