Re: [xslt] Tail-call optimization for call-template possible?

[Nikolai Weibull <now bitwi se>]

On Mon, Jul 26, 2010 at 16:37, Nick Wellnhofer <wellnhofer aevum de> wrote:
> On 26/07/2010 15:58, Nikolai Weibull wrote:

>> I’m using XSL-T for the parsing escaped HTML inside XML.  Yes,
>> terrible, I know, but it’s the best I can do.
>>
>> Anyway, I’m currently running out of stack space when trying to parse
>> longer strings and was wondering if anyone had actually considered
>> “tail-call optimizing” call-template to avoid running out of stack.
>> How much work would it be?

> I once had a look at it and I'd consider it a lot of work. The libxslt code
> uses indirect recursion over three or four functions. It's not trivial to
> convert them all to tail calls. Alternatively, one could use trampolines.
> But that looks pretty complicated, too.

Yeah, I noticed that, too.  :-(

>> On my Cygwin installation the following script works, but changing the
>> test to 5175 causes a silent exit.

> I'd suggest to have a look at the EXSLT string functions. Maybe you can
> solve your problem with them without using recursive template calls.

No, one can’t, I’ve tried.  This really needs a “real” HTML lexer.
The reason I’m doing this is to be able to mark up the escaped HTML as
non-translatable content.  It works perfectly, except for the fact
that it depends so heavily on recursion.  One solution would be to use
for-each on each of the individual characters of the string (through
str:split($text, "")), but there’s no way to save the current state
information of the lexer (that is, are we currently processing an
element name, an attribute name, …).