From: Morbus Iff [mailto:morbus@[...].com]
>   >SAX is a huge advance over the XML::Parser Handler API for a
> =20
>  Thanks for the expose. I'll wax a bit.
> =20
>   > - pluggable - if your code is written to the SAX API you
>   >   can use any SAX parser without changing your code
> =20
>  Not immediately useful to me, since expat is the only library=20
>  currently=20
>  ported (and bundle-able) to every OS I need it to be in (my=20
>  software is one=20
>  of the rare few that turns into a "don't need Perl installed"=20
>  application=20
>  for Mac and Windows).

Note also that XML::SAX comes with an extremely portable parser
written entirely in Perl (XML::SAX::PurePerl).  Unfortunately,
it needs Perl 5.8 to support encodings other than UTF8

>   > - flexible - your data source does not even need to be an
>   >   XML document (eg: you can drive your SAX pipeline from
>   >   a database query
> =20
>  That's kinda neat, although I do some pre-processing before=20
>  sending to=20
>  XML::Simple - enough so that I'm always sending a string of=20
>  XML, not a file.
> =20
>   >I'm not entirely clear on what you're trying to do with
>   >namespaces.  Do you want your hashref keys to be in Clarkian
>   >notation eg: '{http://purl.org/dc/elements/1.1/}date' or
>   >do you want to normalise the prefixes used eg: 'dc:date'?
> =20
>  Nope, not really. The biggest problem is:
> =20
>    - I assume my data is going to be in one data structure, but
>      if someone prefixes the data with a namespace besides the
>      implied default, I get a different structure that breaks
>      my assumption:
> =20
>       assuming:
>        <item><dc:title>boo</dc:title></item>
>        =3D=3D $item->{dc:title}
> =20
>       breaks my thingy:
>        <item><dublincore:title>boo</dublincore:title></item>
>        !=3D $item->{dc:title} but rather $item->{dublinecore:title}

That does look like the normalisation option I referred to.
So if you got a document like this:

<rdf:RDF
 xmlns=3D"http://purl.org/rss/1.0/"
 xmlns:rdf=3D"http://www.w3.org/1999/02/22-rdf-syntax-ns#"
 xmlns:theonetruedublincore=3D"http://purl.org/dc/elements/1.1/" > 
  <theonetruedublincore:date> 2002-10-08</theonetruedublincore:date>
</rdf:RDF> 

then you want to treat it as if it were:

<rdf:RDF
 xmlns=3D"http://purl.org/rss/1.0/"
 xmlns:rdf=3D"http://www.w3.org/1999/02/22-rdf-syntax-ns#"
 xmlns:dc=3D"http://purl.org/dc/elements/1.1/" > 
  <dc:date> 2002-10-08</dc:date>
</rdf:RDF> 

and slurp it into a hash like this:

 {
   'xmlns'     =3D>  'http://purl.org/rss/1.0/',
   'xmlns:rdf' =3D>  'http://www.w3.org/1999/02/22-rdf-syntax-ns#',
   'xmlns:dc'  =3D>  'http://purl.org/dc/elements/1.1/',
   'dc:date'   =3D>  '2002-10-08'
 };

The way I'd see that working with SAX is something like:

  use XML::SAX::Machines qw( :all );
  use XML::Filter::NSNormalise;
  use XML::Simple;

  my $p =3D Pipeline(
    XML::Filter::NSNormalise-> new(
      map =3D>  {=20
        'http://purl.org/dc/elements/1.1/' =3D>  'dc',
        'http://purl.org/rss/1.0/modules/syndication/' =3D>  'syn'
      }
    )
    =3D>  XML::Simple->new(
      keyattr =3D>  {}
    )
  );

  my $ref =3D $p-> parse_uri('./rss.xml');

An off-the-cuff version of XML::Filter::NSNormalise is attached.

Cheers
Grant