Using numpy arrays in Paraview programmable filter

Question

How can I access a field in Paraview's programmable filter as a numpy array?

I want to:

• Import an existing field as a numpy array
• Create a similar array for output
• Register it as a new field for visualization

I'm familiar with the wiki, but it doesn't address the issue.

Given this example:

elev0 = inputs[0].PointData['Elevation']
elev1 = inputs[1].PointData['Elevation']
output.PointData.append(elev1 - elev0, 'difference')

I'd like to do (assuming elev0 and elev1 are numpy arrays):

diff = elev1 - elev0
diff[ diff > 1.0 ] = 1.0
output.PointData.append(diff, 'difference')

but this doesn't work as those are VTKArray objects.

Answer 1

With ParaView 4.2, your original code will work as expected. The following is now acceptable.

output.PointData.append(numpyarray, "name")

Answer 2

Reading the Paraview python API, found the following solution to convert back and forth between VTKArray and numpy arrays. This uses the numpy_support and vtk.dataset_adapter modules :

from paraview.numpy_support import vtk_to_numpy
from paraview.vtk.dataset_adapter import numpyTovtkDataArray, vtkDataArrayToVTKArray
import numpy as np

# get paraview.vtk.dataset_adapter.VTKArray object
vtkArr = inputs[0].PointData['Elevation']
# convert to numpy array, shape is (nNodes,) for scalars, (nNodes,nComponents) for vectors
npArr = vtk_to_numpy(vtkArr)
# do math on a copy so that original field is not altered
npArr2 = npArr.copy()
npArr2 *= 2.0
# convert back to VTKArray
vtkArr2 = vtkDataArrayToVTKArray(numpyTovtkDataArray(npArr2))
# append new field to output
output.PointData.append(vtkArr2, 'New field')