Thread: Sort returns more rows than seq scan?

Sort returns more rows than seq scan?

From

Pavel Stehule

Date:

01 March 2016, 14:15:51

I don't understand how is possible to get more rows from Sort node than in nested seq scan?

http://explain.depesz.com/s/568O

Sec scan returns 1.6M rows, Sort returns 2.7M rows

Regards

Pavel

Re: Sort returns more rows than seq scan?

From

Tom Lane

Date:

01 March 2016, 14:45:27

Pavel Stehule <pavel.stehule@gmail.com> writes:
> I don't understand how is possible to get more rows from Sort node than in
> nested seq scan?

The "extra" rows are re-fetches triggered by mark/restore requests from
the parent mergejoin.  The EXPLAIN instrumentation counts re-fetches as
if they were new rows.
        regards, tom lane

Re: Sort returns more rows than seq scan?

From

Pavel Stehule

Date:

01 March 2016, 14:53:09

2016-03-01 15:45 GMT+01:00 Tom Lane <tgl@sss.pgh.pa.us>:

Pavel Stehule <pavel.stehule@gmail.com> writes:
> I don't understand how is possible to get more rows from Sort node than in
> nested seq scan?

The "extra" rows are re-fetches triggered by mark/restore requests from
the parent mergejoin. The EXPLAIN instrumentation counts re-fetches as
if they were new rows.

aha,

Thank you for explain

Regards

Pavel

regards, tom lane