11 |
|
|
12 |
=cut |
=cut |
13 |
|
|
|
use HTML::TreeBuilder; |
|
|
use WWW::Mechanize; |
|
|
use XML::Feed; |
|
|
|
|
|
=head2 content_have |
|
|
|
|
|
Return regex to match against content |
|
|
|
|
|
=cut |
|
|
|
|
14 |
sub content_have { |
sub content_have { |
15 |
qr/MoinMoin/ |
qr/MoinMoin/ |
16 |
} |
} |
17 |
|
|
|
=head2 fetch |
|
|
|
|
|
$plugin->fetch( $source ); |
|
|
|
|
|
=cut |
|
|
|
|
18 |
sub fetch { |
sub fetch { |
19 |
my $self = shift; |
my $self = shift; |
|
|
|
20 |
my $parent = shift; |
my $parent = shift; |
|
my ($feed,$uri,$q) = ($parent->feed, $parent->uri,$parent->q); |
|
|
die "no uri" unless ($uri); |
|
|
die "feed is not a Grep::Model::Feed but ", ref $feed unless $feed->isa('Grep::Model::Feed'); |
|
|
|
|
|
my $mech = WWW::Mechanize->new(); |
|
|
|
|
|
$mech->get( $uri ); |
|
21 |
|
|
22 |
warn "submit $self form on $uri\n"; |
$parent->scrape( |
23 |
|
submit_form => { |
24 |
$mech->submit_form( |
fields => { |
25 |
fields => { |
value => $parent->q, |
26 |
value => $q, |
}, |
27 |
|
button => 'fullsearch', |
28 |
}, |
}, |
29 |
button => 'fullsearch', |
wrapper => [ qw/div class searchresults/ ], |
30 |
) or die "can't $self submit"; |
results => 'dt', |
31 |
|
scrape => [ qw/div id page/ ], |
32 |
warn "parse result page\n"; |
); |
|
|
|
|
my $tree = HTML::TreeBuilder->new or die "can't create html tree"; |
|
|
$tree->parse( $mech->content ) or die "can't parse fetched content"; |
|
|
|
|
|
my $div = $tree->look_down( '_tag', 'div', sub { |
|
|
( $_[0]->attr('class') || '' ) eq 'searchresults' |
|
|
}); |
|
|
|
|
|
die "can't find div with class searchresults" unless ( $div ); |
|
|
|
|
|
my $max = 5; |
|
|
my $nr = 1; |
|
|
|
|
|
my $base_uri = $uri; |
|
|
$base_uri =~ s!\?.*$!!; |
|
|
|
|
|
foreach my $dt ( $div->look_down( '_tag', 'dt' ) ) { |
|
|
my $a = $dt->look_down( '_tag', 'a', sub { $_[0]->attr('href') } ); |
|
|
if ( $a ) { |
|
|
|
|
|
my $page_uri = $base_uri . $a->attr('href'); |
|
|
$page_uri =~ s!\Q?highlight=\E.*$!!; |
|
|
|
|
|
warn "fetching page: ",$a->as_text," from $page_uri\n"; |
|
|
if ( $mech->follow_link( url => $a->attr('href') ) ) { |
|
|
|
|
|
my $page_tree = HTML::TreeBuilder->new or die "can't create page tree"; |
|
|
$page_tree->parse( $mech->content ) or die "can't parse page at $page_uri"; |
|
|
|
|
|
my $div = $page_tree->look_down( '_tag', 'div', sub { ( $_[0]->attr('id') || '' ) eq 'page' } ); |
|
|
|
|
|
die "can't find <div id=page>" unless ($div); |
|
|
|
|
|
$parent->add_record( |
|
|
in_feed => $feed, |
|
|
title => $mech->title, |
|
|
link => $page_uri, |
|
|
content => $div->as_HTML, |
|
|
# summary => |
|
|
# category => |
|
|
# author => |
|
|
# issued => |
|
|
# modified => |
|
|
); |
|
|
|
|
|
$mech->back; |
|
|
|
|
|
} else { |
|
|
warn "can't follow uri $page_uri: $!\n"; |
|
|
} |
|
|
} |
|
|
|
|
|
last if ($nr == $max); |
|
|
$nr++; |
|
|
} |
|
|
|
|
|
$tree->delete; # clear memory! |
|
33 |
|
|
34 |
} |
} |
35 |
|
|