Trees vs Streams

XML documents are usually modelled as trees:

                    doc
                     |
          -------------------------
          |          |            |
        title     section      section
	  |          |            |
          |     ---------------
	  |     |       |     |
        text  title    para  para
                |       |     |
              text    text    ------------
                              |    |     |
                            text  bold text
                                   |
                                 text

Note that in the tree model text nodes (PCDATA or CDATA) are separate nodes and not attached to the enclosing element.

So there are 2 main models to process those documents: stream and tree

Other models are possible, such as Rax, a record-oriented model, or the XML::Twig model which mixes tree and stream.

Introduction to XML transformation

Stream Processing