r1018@llin: dpavlin | 2006-09-26 12:20:52 +0200
[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 0.04;
11 use WebPAC::Input 0.13;
12 use WebPAC::Store 0.10;
13 use WebPAC::Normalize 0.11;
14 use WebPAC::Output::TT;
15 use WebPAC::Validate 0.06;
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
25 use Proc::Queue size => 1;
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 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 =item --marc-normalize conf/normalize/mapping.pl
71
72 This option specifies normalisation file for MARC creation
73
74 =item --marc-output out/marc/test.marc
75
76 Optional path to output file
77
78 =item --marc-lint
79
80 By default turned on if C<--marc-normalize> is used. You can disable lint
81 messages with C<--no-marc-lint>.
82
83 =item --marc-dump
84
85 Force dump or input and marc record for debugging.
86
87 =item --parallel 4
88
89 Run databases in parallel (aproximatly same as number of processors in
90 machine if you want to use full load)
91
92 =item --only-links
93
94 Create just links
95
96 =item --merge
97
98 Create merged index of databases which have links
99
100 =back
101
102 =cut
103
104 my $offset;
105 my $limit;
106
107 my $clean = 0;
108 my $config_path;
109 my $debug = 0;
110 my $only_filter;
111 my $stats = 0;
112 my $validate_path;
113 my ($marc_normalize, $marc_output);
114 my $marc_lint = 1;
115 my $marc_dump = 0;
116 my $parallel = 0;
117 my $only_links = 0;
118 my $merge = 0;
119
120 my $log = _new WebPAC::Common()->_get_logger();
121
122 GetOptions(
123         "limit=i" => \$limit,
124         "offset=i" => \$offset,
125         "clean" => \$clean,
126         "one=s" => \$only_filter,
127         "only=s" => \$only_filter,
128         "config" => \$config_path,
129         "debug+" => \$debug,
130         "stats" => \$stats,
131         "validate=s" => \$validate_path,
132         "marc-normalize=s" => \$marc_normalize,
133         "marc-output=s" => \$marc_output,
134         "marc-lint!" => \$marc_lint,
135         "marc-dump!" => \$marc_dump,
136         "parallel=i" => \$parallel,
137         "only-links!" => \$only_links,
138         "merge" => \$merge,
139 );
140
141 my $config = new WebPAC::Config( path => $config_path );
142
143 #print "config = ",dump($config) if ($debug);
144
145 die "no databases in config file!\n" unless ($config->databases);
146
147 $log->info( "-" x 79 );
148
149
150 my $estcmd_fh;
151 my $estcmd_path = './estcmd-merge.sh';
152 if ($merge) {
153         open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!");
154         print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/;
155         print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
156         $log->info("created merge batch file $estcmd_path");
157 }
158
159
160 my $validate;
161 $validate = new WebPAC::Validate(
162         path => $validate_path,
163 ) if ($validate_path);
164
165
166 my $use_indexer = $config->use_indexer;
167 if ($stats) {
168         $log->debug("option --stats disables update of indexing engine...");
169         $use_indexer = undef;
170 } else {
171         $log->info("using $use_indexer indexing engine...");
172 }
173
174 # disable indexing when creating marc
175 $use_indexer = undef if ($marc_normalize);
176
177 # parse normalize files and create source files for lookup and normalization
178
179 my $parser = new WebPAC::Parser( config => $config );
180
181 my $total_rows = 0;
182 my $start_t = time();
183
184 my @links;
185
186 if ($parallel) {
187         $log->info("Using $parallel processes for speedup");
188         Proc::Queue::size($parallel);
189 }
190
191 sub create_ds_config {
192         my ($db_config, $database, $input, $mfn) = @_;
193         my $c = dclone( $db_config );
194         $c->{_} = $database || $log->logconfess("need database");
195         $c->{_mfn} = $mfn || $log->logconfess("need mfn");
196         $c->{input} = $input || $log->logconfess("need input");
197         return $c;
198 }
199
200 while (my ($database, $db_config) = each %{ $config->databases }) {
201
202         my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
203         next if ($only_database && $database !~ m/$only_database/i);
204
205         if ($parallel) {
206                 my $f=fork;
207                 if(defined ($f) and $f==0) {
208                         $log->info("Created processes $$ for speedup");
209                 } else {
210                         next;
211                 }
212         }
213
214         my $indexer;
215         if ($use_indexer) {
216
217                 my $cfg_name = $use_indexer;
218                 $cfg_name =~ s/\-.*$//;
219
220                 my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
221                 $indexer_config->{database} = $database;
222                 $indexer_config->{clean} = $clean;
223                 $indexer_config->{label} = $db_config->{name};
224
225                 # force clean if database has links
226                 $indexer_config->{clean} = 1 if ($db_config->{links});
227
228                 if ($use_indexer eq 'hyperestraier') {
229
230                         # open Hyper Estraier database
231                         use WebPAC::Output::Estraier '0.10';
232                         $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
233                 
234                 } elsif ($use_indexer eq 'hyperestraier-native') {
235
236                         # open Hyper Estraier database
237                         use WebPAC::Output::EstraierNative;
238                         $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
239
240                 } elsif ($use_indexer eq 'kinosearch') {
241
242                         # open KinoSearch
243                         use WebPAC::Output::KinoSearch;
244                         $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
245                         $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
246
247                 } else {
248                         $log->logdie("unknown use_indexer: $use_indexer");
249                 }
250
251                 $log->logide("can't continue without valid indexer") unless ($indexer);
252         }
253
254
255         #
256         # store Hyper Estraier links to other databases
257         #
258         if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
259                 foreach my $link (@{ $db_config->{links} }) {
260                         if ($use_indexer eq 'hyperestraier') {
261                                 if ($merge) {
262                                         print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
263                                 } else {
264                                         $log->info("saving link $database -> $link->{to} [$link->{credit}]");
265                                         push @links, sub {
266                                                 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
267                                                 $indexer->add_link(
268                                                         from => $database,
269                                                         to => $link->{to},
270                                                         credit => $link->{credit},
271                                                 );
272                                         };
273                                 }
274                         } else {
275                                 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
276                         }
277                 }
278         }
279         next if ($only_links);
280
281
282         #
283         # now WebPAC::Store
284         #
285         my $abs_path = abs_path($0);
286         $abs_path =~ s#/[^/]*$#/#;
287
288         my $db_path = $config->get('webpac')->{db_path} . '/' . $database;
289
290         if ($clean) {
291                 $log->info("creating new database '$database' in $db_path");
292                 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
293         } else {
294                 $log->info("working on database '$database' in $db_path");
295         }
296
297         my $store = new WebPAC::Store(
298                 path => $db_path,
299                 database => $database,
300                 debug => $debug,
301         );
302
303
304         #
305         # now, iterate through input formats
306         #
307
308         my @inputs;
309         if (ref($db_config->{input}) eq 'ARRAY') {
310                 @inputs = @{ $db_config->{input} };
311         } elsif ($db_config->{input}) {
312                 push @inputs, $db_config->{input};
313         } else {
314                 $log->info("database $database doesn't have inputs defined");
315         }
316
317         foreach my $input (@inputs) {
318
319                 my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));
320
321                 next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
322
323                 my $type = lc($input->{type});
324
325                 die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
326
327                 my $input_module = $config->webpac('inputs')->{$type};
328
329                 my @lookups = $parser->have_lookup_create($database, $input);
330
331                 $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
332                         @lookups ? " creating lookups: ".join(", ", @lookups) : ""
333                 );
334
335                 if ($stats) {
336                         # disable modification of records if --stats is in use
337                         delete($input->{modify_records});
338                         delete($input->{modify_file});
339                 }
340
341                 my $input_db = new WebPAC::Input(
342                         module => $input_module,
343                         encoding => $config->webpac('webpac_encoding'),
344                         limit => $limit || $input->{limit},
345                         offset => $offset,
346                         recode => $input->{recode},
347                         stats => $stats,
348                         modify_records => $input->{modify_records},
349                         modify_file => $input->{modify_file},
350                 );
351                 $log->logdie("can't create input using $input_module") unless ($input);
352
353                 if (defined( $input->{lookup} )) {
354                         $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
355                         delete( $input->{lookup} );
356                 }
357
358                 my $lookup;
359                 my $lookup_coderef;
360
361                 if (@lookups) {
362
363                         my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
364
365                         $lookup_coderef = sub {
366                                 my $rec = shift || die "need rec!";
367                                 my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
368
369                                 WebPAC::Normalize::data_structure(
370                                         row => $rec,
371                                         rules => $rules,
372                                         lookup => $lookup,
373                                         config => create_ds_config( $db_config, $database, $input, $mfn ),
374                                 );
375
376                                 warn "current lookup = ", dump($lookup) if ($lookup);
377                         };
378
379                         WebPAC::Normalize::_set_lookup( undef );
380
381                         $log->debug("created lookup_coderef using:\n$rules");
382
383                 };
384
385                 my $maxmfn = $input_db->open(
386                         path => $input->{path},
387                         code_page => $input->{encoding},        # database encoding
388                         lookup_coderef => $lookup_coderef,
389                         %{ $input },
390                 );
391
392                 my $lookup_data = WebPAC::Normalize::_get_lookup();
393
394                 if (defined( $lookup_data->{$database}->{$input_name} )) {
395                         $log->debug("created following lookups: ", dump( $lookup_data ));
396
397                         foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
398                                 $store->save_lookup( $database, $input_name, $key, $lookup_data->{$database}->{$input_name}->{$key} );
399                         }
400                 }
401
402                 my $report_fh;
403                 if ($stats || $validate) {
404                         my $path = "out/report/${database}-${input_name}.txt";
405                         open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
406
407                         print $report_fh "Report for database '$database' input '$input_name' records ",
408                                 $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
409                         $log->info("Generating report file $path");
410                 }
411
412                 my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
413                         @{ $input->{normalize} } : ( $input->{normalize} );
414
415                 if ($marc_normalize) {
416                         @norm_array = ( {
417                                 path => $marc_normalize,
418                                 output => $marc_output || "out/marc/${database}-${input_name}.marc",
419                         } );
420                 }
421
422                 foreach my $normalize (@norm_array) {
423
424                         my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config");
425
426                         $log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
427
428                         my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
429
430                         $log->info("Using $normalize_path for normalization...");
431
432                         my $marc = new WebPAC::Output::MARC(
433                                 path => $normalize->{output},
434                                 lint => $marc_lint,
435                                 dump => $marc_dump,
436                         ) if ($normalize->{output});
437
438                         # reset position in database
439                         $input_db->seek(1);
440
441                         # generate name of config key for indexer (strip everything after -)
442                         my $indexer_config = $use_indexer;
443                         $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
444
445                         foreach my $pos ( 0 ... $input_db->size ) {
446
447                                 my $row = $input_db->fetch || next;
448
449                                 my $mfn = $row->{'000'}->[0];
450
451                                 if (! $mfn || $mfn !~ m#^\d+$#) {
452                                         $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
453                                         $mfn = $pos;
454                                         push @{ $row->{'000'} }, $pos;
455                                 }
456
457
458                                 if ($validate) {
459                                         if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) {
460                                                 $log->error( "MFN $mfn validation error:\n",
461                                                         $validate->report_error( $errors )
462                                                 );
463                                         }
464                                 }
465
466                                 my $ds = WebPAC::Normalize::data_structure(
467                                         row => $row,
468                                         rules => $rules,
469                                         lookup => $lookup ? $lookup->lookup_hash : undef,
470                                         config => create_ds_config( $db_config, $database, $input, $mfn ),
471                                         marc_encoding => 'utf-8',
472                                 );
473
474                                 $store->save_ds(
475                                         id => $mfn,
476                                         ds => $ds,
477                                         prefix => $input_name,
478                                 ) if ($ds && !$stats);
479
480                                 $indexer->add(
481                                         id => "${input_name}/${mfn}",
482                                         ds => $ds,
483                                         type => $config->get($indexer_config)->{type},
484                                 ) if ($indexer && $ds);
485
486                                 if ($marc) {
487                                         my $i = 0;
488
489                                         while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
490                                                 $marc->add(
491                                                         id => $mfn . ( $i ? "/$i" : '' ),
492                                                         fields => $fields,
493                                                         leader => WebPAC::Normalize::marc_leader(),
494                                                         row => $row,
495                                                 );
496                                                 $i++;
497                                         }
498
499                                         $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
500                                 }
501
502                                 $total_rows++;
503                         }
504
505                         if ($validate) {
506                                 my $errors = $validate->report;
507                                 if ($errors) {
508                                         $log->info("validation errors:\n$errors\n" );
509                                         print $report_fh "$errors\n" if ($report_fh);
510                                 }
511                         }
512
513                         if ($stats) {
514                                 my $s = $input_db->stats;
515                                 $log->info("statistics of fields usage:\n$s");
516                                 print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
517                         }
518
519                         # close MARC file
520                         $marc->finish if ($marc);
521
522                         # close report
523                         close($report_fh) if ($report_fh)
524                 }
525
526         }
527
528         eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
529
530         my $dt = time() - $start_t;
531         $log->info("$total_rows records ", $indexer ? "indexed " : "",
532                 sprintf("in %.2f sec [%.2f rec/sec]",
533                         $dt, ($total_rows / $dt)
534                 )
535         );
536
537
538         # end forked process
539         if ($parallel) {
540                 $log->info("parallel process $$ finished");
541                 exit(0);
542         }
543
544 }
545
546 if ($parallel) {
547         # wait all children to finish
548         sleep(1) while wait != -1;
549         $log->info("all parallel processes finished");
550 }
551
552 #
553 # handle links or merge after indexing
554 #
555
556 if ($merge) {
557         print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
558         close($estcmd_fh);
559         chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
560         system $estcmd_path;
561 } else {
562         foreach my $link (@links) {
563                 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
564                 $link->();
565         }
566 }