[OSM-dev] Looking for guidance on interpreting multipolygon relations

[Karl Newman]

On 11/3/07, Jon Burgess <jburgess777 at googlemail.com> wrote:
> On Sat, 2007-11-03 at 14:25 -0700, Karl Newman wrote:
> > An area calculation would work to
> > find the outermost ring, but after that it's undefined (the "holes"
> > might be separate, not concentric). I would think that true concentric
> > multipolygons would be rare, limited to things like a lake on an
> > island in the middle of a larger lake. But in that case, does the
> > island lake even need to be part of a multipolygon relation? Isn't the
> > exclusion of the island from the larger lake sufficient? Because when
> > the island lake is rendered it will be drawn over the unfilled "hole"
> > (which is the island), right? What am I missing?
>
> If we follow the normal GIS definition of a polygon then you can only
> have an outer ring plus a number of inner rings which are
> non-intersecting and not contained within one another. If you want
> something inside one of the holes then it should be a different
> multipolygon.
>
> In practice you probably want a park with no holes and a lake with holes
> for each of the islands. That matches the current Bettersea Park
> http://www.openstreetmap.org/?lat=51.47919&lon=-0.15541&zoom=16&layers=B0F
>
> The island appears with the park rendering because it is a hole in the
> lake. It does not require any individual tagging.
>
> A different example where someone has tried to tag the islands instead
> of using holes is the park next to Schiphol Airport
> http://www.openstreetmap.org/?lat=52.3074&lon=4.8336&zoom=14&layers=B0F
> This has several rendering problems both in the Mapnik and Osmarender
> layer.
>
>         Jon

Good, so I guess my thinking is correct. In that case, only the
outermost way needs to be given the "outer" role. Any other ways in
the relation would by definition be "inner".

Karl