Robert Haas <robertmhaas@gmail.com> writes:
> On Thu, Mar 15, 2018 at 11:01 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> It might be worth looking at whether we couldn't fix the single-member-
>> Append issue the same way we fix no-op SubqueryScans, ie let setrefs.c
>> get rid of them. That's not the most beautiful solution perhaps, but
>> it'd be very localized and low-risk.
> That's definitely a thought; it's a probably the simplest way of
> saving the run-time cost of the Append node. However, I don't think
> it's a great solution overall because it doesn't get us the other
> advantages that David mentions in his original post. I think that to
> gain those advantages we'll need to know at path-creation time that
> there won't ultimately be an Append node in the finished plan.
Meh. We could certainly know that by inspection ("only one child?
it'll be history"). I remain of the opinion that this is a big patch
with a small patch struggling to get out.
regards, tom lane