[OSM-dev] Reading OSM History dumps

[Peter Körner]

Am 23.08.2010 15:18, schrieb Scott Crosby:
> I have a few comments that might give you a different impementation
> strategy that would work entirely in the database, generating geometry
> data and supporting all the usual indexes. If it works, it'd be a
> *lot* less custom coding and let the database deal with the heavy
> lifting.

The main problem atm. is, that there is no code in place that can 
actually create a database from an history dump. If you throw [1] 
against osmosis, it fails with a none-unique index in its internal 
storage routines, as everything is developed towards a single-id.

In the process of rewriting the pgsql tasks (rewriting, not 
re-inventing), I need to develop a new store that doesn't crash when I 
throw two versions of a node at it, so the step into calculating the 
way-node-versions isn't too far away.

I'll have another more intensive read on your ideas later.


Peter

[1] <http://svn.toolserver.org/svnroot/mazder/osmhist/2-nodemove.osm>