Re: [xsl] Random number generation : requirements

["BR Chrisman brchrisman@xxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>]

I used exslt math:random() in a Monte Carlo simulation about a year ago.
 iirc, it doesn't change seed under libxslt on subsequent applications of a
stylesheet (I guess that's not surprising).
Here I was originally using a C harness to run a bunch of transformations
iteratively without serializing/deserializing a document.
I found that the call was always getting the same random sequence
(ostensibly it was getting the same seed).
I switched that one (of about ten) transformation to a random/evaluation
using C/libXML/XPath, which was oodles faster than applying that one
stylesheet in an otherwise tight loop.
I probably didn't even need to do that, as my simulation had a huge
performance problem in that my calculated time step was orders of magnitude
too small.
The XSLT implementation was *far* simpler than the C/libXML/XPath version.

Just my non-inflation-adjusted $0.02


On Tue, May 6, 2014 at 9:54 AM, Michael Kay mike@xxxxxxxxxxxx <
xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx> wrote:

> The W3C WGs are looking at the idea of introducing a random-number
> function of some kind in XPath 3.1. The challenge of course is making this
> both usable and a pure function with no side-effects. We have various
> design ideas which we need to test against requirements.
>
> If you have any applications that use or need such a function, please
> could I have a brief description of the way it uses random numbers, e.g.
> just wanting a single random number, a random permutation of 52 integers,
> an arbitrary sequence of random numebrs for test data generation, etc.
>
> If you're currently using the EXSLT random-sequence() function, please
> share your experience with it.
>
> Michael Kay
> Saxonica