Re: planner chooses incremental but not the best one

On 18/12/2023 19:53, Tomas Vondra wrote:
> On 12/18/23 11:40, Richard Guo wrote:
> The challenge is where to get usable information about correlation
> between columns. I only have a couple very rought ideas of what might
> try. For example, if we have multi-column ndistinct statistics, we might
> look at ndistinct(b,c) and ndistinct(b,c,d) and deduce something from
>
> ndistinct(b,c,d) / ndistinct(b,c)
>
> If we know how many distinct values we have for the predicate column, we
> could then estimate the number of groups. I mean, we know that for the
> restriction "WHERE b = 3" we only have 1 distinct value, so we could
> estimate the number of groups as
>
> 1 * ndistinct(b,c)
Did you mean here ndistinct(c,d) and the formula:
ndistinct(b,c,d) / ndistinct(c,d) ?

Do you implicitly bear in mind here the necessity of tracking clauses
that were applied to the data up to the moment of grouping?

--
regards,
Andrei Lepikhov
Postgres Professional