Re: [OSM-talk] Mapnik handling of highways that are also landuse...
On Thu, Aug 28, 2008 at 8:47 PM, Jon Burgess <[EMAIL PROTECTED]> wrote: >> FWIW, I've been planning to implement just that at some point, just >> havn't gotten around to it yet. The biggest problem is that the style >> file needs to contain information like "tag X applies only to >> nodes/ways/areas" and this information simply does not exist. >> Collecting this information is a project in itself. > > I did something similar to put the features into multiple tables but I > didn't do any tag filtering. The tag filtering was not important at one > time because there were no common rendering styles shared between linear > and area features. Now some combinations like > highway=pedestrian,area=yes make things things more complicated. In case anybody is wondering, if you know tag X does not apply to nodes ever, osm2pgsql supports that now. For the distinction between lines and polygons there is an attempt but it doesn't quite work right. What I want to do extend it to is that highway=traffic_signals is only for nodes and highway=pedestrian only for ways. > My personal opinion is that features like a road around an area should > use different ways. When mapping features myself I don't even share the > nodes. I find it much clearer to leave a small gap between the area and > surrounding ways. For stuff like name which can be applied to anything it will always get copied to all objects. Being able to have multiple features out of a single OSM object isn't ever going to be easy. But we can do things much better than now... Have a nice day, -- Martijn van Oosterhout <[EMAIL PROTECTED]> http://svana.org/kleptog/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
On Thu, 2008-08-28 at 16:57 +0200, Martijn van Oosterhout wrote: > On Thu, Aug 28, 2008 at 1:14 PM, Dave Stubbs > <[EMAIL PROTECTED]> wrote: > > Well, technically, there are ways of handling this case (the way > would > > need duplicating on import, with some clever tag stripping). But > they > > aren't pretty, or I think, desirable. > > FWIW, I've been planning to implement just that at some point, just > havn't gotten around to it yet. The biggest problem is that the style > file needs to contain information like "tag X applies only to > nodes/ways/areas" and this information simply does not exist. > Collecting this information is a project in itself. I did something similar to put the features into multiple tables but I didn't do any tag filtering. The tag filtering was not important at one time because there were no common rendering styles shared between linear and area features. Now some combinations like highway=pedestrian,area=yes make things things more complicated. I think you'll get stuck with many tags not falling into any of these categories, some obvious ones... * name * layer * access * ref My personal opinion is that features like a road around an area should use different ways. When mapping features myself I don't even share the nodes. I find it much clearer to leave a small gap between the area and surrounding ways. Jon ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
On Thu, Aug 28, 2008 at 3:57 PM, Martijn van Oosterhout <[EMAIL PROTECTED]> wrote: > On Thu, Aug 28, 2008 at 1:14 PM, Dave Stubbs <[EMAIL PROTECTED]> wrote: >> Well, technically, there are ways of handling this case (the way would >> need duplicating on import, with some clever tag stripping). But they >> aren't pretty, or I think, desirable. > > FWIW, I've been planning to implement just that at some point, just > havn't gotten around to it yet. The biggest problem is that the style > file needs to contain information like "tag X applies only to > nodes/ways/areas" and this information simply does not exist. > Collecting this information is a project in itself. > I think this might be one of those Jurassic Park moments ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
On Thu, Aug 28, 2008 at 11:38 AM, Dermot McNally <[EMAIL PROTECTED]> wrote: > Reuse of the way (or > creation of an identical one using the same nodes) ... are conceptually two completely different things. Having one object (the way) representing two different real-world objects is different from two objects (two ways) representing two real-world objects. The former is a recipe for disaster. Cheers, Andy ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
On Thu, Aug 28, 2008 at 1:14 PM, Dave Stubbs <[EMAIL PROTECTED]> wrote: > Well, technically, there are ways of handling this case (the way would > need duplicating on import, with some clever tag stripping). But they > aren't pretty, or I think, desirable. FWIW, I've been planning to implement just that at some point, just havn't gotten around to it yet. The biggest problem is that the style file needs to contain information like "tag X applies only to nodes/ways/areas" and this information simply does not exist. Collecting this information is a project in itself. Have a nice day, -- Martijn van Oosterhout <[EMAIL PROTECTED]> http://svana.org/kleptog/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
On Thu, Aug 28, 2008 at 11:06 AM, Dermot McNally <[EMAIL PROTECTED]> wrote: > 2008/8/28 Steve Chilton <[EMAIL PROTECTED]>: > >> There are two reasons why your "tag combination ought to be able to >> render correctly as is" statement is not valid. > > This is always a risk :) > >> Firstly - as Thomas pointed out - mapnik likes one feature per way. >> Secondly - landuse=grass is not rendered at the moment (because of the >> vast disagreement on the whole landuse/natural/grass thing). > > A further complexity, certainly. The amusing thing is, I only started > using landuse=grass very recently, after deciding that > recreation_ground was probably not a good tag for what we're > describing here. > >> What I am guessing you have on the ground is a residential square with a >> road round the outside and a grassy bit in the middle. > > Correct. > >> There is no way I >> know of for mapnik to interpret this correctly from a double-tagged >> single way like this. It would have to be able to draw a line, a road >> fill, a second line, and then the grass fill. >> Where these occur in areas I have seen (many grass filled squares in C >> London) two parallel ways have been used - and yes I know that people >> are using leisure=park "incorrectly" in order to get them rendered, but >> probably excusable in London as they are often formal park/recreation >> areas. > > Well, a technical limitation is just that - if it can't be done, we > need to live with it. The end result is certainly unintuitive, though, > since the only area tag named (the landuse) isn't the style that gets > applied to the area. > Well, technically, there are ways of handling this case (the way would need duplicating on import, with some clever tag stripping). But they aren't pretty, or I think, desirable. The weirdness is caused by the fact that you have a way which is both a line and a polygon at the same time, as well as being a road and grass at the same time. This kind of duality might be OK for photons, but I'm not sure it makes much sense for data modelling. And if OSM ever gets a native area type like people keep suggesting, then it probably just wouldn't be allowed. There's an easy fix though: just make a 2nd way using the same nodes.. have one be the road, and one the grass. Dave ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse
> -Original Message- > Date: Thu, 28 Aug 2008 00:59:12 +0100 > From: "Thomas Wood" <[EMAIL PROTECTED]> > Subject: Re: [OSM-talk] Mapnik handling of highways that are also > landuse... > To: "Dermot McNally" <[EMAIL PROTECTED]> > Cc: osm > Message-ID: > <[EMAIL PROTECTED]> > Content-Type: text/plain; charset=ISO-8859-1 > > On Thu, Aug 28, 2008 at 12:49 AM, Dermot McNally <[EMAIL PROTECTED]> > wrote: > > Folks - with reference to this: > > > > http://www.openstreetmap.org/?lat=53.72339&lon=- > 6.34273&zoom=17&layers=B00FTF > > > > ...which is a section of the Mapnik render of the outcome of the very > > successful Drogheda Mapping Party in Ireland. Towards the centre of > > the map, you'll see what is represented as an oval area of residential > > highway. I can almost see why, but it struck me that it represents > > unwanted behaviour that could possibly be fixed. > > > > What we have here is a closed way of type highway=residential. > > Importantly, it isn't tagged as an area. It _is_ tagged (the same way) > > as landuse=grass. So without understanding the internals of Mapnik, > > it's as though the landuse, which applies at area-level, infects the > > highway tag and causes it to be considered as an area too. > > > Correct > > > > Clearly, I could simply draw a second way through the same nodes, and > > there are plenty of heated discussions over which approach is the > > saner. But it feels as though this tag combination ought to be able to > > render correctly as is. > > Mapnik still likes the one way per feature way of doing things, (as a > matter of fact, so do I, seems more logical that the two can be > separated if required later) > Personally I prefer to recommend that you definer the area of grass using a separate way that uses the same nodes as the residential road, but which is certainly a separate way from the road. You may prefer to define it as a separate way using separate nodes as this can make editing easier in the short term, however Richard explained yesterday on talk how to use '/' to select from the different ways associated with the same node which I will investigate. Using the approach you have tried is definitely to be discouraged imho, and mixes up two different things into one way. As a longer term discussion I am interested in morphing the 'multi-polygon' relation into a 'polygon' relation so it can be used as an alternative ways of defining areas. The relation would need to allow a number of linear features to form the boundary of the area. The relation would then hold the tags that are associated with the area (in this case 'landuse=grass'). The relation could also be able to refer to zero or more 'inner' areas which can be defined in a similar way to define 'holes' in polygons. This approach allows a single 'edge' to be part of a number of areas (I gave the example of the edge of a park also being the boundary for the borough in a previous post). Currently the approach of using boundary:left=Ipswich for part of the boundary is not compatible with have a single way defining the area of the park. I am also advocating that we dump the current boundary left: and right: tagging in favour of using the 'boundary' relation for boundaries. I might come up with a technical demonstrator for this in the near future so explore how it might work in practice. There is more discussion on polygons and relations here: http://wiki.openstreetmap.org/index.php/Talk:Relation:multipolygon And the boundary relation here: http://wiki.openstreetmap.org/index.php/Relations/Proposed/Boundaries If this approach was used we would be able to run with both coding systems in the short term and possibly then deprecate ways being used for areas and boundaries in the longer term. Any other thoughts? Am I wasting my time on this idea, or do others see value in it? Is so would it be useful to produce some trial rendering or would someone like to make osmarender or Mapnik handle it? Regards, Peter(Ito) ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
> Steve Chilton wrote: > > Dermot > > > > There are two reasons why your "tag combination ought to be able to > > render correctly as is" statement is not valid. > > Firstly - as Thomas pointed out - mapnik likes one feature per way. > > Secondly - landuse=grass is not rendered at the moment (because of the > > vast disagreement on the whole landuse/natural/grass thing). > > What I am guessing you have on the ground is a residential square with a > > road round the outside and a grassy bit in the middle. There is no way I > > know of for mapnik to interpret this correctly from a double-tagged > > single way like this. It would have to be able to draw a line, a road > > fill, a second line, and then the grass fill. > > Dermot, > > this sounds a lot like you're using the same way to form the centre of > the road, and the boundary of the grass. wouldn't it be easier to use a > separate way to represent each? > > i can't imagine the grass extends to the middle of the road > > or am i getting the wrong end of the stick? Well, I'd say that the way does not represent the middle of the road, but the entire road (otherwise you would have to leave streets unconnected at crossings, because the incoming street doesn't extend to the middle of the other street either). But the way it is currently tagged, I would interpret it as an object that is both a road and a grass area at the same time. I would strongly recommend to stick to the "one feature == one way" paradigm. Regards, Marc -- Pt! Schon das coole Video vom GMX MultiMessenger gesehen? Der Eine für Alle: http://www.gmx.net/de/go/messenger03 ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
2008/8/28 robin paulson <[EMAIL PROTECTED]>: > this sounds a lot like you're using the same way to form the centre of > the road, and the boundary of the grass. wouldn't it be easier to use a > separate way to represent each? That is what I'm doing. I don't do this in all cases, but some do lend themselves rather well to it. In my sample case, I have a closed way representing a highway. So no, it wouldn't be "easier" to create a brand new way, either reusing the nodes or staying just a little within them. That's extra work. We can discuss whether your approach (which is one I sometimes use) is more correct, but it certainly isn't easier. > i can't imagine the grass extends to the middle of the road Clearly not - but I'd reiterate what Thomas said. Reuse of the way (or creation of an identical one using the same nodes) is IMHO topologically valid given the intended uses of our data. You would only insist on modelling such green areas, say, 3m away from the road centre if you were also in the business of mapping road boundary instead of centre line. Mapping a centre line is an approximation to reality. A simplified notion of abutting areas that assumes them to extend to that centre line is entirely consistent with that approximation. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
2008/8/28 robin paulson <[EMAIL PROTECTED]>: > Steve Chilton wrote: >> Dermot >> >> There are two reasons why your "tag combination ought to be able to >> render correctly as is" statement is not valid. >> Firstly - as Thomas pointed out - mapnik likes one feature per way. >> Secondly - landuse=grass is not rendered at the moment (because of the >> vast disagreement on the whole landuse/natural/grass thing). >> What I am guessing you have on the ground is a residential square with a >> road round the outside and a grassy bit in the middle. There is no way I >> know of for mapnik to interpret this correctly from a double-tagged >> single way like this. It would have to be able to draw a line, a road >> fill, a second line, and then the grass fill. > > Dermot, > > this sounds a lot like you're using the same way to form the centre of > the road, and the boundary of the grass. wouldn't it be easier to use a > separate way to represent each? > > i can't imagine the grass extends to the middle of the road > > or am i getting the wrong end of the stick? > It's the right end of the stick, since I do it myself. It is my personal view that ways in OSM represent the whole feature, and although they are usually placed along the centreline of that feature, the way does not represent merely the centre line of the feature. Until OSM moves to a very high accuracy method of using areas to define road surfaces, etc, I'll continue using overlapping ways to represent landuse areas contained by roads. -- Regards, Thomas Wood (Edgemaster) ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
Steve Chilton wrote: > Dermot > > There are two reasons why your "tag combination ought to be able to > render correctly as is" statement is not valid. > Firstly - as Thomas pointed out - mapnik likes one feature per way. > Secondly - landuse=grass is not rendered at the moment (because of the > vast disagreement on the whole landuse/natural/grass thing). > What I am guessing you have on the ground is a residential square with a > road round the outside and a grassy bit in the middle. There is no way I > know of for mapnik to interpret this correctly from a double-tagged > single way like this. It would have to be able to draw a line, a road > fill, a second line, and then the grass fill. Dermot, this sounds a lot like you're using the same way to form the centre of the road, and the boundary of the grass. wouldn't it be easier to use a separate way to represent each? i can't imagine the grass extends to the middle of the road or am i getting the wrong end of the stick? ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
2008/8/28 Steve Chilton <[EMAIL PROTECTED]>: > There are two reasons why your "tag combination ought to be able to > render correctly as is" statement is not valid. This is always a risk :) > Firstly - as Thomas pointed out - mapnik likes one feature per way. > Secondly - landuse=grass is not rendered at the moment (because of the > vast disagreement on the whole landuse/natural/grass thing). A further complexity, certainly. The amusing thing is, I only started using landuse=grass very recently, after deciding that recreation_ground was probably not a good tag for what we're describing here. > What I am guessing you have on the ground is a residential square with a > road round the outside and a grassy bit in the middle. Correct. > There is no way I > know of for mapnik to interpret this correctly from a double-tagged > single way like this. It would have to be able to draw a line, a road > fill, a second line, and then the grass fill. > Where these occur in areas I have seen (many grass filled squares in C > London) two parallel ways have been used - and yes I know that people > are using leisure=park "incorrectly" in order to get them rendered, but > probably excusable in London as they are often formal park/recreation > areas. Well, a technical limitation is just that - if it can't be done, we need to live with it. The end result is certainly unintuitive, though, since the only area tag named (the landuse) isn't the style that gets applied to the area. I feel the workaround coming on... Thanks, Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
Dermot There are two reasons why your "tag combination ought to be able to render correctly as is" statement is not valid. Firstly - as Thomas pointed out - mapnik likes one feature per way. Secondly - landuse=grass is not rendered at the moment (because of the vast disagreement on the whole landuse/natural/grass thing). What I am guessing you have on the ground is a residential square with a road round the outside and a grassy bit in the middle. There is no way I know of for mapnik to interpret this correctly from a double-tagged single way like this. It would have to be able to draw a line, a road fill, a second line, and then the grass fill. Where these occur in areas I have seen (many grass filled squares in C London) two parallel ways have been used - and yes I know that people are using leisure=park "incorrectly" in order to get them rendered, but probably excusable in London as they are often formal park/recreation areas. Cheers STEVE Steve Chilton, Learning Support Fellow Manager of e-Learning Academic Development Centre for Learning and Quality Enhancement Middlesex University phone/fax: 020 8411 5355 email: [EMAIL PROTECTED] http://www.mdx.ac.uk/schools/hssc/staff/profiles/technical/chiltons.asp Chair of the Society of Cartographers: http://www.soc.org.uk/ SoC conference 2008: http://www.abdn.ac.uk/cartographers08/ -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Dermot McNally Sent: 28 August 2008 00:49 To: osm Subject: [OSM-talk] Mapnik handling of highways that are also landuse... Folks - with reference to this: http://www.openstreetmap.org/?lat=53.72339&lon=-6.34273&zoom=17&layers=B 00FTF ...which is a section of the Mapnik render of the outcome of the very successful Drogheda Mapping Party in Ireland. Towards the centre of the map, you'll see what is represented as an oval area of residential highway. I can almost see why, but it struck me that it represents unwanted behaviour that could possibly be fixed. What we have here is a closed way of type highway=residential. Importantly, it isn't tagged as an area. It _is_ tagged (the same way) as landuse=grass. So without understanding the internals of Mapnik, it's as though the landuse, which applies at area-level, infects the highway tag and causes it to be considered as an area too. Clearly, I could simply draw a second way through the same nodes, and there are plenty of heated discussions over which approach is the saner. But it feels as though this tag combination ought to be able to render correctly as is. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapnik handling of highways that are also landuse...
On Thu, Aug 28, 2008 at 12:49 AM, Dermot McNally <[EMAIL PROTECTED]> wrote: > Folks - with reference to this: > > http://www.openstreetmap.org/?lat=53.72339&lon=-6.34273&zoom=17&layers=B00FTF > > ...which is a section of the Mapnik render of the outcome of the very > successful Drogheda Mapping Party in Ireland. Towards the centre of > the map, you'll see what is represented as an oval area of residential > highway. I can almost see why, but it struck me that it represents > unwanted behaviour that could possibly be fixed. > > What we have here is a closed way of type highway=residential. > Importantly, it isn't tagged as an area. It _is_ tagged (the same way) > as landuse=grass. So without understanding the internals of Mapnik, > it's as though the landuse, which applies at area-level, infects the > highway tag and causes it to be considered as an area too. > Correct > > Clearly, I could simply draw a second way through the same nodes, and > there are plenty of heated discussions over which approach is the > saner. But it feels as though this tag combination ought to be able to > render correctly as is. Mapnik still likes the one way per feature way of doing things, (as a matter of fact, so do I, seems more logical that the two can be separated if required later) > > Dermot > > -- > -- > Iren sind menschlich > > ___ > talk mailing list > talk@openstreetmap.org > http://lists.openstreetmap.org/listinfo/talk > -- Regards, Thomas Wood (Edgemaster) ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
