8 |
use Carp qw/verbose/; |
use Carp qw/verbose/; |
9 |
use Module::Pluggable search_path => 'Grep::Source', sub_name => 'sources', require => 1; |
use Module::Pluggable search_path => 'Grep::Source', sub_name => 'sources', require => 1; |
10 |
use base qw(Class::Accessor Jifty::Object); |
use base qw(Class::Accessor Jifty::Object); |
11 |
Grep::Source->mk_accessors( qw(feed uri q new_items collection) ); |
Grep::Source->mk_accessors( qw(feed uri q new_items collection search_obj) ); |
12 |
|
|
13 |
use HTML::TreeBuilder; |
use HTML::TreeBuilder; |
14 |
use WWW::Mechanize; |
use WWW::Mechanize; |
108 |
my $class = $self->feed->source || 'Grep::Source::Feed'; |
my $class = $self->feed->source || 'Grep::Source::Feed'; |
109 |
$self->log->debug("using $class"); |
$self->log->debug("using $class"); |
110 |
|
|
111 |
my $parent = $self; |
$self->search_obj( Grep::Search->new() ); |
112 |
$class->fetch( $parent ); |
$self->log->debug("created " . $self->search_obj); |
|
undef $parent; |
|
113 |
|
|
114 |
Grep::Search->finish if $self->new_items; |
$class->fetch( $self ); |
115 |
|
|
116 |
|
$self->search_obj->finish; |
117 |
|
|
118 |
return $self->collection; |
return $self->collection; |
119 |
} |
} |
132 |
sub add_record { |
sub add_record { |
133 |
my $self = shift; |
my $self = shift; |
134 |
|
|
135 |
|
$self->log->confess("no search_obj") unless ($self->search_obj); |
136 |
|
|
137 |
my $i = Grep::Model::Item->new(); |
my $i = Grep::Model::Item->new(); |
138 |
|
|
139 |
my $rec = {@_}; |
my $rec = {@_}; |
140 |
|
|
141 |
warn "resolving links"; |
$self->log->debug("resolving links using base ", $rec->{link}); |
142 |
my $resolver = HTML::ResolveLink->new( base => $rec->{link} ); |
my $resolver = HTML::ResolveLink->new( base => $rec->{link} ); |
143 |
$rec->{content} = $resolver->resolve( $rec->{content} ); |
$rec->{content} = $resolver->resolve( $rec->{content} ); |
144 |
|
|
152 |
|
|
153 |
# is new record? |
# is new record? |
154 |
if ( $msg !~ m/^Found/ ) { |
if ( $msg !~ m/^Found/ ) { |
155 |
Grep::Search->add( $i ); |
$self->search_obj->add( $i ); |
156 |
$self->new_items( ( $self->new_items || 0 ) + 1 ); |
$self->new_items( ( $self->new_items || 0 ) + 1 ); |
157 |
} |
} |
158 |
} else { |
} else { |
200 |
|
|
201 |
my $args = {@_}; |
my $args = {@_}; |
202 |
|
|
203 |
warn "scrape got args ",dump($args); |
$self->log->debug("scrape with args ",dump($args)); |
204 |
|
|
205 |
my ($feed,$uri,$q) = ($self->feed, $self->uri,$self->q); |
my ($feed,$uri,$q) = ($self->feed, $self->uri,$self->q); |
206 |
die "no uri" unless ($uri); |
die "no uri" unless ($uri); |
222 |
$self->save( 'get.html', $mech->content ); |
$self->save( 'get.html', $mech->content ); |
223 |
|
|
224 |
if ( my $form = $args->{submit_form} ) { |
if ( my $form = $args->{submit_form} ) { |
225 |
warn "submit form on $uri with ", dump( $form ),"\n"; |
$self->log->debug("submit form on $uri with ", dump( $form )); |
226 |
$mech->submit_form( %$form ) or die "can't submit form ", dump( $form ); |
$mech->submit_form( %$form ) or die "can't submit form ", dump( $form ); |
227 |
$self->save( 'submit.html', $mech->content ); |
$self->save( 'submit.html', $mech->content ); |
228 |
} |
} |
229 |
|
|
230 |
warn "parse result page\n"; |
$self->log->debug("parse result page"); |
231 |
|
|
232 |
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
233 |
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
235 |
die "wrapper doesn't have 3 elements but ", $#{ $args->{wrapper} } unless ( $#{ $args->{wrapper} } == 2 ); |
die "wrapper doesn't have 3 elements but ", $#{ $args->{wrapper} } unless ( $#{ $args->{wrapper} } == 2 ); |
236 |
my ( $el,$attr,$value ) = @{ $args->{wrapper} }; |
my ( $el,$attr,$value ) = @{ $args->{wrapper} }; |
237 |
|
|
238 |
warn "looking for <$el $attr=\"$value\">"; |
$self->log->debug("looking for <$el $attr=\"$value\">"); |
239 |
|
|
240 |
my $div = $tree->look_down( '_tag', $el, sub { |
my $div = $tree->look_down( '_tag', $el, sub { |
|
warn dump( $_[0]->attr( $attr ) ),$/; |
|
241 |
( $_[0]->attr( $attr ) || '' ) eq $value; |
( $_[0]->attr( $attr ) || '' ) eq $value; |
242 |
}); |
}); |
243 |
|
|
246 |
return; |
return; |
247 |
} |
} |
248 |
|
|
249 |
my $max = 5; |
my $max = 15; |
250 |
my $nr = 1; |
my $nr = 1; |
251 |
|
|
252 |
my $base_uri = $uri; |
my $base_uri = $uri; |
253 |
$base_uri =~ s!\?.*$!!; |
$base_uri =~ s!\?.*$!!; |
254 |
|
|
255 |
foreach my $dt ( $div->look_down( '_tag', $args->{results} ) ) { |
my @r = $div->look_down( '_tag', $args->{results} ); |
256 |
|
$self->log->debug("found ", scalar @r, " <$args->{results}>"); |
257 |
|
|
258 |
|
foreach my $dt ( @r ) { |
259 |
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
260 |
if ( $a ) { |
if ( $a ) { |
261 |
|
|
264 |
$page_uri->query( undef ); |
$page_uri->query( undef ); |
265 |
$page_uri = $page_uri->canonical; |
$page_uri = $page_uri->canonical; |
266 |
|
|
267 |
warn "fetching page: ",$a->as_text," from $page_uri\n"; |
$self->log->debug("fetching page: ",$a->as_text," from $page_uri"); |
268 |
if ( $mech->follow_link( url => $a->attr('href') ) ) { |
if ( $mech->follow_link( url => $a->attr('href') ) ) { |
269 |
|
|
270 |
$self->save( "page-${nr}.html", $mech->content ); |
$self->save( "page-${nr}.html", $mech->content ); |