Re: GSXML

[Richard Frith-Macdonald]

On Monday, May 20, 2002, at 02:54 PM, Nicola Pero wrote:

> > I have to say that in that case I don't know what on earth you are
> > talking about.
>
> My pleasure - try the following code -
>
> #include <Foundation/Foundation.h>
> #include <Foundation/GSXML.h>
>
> /* Parse a NSString, returning the root node as parsed.  */
> GSXMLNode *rootNode (NSString *string)
> {
>   NSData *data = [string dataUsingEncoding: NSUTF8StringEncoding];
>   GSXMLParser *parser = [GSXMLParser parser];
>   if ([parser parse: data] == NO)
>     {
>       NSLog (@"Couldn't parse");
>       return nil;
>     }
>   return [[parser doc] root];

Incorrect code ... returning an object referring to non-retained libxml 
content.

> }
>
> int main (void)
> {
>   CREATE_AUTORELEASE_POOL (pool);
>
>   /* The string we want to parse.  */
>   NSString *string = @"<?xml version=\"1.0\"?><test></test>";
>
>   /* Parse the document and get the root node for later printing.  */
>   GSXMLNode *root = rootNode (string);
>
>   /* Retain it because we need it later.  */
>   RETAIN (root);
>
>   /* The GUI automatically recreates autorelease pools regularly,
>    * so let's simulate what happens in any running GUI program.  */
>   DESTROY (pool);

Here we destroy the libxml data.

>   pool = [NSAutoreleasePool new];
>
>   /* Now print out the details of the parsed root node.  */
>   NSLog (@"Content is :");
>   NSLog ([root content]);

Now we try to print the data we destroyed.

>   NSLog (@"Attributes are :");
>   NSLog ([[root propertiesAsDictionary] description]);
>
>   DESTROY (pool);
>
>   return 0;
> }
>
> If you have been able to use non-SAX GSXML for anything serious - you
> lucky.  I haven't.

But I suspect you are deliberately failing :-)

ie. From earlier discussions I'm pretty certain you know perfectly well 
how the
memory management in libxml works, and therefore know what your problem 
is, but
as you don't like wrapping of the libxml code and the OpenStep 
inconsistent way
it works, you choose not to accept that it does work.

A distaste for the memory management is a perfectly legitimate point of 
view -
I would prefer a more consistent system myself - and I may even get 
round to making it so.

What I think is wrong is that you claim that it does not work and/or is
unreliable when the real case is that it does not work the way you want 
it to,
and you simply have to remember that destroying the GSXMLParser destroys 
all
the underlying data structure - so the parser must be kept in existence 
as long
as you are working with it.

Fixing this memory 'problem' *is* trivial (you just need all GSXML 
objects to retain
the parser) without changing the parsing API as I said earlier.  Perhaps 
I'll spend
an hour on the train doing that, just to prove the point.

I'd recommend using the GSXML parser, not because it's perfect (it 
certainly
isn't), but because it's as good as or better than the alternatives!