Strange over filtering in GenerateEventsFilter

[Owen Arnold]

I'm struggling to explain why the number of events filtered + number of events unfiltered is not equal to the number of events in the original. Then I came up with this test case which uses a fake log, and should filter everything. BUT the number of events in the filtered workspace is a lot less than those in the original workspace.

Hopefully I'm wrong about this, but if not then there are events disappearing for no reason that I can see.

mtd.importAll()

ws = WISH00029057 # Any of the resent wish event workspaces would be fine.
print "Events in original", ws.getNumberEvents() 
start_time=ws.getRun().startTime()
end_time=ws.getRun().endTime() 
duration =  (end_time - start_time).seconds() 
logname= 'fake_log'
try:
        # delete the log if it already exists
        DeleteLog(ws, logname)
except RuntimeError:
        pass
# Make a fake log that extends for the full duration

for i in range(0, duration):
        t = start_time + int(i*1e9)
        AddTimeSeriesLog(ws,logname, Time=str(t).rstrip(), Value=i+1)
        
generator, info = GenerateEventsFilter(InputWorkspace=ws, FastLog=True, LogName=logname, MinimumLogValue=0.0, FilterLogValueByChangingDirection='Both')
FilterEvents(InputWorkspace=WISH00029057, SplitterWorkspace=generator, OutputWorkspaceBaseName='filtering', InformationWorkspace=info, CorrectionToSample='Elastic', FilterByPulseTime=False)
filtered_ws = mtd['filtering_0']
print "Events in filtered ws", filtered_ws.getNumberEvents()

prints

Events in original 928940
Events in filtered ws 886831

[Owen Arnold]

This could be related to #10022. Not completely sure about that.

[Wenduo Zhou]

There are more than 10K spectra that do not have detectors. Therefore, those spectra will be ignored by FilterEvents(). The number of events in those spectra will be shown in log. And it is exactly the difference between the number of events in original workspace and filtered workspace.

[Owen Arnold]

Agreed. I loaded the new WISH IDF via #9980 and no events were lost.

There is however a strange situation which I don't understand. Modifying the duration in the script above does not change the number of events produced? Even when making the duration very small indeed, I never seem to be able to filter events?

[Stuart Campbell]

This ticket has been transferred to github issue ​10886