Talk:Waychains TIGER fixup

From OpenStreetMap Wiki
Jump to: navigation, search

Discuss TIGER fixup/250 cities/waychains here:


Route relations

Many of the interstates have been incorporated into route relations. I've seen conflicting recommendations whether it is necessary to add ref tags to the ways when the proper ref tags are on the relation. I know that for the route relations that I have worked on , I have not look at the ref numbers. Jumbanho 19:02, 13 January 2010 (UTC)

I guess I'm echoing Jumbanho... it seems that the Interstate Highways Relations project takes care of what this idea seems to incorporate. Existing route relation analysis tools (example: click the a or r in this: Relation 282374 (XML, Potlatch2, iD, JOSM, history, analyze, manage, gpx)) will check for continuous route, by roles if used. The ref tag is in the relation, so it doesn't (shouldn't) really matter what ref is on the way itself. I know most map tools other than editors don't even support relations yet, but it feels like we could spend a lot of time fixing refs on the ways just to have them become obsolete before we even get close to done. I'd rather go through and delete them once they aren't needed. For now, having "I5", "I 5", and "I-5" is something I can deal with (though honestly, I've never USED OpenStreetMap data other than for a week or two of Monopoly City Streets, which is what led me here.)
As for adding ref tags when adding ways to relations, I personally add or edit the ref tag (using a space: "I 5"), and verify the name tag is also there, "I-5" if TIGER or a previous user hasn't given it a real name, when I add relations to the way. But traversing a route for the sole purpose of adjusting ref tags seems a bit much. Using ; to enter multiple refs is pretty much required when multiple highways overlap. US Highways often merge with Interstates for long stretches, and even multiple Interstates merge for generally short lengths. Using relations in place of refs takes care of the multiple ref issue nicely (or will, once the refs on relaltions are supported.)
That said, I concede that perhaps I am missing something, or there is some benefit of fixing refs that won't become obsolete with more widespread relation support. I also just verified that the 'a' link above will not catch wrong-way oneways, so that is another benefit... a very big one since wrong ways wreak havoc on routing.
This project WILL find motorway_links improperly tagged as motorways; the first two I looked at from the report were just that, but the next one was a single way with ref "I 5" in the middle of who knows how many ways tagged as "I5" (and oddly, the opposite direction used "I 5"). Perhaps the waychains tool can be adjusted to ignore spaces and "-" in the ref tags ... we'd lose the (maybe unnecessary) ref cleanup aspect, but the improper highway tags and wrongway oneways would get fixed faster. I have been itching for something to do in the 250 cities since the duplicate nodes fizzled out, so I'm going with it, I'm just going to ignore the ref tags if that's the only things wrong with a way.
Wow, that ended up much longer than I intended. Sorry for the essay.
Granack 08:12, 16 January 2010 (UTC)
You probably shouldn't add names like "I-5", since it isn't really a highway name, and is redundant to the ref. --NE2 22:29, 24 January 2010 (UTC)
A lot of people suggested looking at relations back when we were first looking at routing fixup for the 250 cities stuff. Relations are nice but it's kind of like saying the data sucks so lets add a new layer over the top of it. We wanted to encourage people to fix the sucky data.
I guess I have a deep-seated dislike of relations in general. They were always a detraction from the simplicity of the OSM data model, and in general there's too many people fannying around with relations because they can, and not stopping to think whether they should (also not getting on with any real data improvements) The fact that Frederik, who created the relations feature, later had to tell people off for going way overboard is evidence of this.
But OK ...end of rant
It's true the relations could provide a better way of partitioning off sets of ways which we expect to be arranged in a chain. As I've said, the waychains tool is unfortunately mostly finding problems with ref values rather than actual chaining problems. I believe there are already tools for checking sequential arrangement within relations e.g. "Interstate relations check" described below. Maybe to bring the ideas together, we need a tool for checking ref tags on the relations against ref tags on the ways.
Maybe I'll look into it at some point, but from an immediate practical point of view, my 'simplify' osmosis plugin (Osmosis/Detailed_Usage#--simplify) which I use to create us-motorways-simplified.osm, currently ditches all relations data, so I'd have to work on that before I'd even have the data in my waychains ruby script.
-- Harry Wood 15:15, 26 March 2010 (UTC)

east/west role in relations

The tagging method I found most useful in Florida was to go through each area in Potlatch filling in all the refs, then run a XAPI query like http://www.informationfreeway.org/api/0.6/*%5bref=*%5d%5bbbox=-83,27.5,-80,29.5%5d and use JOSM's search to find everything with a ref tag containing "I 4", for example. This made more complicated areas (mainly non-Interstates) easier to deal with. The only problem is that the role (east/west) doesn't get added. Is there an automated tool that will, after analyzing the relation and finding two waychains, allow you to add the appropriate role to each? --NE2 22:29, 24 January 2010 (UTC)

One chain but really two

[1] is actually two chains; they just connect at one end (because the median ends where the motorway ends). --NE2 22:33, 24 January 2010 (UTC)

This doesn't work with overlaps or multiple routes with the same number

For instance, there are several different routes numbered I-275, and the one in Florida overlaps US 19 in the middle. --NE2 22:38, 24 January 2010 (UTC)

Other useful tool : Interstate relations check

Hello,

I maintain the pages User:Nakor/Interstate relations check and User:Nakor/US relations check which scans relations from Interstate_Highways_Relations and United States Numbered Highway Relations and makes sure that there is one segment per role and that that segment has all its oneways in the sane direction. That can be a usefull tool for checking correct "waychains". Nakor 18:32, 26 February 2010 (UTC)

Have I told you you're awesome? :) It might be useful to separate the name and ref tags (so when a common format is finally decided, it's easy to identify ones that need changing) and add the system tag (for the special routes near the bottom). I'm not sure how the "routable" feature works - in particular, how would I identify which oneways are pointing the wrong direction, and what's the difference between no and missing? Does blank mean yes? It would also be useful to have links to your tool in that table. --NE2 20:13, 26 February 2010 (UTC)
Thanks for the feedback. An empty cell means you can "drive" a role from one way to the other (i.e. on that role all oneways are in the same direction), no means at some point if you "drive" you come across a oneway in the wrong direction. The algorithm assumes motorway and motorway_link to be oneway unless explicitly specified oneway=no. Missing means there is a motorway without a oneway. Nakor 03:10, 27 February 2010 (UTC)
Ah. Is there any chance of narrowing down the "bad" ways? Right now the tool just gives you two that are about half the route apart. --NE2 03:44, 27 February 2010 (UTC)
You mean for the directions? I would have to come with a way to detect what is the right direction in the first place. A simple test would be comparing the node coordinates to the direction of the way. Nakor 13:39, 27 February 2010 (UTC)
Good point. That wouldn't work on US 321, but I don't think anything will :) Another possibility is to look which way the majority of oneways are pointing and output any that are the other direction. --NE2 00:27, 28 February 2010 (UTC)