VTK/Remove vtkTemporalDataSet

From KitwarePublic
< VTK
Revision as of 14:38, 25 June 2015 by Mwestphal (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigationJump to search

Previously, VTK pipeline supports multiple time steps in the pipeline: Filters can request data sets of multiple time steps from upstream using the key vtkStreamingDemandDrivenPipeline::UPDATE_TIME_STEPS(); Upstream sources then fulfil the request by wrapping the data sets in a vtkTemporalDataSet object and setting the key vtkDataObject::DATA_TIME_STEPS on the object. Over time, we found that the pipeline logic to handle multiple time steps is too complicated to maintain, especially considering the limited use of it. Therefore, we have reduced the temporal support to handle only single time step requests. Filters that need multiple time steps still work, by using the key vtkStreamingDemandDrivenPipeline::CONTINUE_EXECUTION() to loop the upstream pipeline and caching the resulting data sets inside the filters. We have also created a convenience super class vtkMultipleTimeStepAlgorithm that does exactly that.

In more details, the multiple time request keys:

  • <source lang="cpp"> vtkInformationDoubleVectorKey* vtkStreamingDemandDrivenPipeline::UPDATE_TIME_STEPS() </source>
  • <source lang="cpp"> vtkInformationDoubleVectorKey* vtkDataObject::DATA_TIME_STEPS() </source>

are replaced by single time request keys:

  • <source lang="cpp"> vtkInformationDoubleKey* vtkStreamingDemandDrivenPipeline::UPDATE_TIME_STEP() </source>
  • <source lang="cpp"> vtkInformationDoubleKey* vtkDataObject::DATA_TIME_STEP()</source>

As a consequence of the changes, the following constructs are no longer used and depreciated:

  • <source lang="cpp"> #define VTK_TIME_EXTENT 2 </source>
  • <source lang="cpp"> class vtkTemporalDataSet </source>
  • <source lang="cpp"> class vtkTemporalDataSetAlgorithm </source>

After the changes, to request data sets of multiple time steps, a filter can loop the upstream pipeline by setting the key vtkStreamingDemandDrivenPipeline::CONTINUE_EXECUTION(). The simplest usage example for the key can be found in the class vtkTemporalStatistics. Another example is the class vtkMultiTimeStepAlgorithm, which can request multiple time steps, cache the multiple data sets in a vtkMultiBlockDataSet and pass it to the child class when all the data sets are received. A filter that wants to request multiple time step data can simply inherit this class and set the key vtkMultiTimeStepAlgorithm::UPDATE_TIME_STEPS(). An example is the class vtkTemporalInterpolator. We note that this is no more work than before the pipeline changes.