6 use File::Temp qw/tempdir/;
10 use WebPAC::Common 0.02;
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 0.08;
17 use YAML qw/LoadFile/;
20 use Time::HiRes qw/time/;
24 run.pl - start WebPAC indexing
26 B<this command will probably go away. Don't get used to it!>
34 start loading (all) databases at offset 42
38 limit loading to 100 records
42 remove database and Hyper Estraier index before indexing
44 =item --one=database_name
46 reindex just single database
48 =item --config conf/config.yml
50 path to YAML configuration file
60 my $config = 'conf/config.yml';
66 "offset=i" => \$offset,
68 "one=s" => \$one_db_name,
73 $config = LoadFile($config);
75 print "config = ",Dumper($config) if ($debug);
77 die "no databases in config file!\n" unless ($config->{databases});
82 while (my ($database, $db_config) = each %{ $config->{databases} }) {
84 next if ($one_db_name && $database !~ m/$one_db_name/i);
86 my $log = _new WebPAC::Common()->_get_logger();
89 # open Hyper Estraier database
92 my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration");
93 $est_config->{database} = $database;
94 $est_config->{clean} = $clean;
96 my $est = new WebPAC::Output::Estraier( %{ $est_config } );
101 my $abs_path = abs_path($0);
102 $abs_path =~ s#/[^/]*$#/#;
104 my $db_path = $config->{webpac}->{db_path} . '/' . $database;
107 $log->info("creating new database $database in $db_path");
108 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
110 $log->info("working on $database in $db_path");
113 my $db = new WebPAC::Store(
115 database => $database,
121 # now, iterate through input formats
125 if (ref($db_config->{input}) eq 'ARRAY') {
126 @inputs = @{ $db_config->{input} };
127 } elsif ($db_config->{input}) {
128 push @inputs, $db_config->{input};
130 $log->info("database $database doesn't have inputs defined");
133 my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
135 foreach my $input (@inputs) {
137 my $type = lc($input->{type});
139 die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
141 my $lookup = new WebPAC::Lookup(
142 lookup_file => $input->{lookup},
145 my $input_module = $config->{webpac}->{inputs}->{$type};
147 $log->info("working on input $input->{path} [$input->{type}] using $input_module");
149 my $input_db = new WebPAC::Input(
150 module => $input_module,
151 code_page => $config->{webpac}->{webpac_encoding},
152 limit => $limit || $input->{limit},
156 $log->logdie("can't create input using $input_module") unless ($input);
158 my $maxmfn = $input_db->open(
159 path => $input->{path},
160 code_page => $input->{encoding}, # database encoding
163 my $n = new WebPAC::Normalize::XML(
164 # filter => { 'foo' => sub { shift } },
166 lookup_regex => $lookup->regex,
168 prefix => $input->{name},
171 my $normalize_path = $input->{normalize}->{path};
173 if ($normalize_path =~ m/\.xml$/i) {
175 tag => $input->{normalize}->{tag},
176 xml_file => $input->{normalize}->{path},
178 } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
180 path => $normalize_path,
181 tag => $input->{normalize}->{tag},
185 foreach my $pos ( 0 ... $input_db->size ) {
187 my $row = $input_db->fetch || next;
189 my $mfn = $row->{'000'}->[0];
191 if (! $mfn || $mfn !~ m#^\d+$#) {
192 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
194 push @{ $row->{'000'} }, $pos;
197 my $ds = $n->data_structure($row);
200 id => $input->{name} . "/" . $mfn,
202 type => $config->{hyperestraier}->{type},
210 my $dt = time() - $start_t;
211 $log->info("$total_rows records indexed in " .
212 sprintf("%.2f sec [%.2f rec/sec]",
213 $dt, ($total_rows / $dt)
218 # add Hyper Estraier links to other databases
220 if (ref($db_config->{links}) eq 'ARRAY') {
221 foreach my $link (@{ $db_config->{links} }) {
222 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
226 credit => $link->{credit},