package XML::Descent;

use warnings;
use strict;
use Carp;
use XML::TokeParser;

=head1 NAME

XML::Descent - Recursive descent XML parsing

=head1 VERSION

This document describes XML::Descent version 1.04

=head1 SYNOPSIS

  use XML::Descent;

  # Create parser
  my $p = XML::Descent->new( { Input => \$xml } );

  # Setup handlers
  $p->on(
    folder => sub {
      my ( $elem, $attr ) = @_;

      $p->on(
        url => sub {
          my ( $elem, $attr ) = @_;
          my $link = {
            name => $attr->{name},
            url  => $p->text
          };
        }
      );

      my $folder = $p->walk;
      $folder->{name} = $attr->{name};
    }
  );

  # Parse
  my $res = $p->walk;

=head1 DESCRIPTION

The conventional models for parsing XML are either DOM (a data structure
representing the entire document tree is created) or SAX (callbacks are
issued for each element in the XML).

XML grammar is recursive - so it's nice to be able to write recursive
parsers for it. XML::Descent allows such parsers to be created.

Typically a new XML::Descent is created and handlers are defined for
elements we're interested in

  my $p = XML::Descent->new( { Input => \$xml } );
  $p->on(
    link => sub {
      my ( $elem, $attr ) = @_;
      print "Found link: ", $attr->{url}, "\n";
      $p->walk;    # recurse
    }
  );
  $p->walk;        # parse

A handler provides a convenient lexical scope that lasts until the
closing tag of the element that triggered the handler is reached.

When called at the top level the parsing methods walk, text and
xml parse the whole XML document. When called recursively within a
handler they parse the portion of the document nested inside node that
triggered the handler.

New handlers may be defined within a handler and their scope will be
limited to the XML inside the node that triggered the handler.

=cut

our $VERSION = '1.04';

=head1 INTERFACE 

=head2 C<new( { options } )>

Create a new XML::Descent. Options are supplied has a hash reference.
The only option recognised directly by XML::Descent is C<Input> which
should be reference to the object that provides the XML source. Any
value that can be passed as the first argument to 
C<< XML::TokeParser->new >> is allowed.

The remaining options are passed directly to C<XML::TokeParser>. Consult
that module's documentation for more details.

=cut

sub new {
  my $class = shift;

  my %args = ();
  my @opt  = ();
  for my $arg ( @_ ) {
    if ( 'HASH' eq ref $arg ) {
      %args = ( %args, %$arg );
    }
    else {
      push @opt, $arg;
    }
  }
  croak "Expected a number of name => value pairs"
   if @opt % 2;
  %args = ( %args, @opt );

  my $parser
   = XML::TokeParser->new( delete $args{Input}
     || croak( "No Input arg" ), %args )
   || croak( "Failed to create XML::TokeParser" );

  return bless {
    parser  => $parser,
    context => {
      parent => undef,
      rules  => {},
      obj    => undef
    },
    token => undef,
    path  => [],

  }, $class;
}

sub _get_rule_handler {
  my ( $self, $tos, $elem ) = @_;
  croak "It is not possible to register an explicit handler for '*'"
   if '*' eq $elem;
  while ( $tos ) {
    if ( my $h = $tos->{rules}{$elem} || $tos->{rules}{'*'} ) {
      return $h;
    }
    $tos = $tos->{parent};
  }
  return;
}

sub _depth { scalar @{ shift->{path} } }

=head2 C<walk>

Parse part of the XML document tree triggering any handlers that
correspond with elements it contains. When called recursively within a
handler C<walk> visits all the elements below the element that triggered
the handler and then returns.

=cut

sub walk {
  my $self = shift;

  TOKEN: while ( my $tok = $self->get_token ) {
    if ( $tok->[0] eq 'S' ) {
      my $tos = $self->{context};
      my $handler = $self->_get_rule_handler( $tos, $tok->[1] );
      if ( defined $handler ) {
        my $stopat = $self->_depth;

        # Push context
        $self->{context} = {
          parent => $tos,
          stopat => $stopat,
          obj    => $tos->{obj}
        };

        # Call handler
        $handler->( $tok->[1], $tok->[2], $tos->{obj} );

        # If handler didn't recursively parse the content of
        # this node we need to discard it.
        1 while $self->_depth >= $stopat
           && ( $tok = $self->get_token );

        # Pop context
        $self->{context} = $tos;
      }
      else {
        $self->walk;
      }
    }
    elsif ( $tok->[0] eq 'E' ) {
      last TOKEN;
    }
  }
}

=head2 C<on( [ element names ], handler )>

Register a handler to be called when the named element is encountered.
Multiple element names may be supplied as an array reference. Multiple
handlers may be registered with one call to C<on> by supplying a number
of element, handler pairs.

Calling C<on> within a handler defines a nested local handler whose
scope is limited to the containing element. Handlers are called with
three arguments: the name of the element that triggered the handler, a
hash of the element's attributes and a user defined context value - see
C<context> for more about that.

For example:

  $p = XML::Descent->new( { Input => \$some_xml } );

  # Global handler - trigger anywhere an <options> tag is found
  $p->on(
    options => sub {
      my ( $elem, $attr, $ctx ) = @_;

      # Define a nested handler for <name> elements that only
      # applies within the <options> handler.
      $p->on(
        name => sub {
          my ( $elem, $attr, $ctx ) = @_;
          # Get the inner text of the name element
          my $name = $p->text;
          print "Name: $name\n";
        }
      );

      # Recursively walk elements inside <options> triggering
      # any handlers
      $p->walk;
    }
  );

  # Start parsing
  $p->walk;

A handler may call one of the parsing methods (C<walk>, C<text>, C<xml>
or C<get_token>) to consume any nested XML before returning. If none of
the parsing methods are called nested XML is automatically discarded so
that the parser can properly move past the current element.

Nested handlers temporarily override another handler with the same name.
A handler named '*' will trigger for all elements for which there is no
explicit handler. A nested '*' handler hides all handlers defined in
containing scopes.

As a shorthand you may specify a path to a nested element:

  $p->on( 'a/b/c' => sub {
    print "Woo!\n";
  })->walk;

That's equivalent to:

  $p->on( a => sub {
    $p->on( b => sub {
      $p->on( c => sub {
        print "Woo!\n";
      })->walk;
    })->walk;
  })->walk;

Note that this shorthand only applies to C<on> - not to other methods
that accept element names.

=cut

sub on {
  my $self = shift;
  croak "Please supply a number of path => handler pairs"
   if @_ % 2;

  while ( my ( $spec, $cb ) = splice @_, 0, 2 ) {
    $spec = [$spec] unless ref $spec eq 'ARRAY';
    for my $el ( @$spec ) {
      my ( $name, $tail ) = split /\//, $el, 2;
      if ( defined $tail ) {
        $self->{context}{rules}{$name} = sub {
          $self->on( $tail => $cb )->walk;
        };
      }
      else {
        $self->{context}{rules}{$el} = $cb;
      }
    }
  }
  return $self;
}

=head2 C<inherit( [ element names ] )>

Inherit handlers from the containing scope. Typically used to import
handlers that would otherwise be masked by a catch all '*' handler.

  $p->on(
    'a' => sub {
      my ( $elem, $attr, $ctx ) = @_;
      my $link = $attr->{href} || '';
      my $text = $p->text;
      print "Link: $text ($link)\n";
    }
  );

  $p->on(
    'special' => sub {
      my ( $elem, $attr, $ctx ) = @_;

      # Within <special> we want to handle all
      # tags apart from <a> by printing them out
      $p->on(
        '*' => sub {
          my ( $elem, $attr, $ctx ) = @_;
          print "Found: $elem\n";
        }
      );

      # Get the handler for <a> from our containing
      # scope.
      $p->inherit( 'a' );
      $p->walk;
    }
  );

The inherited handler is the handler that would have applied in the
containing scope for an element with the given name. For example:

  $p->on( '*' => sub { print "Whatever\n"; $p->walk; } );
  $p->on(
    'interesting' => sub {
      # Inherits the default 'Whatever' handler because that's the
      # handler that would have been called for <frob> in the
      # containing scope
      $p->inherit( 'frob' );
      # Handle everything else ourselves
      #p->on('*', sub { $p->walk; });
    }
  );

=cut

sub inherit {
  my $self = shift;
  my ( $path ) = @_;

  $path = [$path] unless ref $path eq 'ARRAY';
  my $par = $self->{context}{parent};
  $self->on( $_, $self->_get_rule_handler( $par, $_ ) ) for @$path;
  return $self;
}

sub _filter {
  my ( $self, $mk_wrapper ) = splice @_, 0, 2;
  croak "Please supply a number of path => handler pairs"
   if @_ % 2;

  my $context = $self->{context};
  while ( my ( $path, $cb ) = splice @_, 0, 2 ) {
    $path = [$path] unless ref $path eq 'ARRAY';
    for my $elem ( @$path ) {
      my $h = $self->_get_rule_handler( $context, $elem )
       or croak "No existing handler for $elem";
      $self->{context}{rules}{$elem} = $mk_wrapper->( $h, $cb );
    }
  }
  return $self;
}

=head2 C<before>

Register a handler to be called before the existing handler for an
element. As with C<on> multiple elements may be targetted by providing
an array ref.

=cut

sub before {
  return shift->_filter(
    sub {
      my ( $h, $cb ) = @_;
      sub { $cb->( @_ ); $h->( @_ ) }
    },
    @_
  );
}

=head2 C<after>

Register a handler to be called after the existing handler for an
element. As with C<on> multiple elements may be targetted by providing
an array ref.

=cut

sub after {
  return shift->_filter(
    sub {
      my ( $h, $cb ) = @_;
      sub { $h->( @_ ); $cb->( @_ ) }
    },
    @_
  );
}

=head2 C<context>

Every time a handler is called a new scope is created for it. This
allows nested handlers to be defined. The current scope contains a user
context variable which can be used, for example, to keep track of an
object that is being filled with values parsed from the XML. The context
value is inherited from the parent scope but may be overridden locally.

For example:

  my $root = {};

  # Set the outermost context
  $p->context( $root );

  # Handle HTML <a href...> links /anywhere/
  $p->on(
    'a' => sub {
      my ( $elem, $attr, $ctx ) = @_;
      my $link = {
        href => $attr->{href},
        text => $p->text
      };
      push @{ $ctx->{links} }, $link;
    }
  );

  # Links in the body are stored in a nested
  # object.
  $p->on(
    'body' => sub {
      my ( $elem, $attr, $ctx ) = @_;
      my $body = {};
      # Set the context
      $p->context( $body );
      $p->walk;
      $ctx->{body} = $body;
    }
  );

  $p->walk;

Note that the handler for <a href...> tags stores its results in the
current context object - whatever that happens to be. That means
that outside of any <body> tag links will be stored in C<$root> but
within a <body> they will be stored in a nested object
(C<< $root->{body} >>). The <a> handler itself need know nothing of
this.

With no parameter C<context> returns the current context. The current
context is also passed as the third argument to handlers.

=cut

sub context {
  my $self = shift;
  $self->{context}->{obj} = shift if @_;
  return $self->{context}{obj};
}

=head2 C<text>

Return any text contained within the current element. XML markup is
discarded.

=cut

sub text {
  my $self = shift;
  my @txt  = ();

  TOKEN: while ( my $tok = $self->get_token ) {
    if ( $tok->[0] eq 'S' ) {
      push @txt, $self->text;
    }
    elsif ( $tok->[0] eq 'E' ) {
      last TOKEN;
    }
    elsif ( $tok->[0] eq 'T' ) {
      push @txt, $tok->[1];
    }
  }

  return join '', @txt;
}

=head2 C<xml>

Return the unparsed inner XML of the current element. For example:

  $p->on(
    'item' => sub {
      my ( $elem, $attr, $ctx ) = @_;
      my $item_source = $p->xml;
      print "Item: $item_source\n";
    }
  );

If <item> contains XHTML (for example) the above handler would correctly
capture it without recursively parsing any elements it contains. Parsing

  <feed>
    <item>This is the <i>first story</i>.</item>
    <item>This is <b>another story</b>.</item>
  </feed>
    
would print

  Item: This is the <i>first story</i>.
  Item: This is <b>another story</b>.

=cut

sub xml {
  my $self = shift;

  my @xml = ();

  TOKEN: while ( my $tok = $self->get_token ) {
    if ( $tok->[0] eq 'S' ) {
      push @xml, $tok->[4], $self->xml, $self->{token}->[2];
    }
    elsif ( $tok->[0] eq 'E' ) {
      last TOKEN;
    }
    elsif ( $tok->[0] eq 'T' || $tok->[0] eq 'C' ) {
      push @xml, $tok->[2];
    }
    elsif ( $tok->[0] eq 'PI' ) {
      push @xml, $tok->[3];
    }
    else {
      die "Unhandled token type: $tok->[0]";
    }
  }

  return join '', @xml;
}

=head2 C<get_path>

Called within a handler returns the path that leads to the current
element. For example:

  $p->on(
    'here' => sub {
      my ( $elem, $attr, $ctx ) = @_;
      print "I am here: ", $p->get_path, "\n";
      $p->walk;
    }
  );

would, if applied to this XML

  <outer>
    <inner>
      <here />
    </inner>
    <here />
  </outer>
    
print

  I am here: /outer/inner/here
  I am here: /outer/here

=cut

sub get_path { '/' . join '/', @{ shift->{path} } }

=head2 C<get_token>

XML::Descent is built on C<XML::TokeParser> which splits an XML document
into a stream of tokens representing start tags, end tags, literal text,
comment and processing instructions. Within an element C<get_token>
returns the same stream of tokens that C<XML::TokeParser> would produce.
Returns C<undef> once all the tokens contained within the current
element have been read (i.e. it's impossible to read past the end of the
enclosed XML).

=cut

sub get_token {
  my $self = shift;
  my $p    = $self->{parser};

  my $tok = $self->{token} = $p->get_token;

  if ( defined( $tok ) ) {
    if ( $tok->[0] eq 'S' ) {
      push @{ $self->{path} }, $tok->[1];
    }
    elsif ( $tok->[0] eq 'E' ) {
      my $tos = pop @{ $self->{path} };
      die "$tos <> $tok->[1]"
       unless $tos eq $tok->[1];
    }
  }

  my $stopat = $self->{context}{stopat};
  return if defined $stopat && $self->_depth < $stopat;
  return $tok;
}

=head2 C<scope_handlers>

Get a list of all handlers that are registered locally to the current
scope. The returned list won't include '*' if a wildcard handler has
been registered.

=cut

sub scope_handlers {
  sort grep { $_ ne '*' } keys %{ shift->{context}{rules} || {} };
}

=head2 C<all_handlers>

Get a list of all registered handlers in all scopes. The returned list
won't include the '*' wildcard handler.

=cut

sub all_handlers {
  my $self = shift;
  my %seen = ();
  my @h    = ();

  my $tos = $self->{context};
  while ( $tos ) {
    push @h, grep { !$seen{$_}++ }
     grep { $_ ne '*' } keys %{ $tos->{rules} || {} };
    $tos = $tos->{parent};
  }

  return sort @h;
}

1;

__END__

=head1 SEE ALSO

L<http://en.wikipedia.org/wiki/Recursive_descent_parser>,
L<XML::TokeParser>,
L<XML::Twig>.

=head1 BUGS AND LIMITATIONS

XML::Descent uses C<XML::TokeParser> to do the actual parsing.
XML::TokeParser can only return start tags, end tags, raw text and
processing instructions. As a result C<xml> called at the root of
an XML document will exclude any <?xml?> declaration.

No bugs have been reported.

Please report any bugs or feature requests to
C<bug-xml-descent@rt.cpan.org>, or through the web interface at
L<http://rt.cpan.org>.

=head1 AUTHOR

Andy Armstrong  C<< <andy@hexten.net> >>

=head1 LICENCE AND COPYRIGHT

Copyright (c) 2006-2009, Andy Armstrong C<< <andy@hexten.net> >>. All
rights reserved.

This module is free software; you can redistribute it and/or
modify it under the same terms as Perl itself. See L<perlartistic>.

=head1 DISCLAIMER OF WARRANTY

BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER
EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE
ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH
YOU. SHOULD THE SOFTWARE PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
NECESSARY SERVICING, REPAIR, OR CORRECTION.

IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE
LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL,
OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE
THE SOFTWARE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING
RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A
FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF
SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES.