WebPAC/Output/KinoSearch.pm

package WebPAC::Output::KinoSearch;

use warnings;
use strict;

use base qw/WebPAC::Common/;

use KinoSearch::InvIndexer;
use KinoSearch::Analysis::PolyAnalyzer;
use Encode qw/from_to/;
use Data::Dump qw/dump/;
use Storable;

=head1 NAME

WebPAC::Output::KinoSearch - Create KinoSearch full text index

=head1 VERSION

Version 0.03

=cut

our $VERSION = '0.03';

=head1 SYNOPSIS

Create full text index using KinoSearch index from data with
type C<search>.

=head1 FUNCTIONS

=head2 new

Open KinoSearch index

 my $est = new WebPAC::Output::KinoSearch(
        index_path => '/path/to/invindex',
        fields => qw/name of all filelds used/,
        database => 'demo',
        label => 'node label',
        encoding => 'iso-8859-2',
        clean => 1,
 );

Options are:

=over 4

=item index_path

path to KinoSearch index to use

=item fields

name of all fields used in this index

=item database

name of database from which data comes

=item label

label for node (optional)

=item encoding

character encoding of C<data_structure> if it's differenet than C<ISO-8859-2>
(and it probably is). This encoding will be converted to C<UTF-8> for
index.

=back

=cut

sub new {
        my $class = shift;
        my $self = {@_};
        bless($self, $class);

        my $log = $self->_get_logger;

        #$log->debug("self: ", sub { dump($self) });

        foreach my $p (qw/index_path fields database/) {
                $log->logdie("need $p") unless ($self->{$p});
        }

        $log->logdie("fields is not ARRAY") unless (ref($self->{fields}) eq 'ARRAY');

        $self->{encoding} ||= 'ISO-8859-2';

        $self->{index_path} .= '/' . $self->{database};

        $self->{clean} = 1 if (! -e $self->{index_path} . '/segments');

        $log->info("using", $self->{clean} ? ' new' : '', " index $self->{index_path} with encoding $self->{encoding}");

        my $analyzer = KinoSearch::Analysis::PolyAnalyzer->new( language => 'en' );

        $self->{invindex} = KinoSearch::InvIndexer->new(
                invindex => $self->{index_path},
                create   => $self->{clean},
                analyzer => $analyzer,
        );

        my $fields_path = $self->{index_path} . '/fields.storable';
        $fields_path =~ s#//#/#g;
        if (-e $fields_path) {
                $self->{fields} = retrieve($fields_path) ||
                        $log->warn("can't open $fields_path: $!");
        } else {
                $log->error("This will be dummy run since no fields statistics are found!");
                $log->error("You will have to re-run indexing to get search results!");
                $self->{dummy_run} = 1;
        }
        $self->{fields_path} = $fields_path;

        foreach my $f (@{ $self->{fields} }) {
                $self->{invindex}->spec_field( 
                        name  => $f,
#                       boost => 10,
                        stored => 1,
                        indexed => 1,
                        vectorized => 0,
                );
        }

        $self ? return $self : return undef;
}


=head2 add

Adds one entry to database.

  $est->add(
        id => 42,
        ds => $ds,
        type => 'display',
        text => 'optional text from which snippet is created',
  );

This function will create  entries in index using following URI format:

  C<file:///type/database%20name/000>

Each tag in C<data_structure> with specified C<type> will create one
attribute and corresponding hidden text (used for search).

=cut

sub add {
        my $self = shift;

        my $args = {@_};

        my $log = $self->_get_logger;

        my $database = $self->{'database'} || $log->logconfess('no database in $self');
        $log->logconfess('need invindex in object') unless ($self->{'invindex'});

        foreach my $p (qw/id ds type/) {
                $log->logdie("need $p") unless ($args->{$p});
        }

        my $type = $args->{'type'};
        my $id = $args->{'id'};

        my $uri = "file:///$type/$database/$id";
        $log->debug("creating $uri");

        my $doc = $self->{invindex}->new_doc( $uri ) || $log->logdie("can't create new_doc( $uri )");

        sub _add_value($$$$$) {
                my ($self,$log,$doc,$n,$v) = @_;
                return unless ($v);

                $self->{value_usage}->{$n}++;
                return if ($self->{dummy_run});

                eval { $doc->set_value($n, $self->convert($v) ) };
                $log->warn("can't insert: $n = $v") if ($@);
        }

        _add_value($self,$log,$doc, 'uri', $uri);

        $log->debug("ds = ", sub { dump($args->{'ds'}) } );

        # filter all tags which have type defined
        my @tags = grep {
                ref($args->{'ds'}->{$_}) eq 'HASH' && defined( $args->{'ds'}->{$_}->{$type} )
        } keys %{ $args->{'ds'} };

        $log->debug("tags = ", join(",", @tags));

        return unless (@tags);

        foreach my $tag (@tags) {

                my $vals = join(" ", @{ $args->{'ds'}->{$tag}->{$type} });

                next if (! $vals);

                $vals = $self->convert( $vals ) or
                        $log->logdie("can't convert '$vals' to UTF-8");

                _add_value($self, $log, $doc, $tag, $vals );
        }

        if (my $text = $args->{'text'}) {
                _add_value($self, $log, $doc, 'bodytext', $text );
        }

        #$log->debug("adding ", sub { $doc->dump_draft } );
        $self->{invindex}->add_doc($doc) || $log->warn("can't add document $uri");

        return 1;
}

=head2 finish

Close index

 $index->finish;

=cut

sub finish {
        my $self = shift;

        my $log = $self->_get_logger();

        $log->info("finish index writing to disk");
        $self->{invindex}->finish;

        $log->info("writing value usage file");

        # add fields from last run
        map { $self->{value_usage}->{$_}++ } @{ $self->{fields} };

        my @fields = keys %{ $self->{value_usage} };
        store \@fields, $self->{fields_path} ||
                $log->warn("can't write $self->{fields_path}: $!");

}

=head2 convert

 my $utf8_string = $self->convert('string in codepage');

=cut

sub convert {
        my $self = shift;

        my $text = shift || return;
        from_to($text, $self->{encoding}, 'UTF-8');
        return $text;
}

=head1 AUTHOR

Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >>

=head1 COPYRIGHT & LICENSE

Copyright 2005 Dobrica Pavlinusic, All Rights Reserved.

This program is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.

=cut

1; # End of WebPAC::Output::Estraier
1	package WebPAC::Output::KinoSearch;
2
3	use warnings;
4	use strict;
5
6	use base qw/WebPAC::Common/;
7
8	use KinoSearch::InvIndexer;
9	use KinoSearch::Analysis::PolyAnalyzer;
10	use Encode qw/from_to/;
11	use Data::Dump qw/dump/;
12	use Storable;
13
14	=head1 NAME
15
16	WebPAC::Output::KinoSearch - Create KinoSearch full text index
17
18	=head1 VERSION
19
20	Version 0.03
21
22	=cut
23
24	our $VERSION = '0.03';
25
26	=head1 SYNOPSIS
27
28	Create full text index using KinoSearch index from data with
29	type C<search>.
30
31	=head1 FUNCTIONS
32
33	=head2 new
34
35	Open KinoSearch index
36
37	my $est = new WebPAC::Output::KinoSearch(
38	index_path => '/path/to/invindex',
39	fields => qw/name of all filelds used/,
40	database => 'demo',
41	label => 'node label',
42	encoding => 'iso-8859-2',
43	clean => 1,
44	);
45
46	Options are:
47
48	=over 4
49
50	=item index_path
51
52	path to KinoSearch index to use
53
54	=item fields
55
56	name of all fields used in this index
57
58	=item database
59
60	name of database from which data comes
61
62	=item label
63
64	label for node (optional)
65
66	=item encoding
67
68	character encoding of C<data_structure> if it's differenet than C<ISO-8859-2>
69	(and it probably is). This encoding will be converted to C<UTF-8> for
70	index.
71
72	=back
73
74	=cut
75
76	sub new {
77	my $class = shift;
78	my $self = {@_};
79	bless($self, $class);
80
81	my $log = $self->_get_logger;
82
83	#$log->debug("self: ", sub { dump($self) });
84
85	foreach my $p (qw/index_path fields database/) {
86	$log->logdie("need $p") unless ($self->{$p});
87	}
88
89	$log->logdie("fields is not ARRAY") unless (ref($self->{fields}) eq 'ARRAY');
90
91	$self->{encoding} \|\|= 'ISO-8859-2';
92
93	$self->{index_path} .= '/' . $self->{database};
94
95	$self->{clean} = 1 if (! -e $self->{index_path} . '/segments');
96
97	$log->info("using", $self->{clean} ? ' new' : '', " index $self->{index_path} with encoding $self->{encoding}");
98
99	my $analyzer = KinoSearch::Analysis::PolyAnalyzer->new( language => 'en' );
100
101	$self->{invindex} = KinoSearch::InvIndexer->new(
102	invindex => $self->{index_path},
103	create => $self->{clean},
104	analyzer => $analyzer,
105	);
106
107	my $fields_path = $self->{index_path} . '/fields.storable';
108	$fields_path =~ s#//#/#g;
109	if (-e $fields_path) {
110	$self->{fields} = retrieve($fields_path) \|\|
111	$log->warn("can't open $fields_path: $!");
112	} else {
113	$log->error("This will be dummy run since no fields statistics are found!");
114	$log->error("You will have to re-run indexing to get search results!");
115	$self->{dummy_run} = 1;
116	}
117	$self->{fields_path} = $fields_path;
118
119	foreach my $f (@{ $self->{fields} }) {
120	$self->{invindex}->spec_field(
121	name => $f,
122	# boost => 10,
123	stored => 1,
124	indexed => 1,
125	vectorized => 0,
126	);
127	}
128
129	$self ? return $self : return undef;
130	}
131
132
133	=head2 add
134
135	Adds one entry to database.
136
137	$est->add(
138	id => 42,
139	ds => $ds,
140	type => 'display',
141	text => 'optional text from which snippet is created',
142	);
143
144	This function will create entries in index using following URI format:
145
146	C<file:///type/database%20name/000>
147
148	Each tag in C<data_structure> with specified C<type> will create one
149	attribute and corresponding hidden text (used for search).
150
151	=cut
152
153	sub add {
154	my $self = shift;
155
156	my $args = {@_};
157
158	my $log = $self->_get_logger;
159
160	my $database = $self->{'database'} \|\| $log->logconfess('no database in $self');
161	$log->logconfess('need invindex in object') unless ($self->{'invindex'});
162
163	foreach my $p (qw/id ds type/) {
164	$log->logdie("need $p") unless ($args->{$p});
165	}
166
167	my $type = $args->{'type'};
168	my $id = $args->{'id'};
169
170	my $uri = "file:///$type/$database/$id";
171	$log->debug("creating $uri");
172
173	my $doc = $self->{invindex}->new_doc( $uri ) \|\| $log->logdie("can't create new_doc( $uri )");
174
175	sub _add_value($$$$$) {
176	my ($self,$log,$doc,$n,$v) = @_;
177	return unless ($v);
178
179	$self->{value_usage}->{$n}++;
180	return if ($self->{dummy_run});
181
182	eval { $doc->set_value($n, $self->convert($v) ) };
183	$log->warn("can't insert: $n = $v") if ($@);
184	}
185
186	_add_value($self,$log,$doc, 'uri', $uri);
187
188	$log->debug("ds = ", sub { dump($args->{'ds'}) } );
189
190	# filter all tags which have type defined
191	my @tags = grep {
192	ref($args->{'ds'}->{$_}) eq 'HASH' && defined( $args->{'ds'}->{$_}->{$type} )
193	} keys %{ $args->{'ds'} };
194
195	$log->debug("tags = ", join(",", @tags));
196
197	return unless (@tags);
198
199	foreach my $tag (@tags) {
200
201	my $vals = join(" ", @{ $args->{'ds'}->{$tag}->{$type} });
202
203	next if (! $vals);
204
205	$vals = $self->convert( $vals ) or
206	$log->logdie("can't convert '$vals' to UTF-8");
207
208	_add_value($self, $log, $doc, $tag, $vals );
209	}
210
211	if (my $text = $args->{'text'}) {
212	_add_value($self, $log, $doc, 'bodytext', $text );
213	}
214
215	#$log->debug("adding ", sub { $doc->dump_draft } );
216	$self->{invindex}->add_doc($doc) \|\| $log->warn("can't add document $uri");
217
218	return 1;
219	}
220
221	=head2 finish
222
223	Close index
224
225	$index->finish;
226
227	=cut
228
229	sub finish {
230	my $self = shift;
231
232	my $log = $self->_get_logger();
233
234	$log->info("finish index writing to disk");
235	$self->{invindex}->finish;
236
237	$log->info("writing value usage file");
238
239	# add fields from last run
240	map { $self->{value_usage}->{$_}++ } @{ $self->{fields} };
241
242	my @fields = keys %{ $self->{value_usage} };
243	store \@fields, $self->{fields_path} \|\|
244	$log->warn("can't write $self->{fields_path}: $!");
245
246	}
247
248	=head2 convert
249
250	my $utf8_string = $self->convert('string in codepage');
251
252	=cut
253
254	sub convert {
255	my $self = shift;
256
257	my $text = shift \|\| return;
258	from_to($text, $self->{encoding}, 'UTF-8');
259	return $text;
260	}
261
262	=head1 AUTHOR
263
264	Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >>
265
266	=head1 COPYRIGHT & LICENSE
267
268	Copyright 2005 Dobrica Pavlinusic, All Rights Reserved.
269
270	This program is free software; you can redistribute it and/or modify it
271	under the same terms as Perl itself.
272
273	=cut
274
275	1; # End of WebPAC::Output::Estraier