Re: [xsl] Preserving numeric character entity reference

[Abel Braaksma <abel.online@xxxxxxxxx>]

Kjetil Kjernsmo wrote:
> Ah, right. And it is even recommendation now... :-) I use libxslt 1.1.17 
> with mod_xslt, so whatever capabilities they have. I assume that means 
> XSLT 1.
>   

Ah, that's a pity. And from the remainder of your post, I understand 
that you want to translate these values hence and forth, using XSLT, 
right? This makes it very hard and an uncommon task to do with XSLT (so 
it may seem trivial at first sight). Mind you, the people behind the 
standards have learned a great deal from the problems of XSLT 1 and 
these, plus many other impossibilities are now part of the language in 
XSLT 2.

If there is no way that I can talk you into using XSLT 2 (i.e., by using 
the cross-platform java-based processor Saxon 8, which is free and also 
available as .NET version), then you are stuck with either a monstrosity 
of a template (ok, only about 92 or so characters need translation, it 
can be done) or with a simple enough extension function.

I am not sure if libxslt supports EXSLT and to what extend, but EXSLT 
has a replace() function, which, if supported, should come to the 
rescue. Please look around at http://www.exslt.org and test some 
functions with your processor (don't trust the support-statements on 
their site, they are out of date).

> Then I put a CDATA around just the character entities. And tried the 
> &, but how would I then go about to translate it back?
>   

See above. This is the hard part in XSLT 1, because a simple replace 
function for strings is not available, let alone a codepoints-to-string 
function, which would be the preferred way here. In XSLT 2 it looks like 
this for any string containing escapes of the literal form { (and 
in XML look like: &123;):

<xsl:value-of select="replace($string, '&amp;#(\d);', 
codepoints-to-string(number('$1')))" />

which is very far from what it will look like in plain XSLT 1: with 
recursive templates and some kind of lookup table.

> Or should I go with a template approach, where I go into the XHTML 
> fragment and do an value-of with disable output escaping?

disable-output-escaping is considered, among most people, as evil. Don't 
go that path. What is more, processors do not have to abide the d-o-e 
(especially when the result is non-conforming XML), so if you can do 
without it, stay without it.

One more thing. If you do decide to go the XSLT 2 way, then all you need 
to do is the obfuscating using xsl:character-maps. The XML you create is 
still legal XML (stay away from illegal entities!) and when read back 
in, all values will look as there normal strings (which is why this kind 
of obfuscation might not scare off people that much: it is too easy to 
read by any XML parser).

Good luck with your efforts,

Cheers,
-- Abel Braaksma