On 04.Jul.2002 -- 08:38 AM, Leszek Gawron wrote:
> > That looks bad, indeed. I will have a look at it but it will take some
> > time to reproduce these values (gimme a week or so). I take your word
> > for it that you haven't changed anything else around it like move the
> > DBMS to another machine, upgrade your network, CPU, RAM, JDK, JDBC
> > driver or anything else.
> 
> I've collected these timings today substituting only cocoon directories

Great!

OK, I've downloaded the indicated files.

> > If you care to do some experiments yourself, the ESQL stuff is fairly
> > self-contained and does not depend on other changes in Cocoon. Thus
> > it's only o.a.c.c.l.m.xsp/Esql*.java and o.a.c.c.l.m.xsp.java/esql.xsl
> > 
> > The resulting java code produced from your XSP should be fairly
> > similar with both versions, actually. 
> > 
> > The main difference should be the introduction of various switch
> > statements to select different results. Those are outside the
> > retrieval loop and so should not account for a big slow-down.
> > 
> > The other major difference should be, that the current version does
> > not check for another result set while the older version did (in a not
> > JDBC compliant way - that was changed to be compliant and later
> > disabled).
> > 
> > Other things happend in the grouping / CallableStatement support which
> > should not be used in your case.
> 
> What exactly should I do ?

Well, if you want to spare me the experimenting you could take the
ESQL components from one version and replace the other versions,
rebuild and retry. If this setup shows the same timings as before
(e.g. improve with old ESQL in new C2 or deteriorate with new ESQL in
old C2), then it would indicate that it's indeed an ESQL problem.
Otherwise something else in C2 is responsible for the performance
loss. 

If it's no ESQL problem, timing differences for other XSPs would be
helpful. If they don't differ, it's likely that it's the Avalon
Excalibur version that accounts for it. Unfortunately, C2 and Avalon
are tightly coupled that it might be difficult to switch versions.

        Chris.
-- 
C h r i s t i a n       H a u l
[EMAIL PROTECTED]
    fingerprint: 99B0 1D9D 7919 644A 4837  7D73 FEF9 6856 335A 9E08


---------------------------------------------------------------------
Please check that your question  has not already been answered in the
FAQ before posting.     <http://xml.apache.org/cocoon/faq/index.html>

To unsubscribe, e-mail:     <[EMAIL PROTECTED]>
For additional commands, e-mail:   <[EMAIL PROTECTED]>

Reply via email to