Re: Calculation of unused columns

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> It's probably true that in MOST of the cases where this comes up, the
> subquery can be flattened, from_collapse_limit permitting. But I
> think there are other cases, too.

Right ... and from_collapse_limit is not relevant here; only the form of
the subquery is. So I'd sure like to see some actual use cases before
we decide to expend planning cycles on this.

Just for fun, I hacked together a first cut at this. It's only about
120 lines but it's a bit cheesy (the limitation to not handling
appendrel members in particular). It passes regression tests and
seems to do what's wanted, but I'm not convinced it's worth the extra
cycles as-is, let alone with the appendrel limitation fixed.

regards, tom lane