Date: Fri, 27 Mar 2026 13:58:39 -0500
From: Nathan Bossart <nathandbossart@gmail.com>
To: Michael Paquier <michael@paquier.xyz>
Cc: Andres Freund <andres@anarazel.de>,
	shihao zhong <zhong950419@gmail.com>,
	PostgreSQL-development <pgsql-hackers@postgresql.org>
Subject: Re: Fixes inconsistent behavior in vacuum when it processes multiple
 relations
Message-ID: <acbTX__yMnHSMIw4@nathan>
References: <aGIzjYsRrMCEADc5@paquier.xyz>
 <rzxpxod4c4la62yvutyrvgoyilrl2fx55djaf2suidy7np5m6c@3l2ln476eadh>
 <ab2iXwhRe3fFNPfp@nathan>
 <aglpy2jbnb75vufdx6hqkgokquzmoshcbnuxlcxhum4k7tyi7m@27m4hbxo2u2g>
 <acMGcoUwA9CiPPU1@nathan>
 <loig4i4hc4vuybn6uv6e2kgqm37sebbzujkrvdf6uhbxq5dsfm@7wbkmucts3eu>
 <acQAXPXov4cbyUx0@nathan>
 <acSYYcbfDk0bSgGq@paquier.xyz>
 <acVZ2O0938nWa73w@nathan>
 <acW7Qg1fjPeaKqeF@paquier.xyz>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <acW7Qg1fjPeaKqeF@paquier.xyz>
Archived-At: <https://www.postgresql.org/message-id/acbTX__yMnHSMIw4%40nathan>
Precedence: bulk

On Fri, Mar 27, 2026 at 08:03:30AM +0900, Michael Paquier wrote:
> Still, I slightly prefer your v2, where the interface of vacuum_rel()
> is leaner with all the other ones.  It comes at the cost of copying
> the input parameters into a temporary "copy" of VacuumParams, but I
> see the fact of marking the input "params" with a const as more
> valuable in the long-run, with less temptation to manipulate it
> directly especially it is not not marked with a const.  One small
> worry with v3 is that people like copy-pasting code around, and I
> suspect that v2 could discourage better the patterns that 2252fcd4276c
> has tried to improve and that 661643dedad9 had to fix.

I disagree with you here.  By passing the struct by-value, we are avoiding
scribbles on the original one without an explicit memcpy and without a big
comment warning folks to only use the copy (which seems like it'd be easy
to miss).  I think using a const pointer in most places makes sense, but
not if we need to immediately copy the contents to a local variable anyway.

-- 
nathan