Re: Improving N-Distinct estimation by ANALYZE

On Wed, 2006-01-04 at 14:49 -0500, Tom Lane wrote:
> Simon Riggs <simon@2ndquadrant.com> writes:
> > [ ... a large amount of analysis based on exactly one test case ... ]

[Hmmm, those are your opinions, not my words. Funny guy ;-) ]

The one test case just happens to be a very common 1:M relationship, an
example of which occurs within the TPC-H/DBT-3 test database. I picked
that so it was an obvious publicly accessible test case, rather than an
isolated customer issue that couldn't be aired fully. I was hoping to
allow the problem to be explored and improved.

> I think you are putting too much emphasis on fixing one case and not
> enough on considering what may happen in other cases ...

It's not just one problem, but knowing that you would accept only a
detailed analysis, I researched that one case so that it was
indisputable. 

> I'm dubious about the rest of this.

Excellent. Much better than "It's Wrong." I'll write some code and run some tests.

Thanks for reading it all; sorry it had to be so long.

Best Regards, Simon Riggs