Useful explanation of PostgreSQL index format:
http://www.faqs.org/docs/ppbook/c13329.htm
I think you are aiming for the wrong thing.
The worst possible index is one with every value the same.
The second worst (still basically useless) is one with only two values.
The greater the differentiation of the data, the more workload is
reduced on a search.
Since it isn't a straight binary tree, I don't think that having highly
dissimilar data in the index should be a problem.
Do you have data or experience that shows otherwise?
-----Original Message-----
From: pgsql-general-owner@postgresql.org
[mailto:pgsql-general-owner@postgresql.org] On Behalf Of TJ O'Donnell
Sent: Tuesday, January 25, 2005 2:19 PM
To: pgsql-general@postgresql.org
Cc: tjo@acm.org
Subject: [GENERAL] visualizing B-tree index coverage
Does anyone know of a tools that allows one to visualize
the tree created by a multi-column B-tree index?
A picture of a tree with branches, showing how "branchy" the
tree is would be great.
I'm wondering how well I've "clustered" the data in my table
using the multi-column index. In other words, do my
multi-columns sufficiently but not overly discriminate rows from each
other?
Do I have too many with the same index? (not enough branches)
Do I have a unique index for each row? (way too many branches)
Thanks,
TJ
---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings