On 4/28/15 1:32 PM, Robert Haas wrote:
>> More than five years have passed since Heikki posted this, and we still
>> >haven't found a solution to the problem -- which neverthless keeps
>> >biting people to the point that multiple "user-space" implementations of
>> >similar techniques are out there.
> Yeah. The problem with solving this with an update is that a
> concurrent "real" update may not see the expected behavior, especially
> at higher isolation levels. Tom also complained that the CTID will
> change, and somebody might care about that. But I think it's pretty
> clear that a lot of people will be able to live with those problems,
> and those who can't will be no worse off than now.
But that's the same thing that would happen during a real update, even
if it was just UPDATE SET some_field = some_field, no? Doesn't
heap_update already do everything that's necessary? Or are you worried
that doing this could be user-visible (which as long as it's a manual
process I think is OK)?
--
Jim Nasby, Data Architect, Blue Treble Consulting
Data in Trouble? Get it in Treble! http://BlueTreble.com