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 search_obj) ); |
Grep::Source->mk_accessors( qw(feed uri q new_items collection search_obj tree) ); |
12 |
|
|
13 |
use HTML::TreeBuilder; |
use HTML::TreeBuilder; |
14 |
use WWW::Mechanize; |
use WWW::Mechanize; |
195 |
|
|
196 |
=cut |
=cut |
197 |
|
|
198 |
|
sub element_by_triplet { |
199 |
|
my $self = shift; |
200 |
|
|
201 |
|
my $args = {@_}; |
202 |
|
|
203 |
|
my $tree = $args->{tree} || die "no tree"; |
204 |
|
my $message = $args->{message}; |
205 |
|
my $fatal = $args->{fatal}; |
206 |
|
die "no templates" unless defined( $args->{templates} ); |
207 |
|
my @templates = @{ $args->{templates} }; |
208 |
|
|
209 |
|
push @templates, ( undef, undef ) if ( $#templates == 0 ); |
210 |
|
|
211 |
|
die "wrapper doesn't have 3 elements but ", $#templates unless ( |
212 |
|
( $#templates + 1 ) % 3 == 0 |
213 |
|
); |
214 |
|
|
215 |
|
my ( $result, $el, $attr, $value ); |
216 |
|
|
217 |
|
my @tags; |
218 |
|
|
219 |
|
while ( @templates ) { |
220 |
|
( $el,$attr,$value ) = splice( @templates, 0, 3 ); |
221 |
|
my $tag = $attr ? "<$el $attr=\"$value\">" : "<$el>"; |
222 |
|
push @tags, $tag; |
223 |
|
$self->log->debug("looking for $message $tag"); |
224 |
|
$result = $tree->look_down( '_tag', $el, sub { |
225 |
|
return 1 unless ( $attr && $value ); |
226 |
|
( $_[0]->attr( $attr ) || '' ) eq $value; |
227 |
|
}); |
228 |
|
last if $result; |
229 |
|
} |
230 |
|
|
231 |
|
if ( ! $result ) { |
232 |
|
my $msg = "can't find $message ", join(" ", @tags); |
233 |
|
die $msg if ( $fatal ); |
234 |
|
warn $msg; |
235 |
|
return; |
236 |
|
} |
237 |
|
|
238 |
|
return $result; |
239 |
|
} |
240 |
|
|
241 |
sub scrape { |
sub scrape { |
242 |
my $self = shift; |
my $self = shift; |
243 |
|
|
275 |
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
276 |
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
277 |
|
|
278 |
die "wrapper doesn't have 3 elements but ", $#{ $args->{wrapper} } unless ( $#{ $args->{wrapper} } == 2 ); |
my $div = $self->element_by_triplet( |
279 |
my ( $el,$attr,$value ) = @{ $args->{wrapper} }; |
tree => $tree, |
280 |
|
templates => $args->{wrapper}, |
281 |
$self->log->debug("looking for <$el $attr=\"$value\">"); |
message => 'wrapper for all results', |
282 |
|
fatal => 1 |
283 |
my $div = $tree->look_down( '_tag', $el, sub { |
); |
|
( $_[0]->attr( $attr ) || '' ) eq $value; |
|
|
}); |
|
|
|
|
|
if ( ! $div ) { |
|
|
warn "can't find results wrapper <$el $attr=\"$value\">"; |
|
|
return; |
|
|
} |
|
284 |
|
|
285 |
my $max = 15; |
my $max = 15; |
286 |
my $nr = 1; |
my $nr = 1; |
288 |
my $base_uri = $uri; |
my $base_uri = $uri; |
289 |
$base_uri =~ s!\?.*$!!; |
$base_uri =~ s!\?.*$!!; |
290 |
|
|
291 |
my @r = $div->look_down( '_tag', $args->{results} ); |
my @r = $self->element_by_triplet( |
292 |
$self->log->debug("found ", scalar @r, " <$args->{results}>"); |
tree => $tree, |
293 |
|
templates => $args->{results}, |
294 |
|
message => 'result element', |
295 |
|
); |
296 |
|
|
297 |
foreach my $dt ( @r ) { |
foreach my $dt ( @r ) { |
298 |
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
310 |
|
|
311 |
my $page_tree = HTML::TreeBuilder->new or die "can't create page tree"; |
my $page_tree = HTML::TreeBuilder->new or die "can't create page tree"; |
312 |
$page_tree->parse( $mech->content ) or die "can't parse page at $page_uri"; |
$page_tree->parse( $mech->content ) or die "can't parse page at $page_uri"; |
313 |
|
$div = $self->element_by_triplet( |
314 |
( $el,$attr,$value ) = @{ $args->{scrape} }; |
tree => $page_tree, |
315 |
$div = $page_tree->look_down( '_tag', $el, sub { ( $_[0]->attr( $attr ) || '' ) eq $value } ); |
message => "result $nr", |
316 |
|
templates => $args->{scrape} |
317 |
die "can't find <$el $attr=\"$value\">" unless ($div); |
); |
318 |
|
|
319 |
$self->add_record( |
$self->add_record( |
320 |
in_feed => $feed, |
in_feed => $feed, |
326 |
# author => |
# author => |
327 |
# issued => |
# issued => |
328 |
# modified => |
# modified => |
329 |
); |
) if ( $div ); |
330 |
|
|
331 |
$mech->back; |
$mech->back; |
332 |
$page_tree->delete; |
$page_tree->delete; |
334 |
} else { |
} else { |
335 |
warn "can't follow uri $page_uri: $!\n"; |
warn "can't follow uri $page_uri: $!\n"; |
336 |
} |
} |
337 |
|
} else { |
338 |
|
$self->log->debug("result $nr doesn't have link inside, ignoring..."); |
339 |
} |
} |
340 |
|
|
341 |
last if ($nr == $max); |
last if ($nr == $max); |