Re: Synchronized scans

Heikki Linnakangas <heikki(at)enterprisedb(dot)com> writes:
> Tom Lane wrote:
>> It occurs to me that there's an actual bug here for catalog access.
>> The code assumes that it can measure rs_nblocks only once and not worry
>> about tuples added beyond that endpoint. But this is only true when
>> using an MVCC-safe snapshot.

> You would only miss tuples inserted after you began the seqscan. After
> you've began the scan, you're going to miss any tuples that are inserted
> before the current position anyway, so stopping the scan early shouldn't
> do any real harm.

Good point.

> It would only be a problem if you do something like:
> heap_beginscan(...)
> Lock
> while(heap_getnext) ...
> Unlock
> And expect to see all tuples inserted before acquiring the lock.

But that could be fixed by taking the lock before the heap_beginscan.
Indeed it's hard to conceive of a situation where you'd want/need to
take the lock afterward; in most cases the beginscan and the actual
scan are right together.

So I withdraw this complaint; it's complexity we don't need. I'll
add a comment about the point though.

regards, tom lane