Re: [docbook-apps] processing modular docbook

[Stefan Kost <ensonic@hora-obscura.de>]

Am 09.04.2010 19:40, schrieb Tim Arnold:
> Hi Bob,
> I'm using a python library (lxml) that wraps the libxml2.6.30 and libxslt1.1.22 libraries. By using the library, I do nearly everything in memory, without having to write the full book.xml and olinkdb.xml; the only time something has to write to disk is when the actual chunking transform takes place.
> 
> However, I do get the same speeds with that as I do with xsltproc on the same machine:
> 
> xsltproc --version
> Using libxml 20632, libxslt 10124 and libexslt 813
> xsltproc was compiled against libxml 20632, libxslt 10124 and libexslt 813
> libxslt 10124 was compiled against libxml 20632
> libexslt 813 was compiled against libxml 20632
> 
> thanks,
> --Tim
> 
>> -----Original Message-----
>> From: Bob Stayton [mailto:bobs@sagehill.net]
>> Sent: Friday, April 09, 2010 12:29 PM
>> To: Tim Arnold; docbook-apps@lists.oasis-open.org
>> Subject: Re: [docbook-apps] processing modular docbook
>>
>> Did you mention what XSLT processor you are using?
>>
>> Bob Stayton
>> Sagehill Enterprises
>> bobs@sagehill.net
>>
>>
>> ----- Original Message -----
>> From: "Tim Arnold" <Tim.Arnold@sas.com>
>> To: <docbook-apps@lists.oasis-open.org>
>> Sent: Friday, April 09, 2010 9:16 AM
>> Subject: RE: [docbook-apps] processing modular docbook
>>
>>
>> hi,
>> I just wanted to reply with the timings I found. For the book that contains
>> about
>> 3,000 pages I used the 1.75.1 namespace aware DocBook XSL stylesheets (xhmtl
>> chunking
>> templates) with
>> 1) no customization
>> 2) my customizations to create the xhtml,
>> 3) my customizations with some additional files created (toc.xml, index.xml,
>> *.hh*
>> files).
>>
>> 1) Straight XHTML: 44 minutes
>> 2) Customized:     55 minutes
>> 3) Custom plus add'l files: 59 minutes
>>
>> My customizations change the chunking/filenaming scheme and puts index
>> primary terms
>> as keywords in each file.
>>
>> For a much smaller book (300 pages) I get:
>> 1) 37 seconds
>> 2) 30 seconds*
>> 3) 43 seconds
>>
>> * I have no idea why my customization is faster here and slower on the
>> bigger book.
>>
>>
>> Do these times sound about right? If so, I guess it's going to take over a
>> couple of
>> hours to do the big 8,000 page book.
>>
>> thanks,
>> --Tim

Tim,

if building libxslt yourself for a test would be okay for you, if would be
interesting to see what you get from my profiling patch.
- clone libxslt git head
  git clone git://git.gnome.org/libxslt
- apply the patch from https://bugzilla.gnome.org/show_bug.cgi?id=612186
  git am 0001-profiling-add-callgraph-report.patch
- build e.g. to $HOME/debug and export LD_LIBRARY_PATH
  ./autogen.sh --prefix=$HOME/debug
- run xsltproc --profile ...
- send us/me the profile

btw. you are saying, you have everything in memory - real memory or is the
system swapping like mad?

Stefan


>>
>>> -----Original Message-----
>>> From: Stefan Kost [mailto:ensonic@hora-obscura.de]
>>> Sent: Tuesday, April 06, 2010 2:44 PM
>>> To: Tim Arnold
>>> Cc: docbook-apps@lists.oasis-open.org
>>> Subject: Re: [docbook-apps] processing modular docbook
>>>
>>> hi Tim,
>>> Am 06.04.2010 21:26, schrieb Tim Arnold:
>>>> Hi,
>>>> I have a document that when printed is about 8000 pages. Because of the
>>> time it takes to process, I'd like to be able to process each chapter
>>> separately, but the more I read, the more I doubt the possibility.
>>>>
>>>> My guess is that if I decide to try it, I'll have to set the numbering
>> for
>>> chapters/tables/examples/figures in each chapter and set up a granular
>> olink
>>> database to handle the cross-refs. And  create the book-level toc, index,
>>> any HTMLHelp files by some other means.
>>>>
>>>> I haven't yet tried processing this particular document, but a 3,000
>> page
>>> doc takes about 2 hours on  my FreeBSD machine. The processing uses only
>> one
>>> of its cpus of course, and neither memory or diskspace is a limitation--
>> the
>>> processing seems to be cpu-bound only.
>>>>
>>>> Any advice ?
>>>>
>>>> thanks,
>>>> --Tim Arnold
>>>>
>>>
>>> one thing that helped me to save a lot of time in bigger api doc project
>> in
>>> gtk-doc is to override the l10n templates in my customization layer (I
>> could
>>> do
>>> it as the docs are only in one language (english)).
>>>
>>> Stefan
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: docbook-apps-unsubscribe@lists.oasis-open.org
>> For additional commands, e-mail: docbook-apps-help@lists.oasis-open.org
>>
>>
>>
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: docbook-apps-unsubscribe@lists.oasis-open.org
> For additional commands, e-mail: docbook-apps-help@lists.oasis-open.org
>