I have continued writing some C++-code to stuff some openstreetmap data into an sqlite database file. The results look promising so far. Though it took 13.5 hours to convert a 20GB europe.osm file, the first tests show amazing speed when selecting some simple data from the dataset. However, I’m curious how more complex queries will perform.
Currently I use one normalized table for all tags, regardless if the tags are linked to nodes, ways or relations. This surely saves a good portion of disk space, but it seems to slow down queries a lot. Maybe I should drop the helper tables, stuff the tags into the table redundantly and just link them directly to the data objects by a foreign key.
Obviously I feel adventurous about all this SQL and database stuff :) .
