Re: Suggestion: provide a "TRUNCATE PARTITION" command

On Fri, Jan 8, 2021 at 10:12 AM Thomas Kellerer <shammat(at)gmx(dot)net> wrote:

> Michael Lewis schrieb am 08.01.2021 um 17:47:
> > > For me, it seems too easily error prone such that a single typo in
> > > the IN clause may result in an entire partition being removed that
> > > wasn't supposed to be targeted.
> >
> > I don't see how this is more dangerous then:
> >
> > delete from base_table
> > where partition_key in (...);
> >
> > which would serve the same purpose, albeit less efficient.
> >
> >
> > Delete has a rollback option, and you can dry-run to see impacted rows
> effectively. Truncate does not.
>
> TRUNCATE can be rolled back as well.
>

My apologies. There are other concerns with concurrent transactions, but
you are correct that it can be rolled back.

Still, no feedback on the effect that a truncate call is having on the DB
and may be doing more than intended fairly easily. I am not in the hackers
group so I couldn't say this feature would not be implemented. It just
seems unlikely given the philosophies of that group.