Re: Re: reloption to prevent VACUUM from truncating empty pages atthe end of relation

On Fri, Apr 5, 2019 at 7:04 PM Robert Haas <robertmhaas@gmail.com> wrote:
>
> On Thu, Apr 4, 2019 at 9:19 PM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
> > As INDEX_CLEANUP option has been added by commit a96c41f, the new
> > option for this feature could also accept zero or one boolean
> > argument, that is SHRINK_TABLE [true|false] and true by default.
> > Explicit options on VACUUM command overwrite options set by
> > reloptions. And if the boolean argument is omitted the option depends
> > on the reloptions.
>
> Yes, I think that's how it should work, because that's how the other
> option works, and there's no compelling reason to be consistent.

Indeed, I totally agree.

> My preference is for "truncate" over "shrink".

I don't really like "shrink" either, but users already have problems
to get the difference between VACUUM and VACUUM FULL, I'm afraid that
"VACUUM TRUNCATE_TABLE" will just make things worse.