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 0.07;
  11 use WebPAC::Input 0.13;
  12 use WebPAC::Store 0.11;
  13 use WebPAC::Normalize 0.22;
  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-lint
  71
  72 By default turned on if normalisation file has C<marc*> directives. You can disable lint
  73 messages with C<--no-marc-lint>.
  74
  75 =item --marc-dump
  76
  77 Force dump or input and marc record for debugging.
  78
  79 =item --parallel 4
  80
  81 Run databases in parallel (aproximatly same as number of processors in
  82 machine if you want to use full load)
  83
  84 =item --only-links
  85
  86 Create just links
  87
  88 =item --merge
  89
  90 Create merged index of databases which have links
  91
  92 =back
  93
  94 =cut
  95
  96 my $offset;
  97 my $limit;
  98
  99 my $clean = 0;
 100 my $config_path;
 101 my $debug = 0;
 102 my $only_filter;
 103 my $stats = 0;
 104 my $validate_path;
 105 my $marc_lint = 1;
 106 my $marc_dump = 0;
 107 my $parallel = 0;
 108 my $only_links = 0;
 109 my $merge = 0;
 110
 111 my $log = _new WebPAC::Common()->_get_logger();
 112
 113 GetOptions(
 114         "limit=i" => \$limit,
 115         "offset=i" => \$offset,
 116         "clean" => \$clean,
 117         "one=s" => \$only_filter,
 118         "only=s" => \$only_filter,
 119         "config" => \$config_path,
 120         "debug+" => \$debug,
 121         "stats" => \$stats,
 122         "validate=s" => \$validate_path,
 123         "marc-lint!" => \$marc_lint,
 124         "marc-dump!" => \$marc_dump,
 125         "parallel=i" => \$parallel,
 126         "only-links!" => \$only_links,
 127         "merge" => \$merge,
 128 );
 129
 130 my $config = new WebPAC::Config( path => $config_path );
 131
 132 #print "config = ",dump($config) if ($debug);
 133
 134 die "no databases in config file!\n" unless ($config->databases);
 135
 136 $log->info( "-" x 79 );
 137
 138
 139 my $estcmd_fh;
 140 my $estcmd_path = './estcmd-merge.sh';
 141 if ($merge) {
 142         open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!");
 143         print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/;
 144         print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
 145         $log->info("created merge batch file $estcmd_path");
 146 }
 147
 148
 149 my $validate;
 150 $validate = new WebPAC::Validate(
 151         path => $validate_path,
 152 ) if ($validate_path);
 153
 154
 155 my $use_indexer = $config->use_indexer;
 156 if ($stats) {
 157         $log->debug("option --stats disables update of indexing engine...");
 158         $use_indexer = undef;
 159 } else {
 160         $log->info("using $use_indexer indexing engine...");
 161 }
 162
 163 # parse normalize files and create source files for lookup and normalization
 164
 165 my $parser = new WebPAC::Parser( config => $config );
 166
 167 my $total_rows = 0;
 168 my $start_t = time();
 169
 170 my @links;
 171
 172 if ($parallel) {
 173         $log->info("Using $parallel processes for speedup");
 174         Proc::Queue::size($parallel);
 175 }
 176
 177 sub create_ds_config {
 178         my ($db_config, $database, $input, $mfn) = @_;
 179         my $c = dclone( $db_config );
 180         $c->{_} = $database || $log->logconfess("need database");
 181         $c->{_mfn} = $mfn || $log->logconfess("need mfn");
 182         $c->{input} = $input || $log->logconfess("need input");
 183         return $c;
 184 }
 185
 186 while (my ($database, $db_config) = each %{ $config->databases }) {
 187
 188         my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
 189         next if ($only_database && $database !~ m/$only_database/i);
 190
 191         if ($parallel) {
 192                 my $f=fork;
 193                 if(defined ($f) and $f==0) {
 194                         $log->info("Created processes $$ for speedup");
 195                 } else {
 196                         next;
 197                 }
 198         }
 199
 200         my $indexer;
 201         if ($use_indexer) {
 202
 203                 my $cfg_name = $use_indexer;
 204                 $cfg_name =~ s/\-.*$//;
 205
 206                 my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
 207                 $indexer_config->{database} = $database;
 208                 $indexer_config->{clean} = $clean;
 209                 $indexer_config->{label} = $db_config->{name};
 210
 211                 # force clean if database has links
 212                 $indexer_config->{clean} = 1 if ($db_config->{links});
 213
 214                 if ($use_indexer eq 'hyperestraier') {
 215
 216                         # open Hyper Estraier database
 217                         use WebPAC::Output::Estraier '0.10';
 218                         $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
 219
 220                 } elsif ($use_indexer eq 'hyperestraier-native') {
 221
 222                         # open Hyper Estraier database
 223                         use WebPAC::Output::EstraierNative;
 224                         $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
 225
 226                 } elsif ($use_indexer eq 'kinosearch') {
 227
 228                         # open KinoSearch
 229                         use WebPAC::Output::KinoSearch;
 230                         $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
 231                         $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
 232
 233                 } else {
 234                         $log->logdie("unknown use_indexer: $use_indexer");
 235                 }
 236
 237                 $log->logide("can't continue without valid indexer") unless ($indexer);
 238         }
 239
 240
 241         #
 242         # store Hyper Estraier links to other databases
 243         #
 244         if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
 245                 foreach my $link (@{ $db_config->{links} }) {
 246                         if ($use_indexer eq 'hyperestraier') {
 247                                 if ($merge) {
 248                                         print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
 249                                 } else {
 250                                         $log->info("saving link $database -> $link->{to} [$link->{credit}]");
 251                                         push @links, sub {
 252                                                 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
 253                                                 $indexer->add_link(
 254                                                         from => $database,
 255                                                         to => $link->{to},
 256                                                         credit => $link->{credit},
 257                                                 );
 258                                         };
 259                                 }
 260                         } else {
 261                                 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
 262                         }
 263                 }
 264         }
 265         next if ($only_links);
 266
 267
 268         #
 269         # now WebPAC::Store
 270         #
 271         my $abs_path = abs_path($0);
 272         $abs_path =~ s#/[^/]*$#/#;
 273
 274         my $db_path = $config->webpac('db_path');
 275
 276         if ($clean) {
 277                 $log->info("creating new database '$database' in $db_path");
 278                 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
 279         } else {
 280                 $log->info("working on database '$database' in $db_path");
 281         }
 282
 283         my $store = new WebPAC::Store(
 284                 path => $db_path,
 285                 debug => $debug,
 286         );
 287
 288
 289         #
 290         # now, iterate through input formats
 291         #
 292
 293         my @inputs;
 294         if (ref($db_config->{input}) eq 'ARRAY') {
 295                 @inputs = @{ $db_config->{input} };
 296         } elsif ($db_config->{input}) {
 297                 push @inputs, $db_config->{input};
 298         } else {
 299                 $log->info("database $database doesn't have inputs defined");
 300         }
 301
 302         foreach my $input (@inputs) {
 303
 304                 my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));
 305
 306                 next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
 307
 308                 my $type = lc($input->{type});
 309
 310                 die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
 311
 312                 my $input_module = $config->webpac('inputs')->{$type};
 313
 314                 my @lookups = $parser->have_lookup_create($database, $input);
 315
 316                 $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
 317                         @lookups ? " creating lookups: ".join(", ", @lookups) : ""
 318                 );
 319
 320                 if ($stats) {
 321                         # disable modification of records if --stats is in use
 322                         delete($input->{modify_records});
 323                         delete($input->{modify_file});
 324                 }
 325
 326                 my $input_db = new WebPAC::Input(
 327                         module => $input_module,
 328                         encoding => $config->webpac('webpac_encoding'),
 329                         limit => $limit || $input->{limit},
 330                         offset => $offset,
 331                         recode => $input->{recode},
 332                         stats => $stats,
 333                         modify_records => $input->{modify_records},
 334                         modify_file => $input->{modify_file},
 335                 );
 336                 $log->logdie("can't create input using $input_module") unless ($input);
 337
 338                 if (defined( $input->{lookup} )) {
 339                         $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
 340                         delete( $input->{lookup} );
 341                 }
 342
 343                 my $lookup_coderef;
 344
 345                 if (@lookups) {
 346
 347                         my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
 348
 349                         $lookup_coderef = sub {
 350                                 my $rec = shift || die "need rec!";
 351                                 my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
 352
 353                                 $store->save_row(
 354                                         database => $database,
 355                                         input => $input_name,
 356                                         id => $mfn,
 357                                         row => $rec,
 358                                 );
 359
 360                                 WebPAC::Normalize::data_structure(
 361                                         row => $rec,
 362                                         rules => $rules,
 363                                         config => create_ds_config( $db_config, $database, $input, $mfn ),
 364                                 );
 365
 366                                 #warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
 367                         };
 368
 369                         WebPAC::Normalize::_set_lookup( undef );
 370
 371                         $log->debug("created lookup_coderef using:\n$rules");
 372
 373                 };
 374
 375                 my $lookup_jar;
 376
 377                 my $maxmfn = $input_db->open(
 378                         path => $input->{path},
 379                         code_page => $input->{encoding},        # database encoding
 380                         lookup_coderef => $lookup_coderef,
 381                         lookup => $lookup_jar,
 382                         %{ $input },
 383                 );
 384
 385                 my $lookup_data = WebPAC::Normalize::_get_lookup();
 386
 387                 if (defined( $lookup_data->{$database}->{$input_name} )) {
 388                         $log->debug("created following lookups: ", dump( $lookup_data ));
 389
 390                         foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
 391                                 $store->save_lookup(
 392                                         database => $database,
 393                                         input => $input_name,
 394                                         key => $key,
 395                                         data => $lookup_data->{$database}->{$input_name}->{$key},
 396                                 );
 397                         }
 398                 }
 399
 400                 my $report_fh;
 401                 if ($stats || $validate) {
 402                         my $path = "out/report/${database}-${input_name}.txt";
 403                         open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
 404
 405                         print $report_fh "Report for database '$database' input '$input_name' records ",
 406                                 $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
 407                         $log->info("Generating report file $path");
 408                 }
 409
 410                 my $marc;
 411                 if ($parser->generate_marc( $database, $input_name )) {
 412                         $marc = new WebPAC::Output::MARC(
 413                                 path => "out/marc/${database}-${input_name}.marc",
 414                                 lint => $marc_lint,
 415                                 dump => $marc_dump,
 416                         );
 417                 }
 418
 419                 my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
 420                 $log->debug("parsed normalize rules:\n$rules");
 421
 422                 # reset position in database
 423                 $input_db->seek(1);
 424
 425                 # generate name of config key for indexer (strip everything after -)
 426                 my $indexer_config = $use_indexer;
 427                 $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
 428
 429                 my $lookup_hash;
 430                 my $depends = $parser->depends($database,$input_name);
 431
 432                 if ($depends) {
 433                         $log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
 434                         $log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
 435
 436                         foreach my $db (keys %$depends) {
 437                                 foreach my $i (keys %{$depends->{$db}}) {
 438                                         foreach my $k (keys %{$depends->{$db}->{$i}}) {
 439                                                 $log->debug("loading lookup $db/$i");
 440                                                 $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
 441                                                         database => $db,
 442                                                         input => $i,
 443                                                         key => $k,
 444                                                 );
 445                                         }
 446                                 }
 447                         }
 448
 449                         $log->debug("lookup_hash = ", dump( $lookup_hash ));
 450                 }
 451
 452
 453                 foreach my $pos ( 0 ... $input_db->size ) {
 454
 455                         my $row = $input_db->fetch || next;
 456
 457                         my $mfn = $row->{'000'}->[0];
 458
 459                         if (! $mfn || $mfn !~ m#^\d+$#) {
 460                                 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
 461                                 $mfn = $pos;
 462                                 push @{ $row->{'000'} }, $pos;
 463                         }
 464
 465
 466                         if ($validate) {
 467                                 if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) {
 468                                         $log->error( "MFN $mfn validation error:\n",
 469                                                 $validate->report_error( $errors )
 470                                         );
 471                                 }
 472                         }
 473
 474                         my $ds = WebPAC::Normalize::data_structure(
 475                                 row => $row,
 476                                 rules => $rules,
 477                                 lookup => $lookup_hash,
 478                                 config => create_ds_config( $db_config, $database, $input, $mfn ),
 479                                 marc_encoding => 'utf-8',
 480                                 load_row_coderef => sub {
 481                                         my ($database,$input,$mfn) = @_;
 482                                         return $store->load_row(
 483                                                 database => $database,
 484                                                 input => $input,
 485                                                 id => $mfn,
 486                                         );
 487                                 },
 488                         );
 489
 490                         $log->debug("ds = ",dump($ds)) if ($ds);
 491
 492                         $store->save_ds(
 493                                 database => $database,
 494                                 input => $input_name,
 495                                 id => $mfn,
 496                                 ds => $ds,
 497                         ) if ($ds && !$stats);
 498
 499                         $indexer->add(
 500                                 id => "${input_name}/${mfn}",
 501                                 ds => $ds,
 502                                 type => $config->get($indexer_config)->{type},
 503                         ) if ($indexer && $ds);
 504
 505                         if ($marc) {
 506                                 my $i = 0;
 507
 508                                 while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
 509                                         $marc->add(
 510                                                 id => $mfn . ( $i ? "/$i" : '' ),
 511                                                 fields => $fields,
 512                                                 leader => WebPAC::Normalize::marc_leader(),
 513                                                 row => $row,
 514                                         );
 515                                         $i++;
 516                                 }
 517
 518                                 $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
 519                         }
 520
 521                         $total_rows++;
 522                 }
 523
 524                 if ($validate) {
 525                         my $errors = $validate->report;
 526                         if ($errors) {
 527                                 $log->info("validation errors:\n$errors\n" );
 528                                 print $report_fh "$errors\n" if ($report_fh);
 529                         }
 530                 }
 531
 532                 if ($stats) {
 533                         my $s = $input_db->stats;
 534                         $log->info("statistics of fields usage:\n$s");
 535                         print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
 536                 }
 537
 538                 # close MARC file
 539                 $marc->finish if ($marc);
 540
 541                 # close report
 542                 close($report_fh) if ($report_fh)
 543
 544         }
 545
 546         eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
 547
 548         my $dt = time() - $start_t;
 549         $log->info("$total_rows records ", $indexer ? "indexed " : "",
 550                 sprintf("in %.2f sec [%.2f rec/sec]",
 551                         $dt, ($total_rows / $dt)
 552                 )
 553         );
 554
 555
 556         # end forked process
 557         if ($parallel) {
 558                 $log->info("parallel process $$ finished");
 559                 exit(0);
 560         }
 561
 562 }
 563
 564 if ($parallel) {
 565         # wait all children to finish
 566         sleep(1) while wait != -1;
 567         $log->info("all parallel processes finished");
 568 }
 569
 570 #
 571 # handle links or merge after indexing
 572 #
 573
 574 if ($merge) {
 575         print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
 576         close($estcmd_fh);
 577         chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
 578         system $estcmd_path;
 579 } else {
 580         foreach my $link (@links) {
 581                 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
 582                 $link->();
 583         }
 584 }