[vtk-developers] vtkMultiPieceDataSet (was Composite Data Redesign)

[John Biddiscombe]

Utkarsh

I've got one use case scenario that fits the MultiPiece dataset....

I have 8 pieces of a volume which are placed into a multiblock dataset. 
Pieces are 0-11, 12-23 along each eaxis.

When I create the data, I put WholeExtent 0-23 and Extent 0-11 or 12-23 
as required and all is well. Then I pass the data into another filter and

reader -> GetDataSet(0,i)->GetWholeExtent(...) returns 0-23, 0,23, 0-23
filter->SetInput(output...) (executive is composite)
filter->Update()
reader -> GetDataSet(0,i)->GetWholeExtent(...) returns 0-11, 0,11, 0,11 
or 12-23, 23-23, 12-23 ie WholeExtent=Extent

The whole extent is always being reset to the actual extents. This is 
bad and I think it is a bug (- I lose precious information that I need 
to use when piecing the data back together later). But if there is a 
reason for it (and I'm trying to chase it down as I write), then I can 
see a case for handling a piece dataset differently.

I say no more until I've investigated further.

JB


> John,
>
> I am done with most of the coding, so now basically waiting for 
> feedback and the dashboard to clean up before committing the changes. 
> Just to clarify, the vtkMultiPieceDataSet part won't get implemented 
> in this pass. There's still no consensus here on the necessity of it  
> given that the same can be achieved by vtkMultiBlockDataSet or whether 
> it should be a new class or simply an ivar (or some other 
> qualification) on vtkMultiBlockDataSet.
>
> Utkarsh
>
> John Biddiscombe wrote:
>> Utkarsh
>>
>> I'm very keen to try out the MultiPiece dataset. I think it will make 
>> a significant difference to the development of several readers for 
>> data that has been dumped into files - one per processor. Do you have 
>> a likely timeframe in which you'll be commiting things?
>>
>> JB
>>
>>> Folks,
>>>
>>> Composite data pipeline is undergoing a redesign to overcome some of 
>>> its current limitations.  Details of the issues, the new design and 
>>> the implementation are available at:
>>>
>>> http://www.vtk.org/Wiki/VTK/Composite_Data_Redesign
>>>
>>> To summarize: vtkCompositeDataSet has been modified to support a 
>>> full tree datastructure instead of a table of tables approach used 
>>> by the previous implementation. It is possible to traverse the tree 
>>> using iterators. The iterators are more powerful than before making 
>>> it possible to not only access the datasets, but also change them. 
>>> vtkHierarchicalDataSet and vtkMultiGroupDataSet have been 
>>> deprecated. The only concrete composite datasets available are 
>>> vtkMultiBlockDataSet, vtkHierarchicalBoxDataSet and vtkTemporalDataSet.
>>>
>>> The new code is not available in CVS yet. I'll wait a few days for 
>>> people to respond before committing it.
>>>
>>> Utkarsh
>>> _______________________________________________
>>> Paraview-developers mailing list
>>> Paraview-developers at paraview.org
>>> http://public.kitware.com/mailman/listinfo/paraview-developers
>>
>>


-- 
John Biddiscombe,                            email:biddisco @ cscs.ch
http://www.cscs.ch/about/BJohn.php
CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.07
Via Cantonale, 6928 Manno, Switzerland      | Fax:  +41 (91) 610.82.82