2011.11.17 Weekly Check In - mattwigway/OpenTripPlanner GitHub Wiki
13:32 <demory> well why don't we go ahead and get started
13:32 <novalis_dt> OK.
13:32 <demory> perhaps Frank will join later
13:32 <demory> i'll check in first
13:33 <demory> a couple different things going on -- some sporadic work on the intermediate places front-end this week
13:33 <JArod_> Hi all, I am here to ask whether we have a package on multiple-attribute routing?
13:33 <demory> i have an interface that allows for selecting / managing multiple places based on map clicks
13:33 <novalis_dt> JArod_, multiple attribute in what way?
13:34 <JArod_> existing routings are based on shortest path
13:34 <JArod_> but multiple-attribute routing is based on pareto optimization
13:34 <novalis_dt> JArod_, Yes, we do do that.
13:34 <JArod_> cool. Where can I find that?
13:34 <novalis_dt> This week, I added supported modes to the API (so that you can query to see if the system has any trains, say), fixed up the start/end naming a bit, and, with some help from Vivien, fixed a bug with JAI (NED), and a couple of other issues
13:34 -!- FrankP [d819d1bd@gateway/web/freenode/ip.216.25.209.189] has joined #opentripplanner
13:34 -!- grant_h [d819d2f8@gateway/web/freenode/ip.216.25.210.248] has joined #opentripplanner
13:35 <novalis_dt> JArod_, that's actually what we do by default.
13:35 <JArod_> novalis_dt, so we can specify multiple attributes when routing
13:35 <novalis_dt> JArod_, presently we only do time and weight in the main code.
13:35 <novalis_dt> JArod_, but andrewbyrd has been working on a more general case
13:36 <andrewbyrd> JArod_, even the basic shortest paths do pareto optimization between weight and time, check out the MultiObjectivepathServiceImpl for the more general case
13:36 <novalis_dt> Oh, and of course in non-trimet news, I did a bunch of work on the OTP setup code.
13:37 <JArod_> novalia_dt, I see. Let me check out the code for this. Thanks novalia_dt
13:37 <andrewbyrd> JARod_, we can talk about it after this meeting if you like
13:37 <JArod_> surelu
13:38 <demory> I had one question about the intermediate places work now that Frank has joined.. is there a way I can plug into the TriMet geocoder for testing?
13:38 <demory> BTW, i have it working but only for map-click based places
13:38 <novalis_dt> demory, can't you use one of the other geocoders?
13:38 <demory> I may go ahead and check that in as a "phase 1" of the work
13:39 <FrankP> sure...I'll send a url.
13:39 <demory> ..but I'd like to get geocoding working eventually
13:39 <demory> novalis_dt which others are you referring to?
13:40 <novalis_dt> demory, geocoder.us, for instance
13:40 <novalis_dt> It is, of course, nowhere near as good.
13:40 <FrankP> ha! lol novalis_dt
13:41 <demory> OK -- for testing purposes I guess that won't really matter
13:41 <novalis_dt> Well, you can't type pdx to zoo, which are the only two places in Portland that are real. The rest is just smoke and mirrors...
13:41 <demory> In any case I'll still go ahead and check in the initial work today, it is functional on its own
13:41 <FrankP> Recently added ohsu..now we have 3 landmarks
13:42 -!- JArod_ [a9ea731b@gateway/web/freenode/ip.169.234.115.27] has quit [Quit: Page closed]
13:42 <demory> Also am I correct in assuming that intermediate places won't be supported anytime soon for transit trips?
13:42 <mele> yeah it would be nice in the long run but that would involve entering layover times and other complex things
13:42 <novalis_dt> demory, I can turn it on any time for ordered trips; for TSP, performance considerations make that a very bad idea
13:43 -!- Peejay_ [d819d36f@gateway/web/freenode/ip.216.25.211.111] has joined #opentripplanner
13:43 <demory> ok. what I've written assumes unordered for now
13:44 <demory> but I should probably disable intermediate places in the front end entirely when transit is selected
13:44 <novalis_dt> Probably a good idea
13:44 <demory> right now if you try it just says the trip can't be completed
13:45 <demory> ok. well look for a commit from me later today/tomorrow
13:45 <novalis_dt> OK. Frank, Andrew, what have you been up to?
13:45 <demory> that's all I have for TriMet-related stuff
13:45 <andrewbyrd> I'm currently making contraction hierarchies use overlaygraphs, and still experimenting with simplifying the graph used by the bidirectional heuristic, which has the potential to speed up multi-objective.
13:46 <novalis_dt> That sounds good.
13:46 <novalis_dt> Does multi-objective depend on that?
13:46 <novalis_dt> Or can it be merged now?
13:48 <andrewbyrd> Multi-objective is already in master, I believe there are no major differences in my side branch.
13:48 <andrewbyrd> Sure, it works OK with the bidi heuristic
13:48 <novalis_dt> I mean, to make it the default.
13:48 -!- JArod__ [a9ea731b@gateway/web/freenode/ip.169.234.115.27] has joined #opentripplanner
13:49 -!- JArod_ [[email protected]] has quit [Remote host closed the connection]
13:49 <andrewbyrd> You still get some weird sets of trips sometimes (because tradeoffs exist where there shouldn't be any) so I wouldn't use it in the trimet demo, but would like to make it default to get feedback from other users
13:50 <novalis_dt> OK, sounds good.
13:50 <novalis_dt> So, does anyone from TriMet have anything for us?
13:51 <mele> not really, just trying to keep up with what y'all are up to :)
13:51 <andrewbyrd> By tradeoffs in bad places, i mean it's possible to have two itineraries that differ only by which streets you walk on, which is undesirable.
13:51 <mele> Frank? What are you working on?
13:51 <novalis_dt> Yes, that is undesirable.
13:51 <novalis_dt> What's being traded off there?
13:53 <andrewbyrd> It's a long story, but basically walk distance vs the part of aggregate weight that comes from walking. Because of turn costs.
13:53 <novalis_dt> Oh.
13:53 <novalis_dt> Maybe we should simplify turn costs to assume all turns are multiples of 90 degrees.
13:54 <andrewbyrd> The best thing I came up with is to soft-limit walk weight (with turns included) and use that in the dominance comparison.
13:54 -!- JArod__ [a9ea731b@gateway/web/freenode/ip.169.234.115.27] has quit []
13:55 <novalis_dt> So, you tried simplifying, or you think that would not matter?
13:56 <andrewbyrd> Simplifying to 90 degrees? That will still allow tradeoffs between walk distance and walk cost. We want one path to always dominate within the walk legs, unless you are biking and want some fancy tradeoffs between hills and danger.
13:57 <novalis_dt> Yeah, fair enough.
13:57 <FrankP> Me work on Printing...making nice headway, and should have something by early next (borrowing liberally from the Velencia Sp app). Leads me to a question out our OSS license: are folks like TriMet or Valencia that host an instance of OTP obligated to share their code with us, where it touches OTP code? (E.g., Valencia's done a lot of custom work...I warned Francisco that I'd be borrowing, and he didn't seem to mind...but
13:58 <novalis_dt> They're not obligated to share back-end code.
13:58 <andrewbyrd> And even then, with a single on-street term so one path always dominates on streets, if different trip sequences are considered incomparable you still get decent hill/danger tradeoffs via alternate trip combinations that use different stops.
13:58 <novalis_dt> But the front-end code, they're actually distribuging.
13:59 <FrankP> Where is there code repo, novalis_dt?
13:59 <novalis_dt> Which?
13:59 <FrankP> Valencia, Spain...
13:59 <FrankP> Front-end code
14:00 <novalis_dt> Oh, I don't l know that they have one -- but you should be able to just get it from their site.
14:00 <FrankP> Yeah, it's there...minimized
14:00 <novalis_dt> Oh. That's not good.
14:01 <novalis_dt> You should ask Francisco for a non-minimized copy
14:01 <FrankP> I'm not pulling things directly...just borrowing ideas of how they do things. I've got what I need for printing. Yeah, did ask about the other system map code they developed, and Franciso says it was another company (he was the sub-contractor on just the trip planner part).
14:02 <FrankP> Anyway, that's it here...printing coming soon.
14:02 <novalis_dt> Is it based on the OTP front-end code?
14:02 <novalis_dt> Yay, printing!
14:03 <demory> Thanks Frank. If there's nothing else on TriMet stuff I can provide some quick updates on other work
14:03 <mele> as long as we hear how Chattanooga went :)
14:03 <demory> Oh yeah, I forgot about that!
14:04 <demory> That went well I thought. They did get CARTA to share the GTFS (after some initial resistance) and there is a basic deployment running now
14:08 <FrankP> demory, did Chata. use the 2-minute precomiled code, or build from scratch?
14:09 <demory> built from scratch. I actually went up there Sat to lead them thru that
14:09 <demory> Also, been working on the long-overdue new OTP community site. Plan is to have basic site structure/content online by the holiday break next week. After that we may bring in professional graphic help to make it look nice
14:10 <demory> And as David mentioned work is underway on the automated setup / hosting tool
14:10 <demory> So expect more updates on those items in the next couple weeks
14:10 <demory> OK that's all I have.. anything else?
14:12 <demory> Guess not. Thanks everyone!
14:12 <mele> thanks
14:12 <novalis_dt> Yep, thanks! Feel free to email any time as always.
14:13 <demory> Oh, obv no call next week due to T-giving holiday. We'll be back on IRC Dec 1