Wrong interpolated value when initialising with grid of ones

[nordam]

I'm trying out the tricubic interpolator, but I have trouble getting it to work. I have constructed the following example:

import numpy as np 
import ARBTools 
from ARBTools.ARBInterp import tricubic

Nx = 3
Ny = 5
Nz = 7

xc = np.linspace(-1, 1, Nx) # x coordinates
yc = np.linspace(-1, 1, Ny) # y coordinates
zc = np.linspace(-1, 1, Nz) # z coordinates

interpolator_array = np.zeros((Nx*Ny*Nz, 4))
ind = 0
for i in range(Nx):
    for j in range(Ny):
        for k in range(Nz):
            interpolator_array[ind,0] = xc[i]
            interpolator_array[ind,1] = yc[j]
            interpolator_array[ind,2] = zc[k]
            interpolator_array[ind,3] = 1.0 # All input values are 1.0
            ind += 1

f = tricubic(interpolator_array)
coords = np.zeros(3)

# Returned interpolated value is 0.0
f.Query(coords)[0]

Here, I pass in a value of 1.0 at all points, but the returned intepolated value at the point (0,0,0) is 0.0, not 1.0.

[nordam]

Since yesterday, I have been reminded that 3 points isn't enough to create a cubic spline, so of course each dimension has to have at least 4 points. With this change, it works as expected. I suppose a warning about invalid input would be handy.

[DurhamDecLab]

Hi, I'm extremely sorry that I did not respond to this issue, I didn't get a notification about it - I've only seen it now because I logged in to upload a slightly updated file.
I'm glad you were able to get things working. I'll look into what you report, the interpolator is designed to return "NaN" if you try to query a point outside the valid region. As you say, you need a minimum number of points to make a spline, in the case of the tricubic interpolator this means that we cannot interpolate any value within the outer layer of cuboid volume elements. I'll reproduce your problem and try and incorporate a warning.
Thank you for the bug report!