From: | Robert Haas <robertmhaas(at)gmail(dot)com> |
---|---|
To: | Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> |
Cc: | Tatsuro Yamada <tatsuro(dot)yamada(dot)tf(at)nttcom(dot)co(dot)jp>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
Subject: | Re: progress report for ANALYZE |
Date: | 2019-07-11 02:23:55 |
Message-ID: | CA+TgmoaYObjivxEmEc-yOeGOFJaOuYd0GPbk_n3T0gr6OFGF-g@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On Wed, Jul 10, 2019 at 9:26 AM Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> wrote:
> On 2019-Jul-10, Robert Haas wrote:
> > On Tue, Jul 9, 2019 at 6:12 PM Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> wrote:
> > > Hmm, ok. In CREATE INDEX, we use the block counters multiple times.
> >
> > Why do we do that?
>
> Because we scan the table first, then the index, then the table again
> (last two for the validation phase of CIC). We count "block numbers"
> separately for each of those, and keep those counters in the same pair
> of columns. I think we also do that for tuple counters in one case.
Hmm. I think I would have been inclined to use different counter
numbers for table blocks and index blocks, but perhaps we don't have
room. Anyway, leaving them set until we need them again seems like the
best we can do as things stand.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
From | Date | Subject | |
---|---|---|---|
Next Message | Thomas Munro | 2019-07-11 02:24:17 | Re: Copy data to DSA area |
Previous Message | Michael Paquier | 2019-07-11 01:11:45 | Re: using explicit_bzero |