git.rot13.org Git - webpac2/blob - 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 0.14;
  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         path => $validate_path,
 181         delimiters => $config->webpac('delimiters'),
 182         delimiters_path => $validate_delimiters_path,
 183 ) if ($validate_path || $validate_delimiters_path);
 184
 185 my $use_indexer = $config->use_indexer;
 186 $stats ||= $validate;
 187 if ($stats) {
 188         $log->debug("disabled indexing for stats collection");
 189         $use_indexer = undef;
 190 } else {
 191         $log->info("using $use_indexer indexing engine...");
 192 }
 193
 194 # parse normalize files and create source files for lookup and normalization
 195
 196 my $parser = new WebPAC::Parser( config => $config );
 197
 198 my $total_rows = 0;
 199 my $start_t = time();
 200
 201 my @links;
 202
 203 if ($parallel) {
 204         $log->info("Using $parallel processes for speedup");
 205         Proc::Queue::size($parallel);
 206 }
 207
 208 sub create_ds_config {
 209         my ($db_config, $database, $input, $mfn) = @_;
 210         my $c = dclone( $db_config );
 211         $c->{_} = $database || $log->logconfess("need database");
 212         $c->{_mfn} = $mfn || $log->logconfess("need mfn");
 213         $c->{input} = $input || $log->logconfess("need input");
 214         return $c;
 215 }
 216
 217 while (my ($database, $db_config) = each %{ $config->databases }) {
 218
 219         my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
 220         next if ($only_database && $database !~ m/$only_database/i);
 221
 222         if ($parallel) {
 223                 my $f=fork;
 224                 if(defined ($f) and $f==0) {
 225                         $log->info("Created processes $$ for speedup");
 226                 } else {
 227                         next;
 228                 }
 229         }
 230
 231         my $indexer;
 232         if ($use_indexer && $parser->have_rules( 'search', $database )) {
 233
 234                 my $cfg_name = $use_indexer;
 235                 $cfg_name =~ s/\-.*$//;
 236
 237                 my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
 238                 $indexer_config->{database} = $database;
 239                 $indexer_config->{clean} = $clean;
 240                 $indexer_config->{label} = $db_config->{name};
 241
 242                 # force clean if database has links
 243                 $indexer_config->{clean} = 1 if ($db_config->{links});
 244
 245                 if ($use_indexer eq 'hyperestraier') {
 246
 247                         # open Hyper Estraier database
 248                         use WebPAC::Output::Estraier '0.10';
 249                         $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
 250
 251                 } elsif ($use_indexer eq 'hyperestraier-native') {
 252
 253                         # open Hyper Estraier database
 254                         use WebPAC::Output::EstraierNative;
 255                         $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
 256
 257                 } elsif ($use_indexer eq 'kinosearch') {
 258
 259                         # open KinoSearch
 260                         use WebPAC::Output::KinoSearch;
 261                         $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
 262                         $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
 263
 264                 } else {
 265                         $log->logdie("unknown use_indexer: $use_indexer");
 266                 }
 267
 268                 $log->logide("can't continue without valid indexer") unless ($indexer);
 269         }
 270
 271
 272         #
 273         # store Hyper Estraier links to other databases
 274         #
 275         if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
 276                 foreach my $link (@{ $db_config->{links} }) {
 277                         if ($use_indexer eq 'hyperestraier') {
 278                                 if ($merge) {
 279                                         print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
 280                                 } else {
 281                                         $log->info("saving link $database -> $link->{to} [$link->{credit}]");
 282                                         push @links, sub {
 283                                                 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
 284                                                 $indexer->add_link(
 285                                                         from => $database,
 286                                                         to => $link->{to},
 287                                                         credit => $link->{credit},
 288                                                 );
 289                                         };
 290                                 }
 291                         } else {
 292                                 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
 293                         }
 294                 }
 295         }
 296         next if ($only_links);
 297
 298
 299         #
 300         # now WebPAC::Store
 301         #
 302         my $abs_path = abs_path($0);
 303         $abs_path =~ s#/[^/]*$#/#;
 304
 305         my $db_path = $config->webpac('db_path');
 306
 307         if ($clean) {
 308                 $log->info("creating new database '$database' in $db_path");
 309                 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
 310         } else {
 311                 $log->info("working on database '$database' in $db_path");
 312         }
 313
 314         my $store = new WebPAC::Store(
 315                 path => $db_path,
 316                 debug => $debug,
 317         );
 318
 319
 320         #
 321         # now, iterate through input formats
 322         #
 323
 324         my @inputs;
 325         if (ref($db_config->{input}) eq 'ARRAY') {
 326                 @inputs = @{ $db_config->{input} };
 327         } elsif ($db_config->{input}) {
 328                 push @inputs, $db_config->{input};
 329         } else {
 330                 $log->info("database $database doesn't have inputs defined");
 331         }
 332
 333         foreach my $input (@inputs) {
 334
 335                 my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));
 336
 337                 next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
 338
 339                 my $type = lc($input->{type});
 340
 341                 die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
 342
 343                 my $input_module = $config->webpac('inputs')->{$type};
 344
 345                 my @lookups = $parser->have_lookup_create($database, $input);
 346
 347                 $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
 348                         @lookups ? " creating lookups: ".join(", ", @lookups) : ""
 349                 );
 350
 351                 if ($stats) {
 352                         # disable modification of records if --stats is in use
 353                         delete($input->{modify_records});
 354                         delete($input->{modify_file});
 355                 }
 356
 357                 my $input_db = new WebPAC::Input(
 358                         module => $input_module,
 359                         encoding => $config->webpac('webpac_encoding'),
 360                         limit => $limit || $input->{limit},
 361                         offset => $offset,
 362                         recode => $input->{recode},
 363                         stats => $stats,
 364                         modify_records => $input->{modify_records},
 365                         modify_file => $input->{modify_file},
 366                 );
 367                 $log->logdie("can't create input using $input_module") unless ($input);
 368
 369                 if (defined( $input->{lookup} )) {
 370                         $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
 371                         delete( $input->{lookup} );
 372                 }
 373
 374                 my $lookup_coderef;
 375
 376                 if (@lookups) {
 377
 378                         my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
 379
 380                         $lookup_coderef = sub {
 381                                 my $rec = shift || die "need rec!";
 382                                 my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
 383
 384                                 WebPAC::Normalize::data_structure(
 385                                         row => $rec,
 386                                         rules => $rules,
 387                                         config => create_ds_config( $db_config, $database, $input, $mfn ),
 388                                 );
 389
 390                                 #warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
 391                         };
 392
 393                         WebPAC::Normalize::_set_lookup( undef );
 394
 395                         $log->debug("created lookup_coderef using:\n$rules");
 396
 397                 };
 398
 399                 my $lookup_jar;
 400
 401                 my $maxmfn = $input_db->open(
 402                         path => $input->{path},
 403                         code_page => $input->{encoding},        # database encoding
 404                         lookup_coderef => $lookup_coderef,
 405                         lookup => $lookup_jar,
 406                         %{ $input },
 407                         load_row => sub {
 408                                 my $a = shift;
 409                                 return $store->load_row(
 410                                         database => $database,
 411                                         input => $input_name,
 412                                         id => $a->{id},
 413                                 );
 414                         },
 415                         save_row => sub {
 416                                 my $a = shift;
 417                                 return $store->save_row(
 418                                         database => $database,
 419                                         input => $input_name,
 420                                         id => $a->{id},
 421                                         row => $a->{row},
 422                                 );
 423                         },
 424
 425                 );
 426
 427                 my $lookup_data = WebPAC::Normalize::_get_lookup();
 428
 429                 if (defined( $lookup_data->{$database}->{$input_name} )) {
 430                         $log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
 431
 432                         foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
 433                                 $store->save_lookup(
 434                                         database => $database,
 435                                         input => $input_name,
 436                                         key => $key,
 437                                         data => $lookup_data->{$database}->{$input_name}->{$key},
 438                                 );
 439                         }
 440                 }
 441
 442                 my $report_fh;
 443                 if ($stats || $validate) {
 444                         my $path = "out/report/${database}-${input_name}.txt";
 445                         open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
 446
 447                         print $report_fh "Report for database '$database' input '$input_name' records ",
 448                                 $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
 449                         $log->info("Generating report file $path");
 450                 }
 451
 452                 my $marc;
 453                 if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) {
 454                         $marc = new WebPAC::Output::MARC(
 455                                 path => "out/marc/${database}-${input_name}.marc",
 456                                 lint => $marc_lint,
 457                                 dump => $marc_dump,
 458                         );
 459                 }
 460
 461                 my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
 462                 $log->debug("parsed normalize rules:\n$rules");
 463
 464                 # reset position in database
 465                 $input_db->seek(1);
 466
 467                 # generate name of config key for indexer (strip everything after -)
 468                 my $indexer_config = $use_indexer;
 469                 $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
 470
 471                 my $lookup_hash;
 472                 my $depends = $parser->depends($database,$input_name);
 473
 474                 if ($depends) {
 475                         $log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
 476                         $log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
 477
 478                         foreach my $db (keys %$depends) {
 479                                 foreach my $i (keys %{$depends->{$db}}) {
 480                                         foreach my $k (keys %{$depends->{$db}->{$i}}) {
 481                                                 my $t = time();
 482                                                 $log->debug("loading lookup $db/$i");
 483                                                 $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
 484                                                         database => $db,
 485                                                         input => $i,
 486                                                         key => $k,
 487                                                 );
 488                                                 $log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
 489                                         }
 490                                 }
 491                         }
 492
 493                         $log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
 494                 }
 495
 496
 497                 foreach my $pos ( 0 ... $input_db->size ) {
 498
 499                         my $row = $input_db->fetch || next;
 500
 501                         $total_rows++;
 502
 503                         my $mfn = $row->{'000'}->[0];
 504
 505                         if (! $mfn || $mfn !~ m#^\d+$#) {
 506                                 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
 507                                 $mfn = $pos;
 508                                 push @{ $row->{'000'} }, $pos;
 509                         }
 510
 511
 512                         if ($validate) {
 513                                 if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
 514                                         $log->error( "MFN $mfn validation error:\n",
 515                                                 $validate->report_error( $errors )
 516                                         );
 517                                 }
 518                                 next;   # validation doesn't create any output
 519                         }
 520
 521                         my $ds = WebPAC::Normalize::data_structure(
 522                                 row => $row,
 523                                 rules => $rules,
 524                                 lookup => $lookup_hash,
 525                                 config => create_ds_config( $db_config, $database, $input, $mfn ),
 526                                 marc_encoding => 'utf-8',
 527                                 load_row_coderef => sub {
 528                                         my ($database,$input,$mfn) = @_;
 529                                         return $store->load_row(
 530                                                 database => $database,
 531                                                 input => $input,
 532                                                 id => $mfn,
 533                                         );
 534                                 },
 535                         );
 536
 537                         $log->debug("ds = ", sub { dump($ds) }) if ($ds);
 538
 539                         $store->save_ds(
 540                                 database => $database,
 541                                 input => $input_name,
 542                                 id => $mfn,
 543                                 ds => $ds,
 544                         ) if ($ds && !$stats);
 545
 546                         $indexer->add(
 547                                 id => "${input_name}/${mfn}",
 548                                 ds => $ds,
 549                                 type => $config->get($indexer_config)->{type},
 550                         ) if ($indexer && $ds);
 551
 552                         if ($marc) {
 553                                 my $i = 0;
 554
 555                                 while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
 556                                         $marc->add(
 557                                                 id => $mfn . ( $i ? "/$i" : '' ),
 558                                                 fields => $fields,
 559                                                 leader => WebPAC::Normalize::_get_marc_leader(),
 560                                                 row => $row,
 561                                         );
 562                                         $i++;
 563                                 }
 564
 565                                 $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
 566                         }
 567                 }
 568
 569                 if ($validate) {
 570                         my $errors = $validate->report;
 571                         if ($errors) {
 572                                 $log->info("validation errors:\n$errors\n" );
 573                                 print $report_fh "$errors\n" if ($report_fh);
 574                         }
 575
 576                         print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n";
 577
 578                         # must be last thing that touches $validate for this input
 579                         $validate->reset;
 580                 }
 581
 582                 if ($stats) {
 583                         my $s = $input_db->stats;
 584                         $log->info("statistics of fields usage:\n$s");
 585                         print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
 586                 }
 587
 588                 # close MARC file
 589                 $marc->finish if ($marc);
 590
 591                 # close report
 592                 close($report_fh) if ($report_fh);
 593         }
 594
 595         eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
 596
 597         my $dt = time() - $start_t;
 598         $log->info("$total_rows records ", $indexer ? "indexed " : "",
 599                 sprintf("in %.2f sec [%.2f rec/sec]",
 600                         $dt, ($total_rows / $dt)
 601                 )
 602         );
 603
 604
 605         # end forked process
 606         if ($parallel) {
 607                 $log->info("parallel process $$ finished");
 608                 exit(0);
 609         }
 610
 611 }
 612
 613 if ($parallel) {
 614         # wait all children to finish
 615         sleep(1) while wait != -1;
 616         $log->info("all parallel processes finished");
 617 }
 618
 619 # save new delimiters if needed
 620 $validate->save_delimiters_templates if ( $validate );
 621
 622 #
 623 # handle links or merge after indexing
 624 #
 625
 626 if ($merge) {
 627         print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
 628         close($estcmd_fh);
 629         chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
 630         system $estcmd_path;
 631 } else {
 632         foreach my $link (@links) {
 633                 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
 634                 $link->();
 635         }
 636 }