2012.01.12 Weekly Check In - atlregional/OpenTripPlanner GitHub Wiki

  • 13:31 <demory> hello everyone
  • 13:31 <novalis_dt> Hi!
  • 13:31 -!- mattwigway [[email protected]] has joined #opentripplanner
  • 13:32 <demory> I have 1:32.. let's get started!
  • 13:32 <andrewbyrd> hi
  • 13:32 <FrankP> hey
  • 13:33 <demory> My check-in is really all OTPsetup-related; let's come back to that at the end
  • 13:33 -!- mattwigway [[email protected]] has quit [Remote host closed the connection]
  • 13:33 <novalis_dt> Frank reported a bug in how patches were applied. I fixed that bug.
  • 13:34 <novalis_dt> (I hope)
  • 13:34 <FrankP> not much on this end ... I'll test things today, Dave (thanks for the help)
  • 13:35 <novalis_dt> Always glad to fix bugs.
  • 13:35 <novalis_dt> Other than that, I haven't really done much.
  • 13:36 <andrewbyrd> I'm working through the problems around analytics raster interpolation, map projections, web map/coverage service zoom levels
  • 13:37 <andrewbyrd> I want to see if I can do our custom interpolation method without necessarily precalculating a specific grid of points using a specific projection
  • 13:38 <andrewbyrd> currently looking at precalculating a voronoi partition, since the slow part of the interpolation is figuring out which road is closest to an arbitrary point
  • 13:40 <novalis_dt> What about just using a STRTree?
  • 13:40 <andrewbyrd> that's what I'm using now for the precalculation
  • 13:41 <andrewbyrd> you then have to scan through all the nearby roads to find the closest ones
  • 13:41 <novalis_dt> Doesn't OpenGeo come with a Voronoi partitioner?
  • 13:41 <novalis_dt> It's got that sweepline thing...
  • 13:41 <andrewbyrd> when you do that at say 10 meter resolution over an entire urban area it takes too long to do in real time to generate your map tiles
  • 13:42 <andrewbyrd> so I was precalculating that pixel/vertex relationship and resampling it which is all fine until you want to support multiple projections
  • 13:44 <novalis_dt> Well, let me know if you want me to take any aspect of this.
  • 13:44 <andrewbyrd> yes geotools will do it, I'm more interested in the various ways of combining the partition with a spatial index to get the nearest neighbor
  • 13:44 <andrewbyrd> because it looks like finding the distance to all the candidate geometries is the slow part
  • 13:45 <andrewbyrd> ok, thanks. right now I
  • 13:45 <andrewbyrd> am just studying some articles on the subject
  • 13:46 <novalis_dt> FrankP, anything new from you?
  • 13:46 <novalis_dt> Or interns?
  • 13:46 <mele> not really with us, as usual these days :) hoping to start on accessibility-related tagging near bus stops soon
  • 13:46 <novalis_dt> ah, excellent
  • 13:47 <demory> Ok thanks all. And Andrew sorry I haven't been too engaged w/ the analytics discussion -- have been focused on OTPsetup this last week. But definitely plan to catch up on this
  • 13:47 <demory> Once OTPsetup is up and running I think my next big focus will be front-end tools for analytics
  • 13:47 <FrankP> Nothing from me...just the alerts thang. (Looking at SOLR configuration for our geocoder...will open source that and the data loader -- and your work Dave on the OSM loader will play a part, in terms of intersections for the geocoder)
  • 13:48 <novalis_dt> FrankP, awesome.
  • 13:48 <novalis_dt> I am very excited about the geocoder
  • 13:49 <novalis_dt> Will it also use openaddresses?
  • 13:49 <andrewbyrd> demory, no problem, I know you guys are concentrating on the automated deployment system.
  • 13:50 <novalis_dt> OK, anything else from anyone?
  • 13:50 <FrankP> Openaddresses ... no plans to, but I guess it could. The Portland data for open addresses is something we feed into SOLR right now, so that loader will be part of things I suppose
  • 13:52 <demory> If there's nothing else I can give a quick update on OTPsetup.
  • 13:52 <novalis_dt> Awesome
  • 13:52 <andrewbyrd> sure
  • 13:53 <demory> So we've have been making good progress -- something should be online to see in the next week
  • 13:53 <demory> Currently GTFS submittal, validation and all aspects of graph building are automated. We have dedicated AWS instances for the controller (always running, includes the web front-end and the message server) and for the validator and graph-builder (which start and stop as needed)
  • 13:54 <demory> We can have multiple concurrent graph-builders going if needed, but initially we'll just have one with a queue
  • 13:54 <demory> Currently working on the last major piece, which is automatic deployment of the built graph (w/ webapp) on a dedicated AWS instance
  • 13:55 <demory> At least that's the initial plan (one instance per deployment) -- but we'd like to allow multiple OTP deployments on a single AWS instance
  • 13:56 <demory> Especially for smaller deployments that don't need a whole machine
  • 13:56 -!- kpw [422c1474@gateway/web/freenode/ip.66.44.20.116] has joined #opentripplanner
  • 13:56 <demory> Question: has anyone worked much w/ the multiple graphs feature?
  • 13:56 <demory> https://github.com/openplans/OpenTripPlanner/wiki/MultipleGraphs
  • 13:57 <demory> Hey Kevin, good timing. We just started talking about OTPsetup progress
  • 13:58 <demory> Anyway we're considering either using the mult. graphs feature or multiple tomcats on a single instance
  • 13:59 <novalis_dt> I've never tried it, but it's worth testing
  • 13:59 <andrewbyrd> My first instinct would be to just run multiple servers and reverse proxy them
  • 13:59 <demory> But again this is not essential to the initial launch so we have some time to do some experimenting
  • 14:00 <kpw> demory: great!
  • 14:00 <demory> So the other near-term tasks are figuring out dns / hostname issues and perhaps sprucing up the web interface a bit
  • 14:01 <demory> But I think we're well on track to have something to show in DC on the 21st
  • 14:01 <novalis_dt> Maybe we should talk to EvanCC about being able to provision foo.otp.org names?
  • 14:01 <andrewbyrd> sounds great, let us know when it's running and I'd be happy to help with some testing and feedback
  • 14:01 <demory> yeah, kpw do you still want to huddle w/ Evan sometime before the weekend?
  • 14:02 <demory> thanks andrewbyrd, will let you know when something is ready to test
  • 14:02 <kpw> yes, let's shoot for tomorrow or later this afternoon
  • 14:02 <kpw> let me know what works best for you all
  • 14:02 <novalis_dt> I'm free either time
  • 14:02 <kpw> and i can check with him
  • 14:03 <demory> great. I am free until around 6:30 today or anytime tomorrow
  • 14:04 <demory> any other questions/comments about this?
  • 14:04 <demory> (or anything else)
  • 14:05 <andrewbyrd> no questions here
  • 14:05 <demory> ok, great. thanks everyone!
  • 14:05 <demory> talk to you next week
  • 14:06 <novalis_dt> later!
  • 14:06 <mele> bye, thanks!
  • 14:06 <andrewbyrd> bye