GitLab

Summary

A Cell Centers filter applied to time-dependent cell data does not update for a different grid geometry. Tested on ParaView 5.8.1-1220-g186faf45 using macOS 10.15.6.

Example

1. Load attached cell_center_noupdate.xdmf2 and load all data.
2. Color by I and advance in time. The cell geometry updates correctly and the coloring also updates correctly. That is, timestep 0 has a 3x3x1 grid and timestep 1 has a 3x3x2 grid.
3. Go to timestep 0.
4. Apply Cell Centers filter, choose 3D Glyphs representation. 3x3 arrows oriented along the $x$ axis should be visible (perhaps colored by I). This is correct.
5. Advance in time.

As a result of advancing in time, I've seen freezes, crashes, or the wrong number of misoriented vectors.

One can also go to timestep 1, apply Cell Centers, glyphs, etc., and observe the vectors correctly but then moving to timestep 0 causes issues (generally displaying six glyphs of various shading including NaN values).

Notes

• My interest is in applying this to vector data; however, for simplicity scalar data is used here. The same sorts of behavior occurs for vector data.
• The "Vertex Cells" toggle in the Cell Centers filter appears to have no effect on this (mis)behavior.
• Avoiding the Cell Centers filter and explicitly creating a grid with the correct number/position of nodes does update correctly, which is what leads me to believe this is a Cell Centers filter issue.

CC: @ethan.stam @patchett2002

cell_center_noupdate.xdmf2

Unclassified per ADC:jmc