1 |
#!/usr/bin/perl -w |
2 |
|
3 |
use strict; |
4 |
use Semantic::API; |
5 |
use DBI; |
6 |
use Data::Dump qw/dump/; |
7 |
use EPrints qw/_x/; |
8 |
use lib '/home/dpavlin/stem-hr/'; |
9 |
use StemHR; |
10 |
|
11 |
warn dump( StemHR->stem('kuæni') ); |
12 |
|
13 |
my $debug = shift @ARGV; |
14 |
|
15 |
my $dbh = EPrints->dbh; |
16 |
my $sth = $dbh->prepare(qq{ |
17 |
SELECT |
18 |
archive_title.eprintid as id, |
19 |
title |
20 |
FROM archive_title |
21 |
}) || die $dbh->errstr(); |
22 |
$sth->execute() || die $sth->errstr(); |
23 |
|
24 |
my $indexer = Semantic::API::Index->new( |
25 |
storage => 'sqlite', |
26 |
database => 'eprints.db', |
27 |
collection => 'EPrints' |
28 |
); |
29 |
|
30 |
|
31 |
$indexer->add_word_filters( minimum_length => 3, |
32 |
too_many_numbers => 10, |
33 |
maximum_word_length => 15 ); |
34 |
|
35 |
# use this encoding for any incoming text |
36 |
$indexer->set_default_encoding( "utf8"); |
37 |
|
38 |
my $total = 0; |
39 |
|
40 |
while (my $row = $sth->fetchrow_hashref ) { |
41 |
EPrints->id( $row->{id} ); |
42 |
my ( $title, $keywords, $abstract ) = ( |
43 |
_x( $row->{title} ), |
44 |
EPrints->lookup( 'keywords' ), |
45 |
EPrints->lookup( 'abstract' ) |
46 |
); |
47 |
my @body = split( /\W*\s+\W*/, "$title $title $title $keywords $keywords $abstract" ); |
48 |
my $body; |
49 |
foreach my $word ( @body ) { |
50 |
$body .= StemHR->stem( $word ) . ' '; |
51 |
} |
52 |
|
53 |
warn "body: $body\n" if $debug; |
54 |
|
55 |
$indexer->index( $row->{id}, join(" ", @body, $body ) ); |
56 |
$total++; |
57 |
print STDERR _x( $row->{id}, " ", $row->{title} ), "\n"; |
58 |
} |
59 |
|
60 |
|
61 |
print STDERR "\nNow adding $total items to the database..."; |
62 |
$indexer->finish(); |
63 |
print STDERR "done!\n"; |
64 |
|