Andrew Gierth <andrew(at)tao11(dot)riddles(dot)org(dot)uk> writes:
> Further experimentation reveals that this change has the effect of
> blocking constant-folding in aggregate input expressions that refer to
> constant outputs of the subquery.
> Is it worth doing anything about that?
Uh ... but I thought the point here is that the outputs aren't really
constant in the presence of grouping sets.
regards, tom lane