"Mark Cave-Ayland" <m.cave-ayland@webbased.co.uk> writes:
> The only reason I kept the Relation parameter
> was because I wasn't sure if there was a historical reason why someone
> would need the relation information as well as the attribute
> information.
I can't think of one, but if someone did, they could extract the
relation OID from the pg_attribute row and re-open it for themselves.
So AFAICS this API does not omit any critical info.
I forgot to email you about the fetch_function revision, but I trust
it meets with your approval. Right now it's just a wrapper around
heap_fetch, but I thought we might conceivably want something different
when we do functional-index stats. The fetch function will give us
wiggle room if we need it.
regards, tom lane