[vtkusers] recent vtkClipDataSet can give non-tets?

[Mathieu Malaterre]

Adrian Umpleby wrote:
>>> I've been using vtkClipDataSet with version 4.2 of vtk to cut off a 
>>> hexahedral mesh at a scalar value of zero. This gives all tets... 
>>> just as the (latest) documentation says...
>>> However, I've recently been using it on some upgraded Linux boxes 
>>> which have version 4.4 of vtk, and now I find a whole load of wedges 
>>> coming out of vtkClipDataSet!
>>> Is this the way vtkClipDataSet is supposed to behave now? Is there a 
>>> way to choose to switch it back to tets? (Couldn't see anything in 
>>> member list.)
>>
>>
>> That's correct. Try to clip a tetra, you'll realize that there are a 
>> bunch of possibilties for the clip to return a wedge.
> 
> 
> Note that I said I'm clipping hexahedra, so the possible polyhedra can 
> be quite mind-boggling - vtkClipDataSet would have to return such 
> polyhedra in a split up configuration just to get down to even mostly 
> hexas again...
> 
> (I'm guessing, though, that vtkClipDataSet first splits up the mesh into 
> tets [using something like vtkDataSetTriangleFilter?] before clipping - 
> so wedges make sense if nothing else is done to the clipped cells. 
> Having said that, I'm curious what happens in the case of quadratic 
> cells? - quad-tets can end up with more than one cut [strictly not even 
> planar!] for a single scalar value... it's interesting to note that even 
> cutting a linear hexa isn't necessarily planar - I'd expect the results 
> from vtkClipDataSet can depend a lot on which way the original hexas are 
> split. But I'm getting off the subject here...)
> 
>> So yes wedge is the general output of clip ( vtkClipDataSet even 
>> return wedges with two edges collapsed (=tetra), but vtkGeometryFilter 
>> handles those 'fake' tetras).
>>
>> What is your problem ?
> 
> 
> Note that it's not actually a problem - I need tets at the end, so all 
> I've done to get around it is put the clipped dataset through 
> vtkDataSetTriangleFilter.
> 
> However, the issue is that the latest documentation says vtkClipDataSet 
> produces an all-tet mesh. This was correct under vtk 4.2, but appears to 
> no longer be correct for vtk 4.4.
> 
> I was simply wondering if this could be acknowledged as "the way it's 
> supposed to be" now (i.e. the docs need updating), or maybe some sort of 
> bizarre bug (resulting mesh appears to conform correctly, so I find this 
> hard to believe), or an oversight after some recent change.

Adrian,

	vtkClipDataSet now returns wedges and tets. This is done because this 
way a case table can be used that is much faster than the previous method.

	In vtkClipVolume there is a flag, Mixed3DCellGeneration, that allows 
you to force the generation of tets. This flag is not found in 
vtkClipDataSet, I suspect it could be added.

HTH
Mathieu