--- Gavin Thomas Nicol <> wrote:
>
> On Apr 25, 2006, at 5:34 PM, Tatu Saloranta wrote:
>
> > But for other use cases -- for example,
> > recursive-descent traversal -- event mode is
> > ass-backwards.
>
> When do you really need to do that? I have an
> extension of
> DefaultHandler that manages a stack of seen elements
> so I can track
> nesting (calls like getPath() etc). For 99% of the
> stuff I do, that
> is enough.
Yes, but only if you actually enjoy having to
implement your stack handlers, dispatchers etc.
Pull-parsing removes requirement for such glue code;
keeping track of context can be done by normal method
calls, hence recursive-descent (but without a need for
an in-memory tree model like DOM).
It would certainly be possible to do something more
evolved on SAX foundation, yacc-style higher-level
parser generation.
Perhaps pipelined operations are one use case where it
would make sense.
> That depends on your purpose. If you're building a
> DOM tree, I'd
> agree. If you're just ripping data out of an XML and
> stuffing it into
> POJOs, I'd disagree. Even simple path-driver data
> extraction (prices
> form orders etc.) can be handled easily using SAX
> without the
> overhead of instantiating DOM trees for XPath
Sure. I am not using an in-memory model, just
switching the driver of parsing from listening to
events, to iterating over similar events.
Thing can be done using SAX; just as one's swiss
pocket knife can be used for screwing. Just not as
well as the real screwdriver.
> engines to operate on.
> Recursive-descent is a waste of time for
> data-binding too IMHO.
Building the tree is, agreed. I am not arguing for
having to build redundant models, but for using more
natural control pattern (iterating over structure vs.
listening to events resulting from someone iterating
over the said structure).
-+ Tatu +-
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
