[mkgmap-dev] oneway reverse patch
From Gerd Petermann gpetermann_muenchen at hotmail.com on Fri Apr 25 07:09:08 BST 2014
Hi Felix, the reversing has a reason. It is needed for routing. Some days ago WanMil asked in a pm if it isn't possible to skip the reversing and change the way how the data is written to the img file. I answered that this is complicated, but maybe I was wrong. The img file has the information that a way is a oneway. We save both the forward arc and the reverse arc between two routing nodes. If I got that right we just have to make sure that we mark the correct arc as forward/reverse. I'll try to create a small test style that creates multiple ways with conflicting oneway tags and see what happens. There are so many places were this matters. If I can't find a good solution, you may send a sample style and test data next week. Gerd Date: Fri, 25 Apr 2014 13:43:51 +0800 From: extremecarver at gmail.com To: mkgmap-dev at lists.mkgmap.org.uk Subject: Re: [mkgmap-dev] oneway reverse patch On 25.04.2014 12:26, Gerd Petermann wrote: Hi Felix, okay, I think that this explains your problems. The current code in mkgmap simply doesn't expect lines based on the same OSM way to go in different directions. I am pretty sure that you get unpredictable results when doing that. In most places I don't see a problem to support that, but I fear that the code for restriction relations will be a problem. Thinking about it... Well I'm still using --no-turn-restrictions anyhow.... However I would see a second possibility. Create another file in the style-file called direction-dependent-lines where one lists all types (0x????? 0x?? 0x???? .... ) that mkmgap may not reverse because the outcome is dependent on the direction of the way due to the layout chosen in the typfile. Or add all tags - but that would be less reliable. Anyhow - the turn-restrictions are still a mystery to me - both in mkgmap as well as in OSM. I'm pretty sure 50% of the turn restrictions in OSM will not be followed by cyclists (either they actually don't have to, or there are quicker possibilities (e.g. using the footway crossing) or simply knowing that noone cares what cyclists do in that case... Especially I would say no cyclists respects right turn restriction in drive-right countries (and some vice-versa for drive-left). Therefore for now I still prefer to completly disrespect them in the map. Gerd Date: Fri, 25 Apr 2014 09:30:54 +0800 From: extremecarver at gmail.com To: mkgmap-dev at lists.mkgmap.org.uk Subject: Re: [mkgmap-dev] oneway reverse patch Oh forgot to mention. I also sometimes set oneway yes. Then continue and oneway -1 then contune. All non routable. I don't use continue with actions so I depend on strict work down of lines file as routable non oneway road copies follow On Apr 25, 2014 9:25 AM, "Felix Hartmann" <extremecarver at gmail.com> wrote: Oh and I of course also add multiple routable lines sometimes. Up to 6 per way in osm as 7 produces crashes on garmins side. Often one routable line has direction but no oneway while the copy is higher importance for routing and oneway (and might be reversed). I need to reverse sometimes to save on the number of lines in type file too - other times its about routing! On Apr 25, 2014 9:19 AM, "Felix Hartmann" <extremecarver at gmail.com> wrote: Yes I both add opposite oneway on the same road as well as only one oneway or no oneway. It is highly important that the order is followed strictly and continue vs continue with actions is strictly enacted in order. All reversing should happen at the time its placed in the style. !! Sometimes I reverse a way 3 times during processing. All I can say right now its a bit of a mess On Apr 25, 2014 2:24 AM, "Gerd Petermann" <gpetermann_muenchen at hotmail.com> wrote: Hi Felix, if your style interprets a tag like a oneway=yes you should add that tag. This might prevent problems caused by the RoadMerger, which might reverse lines which are not oneways. If you find or set oneway=-1, the current implementation will reverse the way. If you add multiple routable lines for one OSM way, one with, one without the oneway tag, you will see unpredictable directions if such a way is modified by the WrongAngleFixer and the type is direction dependent. The WrongAngleFixer assumes that the points in all ways with the same OSM id are equal, so it optimizes one way and copies the points to the others. This will fail if they have different oneway attributes. If you think this could be the cause of the problem, I should be able to provide a fix. Gerd Date: Fri, 25 Apr 2014 01:00:18 +0800 From: extremecarver at gmail.com To: mkgmap-dev at lists.mkgmap.org.uk Subject: Re: [mkgmap-dev] oneway reverse patch Ups - but forgot to say. I think in 99% of all cases - cycleway:left and cycleway:right are used on streets which feature oneway=yes... less often oneway=-1 and even less often no oneway at all. Streets with oneway=yes are fine. I'm talking about no oneway tag from OSM data at all, or oneway=-1 set in style (but no oneway from OSM data) or no oneway at all. Only on those there are problems - so you're not likely to notice them I think.... On 25 April 2014 00:39, Minko <ligfietser at online.nl> wrote: Yes I render cycleway:left and cycleway:right too. And as you say, they are always on the wrong side on the GPS. Interesting to know that Garmin uses asymmetric lines independent of the road direction. If we only could find out how they do this... Felix wrote > I don't think that Minkos style shows cycleways on the left/right side > of a road - am I wrong? - Anyhow they would usually have a oneway tag > already in OSM data. > Also definitely no uphill/downhill arrows - which nearly never > actually have a oneway tag (and only sometimes I add one during > processing). _______________________________________________ mkgmap-dev mailing list mkgmap-dev at lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev _______________________________________________ mkgmap-dev mailing list mkgmap-dev at lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev _______________________________________________ mkgmap-dev mailing list mkgmap-dev at lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev _______________________________________________ mkgmap-dev mailing list mkgmap-dev at lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev _______________________________________________ mkgmap-dev mailing list mkgmap-dev at lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev _______________________________________________ mkgmap-dev mailing list mkgmap-dev at lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.mkgmap.org.uk/pipermail/mkgmap-dev/attachments/20140425/560b57bb/attachment-0001.html>
- Previous message: [mkgmap-dev] oneway reverse patch
- Next message: [mkgmap-dev] Commit: r3206: fix possible problem with --link-pois-to-ways option
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the mkgmap-dev mailing list