[Paraview] writeing vts files
don fisher
dfisher at as.arizona.edu
Mon Feb 28 13:18:00 EST 2005
Brad,
Sorry for the delay in getting back. I answer to the questions you posed:
The source files are from the CVS download.
1) The one that is 4622 lines long is xmlparse.c.
2) There are many defines in the xmlparse.c code, but there is an
entire file called vtl_expat_mangle.h that redefines function names.
This is probably required to accommodate C++ overloading, but it makes
debugging difficult. The debugger tracks compiler mangled names
without any confusion.
3) xmltok_impl.c and vtkSetGet.h are examples of where I thought
inlined functions might be more appropriate than the #defines which
are not expanded by the debugger.
I would like to find documentation on how to add, or extend, a file
reader so that frame sequences contained in one file could be viewed
as an animation. I would also appreciate documentation on the logical
flow of operations, like thinking in the POP world. I need to make my
own arrow glyph where only the length of the shaft is scaled a
function of the vector magnitude.
The numerous documents I purchased from Kitware describe visualization
of petabyte data sets as a goal for Paraview. My goals are large, but
a couple of orders of magnitude less ambitious. An efficient approach
appears to employ the use of memory mapped image files in a 64-bit
environment. My current machine had 16GB of ram, and could be extended
to 32GB. Then the fabrication of IO data sets would not be a concern
to the user. He would just make a huge HDF5 structure and the system
would access the set of components directed by him. Is this type of
approach consistent with your broad objectives?
Thanks for your time
don
Brad King wrote:
> don fisher wrote:
>
>> I have examined the length of the appended data record and it is
>> 34136, the correct length to encode one floating scalar and a three
>> component float vector (dimensions 40X20X2). The output I receive from
>> Paraview consists of 3 occurrences of:
>> ---------------------------------------------------------------------
>> ErrorMessage
>> # Error or warning: There was a VTK Error in file:
>> /usr3/swl/packages/ParaView/VTK/IO/vtkXMLStructuredDataReader.cxx (281)
>> Error reading extent 0 39 0 19 0 1 from piece 0
>> ErrorMessage end
>> ---------------------------------------------------------------------
>
>
> The CVS version of these readers is currently undergoing some changes so
> this may be a bogus message, though it looks like a reasonable error if
> something is wrong with your appended data section.
>
>> I have looked at the code, but it is not easy to read. The parse
>> routines are order 4500 lines long, with mangled names and large
>> #defines rather that inline functions. I does confuse my debugger. Any
>> help would be appreciated.
>
>
> To what source files are you referring?
>
>> The only example on the CD and distribution includes zlib compression.
>> Is that required, or is there another string required to say "no
>> compression".
>
>
> If you leave off the "compressor" attribute from the top-level VTKData
> element then there is no compression.
>
> > I cannot find any details in the books or enclosed CD. The
>
>> terms "binary" and "base64" seem to be overlapped.
>
>
> The "binary" format means that the data are inlined inside the DataArray
> element. Inlined data must be base64 encoded to keep the XML valid. The
> "appended" format means that the data are in the AppendedData element at
> the end of the file. This element has its own "encoding" attribute
> which is usually base64. As a hack to support fast reading we also
> allow the AppendedData section to be pure binary (not encoded), though
> this makes the document invalid XML.
>
> -Brad
>
--
-------------------------------------------------------------------
| Don Fisher dfisher at as.arizona.edu |
| Steward Observatory |
| 933 N. Cherry Ave. VOICE: (520)888-7613 |
| University of Arizona |
| Tucson, AZ 85721 |
-------------------------------------------------------------------
More information about the ParaView
mailing list