Python calculator and programmable filter

From KitwarePublic
Jump to navigationJump to search

Introduction

ParaView has two filters that give a user access to python math functions as well as the underlying VTK library. These are the Python Calculator and the Programmable Filter. This tutorial explains both.


Python Calculator

The Python Calculator allows a user to apply calculations that are available in Python. These include such functions as get volume of cells, get area of cells, get the cross product, dot product, curl, etc. The whole formula must fit on one line. Lets go over the Python Calculator with an example.

Lets create a new point variable, loaded with 5.
  • Open can.exo. All variables on. Apply.
  • Filters → Alphabetical → Python Calculator.
  • Change Expression to 5, Array Association as Point Data, and change Array Name to Calculated Variable. Apply.
  • Paint by Calculated Variable.


Lets create a variable equal to two times displacement.
  • Change Expression to DISPL*2. Apply.
  • Go to last timestep. Rescale to Data Range.


A more complete way to access displacement data. We explicitly pull DISPL from the first input to the filter. Don't forget that the first input is [0], the second is [1], etc.
  • Change Expression to (inputs[0].PointData['DISPL'])*2. Apply. Rescale to Data Range.


Here is how to multiply a vector by a global variable. Lets multiply DISPL by the timestep (TMSTEP)
  • Change expression to inputs[0].FieldData['TMSTEP'][0] * DISPL. Apply. Rescale to Data Range.


To recap, we are pointing to the first input to the filter, point data, the DISPL variable. We are multiplying by 2, and putting this into the Calculated Variable var.
If a function requires a variable input, use the string above. If the function needs the input mesh, use inputs[].
Notice that Python uses square brackets for arrays, and parenthesis to group parts of your formula. Also, to designate a string (i.e., variable name), use either single or double quotation marks.


A common request is to return a cell volume. It is done like this:
  • Change expression to volume(inputs[0]). Apply. Rescale to Data Range.


Numerous functions can be combined in one expression. For instance, a nonsensical expression would be to take a sin of a divergence of a curl. Here is the expression:
  • Change expression to sin(divergence(curl('ACCL'))). Apply. Rescale to Data Range.


Interesting functions available through the Python Calculator:
  • area(dataset)
  • aspect(dataset)
  • cos(array)
  • cross(X,Y) where X and Y are two 3D vector arrays
  • curl(array)
  • divergence(array)
  • dot(a1,a2)
  • eigenvalue and eigenvector(array)
  • gradient(array)
  • max(array)
  • mean(array)
  • min(array)
  • norm(array)
  • sin(array)
  • strain(array)
  • volume(array)
  • vorticity(array)
The complete list can be found in the ParaView Guide, http://www.paraview.org/paraview-guide/

Programmable Source

The Programmable Source is used to either be a source for new data, or to read in data from a file. It is written in Python. Later, filters will be used to modify this data. There is an extensive writeup on the Programmable Source in the ParaView Guide, http://www.paraview.org/paraview-guide/, in chapter 13.2. An example of a Programmable Source would be reading a .csv file into ParaView, once again from the ParaView Guide.

  • Create a .csv file named data.csv. Place the following into this file.
x coord, y coord, z coord, scalar
0,0,0,0
1,0,0,1
0,1,0,2
1,1,0,2
-0.5,-0.5,1,4
-0.5,-0.5,1,5
-0.5,-0.5,1,6
-0.5,-0.5,1,7
  • Sources → Programmable Source
  • Change Output Dataset Type to vtkTable
  • Enter the following into the Script window:
import numpy as np
data = np.genfromtxt("c:/.../data.csv", dtype=None, names=True, delimiter=',', autostrip=True)
for name in data.dtype.names:
array = data[name]
output.RowData.append(array,name)
  • Filters → Table to Points filter. X Column == x_coord, Y Column == y_coord, Z Column == z_coord. Apply.
  • Select the 3d view, and turn on visibility.
Example Programmable Sources in the book are:
  • Reading a CSV file
  • Reading a CSV file series
  • Reading a CSV file with particles
  • Reading binary 2D image
  • Helix source
Programmable Sources are a quick, easy way to prototype readers.

Programmable Filter

The Programmable Filter is used to modify data in the pipeline using Python. There is an extensive writeup about the Programmable Filter in chapter 13.3. Another good source of information is the Programmable Filter wiki page, here: https://www.paraview.org/Wiki/Python_Programmable_Filter. An example of a Programmable Filter dividing a variable by 2:

  • Divide ACCL by 2
    • Open can.exo. All variables on. Apply.
    • Filters → Alphabetical → Programmable Filter. Leave Output Data Set Type as Same as Input.
    • Enter the following into the Script window:
input0=inputs[0]
dataArray=input0.PointData["ACCL"]/2.0
output.PointData.append(dataArray, "ACCL_half")
    • Paint by ACCL_half
  • Subtract two datasets from each other. Use two instances of disk_out_ref.exo as two datasets, subtract GaMe3 from AsH3.
    • Open disk_out_ref.exo. All variables on. Apply.
    • Open disk_out_ref.exo. All variables on. Apply.
    • Highlight both datasets. Use the <ctrl> key with the mouse to select more than one input.
    • Filters → Alphabetical → Programmable Filter. Leave Output Data Set Type as Same as Input.
    • Enter the following into the Script window:
v_0 = inputs[0].PointData['AsH3']
v_1 = inputs[1].PointData['GaMe3']
output.PointData.append(v_1 - v_0, 'difference')
    • Paint by difference
  • Create a tensor
from paraview.vtk.numpy_interface import dataset_adapter as dsa
import numpy
def make_tensor(xx,yy,zz, xy, yz, xz):
t = numpy.vstack([xx,yy,zz,xy, yz, xz]).transpose().view(dsa.VTKArray)
t.DataSet = xx.DataSet
t.Association = xx.Association
return t
xx = inputs[0].PointData["sigma_xx"]
yy = inputs[0].PointData["sigma_yy"]
zz = inputs[0].PointData["sigma_zz"]
xy = inputs[0].PointData["sigma_xy"]
yz = inputs[0].PointData["sigma_yz"]
xz = inputs[0].PointData["sigma_xz"]
output.PointData.append(make_tensor(xx,yy,zz,xy,yz,xz), "tensor")
  • Subtract two timesteps from each other
    • Example from Cory Quammen
    • I think the Force Time filter will come in handy here for this use case. It can be used to "freeze" a dataset at a certain timestep.
      • Load can.ex2
      • Add a Force Time filter. Set the timestep to the one whose displacement you want to be zero. The output of this filter is the

can.ex2 dataset "frozen" at the chosen timestep. It will not change as you advance through time.

      • Select both can.ex2 (using the <ctrl> key) and the ForceTime1 sources in the Pipeline Browser. Note that the original can.ex2 source will update as the

timestep changes, but the ForceTime1 source will not change.

      • Add a Python Calculator filter to substract the displacement in the "frozen" data set from the current timestep in can.ex2.
      • Set the expression to
inputs[0].PointData['DISPL'] - inputs[1].PointData['DISPL']
      • The order of inputs into the Python Calculator is not well defined, so you may need to swap the indices in inputs[0] and inputs[1] to get the

correct sign on the result, but this should work.

      • (If desired,) Finally, add a Plot Selection over Time filter. This filter will run over all time steps, subtracting the data from the current timestep

from the data in the "frozen" timestep produced by the Force Time filter, and plot the result in a graph. The first timestep should have a value of 0.


  • Examples

principleStrainPython

Since the Programmable Source and Programmable Filters work at the server level, paraview.simple cannot be loaded or used.

Documentation(from an e-mail from Utkarsh)

    • Chapter 14 from the ParaView Guide.
http://www.paraview.org/paraview-guide/
    • And the following series of Berk's blog posts:
https://blog.kitware.com/improved-vtk-numpy-integration/
https://blog.kitware.com/improved-vtk-numpy-integration-part-2/
https://blog.kitware.com/improved-vtk-numpy-integration-part-3/
https://blog.kitware.com/mpi4py-and-vtk/
https://blog.kitware.com/improved-vtk-numpy-integration-part-4/
https://blog.kitware.com/improved-vtk-numpy-integration-part-5/
    • Wiki page on the Programmable Filter
http://www.paraview.org/Wiki/Python_Programmable_Filter

Acknowledgements

Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA-0003525.