Re: progress report for ANALYZE

From: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>
To: Tatsuro Yamada <tatsuro(dot)yamada(dot)tf(at)nttcom(dot)co(dot)jp>
Cc: vignesh C <vignesh21(at)gmail(dot)com>, Pg Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: progress report for ANALYZE
Date: 2019-11-05 13:38:50
Message-ID: 20191105133850.GA2494@alvherre.pgsql
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 2019-Nov-05, Tatsuro Yamada wrote:

> ==============
> [Session1]
> \! pgbench -i
> create statistics pg_ext1 (dependencies) ON aid, bid from pgbench_accounts;
> create statistics pg_ext2 (mcv) ON aid, bid from pgbench_accounts;
> create statistics pg_ext3 (ndistinct) ON aid, bid from pgbench_accounts;

Wow, it takes a long time to compute these ...

Hmm, you normally wouldn't define stats that way; you'd do this instead:

create statistics pg_ext1 (dependencies, mcv,ndistinct) ON aid, bid from pgbench_accounts;

I'm not sure if this has an important impact in practice. What I'm
saying is that I'm not sure that "number of ext stats" is necessarily a
useful number as shown. I wonder if it's possible to count the number
of items that have been computed for each stats object. So if you do
this

create statistics pg_ext1 (dependencies, mcv) ON aid, bid from pgbench_accounts;
create statistics pg_ext2 (ndistinct,histogram) ON aid, bid from pgbench_accounts;

then the counter goes to 4. But I also wonder if we need to publish
_which_ type of ext stats is currently being built, in a separate
column.

--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Peter Eisentraut 2019-11-05 13:47:37 Re: v12 and pg_restore -f-
Previous Message amul sul 2019-11-05 13:14:50 Re: [HACKERS] advanced partition matching algorithm for partition-wise join