Minutes for LArSoft coordination meeting on Mar 26, 2019

Present | Vito di Benedetto, Giuseppe Cerati, Lynn Garren, Krzysztof Genser, Herb Greenlee, Robert Hatcher, Kyle Knoepfel, Marc Paterno, Saba Sehrish, Erica Snider, Tingjun Yang |

Remote | Tracy Usher |

Release and Project status report [Erica Snider]

• Last week v08_13_01 and v08_13_02
• Expect to complete art v3.02 + root 6.16 migration this week
• There may a tag for Genie v3 this week.
• caffe support: LArSoft team's proposal is to drop the support, any objections?
• We will continue to support ubuntu. Releases are available upon request.

Conclusion:

Setting Trajectory Point Index in Calorimetry module [Tingjun Yang]

• calo::Calorimetry takes hit charge in ADC, converts to dQ/dx and dE/dx
  • Discards some hits for various reasons
  • Then saves values for each hit into vectors in anab::Calorimetry
  • Can't match back to hits because don't know which hits have been removed
• Can save the hit index, a vector of tpIndex
• cal::Calorimetry just needs to bee modified to store the tpIndeex
• The index
  • hit to trajectory point mapping is trivial (one-to-one)
  • When TrackHitMeta is absent
    • use the index of the hit in the art::FindManyP entry
    • So indexing is simple
  • When TrackHitMeta is present
    • tpIndex is the index returned from a track hits
    • recob::TrackHitMeta::Index()
  • Discussion of whether the handling of the case w/o TrackHitMeta via FindManyP is reasonable, or if something more simple can be done.
  • Needs to be in place by next ProtoDUNE production campaign, which is one to three weeks

Conclusion

• Approve this as it as, and then revisit the first case when there is no TrackHitMeta
  • Will discuss this after the meeting
• Feature branch in larerco: feature/jhugon_caloTPIndex

Simple cleanups to LArSoft [Kyle Knoepfel]

• Motivation: Always see commits to add more code and rarely to remove any
• This talk is about changes that donot relate to the software design
  • Ran cloc on develop branch except larbatch - 300-500K lines of code
  • Comments in LArSoft: mostly is commented out code, talk about this later
  • Guidence: Remove more code than you add, the less code you have its easier to maintain
• Step 1: get rid of the files you dont need
  • art modules shouldn’t have header files - it is unneceessary
• Step 2: Remove unnessary header dependeencies
  • larevt: SimWire_module.cc - showed the effect of headers
• Proposal: LArSoft should adopt a policy where header files include the minimum number of header dependencies
  • It is a maintance issue
• Remove unnessarry link-time dependencies - results in minor savings in build time
• Remove unnecessary functions
  • If no work is done by beginJob etc, donot prrrovide override
• Remove inappropriate preprocessor use
  • Root no longer suppoerts GCCXML
    • replaced by ROOTCLING
  • Do not put header guards in implementation files
• Simplify code
  • Iterating over std::map
• Next steps
  • LArSoft should benefit from adopting several policies
    • When should header dependencies be introduced
    • When should link-time dependencies be introduced
    • What shoudl the header-guard convention be?
    • What about error handling?
      • Seeing cet_enable_asserts() in CMakeLists.txt
        • Lynn: should not be happening
• clang-format can help w whitespace
• clang-tidy might be able to help w some of the code convention things. Have not looked in into this.
• [Krzysztof Genser]: Here is the link to Chris Jones talk on static code analysis.

Conclustion:

• Feature branches
  • feature/knoepfel_cleanups branches in LArSoft in all repositories
• Will provide concrete brnahces to Lynn in a week or 2.

Small updates to the LArSoft Event Display [Tracy Usher]

• Waveform and space point displays
• Proposed changes
  • Waveform drawing part of the 2D event display is mmodified to use art tools, it allows interchangability between Gaussian and Gaussian Skew hits
    • Allows multipel types of same waveform to be overlaid
    • Configurable via FHICL
  • 3D display of SpacePoints
    • Use art tools
    • Allows variations of the params to draw the points
  • In each case, have preserved ability to modify input parames from teh viewer
  • Available for trial on the LArEventDisplay branch feature/user_wiredatadrawer
• Module TQPad is responsible for drawing waveforms
  • re-architected to use the above tools
  • Noted case of DP which has skew gaussian hits. ICARUS too
  • Move all the actual display code out of TQPad
    • ownership of histograms in the tools
    • Remove the if-else block to differentiate DP and SP
  • Drawing toos in subfolder wfHitDrawers
• 3D SpacePoint drawing - provide flexibility how spacepoints are drawn
  • Color coding options
    • Plain white,
    • heat-map coloring by chisq
    • heat-map by charge
  • Configurable
  • All implemented in art tools
• Status
  • all in feature/usher+wiredatadraweer
  • Current w LArSoft v08_13_02
  • Tested for SP detectors
    • ICARUS, ProtoDUNE
  • Have not had a chance to test DP display
  • Should be non-breaking, unless they have private ED fhicl files

Conclusion

• Test with DP and get it done by tomorrow to be included in this week's release.
• ArgoNeuT needs to test. Tingjun can do this before tomorrow.