Follow-up from "Save full hit position information"
The following discussion from !182 (merged) should be addressed:
-
Ensure dR cut is taken from the config and not hard coded (see !282 (merged)) -
Save split cluster probabilities (see !286 (merged)) -
Potentially remove Hit
class, moving the functions toHitDecorator
(see !231 (merged)) -
Use single struct for hits in the config and make it optional -
Add code to remove overlapping clusters in phi (I'd do this probably in a separate class that can be used by HitDecorator
andHitWriter
) (see !309 (merged)) -
Use split probabilities to adjust the number counting of hits (with an extra flag to turn this on/off) - should affect only the decorator since in the writer the full info exists (see !337 (merged)) -
Save deta(hit, jet)
&dphi(hit, jet)
for each hit, to give any MVA a headstart -
Order hits by dR(hit, jet)
, so that if nhits is reduced we keep the most relevant hits. (see !455 (merged)) -
Add radius of each hit to complement the deta
anddphi
variables. Layer is not sufficient when using both barrel and endcap hits (see !455 (merged)) -
Save the track ID for hits that are used on reconstructed tracks (for shared hits we could use the higher pT track for a first go) Make sure dR selection is sensible for VR jets. See #16 (closed)-
Save the truth origin and truth barcode for each hit, based on the truth barcode stored on the cluster