[vtk-developers] New nonlinear celltypes

[David C. Thompson]

> Personally, I would be very pleased to get my hands on 3 of Soeren's
> nonlinear cell types when I move to VTK 5. In particular, it would be
> interesting to convert VTK_QUADRATIC_QUAD cells to VTK_BIQUADRATIC_QUAD
> to see if there is an improvement in the colour mapping.
Have you tried using the Tessellate filter in ParaView? That will create
a smoother approximation of your nonlinear interpolants and even at "0"
subdivisions avoids the triangulation that causes trouble for quadratic
cells.

> One question I have on the use of a generic cell adaptor approach is the
> recovery of the original cell edges for use in wire frame views. In VTK
> 4, I use an unstructured grid surface filter which preserves nonlinear
> faces and saves Face Id's which are used in edge filters to detect
> boundary edges and internally triangulated edges.
There isn't an equivalent filter for the generic dataset yet. The
closest matches are the vtkGenericGeometryFilter (doesn't store cell ID
or face ID with resulting geometry) and the vtkGenericDataSetTessellator
(doesn't take lower-dimensional boundaries, but can store cell IDs with
resulting geometry).

> Perhaps of more general interest is the question of user contributions
> which are documented and tested but not implemented in all VTK classes.
> Insisting on a 100% complete implementation seems to me to be too
> onerous and could potentially stifle all but the smallest of
> contributions. 
Obviously, Kitware has the final say, but here's my 2 cents. It sounds
like Soeren is most of the way there. And while I agree that it makes
the barrier high, VTK would quickly turn into a very difficult piece of
software to use otherwise. It's not that everything has to be there on
the first checkin, but there has to be a commitment to follow through.
If it was simply a filter that only handled a subset of cell types
(consider some of the volume renderers that only work on tetrahedra), it
might be one thing -- but the cell API is deeply involved in the VTK
framework.

	David