Mike Mascari <mascarim(at)yahoo(dot)com> writes:
> If my TRUNCATE TABLE patch was applied as submitted,
> (I haven't downloaded a newer snapshot yet), then
> it falls into category #2...same as VACUUM. It
> commits the current transaction before truncating
> the specified relation, then begins a new transaction.
I took all that out ;-) while updating it to compile against the current
state of heap_open et al. I see no need for multiple transactions in
TRUNCATE. It's really on a par with RENAME TABLE, since both have to
force a buffer flush.
regards, tom lane