Note in-line
--
Regards
Jonathan Lewis
http://www.jlcomp.demon.co.uk/faq/ind_faq.html
The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/seminar.html
Optimising Oracle Seminar - schedule updated Sept 19th
"Christian Antognini" <christian.antognini_at_trivadis.com> wrote in message
news:414d5a4a_at_post.usenet.com...
> > Since outlines are hints, we can infer that a complete
> > and legal set of hints cannot be ignored by the optimiser.
>
> If a complete and legal set of hints is used I basically agree with you.
Why
> basically? Because I already seen situations (fortunately rarely) where it
> is not the case! e.g. by copying outlines from one database (test) to
> another (production).
>
I would be very surprised if there were no differences
between the test and production systems that made
the hints from imported outline illegal on arrival.
There are several parameters documented as "if
you change this one, then an outline may cease
to work". And if you change columns from "not
null" to "nullable", hints may become illegal. Of
course, a outline referencing a non-existent index
is automatically in trouble - and test systems don't
always match indexes properly. It's also possible that
outlines simply don't import if you've used the
'swap outlines' code in 8i to attach an outline to a
statement.
If you'd said "after upgrade" I'd be much more likely to believe
you. There are bugs in some of the optimiser code that allow
hints to be overlooked or mis-used - my favourite was an update
statement which updated the wrong table if you put the relevant
hint into the code.
Received on Sun Sep 19 2004 - 06:37:49 CDT
