Separate database for cross sections of processes

[moritzmolch]

Currently, the cross section has to be filled separately for each sample. This is lots of manual work to do when adding a sample, and it could cause inconsistencies between cross sections for different eras. Also, we do not document the references where the cross sections are coming from.

To overcome these issues, we can collect the cross sections in a central JSON file, serving as the database of cross sections to use for the samples (a separate one for each center-of-mass energy). The entries in this database can follow a schema like this, condensing all important information about the cross section value of a sample.

[
    ...
    {
        "sample_names": [
                "TTto2L2Nu_TuneCP5_13p6TeV_powheg-pythia8",
        ],
        "xsec": 98.03,
        "order": "approx. NNLO(QCD)+NNLL",
        "reference": "https://twiki.cern.ch/twiki/bin/view/LHCPhysics/TtbarNNLO; https://arxiv.org/abs/1112.5675"
    },
    ...
]

The samples are identified by the sample_name, which is the first part of the DAS key (i.e., the DAS key is interpreted as /<sample_name>/<campaign_information>/NANOAODSIM). For some samples, the name changes with the data-taking era, so sample_names should be a list here, containing all different sample names corresponding to the same physics process.

When adding new datasets with the sample_manager, this file could be used to automatically attach a cross section value.

[nshadskiy]

Is there a reason why you want a separate json file for that? I think it would also work to add the order and reference parameters to the detailed json files that we already have.

[moritzmolch]

The cross section would need to be looked up/added again each time you add a sample of the same process for a new era. By this, we could store the cross sections centrally and, if it has already been stored before, attach it automatically when adding the same process for another era to the database.