Re: Switch to multi-inserts for pg_depend

Michael Paquier <michael@paquier.xyz>

From: Michael Paquier <michael@paquier.xyz>
To: Alvaro Herrera <alvherre@2ndquadrant.com>
Cc: Robert Haas <robertmhaas@gmail.com>, Andres Freund <andres@anarazel.de>, Postgres hackers <pgsql-hackers@lists.postgresql.org>, Daniel Gustafsson <daniel@yesql.se>
Date: 2020-08-13T04:40:45Z
Lists: pgsql-hackers
On Wed, Aug 12, 2020 at 07:52:42PM -0400, Alvaro Herrera wrote:
> Yeah.  As I understand, the only reason to have this number is to avoid
> an arbitrarily large number of entries created as a single multi-insert
> WAL record ... but does that really ever happen?  I guess if you create
> a table with some really complicated schema you might get, say, a
> hundred pg_depend rows at once.  But to fill eight complete pages of
> pg_depend entries sounds astoundingly ridiculous already -- I'd say it's
> just an easy way to spell "infinity" for this.  Tweaking one infinity
> value to become some other infinity value sounds useless.
> 
> So I agree with what Andres said.  Let's have just one such define and
> be done with it.

Okay.  Would src/include/catalog/catalog.h be a suited location for
this flag or somebody has a better idea?
--
Michael

Commits

  1. Switch to multi-inserts when registering dependencies for many code paths

  2. Use multi-inserts for pg_depend