[OSM-dev] Nominatim 2.3 release announcement

Sarah Hoffmann lonvia at denofr.de
Sun Oct 5 15:56:36 UTC 2014


On Sun, Oct 05, 2014 at 11:03:39AM +0200, Christoph Hormann wrote:
> On Sunday 05 October 2014, Sarah Hoffmann wrote:
> >
> > There are two new features that are most notable for users of the
> > search engine: there is now support for waterway relations, which
> > improves searching for rivers, [...]
> Nice.
> It seems though they are reported in the same way in the results lists 
> as river lines with 'River' which makes them hard to identify.  

Word of warning: although the instance at nominatim.osm.org (and consequently
the search at osm.org) runs the latest code of Nominatim, it hasn't
been reimported since waterwat relations have been introduced. The
database has only changed to the new schema, where OSM object have
been edited since.

> In principle i would have four major classes for water features:
> * standing water areas (natural=water + water!=river/channel)
> * flowing water areas (waterway=riverbank or natural=water + 
> water=river/channel)
> * waterways (waterway tag but not waterway=riverbank)
> * waterway relations

Well, the idea was to make only the 'river line' of rivers searchable
and complete ignore the areas as they have only limited meanining from
a geolocation point of view. So, it should show only waterway relations,
waterway ways that are not part of a relation (and are not something
potentially interesting like a waterfall) as well as natural=water.
Of course, it's all heuristics because tagging is far from consistent.
For some disussions see https://github.com/twain47/Nominatim/issues/155

I'm dimmly aware that there is a new proposal for water areas aiming
to unify fowing and standing water areas. I can't say I'm very happy
with it from a data processing point of view because it mixes primary
geographic objects that are of interest on its own (lakes etc.) with
secondary objects that only describe the area of another object
(i.e. river and riverline). Having to use a second tag to figure out
the meaning of the first is always problematic. Therefore, Nominatim
ignores the whole water=* issue in the hope that the tagging schema
doesn't really catch on. ;)


More information about the dev mailing list