xml-dev - Re: (more) extensible SAX

[Ken MacLeod <ken@bitsko.slc.ut.us>]

Eric van der Vlist <vdv@dyomedea.com> writes:

> Ken MacLeod wrote:

> > Coincidentally, this is one of the major feature differences with
> > Perl SAX (being descussed in the "SAX Comments" thread).  Perl SAX
> > has always used a node as the argument to SAX events.  In our
> > case, though, it's not a "start element" or "end element" object,
> > but just an element node, a DOM node, to be exact.
> > 
> > For SAX2's namespaces, all that needed adding was Prefix,
> > LocalName, and NamespaceURI properties to the nodes.
> 
> That's interesting (and proves me that I shouldn't be that Java
> biased) especially after the comment from Elliotte Rusty Harold
> about the added complexity of such an architecture.
> 
> What's your feeling after having implemented and used it ?

It's either the difference between strongly and weakly typed languages
or an as-yet-unrecognized difference in design pattern, but the
marriage of SAX and DOM in Perl, using Perl style APIs, results in
fewer classes/interfaces than in just SAX alone in Java.  The same
patterns work very well in Python and Tcl, and even C, but are having
troubles being ported back to Java/C++.

As we move more code into this framework, we've found the resulting
code to be quite a bit simpler overall.

  -- Ken