7 |
use lib 'lib'; |
use lib 'lib'; |
8 |
|
|
9 |
use Jifty; |
use Jifty; |
|
use Lucene; |
|
10 |
use Grep::Search; |
use Grep::Search; |
11 |
use Data::Dump qw/dump/; |
use Data::Dump qw/dump/; |
12 |
|
use Text::DeDuper; |
13 |
|
use Encode; |
14 |
|
|
15 |
|
my $remove_duplicates = 1; |
16 |
|
|
17 |
|
$|=1; |
18 |
|
|
19 |
BEGIN { Jifty->new; }; |
BEGIN { Jifty->new; }; |
20 |
|
|
21 |
my $coll = Grep::Model::ItemCollection->new( results_are_readable => 1 ); |
my $system_user = Grep::CurrentUser->superuser; |
22 |
|
my $coll = Grep::Model::ItemCollection->new( results_are_readable => 1, current_user => $system_user ); |
23 |
$coll->unlimit; |
$coll->unlimit; |
24 |
|
|
25 |
print "indexing ", $coll->count, " items "; |
print "indexing ", $coll->count, " items "; |
26 |
|
|
27 |
|
my $search = Grep::Search->new(); |
28 |
|
my $deduper = Text::DeDuper->new(); |
29 |
|
|
30 |
|
my ( $total, $duplicates ) = ( 0, 0 ); |
31 |
|
|
32 |
while ( my $i = $coll->next ) { |
while ( my $i = $coll->next ) { |
33 |
|
|
34 |
Grep::Search->add( $i ); |
print $i->id; |
35 |
|
|
36 |
|
if ( $remove_duplicates ) { |
37 |
|
|
38 |
|
my $c = encode('utf-8', $i->content); |
39 |
|
|
40 |
print $i->id, ' '; |
my @s = sort $deduper->find_similar( $c ); |
41 |
|
if ( @s ) { |
42 |
|
#warn " similar = ",dump( @s ); |
43 |
|
|
44 |
|
foreach my $id ( @s ) { |
45 |
|
next if $id == $i->id; # keep current |
46 |
|
my $si = Grep::Model::Item->new(); |
47 |
|
$si->load( $id ) or die "can't find similar item $id"; |
48 |
|
print " -$id-"; |
49 |
|
$si->delete; |
50 |
|
$duplicates++; |
51 |
|
} |
52 |
|
} |
53 |
|
|
54 |
|
$deduper->add_doc( $i->id, $c ); |
55 |
|
} |
56 |
|
|
57 |
|
$search->add( $i, $i->in_feed->owner->id ); |
58 |
|
print ' '; |
59 |
|
$total++; |
60 |
} |
} |
61 |
|
|
62 |
print "\n"; |
print "$total records indexed", $remove_duplicates ? " ($duplicates duplicates)" : "", "\n"; |
63 |
|
|
64 |
Grep::Search->finish; |
$search->finish; |