I may be about to commit crimes against CPAN

[Simon Wistow]

So, I've been pondering writing YA XML module. Actually, I hacked up the 
basis over an hour or so yesterday evening just to see if the idea 
worked.

Essentially you define a class per element you may receive so

http://code.google.com/apis/gdata/elements.html#gdComments

might be translated into 

package gdComments;
use strict;
use base qw(XML::Declare::Element);

sub name      { "comments" }
sub namespace { gd => 'http://schemas.google.com/g/2005' }

__PACKAGE__->attributes((
    rel => {
        type     => 'xs:string',
        enum     => [ map { "http://schemas.google.com/g/2005#$_" } 
                    qw(regular reviews) ],
        optional => 1,
    },
));

__PACKAGE__->elements((
    feedLink => {
        type     => 'gdFeedLink',
    },
));

1;


or similar. gdFeedlink is another module. Then you do

my $comments = gdComments->new;
$comments->rel("http://schemas.google.com/g/2005#regular");
my $feedlink = gdFeedLink->new;
$feedlink->countHint(1);
$feedlink->href("http://example.com");

print "Before:\n".$comments->to_string."\n\n";
$comments->feedLink($feedlink);
print "After:\n".$comments->to_string."\n\n";

attributes have types (string, uri, boolean, integer) and can have 
enums. I'm pondering changing things so that attributes and child 
elements don't get automatically generated accessors and instead you 
have to use 

	->set($name, $value) 

or similar instead. 

Parsing and stringification is done via factories. At the moment I have 
a ghetto hacked up thing that does it all with strings but 
writing a XML::SAX or a XML::LibXML Factory would be really easy. In 
fact, it occurs to me that writing a YAML and/or JSON Factory would be 
easy too. Maybe I should call it Data::Declare instead.

So, anyway - it was fun to hack it up to this state and see if the proof 
of concept works but I can't decide whether to just drop it now or if 
it's at all useful.

Thoughts and comments gratefully received,

Simon