"scott.marlowe" <scott.marlowe@ihs.com> writes:
> not being able to use constants in functional indexes is a giant pain as
> well.
ISTM the appropriate way to attack this is to replace "functional
indexes" with "expressional indexes", wherein the value indexed
is anything that can be computed by a simple scalar expression (no
aggregates, no sub-selects, only immutable functions/operators).
This'd be a direct superset of what we can do now.
I haven't studied it in detail but I think this would not be a
terribly large project.
regards, tom lane