Data Structures Task Force, Minutes, Feb. 5, 2002, 08:00 PST, VRVS, Venus
Present: Heather, Toby, Traudl, Michael K, Tracy, Hiro T, Sasha,
Leon
Toby gave a presentation of the GParticle
class.
- GParticle was adopted over the traditional vertex/particle model because
it was thought to be easier to understand and use.
- Inherits from HepLorentzVector
- There are no explicit vertices; these are implied in the links.
- Mass is included because an unstable particle can have a mass other than
the PDG value.
- All links are single, except for parent<->(first) child.
- Methods are important, for example, a smart constructor that does all the
links.
- There are some associated structures, for example, the list if steps if
there was multiple scattering.
Traudl presented her proposal.
She outlines the reasons for preferring this model over the current one (v4r1). Most,
if not all of her objections were with respect to this (current) one, not the
Gismo model.
- In this model the MC tree is represented by the links (particles) and
nodes (vertices), which is supposed to parallel the G4 model..
- Having vertices as separate objects make traversing the structure more
straight-forward.
- Having the same information in more that one place isn't a good idea,
especially if the structures will be used in an interactive environment.
(Could be fixed with smart constructors?)
She also produced a UML
for v4r1. It became pretty obvious during this discussion that we don't really know
what the current model is supposed to mean. Rumor has it that Richard
knows.
- After the meeting, Traudl did
some work on v4r1. Here is the new UML,
and here are her comments.
Traudl asked again how we plan to associate Hits/Digis with particles, and
also reconstructed tracks with particles.
At the core meeting
- there was
some discussion of the use of some kind of relational table (Gaudi
associator, or equivalent in new Gaudi, C++ map?) to do this. Everyone
agreed that the relation should be stored separately from the objects
themselves.
- Also, there was some sentiment in favor of making up some use
cases for the different models. I've made up some for GParticle.
(I'm not positive that they work!)
Action Items:
- Toby will report on the G4 particle/vertex structure
- Traudl will report on the Atlas particle/vertex structure
- Leon will try to find out what the v4r1 particle/vertex structure
really means.
- In view of the item below, we might want to hold some smaller meetings
of the interested parties to try to sort out the MCParticle question.
Next meeting: Our next regular meeting would be on Feb. 19, but this
in in the middle of the G4 Workshop at SLAC, so this might not be such a good
idea. (Who will be attending this meeting?)