Visualisation

Table of Contents

• Writing outputs through the tarch
• Tweaking your pictures (for developers)

• Explore data with PyVista
• Offline (command-line) conversation into VTK file formats
• In-situ visualisation in Paraview
• Peano patch files
• Probing data files

Peano writes different types of files: Some history data (point probes) yield plain csv files, particles typically can dump into VTK files directly, which you can open with tools such as PyVista, Paraview or VisIt. The majority of volumetric Eulerian data in Peano however is dumped in a bespoke Peano data file format. This data format is also used to display "only" meshes or domain decomposition.

Writing outputs through the tarch

Todo

Yet to be written

Tweaking your pictures (for developers)

Peano's default VTK plotters project Peano's tree-based grid onto an unstructured mesh, as VTK does not support tree meshes. The underlying mechanism is straightforward:

1. If a vertex is touched/read the first time, it is dumped as vertex of the unstructured output mesh if it is not refined, i.e. if no other vertex does exist at the same location on a finer mesh level.
2. If a hanging vertex is created, we add a new vertex to the output dump if this hanging vertex has not been written before. The plotters typically maintain one big hash map to bookkeep which hanging vertices have already been written.
3. If the tree traversal runs into an unrefined cell, this cell is made a cell of the unstructured output mesh.

The block format/the conversion tools in contrast generate all the points multiple times. If you have to cells sharing a vertex, the vertex is dumped twice. For patch-based problems, this is okish, as the number of vertices that are shared is way smaller than the total number of vertices in the domain (vertices within a patch are not dumped multiple times). As a consequence of the redundant vertex dumps, isosurfaces, e.g., might not work. If you need them, it is reasonable to project/map the actual data onto a regular mesh first.

If your PDE has $m$ quantities, Peano's block format dumps them into a vector of cardinality $m$ per vertex or cell, respectively. Most visualisation filters/tools however require data to be either scalar or vector data. In Paraview, use the calculator filter to extract only some quantities from the data dump. Alternatively, you can ask the convert tool to extract only some data into the vtu output.

Please note that most visualisation codes (such as Paraview) do interpolate bi-/tri-linearly within the cells. If a cell's face is adjacent to cells on finer levels, hanging vertices on the face exist. If they do not hold the linearly interpolated value, you will discover visualisation artifacts.

For smooth output pictures of meshes with adaptivity, you have to

• set the plotted properties on hanging vertices due to $d$-linear interpolation, and
• you have to add the plotter mapping after you have invoked the mapping that does initialise the hanging vertices.

Please note that the dumped unstructured mesh is a non-conformal mesh. Algorithms such as isosurface identification thus might yield invalid results - it depends on your visualisation algorithms.