[Paraview] [Xdmf] Problem with Xdmf/HyperSlab

[Jens Kleimann]

Hi Ken,

thank you for your looking into it. My actual application will in fact have HDF5 in the second DataItem. But then I encountered problems and tried to figure it out using a simple example first.

Can you confirm that my xmf sample file is syntactically correctly for its designated purpose?

Best, Jens.

On 31.01.2011 17:25, Kenneth Leiter wrote:
> Hi Jens,
> 
> Thanks for the report.  This appears to be a bug caused by incorrect
> handling of the second data item when it exists as XML rather than HDF5.
> I'll see what I can do about fixing it.
> 
> - Ken
> 
> On Mon, 2011-01-31 at 11:30 +0100, Jens Kleimann wrote:
>> Greetings,
>>
>> after spending a couple of days trying to figure out the proper usage of XDMF HyerSlabs, and with no progress whatsoever, things are starting to get frustrating.
>> I have a 3-D curvilinear grid with scalar data, and I need to cut out a sub-grid (in order to discard the boundary). The HyperSlab should do this, but I fail with every try. The smallest example file which reproduces the problem is at [http://www.tp4.rub.de/~jk/science/hyperslab_mintest.xmf]. I have tried various combinations, but I get crashes on Linux, crashes on Windows (using 3.8.1. 64bit, not necessarily for the same cases), and whenever I manage to load it without a crash, the spreadsheet view indicates that the contents of the source array is just copied element-wise to the target array, disregarding the contents of the HyperSlab's first DataItem. (To verify this, increase the count values to from "2 2 2" to something like "2 2 20", and load into PV.)
>> The ability to extract brick-shaped subgrids from larger grids of the same type is absolutely essential to my application. I would very much like to use Paraview for my purposes, but this problem, and the apparent lack of resources which could provide at least a hint towards a solution, is a no-go.
>> Stripping boundary values does not seem an overly exotic task, so I guess someone must have done it in the past.
>> I would be particularly great if someone from the developers could be convinced to have a brief look at it.
>> Any hint would be most welcomed. Anyone...?
>>
>> Jens.