W3C home > Mailing lists > Public > xproc-dev@w3.org > September 2011

Re: Calabash p:import and p:library could be slow?

From: Norman Walsh <ndw@nwalsh.com>
Date: Sat, 17 Sep 2011 17:47:35 +0100
To: XProc Dev <xproc-dev@w3.org>
Message-ID: <m2aaa3m7yw.fsf@nwalsh.com>
Alam Sher <alamsher83@gmail.com> writes:
> I'm writing some long in fact very long XProc pipelines/workflows.

Cool.

> However as I see the calabash code, I assume for every p:import if
> p:library is encountered, whole library is fetched/parsed and lets
> say if I'm writing a main pipeline that only contains two of the
> declared steps in p:library calabash will still parse the whole XML
> structure and read/load step bodies even if those are not going to
> be used in the main pipeline.

Yes, that is the case, but how could it be otherwise? If you load an
XML document, you have to, uh, load the XML document.

> Please correct me If I'm wrong. Also please suggest if there is any
> better strategy, or only the faster executing pipelines are going to
> be those which don't import huge library adapters?

My first question is, have you actually done any performance
measurements? Take one of your pipelines that uses two of the steps
From a big library and time it. Then make a new library that only
contains those two steps and time that. Is the difference really
significant?

If it is, can you divide your big library into several smaller ones,
and is that sufficient?

I suppose if it's really a concern, it would be important to do some
measurments of Calabash and see what takes time.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh
Lead Engineer
MarkLogic Corporation
Phone: +1 413 624 6676
www.marklogic.com

Received on Saturday, 17 September 2011 16:48:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 17 September 2011 16:48:09 GMT