Slow processing when Parsing XML files

Question

Hi
I have written a utility in Java (using JAXP) which parses an .xml file. This xml file
contains references to n number of other xml files. From here I direct the parser to each of these individual xml files where it searches for elements with a particular tag name, lets say 'node'. As in,
<node name= "Parent A">
<entry key="type" value="car"/>....
</node>
<node name= "Parent B">
<entry key="type" value="Truck"/>
</node>
I am collecting all the 'nodes' from these n xml files and then finally building a single xml file which will contain only the 'node' and its attribute 'name' value. As in,
<node name="Parent A"/>
<node name="Parent B"/>
In most cases n is a number greater than 100. Each of these n xml file have LOC exceeding 2 to 3000.
NOW the issue is, the xml parser takes more than an hr to go through just 10 - 15 xml files , collects the 'node', and building a new DOM object which I finally get published using Xml Transformer.
In effect , it beats the whole purpose of writing this utility as no programmer will stick around for an hr to to watch it happen/finish.
Apart from maybe further fine tuning my logic, which I've almost exhusted, is there any 'faster' way of doing the whole thing.
Please reply. Any comment would be greatly appreciated.
I am using JAXP 1.3 specs to parse and build the DOM.

Answer 1

DOM is slow.
Do it yourself using a simple SAX-parser. For all startElement - check if it is a "node", and then write your output!
Xerces is faster than the built in Crimoson SAX-parser in Java.
Parsing with Xerces a 2GHz machine takes manages about 5-6 MB/s of XML files.
Or use STX with Joost - although it's not THAT much faster
http://www.xml.com/pub/a/2003/02/26/stx.html
http://joost.sourceforge.net/
Gil