Cart
[Home] [By Thread] [By Date] [Recent Entries]
----- Original Message ----- From: "John Cowan" <cowan@c...> To: "Karl Waclawek" <karl@w...> Cc: <xml-dev@l...> Sent: Monday, April 12, 2004 7:19 PM > Karl Waclawek scripsit: > > > Actually, the SAX abstraction alone may be rather impractical. > > I would find it hard to map my proprietary object model directly > > to SAX without coming up with some Infoset-like abstraction first, > > even if it is not formalized. > > I would guess that that's because your object model is already tree-ish. Yes, that is likely the reason. > The TagSoup object model is a linear sequence of the following objects: > attribute names, attribute values, minimized attributes, entity references, > element names, end tags, processing instruction targets, processing > instruction bodies, close-of-start-tags, comments, and plain characters. > The TagSoup scanner delivers this object model from the input sequence of > characters using a fairly simple state machine. > > I would be hard put to it to explain *declaratively* (that is, without > using pseudo-code), how this model maps into the Infoset, yet the > result of the TagSoup parser is a sequence of SAX events which is mappable > onto the Infoset in a well-understood way. So, one can map from SAX to the Infoset and back (bijectively, I assume). Then it would be not too far fetched to say that the "SAX abstraction" is isomorphic to the Infoset, or at least some subset of it. Karl
|
