forrest-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From David Crossley <cross...@apache.org>
Subject Re: Profiling Forrest [FOR-572]
Date Mon, 29 Aug 2005 15:21:07 GMT
Ron Blaschke wrote:
> David Crossley wrote:
> > Ron Blaschke wrote:
> >> David Crossley wrote:
> 
> >> Good.  I could provide instructions how to set things up for debugging
> >> with IntelliJ IDEA, and profiling with Yourkit Java Profiler, if you'd
> >> like.
> 
> > That would be good. Perhaps we can have a document on the
> > Project tab of our docs that explains how developers can
> > do the benchmarking themselves. I use IDEA too, and take
> > up the Yourkit offer for opensource licenses.
> 
> How would I submit such a document?  As JIRA issue?

Yes please.

> >> > Good stuff Ron. How shall we proceed? I could get the
> >> > shell of a benchmark site started ('forrest seed-benchmark').
> > Okay done. Running that cammand will now produce a basic site.
> > We need to add some content to exercise the various parts
> > of the forrest processing.
> 
> Excellent!
> 
> >> Should I have a look how the Cocoon profiler block can be integrated?
> >> That is, if it is possible to make a plugin, or if the would need to
> >> put it into the seeded benchmark site.
> 
> > Yes please. I have also started to investigate that,
> > but don't wait for me.
> 
> I've started to look at things, and were able to look at the Cocoon
> profiler things with "forrest run" in site-author/.  Quite a number of
> things were already present within Forrest, but required some tweaking
> on my machine.

If there is anything that we can adjust in SVN, then please
send a patch. The rest could remain as instructions in the doc.

>  Here's what I did:
> 
> - replace the "profiler" matches with "profile.html" in profiler.xmap
> and sitemap.xmap
> 
> - move the "profile.html" match up in the pipeline, before the
> "*.html" match in sitemap.xmap
> 
> - add the profiler roles in forrest-core.xconf
> 
> - replace the caching and noncaching pipelines with their profiling
> counterparts; that is, the pipeline is still named "caching," but uses
> "org.apache.cocoon.components.profiler.ProfilingCachingProcessingPipeline."
> 
> Currently, I am wondering about three things:
> - How can the top level pipes be replaced with their profiling
> counterparts, without too much hassle?

Perhaps this is possible using an input.xmap
in the plugin. However i don't understand what
you mean by "top level".

>  They can, and probably should,
> keep their name "caching" and "noncaching."  The rest would probably
> fit easily within a plugin.
> 
> - The profiling document uses parameters for individual profiling
> results, for example ".../profile.html?key=-3752097465161278089."
> Not sure what this means for static site generation.
> 
> - For static site generation, profile.html would need to be the
> last page generated.

Normally that order cannot be specified. I wonder if
the java code of the LinkGatherer can be enhanced
to put it last.

-David

Mime
View raw message