Cyriaque Dupoirieux wrote:
> Ross Gardler a ?crit :
> >Cyriaque Dupoirieux wrote:
> >>
> >>On the paper it was simple, my problem is the horible performance I
> >>get to generate the skinned HTML.
> >>With the Cocoon Profile I get :
> >>1 file /src=/cocoon://livres/livres.html 3031 3031 3031
> >>2 xslt /src=/resources/stylesheets/php/document2php.xsl
> >>46 46 46
> >>
> >>to get a php file whereas it takes arround 100 to get a "normal"
> >>skinned HTML page...
> >
> >I don't understand how this can be the case since an internal request
> >for "cocoon://{1}{2}.html" is identical to a client request for
> >http://localhost/{1}/{2}.html
> >
> >Your output from the profiler above shows only 46ms for document2php.xsl.
>
> I know that my stylesheet - document2php.xsl - is fast enough, it has a
> single pattern to match...
> My problem is with the 3031ms ! to generate the HTML file before the
> <php> tag cleanning !
>
> >Are you sure you are comparing like with like?
>
> No ;-) ,
You will need to be very careful to read the Profiling
info correctly. I am still experimenting with the facility.
There is discussion about how to interpret the values
somewhere in the forrest-dev archives and some in the
howto-dev.html
The repeated numbers indicate that it is not being cached
(or when small, they might equate to cache-access time).
The large number for livres.html might be the total for
that pipeline, composed of other pieces. Can you follow
through links on that pipeline?
Also see discussion on those dev threads. In some situations
that time also includes time spent in other sitemap components.
Putting LogTransformers before and after that sitemap component
will help to isolate any "extra" time.
This should be addressed on the dev list, but are you
using the new "dispatcher" for this site. If so, then we
already know that there are issues. Please discuss on dev.
-David
