Re: [xslt] Isn't it time to make libxslt multi-threaded?



Дмитрий Грибов, 26.07.2012 18:43:
>> Also note that libxslt has profiling support (see xsltproc --profile).
>> That might help with optimizing your stylesheets.
> 
> That's partially right (any code can be perfected and we do polish it) but
> we have just really, really heavy-logic-overloaded XSL. Any part of the
> page may have 10 or more views depending on user preferences, purchase
> history and so on. Even if we will profile it down to the ground our
> marketing department will make us mess it up back again in a couiple of
> weeks. So we need not only profiling but a broot force to let us make
> mistakes without killing the service.

That sounds like you should really split up the stylesheet into multiple
smaller pieces, each of which does a smaller task, such as selecting,
grouping or transforming data, and eventually styling it up to HTML or
whatever your final format is. Maybe also one in between that merges
multiple intermediate result documents into one - whatever your data flow
looks like.

Having multiple smaller steps would allow you to apply pipelining, and
potentially also to run single steps trivially in parallel against
independent data subsets.

Stefan




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]