[vtk-developers] Parallel Coords Global IDs selection problem in ParaView

[Marcus D. Hanwell]

Hi Eric,

On Fri, Apr 9, 2010 at 2:58 PM, Eric E. Monson <emonson at cs.duke.edu> wrote:

> Hey Marcus,
>
> First, I see you snuck in moving the parallel coordinates axes with the
> middle mouse button -- cool! :)
>

Yes - still experimenting with a few required features. I have refactored
the axis labeling code, and added in support for enumeration on axes.

>
> I was playing with the PC chart in ParaView and came across a strange
> problem: Selections work fine if I'm doing it by ID (set in Selection
> Inspector panel Selection Type), but if I have the Selection Type set to
> Global Node IDs the list of Global IDs is twice as long as it should be, and
> alternates -1 values with the correct IDs.
>

This is where I need to talk to Utkarsh, as I knew that I had not accounted
for all of the selection mechanisms. From looking at the spread sheet view
selection code it appeared that selection ID properties had 2 or 3 elements
per command. When using global node IDs they just have one. Is there a

Here's the pipeline to reproduce with built-in functions (this morning's CVS
> PV):
>
> Point Source (I did 100 pts, 10 rad)
> Generate Ids
> Elevation
> Programmable filter:
>
> from paraview.vtk import vtkAssignAttribute
> pdi = self.GetInputDataObject(0,0)
> pdo = self.GetOutputDataObject(0)
> aa = vtkAssignAttribute()
> aa.SetInput(pdi)
> # aa.Assign()
> # Second arg: 5=GlobalIds, 6=PedigreeIds
> # Third arg: 0=Point, 1=Cell data
> aa.Assign('Ids',5,0)
> aa.Update()
> pdo.ShallowCopy(aa.GetOutputDataObject(0))
>
> PlotData
>
>
> Make a selection in the PC chart with the Selection Inspector visible. The
> SI Selection Type will show IDs. Switch this to Global Node IDs, and the
> list will still look fine. Then, make a different selection in the PC chart
> with Global Node IDs still as the Type. I get lists like this:
>

Thanks for the steps - I see this behavior too, and the single element per
command. The issue is that the selection IDs only match by chance too - they
are still just the row indices. It looks like I would need to get the global
node ids column, and map from table row indices to global node ids (and back
again) in the ParaView proxy/ParaView parallel coordinate view class.

>
> where the values that aren't -1 are correct. If you switch Type back to
> IDs, then back to Global IDs without changing the selection it looks fine
> again, until you make a new selection with Global IDs set as the Type.
>
> Since this is a strange one, let me know if you'd like me to file a bug
> report.
>
> If you could file a bug that would be great. It would be good to get this
working correctly, but I am still learning about selection mechanisms in
ParaView. The -1 values you see are the process ID (not set as multiprocess
is currently not implemented in charts).

Marcus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20100413/c1bf9c84/attachment.html>