Re: pub/sub - specifying optional parameters without values.

Zheng Li <zhengli10(at)gmail(dot)com> writes:
> The behavior is due to the following code
> https://github.com/postgres/postgres/blob/master/src/backend/commands/define.c#L113

Yeah, so you can grep for places that have this behavior by looking
for defGetBoolean calls ... and there are quite a few. That leads
me to the conclusion that we'd better invent a fairly stylized
documentation solution that we can plug into a lot of places,
rather than thinking of slightly different ways to say it and
places to say it. I'm not necessarily opposed to Peter's desire
to fix replication-related commands first, but we have more to do
later.

I'm also not that thrilled with putting the addition up at the top
of the relevant text. This behavior is at least two decades old,
so if we've escaped documenting it at all up to now, it can't be
that important to most people.

I also notice that ALTER SUBSCRIPTION has fully three different
sub-sections with about equal claims on this note, if we're going
to stick it directly into the affected option lists.

That all leads me to propose that we add the new text at the end of
the Parameters <refsect1> in the affected man pages. So about
like the attached. (I left out alter_publication.sgml, as I'm not
sure it needs its own copy of this text --- it doesn't describe
individual parameters at all, just refer to CREATE PUBLICATION.)

regards, tom lane