Extend LoadVTK

[Owen Arnold]

Under #7252, a loader for a specific flavour of vtk legacy image data to MDHistoWorkspaces was created. We currently think that utilising the inherent splitting box structure available in the MDEventWorkspace would be a good solution to providing an increased level of detail in regions of interest would be a highly useful option. It would also allow the data to be reasonably visualised using the VSI tools.

• Change LoadVTK to output IMDWorkspaces
• Split off the current processing route into a separate function
• Add a new loading route to handle MDEventWorkspaces
• Add an option in the loader to choose what type of workspace to output.

[Owen Arnold]

Benchmark performance of existing Load routine using run of performance tests.

time="0.051602"

[Owen Arnold]

refs #7268. Extract processing to new method.

Changeset: 89305cdc5680ec8652ebaa5d800f86ac1dcec3a8

[Owen Arnold]

refs #7268. Change API to allow for MDEventWS.

Changeset: 19a5c42d0c36cfd0c025701384a772c3f6a57c46

[Owen Arnold]

refs #7268. Primative MDEventWorkspaces. Missing splitAllIfNeeded.

Changeset: 2aaf4fb45eddce99db0aeeaf9e8d2cb59408ecad

[Owen Arnold]

refs #7268. Adaptive binning.

A works in progress

Changeset: 34c8e0f1b5996b8a8f38371a2165136f27192e01

[Owen Arnold]

Modifications to iron_protein made using the following paraview filter code:

pdi = self.GetInput()
pdo = self.GetOutput()
newData = vtk.vtkUnsignedShortArray()
newData.SetName("scalar_array") 
npoints = pdi.GetNumberOfPoints() 
for i in range(npoints):
  if i < 10000:
    newData.InsertNextValue(i)
  else:
    newData.InsertNextValue(0) 
pdo.GetPointData().AddArray(newData)

[Owen Arnold]

refs #7268. Fully working with MDEvent workspaces.

Changeset: 0a87bc8707ba3eeae9334e169fec5b9b23b52ba2

[Owen Arnold]

refs #7268. Fix non-universal typedef.

Changeset: d43e173de90544b6832a87611318f42d0b1dea5f

[Martyn Gigg]

Checked using the fly dataset. I can now load it using the adaptive binning option and it has a more coarse structure as expected. It also uses much less memory. Increasing the threshold does indeed bring in more of the sparse signal areas.

[Stuart Campbell]

This ticket has been transferred to github issue ​8114