Re: [xsl] ancestor::* and for-each

[Michael Kay <mike@xxxxxxxxxxxx>]

Just adding a little to what others have said on this.

This is undoubtedly one of the most confusing aspects of the XPath 
specification, and in my opinion it was not a good decision. The way the 
rules are expressed is different between XPath 1.0 and 2.0, because of 
the replacement of node-sets by sequences, but the effect is the same. 
Conceptually an AxisStep (of the form axis::nodetest[predicates]*) is 
evaluated in the following way:

1. All the nodes on the axis are selected, in axis order (that is, 
starting from the context node and moving outwards).

2. This sequence is filtered by retaining only the nodes that satisfy 
the nodetest

3. The sequence is then filtered using each of the predicates in turn. 
Within each predicate, position() refers to the position of a node in 
the sequence that results from the previous stage of filtering

4. The final filtered sequence is sorted into document order.

Apologies that "position" and "document" are trisyllabic, but I hope it 
explains what is going on.

Michael Kay
Saxonica


On 28/07/2010 17:11, Wolfgang Laun wrote:
> I use ancestor-or-self::* to obtain the node sequence from some node
> to the document root.
> The outermost/top element of the document isn't interesting, so I add
> a predicate, and I find that this works:
>     [position() != last()]
> Apparently, the sequence goes from leaf to root, which is the closing
> tag order, as it should be.
>
> Then, I iterate over the sequence, and now the nodes are processed
> from root to leaf, which is strange.
>
> There is a Note in Chapter 7, Repetion, that sounds as if it is
> related to this, but, frankly, I don't understand it.
> What does it mean, in monosyllabic words, please ;-)
>
> I'm using this<xsl:stylesheet version="2.0">, and  saxon9he.jar.
>
> Thanks
> -W