| 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 |
| 163 |
|
|
| 164 |
foreach my $s ( $self->sources ) { |
foreach my $s ( $self->sources ) { |
| 165 |
Jifty->log->debug("testing source class $s"); |
Jifty->log->debug("testing source class $s"); |
| 166 |
if ($s->can('content_have') && $s->content_have( $content ) ) { |
if ( $s->can('content_have') ) { |
| 167 |
Jifty->log->debug("${s}->content_have succesful"); |
my $regex = $s->content_have( $content ) or |
| 168 |
return "$s"; |
die "${s}->content_have didn't return anything"; |
| 169 |
|
die "${s}->content_have didn't return regex but ", dump( $regex ), " ref ", ref( $regex ) |
| 170 |
|
unless ( ref($regex) eq 'Regexp' ); |
| 171 |
|
if ( $content =~ $regex ) { |
| 172 |
|
Jifty->log->debug("${s}->content_have succesful"); |
| 173 |
|
return $s; |
| 174 |
|
} |
| 175 |
|
} |
| 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 mech_warn { |
| 198 |
|
my $m = shift || return; |
| 199 |
|
warn $m; |
| 200 |
|
} |
| 201 |
|
|
| 202 |
|
my $mech = WWW::Mechanize->new( |
| 203 |
|
cookie_jar => {}, |
| 204 |
|
onwarn => \&mech_warn, |
| 205 |
|
onerror => \&mech_warn, |
| 206 |
|
); |
| 207 |
|
|
| 208 |
|
$mech->get( $uri ); |
| 209 |
|
|
| 210 |
|
$self->save( 'get.html', $mech->content ); |
| 211 |
|
|
| 212 |
|
if ( my $form = $args->{submit_form} ) { |
| 213 |
|
warn "submit form on $uri with ", dump( $form ),"\n"; |
| 214 |
|
$mech->submit_form( %$form ) or die "can't submit form ", dump( $form ); |
| 215 |
|
$self->save( 'submit.html', $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 |
|
if ( ! $div ) { |
| 234 |
|
warn "can't find results wrapper <$el $attr=\"$value\">"; |
| 235 |
|
return; |
| 236 |
|
} |
| 237 |
|
|
| 238 |
|
my $max = 5; |
| 239 |
|
my $nr = 1; |
| 240 |
|
|
| 241 |
|
my $base_uri = $uri; |
| 242 |
|
$base_uri =~ s!\?.*$!!; |
| 243 |
|
|
| 244 |
|
foreach my $dt ( $div->look_down( '_tag', $args->{results} ) ) { |
| 245 |
|
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
| 246 |
|
if ( $a ) { |
| 247 |
|
|
| 248 |
|
my $href = $a->attr('href') or die "can't find href inside <", $args->{results}, ">"; |
| 249 |
|
my $page_uri = URI->new_abs( $a->attr('href'), $base_uri ); |
| 250 |
|
$page_uri->query( undef ); |
| 251 |
|
$page_uri = $page_uri->canonical; |
| 252 |
|
|
| 253 |
|
warn "fetching page: ",$a->as_text," from $page_uri\n"; |
| 254 |
|
if ( $mech->follow_link( url => $a->attr('href') ) ) { |
| 255 |
|
|
| 256 |
|
$self->save( "page-${nr}.html", $mech->content ); |
| 257 |
|
|
| 258 |
|
my $page_tree = HTML::TreeBuilder->new or die "can't create page tree"; |
| 259 |
|
$page_tree->parse( $mech->content ) or die "can't parse page at $page_uri"; |
| 260 |
|
|
| 261 |
|
( $el,$attr,$value ) = @{ $args->{scrape} }; |
| 262 |
|
$div = $page_tree->look_down( '_tag', $el, sub { ( $_[0]->attr( $attr ) || '' ) eq $value } ); |
| 263 |
|
|
| 264 |
|
die "can't find <$el $attr=\"$value\">" unless ($div); |
| 265 |
|
|
| 266 |
|
$self->add_record( |
| 267 |
|
in_feed => $feed, |
| 268 |
|
title => $mech->title, |
| 269 |
|
link => $page_uri, |
| 270 |
|
content => $div->as_HTML, |
| 271 |
|
# summary => |
| 272 |
|
# category => |
| 273 |
|
# author => |
| 274 |
|
# issued => |
| 275 |
|
# modified => |
| 276 |
|
); |
| 277 |
|
|
| 278 |
|
$mech->back; |
| 279 |
|
$page_tree->delete; |
| 280 |
|
|
| 281 |
|
} else { |
| 282 |
|
warn "can't follow uri $page_uri: $!\n"; |
| 283 |
|
} |
| 284 |
} |
} |
| 285 |
|
|
| 286 |
|
last if ($nr == $max); |
| 287 |
|
$nr++; |
| 288 |
|
} |
| 289 |
|
|
| 290 |
|
$tree->delete; # clear memory! |
| 291 |
|
|
| 292 |
|
} |
| 293 |
|
|
| 294 |
|
=head2 save |
| 295 |
|
|
| 296 |
|
save( 'name', $content ); |
| 297 |
|
|
| 298 |
|
Save dumps into C</tmp/grep> if writable |
| 299 |
|
|
| 300 |
|
=cut |
| 301 |
|
|
| 302 |
|
sub save { |
| 303 |
|
my $self = shift; |
| 304 |
|
my ( $file, $content ) = @_; |
| 305 |
|
if ( -w '/tmp/grep' ) { |
| 306 |
|
open(my $f, '>', "/tmp/grep/$file") or die "can't open $file: $!"; |
| 307 |
|
print $f $content or die "can't write to $file: $!"; |
| 308 |
|
close $f or die "can't close $file: $!"; |
| 309 |
|
Jifty->log->debug("saved $file ",length($content)," bytes"); |
| 310 |
} |
} |
| 311 |
} |
} |
| 312 |
|
|
Parent Directory
| 
Revision Log
| 
Patch