Skip to content

2012.08.23 Weekly Check In

demory edited this page Aug 30, 2012 · 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!
Clone this wiki locally