On Sun, Jun 15, 2008 at 5:10 PM, Martijn van Oosterhout <[EMAIL PROTECTED]> wrote: > On Wed, Jun 11, 2008 at 3:58 PM, Andy Allan <[EMAIL PROTECTED]> wrote: >> The more complex case, which I'm surprised hasn't been discussed at >> any point in this feature's development, is how I render a river with >> a polygon fill and a polyline riverbank (e.g. a dark blue line down >> each riverbank). As far as I see it osm2pgsql will have to do some >> complex pre-processing to make sure that the sections of riverbank >> which span the river (i.e. aren't actually river banks) don't get >> rendered when rendering an outline of the river. > > If it ever gets decided to do riverbanks as just polylines instead of > complete polygons, the main effect will be that the processing will be > pushed out-of-line just like the coastlines are now. We'll have to > generate seperate riverbank shapefiles for all the riverbanks which > with then have to be processed seperately in the stylesheet.
Sorry, I meant linestring, not polyline. I just mean considering areas as also having edges, not just a fill, as is done for an increasing number of filled areas in our stylesheet. We had similar discussions way back when considering polygons-with-holes, where it was decided that non-existent edges of the area shouldn't exist in the database. So drawing a doughnut-shaped building as one continuous boundary with overlapping edges crossing to the middle and back out was frowned apon, since those edges don't exist. Similarly, the "riverbanks" that cross from one side of the river to the other don't exist either, so they concern me. It's disappointing that this hasn't been considered beyond the point of "does it look right with a solid blue fill". Even non-rendering queries, such as measuring the lengths of riverbank, become much harder to achieve. Cheers, Andy _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
