I have made an NDArray version of the discrete laplace operator "del2",
and have attached it here. As this is a matlab core function. I wanted
to propose this for inclusion in Octave. However, although I can get the
same behavior as the 2D version in octave-forge, the edge points of the
operator differ from the matlab version. It appears that the matlab
version of "del2" calculates the interior points first and then
extrapolates the edge points from this calculation, whereas the
octave-forge version of del2 uses a nearest value approximation.
Should I keep the octave-forge behavior or try and duplicate the Matlab
behavior of del2?