Make Stitch1D easier to use and more robust

[Nick Draper]

1. I'd like to see this move so it can perform a sensible default rebinning across both if they are not binned in advance
2. If the region of overlap is not defined we should guess based on the (pre-rebinning) overlap of the x axes.
3. The handling of zero errors needs to be improved.

#Creating a couple of fake workspaces for this test
CreateWorkspace(OutputWorkspace='bob',DataX='10,20,30',DataY='10,20,30',UnitX='Counts')
CreateWorkspace(OutputWorkspace='bob2',DataX='25,35,45,55,65',DataY='20,30,40,50,60',DataE='5,10,15,20,25',UnitX='Counts')
#You need to rebin both to a common X axis
Rebin(InputWorkspace='bob',OutputWorkspace='rbob1',Params='0,10,100',PreserveEvents='0')
Rebin(InputWorkspace='bob2',OutputWorkspace='rbob2',Params='0,10,100',PreserveEvents='0')
#Then stitch will do its job
Stitch1D(LHSWorkspace='rbob1',RHSWorkspace='rbob2',OutputWorkspace='result',StartOverlap='0.2',EndOverlap='0.4')

[Owen Arnold]

I'm going to use combineMulti.py as a basis for how to do the processing here, and gut the existing Stitch1D to implement these changes. I can then eventually replace processing of combineMulti to use Stitch1D. The stages I'm going to follow are:

• Gut Stitch1D, decouple it from Stitch1DMD.
• Create a test dataset that works with combineMulti, and use that as a benchmark
• Implement Stitch1D so that it meets the criteria determined by the benchmark

[Owen Arnold]

refs #7709. Gut Stitch1D and change API slightly.

This should bring the API to be more in line with combine2 from combineMulti. These changes will break the unit tests.

Changeset: c3ebc452dee2fdb7aab8b78abb28fc30e9c40eac

[Owen Arnold]

refs #7709. Stitching seems to be working ok.

Changeset: 9ef471367613700f8f5ff56a94f5a627caf3e9b5

[Owen Arnold]

refs #7709. Sort out masking ranges.

This is stitching very nicely now with the test data. The most important change here is that I've limited the zero-masking range for the rhs and lhs rebinned workspace, so that the boundaries can be handled properly.

Changeset: 4eed2761616411a39660a4eed2a2e5e2fb5ddd06

[Owen Arnold]

refs #7709. Update tests, and fix algorithm.

Tests updated. Minor changes required to algorithm as a result of problems found during testing. More tests to come.

Changeset: 81b2ac77b7ff8afefb1bfad4a58a259be82a1a93

[Owen Arnold]

refs #7709. Tests and fixes for manual scaling.

Get manual scaling factor aspects of the algorithm working.

Changeset: 6eab90d14fa302929fa3d565784942a84b4bc921

[Owen Arnold]

refs #7709 sanity check inputs.

Unit tests have been added for this too.

Changeset: 7ff02b26c1e1051dc9aff451b8300b3f4461a2a1

[Owen Arnold]

Tester: Obtain the zip file from #7982. Then run the following script. Stitch1D should produce the same results as combine multi, except that the outstanding bug in #7982 (the saw-tooth behaviour) requires aspects of this fix. Otherwise the two generated plots should line-up quite nicely. Also see new unit tests.

from isis_reflgui import combineMulti

import os

DataDir= os.path.expanduser('~/Desktop/Tims')
print os.path.join(DataDir, 'polref8481.dat')
r8481 = Load(Filename= os.path.join(DataDir, 'polref8481.dat'),Unit='MomentumTransfer') 
r8482 = Load(Filename= os.path.join(DataDir, 'polref8482.dat'),Unit='MomentumTransfer')
r8483 = Load(Filename= os.path.join(DataDir, 'polref8483.dat'),Unit='MomentumTransfer')
r8481 = SortXAxis(r8481)
r8482 = SortXAxis(r8482)
r8483 = SortXAxis(r8483)

Rebin(InputWorkspace='r8481',OutputWorkspace='r8481',Params='0.009,-0.02,0.034',PreserveEvents='0') 
Rebin(InputWorkspace='r8482',OutputWorkspace='r8482',Params='0.017,-0.02,0.06',PreserveEvents='0')
Rebin(InputWorkspace='r8483',OutputWorkspace='r8483',Params='0.027,-0.02,0.100',PreserveEvents='0')

ConvertToHistogram(InputWorkspace='r8481',OutputWorkspace='r8481h')
ConvertToHistogram(InputWorkspace='r8482',OutputWorkspace='r8482h')
ConvertToHistogram(InputWorkspace='r8483',OutputWorkspace='r8483h')

wcomb = combineMulti.combineDataMulti(['r8481h' , 'r8482h'],'viaCombineMulti',[0.009, 0.018],[0.03, 0.046],0.009,0.1,binning=-0.02,keep=1)
viaStitch1D = Stitch1D(LHSWorkspace='r8481h', RHSWorkspace='r8482h', StartOverlap=0.02, EndOverlap=0.03, Params='0.009, -0.02, 0.1')
plotSpectrum(['viaCombineMulti', 'viaStitch1D'], 0)

[Owen Arnold]

refs #7709. count_nonzero not portable.

Changeset: 0f41b2732f2cba50b70a85079b1e74e261aab964

[Owen Arnold]

refs #7709. Remove non-portable zero count from stitch1d.

Changeset: 847cf8b6e4b615ebf716cb115a73ee5dca1b092a

[Peter Parker]

Merge remote-tracking branch 'origin/feature/7709_stitch1D_improvements'

[Peter Parker]

Passing.

The given script does indeed show the combineMulti approach and the new Stitch1D approach lining up nicely, except of course for the saw-tooth behaviour.

The algorithm looks good to me. The code seems to do what I would expect. The arguments passed to it are validated, and it cleans up after itself. The Wiki could do with updating to show the Params property, though I imagine you've just not gotten around to that.

The unit tests seem to offer good coverage, and run on my machine fine.

[Stuart Campbell]

This ticket has been transferred to github issue ​8554