Re: Use extended statistics to estimate (Var op Var) clauses

From: Tomas Vondra <tomas(dot)vondra(at)enterprisedb(dot)com>
To: Zhihong Yu <zyu(at)yugabyte(dot)com>
Cc: Mark Dilger <mark(dot)dilger(at)enterprisedb(dot)com>, Robert Haas <robertmhaas(at)gmail(dot)com>, Dean Rasheed <dean(dot)a(dot)rasheed(at)gmail(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: Use extended statistics to estimate (Var op Var) clauses
Date: 2021-12-13 02:04:05
Message-ID: 408649db-702d-8d5d-3294-a6ab673bc71e@enterprisedb.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 8/31/21 00:14, Zhihong Yu wrote:
> Hi,
> For patch 0002,
>
> +                   s1 = statext_clauselist_selectivity(root, clauses,
> varRelid,
> +                                                       jointype,
> sjinfo, rel,
> +
> &estimatedclauses, false);
> +
> +                   estimated = (bms_num_members(estimatedclauses) == 1);
>
> I took a look at clauselist_apply_dependencies() (called by
> statext_clauselist_selectivity) where estimatedclauses is modified.
> Since the caller would not use the returned Selectivity if number of
> elements in estimatedclauses is greater than 1, I wonder
> if a parameter can be added to clauselist_apply_dependencies() which
> indicates early return if the second element is added to estimatedclauses.
>

Hmmm, I'm not sure I understand your point. Are you suggesting there's a
bug in not updating the bitmap, or would this be an optimization? Can
you give an example of a query affected by this?

regards

--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Michael Paquier 2021-12-13 02:05:58 Re: Make pg_waldump report replication origin ID, LSN, and timestamp.
Previous Message Peter Geoghegan 2021-12-13 01:47:18 Defining (and possibly skipping) useless VACUUM operations