W3C home > Mailing lists > Public > xproc-dev@w3.org > November 2012

Re: Embedded Calabash?

From: Florent Georges <fgeorges@fgeorges.org>
Date: Fri, 30 Nov 2012 12:57:58 +0000
Message-ID: <CADyR_r3rROGqYJwzHgbN+vKaEs9q6WCyjqmQGQgRYrBQJuQ_5g@mail.gmail.com>
To: Ichiro Furusato <ichiro.furusato@gmail.com>
Cc: xproc-dev@w3.org
On 30 November 2012 12:38, Ichiro Furusato wrote:
> On Fri, Nov 30, 2012 at 9:16 AM, Florent Georges wrote:

  Hi,

>>     http://code.google.com/p/servlex/source/browse/trunk/servlex/src/java/org/expath/servlex/components/XProcPipeline.java

>>   The biggest constraint I think to use Calabash on a server, is
>> that the compiled pipeline cannot be reused, so it cannot be cached
>> (compiled once and reused many times).  If I remember correctly
>> that is explained in some comment in the above file.

> For my application certainly caching would be nice (there's over 40
> steps) but knowing the difficulties in serialising Saxon stylesheets
> I think that if there were any stability or leak issues we'd prefer
> to take the performance hit.

  What I call caching here has nothing to do with serialization.  It
has to do with reusing the Java object representing the compiled
pipeline.  With Saxon for instance, for XSLT and XQuery, you can just
compile once and keep a reference to an XsltExecutable object (or for
XQuery an XQueryExecutable object), and get out of it an object
XsltTransformer (resp. XQueryEvaluator) for each actual transformation
(resp. query) using it.

  The problem is that you can't do that with Calabash's XPipeline,
which is both the compiled pipeline and its dynamic environment for
execution.  See http://xproc.markmail.org/thread/dhftopkqt6peofcm for
more details.

  Regards,

-- 
Florent Georges
http://fgeorges.org/
http://h2oconsulting.be/
Received on Friday, 30 November 2012 12:58:47 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 30 November 2012 12:58:48 GMT