[Paraview] [Xdmf] How to write "voxel" or "image" data xdmf file.

[Michael Jackson]

On Feb 5, 2013, at 10:23 AM, Jens Kleimann wrote:

> On 05.02.2013 13:27, Michael Jackson wrote:
>> I finally got it to work, at least for Scalar values. The Dimensions in the topology section need to be 1 more than in the DataItem section because the Xdmf reader is creating a Rectilinear Grid and not "ImageData". So once I finally figured that out then I could load up some of my scalar values. 2 Step forward. 
>>   I then tried to load up a "Vector" data set, i.e., at set that has 3 component array at each Cell Center. This failed with various error messages also so I guess I'll try the process of elimination again for that today.
> 
> I might be mistaken, but as far as I recall, vector components have to be on cell NODES, not centers if you want to take advantage of PV's vector field visualization tools (arrows, streamlines and such). I use something like
> 
>  <Attribute Name="my_vec_field" AttributeType="Vector" Center="Node">
>    <DataItem ItemType="Function" Dimensions="128 128 128 3"
>              Function=" JOIN( $0,$1,$2 ) ">
> 	<DataItem Dimensions="128 128 128" NumberType="Float" Format="HDF">
> 	  my_file.h5:/path/comp_x
> 	</DataItem>
> 	<DataItem Dimensions="128 128 128" NumberType="Float" Format="HDF">
> 	  my_file.h5:/path/comp_y
> 	</DataItem>
> 	<DataItem Dimensions="128 128 128" NumberType="Float" Format="HDF">
> 	  my_file.h5:/path/comp_z
> 	</DataItem>
>    </DataItem>
>  </Attribute>
> 
> to load vector data, and it works reasonably well. In your case though, a complication might arise from the possible need to convert a grid of 3D vectors into a set of three separate grids for each components.
> 
> As a side note, it has proven advantageous for debugging purposes to start with XML-only files (replacing the file reference with some (obviously smaller) sample data, and work from there. That way, the dimension mismatch you mentioned earlier would have been detectable to others not in possession of your heavy data file.
> 
>> It would be really helpful if there were more xdmc examples on their web page. Promising technology. Just needs some more documentation.
> 
> That's exactly my impression as well...
> 
> Good luck,
> Jens.

Thanks for the thoughts. I'll take that into consideration when trying to figure out what is going wrong. Now on to see if XDMF will allow me to display my data with my chosen set of colors.

---
Mike Jackson