Re: using extended statistics to improve join estimates

On 20/5/2024 15:52, Andy Fan wrote:
>
> Hi Andrei,
>
>> On 4/3/24 01:22, Tomas Vondra wrote:
>>> Cool! There's obviously no chance to get this into v18, and I have stuff
>>> to do in this CF. But I'll take a look after that.
>> I'm looking at your patch now - an excellent start to an eagerly awaited
>> feature!
>> A couple of questions:
>> 1. I didn't find the implementation of strategy 'c' - estimation by the
>> number of distinct values. Do you forget it?
>
> What do you mean the "strategy 'c'"?
As described in 0001-* patch:
* c) No extended stats with MCV. If there are multiple join clauses,
* we can try using ndistinct coefficients and do what eqjoinsel does.

>
>> 2. Can we add a clauselist selectivity hook into the core (something
>> similar the code in attachment)? It can allow the development and
>> testing of multicolumn join estimations without patching the core.
>
> The idea LGTM. But do you want
>
> + if (clauselist_selectivity_hook)
> + s1 = clauselist_selectivity_hook(root, clauses, varRelid, jointype,
> +
>
> rather than
>
> + if (clauselist_selectivity_hook)
> + *return* clauselist_selectivity_hook(root, clauses, ..)
Of course - library may estimate not all the clauses - it is a reason,
why I added input/output parameter 'estimatedclauses' by analogy with
statext_clauselist_selectivity.

--
regards,
Andrei Lepikhov
Postgres Professional