Far Away Grid Points

The following example computes the maximum distance between grid points, and it looks unusually large to me! I noticed it when testing Hirshfeld weights (instead of Becke weights) for integration. They go unnoticed when using Becke weights for integration. I haven't checked whether that is the case for HORTON2, but we need to add tests for grid points coordinates and weights against HORTON2.

import numpy as np
from scipy.spatial.distance import pdist

from grid.molgrid import MolGrid
from grid.becke import BeckeWeights
from grid.onedgrid import GaussChebyshev
from grid.rtransform import BeckeTF

atnums = np.array([8, 1, 1])
atcoords = np.array([[0.0, 0.0, 0.0], [0.0, 0.0, 1.79523983], [1.69257101, 0.0, -0.59840635]])

oned = GaussChebyshev(30)
rgrid = BeckeTF(1.e-4, 1.5).transform_1d_grid(oned)
grid = MolGrid.horton_molgrid(atcoords, atnums, rgrid, 110, BeckeWeights())

# compute pairwise distance between grid points
pdist = pdist(grid.points)
print('min & max of pdist = ', np.min(pdist), np.max(pdist))

The output is: min & max of pdist = 0.0002980577 4377.9871590172

The issue does not name repository files; start by reproducing the shown MolGrid.horton_molgrid example with its GaussChebyshev, BeckeTF, and BeckeWeights setup. Compare the generated grid point coordinates and weights with HORTON2, then add tests covering the observed discrepancy.