"Greg Sabino Mullane" <greg@turnstep.com> writes:
> This version covers most everything, and includes support for some
> wildcard type matching of schemas and tables,
Like what? I don't much want to reverse-engineer your code to discover
what you think should happen ... and you don't seem to have included any
documentation patches.
In general I'd recommend posting a straw-man proposal for the behavior
before you go to the trouble of writing code. (I had solicited ideas
awhile back, but don't recall that we arrived at any consensus.)
> All references to pg_ tables are now fully qualified
Yeah, I've been wondering whether we shouldn't do that. It will break
psql's \d commands for use with pre-7.3 backends, but of course the
joins against pg_namespace will pretty much do that anyway. Does anyone
want to bite the bullet and make psql backend-version-aware, like
pg_dump is? It would be a lot more work and code bulk, so I'm not sure
it's worth the trouble, but ...
> Currently, it does
> temporary relations first (in other words, temporary schemas), then
> non-public schemas (in order of last creation date), then public schemas,
> and finally the system schemas.
Surely it should pay attention to the search path. I don't think that
\d should show relations that are outside the current path (unless a
qualified name pattern is given).
I wonder in fact whether we could avoid having so much of this logic on
the psql side, and rely more on the backend to do the searching. I'm
not saying that the available name lookup functions are adequate for
psql yet --- but it's not too late to improve them if needed.
regards, tom lane