Present: James Amundson, Bruce Baller, Giuseppe Cerati, Vito Di Benedetto, Lynn Garren, Hebert Greenlee, Thomas Junk, Robert Kutschke, Katherine Lato, Mark Paterno, Gianluca Petrillo, Brian Rebel, Erica Snider, Tingjun Yang, ?

Remote: Saba Sehrish, Dorota Stefan, Robert Sulej

Release and project report [Erica Snider]

• WireCell package integrated (same way as Pandora)
• no requests for feature branch integration

Documentation update [Katherine Lato]

• new material about FHiCL configuration, highly recommended from beginners up

Kalman fitter performance improvements [Giuseppe Cerati]

• a fitter has been available since LArSoft 6.14
• works on a fixed set of hits
  • [Bruce Baller] what happens if the tracks curve and change direction; [Robert Kutschke] this is designed for straight tracks; curly tracks (low energy, stopping tracks) [Giuseppe Cerati] hasn't checked the performances in these cases, but he would be happy to
• [Erica Snider] is the ordering of hits predetermined? [A] the input of the fitter is a recob::Track; that defines a direction, and hits are ordered with respect to it
  • [Thomas Junk] ProtoDUNE will have beam at 45 degrees from z axis [A] can add a parameter to choose the sorting direction
• reusing existing Kalman Filter infrastructure, slow
  • excellent speed up by replacing the UBlas library, tested with ROOT SMatrix
• [Erica Snider] is SMatrix where you want to land?
  • [Robert Kutschke] Eigen was a problem for ATLAS for two reasons: no native symmetric matrices and no way to know if an inversion was successful [James Amundson, after the meeting] has some evidence that that's not true

LArG4 re-factoring project report [Hans Wenzel]

• material properties are specified in two different places (some in FHiCL, some in GDML)
  • [Brian Rebel] the reason is mainly early ignorance rather than fundamental unsatisfied needs
• sensitive detectors are determined by magic names
• pieces of LArG4 come from a Geant4
  • [Brian Rebel] that was a convenience shortcut that we should have removed long time ago
• some validation effort has started (see also Isaac Harris's contribution, presented on LArSoft coordination meeting on August 16, 2016)
• [Erica Snider] you are going to own Geant4 interface, but not the transportation part? [Hans Wenzel] would not mind owning both, so that can be discussed

Conclusions

• limitations coming from material specification in different places and transplant of Geant4 code into LArG4 can (and should) be resolved

TrajCluster changes [Bruce Baller]

• new features, adding new FHiCL parameters
• ArgoNeuT is asking for merging, so that they can start using it

Conclusions

• the merging is approved to happen on next week

Adding metadata to track-hit associaitons for TrackKalmanHit [Tingjun Yang]

• added information at track production in Track3DKalmanHit (association between hit and trajectory point as art::Assns metadata)
• now two associations (with and without metadata)
  • requested art feature 10539 will make this need for two associations vanish
• we would like this to be extended to other trackers
  • [Erica Snider] we are here proposing a new policy for tracking code
  • [Herbert Greenlee] should we wait until the art feature is in? [Mark Paterno] it's not clear why it has not been implemented yet, will follow up
• [Mark Paterno] if the metadata information is such that it can be extracted from the position of the associated item in the associations vector, then additional metadata is not needed
• [Herbert Greenlee] is PMAlgTrackMaker producing two associations now? [A] yes, since a long time
• [Herbert Greenlee] would like a strong policy on associations that are required
  • [Mark Paterno] a vector of art pointer in the data product would enforce that policy

Conclusions

• art team will feed us back with a time estimate for issue #10539
• this metadata is now part of LArSoft track making policy
• the code presented today is ready to go in
• the suggestion of required association information need to be considered