Resample With Dataset

ParaView Support
1

I am attempting to transfer data from a CSV to a geometry with the ResampleWithDataset filter but to no avail.

I’ve tried:

  • using a MultiBlockDataSet as my target geometry
  • using a UnstructuredGrid as my target geometry (applying MergeBlocks to the MBDS)
  • using TableToPoints as my data source (applied to the csv data)
  • using Tetrahedralize as my data source (applied to the TableToPoints)
  • using PointDataToCellData as my data source (applied to the Tetrahedralize) and using the “Pass Cell Arrays” option (otherwise using “Pass Point Arrays”)
  • adjusting the ResampleWithDataset “Tolerance” option
  • adjusting the ResampleWithDataset “Cell Locator” option
  • using AppendReduce with the “move” option to push all my target geometry to the same processor as the source data
  • using DistributePoints on the source data and RedistrubuteDataSeton the target geometry spread each dataset across the same number of processors (RedistributeDataSet on the source data would cause a crash)

I also tried the LegacyResampleWithDataset at one point with no success. I can’t remember the exact parameters of that attempt, though.

The best that I can do is to get the point arrays moved to the geometry but with all the ranges set to [0, 0]

I have a larger/more complicated dataset I’d like to use, but I’ve been testing with the can.ex2 example and can’t even get that to work.

Note that I can’t use AppendAttributes because my real dataset does not have the same number of points as the csv data.

To reproduce:

  • Run PV in parallel (I’ve been using 4 procs)
  • Open can.ex2 and turn on the point arrays
  • Apply ExtractTimeSteps and choose one timestep (I’ve been using 17)
  • Apply ExtractBlock and select a part (I’ve been using block index 3)
  • Save the data from this extract as a CSV. Make sure the point arrays are selected, the Field Association is set to Point Data, and Add Meta Data is checked (so that we get the point coordinates, you could alternatively add the coordinates as an array through the calculator)
  • Apply PassArrays to the extract block and uncheck everything. Now we have a MBDS geometry with no data associated
  • Load the saved CSV. Default options should be fine
  • Apply TableToPoints and set your XYZ columns appropriately
  • Try all the things mentioned above

Note that (at least when running on 4 processors) the geometry and source data will both already be on rank 0. To more closely represent my real dataset scenario you can apply MergeBlocks and then AppendReduce to the PassArrays filter to move the geometry to another processor

1 Like