mbconvert: Cell array GLOBAL_ID with 1 components, has only 91215 tuples but there are 91247 cells

[Nico Schlömer]

When using

mbconvert in.h5m out.vtk

with this input file, ParaView will refuse to open out.vtk with the error message

ERROR: In /home/kitware/Dashboards/MyTests/NightlyMaster/ParaViewSuperbuild-Release/paraview/src/paraview/VTK/Common/DataModel/vtkDataSet.cxx, line 446
vtkUnstructuredGrid (0xa959f0): Cell array GLOBAL_ID with 1 components, has only 91215 tuples but there are 91247 cells


ERROR: In /home/kitware/Dashboards/MyTests/NightlyMaster/ParaViewSuperbuild-Release/paraview/src/paraview/VTK/IO/Parallel/vtkPDataSetReader.cxx, line 923
vtkPDataSetReader (0x386ca00): Attribute Mismatch.


ERROR: In /home/kitware/Dashboards/MyTests/NightlyMaster/ParaViewSuperbuild-Release/paraview/src/paraview/VTK/Common/ExecutionModel/vtkExecutive.cxx, line 784
vtkPVCompositeDataPipeline (0x38433a0): Algorithm vtkFileSeriesReader(0x383fcd0) returned failure for request: vtkInformation (0x3668800)
  Debug: Off
  Modified Time: 161469
  Reference Count: 1
  Registered Events: (none)
  Request: REQUEST_DATA
  FORWARD_DIRECTION: 0
  FROM_OUTPUT_PORT: 0
  ALGORITHM_AFTER_FORWARD: 1

[Iulian Grindeanu]

an mbsize on the file says this:

 mbsize hh.h5m 
File hh.h5m:
   type  count   total            minimum            average                rms            maximum           std.dev.
------- ------ ------- ------------------ ------------------ ------------------ ------------------ ------------------
    Tet  91215     0.2         1.5595e-07         2.1926e-06          2.354e-06         6.2096e-06         8.5677e-07
1D Side 547290 1.5e+04          0.0062461           0.028143           0.028817           0.050842          0.0061952
 Vertex  21218

if I convert to vtk I can still view it with visit

paraview indeed complaints

the issue I think are the one node elements; there are 91215 tetras and (91247-91215) one-node elements

the one-node elements do not have global IDs and paraview is more picky about it. Maybe it is not a correct vtk file (although visit thinks it is )

[Vijay M]

Was this fixed or resolved ?

[Nico Schlömer]

I just tested with the version of the day and it's still buggy. MeshIO does the correct thing, perhaps its output helps debugging.

[Iulian Grindeanu]

it was not resolved :(

[Iulian Grindeanu]

So this pull request
https://bitbucket.org/fathomteam/moab/pull-requests/97/read-and-write-free-nodes-in-vtk-format/diff

introduced this unwanted behavior for paraview;

We need to discuss what is the correct solution to make everybody happy.

So some users want to see free nodes (cloud points?). Paraview does not like cells without data. If we create dummy one-node elements, we have to create dummy data too, for them. It is not a good solution, as other readers might interpret this as good data (even paraview). When we read those vtk files ourselves, with our vtk moab reader, we know to skip the dummy elements (we skip them also because MOAB does not have the concept of one-node elements).

Another solution would be to add a vtk write option, write the one-node dummy cells only when we provide that option. (and know that paraview will not like that file).

Or maybe have the option and also add dummy data too when that is triggered

[Nico Schlömer]

Paraview does not like cells without data.

What do you mean by that? Cell data beyond connectivity data?

[Iulian Grindeanu]

yes, cell data for global id and partition in this particular case