Production Dataset Naming Conventions¶
We expect that we will be adding to the initial samples generated for an analysis campaign as time and computing resources allow. To allow for this, we define a dataset naming scheme that should accommodate this naturally. We will use this convention starting with second analysis.
In the case of data we define epoch or period specific data samples:
The@STREAM@ field indicates the trigger stream (eg numi or comsics). The
TIMESPEC field indicates the data-taking period or epoch involved, for example
epoch3b. Standard NOvA data-taking periods are documneted here. We will also define a
full which removes any constraints on the data-taking period. The optional
SUBVERSION field indicates multiple passes through the processing chain, and is inherited by child datasets. If no subversion is specified, there should be a constraint to subversion '1' in the dataset. The optional
SPECIAL field indicates any special parameters, for example to indicate a sample with the calibration shifted according to systematic its systematic uncertainty. It can be as long as required, and even add more underscores, but it should reflect any other constraints which give the definition meaning.
For Monte Carlo, the convention is:
GEN, which is indicates the generator (eg CRY or genie). For neutrino MC, we also add the
FLAVORSET field (most commonly nonswap, fluxswap or tau) and the
HORN field, indicating the horn polarity (fhc, rhc or 0hc). We also add the
TIMESTAMP field which corresponds to the timestamp in the fcl file names that is passed down to all the descendant files. This naturally allows us to track multiple generation rounds to produce extra statistics. As with
TIMESPEC, we will also define a global set where this parameter is set to
full that includes all statistics.
The Old Convention¶
Historically this has been the old convention for naming production datasets:
FLAVORSET corresponds to GENIE/CRY. For data, it is replaced with the trigger stream. The
SPECIAL bit can be as long as required, and even add more underscores, but it should reflect any other constraints which give the definition meaning.