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=s" => \$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 WebPAC::Normalize::_debug( $debug - 1 ) if $debug > 1;
158 #print "config = ",dump($config) if ($debug);
160 die "no databases in config file!\n" unless ($config->databases);
162 $log->info( "-" x 79 );
164 my $log_file = 'log';
166 if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) {
167 $log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
168 rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
172 my $estcmd_path = './estcmd-merge.sh';
174 open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!");
175 print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/;
176 print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
177 $log->info("created merge batch file $estcmd_path");
181 $validate = new WebPAC::Validate(
182 delimiters => $config->webpac('delimiters'),
183 ) if ($validate_path || $validate_delimiters_path);
185 my $use_indexer = $config->use_indexer;
186 $stats ||= $validate;
188 $log->debug("disabled indexing for stats collection");
189 $use_indexer = undef;
190 } elsif ( $use_indexer ) {
191 $log->info("using $use_indexer indexing engine...");
194 # parse normalize files and create source files for lookup and normalization
196 my ($only_database,$only_input) = split(m#/#, $only_filter) if $only_filter;
198 my $parser = new WebPAC::Parser(
200 only_database => $only_database,
201 only_input => $only_input,
205 my $start_t = time();
210 $log->info("Using $parallel processes for speedup");
211 Proc::Queue::size($parallel);
214 sub create_ds_config {
215 my ($db_config, $database, $input, $mfn) = @_;
216 my $c = dclone( $db_config );
217 $c->{_} = $database || $log->logconfess("need database");
218 $c->{_mfn} = $mfn || $log->logconfess("need mfn");
219 $c->{input} = $input || $log->logconfess("need input");
223 foreach my $database ( sort keys %{ $config->databases } ) {
224 my $db_config = $config->databases->{$database};
226 next if ($only_database && $database !~ m/$only_database/i);
230 if(defined ($f) and $f==0) {
231 $log->info("Created processes $$ for speedup");
238 if ($use_indexer && $parser->have_rules( 'search', $database )) {
240 my $cfg_name = $use_indexer;
241 $cfg_name =~ s/\-.*$//;
243 my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
244 $indexer_config->{database} = $database;
245 $indexer_config->{clean} = $clean;
246 $indexer_config->{label} = $db_config->{name};
248 # force clean if database has links
249 $indexer_config->{clean} = 1 if ($db_config->{links});
251 if ($use_indexer eq 'hyperestraier') {
253 # open Hyper Estraier database
254 require WebPAC::Output::Estraier;
255 $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
257 } elsif ($use_indexer eq 'hyperestraier-native') {
259 # open Hyper Estraier database
260 require WebPAC::Output::EstraierNative;
261 $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
263 } elsif ($use_indexer eq 'kinosearch') {
265 die "no longer supported";
268 $log->logdie("unknown use_indexer: $use_indexer");
271 $log->logdie("can't continue without valid indexer") unless ($indexer);
276 # store Hyper Estraier links to other databases
278 if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
279 foreach my $link (@{ $db_config->{links} }) {
280 if ($use_indexer eq 'hyperestraier') {
282 print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
284 $log->info("saving link $database -> $link->{to} [$link->{credit}]");
286 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
290 credit => $link->{credit},
295 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
299 next if ($only_links);
305 my $store = new WebPAC::Store({
313 my @outputs = force_array( $db_config->{output}, sub {
314 $log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" );
319 foreach my $output ( @outputs ) {
321 #warn '## output = ',dump( $output );
323 my $module = $output->{module} || $log->logdie("need module in output section of $database");
324 $module = 'WebPAC::Output::' . $module unless $module =~ m/::/;
326 $log->debug("loading output module $module");
327 eval "require $module";
329 # add database to arugemnts for output filter
330 $output->{database} = $database;
332 $log->debug("calling $module->new(",dump( $output ),")");
333 my $out = new $module->new( $output );
336 push @output_modules, $out;
341 # now, iterate through input formats
345 my @inputs = force_array( $db_config->{input}, sub {
346 $log->info("database $database doesn't have inputs defined");
349 foreach my $input (@inputs) {
351 my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));
353 next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
355 my $type = lc($input->{type});
357 # FIXME check if input module exists
358 my $input_module = $input->{module};
360 if ( ! $input_module ) {
361 if ( grep(/$type/, $config->webpac('inputs')) ) {
362 $input_module = $config->webpac('inputs')->{$type};
364 $log->logdie("I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!" );
368 my @lookups = $parser->have_lookup_create($database, $input);
370 $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
371 @lookups ? " creating lookups: ".join(", ", @lookups) : ""
375 # disable modification of records if --stats is in use
376 delete($input->{modify_records});
377 delete($input->{modify_file});
380 my $input_db = new WebPAC::Input(
381 module => $input_module,
382 encoding => $config->webpac('webpac_encoding'),
383 limit => $limit || $input->{limit},
385 recode => $input->{recode},
387 modify_records => $input->{modify_records},
388 modify_file => $input->{modify_file},
389 input_config => $input,
391 $log->logdie("can't create input using $input_module") unless ($input);
393 if (defined( $input->{lookup} )) {
394 $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
395 delete( $input->{lookup} );
402 my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
404 $lookup_coderef = sub {
405 my $rec = shift || die "need rec!";
406 my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
408 WebPAC::Normalize::data_structure(
411 config => create_ds_config( $db_config, $database, $input, $mfn ),
414 #warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
417 WebPAC::Normalize::_set_lookup( undef );
419 $log->debug("created lookup_coderef using:\n$rules");
425 my $maxmfn = $input_db->open(
426 path => $input->{path},
427 code_page => $input->{encoding}, # database encoding
428 lookup_coderef => $lookup_coderef,
429 lookup => $lookup_jar,
433 return $store->load_row(
434 database => $database,
435 input => $input_name,
441 return $store->save_row(
442 database => $database,
443 input => $input_name,
451 my $lookup_data = WebPAC::Normalize::_get_lookup();
453 if (defined( $lookup_data->{$database}->{$input_name} )) {
454 $log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
456 foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
458 database => $database,
459 input => $input_name,
461 data => $lookup_data->{$database}->{$input_name}->{$key},
467 if ($stats || $validate) {
468 my $path = "out/report/${database}-${input_name}.txt";
469 open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
471 print $report_fh "Report for database '$database' input '$input_name' records ",
472 $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
473 $log->info("Generating report file $path");
476 $validate->read_validate_file( $validate->fill_in( $validate_path, database => $database, input => $input_name ) ) if ( $validate_path );
477 $validate->read_validate_delimiters_file( $validate->fill_in( $validate_delimiters_path, database => $database, input => $input_name ) ) if ( $validate_delimiters_path );
482 if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) {
483 $marc = new WebPAC::Output::MARC(
484 path => "out/marc/${database}-${input_name}.marc",
490 my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
491 $log->debug("parsed normalize rules:\n$rules");
493 # reset position in database
496 # generate name of config key for indexer (strip everything after -)
497 my $indexer_config = $use_indexer;
498 $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
501 my $depends = $parser->depends($database,$input_name);
504 $log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
505 $log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
507 foreach my $db (keys %$depends) {
508 foreach my $i (keys %{$depends->{$db}}) {
509 foreach my $k (keys %{$depends->{$db}->{$i}}) {
511 $log->debug("loading lookup $db/$i");
512 $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
517 $log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
522 $log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
526 # setup input name for all output filters
527 foreach my $out ( @output_modules ) {
528 if ( $out->can('input') ) {
529 $out->input( $input_name );
531 $log->warn("output filter ",ref($out)," doesn't support input name");
536 foreach my $pos ( 0 ... $input_db->size ) {
538 my $row = $input_db->fetch || next;
542 my $mfn = $row->{'000'}->[0];
544 if (! $mfn || $mfn !~ m{^\d+$}) {
545 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
547 push @{ $row->{'000'} }, $pos;
552 if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
553 $log->error( "MFN $mfn validation error:\n",
554 $validate->report_error( $errors )
557 next; # validation doesn't create any output
560 my $ds = WebPAC::Normalize::data_structure(
563 lookup => $lookup_hash,
564 config => create_ds_config( $db_config, $database, $input, $mfn ),
565 marc_encoding => 'utf-8',
566 load_row_coderef => sub {
567 my ($database,$input,$mfn) = @_;
568 #warn "### load_row($database,$input,$mfn) from data_structure\n";
569 return $store->load_row(
570 database => $database,
577 $log->debug("ds = ", sub { dump($ds) });
582 database => $database,
583 input => $input_name,
589 id => "${input_name}/${mfn}",
591 type => $config->get($indexer_config)->{type},
594 foreach my $out ( @output_modules ) {
595 $out->add( $mfn, $ds ) if $out->can('add');
599 $log->warn("record $pos didn't produce any output after normalization rules!") unless $marc;
605 while (my $fields = WebPAC::Normalize::MARC::_get_marc_fields( fetch_next => 1 ) ) {
607 id => $mfn . ( $i ? "/$i" : '' ),
609 leader => WebPAC::Normalize::MARC::_get_marc_leader(),
615 $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
621 my $errors = $validate->report;
623 $log->info("validation errors:\n$errors\n" );
624 print $report_fh "$errors\n" if ($report_fh);
627 print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n";
629 # must be last thing that touches $validate for this input
634 my $s = $input_db->stats;
635 $log->info("statistics of fields usage:\n$s");
636 print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
640 $marc->finish if ($marc);
643 close($report_fh) if ($report_fh);
646 eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
648 foreach my $out ( @output_modules ) {
649 $out->finish if $out->can('finish');
652 my $dt = time() - $start_t;
653 $log->info("$total_rows records ", $indexer ? "indexed " : "",
654 sprintf("in %.2f sec [%.2f rec/sec]",
655 $dt, ($total_rows / $dt)
662 $log->info("parallel process $$ finished");
669 # wait all children to finish
670 sleep(1) while wait != -1;
671 $log->info("all parallel processes finished");
674 # save new delimiters if needed
675 $validate->save_delimiters_templates if ( $validate_delimiters_path );
678 # handle links or merge after indexing
682 print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
684 chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
687 foreach my $link (@links) {
688 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');