forrest-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Turner <>
Subject Re: scalation
Date Fri, 25 Apr 2003 03:24:03 GMT
On Thu, Apr 24, 2003 at 08:05:07PM +0200, Jan Bromberger wrote:
> Dear forrest dev,
> at the moment I am creating a tool to automatically generate documentation
> out of an LDAP schema. I thought it might be a good idea to produce xdocs and
> have them formated then by forrest. So first off I wanted to state that I
> like your project.
> Then again I only have such much time to write this mail to you, because it
> takes about 45minutes to render all pages. There are some 300 object classes
> and 362 attribute definitions, each having its own xdoc file and one book.xml
> for classes, one for attributes. Although I see that this might be not the
> scope of your software and it actually is quite much of data... well: Do you
> have some tuning tips for me?

I hope you're using the webapp mode ('forrest run') to develop the site?
:)  Waiting 45 mins to see each change would be.. painful.

One of the simplest things to do is remove unused transformers from the
pipeline.  If you're not using site: links, you can remove linkrewriter.
If you don't need automatic creation of @id attributes from <title>
elements, you can remove idgen:

497              <map:generate src="content/xdocs/{1}/{2}.xml" />
499              <map:transform type="linkrewriter" src="cocoon:/{1}/linkmap" />
501              <map:transform type="idgen" />

idgen is slow (~10% of render time) because it builds a DOM.
Linkrewriter is slow because the reload check for the cocoon: source is
done on each link rewrite, and is fairly slow (fixed in CVS Forrest).

Another thing you could try is using a HTTP tool like 'wget' to create a
static version of the site, instead of the Cocoon command-line, which is
quite inefficient.  That involves running 'forrest run', and then 'wget
-r http://localhost:8888'.

Btw, CVS Forrest includes a profiler which lets you isolate how much time
is taken by each step in the pipeline.



> Cheers,
> Jan

View raw message