Thanks, Dimitre for comments.

I was also thinking to propose a simpler syntax, for passing context
to call-template (as follows, please):

<xsl:call-template name="xx" context="nodeReference">
   <!-- template contents -->
</xsl:call-template>

i.e, we can have an attribute on call-template, instead of wrapping it
within something like xsl:context (which looks much verbose).

On Sat, Dec 5, 2009 at 8:06 PM, Dimitre Novatchev <dnovatchev@xxxxxxxxx> wrote:
> Such a feature is not needed, especially in XSLT 2, where the use of
> <xsl:call-template> shouldn't be recommended -- why would anyone
> prefer it to the use of <xsl:function> ?

I think, xsl:function can be only called within XPath 2 expressions,
whereas, call-template can be part of a sequence constructor. IMHO,
both are useful.

> Even in XSLT 1.0 <xsl:call-template> is not necessary -- one can
> always accomplish the same with <xsl:apply-templates>

I largely agree. But I think, xsl:apply-templates is kind of push
processing (and tries to find nodes in the input document, and applies
the match template on them).

But xsl:call-template looks to me, truly a subroutine (but it being
able to inherit the context of caller, makes it different than a true
black box subroutine). It can take certain arguments, and can return a
RTF (in XSLT 1) or a sequence (in 2.0).

btw, I think having following enhancement to 2.0 language:

<xsl:call-template name="xx" inherit-context="no">
   <!-- template contents -->
</xsl:call-template>

(default value of, inherit-context can be "yes")

can make call-template instructions truly black box subroutine calls.
I think, this will be useful to users.

But I agree, that xsl:call-template and xsl:apply-templates can
archive similar objectives, in many cases.



-- 
Regards,
Mukul Gandhi