[xsl] Question about template priority

Subject: [xsl] Question about template priority
From: "Houghton,Andrew" <houghtoa@xxxxxxxx>
Date: Tue, 9 Feb 2010 11:39:59 -0500
I'm a little perplexed by a transform I have.  Basically the input file looks

<ns:foo />

The transform file looks like:

<!--(1)--><xsl:template match="node()[generate-id(.) =
generate-id(root(.)/*[1])]" />
<!--(2)--><xsl:template match="ns:foo" />
<!--(3)--><xsl:template match="/">
            <xsl:apply-templates select="*[1]" />

What template (1) is suppose to do is catch document elements that are not
ns:foo, e.g., someone gave the wrong XML document to the transform.  However,
template (1) is always invoked regardless of whether the correct XML input
document is given or not.

Looking at Michael Kay's book, it says that priority of a template is based

1. select all templates that have a match attribute; selected templates
2. select all templates that have the same mode as the apply-template;
selected templates (1,2,3)
3. select all templates whose pattern matches the node; selected templates
4. select the one with the highest import precedence; selected templates
5. select the one with the highest priority; selected templates (?)

Neither template (1 or 2) have a priority so in order to determine the
template with the highest priority it depends upon the match pattern syntax:

1. Pattern1 | Pattern2 = treat Pattern1 and Pattern2 separately and determine
their default priority
2. QName, @QName, child::QName, attribute::QName, processing-instruction,
(literal) = 0.0 priority
3. NCName:*, @NCName:*, child::NCName:*, attribute::NCName:* = -0.25 priority
4. NodeTest, @NodeTest, child::NodeTest, attribute::NodeTest = -0.5 priority
5. otherwise = 0.5 priority

Template (2) looks to me to be a QName and should therefore get a 0.0

Template (1) looks to me to be a NodeTest and should therefore get a -0.5

However, when I run Saxon, it always selects template (1).

So I added a priority="0" to template (1) and Saxon gives me XTRE0504
ambigious rule match indicating that template (2) priority is in fact zero.
Changing the priority for template (1) to -0.1 and template (2) is executed
which indicates that Saxon thought that the default priority pattern for
template (1) was "otherwise" and selected 0.5 for the priority which is why it
was executed without a priority attribute.

So why isn't template (1) considered a NodeTest?  When I lookup NodeTest in
Michael Kay's book in the examples it clearly shows:

  node()  This *NodeTest* selects all nodes on the relevant axis.

Inquiring minds would like to know, Andy.

Current Thread