RE: Context node traversal inside predicates (was: Nod

> The 'context()' function would be similar to the 'current()' fuction.
> It would recall previous context nodes.
> 
> 'context(n)' would recall the n-th context node (as found 
> when propagating  
> upwards in the predicates of the expression)
> 
> e.g.:
> 'context(0)' is equal to '.'
> 'context(1)' is equal to the context node as constructed by the  
> immmediately outer predicate.
> ...
> 'context(-1)' would be equal to the context node as found in 
> the outermost  
> predicate.

Something like this was in an early draft of XPath 1.0, and was implemented
(IIRC) in WD-xsl, though I've rarely seen it used in practice.

I think the idea of manually addressing the stack of context variables is
very error-prone and counter-intuitive. The proper way to do this is to
declare a range variable. In XQuery you can do this using the "let" clause
of the FLWOR expression; XPath 2.0 has a poor man's version in the "for"
expression, which allows:

$SetA/Element[for $Z in . return $setB/Element[substring-after(.,'.')=$Z]]

In fact you might want to turn the whole thing into an XQuery-like join:

for $Z in $SetA/Element,
    $Y in $setB/Element
return
   if (substring-after($Y, '.') = $Z)
   then $Y else ()

Michael Kay
http://www.saxonica.com/