[postgis-users] looking for some perf data

Paul Ramsey pramsey at refractions.net
Tue Nov 21 10:53:24 PST 2006

It's not so much a matter of "better" as it is a matter of "good 
enough". As long as your data volume remains low enough, you can keep 
your simple non-spatial tables and b-trees. At some point you'll flip 
though and need to move to spatial for "adequate" performance.  It all 
depends on your definitions of "large" and "adequate", none of which are 
  in scope for me :)


Martin Davis wrote:
> Well, those are range queries.  Not too surprising, 99.9% of spatial 
> queries are.
> The real question is: given that every modern spatial database uses some 
> sort of spatial index (R-tree, quad-tree, or grid), why even bother to 
> question whether B-trees might be better?
> raphael Jacquot wrote:
>> typically I'm looking to compare using
>> create table blah1 (
>>     lon    double precision,
>>     lat    double precision
>> )
>> with one index on lon and another one on lat
>> the classic request being
>> lon>constant1 and lon<constant2 and lat>constant3 and lat<constant4
>> and
>> create table blah2 (
>>     position Point
>> )
>> with a gist r-tree index
>> using the @ operator

More information about the postgis-users mailing list