armand pirvu <armand.pi...@gmail.com> writes:
> My question is: I suspect the limit simply limits the fethching to the first
> n-records retrieved and has no implications whatsoever on the planner,
> meaning the planner ignores it. Am I right or wrong ?
You're quite wrong. The presence of a LIMIT causes the planner to prefer
"fast start" plans, since it will then optimize on the basis of picking
the lowest estimated cost to fetch the first N rows. As an example,
you're more likely to get an ordered indexscan than a seqscan-and-sort
for small N, though there are many cases where seqscan-and-sort wins
if the need is to fetch the whole table.
regards, tom lane
