Andrew Welch wrote:

> the question I'm asking though is whether you need to
> workaround potentially expensive match patterns

  From what I've seen in the debugger with Saxon, is that the pattern's
predicate uses a QuantifiedExpression object.  So it *looks like* it is
actually looking at every ancestor::bar/@type, checking if some is
equal to 'A'.

  But honestly, I don't know Saxon's internals enough, maybe there are
other kinds of optimization I missed.  And that's becoming too
Saxon-specific.

  But I would be interested, just out of interest, to know what exact
processing is done under the hood for that pattern.

  Regards,

--drkm






















      _____________________________________________________________________________ 
Envoyez avec Yahoo! Mail. Une boite mail plus intelligente http://mail.yahoo.fr