feat: choose the number of output vtu files

[untereiner]

[Algiane]

I am interested by this feature too: thanks for the PR.

[untereiner]

It works as it is but you should be aware that their is no correlation for time-dependent outputs. It means that:

• it is correct to see a property for each time step
• computing something between two timesteps is not ok because the merging can change the order of the elements.

It should be possible to keep an absolute order by outputting the localToGlobalMap attribute and use it in paraview for example (not tested yet)

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 23.25581% with 33 lines in your changes missing coverage. Please review.

Project coverage is 57.27%. Comparing base (9257440) to head (825ad47). Report is 65 commits behind head on develop.

Files with missing lines	Patch %	Lines
...mponents/fileIO/vtk/VTKPolyDataWriterInterface.cpp	0.00%	33 Missing :warning:

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2637      +/-   ##
===========================================
- Coverage    57.27%   57.27%   -0.01%     
===========================================
  Files         1143     1143              
  Lines        98791    98815      +24     
===========================================
+ Hits         56584    56596      +12     
- Misses       42207    42219      +12     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[untereiner]

To be more precise. It is not possible to keep the coherence of rank merging between each time step. This means that a cell can be in different multi blocks at each time step.

How to keep the consistency between the time steps, meaning: Get a properly from the same cell for each time step ? I propose to use localToGlobalMap to reference the cells (and access to properties) because it is stable and across time.

[herve-gross]

@untereiner is this ready for reviews? Can you explain why indices are shuffled across time steps?

[untereiner]

@herve-gross yes it is ready. The indices are not shuffled. Currently there is an isomorphism between the ranks in memory and the ranks written in a multiblock. There is not possible to enforce which ranks that will be merged. Then for each time step you will have different ranks that will be merged into blocks. So, if you go through the timesteps with paraview the resulting color map will be ok because it loads and renders each timestep independently. But we observed that computing a difference between two timesteps will result in incoherent results because filters are applied per block.

[paveltomin]

Waiting on code owner review from @MelReyCG, @rrsettgast, and/or @wrtobin.

[paveltomin]

requires rebaseline because of newly added numberOfTargetProcesses