[postgis-users] Re: Re bug with GIST indexes

[David Blasby]

Michael,

We have noted the problem you found (see chris's message to the list). 
 We havent been able to figure out why its doing what its doing.

Basically, if you're using the index, you're getting the incorrect 
results.  Everything we've looked at indicates it should be working as 
expected, so we're a bit stuck.

Both chris and I are very busy right now, so we dont have much time to 
spend tracking it down.  Hopefully things will be less busy next month.

I suggest you either (1) dont use the @ and ~ operators (2) only use 
them if you havent indexed the table (3)  re-write your queries so they 
use &&  ("overlap") instead of @/~ or (4) use postgresql 7.1 or 7.2.

Not many people actually use these operators (I think you're one of the 
only people who do), so no one noticed them doing something wrong.  

Next month we hope to release postgis with GEOS support - this will give 
access to robust implementations of the OGC relate() functions.  You'll 
be able to determine all sort of complex relationship between two 
geometries (not just their bounding boxes).

dave