Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
- Original Message - From: Igor Brejc To: David Groom Cc: David Earl ; Artem Pavlenko ; talk Openstreetmap Sent: Monday, February 11, 2008 4:04 PM Subject: Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular David Groom wrote: I'll edit the wiki. Could you put some visual examples, please? Artem, what i had in mind is now shown on http://wiki.openstreetmap.org/index.php/Relations/Proposed/Rivers David G I'm a bit late in entering this discussion, but at the first glance I see at least one problem with this proposal: representing tributaries (http://en.wikipedia.org/wiki/Strahler_Stream_Order), which must be treated separately, since their usually have their own names. Either you would have to define an OSM way splitting the main river and the tributary (which in a sense defeats the idea of this proposal) or you would let the renderer assume that it can draw such a segment by itself (which can sometimes be problematic, I suppose). I'm not against letting renderers do their jobs, but I think we should not presume that all devices will have CPUs powerful enough to provide interactive maps by processing complex geometry algorithms. Drawing something like this, for example: http://en.wikipedia.org/wiki/Image:Mouths_of_amazon_geocover_1990.png :) Thats a good point. I'm reasonably happy with the proposal for defining large rivers as closed areas, as outlined in the proposal http://wiki.openstreetmap.org/index.php/Proposed_features/Large_rivers . The proposal at http://wiki.openstreetmap.org/index.php/Relations/Proposed/Rivers was an attempt to satisfy some peoples concerns about asegment being drawn across the river to close the area, but I think in practice it probably causes more problems than it solves. David Igor -- http://igorbrejc.net ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
I agree with Thomas, we could tag rivers like we do for coastlines. If the river is connected to a coastline it would be rendered by the coastline rendering process. If it's not, it could be rendered the same way lakes are. If we need to draw a way across the river to close it, it's fine by me: that's the end of this extent of the river for OSM even if it doesn't match any real life physical object. Renaud. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
On 8 Feb 2008, at 19:12, David Groom wrote: - Original Message - From: Artem Pavlenko [EMAIL PROTECTED] To: David Earl [EMAIL PROTECTED] Cc: David Groom [EMAIL PROTECTED]; talk Openstreetmap talk@openstreetmap.org Sent: Friday, February 08, 2008 12:24 PM Subject: Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular On 8 Feb 2008, at 12:05, David Earl wrote: On 08/02/2008 11:54, David Groom wrote: You mean like http://wiki.openstreetmap.org/index.php/Relations/Proposed/ Rivers, which would be my ideal, Ah, yes. I was suggesting putting in the connections across the river as well, but there isn't any reason why if the renderer is building its own polygon from the relation that it can't imply a connection from the end of each way to the start of the other. Allowing more than one contiguous way on each bank would also be useful. I'll edit the wiki. Could you put some visual examples, please? Artem, what i had in mind is now shown on http:// wiki.openstreetmap.org/index.php/Relations/Proposed/Rivers David G Yep, we can live with this. Artem David Artem ___ 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] Large Rivers in general, mapnik rendering in Particular
David Groom wrote: I'll edit the wiki. Could you put some visual examples, please? Artem, what i had in mind is now shown on http://wiki.openstreetmap.org/index.php/Relations/Proposed/Rivers David G I'm a bit late in entering this discussion, but at the first glance I see at least one problem with this proposal: representing tributaries (http://en.wikipedia.org/wiki/Strahler_Stream_Order), which must be treated separately, since their usually have their own names. Either you would have to define an OSM way splitting the main river and the tributary (which in a sense defeats the idea of this proposal) or you would let the renderer assume that it can draw such a segment by itself (which can sometimes be problematic, I suppose). I'm not against letting renderers do their jobs, but I think we should not presume that all devices will have CPUs powerful enough to provide interactive maps by processing complex geometry algorithms. Drawing something like this, for example: http://en.wikipedia.org/wiki/Image:Mouths_of_amazon_geocover_1990.png :) Igor -- http://igorbrejc.net ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
On 11/02/2008, Robin Paulson [EMAIL PROTECTED] wrote: On 08/02/2008, David Groom [EMAIL PROTECTED] wrote: The main problem area seems to be that some people do not like the current proposal whereby a river is divided up in to separate closed areas. The reason being that the segment crossing the river to close the area marks a boundary which does not actually exist. Discussion on this could go on why is that a problem? i regard this as analogous to breaking a long road up into shorter ways. there is no property that changes at the join of the two ways, but for practical reasons (very long ways are bad), we break a 200km road into shorter pieces. as a further analogy, to map landuse=residential areas (or commercial, industrial, ...), we don't use one giant area for the whole city, but break it up into smaller areas, along arbitrary lines. breaking a river up similarly is no different at all ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
On 08/02/2008, David Groom [EMAIL PROTECTED] wrote: The main problem area seems to be that some people do not like the current proposal whereby a river is divided up in to separate closed areas. The reason being that the segment crossing the river to close the area marks a boundary which does not actually exist. Discussion on this could go on why is that a problem? i regard this as analogous to breaking a long road up into shorter ways. there is no property that changes at the join of the two ways, but for practical reasons (very long ways are bad), we break a 200km road into shorter pieces. why can this approach not be used for rivers, at least until we come up with something which is more elegant? it will provide a 'recommended' way of tagging wide rivers, something which doesn't exist at the moment. anything is better than nothing ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
On Feb 8, 2008 11:09 AM, Artem Pavlenko [EMAIL PROTECTED] wrote: We can make osm2pgsql or coastline tools to create polygons, but why not create them in the first place ? Can someone enlighten me, please ? If I wanted to draw the rivers as light blue* with dark blue riverbanks, wouldn't storing them as polygons would make this hard? I don't think it would be easy to work out which sections of the polygons are where the river continues as opposed to being the riverbank. If we store the riverbanks, then we can pre-process to our hearts content using osm2pgsql and the like. That way I could have riverbanks as polylines and rivers as polygons and render them as I see fit. The pre-processing could work very similarly to the coastlines, using a left- or right-hand side rule and continuing the riverbank where one way joins onto the next to construct the polygons required. Cheers, Andy * As I think more and more about contours, and semi-transparent renderings and so on I realise that most area-fills will be translucent with edges on my maps, so we need to avoid abutting polygons if we aren't intending to represent an edge. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
On 8 Feb 2008, at 11:27, David Earl wrote: You could do it as a relation. The river bank would be a set of ways (each of which shares its end nodes with the ends of one of the others), and you could have a role for the one or two ways which close the loop which says this is structural, not really part of the river bank. The renderer would have to assemble the polygon from the constituent ways (start with one way, find the end node as the start node of another way and so on), but then rendering would be as per any other polygon. It's a bit fiddly, but it removes the problems of the artificial connections across the water not eing idetifiable while at the same time still providing a complete polygon (albeit indirectly) for the renderer to work on. Yes, this sounds reasonable to me. David Artem ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
- Original Message - From: Martijn van Oosterhout [EMAIL PROTECTED] To: David Groom [EMAIL PROTECTED] Cc: talk@openstreetmap.org Sent: Friday, February 08, 2008 11:10 AM Subject: Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular On Feb 8, 2008 11:39 AM, David Groom [EMAIL PROTECTED] wrote: The main problem area seems to be that some people do not like the current proposal whereby a river is divided up in to separate closed areas. The reason being that the segment crossing the river to close the area marks a boundary which does not actually exist. Discussion on this could go on indefinitely, but it does really need a Mapnik expert to either (i) see if there is a way that Mapnik can render areas which are not closed (ie. comprised of two parallel ways), or (ii) if this is not , and will never be, possible then to state that fact , and we can then have a tag proposal which will render in both Mapnik and [EMAIL PROTECTED] Sure, the same way as coastlines. The question really becomes do we just want to make waterway=river work the same as coastlines. Mapnik can't render incomplete polygons the way you want to, and for that matter neither can osmarender. You get something that vaguely resembles the end result but in general it won't work. *Except* for coastlines, where there is a seperate process that handles them, for both osmarender and mapnik. The main issue in practice is we now have no standard way of tagging rivers, and people are relatively free to do what they like, with the result that large portions of the River Thames disappeared from the Mapnik layer recently http://www.informationfreeway.org/?lat=51.49lon=0.41zoom=11layers=F0B0F The rules are fairly simple: all areas must be closed, except for coastlines. People may not like the results, but it's what works right now. My point was that while a tag is still only at the proposal stage is a bit difficult to talk of rules and tell someone they are doing it wrong. :) David 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] Large Rivers in general, mapnik rendering in Particular
On 8 Feb 2008, at 11:26, Andy Allan wrote: On Feb 8, 2008 11:09 AM, Artem Pavlenko [EMAIL PROTECTED] wrote: We can make osm2pgsql or coastline tools to create polygons, but why not create them in the first place ? Can someone enlighten me, please ? If I wanted to draw the rivers as light blue* with dark blue riverbanks, wouldn't storing them as polygons would make this hard? I don't think it would be easy to work out which sections of the polygons are where the river continues as opposed to being the riverbank. OK, valid point. If we store the riverbanks, then we can pre-process to our hearts content using osm2pgsql and the like. That way I could have riverbanks as polylines and rivers as polygons and render them as I see fit. The pre-processing could work very similarly to the coastlines, using a left- or right-hand side rule and continuing the riverbank where one way joins onto the next to construct the polygons required. Can relations help here ? Cheers, Andy * As I think more and more about contours, and semi-transparent renderings and so on I realise that most area-fills will be translucent with edges on my maps, so we need to avoid abutting polygons if we aren't intending to represent an edge. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
On Feb 8, 2008 11:39 AM, David Groom [EMAIL PROTECTED] wrote: The main problem area seems to be that some people do not like the current proposal whereby a river is divided up in to separate closed areas. The reason being that the segment crossing the river to close the area marks a boundary which does not actually exist. Discussion on this could go on indefinitely, but it does really need a Mapnik expert to either (i) see if there is a way that Mapnik can render areas which are not closed (ie. comprised of two parallel ways), or (ii) if this is not , and will never be, possible then to state that fact , and we can then have a tag proposal which will render in both Mapnik and [EMAIL PROTECTED] Sure, the same way as coastlines. The question really becomes do we just want to make waterway=river work the same as coastlines. Mapnik can't render incomplete polygons the way you want to, and for that matter neither can osmarender. You get something that vaguely resembles the end result but in general it won't work. *Except* for coastlines, where there is a seperate process that handles them, for both osmarender and mapnik. The main issue in practice is we now have no standard way of tagging rivers, and people are relatively free to do what they like, with the result that large portions of the River Thames disappeared from the Mapnik layer recently http://www.informationfreeway.org/?lat=51.49lon=0.41zoom=11layers=F0B0F The rules are fairly simple: all areas must be closed, except for coastlines. People may not like the results, but it's what works right now. 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] Large Rivers in general, mapnik rendering in Particular
On 8 Feb 2008, at 10:39, David Groom wrote: The proposed tag waterway = river, http://wiki.openstreetmap.org/index.php/Proposed_features/ Large_rivers , has been at proposal stage for over 18 months, which seems far too long for a tag which represents such an important feature. The main problem area seems to be that some people do not like the current proposal whereby a river is divided up in to separate closed areas. Representing features (like rivers) as well-formed closed polygon sounds good to me. The reason being that the segment crossing the river to close the area marks a boundary which does not actually exist. Discussion on this could go on indefinitely, but it does really need a Mapnik expert to either (i) see if there is a way that Mapnik can render areas which are not closed (ie. comprised of two parallel ways), Of course there is a way, but I'm not convinced at all we should take this approach. or (ii) if this is not , and will never be, possible then to state that fact , and we can then have a tag proposal which will render in both Mapnik and [EMAIL PROTECTED] We can make osm2pgsql or coastline tools to create polygons, but why not create them in the first place ? Can someone enlighten me, please ? The main issue in practice is we now have no standard way of tagging rivers, and people are relatively free to do what they like, with the result that large portions of the River Thames disappeared from the Mapnik layer recently http://www.informationfreeway.org/? lat=51.49lon=0.41zoom=11layers=F0B0F As a short term solution we can replace problematic coastline tile in London (100x100km vectors) with old one, I guess. David Artem ___ 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] Large Rivers in general, mapnik rendering in Particular
David Groom wrote: Sent: 08 February 2008 10:40 AM To: talk@openstreetmap.org Subject: [OSM-talk] Large Rivers in general, mapnik rendering in Particular The proposed tag waterway = river, http://wiki.openstreetmap.org/index.php/Proposed_features/Large_rivers , has been at proposal stage for over 18 months, which seems far too long for a tag which represents such an important feature. The main problem area seems to be that some people do not like the current proposal whereby a river is divided up in to separate closed areas. The reason being that the segment crossing the river to close the area marks a boundary which does not actually exist. Discussion on this could go on indefinitely, but it does really need a Mapnik expert to either (i) see if there is a way that Mapnik can render areas which are not closed (ie. comprised of two parallel ways), or (ii) if this is not , and will never be, possible then to state that fact , and we can then have a tag proposal which will render in both Mapnik and [EMAIL PROTECTED] The main issue in practice is we now have no standard way of tagging rivers, and people are relatively free to do what they like, with the result that large portions of the River Thames disappeared from the Mapnik layer recently http://www.informationfreeway.org/?lat=51.49lon=0.41zoom=11layers=F0 B0F As time goes on this is going to be an issue that comes up more and more frequently. Thankfully OSM has a much simpler approach to data than a full blown GIS approach where all edge features are tagged. In OSM we accept a certain degree of simplification (roads are created as regular liner features even if their width actually varies). I was looking at the Rotterdam area yesterday: http://www.openstreetmap.org/?lat=51.7578lon=4.7882zoom=13layers=B0FT and it was clear that rivers and other water courses are not ideally served by regular linear rendering. With the exception of canals, which on the whole have pretty regular width with length, rivers, streams and many other water courses do not and we should therefore arguably always think of them as areas. So where we have the required information we should always attempt to create area rendering rather than regular liner lines. Requiring closed areas though is not ideal for many reasons so achieving rendering between defined objects, whether by relationship or otherwise would seem logical to me. Cheers Andy ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular
- Original Message - From: Artem Pavlenko [EMAIL PROTECTED] To: David Earl [EMAIL PROTECTED] Cc: David Groom [EMAIL PROTECTED]; talk Openstreetmap talk@openstreetmap.org Sent: Friday, February 08, 2008 12:24 PM Subject: Re: [OSM-talk] Large Rivers in general, mapnik rendering in Particular On 8 Feb 2008, at 12:05, David Earl wrote: On 08/02/2008 11:54, David Groom wrote: You mean like http://wiki.openstreetmap.org/index.php/Relations/Proposed/Rivers, which would be my ideal, Ah, yes. I was suggesting putting in the connections across the river as well, but there isn't any reason why if the renderer is building its own polygon from the relation that it can't imply a connection from the end of each way to the start of the other. Allowing more than one contiguous way on each bank would also be useful. I'll edit the wiki. Could you put some visual examples, please? Artem, what i had in mind is now shown on http://wiki.openstreetmap.org/index.php/Relations/Proposed/Rivers David G David Artem ___ 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] Large Rivers in general, mapnik rendering in Particular
On 8 Feb 2008, at 12:05, David Earl wrote: On 08/02/2008 11:54, David Groom wrote: You mean like http://wiki.openstreetmap.org/index.php/Relations/Proposed/Rivers, which would be my ideal, Ah, yes. I was suggesting putting in the connections across the river as well, but there isn't any reason why if the renderer is building its own polygon from the relation that it can't imply a connection from the end of each way to the start of the other. Allowing more than one contiguous way on each bank would also be useful. I'll edit the wiki. Could you put some visual examples, please? David Artem ___ 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] Large Rivers in general, mapnik rendering in Particular
You could do it as a relation. The river bank would be a set of ways (each of which shares its end nodes with the ends of one of the others), and you could have a role for the one or two ways which close the loop which says this is structural, not really part of the river bank. The renderer would have to assemble the polygon from the constituent ways (start with one way, find the end node as the start node of another way and so on), but then rendering would be as per any other polygon. It's a bit fiddly, but it removes the problems of the artificial connections across the water not eing idetifiable while at the same time still providing a complete polygon (albeit indirectly) for the renderer to work on. David ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
