Re: make \d pg_toast.foo show its indices

On Tue, May 7, 2019 at 6:03 PM Stephen Frost <sfrost(at)snowman(dot)net> wrote:
> Alright, maybe I'm not the best representation of our user base, but I
> sure type 'select oid,* from pg_class where relname = ...' with some
> regularity, mostly to get the oid to then go do something else. Having
> the relfilenode would be nice too, now that I think about it, and
> reltuples. There's ways to get *nearly* everything that's in pg_class
> and friends out of various \d incantations, but not quite everything,
> which seems unfortunate.
>
> In any case, I can understand an argument that the code it requires is
> too much to maintain for a relatively minor feature (though it hardly
> seems like it would be...) or that it would be confusing or unhelpful to
> users (aka "noise") much of the time, so I'll leave it to others to
> comment on if they think any of these ideas be a useful addition or not.
>
> I just don't think we should be voting down a feature because it'd take
> a bit of extra effort to make our regression tests work with it, which
> is all I was intending to get at here.

I think it's unjustifiable to show this in \d output. But maybe in
\d+ output it could be justified, or perhaps in the \d++ which I seem
to recall Alvaro proposing someplace recently.

I think if we're going to show it, it should be on its own line, with
a clear label, not just in the table header as you proposed.
Otherwise, people won't know what it is.

I suppose the work we'd need to make it work with the regression tests
is no worse than the hide_tableam crock which Andres recently added.
That is certainly a crock, but I can testify that it's a very useful
crock for zheap development.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company