Mailing-List: contact cocoon-dev-help@xml.apache.org; run by ezmlm
Precedence: bulk
Reply-To: cocoon-dev@xml.apache.org
From: "Carsten Ziegeler" <cziegeler@s-und-n.de>
To: <cocoon-dev@xml.apache.org>
Subject: RE: ContentAggregator and multithreating (migrating C2 -> C2.1M2),
 is it thread-safe?
Date: Fri, 23 May 2003 16:34:08 +0200
Message-ID: <LKENJPKDAIKAPKHEOCMMIECGDLAA.cziegeler@s-und-n.de>
MIME-Version: 1.0
Importance: Normal
In-Reply-To: <bala4s$f8f$1@main.gmane.org>
Content-Transfer-Encoding: 7bit
Content-Type: text/plain;
	charset="iso-8859-1"

Christoph Gaffga wrote:
> Just to see if I got it right, you wrote:
> > Have a look at the DefaultIncludeCacheManager - it can include in
> parallel.
> 
> my Aggregator does:
> 
>     // Part extends Thread
>     Part.run() {
>       SourceUtil.toSAX(part.source, buffer);
>     }
> 
>     // Aggregator.generate()
>     for(iterate all parts) {
>       part.start();
>     }
>     for(iterate all parts) {
>       part.join();
>       part.buffer.toSAX(contentHandler);
>     }
> 
> so I need to do something like:
> 
>     // Aggregator.generate()
>     for(iterate all parts) {
>       part.includeCacheManager = new DefaultIncludeCacheManager();
>       part.includeCacheManager.compose(this.manager);
>       part.session = includeCacheManager.getSession(this.par);
>       part.includeCacheManager.load(part.uri, part.session);
>     }
>     for(iterate all parts) {
>       part.includeCacheManager.stream(part.uri, part.session,
> contentHandler);
>     }
> 
> Is this right? Please give me a hint because I am not so familar 
> whith these
> IncludeCacheManager(and Session)-stuff.
Basically, yes. You have to turn on the parallel processing setting
a parameter (it's documented) when calling getSession(Parameter).


> P.S.:
> is there a special reason you are using
>     // wait
>     while (!loader.finished) {
>       try {
>         Thread.sleep(10);
>       } catch (InterruptedException e) { }
>     }
> in DefaultIncludeCacheManger (line 259). Isn't it better to have 
> the thread
> as a class variable (see line 212) and then use thread.join()?
> 
Hmmm, actually I don't know why this has been done in that way.
Perhaps join() is better.

Carsten