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::Normalize::Set;
16 use WebPAC::Output::TT;
17 use YAML qw/LoadFile/;
20 use Time::HiRes qw/time/;
25 run.pl - start WebPAC indexing
27 B<this command will probably go away. Don't get used to it!>
35 start loading (all) databases at offset 42
39 limit loading to 100 records
43 remove database and Hyper Estraier index before indexing
45 =item --only=database_name/input_filter
47 reindex just single database (legacy name is --one)
49 C</input_filter> is optional part which can be C<name>
52 =item --config conf/config.yml
54 path to YAML configuration file
58 force conversion C<< normalize->path >> in C<config.yml> from
63 disable indexing and dump statistics about field and subfield
74 my $config = 'conf/config.yml';
82 "offset=i" => \$offset,
84 "one=s" => \$only_filter,
85 "only=s" => \$only_filter,
88 "force-set" => \$force_set,
92 $config = LoadFile($config);
94 print "config = ",Dumper($config) if ($debug);
96 die "no databases in config file!\n" unless ($config->{databases});
98 my $log = _new WebPAC::Common()->_get_logger();
99 $log->info( "-" x 79 );
101 my $use_indexer = $config->{use_indexer} || 'hyperestraier';
103 $log->debug("option --stats disables update of indexing engine...");
104 $use_indexer = undef;
106 $log->info("using $use_indexer indexing engine...");
110 my $start_t = time();
112 while (my ($database, $db_config) = each %{ $config->{databases} }) {
114 my ($only_database,$only_input) = split(m#/#, $only_filter);
115 next if ($only_database && $database !~ m/$only_database/i);
120 my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
121 $indexer_config->{database} = $database;
122 $indexer_config->{clean} = $clean;
123 $indexer_config->{label} = $db_config->{name};
125 if ($use_indexer eq 'hyperestraier') {
127 # open Hyper Estraier database
128 use WebPAC::Output::Estraier '0.10';
129 $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
131 } elsif ($use_indexer eq 'kinosearch') {
134 use WebPAC::Output::KinoSearch;
135 $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
136 $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
139 $log->logdie("unknown use_indexer: $use_indexer");
142 $log->logide("can't continue without valid indexer") unless ($indexer);
149 my $abs_path = abs_path($0);
150 $abs_path =~ s#/[^/]*$#/#;
152 my $db_path = $config->{webpac}->{db_path} . '/' . $database;
155 $log->info("creating new database '$database' in $db_path");
156 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
158 $log->info("working on database '$database' in $db_path");
161 my $db = new WebPAC::Store(
163 database => $database,
169 # now, iterate through input formats
173 if (ref($db_config->{input}) eq 'ARRAY') {
174 @inputs = @{ $db_config->{input} };
175 } elsif ($db_config->{input}) {
176 push @inputs, $db_config->{input};
178 $log->info("database $database doesn't have inputs defined");
181 my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
183 foreach my $input (@inputs) {
185 next if ($only_input && $input->{name} =~ m#$only_input#i || $input->{type} =~ m#$only_input#i);
187 my $type = lc($input->{type});
189 die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
191 my $lookup = new WebPAC::Lookup(
192 lookup_file => $input->{lookup},
195 my $input_module = $config->{webpac}->{inputs}->{$type};
197 $log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module lookup '$input->{lookup}'");
199 my $input_db = new WebPAC::Input(
200 module => $input_module,
201 code_page => $config->{webpac}->{webpac_encoding},
202 limit => $limit || $input->{limit},
205 recode => $input->{recode},
208 $log->logdie("can't create input using $input_module") unless ($input);
210 my $maxmfn = $input_db->open(
211 path => $input->{path},
212 code_page => $input->{encoding}, # database encoding
215 my $n = new WebPAC::Normalize::XML(
216 # filter => { 'foo' => sub { shift } },
218 lookup_regex => $lookup->regex,
220 prefix => $input->{name},
224 my $normalize_path = $input->{normalize}->{path};
227 my $new_norm_path = $normalize_path;
228 $new_norm_path =~ s/\.xml$/.pl/;
229 if (-e $new_norm_path) {
230 $log->debug("--force-set replaced $normalize_path with $new_norm_path");
231 $normalize_path = $new_norm_path;
233 $log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
237 if ($normalize_path =~ m/\.xml$/i) {
239 tag => $input->{normalize}->{tag},
240 xml_file => $normalize_path,
242 } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
244 path => $normalize_path,
245 tag => $input->{normalize}->{tag},
247 } elsif ($normalize_path =~ m/\.(?:pl)$/i) {
249 $log->info("using WebPAC::Normalize::Set to process $normalize_path");
250 $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
253 foreach my $pos ( 0 ... $input_db->size ) {
255 my $row = $input_db->fetch || next;
257 my $mfn = $row->{'000'}->[0];
259 if (! $mfn || $mfn !~ m#^\d+$#) {
260 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
262 push @{ $row->{'000'} }, $pos;
268 $ds = $n->data_structure($row);
270 $ds = WebPAC::Normalize::Set::data_structure(
273 lookup => $lookup->lookup_hash,
279 prefix => $input->{name},
280 ) if ($ds && !$stats);
284 id => $input->{name} . "/" . $mfn,
286 type => $config->{$use_indexer}->{type},
292 $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
296 eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
298 my $dt = time() - $start_t;
299 $log->info("$total_rows records ", $indexer ? "indexed " : "",
300 sprintf("in %.2f sec [%.2f rec/sec]",
301 $dt, ($total_rows / $dt)
306 # add Hyper Estraier links to other databases
308 if (ref($db_config->{links}) eq 'ARRAY') {
309 foreach my $link (@{ $db_config->{links} }) {
310 if ($use_indexer eq 'hyperestraier') {
311 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
315 credit => $link->{credit},
318 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");