Re: Extended statistics improvement: multi-column MCV missing values

On 6/7/26 18:22, Enrique Sánchez wrote:

> ## v3-0001
> - logic remains the same
> - added a `IsA(clause, RestrictInfo)` check before casting to
> RestrictInfo in `mcv_can_cap()`
> - added some tests

1. mcv_can_cap() reimplements logic already present in
dependency_is_compatible_clause(). Shall we combine the two in order to
avoid code duplication?

2. mcv_can_cap() runs unconditionally before the if (is_or) branch, but
can_cap is only consumed in the else/AND branch. This means
mcv_can_cap() - including get_oprrest, syscache lookups per clause -
runs for every OR query as wasted work. Both can_cap and covered_attnums
should be moved inside the else branch.

>
> ## v3-0002
> - used ndistinct to calculate `non_mcv_sel = (1.0 - mcv_totalsel) /
> (ndistinct - mcv_nitems)` and applied it as an upper bound for non-MCV
> combinations, similarly to what `var_eq_const()` does in `selfuncs.c`

get_ndistinct_for_keys() reimplements the ndistinct item lookup already
present in estimate_multivariate_ndistinct(). Both functions iterate
over MVNDistinct->items match by attributes count.

--
Best regards,
Ilia Evdokimov,
Tantor Labs LLC,
https://tantorlabs.com/