Dear Colleagues! How to add a new multi-component dataset (see figure) to the CellData in the Programmable Filter?
Below is an example. I take component 0 of multi-component dataset Mxx,Myy,Mxy,M1,M2 for some calculation. I would like dataset K to have 3 components too, combined from K1, K2, K3 (in order to work with it as shown in the figure)

Dear @mwestphal , thanks for the help!
I managed to come up with the following algorithm

# Import NumPy
import numpy as np
# VTK data access
input0 = inputs[0]
M = input0.CellData['Mxx,Myy,Mxy,M1,M2']
# Conversation to NumPy, where vtk_index is the index
# of the numerical data array in the VTK object M in the considered model
# (this may be a different vtk_index in different models)
vtk_index = 1
data = np.array(M.Arrays[vtk_index])
rows = data.shape[0]
print(data.shape)
# Performing calculations on the example of the first column of data
k1 = (data[:, 0] * 0.5).reshape(rows, 1)
k2 = (data[:, 0] * 0.25).reshape(rows, 1)
k3 = (data[:, 0] * 0.2).reshape(rows, 1)
print(k1.shape, k2.shape, k3.shape)
k_data = np.hstack((k1, k2, k3))
print(k_data.shape)
# Creating a new VTK object using an existing object as a template
K = M * 0
K.Arrays[vtk_index] = k_data
output.CellData.append(K, "K")
# As a result, a 3-component dataset "K" was created
# using a 5-component dataset "M" as a template (see figures)
# It is important that datasets "M" and "K" refer to the same cells

The question of how to manage the name of dataset components remains relevant. In the graphical interface, they are automatically named depending on the number of components:
X-Y for 2,
X-Y-Z for 3,
XX-YY-ZZ-XY-YZ-XZ for 6
0-1-2-3-4 for another number of components.