WebPAC/Output/SWISH.pm

package WebPAC::Output::SWISH;

use warnings;
use strict;

use lib 'lib';

use base qw/WebPAC::Common Class::Accessor/;
__PACKAGE__->mk_accessors(qw(
        database
        input
        type

        index_path
));

use File::Path qw/mkpath/;
use Data::Dump qw/dump/;
use YAML;
use JSON;


=head1 NAME

WebPAC::Output::SWISH - Create swish-e full text index

=cut

our $VERSION = '0.01';

=head1 SYNOPSIS

Create full text index using swish-e indexer from data with
type C<search>.

=head1 FUNCTIONS

=head2 new

 my $out = new WebPAC::Output::SWISH({
        database => 'demo',
 });

Options are:

=over 4

=item database

name of database from which data comes

=back

Name of database will be used to form URI of documents in index.

=cut

our $dir = 'var/swish';

sub init {
        my $self = shift;

        my $log = $self->_get_logger;

        my $database = $self->database || $log->logdie("need database");

        mkpath $dir if ! -e $dir;

        my $path = "$dir/$database.conf";

        open(my $conf, '>', $path) || die "can't open $path: $!";

        print $conf <<"DEFAULT_SWISH_CONF";
# swish-e config file for $database

IndexDir stdin

# input file definition
DefaultContents XML*

# indexed metatags
MetaNames xml swishdocpath


#XMLClassAttributes type
UndefinedMetaTags auto
UndefinedXMLAttributes auto

IndexFile $dir/$database

# Croatian ISO-8859-2 characters to unaccented equivalents
#TranslateCharacters ¹©ðÐèÈæÆ¾® ssddcccczz

# store data into index
PropertyNames data

# disable output
ParserWarnLevel 0
IndexReport 1

DEFAULT_SWISH_CONF

        close($conf) || die "can't write config $path: $!";

        $self->index_path( "$dir/$database" );

        my $swish = "swish-e -S prog -c $path";
        open( $self->{_swish_fh}, '|-', $swish ) || die "can't open pipe to $swish: $!";

        $log->info( "created $path ", -s $path, " bytes for ", $self->index_path );

        $self->{stats} = {};

        $self ? return $self : return undef;
}

=head2

  my $path = $out->index_path;

=head2 add

  $out->add( 42, $ds );

=cut

my %escape = ('<'=>'&lt;', '>'=>'&gt;', '&'=>'&amp;', '"'=>'&quot;');
my $escape_re  = join '|' => keys %escape;

sub add {
        my ($self,$id,$ds) = @_;

        die "need input" unless $self->input;

        my $log = $self->_get_logger;
        $log->debug("id: $id ds = ",sub { dump($ds) });

        my $database = $self->database || $log->logconfess('no database in $self');

        my $uri = $self->database . '/' . $self->input . "/$id";
        $log->debug("creating $uri");

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

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

        return unless (@tags);

        my $xml = qq{<all>};

        $xml .= "<$_>" . $self->$_ . "</$_>" foreach ( 'database', 'input' );

        my $data;

        foreach my $tag (@tags) {

                my $r = ref $ds->{$tag}->{$type};
                die "tag $tag type $type not ARRAY but '$r' = ",dump( $ds->{$tag}->{$type} ) unless $r eq 'ARRAY';

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

                next if ! $vals;

                $vals =~ s/($escape_re)/$escape{$1}/gs;
                # BW & EW are our markers for tag boundry
                $xml .= qq{<$tag><![CDATA[BW $vals EW]]></$tag>};

                $self->{stats}->{attr}->{$tag}++;
                $self->{stats}->{input}->{ $self->input }->{$tag}++;

                $data->{$tag} = $vals;
        }

        # serialize to JSON instead of YAML because we will loose whitespace
        $data = to_json($data);
        $xml .= qq{<data><![CDATA[$data]]></data>};

        $xml .= qq{</all>\n};

        my $len = length($xml);

        my $fh = $self->{_swish_fh} || die "_swish_fh missing";

        print $fh "Path-Name: $uri\nContent-Length: $len\nDocument-Type: XML\n\n$xml" or
                die "can't add $uri: $@\n$xml";

        $log->debug( $xml );

        return 1;
}

=head2 finish

Dump attributes used on disk

=cut

sub finish {
        my $self = shift;
        my $log = $self->_get_logger();

        my $path = $dir . '/' . $self->{database} . '.yaml';
        YAML::DumpFile( $path, $self->{stats} );
        $log->info("created  $path ", -s $path, " bytes");
        $log->debug( dump( $self->{stats} ) );

        close( $self->{_swish_fh} ) || die "can't close index ", $self->index_path, ": $!";
}

=head1 AUTHOR

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

=head1 COPYRIGHT & LICENSE

Copyright 2004-2009 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;
1	dpavlin	1145	package WebPAC::Output::SWISH;
2
3			use warnings;
4			use strict;
5
6	dpavlin	1147	use lib 'lib';
7	dpavlin	1145
8	dpavlin	1147	use base qw/WebPAC::Common Class::Accessor/;
9			__PACKAGE__->mk_accessors(qw(
10			database
11	dpavlin	1156	input
12	dpavlin	1147	type
13
14			index_path
15			));
16
17	dpavlin	1151	use File::Path qw/mkpath/;
18	dpavlin	1145	use Data::Dump qw/dump/;
19			use YAML;
20	dpavlin	1149	use JSON;
21	dpavlin	1145
22	dpavlin	1147
23	dpavlin	1145	=head1 NAME
24
25			WebPAC::Output::SWISH - Create swish-e full text index
26
27			=cut
28
29			our $VERSION = '0.01';
30
31			=head1 SYNOPSIS
32
33			Create full text index using swish-e indexer from data with
34			type C<search>.
35
36			=head1 FUNCTIONS
37
38			=head2 new
39
40	dpavlin	1147	my $out = new WebPAC::Output::SWISH({
41	dpavlin	1145	database => 'demo',
42	dpavlin	1147	});
43	dpavlin	1145
44			Options are:
45
46			=over 4
47
48			=item database
49
50			name of database from which data comes
51
52			=back
53
54			Name of database will be used to form URI of documents in index.
55
56			=cut
57
58			our $dir = 'var/swish';
59
60	dpavlin	1147	sub init {
61			my $self = shift;
62	dpavlin	1145
63			my $log = $self->_get_logger;
64
65	dpavlin	1147	my $database = $self->database \|\| $log->logdie("need database");
66	dpavlin	1145
67	dpavlin	1151	mkpath $dir if ! -e $dir;
68	dpavlin	1145
69			my $path = "$dir/$database.conf";
70
71			open(my $conf, '>', $path) \|\| die "can't open $path: $!";
72
73			print $conf <<"DEFAULT_SWISH_CONF";
74			# swish-e config file for $database
75
76			IndexDir stdin
77
78			# input file definition
79			DefaultContents XML*
80
81			# indexed metatags
82			MetaNames xml swishdocpath
83
84
85			#XMLClassAttributes type
86			UndefinedMetaTags auto
87			UndefinedXMLAttributes auto
88
89			IndexFile $dir/$database
90
91			# Croatian ISO-8859-2 characters to unaccented equivalents
92	dpavlin	1147	#TranslateCharacters ¹©ðÐèÈæÆ¾® ssddcccczz
93	dpavlin	1145
94	dpavlin	1149	# store data into index
95			PropertyNames data
96	dpavlin	1145
97			# disable output
98			ParserWarnLevel 0
99			IndexReport 1
100
101			DEFAULT_SWISH_CONF
102
103			close($conf) \|\| die "can't write config $path: $!";
104
105	dpavlin	1147	$self->index_path( "$dir/$database" );
106	dpavlin	1145
107			my $swish = "swish-e -S prog -c $path";
108	dpavlin	1147	open( $self->{_swish_fh}, '\|-', $swish ) \|\| die "can't open pipe to $swish: $!";
109	dpavlin	1145
110	dpavlin	1147	$log->info( "created $path ", -s $path, " bytes for ", $self->index_path );
111
112	dpavlin	1145	$self->{stats} = {};
113
114			$self ? return $self : return undef;
115			}
116
117			=head2
118
119			my $path = $out->index_path;
120
121			=head2 add
122
123	dpavlin	1147	$out->add( 42, $ds );
124	dpavlin	1145
125			=cut
126
127			my %escape = ('<'=>'<', '>'=>'>', '&'=>'&', '"'=>'"');
128			my $escape_re = join '\|' => keys %escape;
129
130			sub add {
131	dpavlin	1147	my ($self,$id,$ds) = @_;
132	dpavlin	1145
133	dpavlin	1165	die "need input" unless $self->input;
134
135	dpavlin	1145	my $log = $self->_get_logger;
136	dpavlin	1147	$log->debug("id: $id ds = ",sub { dump($ds) });
137	dpavlin	1145
138	dpavlin	1147	my $database = $self->database \|\| $log->logconfess('no database in $self');
139	dpavlin	1145
140	dpavlin	1156	my $uri = $self->database . '/' . $self->input . "/$id";
141	dpavlin	1145	$log->debug("creating $uri");
142
143			# filter all tags which have type defined
144	dpavlin	1147	my $type = $self->type \|\| 'search';
145	dpavlin	1145	my @tags = grep {
146	dpavlin	1147	ref($ds->{$_}) eq 'HASH' && defined( $ds->{$_}->{$type} )
147			} keys %{ $ds };
148	dpavlin	1145
149			$log->debug("tags = ", join(",", @tags));
150
151			return unless (@tags);
152
153	dpavlin	1161	my $xml = qq{<all>};
154	dpavlin	1145
155	dpavlin	1156	$xml .= "<$_>" . $self->$_ . "</$_>" foreach ( 'database', 'input' );
156
157	dpavlin	1149	my $data;
158
159	dpavlin	1145	foreach my $tag (@tags) {
160
161	dpavlin	1147	my $r = ref $ds->{$tag}->{$type};
162			die "tag $tag type $type not ARRAY but '$r' = ",dump( $ds->{$tag}->{$type} ) unless $r eq 'ARRAY';
163	dpavlin	1145
164	dpavlin	1147	my $vals = join(" ", @{ $ds->{$tag}->{$type} });
165
166	dpavlin	1145	next if ! $vals;
167
168			$vals =~ s/($escape_re)/$escape{$1}/gs;
169	dpavlin	1154	# BW & EW are our markers for tag boundry
170			$xml .= qq{<$tag><![CDATA[BW $vals EW]]></$tag>};
171	dpavlin	1145
172			$self->{stats}->{attr}->{$tag}++;
173	dpavlin	1165	$self->{stats}->{input}->{ $self->input }->{$tag}++;
174	dpavlin	1149
175			$data->{$tag} = $vals;
176	dpavlin	1145	}
177
178	dpavlin	1149	# serialize to JSON instead of YAML because we will loose whitespace
179			$data = to_json($data);
180			$xml .= qq{<data><![CDATA[$data]]></data>};
181
182	dpavlin	1161	$xml .= qq{</all>\n};
183	dpavlin	1145
184			my $len = length($xml);
185
186	dpavlin	1147	my $fh = $self->{_swish_fh} \|\| die "_swish_fh missing";
187	dpavlin	1145
188	dpavlin	1147	print $fh "Path-Name: $uri\nContent-Length: $len\nDocument-Type: XML\n\n$xml" or
189			die "can't add $uri: $@\n$xml";
190	dpavlin	1145
191	dpavlin	1154	$log->debug( $xml );
192	dpavlin	1149
193	dpavlin	1145	return 1;
194			}
195
196			=head2 finish
197
198			Dump attributes used on disk
199
200			=cut
201
202			sub finish {
203			my $self = shift;
204			my $log = $self->_get_logger();
205
206			my $path = $dir . '/' . $self->{database} . '.yaml';
207			YAML::DumpFile( $path, $self->{stats} );
208			$log->info("created $path ", -s $path, " bytes");
209			$log->debug( dump( $self->{stats} ) );
210	dpavlin	1147
211			close( $self->{_swish_fh} ) \|\| die "can't close index ", $self->index_path, ": $!";
212	dpavlin	1145	}
213
214			=head1 AUTHOR
215
216			Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >>
217
218			=head1 COPYRIGHT & LICENSE
219
220			Copyright 2004-2009 Dobrica Pavlinusic, All Rights Reserved.
221
222			This program is free software; you can redistribute it and/or modify it
223			under the same terms as Perl itself.
224
225			=cut
226
227			1;