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 |
|
|
212 |
my $args = {@_}; |
my $args = {@_}; |
213 |
|
|
214 |
my $tree = $args->{tree} || die "no tree"; |
my $tree = $args->{tree} || die "no tree"; |
215 |
my $message = $args->{message}; |
my $message = $args->{message} || ''; |
216 |
my $fatal = $args->{fatal}; |
my $fatal = $args->{fatal}; |
217 |
die "no templates" unless defined( $args->{templates} ); |
die "no triplets" unless defined( $args->{triplets} ); |
218 |
my @templates = @{ $args->{templates} }; |
my @triplets; |
219 |
|
if ( ref( $args->{triplets} ) eq 'ARRAY' ) { |
220 |
|
@triplets = @{ $args->{triplets} }; |
221 |
|
} else { |
222 |
|
@triplets = ( $args->{triplets} ); |
223 |
|
} |
224 |
|
|
225 |
push @templates, ( undef, undef ) if ( $#templates == 0 ); |
push @triplets, ( undef, undef ) if ( $#triplets == 0 ); |
226 |
|
|
227 |
die "wrapper doesn't have 3 elements but ", $#templates unless ( |
die "triplet doesn't have 3 elements but ", $#triplets unless ( |
228 |
( $#templates + 1 ) % 3 == 0 |
( $#triplets + 1 ) % 3 == 0 |
229 |
); |
); |
230 |
|
|
231 |
my ( $result, $el, $attr, $value ); |
my ( $el, $attr, $value ); |
232 |
|
|
233 |
|
my @results; |
234 |
my @tags; |
my @tags; |
235 |
|
|
236 |
while ( @templates ) { |
$self->log->debug("looking for $message ", dump( @triplets )); |
237 |
( $el,$attr,$value ) = splice( @templates, 0, 3 ); |
while ( @triplets ) { |
238 |
|
( $el,$attr,$value ) = splice( @triplets, 0, 3 ); |
239 |
my $tag = $attr ? "<$el $attr=\"$value\">" : "<$el>"; |
my $tag = $attr ? "<$el $attr=\"$value\">" : "<$el>"; |
240 |
push @tags, $tag; |
push @tags, $tag; |
241 |
$self->log->debug("looking for $message $tag"); |
@results = $tree->look_down( '_tag', $el, sub { |
|
$result = $tree->look_down( '_tag', $el, sub { |
|
242 |
return 1 unless ( $attr && $value ); |
return 1 unless ( $attr && $value ); |
243 |
( $_[0]->attr( $attr ) || '' ) eq $value; |
( $_[0]->attr( $attr ) || '' ) =~ m/\b\Q$value\E\b/ |
244 |
}); |
}); |
245 |
last if $result; |
last if @results; |
246 |
} |
} |
247 |
|
|
248 |
if ( ! $result ) { |
if ( ! @results ) { |
249 |
my $msg = "can't find $message ", join(" ", @tags); |
my $msg = "can't find $message " . join(" ", @tags); |
250 |
die $msg if ( $fatal ); |
die $msg if ( $fatal ); |
251 |
warn $msg; |
#warn $msg; |
252 |
return; |
return; |
253 |
} |
} |
254 |
|
|
255 |
return $result; |
$self->log->debug("found ", $#results + 1, " elements"); |
256 |
|
#warn dump( map { $_->as_HTML } @results ); |
257 |
|
|
258 |
|
return @results if wantarray; |
259 |
|
return shift @results; |
260 |
} |
} |
261 |
|
|
262 |
|
=head2 scrape |
263 |
|
|
264 |
|
Create semi-complex L<WWW::Mechanize> rules to scrape page easily |
265 |
|
|
266 |
|
$parent->scrape( |
267 |
|
# if search string isn't part or URI |
268 |
|
submit_form => { |
269 |
|
fields => { |
270 |
|
value => $parent->q, |
271 |
|
}, |
272 |
|
button => 'fullsearch', |
273 |
|
}, |
274 |
|
# element with search results |
275 |
|
wrapper => [ qw/div class searchresults/ ], |
276 |
|
# element (or tripple) for each result with link |
277 |
|
# <a href=".."> inside it to full-text result |
278 |
|
results => 'dt', |
279 |
|
# collect which element on page linked from results |
280 |
|
scrape => [ qw/div id page/ ], |
281 |
|
# when search returns just single hit, it will redirect to result page |
282 |
|
redirect_single_result => 1, |
283 |
|
); |
284 |
|
|
285 |
|
=cut |
286 |
|
|
287 |
sub scrape { |
sub scrape { |
288 |
my $self = shift; |
my $self = shift; |
289 |
|
|
321 |
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
322 |
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
323 |
|
|
324 |
my $div = $self->element_by_triplet( |
my @wrapper_divs = $self->element_by_triplet( |
325 |
tree => $tree, |
tree => $tree, |
326 |
templates => $args->{wrapper}, |
triplets => $args->{wrapper}, |
327 |
message => 'wrapper for all results', |
message => 'wrapper for all results', |
328 |
fatal => 1 |
# on closer recollection, this shouldn't be ever fatal, because |
329 |
|
# "no results found" page might not contain wrapper |
330 |
|
#fatal => $args->{redirect_single_result} ? 0 : 1, |
331 |
); |
); |
332 |
|
|
333 |
my $max = 15; |
my $max = 15; |
336 |
my $base_uri = $uri; |
my $base_uri = $uri; |
337 |
$base_uri =~ s!\?.*$!!; |
$base_uri =~ s!\?.*$!!; |
338 |
|
|
339 |
my @r = $self->element_by_triplet( |
# directly got first result |
340 |
tree => $tree, |
if ( $args->{redirect_single_result} && ! @wrapper_divs ) { |
341 |
templates => $args->{results}, |
|
342 |
message => 'result element', |
my $uri = $mech->uri; $uri->query( undef ); $uri = $uri->canonical; |
343 |
); |
|
344 |
|
my $div = $self->element_by_triplet( |
345 |
|
tree => $tree, |
346 |
|
message => "single result - redirect to $uri", |
347 |
|
triplets => $args->{scrape}, |
348 |
|
fatal => 1, |
349 |
|
); |
350 |
|
|
351 |
|
$self->add_record( |
352 |
|
in_feed => $feed, |
353 |
|
title => $mech->title, |
354 |
|
link => $uri, |
355 |
|
content => $div->as_HTML, |
356 |
|
); |
357 |
|
|
358 |
|
$tree->delete; # clear memory! |
359 |
|
return; |
360 |
|
} |
361 |
|
|
362 |
|
my @r; |
363 |
|
|
364 |
|
foreach my $div ( @wrapper_divs ) { |
365 |
|
|
366 |
|
my @r_here = $self->element_by_triplet( |
367 |
|
tree => $div, |
368 |
|
triplets => $args->{results}, |
369 |
|
message => 'result element', |
370 |
|
); |
371 |
|
|
372 |
|
push @r, @r_here if (@r_here); |
373 |
|
} |
374 |
|
|
375 |
|
$self->log->debug("in total, found ", $#r + 1, " results in ", $#wrapper_divs + 1, " result wrapper elements"); |
376 |
|
|
377 |
foreach my $dt ( @r ) { |
foreach my $dt ( @r ) { |
378 |
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
379 |
if ( $a ) { |
if ( $a ) { |
380 |
|
|
381 |
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}, ">"; |
382 |
my $page_uri = URI->new_abs( $a->attr('href'), $base_uri ); |
|
383 |
|
my $page_uri = URI->new_abs( $href, $base_uri ); |
384 |
$page_uri->query( undef ); |
$page_uri->query( undef ); |
385 |
$page_uri = $page_uri->canonical; |
$page_uri = $page_uri->canonical; |
386 |
|
|
387 |
|
if ( my $item = Grep::Model::Item->link_current( $page_uri ) ) { |
388 |
|
Jifty->log->debug("using cached page for $page_uri"); |
389 |
|
$self->collection->add_record( $item ); |
390 |
|
next; |
391 |
|
} |
392 |
|
|
393 |
$self->log->debug("fetching page: ",$a->as_text," from $page_uri"); |
$self->log->debug("fetching page: ",$a->as_text," from $page_uri"); |
394 |
if ( $mech->follow_link( url => $a->attr('href') ) ) { |
if ( $mech->follow_link( url => $href ) ) { |
395 |
|
|
396 |
$self->save( "page-${nr}.html", $mech->content ); |
$self->save( "page-${nr}.html", $mech->content ); |
397 |
|
|
398 |
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"; |
399 |
$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"; |
400 |
$div = $self->element_by_triplet( |
my @divs = $self->element_by_triplet( |
401 |
tree => $page_tree, |
tree => $page_tree, |
402 |
message => "result $nr", |
message => "result page $nr", |
403 |
templates => $args->{scrape} |
triplets => $args->{scrape} |
404 |
); |
); |
405 |
|
|
406 |
$self->add_record( |
if ( @divs ) { |
407 |
in_feed => $feed, |
|
408 |
title => $mech->title, |
my $html = join("<hr/>\n", map { $_->as_HTML } @divs ); |
409 |
link => $page_uri, |
$self->log->debug("combined ", $#divs + 1, " elements elements in ", length($html), " bytes"); |
410 |
content => $div->as_HTML, |
|
411 |
# summary => |
$self->add_record( |
412 |
# category => |
in_feed => $feed, |
413 |
# author => |
title => $mech->title, |
414 |
# issued => |
link => $page_uri, |
415 |
# modified => |
content => $html, |
416 |
) if ( $div ); |
# summary => |
417 |
|
# category => |
418 |
|
# author => |
419 |
|
# issued => |
420 |
|
# modified => |
421 |
|
); |
422 |
|
|
423 |
|
} else { |
424 |
|
$self->log->debug("NO CONTENT scraped from page $nr"); |
425 |
|
} |
426 |
|
|
427 |
$mech->back; |
$mech->back; |
428 |
$page_tree->delete; |
$page_tree->delete; |