Robert Haas <robertmhaas@gmail.com> writes:
> Perhaps I should RTFP before sticking my neck out too far,
> but... will you serialize EXCLUDE (a =), EXCLUDE (a &&), and EXCLUDE
> (a <some other operator>) differently?
No, and I'm not proposing to expose ASC/DESC/NULLS FIRST/LAST or
nondefault opclasses (to say nothing of non-btree AMs) or index
predicates either. The proposed patch is to my mind just a logical
extension of what we have always done --- namely, to pay attention
to index column names --- to some new cases that were never exposed
before.
We could certainly make it pay attention to all that stuff, but I have
the same feeling you do that it wouldn't produce readable results.
And it would make any compatibility issues a lot worse.
regards, tom lane