On Thu, May 29, 2008 at 11:16 AM, Dave Stubbs <[EMAIL PROTECTED]> wrote: > Not actively as far as I'm aware. > > Martijn did some work to make the osm2pgsql slim mode work which you'd > need for this; this results in tables for ways, nodes, and relations, > as well as the standard feature tables.
Correct, in slim mode it now stores the actual ways and nodes. For making it parse diffs files you basically need the following steps: - Parse the actual diff itself, this shouldn't be too hard - Process the directives in the diff. This shouldn't be too hard either, you need to extend the out/mid api to have commands for modify and delete. This is also fairly straightforward, if a lot of typing. - Correctly update the output tables, this is harder. The basic idea is as follows: If a node is inserted, treat like a normal insert when bulk loading If a node is updated/deleted: - delete from points table with osm_id=this_node - Run new version through output processing (may not appear in new output if tags were removed) - Find all the ways/relations using this node and mark then processed=false For ways the system is the same. There is the additional step of handling processed=false ways, these need to have their output deleted (with matching osm_id). Relations using these need processed=false. Finally, to handle relations the same stuff needs to be repeated again, handling unprocessed relations. The whole deal about processed is because in general all the nodes of a way will be changed when a way is changed and the updates will not appear in order that they can be directly applied. Database changes are: - Indexes on the osm_id fields - Index on the nodes of a way/relations. Needs the intarray contrib module. It's not exceptionally difficult, it just needs a chunk of dedicated coding time... Hope this helps, -- Martijn van Oosterhout <[EMAIL PROTECTED]> http://svana.org/kleptog/ _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
