Set/get MDHistoWorkspace contents as numpy arrays

[Janik Zikovsky]

• Expose methods to get or set the signal and error data of MDHistoWorkspaces as numpy array.

Note from Martyn:

At the moment anything that implements the MatrixWorkspace interface can have a numpy array created from the data. There a several methods that allow you to either create a read-only wrapper, a writeable wrapper and then another for extracting the data out into a 2D array.

PyhonInterface/mantid/api/MatrixWorkspace.cpp shows the export to Python of MatrixWorkspace, which gives you the list of methods there are available there. These use some helper code in the file PythonInterface/mantid/api/WorkspaceToNumpy.cpp in order to actually produce the arrays from our data. I would say that this file could be renamed to MatrixWorkspaceToNumpy.cpp and then you could create another file MDWorkspaceToNumpy and implement the required helpers there. Does that make sense?

I suggest that this should go into the new API only, as we would like to phase the old one out and also for the next iteration.

Cheers,

Martyn

[Nick Draper]

Moved at end of release 2.1

[Nick Draper]

Moved at the end of release 2.2

[Martyn Gigg]

Refactor numpy converters to work through C arrays. Refs #4649

This should simplify the numpy converter code so that everything just funnels it through the C array interface. Future converters should no only have to implement the top level portion an pass c arrays on.

Changeset: e57703a22350472d144eeb3d58fd371977f09931

[Martyn Gigg]

Rename header NumpyWrapMode to WrapWithNumpy. Refs #4649

Changeset: 38539608054608dfa1cb260815d38c499f5cd71b

[Martyn Gigg]

Wrap signal/error arrays from MDHistoWS with numpy. Refs #4649

Changeset: 10f9186f7caa859078e0668f4b9ce26d4db92f57

[Martyn Gigg]

Add setters from numpy array for MDHistoWS from python. Refs #4649

Changeset: 5492e586a830f20da0f2f69c19b4d42a8fff2acc

[Martyn Gigg]

Refactor numpy converters to work through C arrays. Refs #4649

This should simplify the numpy converter code so that everything just funnels it through the C array interface. Future converters should no only have to implement the top level portion an pass c arrays on.

Changeset: e57703a22350472d144eeb3d58fd371977f09931

[Martyn Gigg]

Rename header NumpyWrapMode to WrapWithNumpy. Refs #4649

Changeset: 38539608054608dfa1cb260815d38c499f5cd71b

[Martyn Gigg]

Wrap signal/error arrays from MDHistoWS with numpy. Refs #4649

Changeset: 10f9186f7caa859078e0668f4b9ce26d4db92f57

[Martyn Gigg]

Add setters from numpy array for MDHistoWS from python. Refs #4649

Changeset: 5492e586a830f20da0f2f69c19b4d42a8fff2acc

[Andrei Savici]

Load(Filename='CNCS_7860_event.nxs',OutputWorkspace='CNCS_7860_event')
ConvertUnits(InputWorkspace='CNCS_7860_event',OutputWorkspace='CNCS_7860_event',Target='DeltaE',EMode='Direct',EFixed='3')
ConvertToMD(InputWorkspace='CNCS_7860_event',OutputWorkspace='MD',QDimensions='|Q|',QConversionScales='HKL',MinValues='0,-3',MaxValues='3,3')
BinToMDHistoWorkspace(InputWorkspace='MD',AlignedDim0='|Q|,0,3,30',AlignedDim1='DeltaE,-3,3,30',OutputWorkspace='MDH')

w=mtd['MDH']
w.setSignalArray(-2*w.getSignalArray())

[Stuart Campbell]

This ticket has been transferred to github issue ​5496