Reshape Table

This example will demonstrate how to reshape an input table as though it were a 2D array.

This filter will take a vtkTable object and reshape it. This filter essentially treats vtkTable`s as 2D matrices and reshapes them using `numpy.reshape in a C contiguous manner. Unfortunately, data fields will be renamed arbitrarily because VTK data arrays require a name.

This example demos PVGeo.filters.ReshapeTable

import numpy as np
import vtk
from vtk.numpy_interface import dataset_adapter as dsa
import PVGeo
from PVGeo.filters import ReshapeTable

Create some input table

t0 = vtk.vtkTable()
# Populate the tables
arrs = [None, None, None]
n = 400
ncols = 2
nrows = int(n * len(arrs) / ncols)
titles = ('Array 0', 'Array 1', 'Array 2')
arrs[0] = np.random.random(n)
arrs[1] = np.random.random(n)
arrs[2] = np.random.random(n)

t0.AddColumn(PVGeo.convert_array(arrs[0], titles[0]))
t0.AddColumn(PVGeo.convert_array(arrs[1], titles[1]))
t0.AddColumn(PVGeo.convert_array(arrs[2], titles[2]))

Use the filter to reshape the table

order = 'F'
newtitles = ['Title %d' % i for i in range(ncols)]
output = ReshapeTable(order=order,
                      ncols=ncols,
                      nrows=nrows,
                      names=newtitles).apply(t0)

Check the output against NumPy

wpdi = dsa.WrapDataObject(output)
tarr = np.zeros((nrows, ncols))
for i in range(ncols):
    tarr[:,i] = wpdi.RowData[i]
arrs = np.array(arrs).T
arrs = arrs.flatten()
arrs = np.reshape(arrs, (nrows, ncols), order=order)
assert(tarr.shape == arrs.shape)
assert(np.allclose(tarr, arrs))