Semantic-Engine/EPrints/index.pl

#!/usr/bin/perl -w

use strict;
use Semantic::API;
use Data::Dump qw/dump/;

use EPrints qw/_x/;

my $debug = shift @ARGV;
my $use = {
        score => 1,
        stem => 1,
        slogovi => 1,
};

my $dbh = EPrints->dbh;
my $sth = $dbh->prepare(qq{
SELECT
        archive_title.eprintid as id,
        title
FROM archive_title 
WHERE
        lang = 'hr'
}) || die $dbh->errstr();
$sth->execute() || die $sth->errstr();

my $indexer = Semantic::API::Index->new(
        storage => 'sqlite',
        database => 'eprints.db',
        collection => 'EPrints'
);


$indexer->add_word_filters( minimum_length => 3,
                            too_many_numbers => 10,
                            maximum_word_length => 15 );

# use this encoding for any incoming text
$indexer->set_default_encoding( "iso-8859-2" ); 

my $total = 0;

while (my $row = $sth->fetchrow_hashref ) {
        my $id = $row->{id};
        EPrints->id( $id );
        my $parts = {
                title => [ _x( $row->{title} ), 3 ],
                keywords => [ EPrints->lookup( 'keywords' ), 2 ],
                abstract => [ EPrints->lookup( 'abstract' ), 1 ],
#               content => [ EPrints->fulltext_content, 1 ],
        };

        my $skip = 0;
        foreach my $part ( qw/title keywords abstract/ ) {
                if ( ! $parts->{$part}->[0] ) {
                        warn "skipped $id doesn't have required part $part\n";
                        $skip = 1;
                        last;
                }
        }
        next if $skip;

        my $body = '';

        foreach my $part ( qw/title keywords abstract content/ ) {
                my $content = $parts->{$part}->[0] || next;

                if ( $use->{slogovi} ) {
                        $body .= ' ' . EPrints->slogovi( $content );
                }

                if ( $use->{stem} ) {
                        my $stem = EPrints->stem( $content );
                        warn "stem of '$content' didn't return anything\n" unless $stem;
                        $content = $stem;
                }


                if ( $use->{score} ) {
                        map { $body .= "$content " } 1 .. $parts->{$part}->[1];
                } else {
                        $body .= "$content ";
                }

                warn ">>> $body <<<\n" if $debug;
        }

        $indexer->index( $row->{id}, $body );
        $total++;
        print STDERR "$total: ", $row->{id}, " ", _x( $row->{title} ), " - ", length($body), " bytes\n";
}


print STDERR "\nNow adding $total items to the database...";
$indexer->finish(); 
print STDERR "done!\n"; 

1	dpavlin	1	#!/usr/bin/perl -w
2
3			use strict;
4			use Semantic::API;
5			use Data::Dump qw/dump/;
6	dpavlin	13
7	dpavlin	1	use EPrints qw/_x/;
8	dpavlin	13
9	dpavlin	1	my $debug = shift @ARGV;
10	dpavlin	17	my $use = {
11			score => 1,
12			stem => 1,
13			slogovi => 1,
14			};
15	dpavlin	1
16			my $dbh = EPrints->dbh;
17			my $sth = $dbh->prepare(qq{
18			SELECT
19			archive_title.eprintid as id,
20			title
21			FROM archive_title
22	dpavlin	17	WHERE
23			lang = 'hr'
24	dpavlin	1	}) \|\| die $dbh->errstr();
25			$sth->execute() \|\| die $sth->errstr();
26
27			my $indexer = Semantic::API::Index->new(
28			storage => 'sqlite',
29			database => 'eprints.db',
30			collection => 'EPrints'
31			);
32
33
34			$indexer->add_word_filters( minimum_length => 3,
35			too_many_numbers => 10,
36			maximum_word_length => 15 );
37
38			# use this encoding for any incoming text
39	dpavlin	8	$indexer->set_default_encoding( "iso-8859-2" );
40	dpavlin	1
41			my $total = 0;
42
43			while (my $row = $sth->fetchrow_hashref ) {
44	dpavlin	17	my $id = $row->{id};
45			EPrints->id( $id );
46	dpavlin	14	my $parts = {
47	dpavlin	17	title => [ _x( $row->{title} ), 3 ],
48			keywords => [ EPrints->lookup( 'keywords' ), 2 ],
49			abstract => [ EPrints->lookup( 'abstract' ), 1 ],
50	dpavlin	14	# content => [ EPrints->fulltext_content, 1 ],
51			};
52
53	dpavlin	17	my $skip = 0;
54			foreach my $part ( qw/title keywords abstract/ ) {
55			if ( ! $parts->{$part}->[0] ) {
56			warn "skipped $id doesn't have required part $part\n";
57			$skip = 1;
58			last;
59			}
60			}
61			next if $skip;
62
63	dpavlin	8	my $body = '';
64	dpavlin	1
65	dpavlin	14	foreach my $part ( qw/title keywords abstract content/ ) {
66	dpavlin	17	my $content = $parts->{$part}->[0] \|\| next;
67	dpavlin	8
68	dpavlin	17	if ( $use->{slogovi} ) {
69			$body .= ' ' . EPrints->slogovi( $content );
70			}
71
72			if ( $use->{stem} ) {
73			my $stem = EPrints->stem( $content );
74			warn "stem of '$content' didn't return anything\n" unless $stem;
75			$content = $stem;
76			}
77
78
79			if ( $use->{score} ) {
80			map { $body .= "$content " } 1 .. $parts->{$part}->[1];
81	dpavlin	14	} else {
82	dpavlin	17	$body .= "$content ";
83	dpavlin	14	}
84	dpavlin	17
85			warn ">>> $body <<<\n" if $debug;
86	dpavlin	14	}
87	dpavlin	1
88	dpavlin	14	$indexer->index( $row->{id}, $body );
89	dpavlin	1	$total++;
90	dpavlin	14	print STDERR "$total: ", $row->{id}, " ", _x( $row->{title} ), " - ", length($body), " bytes\n";
91	dpavlin	1	}
92
93
94			print STDERR "\nNow adding $total items to the database...";
95			$indexer->finish();
96			print STDERR "done!\n";
97