gnome-doc-utils l10n stylesheets

[danilo gnome org (Danilo Šegan)]

Hi everybody (i.e. Shaun :),

I've finally started on Serbian localisation of g-d-u stylesheets.
(Yeah, I know you've asked for translators to really get on with this
way earlier, we are now in hard-code freeze and everything, but lets
improve l10n.xml even more now :)

And here are the things I may need.

First off, any list numbering can take also use Serbian alphabet
ordering: 

А Б В Г Д Ђ Е Ж З И Ј К Л Љ М Н Њ О П Р С Т Ћ У Ф Х Ц Ч Џ Ш
а б в г д ђ е ж з и ј к л љ м н њ о п р с т ћ у ф х ц ч џ ш

This is unusable for anything other than Serbian, since all other
Cyrillic languages miss at least some of these letters, and commonly
have some others.  This is probably not easy to add because I even if 
XSL has support for it, libxslt currently doesn't implement it, so you
can't simply do 

  <xsl:template name="db.digit.format.sr" >
    <xsl:param name="digit"/>
    <xsl:number format="а" lang="sr"/><!-- this is lowercase Cyrillic a -->
  </xsl:template>

  ...
  <xsl:when test="$format = 'sr-lower'">
    <xsl:call-template name="db.digit.format.sr">
      <xsl:with-param name="digit" select="$digit"/>
    </xsl:call-template>
  </xsl:when>

but you'd have to implement something like the following (Euclidean?
:) algorithm:

  <xsl:template name="db.digit.format.sr.lower" >
    <xsl:param name="digit"/>
    <xsl:choose> 
      <xsl:when test="$digit&gt;30">
 	<xsl:call-template name="db.digit.format.srk">
	  <xsl:with-param name="digit" select="floor($digit div 30)"/>
	</xsl:call-template>
	<xsl:call-template name="db.digit.format.srk">
	  <xsl:with-param name="digit" select="$digit mod 30"/>
	</xsl:call-template>
      </xsl:when>
      <xsl:when test="$digit&lt;31">
	<xsl:value-of select="substring('абвгдђежзијклљмнњопрстћуфхцчџш',$digit,1)"/>
      </xsl:when>
    </xsl:choose>
  </xsl:template>

(this one is for lowercase, and seems to work in my standalone tests :).

Is it ok to add something like this and introduce a new "format"
"sr-lower" and "sr-upper" that will call-template these ones as
appropriate?  We should probably even generalise this to any alphabet
anyway (i.e. instead of hard-coded substring alphabet, pass that up
too :).




Next, I wonder if we could use different roles for things like
declensions of labels (<xref> stuff), eg:
- "on figure 5" is "na *slici* 5" (locative)
- "from figure 5" is "sa *slike* 5" (genitive)
- "we blame figure 5" is "krivimo *sliku* 5" (accusative)
- "figure 5" is "*slika* 5" (nominative)
... (we have three more, like "with figure 5" [instrumental], "hey,
figure 5, talk to me" [vocative] and "figure 5 misses something"
[dative], but I don't want to bore you with examples; oh, I already
did :)


Now, imagine a following paragraph:

<para>On <xref linkend="figure-sample"/>, if you take a good look,
you'll see something. If we remove <xref linkend="figure-sample"/>,
you'll end up seeing nothing.</para>

would have to be translated to something like the following in
Serbian:

<para>On <xref linkend="figure-sample" xrefstyle="locative"/>, if you
take a good look, you'll see something. If we remove <xref
linkend="figure-sample" xrefstyle="accusative"/>, you'll end up seeing
nothing.</para>

and it should be rendered like:

<p>On <a href="#figure-sample">slici 5</a>, if you take a good look,
you'll see something. If we remove <a href="#figure-sample">sliku
5</a>, you'll end up seeing nothing.</p>

Note the "slici" and "sliku" renderings.  Do l10n stylesheets support
that already (I didn't look :), or do we need to work on that as well?


Cheers,
Danilo