Switch CP algs to ConfigBlock configuration
Follow up to !257 (closed)
This requires a new AthAnalysis release with various fixes. For now it is expected to fail in CI.
Instead of building all calibration sequences independently as before and returning ComponentAccumulators, we have to build multiple ConfigSequences out of the component ConfigBlocks, and return those to cpalgs_config, then add the sequences together and use a ConfigAccumulator to stitch together the full setup. This addresses interdependencies between all the individual collections and MET + OR.
At least for now, we expect to use a separate ConfigAccumulator for the trigger and preselection so as not to have to pass ConfigAccumulators too far up into the main function of the executables. However, it is likely that we will propagate some information from the CP alg sequence to the TTree writing, so it may end up not being that big of an inconvenience to pass a ConfigAccumulator around alongside the flags.
Merge request reports
Activity
added 16 commits
-
ca6d228e...116c8600 - 4 commits from branch
easyjet:master - 116c8600...c79950f0 - 2 earlier commits
- df5e5a85 - Add SystObjectLink config block
- 669ef016 - Janky VR jet ConfigBlocks
- ece45ccf - Make jet sequences work, add view alg creation once more to calib sequence
- 23bb1776 - flake
- 359ca165 - MET in ConfigAccumulator, needs cpalgs_config to combine ConfigAccumulators
- 61db923f - Remove irrelevant event counter
- f152eb0d - Add allcalib jet container name
- 29bfacbe - All calibration via config blocks
- e48a4eb1 - ConfigBlock overlap removal, tree output not working
- 833bf032 - Bugfix PRW
Toggle commit list-
ca6d228e...116c8600 - 4 commits from branch
added 2 commits
mentioned in merge request !263 (merged)
mentioned in merge request !257 (closed)
added 2 commits
added Code:Configuration Code:Refactoring labels
-
Just a small comment, otherwise this looks very good to me. Thanks a lot @khoo for setting this up
Do you want to keep the old analysis sequence comments also on the long term? This can be cleaned up at a later stage of cours.
Many thanks, @tstreble!
Yes, I thought to leave them in for the moment. Partly it was to help with the review actually, so we have some guidance as to whether I overlooked any options. It's true that this will simply be in the diff but I was concerned that there might need to be some follow-up config block adjustments.
Ideally we should do a comparison of the ntuples before/after in case somebody has the capacity to do so, but it's a bit less trivial for the overlap removal changes.
I'll make an issue to take them out, maybe with the 24.2.20 iteration.