[vtk-developers] implementing vtkPolyData::ComputeBounds() ?

[Sebastien BARRE]

At 09:54 31/07/00 -0400, Lisa Sobierajski Avila a écrit:

>I think it is more correct to have the vtkPolyData::ComputeBounds() 
>implemented than to simply use the bounds of all points. Unless someone 
>thinks this isn't the case, I would suggest implementing it, test it, then 
>check it in - and we'll need to keep an eye on the timing for a few days 
>to make sure it doesn't significantly slow anything down. I don't think it 
>should since bounds only need to be recomputed when data is changed.

Done and checked in.

Well, for sure it can not be as fast as vtkPointSet::ComputeBounds(), *but* 
as soon as two objects share the same vtkPoints and have different 
topology, the new vtkPolyData::ComputeBounds will return more accurate 
results. BTW, I do not know a lot about the VTK (point) locators, but if 
some of them use buckets or binary space division, they will probably 
benefit from the fact that the space to divide is now closer to the "real" 
object.

Let's check the timings.
(Lisa, could you please send me (private email) the URL of the dashboard 
and the continuous build results, I'm at home this week, and have not 
bookmarked them).

Thanks