False positive of not closed way for landuse

[robome]

When using multipolygon for combining multiple ways forming a ring as outlined in ​http://wiki.openstreetmap.org/wiki/Talk:Relation:multipolygon#Advanced_multipolygons, the validator does output a warning for landuse forest - not closed way.

While that for itself is correct (the way is really not closed), it's at least irritating because there's not really a problem. JOSM itself as well as Mapnik and Osmarender do draw the forest just fine.

[stoecker]

When learned Mapnik the advanced multipolygons?

[robome]

Replying to stoecker:

When learned Mapnik the advanced multipolygons?

I don't know. But if I'm not wrong it does, see relation 70787 and 30738 for example.

[stoecker]

Actually this is a validator problem and not one of multipoly (p.s. you don't need to set cc, you automatically get mails when using assignment or when you added a comment :-). Seems validator does not yet know about advanced multipolygons. Solution would probably be to turn of the closed way checks for multipolygon relations, as these are anyway checked by josm's drawing code and produce a different warning when wrong.

[bilbo]

But the problem is in multipoly too - it does not know advanced multipolygons either. There is algorithm to detect which way is outer, which works reliably in ordinary cases (one outer way that encompasses any amount of inner ways) but will result in a "guess" in case when there is no single way to be classified as outer way (or if you select some random polygons that won't form multipolygon no matter what roles you use). In such case, one way is guessed as role=outer and the rest are set as role=inner. Result: bad multipolygon

So multipoly plugin should first intrnally "join" all ways before deciding which part of new relation (it could be more than one way) should have role=outer. If validator detects this as bug, then this could be correct detection.

Which means I have to fix the plugin.

[bilbo]

I forgot I add check there so the multipoly first checks that all ways are closed. So in this case it won't create invalid multipolygon, but will just report error and won't do anything. But still I could add advanced multipolygon support, since this is required if outer way is over 2000 nodes. I know many places where such ways are used, mostly large forests, about 50 - 100 km large in reality.

[stoecker]

You may take the joining code from josm's mappaint class.

[stoecker]

In [o18088].

[stoecker]

BTW: 70787 and 30738 are no advanced multipolygons. Mapnik still does not support them as I have seen today after creating one. And Osmarender has also troubles when the area is really large. It gets time these issues get fixed in the renderers as we have a 2000 points limit for ways now.