2012.02.16 Weekly Check In

• 13:31 <demory> Hello everyone
• 13:31 <FrankP> hey David
• 13:31 <demory> FrankP you expecting anyone else from TriMet?
• 13:32 <FrankP> I thought they'd be on by now...we can start w/out them...
• 13:33 -!- PJ____ [ae199ddb@gateway/web/freenode/ip.174.25.157.219] has joined #opentripplanner
• 13:33 <demory> sure, let's get started
• 13:33 <novalis_dt> OK. I've been fighting with this route-street matching stuff, but it's maybe starting to get into shape.
• 13:33 <novalis_dt> (as it were)
• 13:33 <novalis_dt> Also misc troubleshooting, of course.
• 13:34 <novalis_dt> FrankP, I wanted to ask you about Matt's mail from Saturday re #551
• 13:34 <FrankP> Okay...bookmarks
• 13:36 <demory> novalis_dt is your route-street matching code online / ready for others to look at? I may have some time to help w/ that over the next week now that the OTPsetup is wrapping up
• 13:36 <novalis_dt> demory, a version is online
• 13:36 <novalis_dt> demory, but I have made changes since
• 13:36 <demory> ok
• 13:36 <demory> well i'd be curious to look it over at some point. no hurry though
• 13:37 <demory> since it really fits in w/ a lot of the stylized mapping work i've been doing on the side
• 13:37 <novalis_dt> demory, cool, let me know when you have time, and I'll make sure the latest version is up. Or if you'll be in NYC, we can pair.
• 13:38 <demory> in nyc week after next. i'll let you know if i have time next week though
• 13:38 <demory> My OTPsetup update -- we ran into some unexpected AWS difficulties this week, which prompted some changes to the VPC architecture, but that's all resolved now and I believe the coding work is basically done. The last step is setting up the DNS routing-- we're going to use "deployer.otp.org" as the public gateway for the project though it isn't up yet. That and the visual design work is really all that's left. So look for an official announcement about
• 13:38 <demory> this very soon..
• 13:40 <kpw> great!
• 13:40 <novalis_dt> FrankP, can you explain some of the solr stuff you've been checking in?
• 13:40 <FrankP> I've been working (every so slowly) on a SOLR ComboBox. In combo with a SOLR (search engine) backend, it will give the from & to inputs a Google Suggest type feel.
• 13:41 <novalis_dt> Do you have a solr backend that you could check in?
• 13:43 <FrankP> No backend itself. I'd like to get something out there as a project eventually. But creating a SOLR schema that fits these ComboBoxes is trivial -- hard part is populating SOLR.
• 13:43 <novalis_dt> Yep.
• 13:45 <novalis_dt> KPW, anything from you?
• 13:45 <kpw> Not too much to report: i've just been following from the sidelines with the otpsetup work that demory's been leading
• 13:46 <kpw> also, putting together some proposal materials regarding the analyst functionality
• 13:46 <novalis_dt> Cool.
• 13:46 <demory> Yeah kpw and EvanCC were a huge help w/ the AWS stuff this week, thanks!
• 13:46 <novalis_dt> Great. Anyone have anything for OpenPlans?
• 13:46 <kpw> yeah, looking forward to having that online
• 13:46 <kpw> related:
• 13:46 <kpw> i'd love to put more thought into performance benchmarking
• 13:47 <PJ____> the other interns are at a meeting, but we don't really have anything to mention right now
• 13:47 <kpw> i'd really like to get a sense of how much CPU/RAM it takes to run OTP in production
• 13:47 <kpw> we'll be conducting some benchmarks on AWS
• 13:47 <novalis_dt> kpw, it's hard to know, since it will depend on the particular graph.
• 13:47 <kpw> right
• 13:47 <kpw> but i'd like to know if there are generalities
• 13:48 <novalis_dt> Sure, fair enough.
• 13:48 <kpw> e.g. is there a rough correlation between number of trips and the CPU/memory requirements
• 13:48 <kpw> know things like that will really help inform the hosting side of the deployment
• 13:48 <novalis_dt> Well, memory should be pretty flat across load.
• 13:48 <novalis_dt> But I guess we'll see once we have data.
• 13:49 <kpw> I think it would be useful to compare notes with TriMet as well
• 13:49 -!- mattwigway [~[email protected]] has joined #opentripplanner
• 13:49 <novalis_dt> mattwigway, before you got here, I asked Frank about #551 -- he's done most of the front-end work, so he would be the person to talk to.
• 13:49 <kpw> the reason for all is this is that folks will want cost figures for deployments. I'm not sure we know enough yet to generate an estimate
• 13:49 <novalis_dt> kpw, yeah, that's fair.
• 13:50 <demory> yeah, and a future step for otpsetup is looking at loading multiple deployments on a single instance, so this would be helpful for that too
• 13:50 <mattwigway> novalis, thanks. Maybe a little later on he and I can have a little chat about that
• 13:50 <novalis_dt> cool.
• 13:50 <FrankP> Later this Summer, Kevin...when we move rtp into a more prime position (replacing ride.trimet.org), we'll get traffic that will tell us a lot
• 13:51 <kpw> right!
• 13:51 <kpw> i'm hoping that we can do some work ahead of that to simulate load/benchmarks
• 13:51 <kpw> so we'll have some useful estimates
• 13:51 <kpw> but real-world data will be key
• 13:53 <demory> alright, anything else for today?
• 13:54 <novalis_dt> Nothing here.
• 13:54 <FrankP> I stress tested RTP before it went into production...I have a 24G machine. I give 1/2 the memory to 4 instances of TC/OTP (3 G each). The memory starts at 2.4G at startup, and has crept up and down, going as high as 2.8G .
• 13:54 <mattwigway> If FrankP has some time I'd like to revisit #551, otherwise no problem.
• 13:55 <FrankP> https://github.com/openplans/OpenTripPlanner/issues/551
• 13:55 <mattwigway> I did some work on it at https://github.com/mattwigway/OpenTripPlanner/compare/master...dynamic-bookmarking
• 13:58 <FrankP> does it work, just not on IE?
• 13:58 <mattwigway> Woops, forgot to push my changes... doing that now. done.
• 13:59 <mattwigway> Yes, it works in IE (7 or 8, don't remember), FF 9and Chromium.
• 14:00 <mattwigway> Oh, the "can't set title" commit just changes from using Ext to set title to directly setting window.title, so that it works in all browsers.
• 14:00 <mattwigway> I haven't minified the code yet to make sure it still works, I'm not exactly sure how to do that.
• 14:01 <novalis_dt> The build process does it for you, I think
• 14:01 <mattwigway> The other thing this opens the door to is history management, so that the back button works.
• 14:01 <FrankP> yeah, build process creates otp-min.js
• 14:01 <mattwigway> novalis, it doesn't seem to, I'm seeing the actual files in Firebug.
• 14:01 <novalis_dt> I think it's part of mvn package
• 14:02 <FrankP> there's another .html file called min.html
• 14:02 <FrankP> (as opposed to index.html)
• 14:02 <mattwigway> Oh, thanks FrankP, I'll test that.
• 14:02 <novalis_dt> ah, I did not realize that
• 14:03 <FrankP> min.html just uses the otp-min.js -- I use the otp-min.js on rtp.trimet.org
• 14:03 <mattwigway> The reason RTP looks different is just CSS, yes... it's the same webapp, right?
• 14:04 <novalis_dt> there are some changes to support TriMet's system map.
• 14:04 <novalis_dt> i.e. routes
• 14:04 <FrankP> Pretty much...there's more .js in rtp -- custom stuff for TriMet ... but the trip planner is pure otp-min.js, with different css
• 14:05 <mattwigway> Yeah, it works in FF9 minified, generates URLs like: http://localhost:8080/opentripplanner-webapp/min.html#/submit&fromPlace=53.534290,9.968911&toPlace=53.550815,9.969683&mode=WALK&min=QUICK&maxWalkDistance=840&time=11:04%20am&date=02/16/2012&arr=Depart&itinID=1
• 14:05 -!- FrankX [1815504e@gateway/web/freenode/ip.24.21.80.78] has joined #opentripplanner
• 14:06 <novalis_dt> Awesome.
• 14:06 <FrankX> Pretty much...there's more .js in rtp -- custom stuff for TriMet ... but the trip planner is pure otp-min.js, with different css
• 14:07 <FrankX> Sorry...I got disconnected...what's Awesome? Back Button? (Awsome :0)
• 14:07 <mattwigway> So I guess the question is what this means for link-to-trip functionality. If one can just copy the address, do we need link-to-trip?
• 14:08 <mattwigway> Frank, I just noted that it works with the minified build in FF9.
• 14:08 <mattwigway> And makes URLs like http://localhost:8080/opentripplanner-webapp/min.html#/submit&fromPlace=53.534290,9.968911&toPlace=53.550815,9.969683&mode=WALK&min=QUICK&maxWalkDistance=840&time=11:04%20am&date=02/16/2012&arr=Depart&itinID=1
• 14:08 <novalis_dt> That sounds pretty cool to me. I think we should include it
• 14:09 -!- FrankP [1815504e@gateway/web/freenode/ip.24.21.80.78] has quit [Ping timeout: 245 seconds]
• 14:09 <mattwigway> I changed link-to-trip to generate the #/ instead of ? URLs, but I set up PArseUrlParams to handle ? iff a #/ string is not present.
• 14:09 <mattwigway> That way already generated links continue to work, but if someone gets a ? link, modifies and sends it on, it will still behave as expected.
• 14:10 <mattwigway> I haven't tested on older Firefoxes or on Safari.
• 14:10 <novalis_dt> That's probably worth doing. What about IE?
• 14:10 <mattwigway> I tested with IE, I think 8 but maybe 7, whichever came with my old Windows XP box.,
• 14:11 <novalis_dt> OK, sounds good.
• 14:11 <FrankX> Yeah, I'm +1. BTW, we can turn Link to Trip off in config.js without changing more than a line of code.
• 14:11 <mattwigway> In fact, I think it's even off on rtp, yes?
• 14:11 <mattwigway> Yes, IE8.
• 14:11 <FrankX> Yup, it's off there. BTW, I can test on IE 9 tomorrow.
• 14:12 <mattwigway> Cool, thanks. The other thing it does is dynamically change the window title so bookmarks are something meaningful.
• 14:12 <mattwigway> I shared code with link-to-trip, so enhancements to that will also enhance this.
• 14:13 <mattwigway> Basically it just sticks the GET query string after #/ instead of ?
• 14:13 <mattwigway> For instance, when we fix #599 that should work in both places.
• 14:14 <novalis_dt> Nice.
• 14:14 <mattwigway> The other little wrinkle for a merge is that I stupidly committed some indentation fixes in master instead of elevators-new and then forked that, so there's a commit-revert in there.
• 14:15 <mattwigway> I pushed before I realized I was on the wrong branch, and by the time I realized it Andrew (or someone) pulled my change.
• 14:15 <mattwigway> See the first two commits on https://github.com/mattwigway/OpenTripPlanner/compare/openplans:master...dynamic-bookmarking
• 14:16 <novalis_dt> So, 9f334c00 simply reverts c61da22f?
• 14:16 <mattwigway> As to link-to-trip, what's the reason it's off on RTP?
• 14:16 <mattwigway> novalis, yes.
• 14:16 <novalis_dt> In general, what I do in these cases is to simply use git rebase to eliminate the commits
• 14:17 <mattwigway> It was in a api-webapp file, you'll see there are no actual changes in the diff but the commits are there.
• 14:17 <novalis_dt> But if we've already merged from there, it's fine to keep it.
• 14:17 <mattwigway> I thought about that, no I don't think it's been merged but I think Andrew (probably) has pulled it.
• 14:17 <mattwigway> I'm not a git wizard and I was chicken to make a mess...
• 14:18 <novalis_dt> Yeah, someone described interactive rebase as a chainsaw
• 14:18 <novalis_dt> Which admittedly only made me more eager to learn it :)
• 14:18 <novalis_dt> But I'm that kind of guy.
• 14:18 <mattwigway> I can see one of those commits here: https://github.com/openplans/OpenTripPlanner/commit/9f334c00f3083908d18b3f10306471e9dcbd5542
• 14:19 <mattwigway> which makes me suspect it's been merges somewhere.
• 14:19 <FrankX> I originally created it as a more of a test tool to easily compare Google (and maps.trimet.org) with OTP. Thought it might evolve into an email / SMS / Share This! type of functionality, but never got it done for RTP ... so turned it off to avoid customer confusion
• 14:19 <mattwigway> That's a good point, Frank, we might want to keep link-to-trip so that user can compare with Google.
• 14:20 <mattwigway> If I had one complaint about RTP, not having link-to-trip would be it.
• 14:20 <mattwigway> Frequently friends ask me for transit directions (me being the transit geek and all). Also, when I worked at 511, having a link to a trip made bug-tracking far easier.
• 14:20 <novalis_dt> Maybe it could be a hidden feature -- it would only appear if you enter the Konami code...
• 14:21 <mattwigway> Which begs the question, should there a be a config feature to turn of dynamic URLs?
• 14:21 <mattwigway> I don't really see it confusing people, I think they're used to seeing strange URLs full of numbers on news sites &c.
• 14:22 <novalis_dt> I think it's probably fine not to have that feature until someone requests it
• 14:23 <mattwigway> Having integration with a short URL service (either our own or someone else's) might also be useful in link to trip long term since full URLs get long.
• 14:23 <FrankX> Actually, there is a link to trip hidden in the feedback button (plan a trip, then the email link will have your trip link). More of the confusion would be 'why are you linking to Google' questions...
• 14:23 <mattwigway> Yes, 511 is the same way.
• 14:24 <mattwigway> Should say Bay Area 511 since not everyone thinks of 511 as the same thing.
• 14:24 <novalis_dt> NY 511 has the world's worst trip planner
• 14:24 <novalis_dt> I use one of its trip plans in a comment in OTP explaining what not to do.
• 14:25 <FrankX> "should there a be a config feature to turn of dynamic URLs?" -- I'm the wrong person to ask, since I love switches to turn things on & off. So I'd say yes. (And would probably write one for the dynamic page naming -- just in case the graphic designers who rule oure website get upset.
• 14:26 <mattwigway> novalis, I guess that means someone has requested it.
• 14:26 <mattwigway> I'm kind of buried right now, so it'll probably be a little while before I can do that.
• 14:26 <novalis_dt> Well, we can merge without it. Or are we already fully merged?
• 14:26 <mattwigway> No, dynamic-bookmarking has not been merged. Should I make a pull request?
• 14:27 <novalis_dt> Sure. Or if you already have commit access, you can just push it
• 14:27 <novalis_dt> And if you don't, would you like it?
• 14:27 <mattwigway> No, I don't have commit access. Sure!
• 14:27 <novalis_dt> OK.
• 14:27 <novalis_dt> Added.
• 14:28 <novalis_dt> OK, I'm excited to get back to working on the route matching -- anyone have anything else?
• 14:29 <mattwigway> Thanks. I don't know what OpenPlans philosophy is, but I like to use Pull Reqs on my own projects for ease of code review and for historical purposes
• 14:29 <mattwigway> I'll wait for Frank to test IE9 before merging, do we want to test olde Firefoxes &c?
• 14:30 <mattwigway> Shall I try to rebase, or just leave the commit-revert pair?
• 14:30 <novalis_dt> We tend to just give commit access to anyone who has made serious contributions.
• 14:30 <novalis_dt> Leaving it is fine.
• 14:30 <novalis_dt> It's just worth knowing for the future that it is a fixable problem if you catch it before someone else does
• 14:31 <mattwigway> thanks.
• 14:31 -!- PJ____ [ae199ddb@gateway/web/freenode/ip.174.25.157.219] has quit [Quit: Page closed]
• 14:32 <FrankX> Hey Matt ... if it's working IE 8, it probably works on IE 9. I'm okay (and kinda prefer) that you merge sooner than later...that way, I'll just test trunk
• 14:32 <mattwigway> Ok, I'll merge now then, and we can test when we test other features...
• 14:37 <FrankX> Testing all the versions of FF and IE is kinda difficult, and I don't have a good solution for that ... I usually test a single version of FF, IE, Chrome and Safari (on both Mac & Windows). If things work in IE, then I'm confident I didn't break anything too badly.
• 14:38 <mattwigway> Hey Frank, do we know the latest trunk works in IE8... when minified I get otp.planner is not defined, but of course no useful error messages. It works when unminified.
• 14:38 <mattwigway> I'm primarily a Linux user, I basically start up Windows to test with IE, run ArcGIS and run Excel.
• 14:39 <FrankX> Grrr... Not sure. RTP was working in IE 8...that's the last I tested. Also, IE9 has an IE8/IE7 mode that was working a month back.
• 14:40 <FrankX> I'll take a look today/tomorrow to see if I can find a problem.
• 14:41 <mattwigway> Should I go ahead and merge? I don't think it's my changes, but I suppose it's possible,
• 14:41 <FrankX> Yes please.
• 14:42 <mattwigway> Actually it's really easy for me to test the latest master, I'll do that, if it doesn't work also I'll merge.
• 14:43 <mattwigway> Yep, same with master. I'll merge.
• 14:43 <mattwigway> Incidentally, rtp seems to still work with IE8.
• 14:44 <mattwigway> OK, done.
• 14:44 * novalis_dt pulls
• 14:45 <novalis_dt> Interesting -- git log -p doesn't show diffs on merge commits
• 14:45 <mattwigway> I think you have to specify a parent branch, it doesn't know what to diff against.
• 14:46 <novalis_dt> well, it would be diffing against master @ the prior commit.
• 14:46 <mattwigway> That would be what you would expect, wouldn't you.
• 14:46 <novalis_dt> Yep.
• 14:46 <novalis_dt> But the ways of git are sometimes mysterious
• 14:46 <mattwigway> There definitely are a few changes in there.
• 14:50 <novalis_dt> These commits confuse me, but hopefully they make sense in the end.
• 14:50 <mattwigway> Something specific?
• 14:51 <novalis_dt> well, one says it changes 19 files, and the other 21
• 14:51 <novalis_dt> But I would expect to see about two files changed.
• 14:51 <mattwigway> That can't be right, it should only change maybe two or three.
• 14:51 <novalis_dt> Yeah.
• 14:51 <mattwigway> Yeah, config, params, planner and triptab, so 4
• 14:51 <novalis_dt> I'm looking on github, at https://github.com/openplans/OpenTripPlanner/commits/master
• 14:52 <novalis_dt> And then clicking on the version number
• 14:52 <mattwigway> Oh, it's because I merged master -> dynamic-bookmarking and then dynaimc-bookmarking -> master
• 14:53 <novalis_dt> The order is the opposite, according to github.
• 14:53 <mattwigway> I see that, hmmm...
• 14:53 <mattwigway> Does the code at least work, or have you gotten that far.
• 14:54 <novalis_dt> I actually am in the middle of a different bit of debugging, so I haven't looked
• 14:54 <mattwigway> Ok, cool.
• 14:54 <mattwigway> The changes do seem to have been applied to master, see e.g. line 49 of config.js
• 14:56 <novalis_dt> Yep
• 14:56 <mattwigway> I see what's going on... in my master I did git pull upstream master, then git checkout dynamic-bookmarking, then git merge master.
• 14:56 <mattwigway> I should have only done one merge.
• 14:56 <mattwigway> The results are the same though; I should have used a remote-tracking branch though
• 14:57 <novalis_dt> Well, if the results are good, then all is well