--- lib/Grep/Source.pm	2007/02/23 17:16:33	74
+++ lib/Grep/Source.pm	2007/02/24 11:16:05	92
@@ -10,6 +10,11 @@
 use base qw(Class::Accessor);
 Grep::Source->mk_accessors( qw(feed uri q new_items collection) );
 
+use HTML::TreeBuilder;
+use WWW::Mechanize;
+use XML::Feed;
+use URI;
+
 use Data::Dump qw/dump/;
 
 =head1 NAME
@@ -158,10 +163,150 @@
 
 	foreach my $s ( $self->sources ) {
 		Jifty->log->debug("testing source class $s");
-		if ($s->can('content_have') && $s->content_have( $content ) ) {
-			Jifty->log->debug("${s}->content_have succesful");
-			return "$s";
+		if ( $s->can('content_have') ) {
+			my $regex =	$s->content_have( $content ) or
+				die "${s}->content_have didn't return anything";
+			die "${s}->content_have didn't return regex but ", dump( $regex ), " ref ", ref( $regex )
+				unless ( ref($regex) eq 'Regexp' );
+			if ( $content =~ $regex ) {
+				Jifty->log->debug("${s}->content_have succesful");
+				return $s;
+			}
+		}
+	}
+}
+
+=head2 scrape
+
+Create semi-complex L<WWW::Mechanize> rules to scrape page
+
+
+=cut
+
+sub scrape {
+	my $self = shift;
+
+	my $args = {@_};
+
+	warn "scrape got args ",dump($args);
+
+	my ($feed,$uri,$q) = ($self->feed, $self->uri,$self->q);
+	die "no uri" unless ($uri);
+	die "feed is not a Grep::Model::Feed but ", ref $feed unless $feed->isa('Grep::Model::Feed');
+
+	sub mech_warn {
+		my $m = shift || return;
+		warn $m;
+	}
+
+	my $mech = WWW::Mechanize->new(
+		cookie_jar => {},
+		onwarn => \&mech_warn,
+		onerror => \&mech_warn,
+	);
+
+	$mech->get( $uri );
+
+	$self->save( 'get.html', $mech->content );
+
+	if ( my $form = $args->{submit_form} ) {
+		warn "submit form on $uri with ", dump( $form ),"\n";
+		$mech->submit_form( %$form ) or die "can't submit form ", dump( $form );
+		$self->save( 'submit.html', $mech->content );
+	}
+
+	warn "parse result page\n";
+
+	my $tree = HTML::TreeBuilder->new or die "can't create html tree";
+	$tree->parse( $mech->content ) or die "can't parse fetched content";
+
+	die "wrapper doesn't have 3 elements but ", $#{ $args->{wrapper} } unless ( $#{ $args->{wrapper} } == 2 );
+	my ( $el,$attr,$value ) = @{ $args->{wrapper} };
+
+	warn "looking for <$el $attr=\"$value\">";
+
+	my $div = $tree->look_down( '_tag', $el, sub {
+			warn dump( $_[0]->attr( $attr ) ),$/;
+			( $_[0]->attr( $attr ) || '' ) eq $value;
+	});
+
+	if ( ! $div ) {
+		warn "can't find results wrapper <$el $attr=\"$value\">";
+		return;
+	}
+
+	my $max = 5;
+	my $nr = 1;
+
+	my $base_uri = $uri;
+	$base_uri =~ s!\?.*$!!;
+
+	foreach my $dt ( $div->look_down( '_tag', $args->{results} ) ) {
+		my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } );
+		if ( $a ) {
+
+			my $href = $a->attr('href') or die "can't find href inside <", $args->{results}, ">";
+			my $page_uri = URI->new_abs( $a->attr('href'), $base_uri );
+			$page_uri->query( undef );
+			$page_uri = $page_uri->canonical;
+
+			warn "fetching page: ",$a->as_text," from $page_uri\n";
+			if ( $mech->follow_link( url => $a->attr('href') ) ) {
+
+				$self->save( "page-${nr}.html", $mech->content );
+
+				my $page_tree = HTML::TreeBuilder->new or die "can't create page tree";
+				$page_tree->parse( $mech->content ) or die "can't parse page at $page_uri";
+
+				( $el,$attr,$value ) = @{ $args->{scrape} };
+				$div = $page_tree->look_down( '_tag', $el, sub { ( $_[0]->attr( $attr ) || '' ) eq $value } );
+
+				die "can't find <$el $attr=\"$value\">" unless ($div);
+
+				$self->add_record(
+					in_feed => $feed,
+					title => $mech->title,
+					link => $page_uri,
+					content => $div->as_HTML,
+#					summary => 
+#					category => 
+#					author => 
+#					issued => 
+#					modified => 
+				);
+
+				$mech->back;
+				$page_tree->delete;
+
+			} else {
+				warn "can't follow uri $page_uri: $!\n";
+			}
 		}
+
+		last if ($nr == $max);
+		$nr++;
+	}
+
+	$tree->delete; # clear memory!
+
+}
+
+=head2 save
+
+  save( 'name', $content );
+
+Save dumps into C</tmp/grep> if writable
+
+=cut
+
+sub save {
+	my $self = shift;
+	my ( $file, $content ) = @_;
+	if ( -w '/tmp/grep' ) {
+		open(my $f, '>', "/tmp/grep/$file") or die "can't open $file: $!";
+		print $f $content or die "can't write to $file: $!";
+		close $f or die "can't close $file: $!";
+		Jifty->log->debug("saved $file ",length($content)," bytes");
 	}
 }