Re: [xslt] key() in match pattern of xsl:key

[Daniel Veillard <veillard redhat com>]

On Sat, Dec 10, 2005 at 01:32:49PM -0500, Joel E. Denny wrote:
> On Sat, 10 Dec 2005, Joel E. Denny wrote:
> 
> > On Thu, 8 Dec 2005, Joel E. Denny wrote:
> > 
> > > > I recommend asking the people of the XSL list [1] about how the most
> > > > XSTL processors have implemented this. Would be good to go the de facto
> > > > way here if this is an 'implementation defined' issue.
> > > > 
> > > > [1] xsl-list lists mulberrytech com
> > > 
> > > Thanks for the suggestion. I will pursue this.
> > 
> > I have started a thread over there.  It has the same subject and is dated 
> > Dec 9.
> 
> Daniel, I have received a response from Michael Kay.  He claimed that both 
> forward and back references among keys are legal in XSLT 1.0 by virtue of 
> not being specifically banned.  On the other hand, he noted that he hasn't 
> seen it done very often.  I'm not sure how relevant this is to libxslt 
> right now, but he also pointed out that XSLT 2.0 discusses such references 
> explicitly.

  I think Saxon use an object model where evaluating an object initialize
its dependancies independantly of the order in which they appear. libxslt
initialize all types synchronously whithout trying to build a causal
dependancy chain.

> I found it encouraging that libxslt's forward key references should be 
> portable.  Although back references would offer some organizational 
> flexibility, I'm not sure they would offer any additional functionality, 
> so I'm happy without them.
> 
> I would like to see forward key references remain a permanent feature.  
> If you believe this is not right for libxslt, let me know and I will not 
> ask again.  Otherwise, how can I help?  If you want a regression test, do 
> you prefer a simple combination of XML, XSL, and expected output?  Or do 
> you prefer a cvs diff?

  If a diff is needed, send a diff, it will be reviewed. It would be nicer
if no diff were needed, ina any case, yes providing regression tests
would help garantee we won't break your expectations by mistake.

Daniel

-- 
Daniel Veillard      | Red Hat http://redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/