[vtk-developers] Information visualization data structures

[Tim Tautges]

I've been working with the Terascale Simulation Tools & Technologies 
(TSTT) DOE SciDAC project, and as part of that, have helped develop a 
data model for mesh data.  One of the things that has proven extremely 
powerful in our data model is the ability to define arbitrary sets of 
entities (including other sets), and also to define parent/child 
relationships between sets.  Essentially, this allows sets to be nodes 
in an arbitrary directed graph.  Using this construct, I've been able to 
embed, in just the mesh definition, the geometric model topology from 
which a mesh was generated, a processor partition (and partition 
interfaces) for parallel FE, and groupings of elements defining the dual 
of a hexahedral mesh.  I'm also in the process of implementing an OBB 
tree using this data model.  Having the ability to browse these graphs 
in paraview is something I've wanted for some time (so, that's a little 
different from visualizing the graphs themselves, but also very useful).

I think if you allowed for the possibility of your nodes to be sets, and 
to allow the contents of those sets to be data from various other 
datasets in the model, you'd have a much more powerful capability.  I 
think vtkExtractCell is a good start at a set capability; having your 
graph nodes derive from that, adding parent/child links, would give you 
what you needed (an undirected graph could simply use the parent links 
only, or wouldn't distinguish parent from child links).

Again, this is slightly different from what you're proposing, in that 
the data is still mesh-type data, but there are graph(s) embedded in 
those data too.  It might be possible to represent both types of graphs 
with the same data structure, which would be easier to maintain and more 
broadly applicable.

- tim

Jeff Baumes wrote:
> We would like to propose new classes for a VTK information
> visualization architecture.  VTK has been used mainly for scientific
> visualization, but it makes sense to leverage its capabilities to use
> in the information visualization arena.  In order to do so requires a
> new set of data structures for storing tables, graphs, and trees.  The
> attached document describes the proposed API.
> 
> Jeff Baumes, Berk Geveci, Will Schroeder
> Kitware, Inc.
> 
> Patricia Crossno, Ken Moreland, Andrew Wilson, Brian Wylie
> Sandia National Laboratories
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> vtk-developers mailing list
> vtk-developers at vtk.org
> http://www.vtk.org/mailman/listinfo/vtk-developers

-- 
================================================================
"You will keep in perfect peace him whose mind is
  steadfast, because he trusts in you."               Isaiah 26:3

             Tim Tautges            Sandia National Laboratories
         (tjtautg at sandia.gov)       (telecommuting from UW-Madison)
         phone: (608) 263-8485      1500 Engineering Dr.
           fax: (608) 263-4499      Madison, WI 53706