Transferring PolyData Faces into a Numpy Array

A fast way to transfer the face data, as stored in the Polydata object, into a Numpy object array containing the indexes of all the points of each simplex.

This should be a fairly simple for loop in C, I would add it myself, but I A) don’t know C yet, and B) don’t know how to integrate C and Python. However, the pseudocode in Python could look something like this:

simplexes = []
index = 0
while index < face_array.shape[0]:
    number_of_points = face_array[index]
    points = face_array[index+1:index+1+number_of_points]
    simplexes.append(points)```

[darikg]

It’d definitely be nice to have an official solution for this.

Maybe:

from typing import List
from pyvista import _vtk, PolyData
from numpy import split, ndarray

def face_arrays(mesh: PolyData) -> List[ndarray]:
    cells = mesh.GetPolys()
    c = _vtk.vtk_to_numpy(cells.GetConnectivityArray())
    o = _vtk.vtk_to_numpy(cells.GetOffsetsArray())
    return split(c, o[1:-1])

[banesullivan]

For meshes with uniform cell types (all tirangles or all quads), we could modify @darikg’s to the following wich would return an ndarray with int64 dtype

from typing import List
from pyvista import _vtk, PolyData
from numpy import split, ndarray

def face_arrays(mesh: PolyData) -> List[ndarray]:
    cells = mesh.GetPolys()
    c = _vtk.vtk_to_numpy(cells.GetConnectivityArray())
    o = _vtk.vtk_to_numpy(cells.GetOffsetsArray())
    faces = split(c, o[1:-1])
    if len(np.unique(np.diff(o))) == 1:
        return np.array(faces)
    return faces

>>> import pyvista as pv
>>> face_arrays(pv.Sphere()).dtype
dtype('int64')