Re: [Xsltforms-support] "xps" template and my old xalan problem.

[Christian M. <chr...@mu...>]

Thanks! I am a happy XML geek now!

On Tue, Mar 9, 2010 at 11:01 PM, COUTHURES Alain <
ala...@ag...> wrote:

> Christian,
>
>  I think I have located the problem why xalan gets a stack overflow error
>> when transforming a xform with a lot of xpath expressions. As fas as I
>> understand the xps template gets a long string which contains all
>> concatenated xpath expressions to transform to javascript. the xps template
>> gets recursively called with the remaining rest of the string. The parameter
>> containing the string has to be created each recursion on the stack. and
>> eventually it gets to small or something else happens such that the
>> recursion doesn't finish correctly.
>>
>> I have tested that the occurrence of this problem depends only on the
>> number of xpath expressions not on the kind. There is a point on which the
>> transform fails if I add any attribute to the form.
>>
>> If I remove the recursion the transform finishes (giving an unusable
>> result). I do not understand the code enough yet to replace it with
>> something that would work for me. But I expect that this recursion could be
>> transformed into a for-each loop by using something like tokenize(). or
>> store the xpaths into a result tree fragment and using the node-set()
>> function.
>>
>> I really want this to work on existdb and xalan because then I do not have
>> to care about the performance any more as I would be able to serve a
>> precompiled html-javascript page instead of doing the very slow
>> xslttransform. If it takes a minute to redo the transform when I change the
>> xforms -- I think I can live with that. Saxon9 is unfortunately not saving
>> me because it fails too (and faster). Only saxon6.5 comes back with the
>> result but Saxon6.5 is not supported by existdb (as far as I know).
>> Thanks for all the help so far, and keep up the great work.
>>
> node-set() function can be used to avoid deep recursion. This function was
> not supported by FireFox 2.0 but it doesn't sound important anymore.
>
> This is committed now.
>
> Thank you for your feedbacks!
>
> -Alain
>