[vtk-developers] Tuple ordering of 2nd order tensors and changes to vtkCellDerivatives

[Andy Bauer]

Hi,

I wanted to inform everyone that there is some inconsistencies in the way
2nd order tensors are represented in VTK data arrays. The vtkGradientFilter
outputs in C/row-major ordering while the vtkCellDerivatives filter outputs
in Fortran/column-major ordering. There are other inconsistencies to this
as well. Officially, VTK uses C/row-major ordering for tuples. An example
of this is the vtkCell::Derivatives() method (
http://www.vtk.org/doc/nightly/html/classvtkCell.html#aff3d8332e9d7d556a9d2e9f91173d068).


Martin Genet has brought this inconsistency to light with his work on
vtkCellDerivatives in changing this to row-major ordering output. 40e05b6a
is his commit that fixed this issue for the filter, its documentation and
how vector_gradient and strain are computed in the VTK numpy interface. He
additionally removed the vtkTensor which was only used by
vtkCellDerivatives to transpose the 2nd order tensor's output.

There may be other 2nd order tensor ordering inconsistencies in VTK that
should be cleaned up as well and we hope to address those ASAP. The most
difficult of these will likely be in the readers that bring in 2nd order
tensors. I would encourage those that are familiar with specific readers to
spend a couple of minutes verifying the correct ordering from them.

I'll be adding in notes about this fix to
Documentation/Doxygen/ChangesVTK-7-1.md shortly.

Best regards,
Andy

ps. A big thanks to Martin for his fixes and his patience on getting this
through! This is his second and third commit to VTK.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20160203/6a16848f/attachment.html>