NikhilS <nikkhils@gmail.com> writes:
> This can be handled by setting index->isconstraint appropriately inside
> generateClonedIndexStmt().
Done.
> The fundamental question though is should we allow primary, unique
> CONSTRAINTS which use the index mechanism just as an implementation to be
> created using the "INCLUDING INDEXES" mechanism.
Yeah, this bizarreness was foreseen and agreed to back when we set up
LIKE INCLUDING CONSTRAINTS the way it was defined (ie, copying only
CHECK constraints and not other things called constraints). I was never
very thrilled with that definition myself, but it's a bit too late to
revisit it.
regards, tom lane