XSLT 2.0 and XPath 2.0 Programmer's Reference, 4th Edition (144 page)

Note that this is taking advantage of several new features of XSLT 2.0. The template uses

to return a reference to an existing node, rather than creating a copy of the node using

. It also declares the type of the parameters expected by the template, and the type of the result, which is useful documentation, and provides information that the XSLT processor can use for generating optimized code. I also found that while I was developing the stylesheet, many of my errors were trapped by the type checking. Note that the form
as=“element(SPEECH)”
can be used even when there is no schema. The example could have been rewritten to make much heavier use of XSLT 2.0 features; for example, it could have been written using

rather than

, and the

instruction could have been replaced by an XPath 2.0
if
expression. The result would have occupied fewer lines of code, but it would not necessarily have been any more readable or more efficient.

There is another solution to this problem that may be more appropriate depending on the circumstances. This involves sorting the node-set, and taking the first or last element. It goes like this:

In principle, the recursive solution should be faster, because it only looks at each node once, whereas sorting all the values requires more work than is strictly necessary to find the largest. In practice, though, it rather depends on how efficiently recursion is implemented in the particular processor.

Another case where recursion has traditionally been useful is processing of a list presented in the form of a string containing a list of tokens. In XSLT 2.0, most such problems can be tackled much more conveniently using the XPath 2.0
tokenize()
function, which breaks a string into a sequence by using regular expressions, or by using the

instruction described on page 230. But although these functions are excellent at breaking a string into a sequence of substrings, they don't by themselves provide any ability to process the resulting sequence in a nonlinear way. Sometimes recursion is still the best way of tackling such problems.