Skip to content

Weekly check in 2012.08.23

Andrew Byrd edited this page Dec 17, 2014 · 1 revision
  • 13:33 <novalis_dt> Shall we get started on a meeting?
  • 13:33 <demory> sure. mattwigway let's come back to the aws setup after
  • 13:33 <demory> hello everyone
  • 13:34 <demory> great news on the kickstarter campaign! thanks for everyone's support there
  • 13:35 <demory> my update: mostly working on data workflow stuff for the iOS / OTP North America app
  • 13:35 <novalis_dt> I've been getting RAPTOR ready to merge, by doing a bunch of cleanup and also adding in the missing features. The current performance story is that it is 3x as fast in Portland, and even better in NYC.
  • 13:35 <novalis_dt> mele, would you be willing to run some tests on it?
  • 13:35 <mele> Absolutely!
  • 13:35 <mattwigway> novalis_dt, does it get faster (relative to A*) the larger the area?
  • 13:36 <novalis_dt> mattwigway, yes
  • 13:36 <demory> also TriMet folks i haven't forgotten about about the elevation / scale improvements but kind of got sidetracked this week -- hope to have something ready by next week
  • 13:37 <mele> No problem, we're pretty much on hold here until Frank gets back next week anyway
  • 13:37 <novalis_dt> mele, awesome. The thing that's stopping me from merging right now is that I've adjusted the graph structure, and I don't know yet whether to try to support both the old and the new structures, or just the new
  • 13:37 <mattwigway> novalis_dt, how have you adjusted?
  • 13:37 <novalis_dt> mattwigway, I've removed the edge-based graph
  • 13:38 <mele> novalis_dt well, just let me know how to access it and what kind of tests you'd like me to run and I'd be happy to do it
  • 13:38 <novalis_dt> And moved turn restrictions/penalties into PlainStreetEdge
  • 13:38 <novalis_dt> mele, let me figure out how to set up a test instance
  • 13:38 <mele> novalis_dt no problem, you know how to reach me when you're ready
  • 13:38 <novalis_dt> Ok, awesome
  • 13:39 <novalis_dt> mele, maybe I'll ask Frank if he is willing to set one up
  • 13:39 <novalis_dt> Once he's back from vacation
  • 13:39 <mele> ok
  • 13:39 <novalis_dt> Just got Andrew's email agreeing with removing TurnEdge etc
  • 13:39 <novalis_dt> Any objections here?
  • 13:39 <mattwigway> that makes long-distance walk/car routing slower, right?
  • 13:40 <abyrd> novalis_dt, also interested in seeing how this will affect the in-memory and on-disk graph size
  • 13:40 -!- kpw [~[email protected]] has quit [Ping timeout: 260 seconds]
  • 13:40 <abyrd> mattwigway, why?
  • 13:40 <mattwigway> I thought someone had said that.
  • 13:41 <novalis_dt> It does preclude contraction hierarchies
  • 13:41 <novalis_dt> Which would make long-distance routing faster
  • 13:41 <abyrd> it prevents us from doing contraction hierarchies, which can certainly speed up pure on-street searches
  • 13:41 <mattwigway> oh, but no one uses those anyhow.
  • 13:41 <novalis_dt> Yeah, if you want that, OSRM might be a better bet
  • 13:41 <abyrd> but those are already our fastest search case, and are much faster than transit searches
  • 13:41 <abyrd> and besides there are libraries optimized specifically for that
  • 13:42 <mattwigway> in that case never mind, no objection.
  • 13:42 <abyrd> right, OSRM
  • 13:43 <abyrd> I've been trying variations on how we store transit timetables
  • 13:43 <abyrd> looking for a balance between search speed, clarity, and good handling of concurrent reading/writing
  • 13:44 <abyrd> the last bit was the real motivation for this work. we've now got predicted stop times coming in real-time from most of the busses in the Netherlands
  • 13:45 <abyrd> and those are updating the timetables in real-time, while each individual search sees a snapshot of the whole set of timetables at the time it began
  • 13:45 <abyrd> without searches or rewriting threads waiting on one another
  • 13:46 <abyrd> so I've been watching the failures as the data comes in and adding a few clean-up steps
  • 13:47 <mattwigway> I fixed tickets 799 (roundabout issues) and 803 (biking on elevators) earlier this week.
  • 13:47 <mattwigway> Unfortunately they're in master not stable yet.
  • 13:48 <mattwigway> because they depend on edge unwrapping.
  • 13:48 <mattwigway> #799 could be cherry-picked, #803 would be harder.
  • 13:48 -!- atogle [~[email protected]] has joined #opentripplanner
  • 13:49 <mattwigway> And for the last few days I've been working on OTP North America stuff.
  • 13:50 <mele> brb
  • 13:50 -!- mele [d819d421@gateway/web/freenode/ip.216.25.212.33] has quit [Quit: Page closed]
  • 13:51 <demory> anything else for the check-in?
  • 13:52 <grant_h> I'll just add that I've been helping to organize OpenStreetMap's State of the Map US conference
  • 13:53 <grant_h> it will be in Portland October 13th and 14th if any of you are interested in attending
  • 13:53 -!- mele [d819d2a5@gateway/web/freenode/ip.216.25.210.165] has joined #opentripplanner
  • 13:53 <novalis_dt> grant_h, you should let me know if I have a speaking slot so I can get tickets.
  • 13:55 <grant_h> novalis_dt: I'll check with Bonnie Bogle from Mapbox (she's the lead organizer), but I can pretty much guarentee you'll have a spot
  • 13:56 <novalis_dt> OK, thanks.
  • 13:57 <demory> ok, anything else?
  • 13:57 <novalis_dt> Nothing here
  • 13:57 <demory> btw, i will miss next week's chat so if someone can send me the transcript afterward i'd appreciate it
  • 13:58 <demory> ok, guess that's it. thanks everyone!

The documentation on this wiki is outdated and should not be used

unless you are intentionally working with legacy versions of OpenTripPlanner. Please consult the current documentation at readthedocs

Clone this wiki locally