[vtkusers] vtk.vtkExodusIIReader: information in GetOutput()?

[Nico Schlömer]

I've written and alternative implementation for the Exodus
reader/writer using vtkExodusReader (as opposed to the
vtkExodusIIReader) and found that things work out much more easily.

It's basically enough to

    reader.ExodusModelMetadataOn()

to get information on all the time steps, for example. With the
vtkExodusIIReader (where there is no ModelMetadata), I could get the
same to work only with a loop over [0:GetNumberOfTimeSteps()], then
SetTimeStep(k) and subsequently retrieve a vtkUnstructuredGrid for
each step with the corresponding point data.

The only thing I need to do now is to figure out how to extract the
individual time step arrays from the vtkUnstructuredGrid, work on
them, and inject them back to be written to a file.

Cheers,
Nico



On Sun, Dec 12, 2010 at 5:11 PM, Nico Schlömer <nico.schloemer at gmail.com> wrote:
> Hi,
>
> I was able to get it all together with
>
>    reader.UpdateInformation()
>    for k in xrange( reader.GetNumberOfPointResultArrays() ):
>        arr_name = reader.GetPointResultArrayName( k )
>        reader.SetPointResultArrayStatus( arr_name, 1 )
>
> Then, after Update() has been called, the vtkUnstructuredGrid will
> magically containt the array. Great!
>
>> Note that because "_R" and "_Z" imply cylindrical coordinates, the two scalars
>> in the exodus file will be combined into a 2-component vector by the reader and
>> the result will be named either "x0" or "x0_" (I forget whether the trailing underscore
>> is omitted or not).
>
> Cylindrical coordinates, alright? Well, I got the data in there by
> providing a complex valued array, but I guess that doesn't really
> matter. The underscore is preserved by the way, so I'll just go ahead
> and cut it off.
>
> Thanks for the help!
> --Nico
>
>
>
> On Sun, Dec 12, 2010 at 3:30 AM, Thompson, David C <dcthomp at sandia.gov> wrote:
>> Hi Nico,
>>
>>> Alright, so this
>>> ...
>>> gives me all vtkUnstructuredGrid in the vtkExodusIIReader output. Nice!
>>> However, there's more data in the Exodus file I'd like to retrieve,
>>> for example a data array for point data. In the ExodusII file, the
>>> values are declared as "vals_nod_var1=..." and "name_nod_var =
>>> "x0_R",  "x0_Z" ;", but those are certainly not contained in the block
>>> hierarchy above.
>>
>> Each unstructured grid that is output may have point and cell
>> arrays (I forget whether they are read by default or whether you
>> must call SetObjectArrayStatus() to tell the reader you want them
>> loaded). For the point variables "x0_R" and "x0_Z" in your example
>> above, you would call
>>
>>  rdr = vtkExodusIIReader()
>>  rdr.SetFileName( 'foo.exo' ) # or whatever
>>  rdr.UpdateInformation() # this tells the reader to fetch metadata from the file
>>  # Iterate over all the point arrays:
>>  for i in range( rdr.GetNumberOfObjectArrays( rdr.NODAL ) ):
>>    nname = rdr.GetObjectArrayName( rdr.NODAL, i )
>>    print 'Array %d is "%s"' % (i,nname)
>>    # if an array name matches, tell the reader we want to load it:
>>    if nname == 'x0' or nname == 'x0_':
>>      # 0 means "don't load", 1 means "load" the i-th nodal array
>>      rdr.SetObjectArrayStatus( rdr.NODAL, i, 1 )
>>  rdr.Update()
>>  out = rdr.GetOutput()
>>  # Assuming you have an element block, its point data will have
>>  # the x0 array.
>>  elem_blks = out.GetBlock( 0 )
>>  elem_blk0 = elem_blks.GetBlock( 0 )
>>  x0 = elem_blk0.GetPointData().GetArray( 'x0' )
>>
>> Note that because "_R" and "_Z" imply cylindrical coordinates, the two scalars
>> in the exodus file will be combined into a 2-component vector by the reader and
>> the result will be named either "x0" or "x0_" (I forget whether the trailing underscore
>> is omitted or not).
>>
>> Because Exodus defines a single, global set of nodes for all element/face/edge
>> blocks, any nodal variables you request will be present on every output block.
>> However, because the cells in a block don't necessarily reference every point,
>> the point coordinates and point data for each block are subsets of the global
>> arrays. For instance, if you have 10 points defined in a mesh but the first
>> element block contains a single quadrilateral (4 points), there will only be 4
>> values in the 'x0' array attached to that element block's unstructured grid.
>>
>>    David
>>
>>
>> On Mon, Dec 6, 2010 at 9:21 PM, David Thompson <dcthomp at sandia.gov> wrote:
>>>> ... when reading an ExodusII file with
>>>>  reader = vtk.vtkExodusIIReader()
>>>>  reader.Update()
>>>>  out = reader.GetOutput()
>>>> 'out' would contain a whole lot of output the significance of which I
>>>> don't quite understant;
>>>
>>> Hi Nico,
>>>
>>> The old vtkExodusReader used to output a single unstructured grid. This made
>>> it difficult to handle the case when a variable was defined on some cells
>>> (in one element block) but not on others (in a different element block). It
>>> was also awkward to provide information about node sets and face sets. At
>>> the time, there was also no concept of pedigree and global ID arrays in
>>> vtkDataSetAttributes. This made writing Exodus datasets back out to a file
>>> after some manipulation difficult since much of the information in the
>>> original file could not be preserved. The vtkExodusModel and
>>> vtkModelMetadata classes were attempts to encapsulate some of that
>>> information for later access by a writer.
>>>
>>> The new vtkExodusIIReader outputs a multiblock dataset. Each multiblock
>>> dataset may have an arbitrary number of blocks, where each block is a
>>> arbitary dataset (an unstructured grid, an image, or even another multiblock
>>> dataset). The new reader outputs a multiblock dataset where each of the
>>> top-level blocks is itself a multiblock dataset holding unstructured grids
>>> of a single type (element blocks, face blocks, edge blocks, element sets,
>>> side sets, face sets, edge sets, node sets). When you "print out" you are
>>> seeing all of the internal variables associated with the one toplevel
>>> multiblock dataset and its children.
>>>
>>> This should print some more concise information about the blocks:
>>> ====================== *snip* ======================
>>> for i in range( out.GetNumberOfBlocks() ):
>>>  print out.GetMetaData( i ).Get( vtkCompositeDataSet.NAME() )
>>>  blk = out.GetBlock( i )
>>>  for j in range( blk.GetNumberOfBlocks() ):
>>>    print '  ' + blk.GetMetaData( j ).Get( vtkCompositeDataSet.NAME() )
>>>
>>> ====================== *snip* ======================
>>> When the reader is told to read a block or set (only element blocks are read
>>> by default), then an unstructured grid will be inserted into the appropriate
>>> sub-block of the main reader's output.
>>>
>>>        Hope this helps,
>>>        David
>>>
>>>
>>>
>>>
>>
>>
>>
>