Re: [PATCH 0/2] Improve ndistinct estimates for flattened UNION ALL

On Oct 16, 2026 05:36, David Rowley <dgrowleyml(at)gmail(dot)com> wrote
>> overstate Memoize hit rates.
>
> If that's true for Memoize, please provide a test case that shows a
> Memoize node being used on a DEFAULT_NUM_DISTINCT estimate.

Thanks for pointing this out. You are right.

I do not have a test case showing a Memoize node being used on a
DEFAULT_NUM_DISTINCT estimate. In fact, for Memoize I am trying to
address the opposite problem: when a join contains a UNION ALL
subquery, the planner detects DEFAULT_NUM_DISTINCT and avoids using
Memoize. This patch tries to widen that boundary. I still use the hi
estimate (the sum of all member estimates, when available) to avoid
overstating the hit rate.

Below are the cases I have tested where Memoize is used more
effectively. I have not yet found a counterexample in which Memoize
caused a worse plan.

Thanks.