Normals generation is wrong

[microy]

Hello,

the code used to generate the (vertex) normals gives a wrong result.

Original code :

def generateNormals(self):
    """If :attr:`normals` is `None` or you wish for normals to be
    recomputed, call this method to recompute them."""
    norms = numpy.zeros( self._vertex.shape, dtype=self._vertex.dtype )
    tris = self._vertex[self._vertex_index]
    n = numpy.cross( tris[::,1] - tris[::,0], tris[::,2] - tris[::,0] )
    normalize_v3(n)
    norms[ self._vertex_index[:,0] ] += n
    norms[ self._vertex_index[:,1] ] += n
    norms[ self._vertex_index[:,2] ] += n
    normalize_v3(norms)

    self._normal = norms
    self._normal_index = self._vertex_index

Problem :

    norms[ self._vertex_index[:,0] ] += n
    norms[ self._vertex_index[:,1] ] += n
    norms[ self._vertex_index[:,2] ] += n

Solution (naive implementation) :

   for i, f in enumerate( self._vertex_index ) :
         norms[ f ] += n[ i ]

The problem comes from the implementation norms[ self._vertex_index[:,0] ] += n where the face normal vector n is only added once per vertex (instead of a certain number).

Best regards !

[jterrace]

Can you clarify what you mean? These are per vertex normals.

[microy]

Yes but the implementation gives wrong result (as the vertex normal is the sum of the surrounding face normals).

The problem is the use of the integer indexing of the numpy array.