In-situ visualisation in Paraview

You will need to have Paraview installed. On Ubuntum, this can be done through

sudo apt install paraview

Peano's postprocessing scripts use Python 3. This means you need a Paraview version with Python 3 support (oldish Paraview installations only support Python 2.7, and some Paraview downloads don't come along with Python support at al).

You can display Peano's patch files directly within Paraview. Again, there are different ways. For all of them, Peano's Python environment has to be available within Paraview:

      export PYTHONPATH=Peano-dir/src

Start Paraview's Python environment

You can either manipulate all data directly within Paraview or use Paraview's Python terminal to convert data as a batch from Peano's native format into native Paraview data. I recommend the direct Paraview terminal route to interactively explore data. The actual data processing (of time-dependent data) however can be very slow. Once I know what information I need, I thus usually switch to Paraview's Python interpreter to convert all data into Paraview's native file formats in one rush.

For the batch mode, open the terminal pvpython. For the interactive GUI mode, open Paraview and activate the Python Shell by clicking the check box in the View menu.

Both visualisation routes first require you to load Peano's Python tools - either within the pvpython terminal or the Paraview Python window:

      import tarch.plotter.api

Work with a whole dataset

Peano provides a Python class tarch.plotter.api.Visualiser which allows you to navigate through the whole data.

import tarch.plotter.api
visualiser = tarch.plotter.api.Visualiser( ”solution-Euler.peano-patch-file” )
visualiser.display()
visualiser.append filter(tarch.plotter.api.ExtractFineGridFilter())

The snippet displays the first dataset (time step) from a  data dump. Via

visualiser.select dataset(10)
visualiser.reload()

you can, for example, select the 11th data set and display this one.

Running through data sets (time steps) manually is cumbersome in many cases. We have not yet managed to integrate our Python-based, bespoke postprocessing into Paraview's video replay features, and we also experience that it is relatively slow. In such a case, it is more convenient to convert all data in one rush into Paraview's native, binary file format vtu:

visualiser.write vtu time series()

Don't use pvpython for this. Converting a whole time series to vtu this way runs through the exact same code as the standalone render.py script, and pvpython's bundled VTK/ParaView Python has repeatedly crashed on us doing this - anything from a BrokenProcessPool error to a silent segmentation fault with no Python traceback. Prefer running render.py viaplain python3" for bulk/batch conversion instead of triggering it from here; only use this pvpython route if you specifically need to stay inside a live Paraview session.

Once this routine terminates, you'll find a file with the extension .pvd in your directory which you can open and play as video. Obviously, this conversion can be triggered in Paraview's Python interpreter which is typically faster than Paraview's built-in terminal.

You can trigger the Python postprocessing routines while the simulation is running.

Display a single file

Load the file you are interested in via

data = tarch.plotter.api.render single file( ”solution-Euler-tree-0-0.peano-patch-file”, identifier=”MyQName” )

This is only one snapshot file, i.e. one file written by one thread in one step. We can display the file with

tp = TrivialProducer() 
tp.GetClientSideObject().SetOutput(data) 
Show(tp)

If we write large parallel files or time series, Peano will typically create one file like solution-Euler.peano-patch-file which links to all the files such as solution-Euler-tree-0-0.peano-patch-file. The meta file (with the links) is plain text, so you can study it via a text editor. Instead of picking a particular file, you can display a particular data set (snapshot) from the meta file:

import tarch.plotter.api
data = tarch.plotter.api.render dataset( ”solution-Euler.peano-patch-file”, 
                                            display_as_tree=False,
                                            filter=[tarch.plotter.api.ExtractFineGridFilter()],
                                            dataset_number=0, 
                                            identifier=”MyQName” )
tp = TrivialProducer()
tp.GetClientSideObject().SetOutput(data)
Show(tp)

Filters

Peano's snapshots are relatively complex. In practice, you typically want to filter data before you pipe it into Paraview. The most popular filter is the one which eliminates the coarse grid levels: Peano holds all data in multiple resolution. If your code does not exploit this feature, you might not be interested in the coarser resolutions, so removing it makes sense.

There are more filters or you can write your own custom filters. Study the source code for further details.