How Do You Handle XML in Your Java Applications?

I am using SAX and DOM but also joost which is a fast and memory efficient 'xslt' transformator and see a list of xml processing/serializers where simple looks promising. Another nice software is jaxen.

One approach I implemented myself (similar to this) does not hold the entire xml in memory but can be parsed via DOM. E.g.you have the xml file

<products> 

   <product id="1">  CONTENT1 </product>

   <product id="2">  CONTENT2 </product>

   <product id="3">  CONTENT3 </product>

    ... 

</prodcuts> 

 Then you can parse it product by product via:

ArrayList list = ... 

ContentHandler productHandler = new GenericXDOMHandler("/products/product") {

            @Override
            public void writeDocument(String localName, Element element) throws Exception {

                // use stupid simple DOM here 

                list.add(item) 

            }

} 

(If someone wants the production ready source I could blog about this somewhen)

I've found myself using that type of approach too Peter...It would be great to see a blog/more detail.

James

JAXB if I need XML Schema validation.

XStream if I want something quick, lightweight and don't require XML Schema validation.

I second XMLBeans. I would rather work with Objects than XPath expressions, so XMLBeans fits my comfort level.

Mix of xpath, jaxb and string manipulation mostly

The choice is really divided in producing and consuming here. For producing xml, I prefer a template based solution although sadly I am stuck with jaxb currently. There's a reason why there is no serverside dom for html pages in most web servers: it's a waste of time to write the boilerplate pojo code and memory space. The same holds true for most non trivial xml documents in my view. Your db model classes tend to be different than your jaxb classes. So you end up with a lot of stupid conversion: request xml -> jaxb dtos ->db model->sql -> result set -> db model -> jaxb dtos -> response xml. I have  a lot of code and tests that are about babysitting data through this chain instead of actual business logic. 

For consuming xml, I believe in duck typing and jaxb tends to be too anal about little details for my taste. Xpath gives me all I need here and allows me to be much more relaxed about the content I can accept (which in most real life scenarios is a good thing).

 Dom and sax are too low level to be of use. If you want that level of detail, you really want to be using pull parsers like STAX probably. Sax and DOM are there for when you run out of other options, which in Java means you don't need to work with them directly.

When the files get big, you might want to consider hybrid approaches. One case where I had files that did not fit in memory (except with multi GB heap), I used a simple String.indexOf() to examine each line in order to consume chunks of xml that I could feed to Jaxb (convenient because the object model was already there). I actually multi threaded the whole thing to queue chunks and process them with 30 threads. BTW. If you ever find yourself writing this kind of code, take a good look at Spring Batch before you continue with your home grown batch processing system.

Alternatively, use a more batch friendly language like python. It does xml too.

Now I blogged about Memory Efficient XML Processing.

Comments are appreciated!