6 use File::Temp qw/tempdir/;
9 use WebPAC::Common 0.02;
10 use WebPAC::Parser 0.08;
11 use WebPAC::Input 0.16;
12 use WebPAC::Store 0.15;
13 use WebPAC::Normalize 0.22;
14 use WebPAC::Output::TT;
15 use WebPAC::Validate 0.11;
16 use WebPAC::Output::MARC;
20 use Time::HiRes qw/time/;
22 use Data::Dump qw/dump/;
23 use Storable qw/dclone/;
24 use Pod::Usage qw/pod2usage/;
26 use Proc::Queue size => 1;
27 use POSIX ":sys_wait_h"; # imports WNOHANG
31 run.pl - start WebPAC indexing
33 B<this command will probably go away. Don't get used to it!>
41 start loading (all) databases at offset 42
45 limit loading to 100 records
49 remove database and Hyper Estraier index before indexing
51 =item --only=database_name/input_filter
53 reindex just single database (legacy name is --one)
55 C</input_filter> is optional part which can be C<name>
58 =item --config conf/config.yml
60 path to YAML configuration file
64 disable indexing, modify_* in configuration and dump statistics about field
65 and subfield usage for each input
67 =item --validate path/to/validation_file
69 turn on extra validation of imput records, see L<WebPAC::Validation>
71 You can use special variables C<$database> and $C<$input> in this parametar
72 like C<--validate 'conf/validate/$database-$input'> to construct filename
74 =item --validate-delimiters path/to/validate_delimiters_file
76 this option is used with C<--validate> to turn on extra validation of
77 delimiters. If file is non existant, it will be created on first run.
81 Generate MARC file. This will automatically be on if file contains C<marc*> directives.
82 You can use this option as C<--no-marc-generate> to disable MARC generation.
86 By default turned on if normalisation file has C<marc*> directives. You can disable lint
87 messages with C<--no-marc-lint>.
91 Force dump or input and marc record for debugging.
95 Run databases in parallel (aproximatly same as number of processors in
96 machine if you want to use full load)
104 Create merged index of databases which have links
119 my $validate_delimiters_path;
120 my $marc_generate = 1;
128 my $log = _new WebPAC::Common()->_get_logger();
131 "limit=i" => \$limit,
132 "offset=i" => \$offset,
134 "one=s" => \$only_filter,
135 "only=s" => \$only_filter,
136 "config" => \$config_path,
139 "validate=s" => \$validate_path,
140 "validate-delimiters=s" => \$validate_delimiters_path,
141 "marc-generate!" => \$marc_generate,
142 "marc-lint!" => \$marc_lint,
143 "marc-dump!" => \$marc_dump,
144 "parallel=i" => \$parallel,
145 "only-links!" => \$only_links,
150 $marc_generate = 0 if ( $validate_delimiters_path );
152 pod2usage(-verbose => 2) if ($help);
154 my $config = new WebPAC::Config( path => $config_path );
156 #print "config = ",dump($config) if ($debug);
158 die "no databases in config file!\n" unless ($config->databases);
160 $log->info( "-" x 79 );
162 my $log_file = 'log';
164 if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) {
165 $log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
166 rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
170 my $estcmd_path = './estcmd-merge.sh';
172 open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!");
173 print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/;
174 print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
175 $log->info("created merge batch file $estcmd_path");
179 $validate = new WebPAC::Validate(
180 delimiters => $config->webpac('delimiters'),
181 ) if ($validate_path || $validate_delimiters_path);
183 my $use_indexer = $config->use_indexer;
184 $stats ||= $validate;
186 $log->debug("disabled indexing for stats collection");
187 $use_indexer = undef;
188 } elsif ( $use_indexer ) {
189 $log->info("using $use_indexer indexing engine...");
192 # parse normalize files and create source files for lookup and normalization
194 my $parser = new WebPAC::Parser( config => $config );
197 my $start_t = time();
202 $log->info("Using $parallel processes for speedup");
203 Proc::Queue::size($parallel);
206 sub create_ds_config {
207 my ($db_config, $database, $input, $mfn) = @_;
208 my $c = dclone( $db_config );
209 $c->{_} = $database || $log->logconfess("need database");
210 $c->{_mfn} = $mfn || $log->logconfess("need mfn");
211 $c->{input} = $input || $log->logconfess("need input");
215 foreach my $database ( sort keys %{ $config->databases } ) {
216 my $db_config = $config->databases->{$database};
218 my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
219 next if ($only_database && $database !~ m/$only_database/i);
223 if(defined ($f) and $f==0) {
224 $log->info("Created processes $$ for speedup");
231 if ($use_indexer && $parser->have_rules( 'search', $database )) {
233 my $cfg_name = $use_indexer;
234 $cfg_name =~ s/\-.*$//;
236 my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
237 $indexer_config->{database} = $database;
238 $indexer_config->{clean} = $clean;
239 $indexer_config->{label} = $db_config->{name};
241 # force clean if database has links
242 $indexer_config->{clean} = 1 if ($db_config->{links});
244 if ($use_indexer eq 'hyperestraier') {
246 # open Hyper Estraier database
247 require WebPAC::Output::Estraier;
248 $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
250 } elsif ($use_indexer eq 'hyperestraier-native') {
252 # open Hyper Estraier database
253 require WebPAC::Output::EstraierNative;
254 $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
256 } elsif ($use_indexer eq 'kinosearch') {
258 die "no longer supported";
261 $log->logdie("unknown use_indexer: $use_indexer");
264 $log->logdie("can't continue without valid indexer") unless ($indexer);
269 # store Hyper Estraier links to other databases
271 if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
272 foreach my $link (@{ $db_config->{links} }) {
273 if ($use_indexer eq 'hyperestraier') {
275 print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
277 $log->info("saving link $database -> $link->{to} [$link->{credit}]");
279 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
283 credit => $link->{credit},
288 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
292 next if ($only_links);
298 my $store = new WebPAC::Store({
306 my @outputs = force_array( $db_config->{output}, sub {
307 $log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" );
312 foreach my $output ( @outputs ) {
314 warn '## output = ',dump( $output );
316 my $module = $output->{module} || $log->logdie("need module in output section of $database");
317 $module = 'WebPAC::Output::' . $module unless $module =~ m/::/;
319 $log->debug("loading output module $module");
320 eval "require $module";
322 $output->{database} = $database;
324 $log->debug("calling $module->new(",dump( $output ),")");
325 my $out = new $module->new( $output );
328 push @output_modules, $out;
333 # now, iterate through input formats
337 my @inputs = force_array( $db_config->{input}, sub {
338 $log->info("database $database doesn't have inputs defined");
341 foreach my $input (@inputs) {
343 my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));
345 next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
347 my $type = lc($input->{type});
349 die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
351 my $input_module = $config->webpac('inputs')->{$type};
353 my @lookups = $parser->have_lookup_create($database, $input);
355 $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
356 @lookups ? " creating lookups: ".join(", ", @lookups) : ""
360 # disable modification of records if --stats is in use
361 delete($input->{modify_records});
362 delete($input->{modify_file});
365 my $input_db = new WebPAC::Input(
366 module => $input_module,
367 encoding => $config->webpac('webpac_encoding'),
368 limit => $limit || $input->{limit},
370 recode => $input->{recode},
372 modify_records => $input->{modify_records},
373 modify_file => $input->{modify_file},
374 input_config => $input,
376 $log->logdie("can't create input using $input_module") unless ($input);
378 if (defined( $input->{lookup} )) {
379 $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
380 delete( $input->{lookup} );
387 my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
389 $lookup_coderef = sub {
390 my $rec = shift || die "need rec!";
391 my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
393 WebPAC::Normalize::data_structure(
396 config => create_ds_config( $db_config, $database, $input, $mfn ),
399 #warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
402 WebPAC::Normalize::_set_lookup( undef );
404 $log->debug("created lookup_coderef using:\n$rules");
410 my $maxmfn = $input_db->open(
411 path => $input->{path},
412 code_page => $input->{encoding}, # database encoding
413 lookup_coderef => $lookup_coderef,
414 lookup => $lookup_jar,
418 return $store->load_row(
419 database => $database,
420 input => $input_name,
426 return $store->save_row(
427 database => $database,
428 input => $input_name,
436 my $lookup_data = WebPAC::Normalize::_get_lookup();
438 if (defined( $lookup_data->{$database}->{$input_name} )) {
439 $log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
441 foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
443 database => $database,
444 input => $input_name,
446 data => $lookup_data->{$database}->{$input_name}->{$key},
452 if ($stats || $validate) {
453 my $path = "out/report/${database}-${input_name}.txt";
454 open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
456 print $report_fh "Report for database '$database' input '$input_name' records ",
457 $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
458 $log->info("Generating report file $path");
461 $validate->read_validate_file( $validate->fill_in( $validate_path, database => $database, input => $input_name ) ) if ( $validate_path );
462 $validate->read_validate_delimiters_file( $validate->fill_in( $validate_delimiters_path, database => $database, input => $input_name ) ) if ( $validate_delimiters_path );
467 if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) {
468 $marc = new WebPAC::Output::MARC(
469 path => "out/marc/${database}-${input_name}.marc",
475 my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
476 $log->debug("parsed normalize rules:\n$rules");
478 # reset position in database
481 # generate name of config key for indexer (strip everything after -)
482 my $indexer_config = $use_indexer;
483 $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
486 my $depends = $parser->depends($database,$input_name);
489 $log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
490 $log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
492 foreach my $db (keys %$depends) {
493 foreach my $i (keys %{$depends->{$db}}) {
494 foreach my $k (keys %{$depends->{$db}->{$i}}) {
496 $log->debug("loading lookup $db/$i");
497 $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
502 $log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
507 $log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
511 foreach my $pos ( 0 ... $input_db->size ) {
513 my $row = $input_db->fetch || next;
517 my $mfn = $row->{'000'}->[0];
519 if (! $mfn || $mfn !~ m{^\d+$}) {
520 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
522 push @{ $row->{'000'} }, $pos;
527 if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
528 $log->error( "MFN $mfn validation error:\n",
529 $validate->report_error( $errors )
532 next; # validation doesn't create any output
535 my $ds = WebPAC::Normalize::data_structure(
538 lookup => $lookup_hash,
539 config => create_ds_config( $db_config, $database, $input, $mfn ),
540 marc_encoding => 'utf-8',
541 load_row_coderef => sub {
542 my ($database,$input,$mfn) = @_;
543 return $store->load_row(
544 database => $database,
551 $log->debug("ds = ", sub { dump($ds) }) if ($ds);
554 database => $database,
555 input => $input_name,
558 ) if ($ds && !$stats);
561 id => "${input_name}/${mfn}",
563 type => $config->get($indexer_config)->{type},
564 ) if ($indexer && $ds);
569 while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
571 id => $mfn . ( $i ? "/$i" : '' ),
573 leader => WebPAC::Normalize::_get_marc_leader(),
579 $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
582 foreach my $out ( @output_modules ) {
583 $out->add( $mfn, $ds ) if $out->can('add');
589 my $errors = $validate->report;
591 $log->info("validation errors:\n$errors\n" );
592 print $report_fh "$errors\n" if ($report_fh);
595 print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n";
597 # must be last thing that touches $validate for this input
602 my $s = $input_db->stats;
603 $log->info("statistics of fields usage:\n$s");
604 print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
608 $marc->finish if ($marc);
611 close($report_fh) if ($report_fh);
614 eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
616 foreach my $out ( @output_modules ) {
617 $out->finish if $out->can('finish');
620 my $dt = time() - $start_t;
621 $log->info("$total_rows records ", $indexer ? "indexed " : "",
622 sprintf("in %.2f sec [%.2f rec/sec]",
623 $dt, ($total_rows / $dt)
630 $log->info("parallel process $$ finished");
637 # wait all children to finish
638 sleep(1) while wait != -1;
639 $log->info("all parallel processes finished");
642 # save new delimiters if needed
643 $validate->save_delimiters_templates if ( $validate_delimiters_path );
646 # handle links or merge after indexing
650 print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
652 chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
655 foreach my $link (@links) {
656 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');