Re: Multivariate MCV list vs. statistics target

From: Dean Rasheed <dean(dot)a(dot)rasheed(at)gmail(dot)com>
To: Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>
Cc: Thomas Munro <thomas(dot)munro(at)gmail(dot)com>, "Jamison, Kirk" <k(dot)jamison(at)jp(dot)fujitsu(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: Multivariate MCV list vs. statistics target
Date: 2019-08-01 14:42:35
Message-ID: CAEZATCVmHgG7LVCLHmyNj8TKQVx_tFwbRvfaxXwUNn8UO=-4Zw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Thu, 1 Aug 2019 at 11:30, Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
>
> I'll move it to the next CF. Aside from the issues pointed by Kyotaro-san
> in his review, I still haven't made my mind about whether to base the use
> statistics targets set for the attributes. That's what we're doing now,
> but I'm not sure it's a good idea after adding separate statistics target.
> I wonder what Dean's opinion on this is, as he added the current logic.
>

If this were being released in the same version as MCV stats first
appeared, I'd say that there's not much point basing the default
multivariate stats target on the per-column targets, when it has its
own knob to control it. However, since this won't be released for a
year, those per-column-based defaults will be in the field for that
long, and so I'd say that we shouldn't change the default when adding
this, otherwise users who don't use this new feature might be
surprised by the change in behaviour.

Regards,
Dean

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Fabien COELHO 2019-08-01 14:48:35 Re: refactoring - share str2*int64 functions
Previous Message Tom Lane 2019-08-01 14:37:06 Re: Avoid full GIN index scan when possible