[OSM-dev] Area data type -- again!
steggink at steggink.org
steggink at steggink.org
Tue Jun 24 09:52:43 UTC 2014
Hi Ilya,
Can you please explain how ways referencing areas "work"? Preferably
with some illustrations. Either I don't get it, or it doesn't make
sense. The way as I see it now you'll get some closed loops
along/inside ways.
I agree on introducing a new primitive for areas, in order to resolve
the ambiguity with closed ways, and also to make it explicit that the
multipolygon relation is not really a relation like other things (like
routes). That way all 2D features can be represented with a single
datatype. I also think that the ring concept should be introduced,
which is also used by the OGC and many other GIS- and DB-systems.
However, I'd leave the multipolygon concept partially intact. It would
be a huge amount of work to trace nodes repeatedly when many areas
share the same boundary. Take a country boundary as an example. It is
also used by regions, municipalities, villages, on both side of the
boundary. In this case a ring would be a series of ways.
In other cases it won't make sense to reuse existing ways. So in that
case one would have a ring which consists of nodes entirely.
Eventually a ring could become a hybrid of nodes and ways, but that
might become messy/confusing, thus hard to maintain.
As for outer/inner: either tag it explicitly (as in the multipolygon
relation; although there is no ring concept there), or have it
implicitly, by treating the first ring as the outer ring. I'd favor
the second option. I see no point in having the role depend on the
node order. IMO that is an artifact from old-fashioned GIS-systems.
Regarding maintenance: the editors could become more user friendly.
When it is clear which ways are to be used in a ring (in an area) the
user should not be confronted with putting them in the right order.
This is at least the case for a ring built by ways. Since the area is
not a "multipolygon relation" anymore, relation constraints do no
longer apply When a ring is based on nodes, this is a different story
(and very similar to current ways).
My two cents,
Frank
Quoting Ilya Zverev <zverik at textual.ru>:
> Hi all! There has been no discussion of areas since November 2012, and
> the last EWG meeting made me think of them for some reason. I believe
> a proper area type can be done, and it would nicely represent every
> kind of areas — from simple buildings to the coastline.
>
> Please see http://wiki.openstreetmap.org/wiki/User:Zverik/Areas
>
> It suggests having an area as both an atomic object - much like ways:
> list of lists of nodes, - and as a virtual object to which ways refer.
> Like multipolygon relations, but reversed: it does not reference ways
> (there could be too many: e.g. the coastline), but ways reference it.
> It allows for better consistency checks and smarter partial editing.
>
> IZ
>
>
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/dev
More information about the dev
mailing list