On 2/13/2011 7:51 AM, Jonathan Lewis wrote:
My point is that the hint should supply the missing information (e.g.
selectivity) and not dictate the plan.
It's still a hint.

BTW, the IBM position (which may or may not match mine) is that the user
tells support (i.e. last resort) that they are stuck and support
provides the hint (i.e. no need to wait for a patch).
If hints are truly rarely needed then this ought to be an acceptable
compromise between idealism and pragmatism.

Cheers
Serge

--
Serge Rielau
SQL Architect DB2 for LUW, IBM Toronto Lab
Blog: tinyurl.com/SQLTips4DB2
Wiki: tinyurl.com/Oracle2DB2Wiki
Twitter: srielau

On Sun, 13 Feb 2011 09:19:06 -0500, Serge Rielau wrote:

That is true, that's why DB2 supports hints, in the first place:
http://tinyurl.com/48fv7w7
Hints are documented, so the users can use them if they think that hints
are necessary. Nice thing to know that support also advises the use of
hints.



--
http://mgogala.byethost5.com

Serge Rielau wrote,on my timestamp of 14/02/2011 1:19 AM:


And that is perfectly acceptable and makes a lot of sense.

"Serge Rielau" <srielau (AT) ca (DOT) ibm.com> wrote

Fair enough, but my comment was addressed to the argument that "the
optimizer doesn't need hints".

Mind you, the "missing information" is a lot harder than your response to
my first trivial example.

Take a 4 table join, with simple filter predicates and join predicates.

You may be able to say
selectivity(table1 xxx) ... selectivity(table4 yyyy)
but then you need to be able to say
selectivity(table1,table2 nnnn) - because you know it's much smaller
than the optimizer things
but you also may have to say
selectivity(table1,table3 mmmm)
selectivity(table1,table4 mmmm)
because you know those are much larger than the optimizer things
then you have to worry about three table selectivities to stop the picking
the wrong third table .. and so on.
Or you might decide that you need to be able to say:
selectivity (t1, t2, based on colX = coly, selectivity) -- or something
similar.

Ultimately it's much harder (and much less reasonable) for the DBA to
decide what the right information is for a particular query than it is to
know enough to determine a sensible path.

--
Regards

Jonathan Lewis
http://jonathanlewis.wordpress.com

On Feb 14, 5:32*am, "Jonathan Lewis" <jonat... (AT) jlcomp (DOT) demon.co.uk>
wrote:

I don't know if people outside of the US are following it, but IBM's
Watson computer is playing on Jeopardy. It's kind of an infomercial
for IBM, not that that is a bad thing. Big roomful of servers, with
an icon on a screen in the player's area showing how much it is
"sweating." Last night's first round ended in a tie between Watson
and one of the humans.

The funniest part was when Watson said "Harry Potter" when the correct
question was "Voldemort." 2880 POWER7 processor cores and 16
Terabytes of RAM, got it exactly backwards in 3 seconds.

jg
--
@home.com is bogus.
http://www.nbc11news.com/news/headli...9.html?ref=689

On Tue, 15 Feb 2011 09:14:56 -0800, joel garry wrote:

It's magic.



--
http://mgogala.byethost5.com