On 23 March 2015 21:39, Robert Haas
> On Mon, Mar 23, 2015 at 6:01 AM, Rajeev rastogi
> wrote:
> > The cost of re-scan of SeqScan node is considered to be same scan of
> > SeqScan node, which always assumes that the records is fetched from
> > disk and hence disk access cost is added (As we don’
On Mon, Mar 23, 2015 at 6:01 AM, Rajeev rastogi
wrote:
> The cost of re-scan of SeqScan node is considered to be same scan of SeqScan
> node, which always assumes that the records is fetched from disk and hence
> disk access cost is added (As we don’t know really how much memory will be
> availabl
During my routine work, I observed that incase of execution of plan having
inner node of NLJ as materialized node (on top of SeqScan) is slower compared
to non-materialized SeqScan node. This happens only if "Work_mem is not big
enough to hold all tuples in memory."
To make test easy and faster
