10 |
use base qw(Class::Accessor); |
use base qw(Class::Accessor); |
11 |
Grep::Source->mk_accessors( qw(feed uri q new_items collection) ); |
Grep::Source->mk_accessors( qw(feed uri q new_items collection) ); |
12 |
|
|
13 |
|
use HTML::TreeBuilder; |
14 |
|
use WWW::Mechanize; |
15 |
|
use XML::Feed; |
16 |
|
use URI; |
17 |
|
|
18 |
use Data::Dump qw/dump/; |
use Data::Dump qw/dump/; |
19 |
|
|
20 |
=head1 NAME |
=head1 NAME |
176 |
} |
} |
177 |
} |
} |
178 |
|
|
179 |
|
=head2 scrape |
180 |
|
|
181 |
|
Create semi-complex L<WWW::Mechanize> rules to scrape page |
182 |
|
|
183 |
|
|
184 |
|
=cut |
185 |
|
|
186 |
|
sub scrape { |
187 |
|
my $self = shift; |
188 |
|
|
189 |
|
my $args = {@_}; |
190 |
|
|
191 |
|
warn "scrape got args ",dump($args); |
192 |
|
|
193 |
|
my ($feed,$uri,$q) = ($self->feed, $self->uri,$self->q); |
194 |
|
die "no uri" unless ($uri); |
195 |
|
die "feed is not a Grep::Model::Feed but ", ref $feed unless $feed->isa('Grep::Model::Feed'); |
196 |
|
|
197 |
|
sub save_html { |
198 |
|
my ( $file, $content ) = @_; |
199 |
|
if ( -w '/tmp/grep' ) { |
200 |
|
open(my $f, '>', "/tmp/grep/${file}.html") or die "can't open $file: $!"; |
201 |
|
print $f $content or die "can't write to $file: $!"; |
202 |
|
close $f or die "can't close $file: $!"; |
203 |
|
} |
204 |
|
} |
205 |
|
|
206 |
|
my $mech = WWW::Mechanize->new(); |
207 |
|
|
208 |
|
$mech->get( $uri ); |
209 |
|
|
210 |
|
save_html( 'get', $mech->content ); |
211 |
|
|
212 |
|
if ( $args->{submit_form} ) { |
213 |
|
warn "submit form on $uri\n"; |
214 |
|
$mech->submit_form( %{ $args->{submit_form} } ) or die "can't submit form"; |
215 |
|
save_html( 'submit', $mech->content ); |
216 |
|
} |
217 |
|
|
218 |
|
warn "parse result page\n"; |
219 |
|
|
220 |
|
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
221 |
|
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
222 |
|
|
223 |
|
die "wrapper doesn't have 3 elements but ", $#{ $args->{wrapper} } unless ( $#{ $args->{wrapper} } == 2 ); |
224 |
|
my ( $el,$attr,$value ) = @{ $args->{wrapper} }; |
225 |
|
|
226 |
|
warn "looking for <$el $attr=\"$value\">"; |
227 |
|
|
228 |
|
my $div = $tree->look_down( '_tag', $el, sub { |
229 |
|
warn dump( $_[0]->attr( $attr ) ),$/; |
230 |
|
( $_[0]->attr( $attr ) || '' ) eq $value; |
231 |
|
}); |
232 |
|
|
233 |
|
die "can't find results wrapper <$el $attr=\"$value\">" unless ( $div ); |
234 |
|
|
235 |
|
my $max = 5; |
236 |
|
my $nr = 1; |
237 |
|
|
238 |
|
my $base_uri = $uri; |
239 |
|
$base_uri =~ s!\?.*$!!; |
240 |
|
|
241 |
|
foreach my $dt ( $div->look_down( '_tag', $args->{results} ) ) { |
242 |
|
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
243 |
|
if ( $a ) { |
244 |
|
|
245 |
|
my $href = $a->attr('href') or die "can't find href inside <", $args->{results}, ">"; |
246 |
|
my $page_uri = URI->new_abs( $a->attr('href'), $base_uri ); |
247 |
|
$page_uri->query( undef ); |
248 |
|
$page_uri = $page_uri->canonical; |
249 |
|
|
250 |
|
warn "fetching page: ",$a->as_text," from $page_uri\n"; |
251 |
|
if ( $mech->follow_link( url => $a->attr('href') ) ) { |
252 |
|
|
253 |
|
save_html( "page-${nr}", $mech->content ); |
254 |
|
|
255 |
|
my $page_tree = HTML::TreeBuilder->new or die "can't create page tree"; |
256 |
|
$page_tree->parse( $mech->content ) or die "can't parse page at $page_uri"; |
257 |
|
|
258 |
|
my ( $el,$attr,$value ) = @{ $args->{scrape} }; |
259 |
|
my $div = $page_tree->look_down( '_tag', $el, sub { ( $_[0]->attr( $attr ) || '' ) eq $value } ); |
260 |
|
|
261 |
|
die "can't find <$el $attr=\"$value\">" unless ($div); |
262 |
|
|
263 |
|
$self->add_record( |
264 |
|
in_feed => $feed, |
265 |
|
title => $mech->title, |
266 |
|
link => $page_uri, |
267 |
|
content => $div->as_HTML, |
268 |
|
# summary => |
269 |
|
# category => |
270 |
|
# author => |
271 |
|
# issued => |
272 |
|
# modified => |
273 |
|
); |
274 |
|
|
275 |
|
$mech->back; |
276 |
|
$page_tree->delete; |
277 |
|
|
278 |
|
} else { |
279 |
|
warn "can't follow uri $page_uri: $!\n"; |
280 |
|
} |
281 |
|
} |
282 |
|
|
283 |
|
last if ($nr == $max); |
284 |
|
$nr++; |
285 |
|
} |
286 |
|
|
287 |
|
$tree->delete; # clear memory! |
288 |
|
|
289 |
|
} |
290 |
|
|
291 |
1; |
1; |