SmoothNeighbours: generalize the algorithm

[Janik Zikovsky]

11:22:15 AM EDT] * Owen Arnold invited janikzikovsky [11:24:18 AM EDT] Owen Arnold: Janik, we need a generic SumNeighbours algorithm. It's quite high priority at this end. Did you write the original one? [11:24:45 AM EDT] Janik Zikovsky: Yes I did [11:25:31 AM EDT] Janik Zikovsky: What are the requirements? [11:26:28 AM EDT] Owen Arnold: Pascal wants to be able to find nearest neighbours within a radius and then plot summed counts against tof. [11:27:24 AM EDT] Owen Arnold: I think he'd be quite flexible on how the radius is specified. (like the current algorithm). [11:27:39 AM EDT] Janik Zikovsky: EventWorkspace and/or Workspace2D? [11:27:45 AM EDT] Owen Arnold: Workspace2D [11:28:16 AM EDT] Owen Arnold: I think the algorithm already works EventWorkspace, I guess SumNeighbours should do both? [11:28:41 AM EDT] Janik Zikovsky: It doesn't handle Workspace2D now; it should though [11:28:54 AM EDT] Owen Arnold: According to the wiki it doesn't [11:29:10 AM EDT] Janik Zikovsky: I'm assuming that he wants a moving average, i.e. same # of spectra in the output as in the input? [11:29:54 AM EDT] Owen Arnold: Yes [11:30:32 AM EDT] Janik Zikovsky: OK will do

[Janik Zikovsky]

[11:22:15 AM EDT] * Owen Arnold invited janikzikovsky
[11:24:18 AM EDT] Owen Arnold: Janik, we need a generic SumNeighbours algorithm. It's quite high priority at this end. Did you write the original one?
[11:24:45 AM EDT] Janik Zikovsky: Yes I did
[11:25:31 AM EDT] Janik Zikovsky: What are the requirements?
[11:26:28 AM EDT] Owen Arnold: Pascal wants to be able to find nearest neighbours within a radius and then plot summed counts against tof.
[11:27:24 AM EDT] Owen Arnold: I think he'd be quite flexible on how the radius is specified. (like the current algorithm).
[11:27:39 AM EDT] Janik Zikovsky: EventWorkspace and/or Workspace2D?
[11:27:45 AM EDT] Owen Arnold: Workspace2D
[11:28:16 AM EDT] Owen Arnold: I think the algorithm already works EventWorkspace, I guess SumNeighbours should do both?
[11:28:41 AM EDT] Janik Zikovsky: It doesn't handle Workspace2D now; it should though
[11:28:54 AM EDT] Owen Arnold: According to the wiki it doesn't
[11:29:10 AM EDT] Janik Zikovsky: I'm assuming that he wants a moving average, i.e. same # of spectra in the output as in the input?
[11:29:54 AM EDT] Owen Arnold: Yes
[11:30:32 AM EDT] Janik Zikovsky: OK will do
[11:31:31 AM EDT] Owen Arnold: Also should work for Non-rectangular detectors. should just be able to call getNearestNeighbours on MatrixWS.
[11:31:56 AM EDT] Owen Arnold: Also, as a warning, I found out today that getNearestNeighours is not threadsafe :)
[11:32:26 AM EDT] Janik Zikovsky: Oh really? That's good to know

[Janik Zikovsky]

Refs #4025 starting smoothNeighbors

Changeset: cedd5b15679dc64399bff21dd2d8ca540fb97f7f

[Janik Zikovsky]

Refs #4025 SmoothNeighbours much more general. Much faster method for finding neighbours added on MatrixWorkspace

Changeset: ba12917b6be2d07b9665abbb3ad86f4341e7900d

[Janik Zikovsky]

Refs #4025 wiki

Changeset: 71c7b5e71c446e5af1c34da2bb1340d86acc36bd

[Janik Zikovsky]

Refs #4025 fix for windows that really needs to update their openMP

Changeset: c7373722d7429aaf268b707cf4e2f67a61b8ee6f

[Owen Arnold]

I've been using this algorithm heavily over the last iteration and it definitely does work on instruments with non-rectangular detectors and take input workspace of type Matrix as well as Event.

[Stuart Campbell]

This ticket has been transferred to github issue ​4872