On Wed, Oct 13, 2010 at 11:25 AM, Olivier Jeulin
<olivier.jeulin@xxxxxxxxx>  wrote:

Le 13/10/2010 19:50, Dimitre Novatchev a C)crit :

On Wed, Oct 13, 2010 at 8:23 AM, Scott Trenda<Scott.Trenda@xxxxxxxx>
  wrote:

New XSLT 2.1 feature? :)

Could be something as:

keys(keyname as xs:string) as item()*

and an additional overload:

keys(keyname as xs:string, document as document-node()) as item()*

Yes, it looks fine. But keys are atomized, so shouldn't keys() return
xs:anyType* instead of item()* ?

Did you mean xs:anyAtomicType* ? Yes, this is precise.