Re: More On 7.2 Distributions - Estimates For Number Distinct < 0

Mark kirkwood <markir@slingshot.co.nz> writes:
> everything looks good except for n_distinct ( its negative - should be 5)

This is okay; see the documentation about what n_distinct means.

I'd be interested in a better rule-of-thumb for estimating whether the
number of distinct values in a column should be assumed to scale with
the total table size or not.  That 10% threshold was just pulled out
of the air...

            regards, tom lane