[vtk-developers] new guy...

[Lorensen, William E (CRD)]

Jeff,
Many tests are crashing. I suspect your changes. In fact many tests that use unstructured grid are
crashing or failing. When you fix the problem, please run as many of the regression tests as you can.
 
For example, graphics/examplesTcl/appendCells.tcl
 
http://public.kitware.com/vtk/quality/MostRecentResults/
<http://public.kitware.com/vtk/quality/MostRecentResults/> 
 
Bill

-----Original Message-----
From: Jeff Lee [mailto:jeff at cdnorthamerica.com]
Sent: Wednesday, March 14, 2001 4:56 PM
To: vtk developers
Subject: [vtk-developers] new guy...


Hi, 
I have some changes I'd like to check in for vtkUnstructuredGrid.  They're not earth-shattering, just
some simplification and addition of a couple of features.  First, I removed vtkCellTypes and added in
its place a vtkUnsignedCharArray to represent the cell types, and a vtkIntArray to represent the cell
Location (i.e. given point id, return the offset in the vtkCellArray so you can retrieve the cell).
It doesn't affect the interface in any way except to add another way to SetCells directly.  If you
choose to SetCells with just 
(int* types, vtkCellArray cells), then the CellTypes and Locations arrays are built automatically
from them.  If you choose to SetCells with (vtkUnsignedCharArray *cellTypes, vtkIntArray
*cellLocations,  vtkCellArray *cells), then the UnstructuredGrid can use your memory directly without
copying (My motivation was to provide a way to let the UnstructuredGrid use my solver's memory
directly without copying). 

I also made changes to vtkCellTypes.  I basically replaced the struct _vCell_s with two separate
arrays.  I primarily did that to save memory - my thinking is that a struct of unsigned char and
integer cannot be compact because structs line up on word boundaries, but mabey some compilers can
optimize that (i doubt it). 
Anyway, does anyone have an interest in this stuff?  I have been using these changes for about a
month and haven't noticed any problems with examples, etc...  I should probably try some regression
tests.  Any thoughts? 
-Jeff 

-- 

J.A. Lee

Software Engineer

Computational Dynamics North America

3 Schoolhouse Lane

Etna NH, 03750

ph: 603-643-9993 x109, fax: 603-643-9994