Re: Fav. Urban Legend...

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Wed, 13 Mar 2002 15:08:57 -0800
Message-ID: <F001.00428681.20020313150857@fatcity.com>

Interesting stuff, but an interesting example of how myths take off. Did you notice the directive:

"Rewrite all correlated subqueries using in-line views".

But you can't guarantee that a rewrite can be achieved that will produce the same results.

If you don't believe that, look at the Niemiec book, page 48, where one of his performance successes is to turn a join into a correlated subquery --which in general WILL NOT RETURN THE SAME RESULT !!! If you doubt that,
think about the case for the 'real-life' example where an order has two lines for the same value.

So, what's the bet that two years from now, you'll be sitting in front of someone saying:

    Yes, I KNOW that Gaja says ... but your     rewrite was wrong, and in this case a     rewrite is not possible.

BTW - Oracle 9 will do the transform internally anyway into the 'corrected' solution, and Oracle 8.1.7 will do the same if you set

    "_unnest_subquery=true".

Of course, another argument for being careful how you state such advice (perhaps "Look for opportunities to .... because they can ...." is better than a firm directive) is that there are always other solutions, such as:

select

        dept_no,
        emp_no,
        sal
from    (
        select
                dept_no,
                emp_no,
                sal,
                avg(sal) over(partition by dept_no) as avg_sal
        from emp
        )               emp_avg
where
        sal > avg_sal

;

With a test case I set up, the CPU usage was slightly higher, but the logical I/O and latching significantly lower using this approach when compared to the hash-join method.

Jonathan Lewis
http://www.jlcomp.demon.co.uk

Next Seminar - UK, April 3rd - 5th
http://www.jlcomp.demon.co.uk/seminar.html

Host to The Co-Operative Oracle Users' FAQ http://www.jlcomp.demon.co.uk/faq/ind_faq.html

Author of:
Practical Oracle 8i: Building Efficient Databases

• Original Message ----- To: Multiple recipients of list ORACLE-L <mailto:ORACLE-L_at_fatcity.com> Sent: Wednesday, March 13, 2002 11:28 AM

Hey Guys,

Did anyone have a look at Gaja's new paper on myths and folklore about
Oracle at Craig Shallahamer's site? Wonderful reading.

--

Please see the official ORACLE-L FAQ: http://www.orafaq.com
--

Author: Jonathan Lewis
  INET: jonathan_at_jlcomp.demon.co.uk

Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------

To REMOVE yourself from this mailing list, send an E-Mail message to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in the message BODY, include a line containing: UNSUB ORACLE-L (or the name of mailing list you want to be removed from). You may also send the HELP command for other information (like subscribing). Received on Wed Mar 13 2002 - 17:08:57 CST