...
<xsl:variable name='auth' select='$biogs/key("biog", $name)'/>
...
</xsl:template>
That is, key() is used directly in the expression, and has its
context supplied by the previous expression step. Assuming I'm
understanding all this correctly, then I like this approach more than
the three arg key() call, because it seems to fit better with
typical XSLT usage patterns.
There is actually one use case where the 3-argument approach is much more
convenient, namely when the second argument is a node in a different
document from the one you want to search:
select="key('k', @code, document('codes.xml'))"
The above, however, didn't work in my XSLT 1.0 processor. No
surprise, I guess. No need for the for-each hack if you could do
that. But I've been looking around for a succint explaination of
what exactly changed in XSLT 2.0 to make this possible, and I haven't
been able to find it yet.
The change was actually in XPath 2.0, which now allows any expression on the
rhs of the "/" operator. XPath 1.0 restricted this to be an axis step.
So I guess my questions are:
* is it true that the above ($biogs/key( ...)) really is illegal in
XSLT 1.0?
Yes.
* what restriction was relaxed to make this possible in XSLT 2.0?
See above.
Michael Kay