On Wednesday 1 August 2007 Gregory Stark's cat, walking on the keyboard,
wrote:
> You're right, the index contains pointers to *every* version of the tuple.
> So in a regular SELECT statement you don't need to look at the update chain
> at all.
>
> The main use of the update chain is when you want to perform an UPDATE or
> DELETE. In that case when you come across a record which is being updated
> by another transaction you must wait until that other transaction finishes
> and then update the resulting record (if you're in read-committed mode).
So, just to see if I got this, if the HeapTupleSatisfiesUpdate returns
HeapTupleUpdated the chain update must be walked, right?
> > Third, I read about not valid MVCC snapshots (e.g., SnapshotNow). What is
> > the meaning of such "strange" snapshots? Because postgresql should always
> > guarantee at least read committed isolation, and thus this should be done
> > thru MVCC.....
>
> They're needed for things like the above UPDATE chain following
uhm...so first I get a tuple version using HeapSatisfiesUpdate, then if the
tuple is HeapTupleUpdated I follow the update chain, then (may be) recheck
with HeapTupleSatisfiesSnapshot with a SnapshotNow?? Something like this?
Thanks,
Luca
