188 |
} |
} |
189 |
} |
} |
190 |
|
|
|
=head2 scrape |
|
191 |
|
|
192 |
Create semi-complex L<WWW::Mechanize> rules to scrape page |
=head2 element_by_triplet |
193 |
|
|
194 |
|
Helper method to select element(s) using C<element/attribute/value> triplet using |
195 |
|
L<HTML::TreeBuilder> trees. |
196 |
|
|
197 |
|
my $el = $self->element_by_triplet( |
198 |
|
tree => $tree_or_element, |
199 |
|
triplets => [ qw/ |
200 |
|
div id target |
201 |
|
div class another |
202 |
|
/ ], |
203 |
|
message => 'find search result element', |
204 |
|
fatal => 1, # die instead of warn |
205 |
|
); |
206 |
|
|
207 |
=cut |
=cut |
208 |
|
|
224 |
|
|
225 |
push @triplets, ( undef, undef ) if ( $#triplets == 0 ); |
push @triplets, ( undef, undef ) if ( $#triplets == 0 ); |
226 |
|
|
227 |
die "wrapper doesn't have 3 elements but ", $#triplets unless ( |
die "triplet doesn't have 3 elements but ", $#triplets unless ( |
228 |
( $#triplets + 1 ) % 3 == 0 |
( $#triplets + 1 ) % 3 == 0 |
229 |
); |
); |
230 |
|
|
258 |
return shift @results; |
return shift @results; |
259 |
} |
} |
260 |
|
|
261 |
|
=head2 scrape |
262 |
|
|
263 |
|
Create semi-complex L<WWW::Mechanize> rules to scrape page easily |
264 |
|
|
265 |
|
$parent->scrape( |
266 |
|
# if search string isn't part or URI |
267 |
|
submit_form => { |
268 |
|
fields => { |
269 |
|
value => $parent->q, |
270 |
|
}, |
271 |
|
button => 'fullsearch', |
272 |
|
}, |
273 |
|
# element with search results |
274 |
|
wrapper => [ qw/div class searchresults/ ], |
275 |
|
# element (or tripple) for each result with link |
276 |
|
# <a href=".."> inside it to full-text result |
277 |
|
results => 'dt', |
278 |
|
# collect which element on page linked from results |
279 |
|
scrape => [ qw/div id page/ ], |
280 |
|
# when search returns just single hit, it will redirect to result page |
281 |
|
redirect_single_result => 1, |
282 |
|
); |
283 |
|
|
284 |
|
=cut |
285 |
|
|
286 |
sub scrape { |
sub scrape { |
287 |
my $self = shift; |
my $self = shift; |
288 |
|
|
324 |
tree => $tree, |
tree => $tree, |
325 |
triplets => $args->{wrapper}, |
triplets => $args->{wrapper}, |
326 |
message => 'wrapper for all results', |
message => 'wrapper for all results', |
327 |
fatal => 1 |
fatal => $args->{redirect_single_result} ? 0 : 1, |
328 |
); |
); |
329 |
|
|
330 |
my $max = 15; |
my $max = 15; |
333 |
my $base_uri = $uri; |
my $base_uri = $uri; |
334 |
$base_uri =~ s!\?.*$!!; |
$base_uri =~ s!\?.*$!!; |
335 |
|
|
336 |
|
# directly got first result |
337 |
|
if ( $args->{redirect_single_result} && ! $div ) { |
338 |
|
|
339 |
|
my $uri = $mech->uri; $uri->query( undef ); $uri = $uri->canonical; |
340 |
|
|
341 |
|
$div = $self->element_by_triplet( |
342 |
|
tree => $tree, |
343 |
|
message => "single result - redirect to $uri", |
344 |
|
triplets => $args->{scrape}, |
345 |
|
fatal => 1, |
346 |
|
); |
347 |
|
|
348 |
|
$self->add_record( |
349 |
|
in_feed => $feed, |
350 |
|
title => $mech->title, |
351 |
|
link => $uri, |
352 |
|
content => $div->as_HTML, |
353 |
|
); |
354 |
|
|
355 |
|
$tree->delete; # clear memory! |
356 |
|
return; |
357 |
|
} |
358 |
|
|
359 |
my @r = $self->element_by_triplet( |
my @r = $self->element_by_triplet( |
360 |
tree => $tree, |
tree => $div, |
361 |
triplets => $args->{results}, |
triplets => $args->{results}, |
362 |
message => 'result element', |
message => 'result element', |
363 |
); |
); |
367 |
if ( $a ) { |
if ( $a ) { |
368 |
|
|
369 |
my $href = $a->attr('href') or die "can't find href inside <", $args->{results}, ">"; |
my $href = $a->attr('href') or die "can't find href inside <", $args->{results}, ">"; |
370 |
my $page_uri = URI->new_abs( $a->attr('href'), $base_uri ); |
|
371 |
|
my $page_uri = URI->new_abs( $href, $base_uri ); |
372 |
$page_uri->query( undef ); |
$page_uri->query( undef ); |
373 |
$page_uri = $page_uri->canonical; |
$page_uri = $page_uri->canonical; |
374 |
|
|
375 |
$self->log->debug("fetching page: ",$a->as_text," from $page_uri"); |
$self->log->debug("fetching page: ",$a->as_text," from $page_uri"); |
376 |
if ( $mech->follow_link( url => $a->attr('href') ) ) { |
if ( $mech->follow_link( url => $href ) ) { |
377 |
|
|
378 |
$self->save( "page-${nr}.html", $mech->content ); |
$self->save( "page-${nr}.html", $mech->content ); |
379 |
|
|