Alvaro Herrera <[EMAIL PROTECTED]> writes:
> No, a concurrent vacuum can't change that because vacuum can't change
> the page unless it can get a super-exclusive lock on it (which means
> nobody else can have a scan stopped at that page, which is exactly
> what this cursor has).
More to the point, vacuum certainly may not delete a row that's still
visible to any open transaction, which this row would be by definition.
And VACUUM FULL couldn't move it, because it couldn't get exclusive
lock on the table.
You'd probably have to forbid use of WHERE CURRENT for a cursor WITH HOLD
though, since that quite possibly would contain rows that don't exist
anymore.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
