It should do whatever a "foo" element would do in E.xsl by itself, which may or may not be nothing. It could be the default template rule. That is how I would interpret the spec anyway. I am pretty sure it should not invoke the template from B.xsl.
Chris Morten Jorgensen wrote: >I have spent some time on cleaning up XSLTC's implementation of >the xsl:apply-imports element and I have come across a scenario >that seems to be handled differently by the various XSLT processors. >I'd like some help with determing what the correct behaviour for >XSLTC is. Consider that you have these stylesheets: > > A.xsl imports B.xsl, which imports C.xsl > A.xsl imports D.xsl, which imports E.xsl > >Stylesheets A, B, C and D all have a template that matches on >an element "foo" - stylesheet E does _not_. This matching >template looks like this (in all of the first 4 stylesheets): > > <xsl:template match="foo"> > <A><xsl:apply-imports/></A> > </xsl:template> > >The xsl:apply-imports elements is similar to a call to "super()" >in Java, so the template in A.xsl triggers the template in D.xsl >(because this is the imported template with the highest import >precedence). There are no matching templates in E.xsl, so what >should the template in D.xsl do? > > A) Nothing? > B) Trigger the matching template in B.xsl? > >Saxon goes with option A), while Xalan (and currently XSLTC) goes >with option B). I am tempted to say that Saxon is correct, as >template D.xsl does not "inherit" any behaviour from template B.xsl. > >Any input welcome! > >Morten Jørgensen, >XML Technology Centre, >Sun Microsystems Ireland ltd. >
