WG: [Paraview] VTU File Format wih raw Data
Kent Eschenberg
eschenbe at psc.edu
Fri Oct 19 16:13:01 EDT 2007
I had an error in my earlier message. The cell types should be UInt8; the cell
offsets should be Int32. However it doesn't seem to make a difference.
I've looked at 3 versions of each case: yours, mine unfixed, mine fixed where
"fixed" means I changed the cell types to UInt8:
1 cell: 1032 1011 1008 bytes
2 cell: 1206 1184 1178 bytes
My fixed case should be 3 bytes less because UInt8 is 3 bytes smaller than
Int32. I'm not sure why your cases are larger.
Of these 6 files the only one that causes an error is your 2 cell case. In the
debugger I can see that the reader thinks the offsets are 2048 and 4096 while
the offset for your one cell case is 8.
I'll send all to you separately. You might look at these carefully with "od" or
some other utility to see what is different.
FYI I'm using PV 2.6.1 on a Fedora Core 6 system and Pentium 4 processor. My
compiler is Redhat's version of g++ 4.1.2.
Kent
Sören Freudiger wrote:
Hi
We really tried to do it. But it still doesn't work. And we cannot find the
problem. It looks quite easy to me. But something with the offsets must be
tricky. Our ascii-vtu file works fine (without appended data).
But the appended version with "raw" data couldn't be opened. Attached there's a
small program to show the problem. It's very simple and justs buts one or two
octants with some node data into a vtu-file. If we only have one cell it seems
to work. It can be opened with paraview 2.6.x and 3.x. But the two cell version
always ends with the error message:
" ERROR: In ..\..\..\ParaView3\VTK\IO\vtkXMLUnstructuredDataReader.cxx, line 646
vtkXMLUnstructuredGridReader (051948A0): Cannot read cell connectivity from
Cells in piece 0 because the "connectivity" array is not long enough. "
But I cannot find an error at the connectivities (I just wrote it to ASCII and
the connectivity for each Oct was 8 and the datalength was #cells*4(byte) ).
The type sizes are correct. Float32=4bytes, Int32=4 bytes...
More information about the ParaView
mailing list