[vtkusers] Streamlines Debugging tips?

tom fogal tfogal at apollo.sr.unh.edu
Thu Aug 26 17:56:47 EDT 2004


 <412E5306.5000707 at ornl.gov>Stewart Dickson writes:
<snip>

>When I run my scri[pt through Vtk, I get the following messages:
>ERROR: In /usr/local/src/VTK/Common/vtkPolyLine.cxx, line 101
>vtkPolyLine (0x892cfa8): Coincident points in polyline...can't compute 
>normals
>
>Warning: In /usr/local/src/VTK/Graphics/vtkTubeFilter.cxx, line 195
>vtkTubeFilter (0x80ef9a0): No normals for line!
>(Repeated 4 times)
>And no geometry displayed in the vtkRenderWindow (OpenGL).
>
>I suspect that vtkStreamLine  Set[Integration]StepLength [0.05]0.5 
>is/are wrong for the dimensions of my space and the velocity of my 
>vector data.

StreamLine? vtkStreamLine is deprecated, you should be using
vtkStreamTracer instead. This bit me as well, it seems the only place
to find out currently is this mailing list =). Gotcha to note while you
switch over: make sure to set the MaximumPropagation to something
large.
I'm using vtkStreamTracer with ridiculously small integration steps
(like 1e-6) and having no troubles.

>Will randomly-generated points landing in a region of zero velocity
>cause the streamline generator to puke? (Cannot generate normals on 
>zero-length segments)

Looking through the code, it appears this is correct.

>Are there any published rules-of-thumb for setting integration 
>parameters appropriate to the data fields under observation?
>Are there any resommended methods of debugging the integrator/streamline 
>generator?

Hrm.. good question. Maybe someone smarter than me can answer =)

Try switching to StreamTracer and see if that helps. Report back if it
does not.

-tom



More information about the vtkusers mailing list