xml-dev - Re: XML API specification

[Derek Denny-Brown <ddb@criinc.com>]

At 10:14 AM 2/27/97 -0500, you wrote:
>I do not think our *parser* API should be the grove that results
>from parsing an XML document though. In my JAVA/C++ API, I do something
>like the following:
>
>   public class Property {
>      public String   m_name;
>      public Node     m_value;
>   }
>
>   public class Node {
>	public	Node         m_parent;
>	public  Node         m_left_sibling;
>	public  Node         m_right_sibling;
>	public  Node         m_first_child; 
>	public  Vector       m_properties;
>   }
>
>   public interface GroveEventHandler {
>	public OnStartNode(Node node);
>	public OnEndNode(Node node);
>   }

hmmm...

How does is a Property which is a NodeList represented?
What about a integer/string/boolean Property? 
Or even worse, and integer list or a string list!  

Unfortunately, at one point I spent over a week trying to come up with a
simple but powerful grove interface.  It just ain't that easy.

I think it would go more like...


public class Value {
	public String m_string;
	public int m_integer;
	public bool m_bool;
	public Vector m_list;
	public int m_type; // what kind of data does this value contain
}

public class Property {
	public String m_name;
	public Value  m_value;
}

public class Node {

 .... as listed ...
}


The other minor problem is that the child(ren) are defined by the value of
a property (which property is specified on the classdef).


I am not convinced that it is really worth implimenting groves this way
though.  Groves are designed such that it is often easier to duplicate
information, in a slightly different form, than to just reference it.
Sometimes you have to.  For example each element has it's GI as a string
valued property.  It also has the same information encoded as a reference
to the element type.  When I was working on a general Grove interface I
came to decide that it really should be a grove specific set of classes
which build on some more primitive interfaces (such as node,
property_value), and which impliment properties as methods.  This also
opened the door to such functionality as concurent construction (a property
value may be 'not-available-yet') which is necessary for an online
environment.

A the time I also had a SP Application which parsed property sets into some
C++ data structures, which I then used to generate teh stubs for the grove
interface.  The project never was really finished, and I have since moved
jobs... so I can't really go back an look at it, though I may be able to
get a copy of the SP app.

-derek
-- Remember: Computers are here to make our lives easier --
ddb@criinc.com // software-engineer // www/sgml/java/perl/etc.



xml-dev: A list for W3C XML Developers
Archived as: http://www.lists.ic.ac.uk/hypermail/xml-dev/
To unsubscribe, send to majordomo@ic.ac.uk the following message;
unsubscribe xml-dev
List coordinator, Henry Rzepa (rzepa@ic.ac.uk)