XOM API for Processing XML with Java

[Elliotte Harold <elharo AT metalab.unc.edu>]

Michael Kay wrote:


> That's a weird design. To count how many empty paragraphs there are, I
> evaluate the expression count(//para[not(node())]), catch the exception, and
> extract the number?
>
> Can you explain the thinking?

It's not meant to count the paragraphs at all. The way to do that is 
node.query("//p").size(). I'm not even willing to promise that what 
comes out of getReturnValue() for node.query("count(//p)") is a number.

This mostly helps distinguish between problems that arise due to XPath 
syntax errors and problems that arise because the XPath expressions go 
outside the bounds of what XOM supports.

Possibly I should add a third exception to account for bugs in the XPath 
engine, though hopefully I'll have most of those resolved or worked 
around before shipping.

Once I'd added this distinction, it was pretty straight-forward to add 
the additional method without cluttering the primary interfaces. The 
immediate impetus was simply to help me track down a few bugs in Jaxen.

However, it's still my opinion that the primary use case for XPath 
operating over XOM is to select node-sets. It is not to do simple 
arithmetic or string manipulation. Java code is far more effective and 
efficient at these tasks, and the interface reflects that. Remember, XOM 
tries to make it easy to do the right thing, and make it hard to 
impossible to do the wrong thing. using XPath to do arithmetic and 
string manipulation inside a Java program is the wrong thing.

--
Elliotte Rusty Harold  elharo AT metalab.unc.edu
XML in a Nutshell 3rd Edition Just Published!
http://www.cafeconleche.org/books/xian3/
http://www.amazon.com/exec/obidos/ISBN=0596007647/cafeaulaitA/ref=nosim