Re: extensible options syntax for replication parser?

On Thu, Oct 8, 2020 at 10:33 AM Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> That patch set includes this patch, and the reason for the behavior
> difference turned out to be that I had gotten an if-test that is part
> of this patch backwards. Here is v3, fixing that. It is a little
> disappointing that this mistake didn't cause any existing regression
> tests to fail.

I'm returning to this topic ~11 months later with a more definite
intent to get something committed, since my "refactoring basebackup.c"
patch set - that also adds server-side compression and server-side
backup - needs to add more options to BASE_BACKUP, and doubling down
on the present options-parsing strategy seems like a horrible idea.
I've now split this into two patches, one for BASE_BACKUP, and the
other for CREATE_REPLICATION_SLOT. I've rebased the patches and added
documentation as well. The CREATE_REPLICATION_SLOT patch now unifies
EXPORT_SNAPSHOT, NOEXPORT_SNAPSHOT, and USE_SNAPSHOT, which are
mutually exclusive choices, into SNAPSHOT { 'export' | 'use' |
'nothing' } which IMHO is clearer.

Last call for complaints about either the overall direction or the
specific implementation choices...

--
Robert Haas
EDB: http://www.enterprisedb.com