Child algorithms don't propagate their time estimates up

[Martyn Gigg]

When testing #3620 it became apparent that the time estimation functionality on the progress notifications does not work when using child algorithms.

A test example is:

• Use the Load algorithm to load file that takes some time, e.g. Data/SANS2D/SANS2D00005512.nxs and the progress is reported as just a percentage;
• Now use LoadISISNexus to load the same file and the percentage also comes with a time estimate.

[Nick Draper]

Moved to iteration 33 at iteration 32 code freeze

[Nick Draper]

Moved at end of release 2.1

[Nick Draper]

http://www.mantidproject.org/Release_Notes_2_2_Full_List

[Nick Draper]

Moved to milestone 2.4

[Karl Palmen]

It looks as though Algorithm::handleChildProgressNotification will need modifying to do this. The estimated time used by the parent could be equal to

childEstimatedTime + AfterChildEstimatedTime

where

AfterChildEstimatedTime =
((1 - m_endChildProgess)/ (m_endChildProgress - m_startChildProgress)) * EstimatedTimeOfWholeChildAlgorithm

This issue of how EstimatedTimeOfWholeChildAlgorithm is calculated still needs to be decided. It could be simply extrapolated from the estimatedTime of the child progress, but this might be unstable and could be stabilised by using data from earlier progress reports of the child algorithm.

Also child algorithm may not have an estimated time to propagate and I do not know how to determine whether this is so.

[Karl Palmen]

Another approach would be to ignore the time estimate in the child algorithm and get the parent algorithm to do it's own timing whenever the child algorithm reports its progress.

[Karl Palmen]

Pass Child Estimated Time if Final re #3762

Passing the unmodified estimated time of a child algorithm that finishes when its parent finishes, fixes this ticket for the use case of SANS2D00005512.nxs described in this ticket.

Signed-off-by: Karl Palmen <karl.palmen@…>

Changeset: f80fc24e1c4da6bec1dde8e1eb20c772ef69fcc9

[Karl Palmen]

Reverse change cos of side effect in SaveNexus re #3762

Signed-off-by: Karl Palmen <karl.palmen@…>

Changeset: 6ed752a52bfe5a86e9952f90a1e4407caada3a8e

[Karl Palmen]

Pass Child Estimated Time if Final re #3762

Passing the unmodified estimated time of a child algorithm that finishes when its parent finishes, fixes this ticket for the use case of SANS2D00005512.nxs described in this ticket.

Signed-off-by: Karl Palmen <karl.palmen@…>

Changeset: e7afbcfcb14abe6d69cf1fc1c2d878bc88040285

[Karl Palmen]

Reverse change cos of side effect in SaveNexus re #3762

Signed-off-by: Karl Palmen <karl.palmen@…>

Changeset: b0d5356bd2a5ad2d7599f939eb5c79aea8cee337

[Karl Palmen]

I think a major change in progress reporting may be needed for this. I think we need a way to share a single progress object with subalgorithms or functions within a algorithm in such a manner that to the subalgorithm or function, it behaves as though it were created locally, as in current practice.

[Nick Draper]

Moved to the Backlog after the code freeze for R2.6

[Nick Draper]

bulk move to assigned at the into of the triage step

[Stuart Campbell]

This ticket has been transferred to github issue ​4609