[jdom-interest] jdom 1.1

[Jason Hunter]

With this test code at http://jira.activemath.org/browse/OM-12 (based  
on the original report) I don't see any difference between JDOM 1.0  
and JDOM 1.1.  After upping the memory given to Java, the load of the  
25M file happens in 6 seconds under both versions.  The full walk  
takes 84 additional seconds, same on both versions.  This is with  
"1.5.0_13" on a modern MacBook Pro.

The "walking code" is slow at 84s but that's due to the test program's  
code not being optimized:

         List bookList=booksElement.getChildren("book");
         int numberOfBooks=bookList.size();
         for(int i=0;i<numberOfBooks;i++){
             Element bookElement=(Element)bookList.get(i);
             readPriceList(bookElement.getChild("price"));
         }

It would be much, MUCH faster for user code to iterate over the books  
directly rather than looking them up each time by index.  Each numeric  
index lookup requires a child-walk to find the item.  It's an n- 
squared algorithm.  That's not a big deal with 5 items but with 50,000  
items as with this test file, that's going to be slow.  It's the same  
on JDOM 1.0 and JDOM 1.1 both for me.

I rewrote the example to fix two main iterator loops and the full walk  
now takes 0.1 seconds.

-jh-

P.S.  The test code also uses == to compare Strings, which is a  
potential unrelated problem:

         List children=element.getChildren();
         if(children !=null){
             Iterator iterator=children.iterator();
             while (iterator.hasNext()) {
                 Element child = (Element) iterator.next();
                 if(child.getName()=="date_created"){
                     readDate(child);
                 }else if(child.getName()=="date_modified"){
                     readDate(child);



On Sep 3, 2008, at 2:20 AM, katja.radelicki at colruyt.be wrote:

>
> Hello,
> I simplified my code and tested it again.
> You can find it here: http://jira.activemath.org/browse/OM-12
> It seems walking trough the JDOMTree generates the problem and not  
> creating the JDOMTree.
> You can can find two programs:
>
> 1.JDomReaderSaxCreateJdomTree.java
> this program only reads the xml-document and creates the JDOMTree.  
> Here seems to be no problem!
>
> 2.JDomReaderSax.java
> this program reads the xml-document,creates the JDOMTree and walks  
> trough the jdomtree. I suppose walking trough the tree generates the  
> problem in time.
>
> There are also 3 example xml files.
>
> I hope you can test the code. If there is a problem, just tell me.
>
> Thank you very much.
>
>
> ----- Original Message ---------------------------------
>
> Can someone tell me why jdom 1.1 is slower than jdom 1.0? Which  
> changes are made that causes this?
>
> And why is jdom 1.1 is more memory intensive than jdom 1.0?
>
> Thanks a lot!