[Paraview] paraviw + discontinuous/non-linear functionswithin cells

[David C Thompson]

> Hi, thanks for the response.  I'm sorry, perhaps my email wasn't worded 
> carefully enough.  I believe the basis functions are continuous within the 
> cell itself.  I just meant to say that discontinuities can occur at any 
> boundar between cells.  It sounds like that you're of the opinion that 
> subdivision may work.  Thanks again.  
1. You need not perform subdivision at all if (over a single cell) the
approximation is trilinear. I haven't seen commercial DG solvers do
anything but trilinear interpolation, so you should be fine just by
repeating vertices in the mesh so that each cell can have unique nodal
values.

2. If you have quadratic interpolants within a cell, ParaView's existing
tessellation filter will most probably suit your needs. By forcing a
single subdivision along each axis (r, s, and t), you'll capture the
behavior of the interpolant properly with the resulting linear
approximation.

3. If you have interpolants that are higher order than quadratic, you'll
need to develop a mesh adaptor (á la the vtkBridgeDataSet
http://www.vtk.org/doc/nightly/html/classvtkBridgeDataSet.html). A
version of the tessellation filter is in ParaView so that you can create
a linear approximation. Sandia has some code (not in ParaView yet) that
locates critical points and creates more accurate approximations, but
what's there is most probably fine for your needs.

	David