Jeff Davis wrote:
On Mon, 2008-01-28 at 23:13 +0000, Heikki Linnakangas wrote:
It's a good point that we don't want pg_dump to screw up the cluster
order, but that's the only use case I've seen this far for disabling
sync scans. Even that wouldn't matter much if our estimate for
"clusteredness" didn't get screwed up by a table that looks like this:
"5 6 7 8 9 1 2 3 4"
It doesn't seem like there is any reason for the estimate to get
confused, but it apparently does. I loaded a test table with a similar
distribution to your example, and it shows a correlation of about -0.5,
but it should be as good as something near -1 or +1.
I am not a statistics expert, but it seems like a better measurement
would be: "what is the chance that, if the tuples are close together in
index order, the corresponding heap tuples are close together?".
The answer to that question in your example is "very likely", so there
would be no problem.
Is there a reason we don't do this?
It has been discussed before, but no-one has come up with a good
measurement for that.
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
