Re: add more frame types in window functions (ROWS)

Andrew Gierth <andrew@tao11.riddles.org.uk>

From: Andrew Gierth <andrew@tao11.riddles.org.uk>
To: Hitoshi Harada <umi.tanuki@gmail.com>
Cc: pgsql-hackers@postgresql.org
Date: 2009-12-05T10:53:35Z
Lists: pgsql-hackers
>>>>> "Hitoshi" == Hitoshi Harada <umi.tanuki@gmail.com> writes:

 Hitoshi> One thing for rule test, I checked existing regression test
 Hitoshi> cases and concluded DROP VIEW is necessary, or even VIEW
 Hitoshi> test for a specific feature is not needed. I remember your
 Hitoshi> aggregate ORDER BY patch contains "rules" test
 Hitoshi> changes. However, since processing order of regression tests
 Hitoshi> is not predictable and may change AFAIK, I guess it
 Hitoshi> shouldn't add those changes in rules.out.

Actually, looking more closely, the way you have it currently works only
by chance - "rules" and "window" are running in parallel, therefore the
view creation in "window" can break the output of "rules".

The order of regression tests is set in parallel_schedule and
serial_schedule; it's unpredictable only for tests within the same
parallel group.

I think a modification of the schedule is needed here; the only other
option would be to move the view creation into a different test.

-- 
Andrew.