Re: comment on 'mode' and also Re: Processing modes

[James Clark <jjc@xxxxxxxxxx>]

Alistair MacDonald wrote:
> 
> >>> Lars Marius Garshol <larsga@xxxxxxxxxx> 04/15 8:19 am >>>
> > My understanding of processing modes is that
> >    <xsl:apply-templates modes="by-number"/>
> > should only apply template rules that have the by-number mode and
> > ignore all template rules that have no mode.
> 
> That should be 'mode', not 'modes'. The match sequence will be a template that
> HAS the required 'mode' and, if that fails a template with NO mode.

If you specify a mode on xsl:apply-template, then template rules in the
stylesheet with no mode are not considered: only the built-in rules are
applied.


> Why ? Because, as I said, modes aren't sticky and have to be
> *EXPLICITLY* passed on. However, <foo> was matched by the IMPLICIT
> rule, which didn't explicitly pass on the mode. This, compined
> with the fact that there is no way of find the *XSL* parameters
> to the triggering <xsl:apply-templates ... /> means that 'mode'
> is *almost* useless. (It *DOES* have uses, but trying to use it
> effectively means implementing lots of "do-nothing" templates
> which just call apply-templates with the same mode as they had!)

Why do you need lots of "do-nothing" template rules. You need one
"do-nothing" template rule:

<xsl:template match="*" mode="by-number">
  <xsl:apply-templates mode="by-number"/>
</xsl:template>

> There is no reason why 'mode' should not be sticky, since it
> could easily be cancelled ('mode=""') and, if no mode is matched
> it defaults to the empty mode.

There are two separate issues:

(a) whether modes should be sticky

(b) whether the built-in rule remains in the same mode or switches back
to the initial mode

The sticky versus non-sticky modes issue is very similar to the dynamic
versus lexical scoping issue.  Sticky modes like dynamic scoping is
seductive but a bad idea. With non-sticky modes you can tell just by
looking at the stylesheet which templates are applicable for a given use
of xsl:apply-templates.

On the built-in rule issue, I think there are arguments both ways. 
Whichever way you do it, you can get the other behaviour by adding a
single additional template rule for each mode.

James


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list