[
Lists Home |
Date Index |
Thread Index
]
> > This just uses an inbuilt 'identity transform' stylesheet,
> > that roughly
> > outputs a copy of it's input. For this to happen it needs
> to build a
> > tree of the source, so the performance benefits of using SAX are
> > effectively lost at that point.
>
> You're making assumptions about the implementation.
>
> I don't know what Xalan does, but Saxon's implementation of
> the JAXP identity transform does not build a tree
> representation of the source document.
That's interesting.
I've often had the requirement to alter/add to the xml as it passed
through a set of SAXFilters. I've always been under the impression that
to write out the xml at the end of the filter chain I either needed to:
A) use DM's XMLWriter or similar
or
B) use a SAXTransformerFactory which allowed me to use setResult() to
give it a Writer.
Other than that I thought the SAX events were 'lost' by the handler,
which I've always considered odd. For me, the handler should have the
ability to write out the xml inbuilt into it, without the need for
XMLWriter or SAXTransformerFactory.
Reading here,
http://java.sun.com/j2se/1.4.2/docs/api/javax/xml/transform/sax/SAXTrans
formerFactory.html#newTransformerHandler()
you can see how I made my assumption. How does Saxon go from sax events
to a Result using an empty TransformerHandler()? Is there anything like
the performance costs of building a tree for the source and performing a
transform?
thanks
andrew
|
