Move HltEffChecker scripts onto RDataFrame
This MR replaces the existing ROOT TTree interface of the HltEfficiencyChecker scripts (e.g. hlt_calculate_rates.py) with RDataFrame for speedup. This MR also contains a reworking of the fitting to allow the user to specify if plots should be produced per denominator or with all denominators shown together (thanks @nschulte for the idea and implementation!).
The performance numbers can be found posted below and performance checks are summarised in the checklist below:
-
Performance comparable with master
-
Efficiencies consistent with master
-
Overlaps consistent with master
-
Rates consistent with master
Merge request reports
Activity
added RTA label
assigned to @jagoodin
added 1 commit
- 8b370d7f - Reduce historgamming to two event loops in efficiencies script
added 4 commits
Toggle commit list- Resolved by Jamie Gooding
I've run the tests locally 5 times on the master branch and 5 times on the branch from this MR; this is how the new branch compares to master:
Edited by Jamie Gooding
Hi @rjhunter, this MR is ready for review
I've posted the performance with respect to master master on local tests in the thread above; looks fairly comparable to master on the smaller event numbers/line numbers of the tests, but would expect the parallel behaviour to kick in on larger samples.