[vtkusers] about vtkStreamer

[Petru Pau]

I wonder if the people who implemented vtkStreamer are still with us...

...Because I could use some advice.

I needed to generate stream lines on the surfaces of some solid bodies. 
It is
known that the initial algorithm allows the streamlines to escape from 
the surface,
if the vector field requires it.

So I had to constrain the integration process to the surface of each cell.

There were two problems to be solved:
- using only the projection of the velocity vector;
- shifting the integration process to next cell, when a cell boundary is 
met.

I hacked in the code of vtkStreamer (it seemed to much bother to make a new
set of classes - a set because it's not enough to modify only vtkStreamer),
and added some code that works nicely in many cases.

However...

For some situations the algorithm never ends. It is hard to understand 
why, especially
considering that the infinite (well...) loop generates points that stay 
on the same cell.

For other situations the streamline escapes in space. I reckon this is 
due to a bug.

For other situations, the integration ends when the streamline touches a 
cell boundary.

Anyway, things work somehow satisfactory. Almost always. Unfortunately 
not in the
real situations; the solids for which I need this the integration 
process generates
millions of points on streamlines. As always, the exceptional situations 
tend to be
"general cases". I was able to detect a few, but not all - after all, I 
had to guess
the functionality of the code, in vtkStreamer.

I attach to this message a relevant image. If I change the vectors, 
there will be other
lines that go over all three faces.

Petru
-------------- next part --------------
A non-text attachment was scrubbed...
Name: simple_ex.jpg
Type: image/jpeg
Size: 14470 bytes
Desc: not available
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20040730/1d249f59/attachment.jpg>