Weekly check in 2009.10.16

Call details

• When: Friday, October 16, 2009 at 1:30pm PST/4:30pm EST

• Number: 917-388-9050

• Room: 8

• PIN: 1234

If possible, you should join #opentripplanner on freenode so we can pass relevant links, etc around during the call.

Agenda

• check-in

• discussion of tickets

Minutes

Check-in:

• everyone pleased with outcomes from last meeting

Discussion of #21

• Brian has all tests passing on branch

• Code is well-tested both from having been used for some time and has unit tests

• fully handles GTFS spec

• TODO: Brian will merge into trunk in the next two days

Discussion of #23

• DECISION: agreed to jump to Java 1.6 as default

Discussion of a few different trip planer problems:

• how do we want to handle different projections? Should just do everything in lat/lon internally and reproject in the output?

• things to consider:

• performance hit from distance calculations (can be improved by pre-calculating)

• routing across large geographic distances

• mapping form lat/lon to Vertex (#22) - can this be handled with an existing lib? (GeoTools, JTS, ?)

• why are there non-spatial vertices?

• Brandon: requires a long answer, will send out message to listserv. Short answer: encodes user state in the graph (tradeoff between storing in State, which leads to non-optimal trips from Dijkstra's algorithm, and storing in graph, which increases the size of the graph)

Lack of documentation

• it'd be great to have both more documentation for the code and the trip planning problem in general

• DECISION: start adding Javadoc comments

• TODO: Nick will stub out documentation pages