Re: [xsl] variable question

[Bruce D'Arcus <bdarcus@xxxxxxxxxxxxx>]

On Aug 29, 2004, at 4:32 PM, Jeni Tennison wrote:

> For XSLT 2.0, I think I'd turn the reftype template into a function
> instead, since it just returns a string. I'd keep the title-before
> template as a template since it might return some XML when you revise
> it in the future. So it would look more like:
>
> <xsl:template match="mods:titleInfo[not(@type='abbreviated')]"
>               mode="bibliography">
>   <span class="{if (../mods:relatedItem[@type = 'host'])
>                 then 'title'
>                 else 'title italic'}">
>     <xsl:apply-templates select="." mode="title-before"/>
>    	<xsl:apply-templates select="mods:title"/>
>    	<xsl:apply-templates select="mods:subTitle"/>
>    	<xsl:apply-templates select="." mode="title-after"/>
>   </span>
> </xsl:template>
>
> <xsl:template match="mods:titleInfo" mode="title-before">
>   <xsl:if test="mods:reftype(parent::mods:relatedItem) = 'chapter'">
>     <xsl:text>"</xsl:text>
>   </xsl:if>
> </xsl:template>
>
> <xsl:function name="mods:reftype" as="xs:string">
>   <xsl:param name="relatedItem" as="element(mods:relatedItem)" />
>   <xsl:choose>
>     <xsl:when test="$relatedItem/@type = 'host'">
>       <xsl:variable name="issuance" as="xs:string"
>         select="$relatedItem/mods:originInfo/mods:issuance" />
>       <xsl:choose>
>         <xsl:when test="$issuance = 'continuing'">article</xsl:when>
>         <xsl:when test="$issuance = 'monographic'">chapter</xsl:when>
>       </xsl:choose>
>     </xsl:when>
>     <xsl:otherwise>book</xsl:otherwise>
>   </xsl:choose>
> </xsl:function>

Interesting.  Thanks for this Jeni; very informative as always!

One of things I'm doing in this system that is new (and IMHO 
innovative) is trying to eliminate the typical typing system you see in 
existing bibliographic systems like bibtex.

My argument is that the recursive structure of the data is a nice match 
to the processing logic of XSLT, and that to force, for example, 
separate templates for every single minor variation of type (books, 
report, journal article, magazine article, blah, blah, blah) is both 
inefficient and error prone (though no doubt easier to code!).

So ... where the stylesheets currently are is there are three "types" 
which are in essence proxies for structural features.  They are the 
fallbacks.  That's what you see, and ultimately what the user will deal 
with (easier to think in terms of "book" rather than "record without 
relatedItem and with monographic issuance"!).

However, when I ultimately figure out how to add customizability, it 
will need to be possible for the user to specify additional "types" 
where the default's are for some reason insufficient.  So, in your 
function, the "otherwise" choice will probably be to look at the value 
of the <mods:genre> element.  So, style file has an element, say, 
<reftype name="memo">, and XSLT knows that for those records that have 
a genre value of the same name, it will use the formatting specified 
there to process the record.

I'm curious if you think that changes how I ought to approach 
constructing the function and related templates?    Or maybe I'm not 
thinking about this right to begin with?

Bruce