Re: More stable query plans via more predictable column statistics

Alex Shulgin <alex(dot)shulgin(at)gmail(dot)com> writes:
> On Mon, Apr 4, 2016 at 1:06 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> I'm inclined to
>> revert the aspect of 3d3bf62f3 that made us work from "d" (the observed
>> number of distinct values in the sample) rather than stadistinct (the
>> extrapolated estimate for the table). On reflection I think that that's
>> inconsistent with the theory behind the old MCV-cutoff rule. It wouldn't
>> matter if we were going to replace the cutoff rule with something else,
>> but it's beginning to sound like that won't happen for 9.6.

> Please feel free to do what you think is in the best interest of the people
> preparing 9.6 for the freeze. I'm not all that familiar with the process,
> but I guess reverting this early might save some head-scratching if some
> interesting interactions of this change combined with some others are going
> to show up.

I've reverted that bit; so we still have the improvements associated with
ignoring nulls, but nothing else at the moment. I'll set this commitfest
item back to Waiting on Author, just in case you are able to make some
more progress before the end of the week.

regards, tom lane