[mkgmap-dev] [MP] Cannot join ways to closed polygons

Tue Jan 19 19:12:07 GMT 2010

Hi WanMil,

On Tue, Jan 19, 2010 at 07:04:54PM +0100, WanMil wrote:
> > This seems to be another bogus warning:
> >
> > 2010/01/19 09:26:47 WARNING (MultiPolygonRelation): Cannot join the following ways to closed polygons. MP-Relation http://www.openstreetmap.org/browse/relation/302408
> >
> > I did
> > josm 'http://www.openstreetmap.org/?lat=64.73925&lon=29.49075&zoom=18'
> > selected the relation of the node and downloaded all members in the
> > relation editor.  Each 'inner' member (island) consists of a closed way
> > according to the relation editor.  (It shows closed loops next to the
> > members.)  The 'outer' member (coastline) consists of 7 ways, which
> > form a closed loop.  6 of the ways are 500 nodes and the last one
> > is 218 nodes.
> >
> > To me, the data seems fine.
> >
> > 	Marko
> 
> Marko,
> 
> there are some problems with mp code and osmosis/tile splitter. In your 
> case I think it's the tile splitter.

In that run, I had split finland.osm.bz2 into three tiles, at two latitudes:
lat=62.578125 and lat=64.819336.  Later, to get rid of a data structure
overflow in the southern tile, I moved the southern boundary to lat=62.226562.
You can get my files at http://www.polkupyoraily.net/osm/.

> The splitter puts only parts of the ways that crosses the tile bounds to 
> the resulting tile file. The handling for these incomplete mp ways on 
> the tile border is on the list which must be implemented next.

That is a plausible explanation.  The lake runs from lat=64.6895 to
lat=64.8580, crossing my northern tile boundary of lat=64.819336.

> By the way: did you see that lake with the old mp code and the same 
> splitter bounds?

There are many missing lakes in Northern Finland, even large reservoirs
that are shown on Garmin's base map.  I have not paid much attention to
lakes before.  Many lakes are displayed just as coastlines, but I hope
that I can soon enable sea polygons.  Finland's thousands of lakes and
islands should be the ultimate "acid test" for multipolygon handling.

Best regards,

	Marko