webpac2/trunk/run.pl

#!/usr/bin/perl -w

use strict;

use Cwd qw/abs_path/;
use File::Temp qw/tempdir/;
use Data::Dumper;
use lib './lib';

use WebPAC::Common 0.02;
use WebPAC::Lookup;
use WebPAC::Input 0.03;
use WebPAC::Store 0.03;
use WebPAC::Normalize::XML;
use WebPAC::Output::TT;
use WebPAC::Output::Estraier 0.05;
use YAML qw/LoadFile/;
use Getopt::Long;
use File::Path;

=head1 NAME

run.pl - start WebPAC indexing

B<this command will probably go away. Don't get used to it!>

Options:

=over 4

=item --offset 42

start loading (all) databases at offset 42

=item --limit 100

limit loading to 100 records

=item --clean

remove database and Hyper Estraier index before indexing

=item --config conf/config.yml

path to YAML configuration file

=back

=cut

my $offset;
my $limit;

my $clean = 0;
my $config = 'conf/config.yml';
my $debug = 0;

GetOptions(
        "limit=i" => \$limit,
        "offset=i" => \$offset,
        "clean" => \$clean,
        "config" => \$config,
        "debug" => \$debug,
);

$config = LoadFile($config);

print "config = ",Dumper($config) if ($debug);

die "no databases in config file!\n" unless ($config->{databases});

my $total_rows = 0;

while (my ($database, $db_config) = each %{ $config->{databases} }) {

        my $log = _new WebPAC::Common()->_get_logger();

        #
        # open Hyper Estraier database
        #

        my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration");
        $est_config->{database} = $database;

        my $est = new WebPAC::Output::Estraier(
                %{ $est_config },
        );

        if ($clean) {
                $log->warn("creating new empty index $database");
                $est->master( action => 'nodedel', name => $database );
                $est->master( action => 'nodeadd', name => $database, label => $database );
        }

        #
        # now WebPAC::Store
        #
        my $abs_path = abs_path($0);
        $abs_path =~ s#/[^/]*$#/#;

        my $db_path = $config->{webpac}->{db_path} . '/' . $database;

        if ($clean) {
                $log->info("creating new database $database in $db_path");
                rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
        } else {
                $log->info("working on $database in $db_path");
        }

        my $db = new WebPAC::Store(
                path => $db_path,
                database => $database,
                debug => $debug,
        );


        #
        # now, iterate through input formats
        #

        my @inputs;
        if (ref($db_config->{input}) eq 'ARRAY') {
                @inputs = @{ $db_config->{input} };
        } elsif ($db_config->{input}) {
                push @inputs, $db_config->{input};
        } else {
                $log->info("database $database doesn't have inputs defined");
        }

        my @supported_inputs = keys %{ $config->{webpac}->{inputs} };

        foreach my $input (@inputs) {

                my $type = lc($input->{type});

                die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));

                my $lookup = new WebPAC::Lookup(
                        lookup_file => $input->{lookup},
                );

                my $input_module = $config->{webpac}->{inputs}->{$type};

                $log->info("working on input $input->{path} [$input->{type}] using $input_module");

                my $input_db = new WebPAC::Input(
                        module => $input_module,
                        code_page => $config->{webpac}->{webpac_encoding},
                        limit => $limit || $input->{limit},
                        offset => $offset,
                        lookup => $lookup,
                );
                $log->logdie("can't create input using $input_module") unless ($input);

                my $maxmfn = $input_db->open(
                        path => $input->{path},
                        code_page => $input->{encoding},        # database encoding
                );

                my $n = new WebPAC::Normalize::XML(
                #       filter => { 'foo' => sub { shift } },
                        db => $db,
                        lookup_regex => $lookup->regex,
                        lookup => $lookup,
                        prefix => $input->{name},
                );

                my $normalize_path = $input->{normalize}->{path};

                if ($normalize_path =~ m/\.xml$/i) {
                        $n->open(
                                tag => $input->{normalize}->{tag},
                                xml_file => $input->{normalize}->{path},
                        );
                } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
                        $n->open_yaml(
                                path => $normalize_path,
                                tag => $input->{normalize}->{tag},
                        );
                }

                foreach my $pos ( 0 ... $input_db->size ) {

                        my $row = $input_db->fetch || next;

                        my $mfn = $row->{'000'}->[0];

                        if (! $mfn || $mfn !~ m#^\d+$#) {
                                $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
                                $mfn = $pos;
                                push @{ $row->{'000'} }, $pos;
                        }

                        my $ds = $n->data_structure($row);

                        $est->add(
                                id => $input->{name} . "/" . $mfn,
                                ds => $ds,
                                type => $config->{hyperestraier}->{type},
                        );

                        $total_rows++;
                }

        };

        $log->info("$total_rows records indexed");

        #
        # add Hyper Estraier links to other databases
        #
        if (ref($db_config->{links}) eq 'ARRAY') {
                foreach my $link (@{ $db_config->{links} }) {
                        $log->info("adding link $database -> $link->{to} [$link->{credit}]");
                        $est->add_link(
                                from => $database,
                                to => $link->{to},
                                credit => $link->{credit},
                        );
                }
        }

}

1	#!/usr/bin/perl -w
2
3	use strict;
4
5	use Cwd qw/abs_path/;
6	use File::Temp qw/tempdir/;
7	use Data::Dumper;
8	use lib './lib';
9
10	use WebPAC::Common 0.02;
11	use WebPAC::Lookup;
12	use WebPAC::Input 0.03;
13	use WebPAC::Store 0.03;
14	use WebPAC::Normalize::XML;
15	use WebPAC::Output::TT;
16	use WebPAC::Output::Estraier 0.05;
17	use YAML qw/LoadFile/;
18	use Getopt::Long;
19	use File::Path;
20
21	=head1 NAME
22
23	run.pl - start WebPAC indexing
24
25	B<this command will probably go away. Don't get used to it!>
26
27	Options:
28
29	=over 4
30
31	=item --offset 42
32
33	start loading (all) databases at offset 42
34
35	=item --limit 100
36
37	limit loading to 100 records
38
39	=item --clean
40
41	remove database and Hyper Estraier index before indexing
42
43	=item --config conf/config.yml
44
45	path to YAML configuration file
46
47	=back
48
49	=cut
50
51	my $offset;
52	my $limit;
53
54	my $clean = 0;
55	my $config = 'conf/config.yml';
56	my $debug = 0;
57
58	GetOptions(
59	"limit=i" => \$limit,
60	"offset=i" => \$offset,
61	"clean" => \$clean,
62	"config" => \$config,
63	"debug" => \$debug,
64	);
65
66	$config = LoadFile($config);
67
68	print "config = ",Dumper($config) if ($debug);
69
70	die "no databases in config file!\n" unless ($config->{databases});
71
72	my $total_rows = 0;
73
74	while (my ($database, $db_config) = each %{ $config->{databases} }) {
75
76	my $log = _new WebPAC::Common()->_get_logger();
77
78	#
79	# open Hyper Estraier database
80	#
81
82	my $est_config = $config->{hyperestraier} \|\| $log->logdie("can't find 'hyperestraier' part in confguration");
83	$est_config->{database} = $database;
84
85	my $est = new WebPAC::Output::Estraier(
86	%{ $est_config },
87	);
88
89	if ($clean) {
90	$log->warn("creating new empty index $database");
91	$est->master( action => 'nodedel', name => $database );
92	$est->master( action => 'nodeadd', name => $database, label => $database );
93	}
94
95	#
96	# now WebPAC::Store
97	#
98	my $abs_path = abs_path($0);
99	$abs_path =~ s#/[^/]*$#/#;
100
101	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
102
103	if ($clean) {
104	$log->info("creating new database $database in $db_path");
105	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
106	} else {
107	$log->info("working on $database in $db_path");
108	}
109
110	my $db = new WebPAC::Store(
111	path => $db_path,
112	database => $database,
113	debug => $debug,
114	);
115
116
117	#
118	# now, iterate through input formats
119	#
120
121	my @inputs;
122	if (ref($db_config->{input}) eq 'ARRAY') {
123	@inputs = @{ $db_config->{input} };
124	} elsif ($db_config->{input}) {
125	push @inputs, $db_config->{input};
126	} else {
127	$log->info("database $database doesn't have inputs defined");
128	}
129
130	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
131
132	foreach my $input (@inputs) {
133
134	my $type = lc($input->{type});
135
136	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
137
138	my $lookup = new WebPAC::Lookup(
139	lookup_file => $input->{lookup},
140	);
141
142	my $input_module = $config->{webpac}->{inputs}->{$type};
143
144	$log->info("working on input $input->{path} [$input->{type}] using $input_module");
145
146	my $input_db = new WebPAC::Input(
147	module => $input_module,
148	code_page => $config->{webpac}->{webpac_encoding},
149	limit => $limit \|\| $input->{limit},
150	offset => $offset,
151	lookup => $lookup,
152	);
153	$log->logdie("can't create input using $input_module") unless ($input);
154
155	my $maxmfn = $input_db->open(
156	path => $input->{path},
157	code_page => $input->{encoding}, # database encoding
158	);
159
160	my $n = new WebPAC::Normalize::XML(
161	# filter => { 'foo' => sub { shift } },
162	db => $db,
163	lookup_regex => $lookup->regex,
164	lookup => $lookup,
165	prefix => $input->{name},
166	);
167
168	my $normalize_path = $input->{normalize}->{path};
169
170	if ($normalize_path =~ m/\.xml$/i) {
171	$n->open(
172	tag => $input->{normalize}->{tag},
173	xml_file => $input->{normalize}->{path},
174	);
175	} elsif ($normalize_path =~ m/\.(?:yml\|yaml)$/i) {
176	$n->open_yaml(
177	path => $normalize_path,
178	tag => $input->{normalize}->{tag},
179	);
180	}
181
182	foreach my $pos ( 0 ... $input_db->size ) {
183
184	my $row = $input_db->fetch \|\| next;
185
186	my $mfn = $row->{'000'}->[0];
187
188	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
189	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
190	$mfn = $pos;
191	push @{ $row->{'000'} }, $pos;
192	}
193
194	my $ds = $n->data_structure($row);
195
196	$est->add(
197	id => $input->{name} . "/" . $mfn,
198	ds => $ds,
199	type => $config->{hyperestraier}->{type},
200	);
201
202	$total_rows++;
203	}
204
205	};
206
207	$log->info("$total_rows records indexed");
208
209	#
210	# add Hyper Estraier links to other databases
211	#
212	if (ref($db_config->{links}) eq 'ARRAY') {
213	foreach my $link (@{ $db_config->{links} }) {
214	$log->info("adding link $database -> $link->{to} [$link->{credit}]");
215	$est->add_link(
216	from => $database,
217	to => $link->{to},
218	credit => $link->{credit},
219	);
220	}
221	}
222
223	}
224