Project Meeting 2022.03.03 - ActivitySim/activitysim GitHub Wiki

Agenda

  1. Admin Items
  2. Issue Review
  3. CS continued presentation on sharrow

Action Items

  • Joe to email Caitlin (AMPO) to see if the ActivitySim Consortium needs to submit a formal proposal.
  • MTC to update transit skim documentation, particularly the simple skims page (Michelle Bina to assign to Flavia)
  • Jeff Newman to clean up issues and nominate reviewers for outstanding issues

Key Decisions

  • Closing issues requires a second person to review. This person can be anyone, it could be a partner or a bench contractor.

Notes

Consortium presentation at AMPO Conference

  • Minneapolis, October 25-28
  • Suggested presentation topic: status update on real world deployments / success stores from MPOs. The presentation should focus on practical features, maybe the development work done/to be done in Phases 6 and 7.
  • Possible attendees
    • Guy volunteered to present on ARC, he likely will attend
    • Jonathan Ehrlich and Dennis Farmer will likely attend
    • MWCOG might attend

Issues Review

  • #541

  • #538

    • Unable to change default parameters In location choice estimation component models.
    • This is addressed in #539.
  • Resolving Commits

    • Question: Do we want to invoke a formalized review process for even simple code changes?
    • Decisions was made to require a second person to review. This person can be anyone, it could be a partner or a bench contractor. Jeff will nominate someone that likely has the expertise to review. This person can decline and Jeff will go to someone else, or the decliner can nominate someone else.
    • There is bunch for each of the bench contractors to review code.

Technical Presentation: Sharrow (focused on data trees and flow)

  • Merging data
  • Two types of relationships: @ and -> (these are defined in the documentation)
    • @ indexing by lookup, label-based matching
    • -> indexing by position; this is faster but does not work if the files are not in order
  • All mapped relationship need to be on the right side of the relationships need to be unique destination. This allows many-to-one but not-one-to many relationships.
  • Zero-based mapping
    • Zero-based TAZ indexes have been made to MTC example to make things run faster, utilize the -> relationship; remap TAZ numbers at the end to be compatible with other components that expect labels and not positions
    • Any set of skims you provide get mapped to be zero-based contiguous index
  • Data tree is defined for each model component. Many are similar so the code can be reused. Jeff has already done this for all the components of the mtc example code.