[vtkusers] Re: vtkFeatureEdges AND vtkPointLocator ...

[Will Schroeder]

Hi Sebastien-

The algorithm could easily be written so that it does not need 
vtkPointLocator. The reason this was done was to squeeze down the size of 
the output. For example, say the input is a mesh with 10,000,000 points, 
and the output of (the current) vtkFeatureEdges is one line with two 
points. Without using the vtkPointLocator the output would still contain 
10,000,000 points. (But because of reference counting maybe it doesn't 
matter because the points are not copied.) Anyway, it looks like it was 
somewhat of an arbitrary implementation decision. It would be quite easy to 
rewrite the filter to not use a vtkPointLocator, or to add a flag to 
provide an option to the user.

Are you going to take a crack at it :-) Please send the changes.

Will


At 01:58 PM 5/28/2003 -0400, Sebastien Auclair wrote:
>Why does vtkFeatureEdges needs to merge points in the first place.
>In our case, the dataset we display don't have duplicated points.
>Is there a way to go around this merging stuff ?
>
>What appends if i use SetLocator( vtkPointLocator* locator) set with
>Divisions  == 1x1x1 ?
>
>Thanks !
>______________________________
>Seb