Re: [OSM-talk] Mulltipolygons and Mapnik
On Sat, May 31, 2008 at 9:32 PM, Jon Burgess [EMAIL PROTECTED] wrote: I believe the sites are both using very similar code (unless Kleptog has some private changes for the NL site). I think the difference may be due to the NL site using the --slim mode. In this case this is probably allowing it to keep the untagged inner way that would have been discarded in non-slim mode. Correct, the --slim mode makes all the difference. The basic problem is that to manage such relations it needs to remember the relevent ways for when it processes the relations. When processing a whole planet dump it obviously can't rememeber all the ways in memory, which is why non-slim mode doesn't get it. Slim mode remembers everything and so it always works. It would be possible to get non-slim mode to work also by processing relations first. As usual, needs a coder. Have a nice day, -- Martijn van Oosterhout [EMAIL PROTECTED] http://svana.org/kleptog/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
On Saturday 31 May 2008 19:07:31 Dermot McNally wrote: Multipolygons with untagged inner elements (per wiki): Don't render on Mapnik Multipolygons with inner elements tagged as the outer ones: Do render on Mapnik At this point, we _could_ define the current behaviour as a standard, update the wiki and change whatever data isn't compliant. *or* we can see if the Mapnik brains are willing/able to support multipolygons according to the current wiki instructions. Here in Utrecht I have drawn a building with a courtyard. The inner polygon does not have any tags. * The hole shows OK in [EMAIL PROTECTED]/Osmarender http://tile.openstreetmap.nl/?zoom=17lat=52.08396lon=5.10485layers=00BF * The hole does not show in the osm2pgsql/Mapnik generated tiles from openstreetmap.org. http://tile.openstreetmap.nl/?zoom=17lat=52.08396lon=5.10485layers=0B0F * The hole shows OK in the osm2pgsql/Mapnik generated tiles from openstreetmap.nl. http://tile.openstreetmap.nl/?zoom=17lat=52.08396lon=5.10485layers=B00F I think openstreetmap.nl uses a newer version of osm2pgsql than the main site. If that is true than it is just a question of time till multipolygons defined as per wiki description will show OK on the main site too. A multipolygon with the same tags on inner and outer shows OK at all three: http://tile.openstreetmap.nl/?zoom=15lat=52.05845lon=5.16772layers=00BF http://tile.openstreetmap.nl/?zoom=15lat=52.05845lon=5.16772layers=0B0F http://tile.openstreetmap.nl/?zoom=15lat=52.05845lon=5.16772layers=B00F -- m.v.g., Cartinus ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
I forgot to mention: A hole filled with something else IOW different tags on the outer and inner polygons only works in [EMAIL PROTECTED]/Osmarender http://tile.openstreetmap.nl/?zoom=17lat=52.06243lon=5.10283layers=00BF http://tile.openstreetmap.nl/?zoom=17lat=52.06243lon=5.10283layers=0B0F http://tile.openstreetmap.nl/?zoom=17lat=52.06243lon=5.10283layers=B00F (You can just use the layer switcher on the page in stead of opening each link separately.) -- m.v.g., Cartinus ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
2008/5/31 Chris Hill [EMAIL PROTECTED]: I've already changed the wiki to match this situation. This is the way it was, so we are just back to the same position before it was changed (in my view) erroneously. Hmm. Could you have missed a bit? Tags on inner ways describe the hole. If it's just a hole, leave it untagged. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
Dermot McNally wrote: 2008/5/31 Chris Hill [EMAIL PROTECTED]: I've already changed the wiki to match this situation. This is the way it was, so we are just back to the same position before it was changed (in my view) erroneously. Hmm. Could you have missed a bit? Tags on inner ways describe the hole. If it's just a hole, leave it untagged. Dermot Er, yeah, so I had another go with my eyes open this time. Chris ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
On Sat, 2008-05-31 at 20:08 +0200, Cartinus wrote: I forgot to mention: A hole filled with something else IOW different tags on the outer and inner polygons only works in [EMAIL PROTECTED]/Osmarender http://tile.openstreetmap.nl/?zoom=17lat=52.06243lon=5.10283layers=00BF http://tile.openstreetmap.nl/?zoom=17lat=52.06243lon=5.10283layers=0B0F http://tile.openstreetmap.nl/?zoom=17lat=52.06243lon=5.10283layers=B00F I'm not certain which polygons you are talking about here. If you are talking about the ones to the NW, then those look like they are marked with area=yes. Support for area=yes was only recently added to the osm2pgsql code and the rendering styles have not been implemented yet. Mapnik will render inner areas with different tags. It uses the area of the polygons to figure out the render ordering. Smaller areas always render on top of larger areas of the same layer. If the layer tags differ then this will take precedence. Jon ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
On Sat, 2008-05-31 at 19:54 +0200, Cartinus wrote: Here in Utrecht I have drawn a building with a courtyard. The inner polygon does not have any tags. * The hole shows OK in [EMAIL PROTECTED]/Osmarender http://tile.openstreetmap.nl/?zoom=17lat=52.08396lon=5.10485layers=00BF * The hole does not show in the osm2pgsql/Mapnik generated tiles from openstreetmap.org. http://tile.openstreetmap.nl/?zoom=17lat=52.08396lon=5.10485layers=0B0F * The hole shows OK in the osm2pgsql/Mapnik generated tiles from openstreetmap.nl. http://tile.openstreetmap.nl/?zoom=17lat=52.08396lon=5.10485layers=B00F I think openstreetmap.nl uses a newer version of osm2pgsql than the main site. If that is true than it is just a question of time till multipolygons defined as per wiki description will show OK on the main site too. I believe the sites are both using very similar code (unless Kleptog has some private changes for the NL site). I think the difference may be due to the NL site using the --slim mode. In this case this is probably allowing it to keep the untagged inner way that would have been discarded in non-slim mode. A multipolygon with the same tags on inner and outer shows OK at all three: http://tile.openstreetmap.nl/?zoom=15lat=52.05845lon=5.16772layers=00BF http://tile.openstreetmap.nl/?zoom=15lat=52.05845lon=5.16772layers=0B0F http://tile.openstreetmap.nl/?zoom=15lat=52.05845lon=5.16772layers=B00F Right, this is normally the best approach. Jon ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
On May 31, 2008, at 19:27, Chris Hill wrote: I've already changed the wiki to match this situation. This is the way it was, so we are just back to the same position before it was changed (in my view) erroneously. I also thing this makes logical sense too. The outer marks the edge of the water and so does the inner. There is a short discuusion on the wiki talk page. As far as I can tell, the main argument for this reversion is that's what the renderers want. I wonder why those who claim the change was wrong didn't speak up in the lengthy discussions in March? (Starting at http://lists.openstreetmap.org/pipermail/talk/2008-March/023876.html and http://lists.openstreetmap.org/pipermail/dev/2008-March/009345.html , and I do remember one opposing opinion which I can't find.) Cheers Robert ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
On Sat, May 24, 2008 at 1:01 AM, Dermot McNally [EMAIL PROTECTED] wrote: It's worth a try, and we'll know in a week, but I'm not so sure this will cure the problem. Otherwise, how do you explain the island that already was tagged the same but still didn't render on the mapnik layer? I'm also about 90% sure that I've seen untagged islands with an inner role in the relations showing up OK in the past. Also, when reporting these kinds of issues it's very helpful to provide the ID of the relation, because then it only takes seconds to see if its tagged properly. Don't discount the possibility of bugs in renderers Have a nice day, -- Martijn van Oosterhout [EMAIL PROTECTED] http://svana.org/kleptog/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
I noticed this also. Without investigating it any further, my hunch was that only the first island is rendered. This lake has two islands, but only one shows up in Mapnik: http://geo.topf.org/comparison/index.html?mt0=tahmt1=mapnikx=1959y=1305z=12 Ludwig 2008/5/23 Dermot McNally [EMAIL PROTECTED]: Folks, I've just noticed that the Mapnik layer isn't showing islands correctly in some lakes where I had modelled them as holes via multipolygon relations. Example: http://geo.topf.org/comparison/index.html?mt0=tahmt1=mapnikx=1959y=1305z=12 I thought that Mapnik was able to handle this. Am I mistaken, or might there be recent breakage? Thanks, Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
Dermot McNally dermotm at gmail.com writes: Unfortunately, I don't think that's the problem: http://geo.topf.org/comparison/index.html?mt0=tahmt1=mapniklon=-8.1250762lat=53.9649733z=13 This example has only one island, and that doesn't show either. Dermot Hi, Have you tried adding the lake tag (natural=water)also for the holes? I know it works with buildings but I do not know if it is necessary. -Jukka Rahkonen- ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
The holes, at least for my example, are land (so natural=land). 2008/5/23 Jukka Rahkonen [EMAIL PROTECTED]: Dermot McNally dermotm at gmail.com writes: Unfortunately, I don't think that's the problem: http://geo.topf.org/comparison/index.html?mt0=tahmt1=mapniklon=-8.1250762lat=53.9649733z=13 This example has only one island, and that doesn't show either. Dermot Hi, Have you tried adding the lake tag (natural=water)also for the holes? I know it works with buildings but I do not know if it is necessary. -Jukka Rahkonen- ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
2008/5/23 Jukka Rahkonen [EMAIL PROTECTED]: Have you tried adding the lake tag (natural=water)also for the holes? I know it works with buildings but I do not know if it is necessary. It won't help here either - my second example already has the holes tagged. Good thought, though. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
2008/5/23 Ludwig [EMAIL PROTECTED]: The holes, at least for my example, are land (so natural=land). It's not what's causing the problem, but to me that's just incorrect. Valid tagging to me could be: Untagged: It's just a hole in whatever its parent polygon is. Tagged water: It's a hole in something that is water, so it identifies itself as a water hole To tag it as land _and_ make it part of a multipolygon seems to be a double negative. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
I was just trying to follow the rules from http://wiki.openstreetmap.org/index.php/Map_Features: - Land that exists within another area, such as a lake. (i.e an island). Keep water on the right and land on the left side in relation to sequence of nodes in the Way. Layering may also be required. See Relations/Multipolygonhttp://wiki.openstreetmap.org/index.php/Relations/Multipolygonfor islands in lakes I agree that the land-tag itself is redundant, so maybe the instructions should change. Ludwig 2008/5/23 Dermot McNally [EMAIL PROTECTED]: 2008/5/23 Ludwig [EMAIL PROTECTED]: The holes, at least for my example, are land (so natural=land). It's not what's causing the problem, but to me that's just incorrect. Valid tagging to me could be: Untagged: It's just a hole in whatever its parent polygon is. Tagged water: It's a hole in something that is water, so it identifies itself as a water hole To tag it as land _and_ make it part of a multipolygon seems to be a double negative. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
On Friday 23 May 2008 21:18:40 Ludwig wrote: - Land that exists within another area, such as a lake. (i.e an island). Keep water on the right and land on the left side in relation to sequence of nodes in the Way. Layering may also be required. See Relations/Multipolygonhttp://wiki.openstreetmap.org/index.php/Relations/M ultipolygonfor islands in lakes You either hack around with natural=land and layers (which is bad practice IMHO) or you use multipolygons. Trying to do both at once is just weird (and not really what that text is trying to say.) - - - - - - - - After the last round of discussions about how to tag multipolygons I left some tagged in different ways around my home city. I just checked the current state of their renders. 1) outer tagged with feature X inner tagged with feature X - Works with osm2pgsql used on the main site (i.e. the mapnik layer) - Works with Osmarender 2) outer tagged with feature X inner not tagged at all - Does not work with osm2pgsql used on the main site (i.e. the mapnik layer) - Works with Osmarender 3) outer tagged with feature X inner tagged with feature Y - Does not work with osm2pgsql used on the main site (i.e. the mapnik layer) - Does not work with Osmarender So your first example doesn't show the islands because they are not having tags. I don't know why the island in the second example doesn't work. The relation and both the inner and outer polygon look OK. -- m.v.g., Cartinus ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Mulltipolygons and Mapnik
2008/5/23 Chris Hill [EMAIL PROTECTED]: I've worked with islands before. It seems that the key points are: 1. make sure that the same tag is used on the inner and outer, in this case natural=water 2. Make sure that the outer is clockwise and inner(s) are anticlockwise I took the liberty of changing the islands to match their outer so I expect it to render. It's worth a try, and we'll know in a week, but I'm not so sure this will cure the problem. Otherwise, how do you explain the island that already was tagged the same but still didn't render on the mapnik layer? I'm also about 90% sure that I've seen untagged islands with an inner role in the relations showing up OK in the past. Regardless, we need to decide: * What version of this tagging we think _should_ be considered correct. * Ensure both main renderers will accept the valid form. * Adapt the wiki as required to reflect reality Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk