RE: DOCBOOK-APPS: Sorting and non-en_US indexes

[David Cramer <dcramer at broadjump dot com>]

For the record here's the correct way to do what I was describing in a
previous post
<http://lists.oasis-open.org/archives/docbook-apps/200209/msg00283.html>
. The code in the previous post had some stupidities. For this to work,
you need to have the alphabet in the gentext language files
(common/en.xml etc) for your target languages (even when they lack
upper/lowercase distinction), tho it should fallback to English if
there's no key in the language specified. As I understand it, the
language is going to come from the value of the lang attribute on the
index's parent, correct (the value that comes from <xsl:call-template
name="l10n.language"/>  in common/l10n.xsl)? Should that same value be
used to set the lang attributes on the xsl:sorts in autoidx.xsl? 

<!-- Change the following entities in the internal subset of
autoidx.xsl-->
<!ENTITY lowercase "f:lowercase()">
<!ENTITY uppercase "f:uppercase()">

<!-- SNIP -->

<!-- add some stuff to the xsl:stylesheet element-->
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform";
                xmlns:fo="http://www.w3.org/1999/XSL/Format";
  xmlns:saxon="http://icl.com/saxon";
  xmlns:f="http://www.broadjump.com/TechPubs/functions"; 
  extension-element-prefixes="saxon"
  exclude-result-prefixes="f"
                version="1.0">

<!-- add these two elements to the top level of the autoidx.xsl
stylesheet -->
  <saxon:function name="f:uppercase">
	<xsl:variable name="uppercase">
	  <xsl:call-template name="gentext.element.name">
		<xsl:with-param
name="element.name">uppercase.alpha</xsl:with-param>
	  </xsl:call-template>
	</xsl:variable>
	<saxon:return select="normalize-space($uppercase)"/>
  </saxon:function>

  <saxon:function name="f:lowercase">
	<xsl:variable name="lowercase">
	<xsl:call-template name="gentext.element.name">
	  <xsl:with-param
name="element.name">lowercase.alpha</xsl:with-param>
	</xsl:call-template>
	</xsl:variable>
	  <saxon:return select="normalize-space($lowercase)"/>
  </saxon:function>

<!-- SNIP -->

As useful as they sound, I haven't used xml catalog files yet. Instead
of having several autoidx.xsls (which sounds unpleasant), could you have
one and get the alphabet with document('[level of indirection here,
resolved with catalog files to en.xml or target
lang.xml]')//l:gentext[@key = 'lowercase.alpha']/@text? But then you
lose the fallback mechanism to English if that language doesn't have
those keys. Or you could have a bunch of language.ent files and pull in
the right one as a parameter entity using catalog files. 

The <saxon:function>s are ignored by non-Saxon processors, correct? So
they could be included in the distribution and it could be documented
that you can use them with <!ENTITY lowercase "f:lowercase()"> <!ENTITY
uppercase "f:uppercase()">. Everything seems a little messy--not sure
what a good general solution is. I still have some work to do in pulling
everything together myself. 

David

> -----Original Message-----
> From: Jirka Kosek [mailto:jirka@kosek.cz]
<snip>
> I have one another idea how to make this processor idependat. You can
> create several versions of autoidx.xsl with different upper and
> lower-case entites and lang in xsl:sort. When invoking transformation
> you provide different XML catalag for each language to XSLT procesor.
> This catalog can redirect request for autoidx.xsl to proper localized
> version.
<snip>