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
        type

        index_path
));

use File::Path qw/make_path/;
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");

        make_path $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) = @_;

        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 = $self->database . "/$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{<xml>};

        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;
                $xml .= qq{<$tag><![CDATA[$vals]]></$tag>};

                $self->{stats}->{attr}->{$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{</xml>\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";

        warn "$xml\n";

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