[OSM-dev] Binary OSM db (was: Re: on-disk indexing of geodata)

Freek freek_osm at vanwal.nl
Mon Oct 27 19:47:06 GMT 2008


On Wednesday 22 October 2008, Sascha Silbe wrote:
> > What is the advantage of having separate internal IDs?
>
> OSM ids are sparse (at most one object per id). Internal ids always
> point to exactly one object (at least one, at most one).

For static data that seems indeed to be advantageous. In a second, dynamic 
index only OSM IDs may be relevant, except perhaps for cross-reference to the 
static index.

> > Ah, nice. By the way, is all this still in (early) development, or is
> > there already a description/website/demo online?
>
> There's no website or demo, but it works quite fine in router prototype
> (routing works fine but CLI and postprocessing are not implemented yet).
> It's available (license is GPL) in my 2008 GNU arch repository [1] as
> osmbindb--devel--0.1 (the router is osmroute--devel--2.0 in my 2007
> repository [2]). [...]

Thanks, I hope I have some time to have a look at it shortly.

> [Geocoding with Gosmore]
> > I remembered that it did. The feature list on the wiki says:
> > "Incremental search of all tags. Results are ordered from nearest to
> > farthest."
>
> Judging from Nic Roets' mails on routing at osm, there is no spatial index
> involved.

Why would you need a spatial index to search for textual values?
Do you refer to this mail? 
http://lists.openstreetmap.org/pipermail/routing/2008-October/000564.html

It's an interesting topic (finding PoIs and such around some given location), 
Nic's solution sounds like a worthwhile idea. Generically, the idea would be 
that for larger sets of objects represented by the same tag (such as 
amenity=fuel, amenity=post_box, tourism=hotel, and perhaps even name=High 
Street, name=Bahnhofstraße) you would have a secondary spatial index (of any 
type).

-- 
Freek




More information about the dev mailing list