Additional complex jet matching challenge?

[benkrikler]

I recently had to tackle a pretty complicated reco-gen level matching problem, that IMHO demonstrated that a columnar array-based language will be extremely difficult to make expressive enough.

The challenge involved building a ROC curve for a jet-tagging algo, so it involved making histograms of the discimrinator score for reconstructed jets which were matched or not to a gen-level W. The gen-level W had to be within deltaR of 0.8 but if a parent of the gen-level W was a top, the sister gen-level b quark had to be away from the W by at least deltaR of 0.8.

I'm not sure if that's a complete enough challenge description, but it was a challenge I couldn't easily build using simple jagged array manipulations, so I'm curious how we can handle this in an ADL.