Ok, so this should be a pretty quick "whether XSLT can do this or not", it seems like it really should be able to, but I can't find any examples of it:

I have a stylesheet that looks like this:

<?xml version="1.0"?>
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform" version="1.0">

<xsl:template match="foo">
<b> foo's bold text </b>
</xsl:template>

<xsl:template match="b">
<i> <xsl:value-of select="."/> </i>
</xsl:template>

</xsl:stylesheet>


My input is this:

<foo/>
<b> bold text outside of foo </b>

My output is this:
<?xml version="1.0"?>
<b> foo's bold text </b>
<i> bold text outside of foo </i>

My desired output is this:
<?xml version="1.0"?>
<i> foo's bold text </i>
<i> bold text outside of foo </i>


Essentially what I want is for my rule for <b> tags to be applied in the definition of my rule for <foo> tags. I know I could do a second pass, but that seems like a hack. Is this possible to do in one pass? It seems like there should be some random trick to do this since it seems like a common operation.

just an FYI, I'm using libxml 20616, libxslt 10111 and libexslt 809 on a standard linux install

In XSLT 2.0 you write:

<xsl:template match="foo">
<xsl:variable name="temp">
  <b> foo's bold text </b>
</xsl:variable>
<xsl:apply-templates select="$temp/*"/>
</xsl:template>

Unfortunately XSLT 1.0 has a (crazy) restriction that you can't apply-templates to a "result tree fragment". Most processor have got around this by providing an extension function, xx:node-set() where the namespace assigned to prefix xx varies from processor to another. This allows you to write it as:

<xsl:template match="foo">
<xsl:variable name="temp">
  <b> foo's bold text </b>
</xsl:variable>
<xsl:apply-templates select="xx:node-set($temp/*)"/>
</xsl:template>


Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference

thanks Michael, that did the trick! took me a while to figure out that I had to add the exsl xmlns namespace for all this to work, but other than that, excellent and thorough response.

Does it really? What exactly are the restrictions on it?

The reason I ask is because I found a stylesheet yesterday that does precisely that. I made a small example to test it:


xml:

xslt:


output:
Needless to say, this intrigued me. A lot. (Some of us aren't fortunate enough to have access to an XSLT 2.0 processor at work yet. T_T) But this works in all the 1.0 processors I've tried, and I can't find more explicit information about it. The original stylesheet is only a few weeks old, but I suspect the concept has been around for a lot longer.

So... concerning result-tree-fragments vs. node-sets, I now assume the only differences are:
1) you can't apply xpath expressions when selecting what nodes to apply the templates to - it's a one-shot apply from the root node on downward
2) you have to make damn sure that your templates matching a root-node know exactly what they're doing, or otherwise it'll catch the variable's root. (i.e. always start with <xsl:template match="/document-root-node"> rather than <xsl:template match="/">, or always use a mode like so: <xsl:apply-templates select="$rtf-to-traverse" mode="rtf"/>)



In the basic example I've made, and even in the FizzBuzz example, it's not really worth arguing, seeing that one could simply use a key to achieve the same thing. However, when you can continue and apply this to variables consisting of complex node-sets, nested parameter calls, and so on with the freedom of XSLT 2.0 (only a lot more controlled and RTF-aware) - it sounds like it could open up a lot of doors for those of us still confined to XSLT 1.0.

Anything I'm mistaken about here? I'd like to know of any shortcomings in my logic before I start relying on this functionality in the future.

I don't see a result tree fragment anywhere in your stylesheet. Result tree fragments are created when you have an xsl:variable, xsl:with-param, or xsl:param element with child instructions.

Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference

Gotcha; I had known that variable/param/with-param content was treated as a RTF, but all of the documentation I've seen so far has been really dodgy on what the content is treated as if there's an XPath expression in the select attribute. So if I keep it constrained to variables and parameters that pass node-sets, I can legally perform the RTF-forbidden XPath operators on them?

If you use the select attribute to specify the contents of a variable, param or with-param then you'll have no problems. Unfortunately some content is too complex to create in this way and, in version 1.0 at least, you're forced to use the node-set extension function.

--

Joe (Microsoft MVP - XML)