Lists Home |
Date Index |
> Reminds me of long threads, elsewhere,... we could argue that node sets
> are not -sets- in the sense that they have an order: if I affect
> "preceding-sibling::*" to a variable "var", then I can ask for $var
> which would have no meaning if this was a -set- and the order used to
> evaluate $var is the forward order even though the node set has been
> constructed using an axis with a reverse order.
Yah, though setting to a variable is outside the scope of the XPath REC, no?
We (at jaxen.org) have had many discussions, particular regarding
//foo and our breadth-first walk of the tree instead of depth-first.
We sometimes end up with nodes out-of-document-order.
Technically, I think we're compliant, but it sure can annoy the user. ;)