Skip to content

Weekly check in 2012.01.26

Andrew Byrd edited this page Dec 17, 2014 · 1 revision
  • 13:31 <novalis_dt> Hi everyone.
  • 13:31 <demory_> Hey there
  • 13:31 <demory_> We expecting anyone else?
  • 13:31 <FrankP> hey
  • 13:31 <andrewbyrd> hello
  • 13:32 <demory_> we let's get started. I'll go first..
  • 13:32 <mattwigway> Hi
  • 13:33 -!- grant_h [d819d277@gateway/web/freenode/ip.216.25.210.119] has joined #opentripplanner
  • 13:33 <demory_> First, a quick TranspoCamp recap. We had a good session on OTP Saturday morning where we focused on the TriMet RTP launch, the OTPSetup (automated deployment) project, and the most recent work on analytics. Thanks Andrew for the last minute help there, we did get the accessibility raster working for Portland and it looked really good.
  • 13:33 <demory_> Overall I thought it was all well received and and there were several people from agencies/government that seemed interested in what we're doing. So perhaps there will be some follow ups from that to report in the future..
  • 13:34 <andrewbyrd> Great, I'm glad I got it to you in time.
  • 13:34 <demory_> Otherwise, I'm currently working primarily on debugging/testing OTPSetup.
  • 13:34 <novalis_dt> I springified otp-analyst, and spent a bunch of time fighting with Maven.
  • 13:35 <demory_> That sounds fun
  • 13:35 <FrankP> Ah, Maven...
  • 13:35 <novalis_dt> Well, actually, the spring stuff wasn't too bad. A bunch of wide-ranging changes, but that was good since it let me take a tour of the codebase.
  • 13:36 <novalis_dt> But Maven, well, yeah
  • 13:36 <andrewbyrd> I wonder what is the average percentage of development time spent on making Maven work.
  • 13:36 <novalis_dt> Now I'm looking into the layering issue in otp-analyst
  • 13:36 <demory_> Well thanks for doing that. Once I'm to a stoppiing point w/ OTPsetup I'd like to take a closer look at that code
  • 13:37 <andrewbyrd> novalis_dt: thanks for springifying, it works great.
  • 13:37 <novalis_dt> I feel more confident with Spring these days, which is nice.
  • 13:37 <andrewbyrd> novalis_dt: you mean the base map tiles popping over the travel time raster?
  • 13:38 <novalis_dt> yeah
  • 13:38 <demory_> novalis_dt / andrewbyrd: Specifically I'd like to replace the demo at analyst.opentripplanner.org w/ the raster-based one -- is it to a point where that would make sense?
  • 13:39 <andrewbyrd> demory: Yes, the tile-based version works well now.
  • 13:39 <novalis_dt> It looks to me like there's basically no front-end for the tile-based version?
  • 13:39 <demory_> ok, great
  • 13:39 <novalis_dt> That is, analyst.otp.org has the whole purple line thing going
  • 13:39 -!- Pj__________ [617375b0@gateway/web/freenode/ip.97.115.117.176] has joined #opentripplanner
  • 13:40 <demory_> yeah I might have to spend some time on a simple front end
  • 13:40 <demory_> but if it's wms-based it should be pretty simple to plug into somethign like leaflet
  • 13:41 <andrewbyrd> I've been cleaning up WMS and an OSM/Google-style X/Z/Y tile service for analyst.
  • 13:41 <andrewbyrd> demory_: there's already a very simple leaflet client
  • 13:41 <novalis_dt> andrewbyrd, how does that work with other configuration options -- traverse mode, etc
  • 13:41 <andrewbyrd> you can use either a WMS layer or a tile layer, both should work
  • 13:43 <andrewbyrd> novalis_dt: you can't send the traverse options in, at least as is. I suppose they can just be passed in as query parameters. That defies the standard, but both client libraries I've used allow it.
  • 13:43 <demory_> andrewbyrd: great, I'll take a look at it
  • 13:44 <andrewbyrd> demory_: definitely do take a look, it's nicer than what you used on Saturday.
  • 13:44 <demory_> i'm sure it is
  • 13:44 <andrewbyrd> Other than that I'm just making the graph/graphbuilder not use vertex labels before I merge in my branch.
  • 13:45 <mattwigway> andrewbyrd, where do the vertex labels get removed? I use vertex labels for elevators, we may need to change my code a bit.
  • 13:47 <andrewbyrd> mattwigway: If I break anything, I will fix it up when the time comes. Thanks for pointing that out though.
  • 13:48 <mattwigway> Ok, thanks.
  • 13:49 <demory_> Great. So, anything else? TriMet folks, anything to report?
  • 13:50 <novalis_dt> How's that OSMSplit treating you?
  • 13:50 <grant_h> Not too much here, I'm almost done with the presentation that proposes some features we'd like to implement for users with disabilities
  • 13:51 <grant_h> Bibiana wanted me to run it by here before I send it out and she's out of town till Wednesday, so I should have it for you this time next week
  • 13:51 <demory_> grant_h: thanks, I'd be interested in seeing that once it's ready
  • 13:51 <novalis_dt> Cool!
  • 13:52 <FrankP> Not much here (been sick this week). A few weeks back, C-Tran/SMART/Cherriotts/TriMet met up to discuss what it would mean to have a single trip planner for northern Oregon/s. Washington (I-5). Nothing imminent beyond building a test graph with all agencies included. But challenges of supporting such a system (e.g., testing multiple agencies data; supporting customers; etc...) are a bit of a worry.
  • 13:53 <demory_> Ok, thanks. Let us know how the testing goes
  • 13:53 <novalis_dt> Are there tools we could build that would make the data analysis easier?
  • 13:54 <FrankP> For multiple agencies, Dave?
  • 13:54 <novalis_dt> Yeah
  • 13:55 <novalis_dt> I'm not thinking of anything specific, but if there's something that sounds useful, let us know
  • 13:56 <andrewbyrd> What do you think, should we make the analytics WMS server accept all the usual OTP query parameters? Maybe just date/time, origin point, and mode. I believe optional enum-like sample dimensions are allowed by the WMS standard.
  • 13:56 <novalis_dt> andrewbyrd, well, I would like to have banned routes for the purple line demo.
  • 13:56 <FrankP> Okay. It's more that we have a trip planner that other agencies would need to test (and customers to support)...not sure how that would work, and not sure how far some of the agencies (who feel strapped for resources) would step up.
  • 13:56 <mattwigway> FrankP, 511/MTC here in the Bay Area supports lots of agencies, they might have some insights on multiagency trip planning.
  • 13:57 <demory_> novalis_dt, andrewbyrd: +1 on banned routes for the analyst
  • 13:57 <mattwigway> Then again, they're a regional body not a single agency.
  • 13:57 <FrankP> Yeah, we should talk to those folks, mattwigway. Bibi and I have talked with them before (about GeoServer), and she knows the manager pretty well.
  • 13:57 <andrewbyrd> novalis_dt: for a single purpose demo (purple line) we can pass in anything we want... but in the long term I'm afraid we'll run into clients that don't support arbitrary query parameters.
  • 13:58 <novalis_dt> Is there another way to do it?
  • 14:00 <novalis_dt> Tile layers look possible to me
  • 14:00 <andrewbyrd> novalis_dt: a whole lot of things can be expressed as sample dimensions - lists or ranges in different units, or named subsets. I guess as long as the extended parameters are optional (ie server still returns a tile without them) a strict client is still usable.
  • 14:01 <novalis_dt> So, the other thing that we should think about as we go forward is resurrecting the system map code so that we can overlay routes.
  • 14:01 <novalis_dt> Because I feel like that would be useful.
  • 14:01 <andrewbyrd> I guess we can just do http://server.tld/tile/Z/X/Y?{OTP query string}
  • 14:02 <novalis_dt> Sounds good.
  • 14:03 <demory_> novalis_dt: has anyone looked at the system map code lately? do we even know if it still works?
  • 14:03 <andrewbyrd> if you mean overlaying the system map over analytics tiles, yes that would be great. It would make interpreting the shape of the isochrones easier.
  • 14:03 <novalis_dt> demory, nope.
  • 14:04 <novalis_dt> demory, I could volunteer to take a look at that, if you like.
  • 14:04 <demory_> but I agree it would be nice to have some system map support
  • 14:05 <novalis_dt> I'm just not clear on what the overall architecture of the client ought to be
  • 14:05 <demory_> novalist_dt: sure, if you have time. i'll likely have my hands full w/ other things the next few weeks
  • 14:05 <novalis_dt> OK, then that's what I'll do.
  • 14:05 <andrewbyrd> novalis_dt: the system map will just return a vector layer, right?
  • 14:05 <novalis_dt> andrewbyrd, IIRC, yes
  • 14:06 <novalis_dt> Also, tabular data for (scheduled) next departures
  • 14:06 <novalis_dt> Don't know how that works
  • 14:07 <demory_> fyi, there is still this ancient demo at http://demo.opentripplanner.org/ that includes the system map layer. it's being very sluggish for me right now though
  • 14:07 <novalis_dt> demory, yeah, we should replace that
  • 14:08 <demory_> yeah, we should. I'll follow up on that
  • 14:10 <demory_> ok, anything else for today?
  • 14:10 <andrewbyrd> One other thing I forgot to mention - I am also working making the batch/aggregate classes agree with the new analytics code. The idea is to associate an accessibility variable with an existing data set for analysis in R etc.
  • 14:12 <novalis_dt> Actually, andrewbyrd, looking at the api-extended code, it looks to me like it basically loads the relevant portions of the dataset into RAM. But we already have this loaded as part of the graph. Do you think that the GraphVertex refactor, combined with the transit index, could make it possible to do various queries directly?
  • 14:12 <novalis_dt> For instance, "next departures from stop"
  • 14:13 <andrewbyrd> If I understand correctly, yes we should be able to handle queries directly. I'll take a look at it.
  • 14:13 <novalis_dt> andrewbyrd, well, I can just wait until your code is in place, assuming that will be reasonably soon
  • 14:14 -!- mattwigway [~matthewc@adsl-75-36-157-173.dsl.pltn13.sbcglobal.net] has quit [Remote host closed the connection]
  • 14:15 <andrewbyrd> novalis_dt: yes I won't be implementing the whole laundry list of changes before the merge. Should be done soon,
  • 14:15 <andrewbyrd> .
  • 14:16 <novalis_dt> Ok. I'm going to come up with a plan for how to rewrite and the system map code; there will be a tradeoff between simplicity of code and simplicity of setup, and I intend to go for simple setup, since presently the setup is so complicated that we don't even use the code.
  • 14:18 <demory_> novalis_dt: sounds good to me. let me know if you want to chat once you've had some time to think about it. we'll probably want to loop in kpw too since i know he's interested in mapping
  • 14:18 <novalis_dt> Ok.
  • 14:20 <demory_> ok, great. anything else?
  • 14:20 <andrewbyrd> novalis_dt: I see your comment on https://github.com/CloudMade/Leaflet/issues/52 . Leaflet doesn't look too complex, maybe we can patch this an submit it.
  • 14:20 <novalis_dt> andrewbyrd, I'm willing to take a look at that.
  • 14:21 <andrewbyrd> great. I think it would really improve the perception from the user end.
  • 14:22 <novalis_dt> yep
  • 14:24 <novalis_dt> OK, next week, then?
  • 14:25 <FrankP> take care
  • 14:25 <demory_> yep, I think we're done for today. thanks everyone!
  • 14:26 <andrewbyrd> all right, good {morning, afternoon, evening} everyone.
Clone this wiki locally