Consider b tagging 2017 procedure, and generally handling 2016/2017 differences

[watson-ij]

BTagging in 2017 will use DeepCSV while 2016 uses CSVv2. I would like to keep the repository unified (i.e. don't split off a separate branch for 2017 if we can help it). How can we do this if the structure of the nanoAOD changes?

[jshlee]

keep both variables in header??

[watson-ij]

We need to add some 2016/17 switch in the code though, no one has worked on 2017 as far as I know, so this is still tbd

[quark2]

Um, I don't understand the situ. We already have variables for both of CSVv2 and DeepCSV; see https://github.com/CPLUOS/nano/blob/master/analysis/interface/Events_2016v4.h#L190-L193. Is there no CSVv2 in nanoAOD production codes for 2017 (although it might not cause a serious problem, just a unharmful warning and that's all, I think)?

[jshlee]

@watson-ij how did you want to handle the difference?
i think in 2017 we will use deep B tags.
we can either have the analysis code tell the difference between 2016 and 2017 or have a loose/medium/tight btag cut which changes for 2016 and 2017 during production