[OSM-dev] Gisify relations

[Matt Amos]

On Tue, Jun 16, 2009 at 10:04 AM, Frederik Ramm<frederik at remote.org> wrote:
> Andreas Kalsch wrote:
>> Conclusion: Make a
>> robot change all multipolygon relations to fit into the new model - so
>> the software built on top of OSM (starting with the Map renderers)
>> doesn't need to care any more of different representations.
>
> I don't think the new model mandates proper ordering of elements. It may
> *suggest* it, but not require it. My take is that if you edit
> multipolygons you should try to order the elements properly but if you
> read the data you should not assume they are ordered. Because even if we
> ran a robot that ordered things, this still does not guarantee that
> everything is always correct.

as you say, the new model merely allows ordering and it can safely be
ignored in most cases - including multipolygons. the outer/inner roles
should probably be considered hints, but nothing more.

> Maybe a relatively fast algorithm can be devised that tries to process
> things as-is but if that fails (e.g. if there is a gap in a ring) then
> switches back to the more computationally expensive method.

indeed. but i'd be very surprised if JTS/GEOS's algorithms aren't
already fast enough (e.g:
http://www.vividsolutions.com/jts/javadoc/com/vividsolutions/jts/operation/polygonize/Polygonizer.html
or http://download.osgeo.org/geos/doxygen/classgeos_1_1operation_1_1polygonize_1_1Polygonizer.html
).

cheers,

matt