Re: [OSM-talk] Mapping huge lakes as coastline
On Apr 14, 2011, at 1:52 AM, Jukka Rahkonen wrote: > Teemu Koskinen mbnet.fi> writes: > >> >> I converted a few of the biggest lakes in Finland a few years ago to >> coastlines, and they worked fine, until last year some other user converted >> them to multipolygons with natural=water -tags. He also splitted the biggest >> lake (Päijänne) in pieces, which created arbitrary lines across the lakes at >> random where the lake was divided to different polygons. >> >> The biggest lakes in Finland have tens of thousands (or even hundreds of >> thousands) nodes and a LOT of islands, so it's not practical to represent >> them >> as (multi)polygons IMO. > > It is not practical, either, to represent them as coastlines. For example > osm2pgsql is not importing coastlines into PostGIS at all but users must use > the > processed land polygons as shapefiles for rendering these coastline lakes. One > may say it works fine with Mapnik rendering because of this shapefile > workaround. Some could call it as a dirty hack. For example, it gets > complicated > when somebody wants to add tags for the lakes and islands. FWIW, Dane added a --keep-coastlines flag to recent versions of osm2pgsql. It hasn't fully propagated out to various package managers and things, but it's a big help in these situations. -mike. michal migurski- m...@stamen.com 415.558.1610 ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapping huge lakes as coastline
Jukka Rahkonen latuviitta.fi> writes: > By the way, i checked that the biggest lake polygon in the data of the > National > land survey of Finland is the lake Saimaa, and it has exactly 287273 vertices > and more than 5000 islands. It is a bit heavy to handle in PostGIS and Oracle > Spatial and with GIS programs but not at all impossible. Sorry, I made a wrong query and the correct number is 820357 vertices. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapping huge lakes as coastline
Teemu Koskinen mbnet.fi> writes: > > I converted a few of the biggest lakes in Finland a few years ago to > coastlines, and they worked fine, until last year some other user converted > them to multipolygons with natural=water -tags. He also splitted the biggest > lake (Päijänne) in pieces, which created arbitrary lines across the lakes at > random where the lake was divided to different polygons. > > The biggest lakes in Finland have tens of thousands (or even hundreds of > thousands) nodes and a LOT of islands, so it's not practical to represent > them > as (multi)polygons IMO. It is not practical, either, to represent them as coastlines. For example osm2pgsql is not importing coastlines into PostGIS at all but users must use the processed land polygons as shapefiles for rendering these coastline lakes. One may say it works fine with Mapnik rendering because of this shapefile workaround. Some could call it as a dirty hack. For example, it gets complicated when somebody wants to add tags for the lakes and islands. By the way, i checked that the biggest lake polygon in the data of the National land survey of Finland is the lake Saimaa, and it has exactly 287273 vertices and more than 5000 islands. It is a bit heavy to handle in PostGIS and Oracle Spatial and with GIS programs but not at all impossible. There is a wiki page about the future of areas in OSM. Handling big lakes is one more thing to be discussed there, see http://wiki.openstreetmap.org/wiki/The_Future_of_Areas In the data ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapping huge lakes as coastline
On 14 April 2011 03:24, M∡rtin Koppenhoefer wrote: > This results in bad rendering for low zoom tiles, with the lake > showing up on zoom6 but not on zoom5 (in Mapnik). Wouldn't it be better to fix the rendering side of things, than incorrectly mapping just so it renders how you expect it to? In terms of fixing the rendering, all that would need to be done is have polygons of a certain area render sooner. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Mapping huge lakes as coastline
On Wednesday 13 April 2011 20:24:36 M∡rtin Koppenhoefer wrote: > I wonder if it would not be better to map really big lakes as > coastline. This is done somewhere, e.g. here > http://www.openstreetmap.org/browse/relation/555716 > (Baikal lake) > > but it is not done here: > http://www.openstreetmap.org/browse/relation/1308279 > (Lake Onega) > > This results in bad rendering for low zoom tiles, with the lake > showing up on zoom6 but not on zoom5 (in Mapnik). > > A few similar cases (quite big lakes modeled as water and not as > coastline) can be found all over the world, e.g. > - http://www.openstreetmap.org/browse/relation/1239458 > (Vänern in Sweden, which by the way has a warning attached that > suggests it once was a coastline ;-) ) > > - http://www.openstreetmap.org/browse/relation/38997 > (Lago de Nicaragua) > > - some lakes in Canada, ... > > Modelling these as multipolygons also might slower the rendering speed > the more complex the polygons get by adding further detail. > > cheers, > Martin > I converted a few of the biggest lakes in Finland a few years ago to coastlines, and they worked fine, until last year some other user converted them to multipolygons with natural=water -tags. He also splitted the biggest lake (Päijänne) in pieces, which created arbitrary lines across the lakes at random where the lake was divided to different polygons. The biggest lakes in Finland have tens of thousands (or even hundreds of thousands) nodes and a LOT of islands, so it's not practical to represent them as (multi)polygons IMO. Teemu Koskinen ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk