Re: [HACKERS] Custom compression methods

On Tue, Mar 9, 2021 at 1:13 PM Justin Pryzby <pryzby(at)telsasoft(dot)com> wrote:
>
> On Tue, Mar 09, 2021 at 01:04:10PM +0530, Dilip Kumar wrote:
> > On Tue, Mar 9, 2021 at 2:45 AM Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> > >
> > Yeah, vacuum full or cluster will not re-compress the data. How about
> > providing syntax ALTER TABLE <tab-name> ALTER COLUMN <col_name> SET
> > COMPRESSION <com_name> REWRITE?
>
> It'd be strange to me if "rewrite" were associated with a column.
>
> Depending on what data strucutures you use, you might accidentally fail to
> rewrite the table if someone wrote something like:
> postgres=# alter table t alter a set compression pglz REWRITE, alter a set compression pglz;

I think that should not be a problem because in ATControlle, first we
execute all the AlterExec* command and therein we set various flags to
AlteredTableInfo->rewrite, so we can set one flag to compression
rewrite as well if rewrite is given for any column while updating the
compression method.

> Although I hope that's something to support someday for columnar AMs, I think
> table-rewriting now is done in entirety, and the syntax should reflect that.

I thought it would be good to give the rewrite along with the set
compression, and since we are setting compression column wise so
better to give rewrite also there. If we give rewrite as alter table
syntax and not with the compression then it will be like we can always
give the rewrite option for the table whether we are setting the
compression method or not.

--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com