Re: --sync-method isn't documented to take an argument

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Wed, Oct 4, 2023 at 11:27 AM Nathan Bossart <nathandbossart(at)gmail(dot)com> wrote:
>> Here's a patch. I didn't address the class="parameter" stuff at all. I
>> figured it would be best to handle that separately.

> I guess I'll vote for including class=parameter in this addition for
> now, as that appears to be the majority position in the documentation
> today. If we get a consensus to change something, so be it. But also,
> if you don't want to do that, so be it.

FWIW, I just did a little sed hacking to count the instances of the
different cases in the docs as of today. I found

4038 <replaceable>
3 <replaceable class="command">
4017 <replaceable class="parameter">

The three with "command" are all in plpgsql.sgml, and are marking up
query strings in the synopses of EXECUTE and variants. I'm inclined
to argue that those three are actually wrong, on the grounds that

(1) From the perspective of EXECUTE, you could equally well say that
the string to be executed is a parameter;

(2) Our general convention elsewhere is that "command" refers to a
command type such as SELECT or UPDATE, not to a complete query string.

In any case, trying to standardize this looks like it would be a
huge amount of churn for very little gain. I'd recommend making
your markup look similar to what's immediately adjacent, if possible,
and not sweating too much otherwise.

regards, tom lane