For the reasons you point out, there's some drawbacks. However, the spatial index does provide you a means to get spatial subsets much more quickly than a full table scan. With 1.4 you could also store your points as a geohash, which is pretty compact and can be indexed and searched with a b-tree.
The question of what indexing buys you does rely a good deal on use case. If you're doing random access on 400M points, the fastest bet by far is to convert from single POINTs to MULTIPOINT patches of a few hundred points. kneufeld can elaborate on the multifarious benefits. P. On Tue, May 5, 2009 at 4:02 PM, <[email protected]> wrote: > > Hi, > > I'm currently working with a table storing around 300,000,000 point > geometries. No, it's not particularly fast :-) > > I'm unsure of the value in creating a spatial index effectively comprising a > BBOX on these, as the index then requires twice as many coordinates as the > actual data. > > Can anyone advise on the merits of spatial indexes on points? > > Thanks, > > Brent Wood > _______________________________________________ > postgis-users mailing list > [email protected] > http://postgis.refractions.net/mailman/listinfo/postgis-users > _______________________________________________ postgis-users mailing list [email protected] http://postgis.refractions.net/mailman/listinfo/postgis-users
