/[webpac2]/trunk/run.pl
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Contents of /trunk/run.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 936 - (show annotations)
Wed Oct 31 12:14:03 2007 UTC (16 years, 5 months ago) by dpavlin
File MIME type: text/plain
File size: 17363 byte(s)
 r1413@llin:  dpavlin | 2007-10-31 13:13:58 +0100
 Add input name to all output filters

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

Properties

Name Value
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26