On Sun, Aug 25, 2019 at 2:40 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > I was thinking of stashing an "equality is precise" flag in the
> > metapage of each nbtree index, since we will only determine this once,
> > at CREATE INDEX time.
>
> Sure.
I suppose that we'd add something new to CREATE OPERATOR CLASS to make
this work? My instinct is to avoid adding things that are only
meaningful for a single AM to interfaces like CREATE OPERATOR CLASS,
but the system already has numerous dependencies on B-Tree opclasses
that seem comparable to me.
There is a single case where nbtree stores a type that differs from
the type actually being indexed by the operator class: the "name"
case, where the underlying storage type is actually cstring. I'm not
sure whether or not this needs to be treated as its own kind of
special case. I suppose that we can ignore it completely, because
we're not directly concerned with the physical representation used
within an index. In fact, a major goal for this new infrastructure is
that nbtree gets to fully own the representation (it just needs to
know about the high level or logical requirements).
--
Peter Geoghegan
