implement skip on inputs
[webpac2] / run.pl
1 #!/usr/bin/perl -w
2
3 use strict;
4
5 use Cwd qw/abs_path/;
6 use File::Temp qw/tempdir/;
7 use lib './lib';
8
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;
17 use WebPAC::Config;
18 use Getopt::Long;
19 use File::Path;
20 use Time::HiRes qw/time/;
21 use File::Slurp;
22 use Data::Dump qw/dump/;
23 use Storable qw/dclone/;
24 use Pod::Usage qw/pod2usage/;
25
26 use POSIX ":sys_wait_h"; # imports WNOHANG
27
28 =head1 NAME
29
30 run.pl - start WebPAC indexing
31
32 B<this command will probably go away. Don't get used to it!>
33
34 =head1 OPTIONS
35
36 =over 4
37
38 =item --offset 42
39
40 start loading (all) databases at offset 42
41
42 =item --limit 100
43
44 limit loading to 100 records
45
46 =item --clean
47
48 remove database and Hyper Estraier index before indexing
49
50 =item --only=database_name/input_filter
51
52 reindex just single database (legacy name is --one)
53
54 C</input_filter> is optional part which can be C<name>
55 or C<type> from input
56
57 =item --config conf/config.yml
58
59 path to YAML configuration file
60
61 =item --stats
62
63 disable indexing, modify_* in configuration and dump statistics about field
64 and subfield usage for each input
65
66 =item --validate path/to/validation_file
67
68 turn on extra validation of imput records, see L<WebPAC::Validation>
69
70 You can use special variables C<$database> and $C<$input> in this parametar
71 like C<--validate 'conf/validate/$database-$input'> to construct filename
72
73 =item --validate-delimiters path/to/validate_delimiters_file
74
75 this option is used with C<--validate> to turn on extra validation of
76 delimiters. If file is non existant, it will be created on first run.
77
78 =item --marc-generate
79
80 Generate MARC file. This will automatically be on if file contains C<marc*> directives.
81 You can use this option as C<--no-marc-generate> to disable MARC generation.
82
83 =item --marc-lint
84
85 By default turned on if normalisation file has C<marc*> directives. You can disable lint
86 messages with C<--no-marc-lint>.
87
88 =item --marc-dump
89
90 Force dump or input and marc record for debugging.
91
92 =item --parallel 4
93
94 Run databases in parallel (aproximatly same as number of processors in
95 machine if you want to use full load)
96
97 =item --only-links
98
99 Create just links
100
101 =item --merge
102
103 Create merged index of databases which have links
104
105 =back
106
107 =cut
108
109 my $offset;
110 my $limit;
111
112 my $clean = 0;
113 my $config_path;
114 my $debug = 0;
115 my $only_filter;
116 my $stats = 0;
117 my $validate_path;
118 my $validate_delimiters_path;
119 my $marc_generate = 1;
120 my $marc_lint = 1;
121 my $marc_dump = 0;
122 my $parallel = 0;
123 my $only_links = 0;
124 my $merge = 0;
125 my $help;
126
127 my $log = _new WebPAC::Common()->_get_logger();
128
129 GetOptions(
130         "limit=i" => \$limit,
131         "offset=i" => \$offset,
132         "clean" => \$clean,
133         "one=s" => \$only_filter,
134         "only=s" => \$only_filter,
135         "config=s" => \$config_path,
136         "debug+" => \$debug,
137         "stats" => \$stats,
138         "validate=s" => \$validate_path,
139         "validate-delimiters=s" => \$validate_delimiters_path,
140         "marc-generate!" => \$marc_generate,
141         "marc-lint!" => \$marc_lint,
142         "marc-dump!" => \$marc_dump,
143         "parallel=i" => \$parallel,
144         "only-links!" => \$only_links,
145         "merge" => \$merge,
146         "help" => \$help,
147 );
148
149 $marc_generate = 0 if ( $validate_delimiters_path );
150
151 pod2usage(-verbose => 2) if ($help);
152
153 my $config = new WebPAC::Config( path => $config_path );
154
155 WebPAC::Normalize::_debug( $debug - 1 ) if $debug > 1;
156
157 #print "config = ",dump($config) if ($debug);
158
159 die "no databases in config file!\n" unless ($config->databases);
160
161 $log->info( "-" x 79 );
162
163 my $log_file = 'log';
164
165 if (-e $log_file ) {    # && -s $log_file > 5 * 1024 * 1024) {
166         $log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
167         rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
168 }
169
170 my $estcmd_fh;
171 my $estcmd_path = './estcmd-merge.sh';
172 if ($merge) {
173         open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!");
174         print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/;
175         print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
176         $log->info("created merge batch file $estcmd_path");
177 }
178
179 my $validate;
180 $validate = new WebPAC::Validate(
181         delimiters => $config->webpac('delimiters'),
182 ) if ($validate_path || $validate_delimiters_path);
183
184 my $use_indexer = $config->use_indexer;
185 $stats ||= $validate;
186 if ($stats) {
187         $log->debug("disabled indexing for stats collection");
188         $use_indexer = undef;
189 } elsif ( $use_indexer ) {
190         $log->info("using $use_indexer indexing engine...");
191 }
192
193 # parse normalize files and create source files for lookup and normalization
194
195 my ($only_database,$only_input) = split(m#/#, $only_filter) if $only_filter;
196
197 my $parser = new WebPAC::Parser(
198         config => $config,
199         only_database => $only_database,
200         only_input => $only_input,
201 );
202
203 my $total_rows = 0;
204 my $start_t = time();
205
206 my @links;
207
208 if ($parallel) {
209         eval 'use Proc::Queue size => 1;';
210         die $@ if $@;
211         $log->info("Using $parallel processes for speedup");
212         Proc::Queue::size($parallel);
213 }
214
215 sub create_ds_config {
216         my ($db_config, $database, $input, $mfn) = @_;
217         my $c = dclone( $db_config );
218         $c->{_} = $database || $log->logconfess("need database");
219         $c->{_mfn} = $mfn || $log->logconfess("need mfn");
220         $c->{input} = $input || $log->logconfess("need input");
221         return $c;
222 }
223
224 foreach my $database ( sort keys %{ $config->databases } ) {
225         my $db_config = $config->databases->{$database};
226
227         next if ($only_database && $database !~ m/$only_database/i);
228
229         if ($parallel) {
230                 my $f=fork;
231                 if(defined ($f) and $f==0) {
232                         $log->info("Created processes $$ for speedup");
233                 } else {
234                         next;
235                 }
236         }
237
238         my $indexer;
239         if ($use_indexer && $parser->have_rules( 'search', $database )) {
240
241                 my $cfg_name = $use_indexer;
242                 $cfg_name =~ s/\-.*$//;
243
244                 my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
245                 $indexer_config->{database} = $database;
246                 $indexer_config->{clean} = $clean;
247                 $indexer_config->{label} = $db_config->{name};
248
249                 # force clean if database has links
250                 $indexer_config->{clean} = 1 if ($db_config->{links});
251
252                 if ($use_indexer eq 'hyperestraier') {
253
254                         # open Hyper Estraier database
255                         require WebPAC::Output::Estraier;
256                         $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
257                 
258                 } elsif ($use_indexer eq 'hyperestraier-native') {
259
260                         # open Hyper Estraier database
261                         require WebPAC::Output::EstraierNative;
262                         $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
263
264                 } elsif ($use_indexer eq 'kinosearch') {
265
266                         die "no longer supported";
267
268                 } else {
269                         $log->logdie("unknown use_indexer: $use_indexer");
270                 }
271
272                 $log->logdie("can't continue without valid indexer") unless ($indexer);
273         }
274
275
276         #
277         # store Hyper Estraier links to other databases
278         #
279         if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
280                 foreach my $link (@{ $db_config->{links} }) {
281                         if ($use_indexer eq 'hyperestraier') {
282                                 if ($merge) {
283                                         print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
284                                 } else {
285                                         $log->info("saving link $database -> $link->{to} [$link->{credit}]");
286                                         push @links, sub {
287                                                 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
288                                                 $indexer->add_link(
289                                                         from => $database,
290                                                         to => $link->{to},
291                                                         credit => $link->{credit},
292                                                 );
293                                         };
294                                 }
295                         } else {
296                                 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
297                         }
298                 }
299         }
300         next if ($only_links);
301
302
303         #
304         # now WebPAC::Store
305         #
306         my $store = new WebPAC::Store({
307                 debug => $debug,
308         });
309
310
311         #
312         # prepare output
313         #
314         my @outputs = force_array( $db_config->{output}, sub {
315                 $log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" );
316         } );
317
318         my @output_modules;
319
320         foreach my $output ( @outputs ) {
321
322 #warn '## output = ',dump( $output );
323
324                 my $module = $output->{module} || $log->logdie("need module in output section of $database");
325                 $module = 'WebPAC::Output::' . $module unless $module =~ m/::/;
326         
327                 $log->debug("loading output module $module");
328                 eval "require $module";
329
330                 # add database to arugemnts for output filter
331                 $output->{database} = $database;
332                 $output->{clean} = $clean;
333
334                 $log->debug("calling $module->new(",dump( $output ),")");
335                 my $out = new $module->new( $output );
336                 if ( $out->init ) {
337                         push @output_modules, $out;
338                 } else {
339                         $log->warn("SKIPPED $module");
340                 }
341         }
342
343
344         #
345         # now, iterate through input formats
346         #
347
348
349         my @inputs = force_array( $db_config->{input}, sub {
350                 $log->info("database $database doesn't have inputs defined");
351         } );
352
353         if ( -e 'out/debug' ) { # FIXME flag?
354                 my $out;
355                 foreach my $i ( @inputs ) {
356                         warn dump( $i );
357                         next unless defined $i->{normalize};
358                         warn dump( $i->{normalize} );
359                         foreach my $normalize ( @{ $i->{normalize} } ) {
360                                 my $path = $normalize->{path};
361                                 $out .= qq/\n##\n## $path\n##\n\n/;
362                                 $out .= read_file( $path );
363                         }
364                 }
365                 my $all = "out/debug/all-normalize.pl";
366                 write_file( $all, $out );
367                 warn "### all normalize for this input saved to: $all";
368         };
369
370         foreach my $input (@inputs) {
371
372                 my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));
373
374                 if ( $input->{skip} ) {
375                         $log->info("skip $input_name");
376                         next;
377                 }
378
379                 next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
380
381                 my $type = lc($input->{type});
382
383                 # FIXME check if input module exists
384                 my $input_module = $input->{module};
385
386                 if ( ! $input_module ) {
387                         if ( grep(/$type/, $config->webpac('inputs')) ) {
388                                 $input_module = $config->webpac('inputs')->{$type};
389                         } else {
390                                 $log->logdie("I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!" );
391                         }
392                 }
393
394                 my @lookups = $parser->have_lookup_create($database, $input);
395
396                 $log->info("working on $database/$input_name with $input_module on $input->{path}",
397                         @lookups ? " creating lookups: ".join(", ", @lookups) : ""
398                 );
399
400                 if ($stats) {
401                         # disable modification of records if --stats is in use
402                         delete($input->{modify_records});
403                         delete($input->{modify_file});
404                 }
405
406                 my $input_db = new WebPAC::Input(
407                         module => $input_module,
408                         limit => $limit || $input->{limit},
409                         offset => $offset,
410                         recode => $input->{recode},
411                         stats => $stats,
412                         modify_records => $input->{modify_records},
413                         modify_file => $input->{modify_file},
414                         input_config => $input,
415                 );
416                 $log->logdie("can't create input using $input_module") unless ($input);
417
418                 if (defined( $input->{lookup} )) {
419                         $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
420                         delete( $input->{lookup} );
421                 }
422
423                 my $lookup_coderef;
424
425                 if (@lookups) {
426
427                         my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
428
429                         $lookup_coderef = sub {
430                                 my $rec = shift || die "need rec!";
431                                 my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
432
433                                 WebPAC::Normalize::data_structure(
434                                         row => $rec,
435                                         rules => $rules,
436                                         config => create_ds_config( $db_config, $database, $input, $mfn ),
437                                 );
438
439                                 #warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
440                         };
441
442                         WebPAC::Normalize::_set_lookup( undef );
443
444                         $log->debug("created lookup_coderef using:\n$rules");
445
446                 };
447
448                 my $lookup_jar;
449
450                 my $maxmfn = $input_db->open(
451                         path => $input->{path},
452                         input_encoding => $input->{encoding},   # database encoding
453                         lookup_coderef => $lookup_coderef,
454                         lookup => $lookup_jar,
455                         %{ $input },
456                         load_row => sub {
457                                 my $a = shift;
458                                 return $store->load_row(
459                                         database => $database,
460                                         input => $input_name,
461                                         id => $a->{id},
462                                 );
463                         },
464                         save_row => sub {
465                                 my $a = shift;
466                                 return $store->save_row(
467                                         database => $database,
468                                         input => $input_name,
469                                         id => $a->{id},
470                                         row => $a->{row},
471                                 );
472                         },
473
474                 );
475
476                 my $lookup_data = WebPAC::Normalize::_get_lookup();
477
478                 if (defined( $lookup_data->{$database}->{$input_name} )) {
479                         $log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
480
481                         foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
482                                 $store->save_lookup(
483                                         database => $database,
484                                         input => $input_name,
485                                         key => $key,
486                                         data => $lookup_data->{$database}->{$input_name}->{$key},
487                                 );
488                         }
489                 }
490
491                 my $report_fh;
492                 if ($stats || $validate) {
493                         my $path = "out/report/${database}-${input_name}.txt";
494                         open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
495
496                         print $report_fh "Report for database '$database' input '$input_name' records ",
497                                 $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
498                         $log->info("Generating report file $path");
499
500                         if ( $validate ) {
501                                 $validate->read_validate_file( $validate->fill_in( $validate_path, database => $database, input => $input_name ) ) if ( $validate_path );
502                                 $validate->read_validate_delimiters_file( $validate->fill_in( $validate_delimiters_path, database => $database, input => $input_name ) ) if ( $validate_delimiters_path );
503                         }
504                 }
505
506                 my $marc;
507                 if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) {
508                         $marc = new WebPAC::Output::MARC(
509                                 path => "out/marc/${database}-${input_name}.marc",
510                                 lint => $marc_lint,
511                                 dump => $marc_dump,
512                         );
513                 }
514
515                 my $rules = $parser->normalize_rules($database,$input_name);
516                 $log->logwarn("no normalize rules for $database/$input_name") unless $rules;
517
518                 $log->debug("parsed normalize rules:\n$rules");
519
520                 # reset position in database
521                 $input_db->seek(1);
522
523                 # generate name of config key for indexer (strip everything after -)
524                 my $indexer_config = $use_indexer;
525                 $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
526
527                 my $lookup_hash;
528                 my $depends = $parser->depends($database,$input_name);
529         
530                 if ($depends) {
531                         $log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
532                         $log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
533
534                         foreach my $db (keys %$depends) {
535                                 foreach my $i (keys %{$depends->{$db}}) {
536                                         foreach my $k (keys %{$depends->{$db}->{$i}}) {
537                                                 my $t = time();
538                                                 $log->debug("loading lookup $db/$i");
539                                                 $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
540                                                         database => $db,
541                                                         input => $i,
542                                                         key => $k,
543                                                 );
544                                                 $log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
545                                         }
546                                 }
547                         }
548
549                         $log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
550                 }
551
552
553                 # setup input name for all output filters
554                 foreach my $out ( @output_modules ) {
555                         if ( $out->can('input') ) {
556                                 $out->input( $input_name );
557                         } else {
558                                 $log->warn("output filter ",ref($out)," doesn't support input name");
559                         }
560                 }
561
562
563                 foreach my $pos ( 0 ... $input_db->size ) {
564
565                         my $row = $input_db->fetch || next;
566
567                         $total_rows++;
568
569                         my $mfn = $row->{'000'}->[0];
570
571                         if (! $mfn || $mfn !~ m{^\d+$}) {
572                                 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
573                                 $mfn = $pos;
574                                 push @{ $row->{'000'} }, $pos;
575                         }
576
577
578                         if ($validate) {
579                                 if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
580                                         $log->error( "MFN $mfn validation error:\n",
581                                                 $validate->report_error( $errors )
582                                         );
583                                 }
584                                 next;   # validation doesn't create any output
585                         }
586
587                         if ($rules) {
588
589                                 my $ds = WebPAC::Normalize::data_structure(
590                                         row => $row,
591                                         rules => $rules,
592                                         lookup => $lookup_hash,
593                                         config => create_ds_config( $db_config, $database, $input, $mfn ),
594                                         marc_encoding => 'utf-8',
595                                         load_row_coderef => sub {
596                                                 my ($database,$input,$mfn) = @_;
597 #warn "### load_row($database,$input,$mfn) from data_structure\n";
598                                                 return $store->load_row(
599                                                         database => $database,
600                                                         input => $input,
601                                                         id => $mfn,
602                                                 );
603                                         },
604                                 );
605
606                                 $log->debug("ds = ", sub { dump($ds) });
607
608                                 if ( $ds ) {
609
610                                         $store->save_ds(
611                                                 database => $database,
612                                                 input => $input_name,
613                                                 id => $mfn,
614                                                 ds => $ds,
615                                         ) if !$stats;
616
617                                         $indexer->add(
618                                                 id => "${input_name}/${mfn}",
619                                                 ds => $ds,
620                                                 type => $config->get($indexer_config)->{type},
621                                         ) if $indexer;
622
623                                         foreach my $out ( @output_modules ) {
624                                                 $out->add( $mfn, $ds ) if $out->can('add');
625                                         }
626
627                                 } else {
628                                         $log->warn("record $pos didn't produce any output after normalization rules!") unless $marc;
629                                 }
630                         }
631
632                         if ($marc) {
633                                 my $i = 0;
634
635                                 while (my $fields = WebPAC::Normalize::MARC::_get_marc_fields( fetch_next => 1 ) ) {
636                                         $marc->add(
637                                                 id => $mfn . ( $i ? "/$i" : '' ),
638                                                 fields => $fields,
639                                                 leader => WebPAC::Normalize::MARC::_get_marc_leader(),
640                                                 row => $row,
641                                         );
642                                         $i++;
643                                 }
644
645                                 $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
646                         }
647
648                 }
649
650                 if ($validate) {
651                         my $errors = $validate->report;
652                         if ($errors) {
653                                 $log->info("validation errors:\n$errors\n" );
654                                 print $report_fh "$errors\n" if ($report_fh);
655                         }
656
657                         print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n";
658
659                         # must be last thing that touches $validate for this input
660                         $validate->reset;
661                 }
662
663                 if ($stats) {
664                         my $s = $input_db->stats;
665                         $log->info("statistics of fields usage:\n$s");
666                         print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
667                 }
668
669                 # close MARC file
670                 $marc->finish if ($marc);
671
672                 # close report
673                 close($report_fh) if ($report_fh);
674         }
675
676         $indexer->finish if $indexer && $indexer->can('finish');
677
678         foreach my $out ( @output_modules ) {
679                 $out->finish if $out->can('finish');
680         }
681
682         my $dt = time() - $start_t;
683         $log->info("$total_rows records ", $indexer ? "indexed " : "",
684                 sprintf("in %.2f sec [%.2f rec/sec]",
685                         $dt, ($total_rows / $dt)
686                 )
687         );
688
689
690         # end forked process
691         if ($parallel) {
692                 $log->info("parallel process $$ finished");
693                 exit(0);
694         }
695
696 }
697
698 if ($parallel) {
699         # wait all children to finish
700         sleep(1) while wait != -1;
701         $log->info("all parallel processes finished");
702 }
703
704 # save new delimiters if needed
705 $validate->save_delimiters_templates if ( $validate_delimiters_path );
706
707 #
708 # handle links or merge after indexing
709 #
710
711 if ($merge) {
712         print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
713         close($estcmd_fh);
714         chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
715         system $estcmd_path;
716 } else {
717         foreach my $link (@links) {
718                 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
719                 $link->();
720         }
721 }