This is probably a simple question, but I’m not very good at PostGIS and don’t fully grok all of this.
Basically I have a table (nodes) with a POINT column (point). I have created an index on this column
create index nodes__points on nodes using gist (point)
The column was created with
select addgeometrycolumn('nodes', 'points', 'POINT', 4326, 2)
I am using srid 4326 because I’m adding data that’s in the form (latitude, longitude). (i.e. the co-ordinate system where the position of Dublin, Ireland is lat=53.353 lon=-6.264 (which I’ve added with GeomFromText('POINT(-6.264 53.535)'))).
For each point, I want to find all points that are roughly within a 1km box centred on that point (so selcet a.id, count(*) from nodes as a, nodes as b where SOME_DISTANCE_FUNCTION_HERE(a.point, b.point, 1000) group by a.id;. It doesn’t have to be exact, just a rough hueristic figure. a 1km bbox is fine, a 1km circle is fine. It doesn’t have to be exactly 1km, just that order of magnitude.
The ST_Distance/ST_DWithin/etc. all use the units of the SRID, which for 4326/WGS64 is degrees (so 1 = 1 degree of lattitude/longitude). But I want to use metres.
I tried ST_distance_sphere and st_dwithin which can use metres, but if I do that, the explain shows that the index isn’t being used.
How can I get roughly what I want, and use the geographical index?
UPDATE: This is on PostgreSQL 9.1 and PostGIS 2.0 svn build.
Since writing this, I have discovered the “geographic” as opposed to the “geometry” type in PostGIS, which might do exactly what I want.