> If the system needs to fetch more than a small percentage of the
> records, then seqscan *will* be faster. The issue you are dealing
> with seems to be misestimation of the retrieval percentage for this
> particular query, causing the planner to guess wrong about which
> kind of plan to use.
no worries...i'll try building a subset of the data and see if there is
some "threshhold" value
or...maybe its time i actually contributed some code to the project :)
i built an OO database engine a few years ago (in objective-c) that used a
modified N-tree approach to indicies that massively accelerated the
retrieval of a lot of "highly similar" data items
-duck
-----------------------------------------------------------------------------
david@backpack.com BackPack Software, Inc. www.backpack.com
+1 651.645.7550 voice "Life is an Adventure.
+1 651.645.9798 fax Don't forget your BackPack!"
-----------------------------------------------------------------------------