Tom Lane wrote:
>
> "Hiroshi Inoue" <Inoue@tpf.co.jp> writes:
> > As for HeapTupleSatisfies() there seems to be another choise to
> > let HeapTupleSatisfiesAny() be equivalent to HeapTupleSatisfiesNow()
> > other than always returning true.
>
> Wouldn't that break the other uses of SnapshotAny?
In theory no because HeapTupleSatisfies...() only touches
hint bits. What I mean is to implement a new function
HeapTupleSatisfiesAny() as
bool
HeapTupleSatisfiesAny(HeapTupleHeader tuple)
{HeapTupleSatisfiesNow(tuple);return true;
}
.
> I'm not sure
> it's what nbtree.c wants, either, because then the heap_getnext
> call wouldn't return recently-dead tuples at all.
>
nbtree.c has to see all(including dead) tuples and judge
if the tuples are alive, dead or removable via unified
time qualification.
regards,
Hiroshi Inoue