d

> -----Original Message-----
> From: Shapira, Yoav [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, August 21, 2002 8:15 AM
> To: Tomcat Users List
> Subject: RE: Tomcat Crash
> 
> 
> Hi,
> I suppose you can't post code for confidentiality reasons, so I won't
> even ask for that.  But I have a couple of questions that may help us
> understand the problem better.
> 
> If you're reading from a CSV file, why do you need a DB connection?
> Couldn't you just open a FileReader, read a line at a time, use
> StringTokenizer to get your tokens?  This approach will be 
> very fast and
> use very little memory, even for huge CSV files, because only one line
> of the file has to be in memory at a time and the 
> StringTokenizer for a
> String is a small object.

This is what I'm doing. I actually used jakarta-oro to split the lines.




> 
> In fact, you problem statement says "No database work is done at this
> point..." so why all the discussion about Oracle connection pooling?

Jumped the gun a bit I guess, retrieving db connections and inserting into a database 
is the last thing I do and is when I notice the Tomcat process is about to die.

> 
> The next part would be the connection pooling implementation 
> questions.
> 
> 
> Did you try to get a new connection each time and not go through the
> pool?  That would degrade performance significantly (although 
> you could
> just get one connection at the beginning and use it until 
> you're done),
> but maybe memory behavior will be better?


I will be removing the pooling and just getting new connections to see if this makes a 
difference today, and will post results.


> 
> Did you try an alternative connection pooling implementation 
> instead of
> Oracle's?  One such as commons-dbcp or PoolMan?  If nothing else, it
> would allow you to pinpoint the blame on the Oracle connection pooling
> implementation.


I will be doing some performance tests with different pooling implementations soon. 
Will post my results if anyone would like to see them.  If anyone knows of benchmarks 
already available, please let me know.

> 
> And finally, are you trying to do any fancy cloning of data 
> in the file
> / objects in the ImportManager?

no

> 
> This shouldn't be a tough problem ;)  I'm sure we can figure 
> this out...
> 
> Yoav Shapira
> Millennium ChemInformatics

Dave

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

Reply via email to