ConvertToMDEvents does not keep track of 0s

[Andrei Savici]

Try

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')
Rebin(InputWorkspace='CNCS_7860_event',OutputWorkspace='CNCS_7860_event',Params='-3,0.1,3',PreserveEvents='0')
SetGoniometer(Workspace='CNCS_7860_event',Axis0='huber,0,1,0,1')
SetUB(Workspace='CNCS_7860_event',a='2',b='3',c='4')
ConvertToMDEvents(InputWorkspace='CNCS_7860_event',OutputWorkspace='md',QDimensions='Q3D',QConversionScales='HKL',MinValues='-3,-3,-3,-3',MaxValues='3,3,3,3')

There should be 51200*60 MDEvents, but I see only 57715. I saved the MD workspace to a NeXus file, and I do not see events with weight 0.

[Alex Buts]

there is condition in the convertToMDEvents which drops empty events. I have reviseted the code and see no reason why it is there.

ignoreNaN should be there instead.

[Andrei Savici]

The problem is that you drop so called "empty events" in line 131 of ConvertToMDEventsHistoWS.h Of course, that is physically wrong. If you measure something and you obtain that it is 0, you should keep it. Also, your code would get rid of events coming from bins with negative weights. You need to keep those as well (some people subtract background)

Note that there is a system test that will fail when you fix this algorithm

[Alex Buts]

refs #5318 This should fix it

Changeset: c77edbf0c53ef996de0f948e03b36718d1801ddc

[Alex Buts]

the problem has been fixed for old ConvertToMDEvents Code but also should be fixed (not appear in) new ConvertToMD code

[Andrei Savici]

Works in the new code

[Stuart Campbell]

This ticket has been transferred to github issue ​6164