|
[Rivet] Suggestion: change to compare-histos default behaviourAndy Buckley andy.buckley at ed.ac.ukTue Jul 27 11:48:45 BST 2010
Hi everyone, I've encountered quite a few confused new Rivet users recently who have tried writing their own MC analysis and then find that compare-histos will not write out any histogram .dat files from their analysis' .aida output file. The "correct" solution is to use the --all flag, or one of its less memorable synonyms, but I think we should actually now change the default behaviour to be the --all mode. The reason for the problem, obvious to anyone involved in the development but opaque to outsiders, is that compare-histos won't write out plots for which there is only one data line, since it assumes that this is a symptom of a multi-energy analysis writing out unfilled histograms for energies other than that used in the current run. That assumption was reasonable and convenient when we were only writing analyses with reference data, and when analyses had no way at the init() histo-booking stage of determining what the run energy was going to be. Both these assumptions are now wrong: we have lots of MC-only analyses with no ref data, and Rivet is being widely used as a prototyping and MC systematics tool for new measurements; and the normal analysis design is now to only book those histograms which will be used, based on init()-time knowledge of the beam conditions. Making simple things simple, and making new users feel in control is, I think, more important than exact interface backwards compatibility at this stage in Rivet's lifetime. So I suggest that for 1.2.2 we (== I) convert compare-histos to always plot all the histograms in the files it is given, unless explicitly disabled. I suggest for now that we keep the defunct --all etc. flags for compatibility, but make them print a "deprecated flag" warning. Opinions? Andy -- Dr Andy Buckley SUPA Advanced Research Fellow Particle Physics Experiment Group, University of Edinburgh
More information about the Rivet mailing list |