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_duplicate = 1; |
16 |
|
|
17 |
|
$|=1; |
18 |
|
|
19 |
BEGIN { Jifty->new; }; |
BEGIN { Jifty->new; }; |
20 |
|
|
23 |
|
|
24 |
print "indexing ", $coll->count, " items "; |
print "indexing ", $coll->count, " items "; |
25 |
|
|
26 |
|
my $search = Grep::Search->new(); |
27 |
|
my $deduper = Text::DeDuper->new(); |
28 |
|
|
29 |
|
my ( $total, $duplicates ) = ( 0, 0 ); |
30 |
|
|
31 |
while ( my $i = $coll->next ) { |
while ( my $i = $coll->next ) { |
32 |
|
|
33 |
Grep::Search->add( $i, $i->in_feed->owner->id ); |
my $c = encode('utf-8', $i->content); |
34 |
|
|
35 |
print $i->id, ' '; |
if ( $remove_duplicate && $deduper->find_similar( $c ) ) { |
36 |
|
$i->delete; |
37 |
|
print "-",$i->id,"- "; |
38 |
|
$duplicates++; |
39 |
|
} else { |
40 |
|
$search->add( $i, $i->in_feed->owner->id ); |
41 |
|
print $i->id; |
42 |
|
$deduper->add_doc( $i->id, $c ); |
43 |
|
print ' '; |
44 |
|
} |
45 |
|
$total++; |
46 |
} |
} |
47 |
|
|
48 |
print "\n"; |
print "$total records indexed", $remove_duplicate ? " ($duplicates duplicates)" : "", "\n"; |
49 |
|
|
50 |
Grep::Search->finish; |
$search->finish; |