Re: [Sax-devel] Should I use sax?

[Chris B. <d0...@hu...>]

David Megginson wrote:
> On 22/05/06, Michael Glavassevich <mrg...@ca...> wrote:
>> If you're writing new code, I would avoid using Xerces' native
>> serializers. They are non-standard, not particularly well =20
>> maintained and
>> likely to be deprecated [1] soon (perhaps in the next release). =20
>> JAXP's
>> TransformerHandler [2] is a much better choice for serializing SAX =20=

>> events.
>
> Speaking of new code, I should mention that another option is to use
> StAX instead of SAX. Like SAX, StAX is a streaming API, so in theory,
> it should be as efficient as SAX (I haven't run side-by-side tests).
> However, since StAX is a pull parser, it leaves you in control of the
> central loop instead of having to write event handlers.  Here's an
> intro:
>
> http://www.xml.com/pub/a/2003/09/17/stax.html
>
> StAX also covers writing.  Does anyone know how stable and efficient
> current StAX implementations are?

The GNU Classpath default XML parser implementation is entirely based =20=

on an underlying StAX parser. The SAX parser pulls events off this =20
StAX parser and delivers them as SAX events, and DOM trees are built =20
from this.

This parser implementation is used in nearly all free Java runtimes =20
including gcj, kaffe, jamvm, cacao etc, and gets a lot of exercise in =20=

free (gcj compiled-to-native) Eclipse, which makes extensive use of XML.

Regarding efficiency, the SAX-over-StAX parser, despite delegating to =20=

another implementation, takes approximately 48 seconds to parse a =20
typical set of documents that takes 60 seconds to parse with =C3=86lfred2.=
 =20
The same set of documents takes 42 seconds to parse with Xerces2's =20
SAX parser. I don't have straightforward comparison figures with =20
other StAX implementations like codehaus; I would suspect that it is =20
not as performant but is more conformant - attention to XML and SAX =20
conformance was the driving force behind this parser, and conformance =20=

results can be found at

   http://builder.classpath.org/xml/SAXTest/
--=20
=E7=8A=AC Chris Burdess
   "They that can give up essential liberty to obtain a little safety
   deserve neither liberty nor safety." - Benjamin Franklin