patch: pg_comments system view

On Fri, Jul 15, 2011 at 4:48 PM, Josh Berkus <josh(at)agliodbs(dot)com> wrote:
> I am unable to figure out the status of the pg_comments patch from this
> thread.  What's going on with it?

I don't blame you :-)

I think this thread got so confusing because two separate topics were
intertwined. (I'm going to try to change this thread subject to
reflect the fact that we're really talking about pg_comments here
now.)

First, the original post, with a small patch to fix one of the many
problems with psql's \dd command. That patch was rejected because it
only plugged one of the many problems with \dd. I have since started
another thread[1] with a plausible fix for \dd and several other
backslash commands, so that we will have working displays of all
comments with minimal duplication.

Second, we have the pg_comments view (latest version is the
"v10.WIP"). Despite its WIP tag, I think it is actually pretty close
to being complete at this point. The first concern which I raised a
concern in that thread[2]:

> 1.) For now, I'm just ignoring the issue of visibility checks;

is the only big issue I see as still outstanding. At first, I was
assuming that \dd should naturally read from pg_comments to fetch the
object comments it is interested in. But that would mean that we'd
need some way to duplicate those "visibility checks" \dd was doing,
either in \dd or in another "is_visible" column in pg_comments.

I haven't tried either of those options out yet, but I was worried
they'd both be tricky/ugly. Which leads me to think, maybe it's not so
bad if \dd stays the way I've suggested in thread[1], i.e. just
querying pg_[sh]description for the five object types it needs
directly. After all, \dd will IMO be close to useless/deprecated once
we have pg_comments; it'll be much easier to just query pg_comments
for what you're looking for directly, and \dd will only display five
funky object types, anyway.

How do folks feel about this issue?

The second concern I raised with the last pg_comments patch,

> I think now might be a good time to
> re-examine what types of objects are displayed by \dd.

should be handled by thread[1], and the third concern is just about
whitespace. Oh, and docs need some adjusting too, and it'd be nice if
someone sanity checked my guesses for the "is_system" column (or if
it's not needed, that's OK too).

So that's where the pg_comments patch stands, at least AIUI. Clear as
mud yet? :)

Josh

[1] http://archives.postgresql.org/pgsql-hackers/2011-07/msg00459.php
[2] http://archives.postgresql.org/message-id/CAK3UJRGNwKq0c2VsSYV-Mg55Y_kvZE=8FMR_Xt8Rzp__1LoNBQ@mail.gmail.com