Hi WanMil, On Mon, Feb 01, 2010 at 09:12:24PM +0100, WanMil wrote:

...

I checked http://www.openstreetmap.org/browse/relation/4370 and it looks OK to me. Three edges in the multipolygon are close to each other (two of them within the inner way 21252252) but the lines are not intersecting.

Marko

Marko,

the nodes http://www.openstreetmap.org/browse/node/228621224 and http://www.openstreetmap.org/browse/node/228621268 are mapped to the same mkgmap coordinate 2804010/1161649.

That's what I suspected. Would it be feasible to process the multipolygons before converting to mkgmap coordinates? Or how hard would it be to process multipolygons that contain intersecting lines? Marko

Hi WanMil, On Mon, Feb 01, 2010 at 10:59:47PM +0100, WanMil wrote:

If you have intersecting lines the mp code cannot decide which polygon is contained by another polygon. From my point of view this cannot be fixed without major drawbacks.

Overlapping lines are not allowed for multipolygons but are sometimes used in OSM data. One could debate if they should be allowed or not but in the end they exist. To detect overlapping lines the mp code must be changed. At the moment I don't have a good and performant algorithm to do that, so I cannot say for sure if that's possible or not.

As there seem to be fewer than 10 such problematic multipolygons in Finland, I think that it will be easier to split the problematic multipolygons by hand. For instance, the Japanese garden http://www.openstreetmap.org/browse/relation/48542 does not need to be a multipolygon at all. I split the building multipolygon to a multipolygon and a separate polygon at the NE edge. According to municipal aerial imagery and data on property boundaries, the building mass consists of several separate buildings, so my resolution of the problem should better reflect the reality. Marko