On Fri, Feb 18, 2022 at 6:16 PM Bryn Llewellyn <bryn@xxxxxxxxxxxx> wrote:
(4) The clue to the whole thing is the semantics of the LoV for "pg_depend. deptype" (see below) — but you all kept this a secret from me!
I didn't even think to look at the system catalogs for guidance in this kind of thing. The catalogs are not what I consider end-user facing documentation; I'd want the core documentation (SQL Command Reference and exposition chapters for different features) to be sufficient for someone to understand how these things work. Now, I would probably have been a bit less certain of myself had I gone and looked at the catalogs early on. Seeing the specification for DEPENDENCY_AUTO (a) would have reset my internal consistency trigger.
David wrote:
This must be a typo: "depends on" and "is dependent upon" mean the same. I’m guessing that this was meant: « An extension neither depends on nor is *depended* upon by its members. » If so, then it’s wrong. Yes: an extension doesn’t depend on its members. This query (which runs without error) shows that an extension depends upon only the schema nominated by "create extension… with schema…".
But no: the objects do depend upon the extension that creates them, as this query shows:
Yeah, I seem to have had a brain fade there.
David J.
