Project Meeting 2016.04.08

Project Administration

1. Ben (@bstabler) waiting to hear back from OSG about getting started
2. We'll have at least draft deliverables for task 2 by 4/22
3. Jeff (@toliwaga) out next week

Updates on Current Milestones - Task 2

0. In addition to the items below, RSG's goal for this task is to understand how things work, how far the current implementation is from full scale, to identify holes, etc. - and we're making good progress on this knowledge transfer goal

1. Getting Started Tutorial

   • Ben debugged/traced the entire example and drafted a few documents:
     • data flow through the example modules
     • data schema, when and where tables and columns are created
     • major items skipped in the existing example, which is similar to Progress-Report
   • Ben also working on an Example section for the documentation
     • overview of the system design
     • setup - config files and expression files
     • data schema (a central place to look for table and column definitions) and how data I/O works
   • Example wasn't broken, just needs pandas 0.18
   • Ben working on resolving conflicts with the [trip mode choice branch] (https://github.com/activitysim/activitysim/pull/55) and updating the example documentation as a result
   • Ben will update the Progress-Report
   • We discussed migrating to a 20 zone test system in addition to what we have now for a better example and test bed and agreed to create a zone 1-190 (SF county) test.
   • Ben write script to create 1-190 test inputs
   • Wu (@wusun2) - would like support for generic labels other than 'TAZ'
   • Will push draft of materials before 4/22 for team to review

2. Large Scale Performance Test

   • Ben and Jeff wrote little helper script to convert matrices to OMX and merge the TM1 skims into one file
   • Jeff updated skim(s) classes to use all the skims
   • Jeff updating code to process all tour types and will turnoff household sampling as well for testing
   • Jeff working on fixing destination TAZ bug
   • Will need to read all 5 time period skims as well for test
   • Reading OMX matrices from disk into Python objects is relatively slow; may try to first pickle all the skims so they are closer to Python native format when being re-read from disk