Re: Extended multivariate statistics are ignored (potentially related to high null fraction, not sure) - Mailing list pgsql-general

From Danny Shemesh
Subject Re: Extended multivariate statistics are ignored (potentially related to high null fraction, not sure)
Date
Msg-id CAFZC=Qrj00c8m3BPw2934xwCf01Ou6AA7g04C3=7ggRrGOmZ7Q@mail.gmail.com
Whole thread Raw
In response to Re: Extended multivariate statistics are ignored (potentially related to high null fraction, not sure)  (Bruce Momjian <bruce@momjian.us>)
List pgsql-general
Hey bruce, thanks for the prompt reply !

It reproduces in 12.11 and 13.7, we use a managed offering that is yet to include pg 14,
so sadly I can't try expressional extended statistics as of yet, nor can I attach gdb to the process
to debug the flow.

If any other directions come to mind, I'd really appreciate hearing them.
Thanks again !

On Wed, Jun 1, 2022 at 11:08 PM Bruce Momjian <bruce@momjian.us> wrote:
On Wed, Jun  1, 2022 at 07:28:58PM +0300, Danny Shemesh wrote:
> Hey everyone,
>
> I'm working on improving gathered statistics on several large tables (2TB /
> 500M records);
> I've created extended stats on correlated columns, ran analyze, compared the
> pre and post explains, and it seems as though the extended statistics are
> ignored - 
> the estimation doesn't change much, and I can accurately derive the est. rows
> from multiplying the single-column freqs from pg_stats with the supplied
> values.
>
>
> Context:
> - pg 12.8

You should be on the most recent version of 12.X, and I do see some
extended statistics change in the later releases you are missing.

--
  Bruce Momjian  <bruce@momjian.us>        https://momjian.us
  EDB                                      https://enterprisedb.com

  Indecision is a decision.  Inaction is an action.  Mark Batterson

pgsql-general by date:

Previous
From: Laurenz Albe
Date:
Subject: Re: How to display complicated Chinese character: Biang.
Next
From: Garfield Lewis
Date:
Subject: Accessing composite type elements