Re: psql's \d and \dt are sending their complaints to different output files

Daniel Gustafsson <daniel(at)yesql(dot)se> writes:
>> On 19 Jun 2017, at 17:32, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> So, if we're getting into enforcing consistency in describe.c, there's
>> lots to do.

> Addressed in attached patch, see list of patches below.

I've pushed most of this. There are a couple of remaining
inconsistencies:

listTables and listDbRoleSettings print a custom message rather than
an empty table for no matches (but in QUIET mode they just do the
latter). I think that's actually a good decision for listDbRoleSettings,
because the user might be confused about which pattern is which, and
we can straighten him out with a custom error message. But the only
good argument for listTables behaving that way is that people are used
to it. Should we override backwards compatibility to the extent of
dropping the custom messages in listTables, and just printing an empty
table-of-tables?

> 0004 - Most verbose metacommands include additional information on top of what
> is in the normal output, while \dRp+ dropped two columns (moving one to the
> title). This include the information from \dRp in \dRp+. Having the pubname
> in the title as well as the table is perhaps superfuous, but consistent with
> other commands so opted for it.

I'm not sure about this one. It definitely seems bogus that \dRp+ is
omitting the owner column, but I'm less excited about duplicating the
pubname. We'd better make up our minds before v10 freezes, though.
Anybody else have an opinion?

regards, tom lane