[vtkusers] vtkAppendFilter with double precision points (UNCLASSIFIED)

[Dominik Szczerba]

Berk Geveci wrote:
>> Single precision is hard coded - and is so in many filters. I have long
>> argued to let the user specify the type, but nobody seemed interested even
>> in my patches.
> 
> You probably caught us in a time we were very busy (which is pretty
> much always). Such patches would be of interest. However, I would argue
> that instead of asking the user, the append filter should pick an appropriate
> precision. For example, if at least one of the inputs is double, it should
> probably produce double as well. I would argue the same for most filters
> except those that produce new arrays independent of their input arrays.

I am not at this level of sophistication in my changes... just setting 
the precision by hand as desired. As to an automated choice, I would say 
there might be situations where you still want float, even if the input 
is double. Therefore I think a manual solution is most universal and 
easiest to add with smallest risk of introducing new bugs (let's keep 
the hardcoded float as default). Just drop me a line if my effort would 
make sense and I will submit some patches to the tracker.

Regards,
Dominik

> 
> I cc'ed vtk-developers since this may be relevant to other developers as
> well.
> 
> -berk
> 
> On Thu, Dec 10, 2009 at 3:29 PM, Dominik Szczerba <dominik at itis.ethz.ch> wrote:
>> Single precision is hard coded - and is so in many filters. I have long
>> argued to let the user specify the type, but nobody seemed interested even
>> in my patches.
>>
>> My workaround is to maintain vtkAppendFilter2 (with an appropriate function
>> to set the precision) under vtkmy structure.
>>
>> Once again, if interest, I can contribute a few patches.
>>
>> Dominik
>>
>> Leiter, Kenneth (Cont, ARL/CISD) wrote:
>>>
>>> Classification:  * UNCLASSIFIED* **
>>> Caveats: NONE
>>>
>>> Hello all,
>>>
>>> I am reading an exodus II file into vtk using the vtkExodusIIReader and am
>>> converting the resulting vtkMultiBlockDataSet to a vtkUnstructuredGrid for
>>> further processing.  I am using a vtkAppendFilter to get from the
>>> vtkMultiBlockDataSet to a vtkUnstructuredGrid.
>>>
>>> However, I am losing precision for the point values after running through
>>> the vtkAppendFilter.  I have verified that the points are double precision
>>> before entering the filter and single precision afterwards.  I am not
>>> surprised that I lose precision as it looks like a new vtkPoints (default =
>>> single precision) is allocated in the vtkAppendFilter class and there
>>> appears to be no way to change this behavior.
>>>
>>> Does anyone have a workaround or another way to accomplish what I need to
>>> do?  It was difficult enough to get the vtkExodusIIReader to read in double
>>> precision points without converting them to floats!
>>>
>>> Thanks,
>>> Ken
>>>
>>> ---------------------------------------------------------
>>> Kenneth Leiter
>>> Scientific Visualization Team
>>> Lockheed Martin / ARL DoD Supercomputing Resource Center
>>>
>>>  Classification:  * UNCLASSIFIED * **
>>> Caveats: NONE
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> Powered by www.kitware.com
>>>
>>> Visit other Kitware open-source projects at
>>> http://www.kitware.com/opensource/opensource.html
>>>
>>> Please keep messages on-topic and check the VTK FAQ at:
>>> http://www.vtk.org/Wiki/VTK_FAQ
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://www.vtk.org/mailman/listinfo/vtkusers
>> _______________________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Please keep messages on-topic and check the VTK FAQ at:
>> http://www.vtk.org/Wiki/VTK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.vtk.org/mailman/listinfo/vtkusers
>>
>