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 0.06;
  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 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 =item --marc-lint
  72
  73 By default turned on if normalisation file has C<marc*> directives. You can disable lint
  74 messages with C<--no-marc-lint>.
  75
  76 =item --marc-dump
  77
  78 Force dump or input and marc record for debugging.
  79
  80 =item --parallel 4
  81
  82 Run databases in parallel (aproximatly same as number of processors in
  83 machine if you want to use full load)
  84
  85 =item --only-links
  86
  87 Create just links
  88
  89 =item --merge
  90
  91 Create merged index of databases which have links
  92
  93 =back
  94
  95 =cut
  96
  97 my $offset;
  98 my $limit;
  99
 100 my $clean = 0;
 101 my $config_path;
 102 my $debug = 0;
 103 my $only_filter;
 104 my $stats = 0;
 105 my $validate_path;
 106 my $marc_lint = 1;
 107 my $marc_dump = 0;
 108 my $parallel = 0;
 109 my $only_links = 0;
 110 my $merge = 0;
 111 my $help;
 112
 113 my $log = _new WebPAC::Common()->_get_logger();
 114
 115 GetOptions(
 116         "limit=i" => \$limit,
 117         "offset=i" => \$offset,
 118         "clean" => \$clean,
 119         "one=s" => \$only_filter,
 120         "only=s" => \$only_filter,
 121         "config" => \$config_path,
 122         "debug+" => \$debug,
 123         "stats" => \$stats,
 124         "validate=s" => \$validate_path,
 125         "marc-lint!" => \$marc_lint,
 126         "marc-dump!" => \$marc_dump,
 127         "parallel=i" => \$parallel,
 128         "only-links!" => \$only_links,
 129         "merge" => \$merge,
 130         "help" => \$help,
 131 );
 132
 133 pod2usage(-verbose => 2) if ($help);
 134
 135 my $config = new WebPAC::Config( path => $config_path );
 136
 137 #print "config = ",dump($config) if ($debug);
 138
 139 die "no databases in config file!\n" unless ($config->databases);
 140
 141 $log->info( "-" x 79 );
 142
 143 my $log_file = 'log';
 144
 145 if (-e $log_file ) {    # && -s $log_file > 5 * 1024 * 1024) {
 146         $log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
 147         rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
 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 $stats ||= $validate;
 168 if ($stats) {
 169         $log->debug("disabled indexing for stats collection");
 170         $use_indexer = undef;
 171 } else {
 172         $log->info("using $use_indexer indexing engine...");
 173 }
 174
 175 # parse normalize files and create source files for lookup and normalization
 176
 177 my $parser = new WebPAC::Parser( config => $config );
 178
 179 my $total_rows = 0;
 180 my $start_t = time();
 181
 182 my @links;
 183
 184 if ($parallel) {
 185         $log->info("Using $parallel processes for speedup");
 186         Proc::Queue::size($parallel);
 187 }
 188
 189 sub create_ds_config {
 190         my ($db_config, $database, $input, $mfn) = @_;
 191         my $c = dclone( $db_config );
 192         $c->{_} = $database || $log->logconfess("need database");
 193         $c->{_mfn} = $mfn || $log->logconfess("need mfn");
 194         $c->{input} = $input || $log->logconfess("need input");
 195         return $c;
 196 }
 197
 198 while (my ($database, $db_config) = each %{ $config->databases }) {
 199
 200         my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
 201         next if ($only_database && $database !~ m/$only_database/i);
 202
 203         if ($parallel) {
 204                 my $f=fork;
 205                 if(defined ($f) and $f==0) {
 206                         $log->info("Created processes $$ for speedup");
 207                 } else {
 208                         next;
 209                 }
 210         }
 211
 212         my $indexer;
 213         if ($use_indexer && $parser->have_rules( 'search', $database )) {
 214
 215                 my $cfg_name = $use_indexer;
 216                 $cfg_name =~ s/\-.*$//;
 217
 218                 my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
 219                 $indexer_config->{database} = $database;
 220                 $indexer_config->{clean} = $clean;
 221                 $indexer_config->{label} = $db_config->{name};
 222
 223                 # force clean if database has links
 224                 $indexer_config->{clean} = 1 if ($db_config->{links});
 225
 226                 if ($use_indexer eq 'hyperestraier') {
 227
 228                         # open Hyper Estraier database
 229                         use WebPAC::Output::Estraier '0.10';
 230                         $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
 231
 232                 } elsif ($use_indexer eq 'hyperestraier-native') {
 233
 234                         # open Hyper Estraier database
 235                         use WebPAC::Output::EstraierNative;
 236                         $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
 237
 238                 } elsif ($use_indexer eq 'kinosearch') {
 239
 240                         # open KinoSearch
 241                         use WebPAC::Output::KinoSearch;
 242                         $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
 243                         $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
 244
 245                 } else {
 246                         $log->logdie("unknown use_indexer: $use_indexer");
 247                 }
 248
 249                 $log->logide("can't continue without valid indexer") unless ($indexer);
 250         }
 251
 252
 253         #
 254         # store Hyper Estraier links to other databases
 255         #
 256         if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
 257                 foreach my $link (@{ $db_config->{links} }) {
 258                         if ($use_indexer eq 'hyperestraier') {
 259                                 if ($merge) {
 260                                         print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
 261                                 } else {
 262                                         $log->info("saving link $database -> $link->{to} [$link->{credit}]");
 263                                         push @links, sub {
 264                                                 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
 265                                                 $indexer->add_link(
 266                                                         from => $database,
 267                                                         to => $link->{to},
 268                                                         credit => $link->{credit},
 269                                                 );
 270                                         };
 271                                 }
 272                         } else {
 273                                 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
 274                         }
 275                 }
 276         }
 277         next if ($only_links);
 278
 279
 280         #
 281         # now WebPAC::Store
 282         #
 283         my $abs_path = abs_path($0);
 284         $abs_path =~ s#/[^/]*$#/#;
 285
 286         my $db_path = $config->webpac('db_path');
 287
 288         if ($clean) {
 289                 $log->info("creating new database '$database' in $db_path");
 290                 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
 291         } else {
 292                 $log->info("working on database '$database' in $db_path");
 293         }
 294
 295         my $store = new WebPAC::Store(
 296                 path => $db_path,
 297                 debug => $debug,
 298         );
 299
 300
 301         #
 302         # now, iterate through input formats
 303         #
 304
 305         my @inputs;
 306         if (ref($db_config->{input}) eq 'ARRAY') {
 307                 @inputs = @{ $db_config->{input} };
 308         } elsif ($db_config->{input}) {
 309                 push @inputs, $db_config->{input};
 310         } else {
 311                 $log->info("database $database doesn't have inputs defined");
 312         }
 313
 314         foreach my $input (@inputs) {
 315
 316                 my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));
 317
 318                 next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
 319
 320                 my $type = lc($input->{type});
 321
 322                 die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
 323
 324                 my $input_module = $config->webpac('inputs')->{$type};
 325
 326                 my @lookups = $parser->have_lookup_create($database, $input);
 327
 328                 $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
 329                         @lookups ? " creating lookups: ".join(", ", @lookups) : ""
 330                 );
 331
 332                 if ($stats) {
 333                         # disable modification of records if --stats is in use
 334                         delete($input->{modify_records});
 335                         delete($input->{modify_file});
 336                 }
 337
 338                 my $input_db = new WebPAC::Input(
 339                         module => $input_module,
 340                         encoding => $config->webpac('webpac_encoding'),
 341                         limit => $limit || $input->{limit},
 342                         offset => $offset,
 343                         recode => $input->{recode},
 344                         stats => $stats,
 345                         modify_records => $input->{modify_records},
 346                         modify_file => $input->{modify_file},
 347                 );
 348                 $log->logdie("can't create input using $input_module") unless ($input);
 349
 350                 if (defined( $input->{lookup} )) {
 351                         $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
 352                         delete( $input->{lookup} );
 353                 }
 354
 355                 my $lookup_coderef;
 356
 357                 if (@lookups) {
 358
 359                         my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
 360
 361                         $lookup_coderef = sub {
 362                                 my $rec = shift || die "need rec!";
 363                                 my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
 364
 365                                 WebPAC::Normalize::data_structure(
 366                                         row => $rec,
 367                                         rules => $rules,
 368                                         config => create_ds_config( $db_config, $database, $input, $mfn ),
 369                                 );
 370
 371                                 #warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
 372                         };
 373
 374                         WebPAC::Normalize::_set_lookup( undef );
 375
 376                         $log->debug("created lookup_coderef using:\n$rules");
 377
 378                 };
 379
 380                 my $lookup_jar;
 381
 382                 my $maxmfn = $input_db->open(
 383                         path => $input->{path},
 384                         code_page => $input->{encoding},        # database encoding
 385                         lookup_coderef => $lookup_coderef,
 386                         lookup => $lookup_jar,
 387                         %{ $input },
 388                         load_row => sub {
 389                                 my $a = shift;
 390                                 return $store->load_row(
 391                                         database => $database,
 392                                         input => $input_name,
 393                                         id => $a->{id},
 394                                 );
 395                         },
 396                         save_row => sub {
 397                                 my $a = shift;
 398                                 return $store->save_row(
 399                                         database => $database,
 400                                         input => $input_name,
 401                                         id => $a->{id},
 402                                         row => $a->{row},
 403                                 );
 404                         },
 405
 406                 );
 407
 408                 my $lookup_data = WebPAC::Normalize::_get_lookup();
 409
 410                 if (defined( $lookup_data->{$database}->{$input_name} )) {
 411                         $log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
 412
 413                         foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
 414                                 $store->save_lookup(
 415                                         database => $database,
 416                                         input => $input_name,
 417                                         key => $key,
 418                                         data => $lookup_data->{$database}->{$input_name}->{$key},
 419                                 );
 420                         }
 421                 }
 422
 423                 my $report_fh;
 424                 if ($stats || $validate) {
 425                         my $path = "out/report/${database}-${input_name}.txt";
 426                         open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
 427
 428                         print $report_fh "Report for database '$database' input '$input_name' records ",
 429                                 $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
 430                         $log->info("Generating report file $path");
 431                 }
 432
 433                 my $marc;
 434                 if ($parser->have_rules( 'marc', $database, $input_name )) {
 435                         $marc = new WebPAC::Output::MARC(
 436                                 path => "out/marc/${database}-${input_name}.marc",
 437                                 lint => $marc_lint,
 438                                 dump => $marc_dump,
 439                         );
 440                 }
 441
 442                 my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
 443                 $log->debug("parsed normalize rules:\n$rules");
 444
 445                 # reset position in database
 446                 $input_db->seek(1);
 447
 448                 # generate name of config key for indexer (strip everything after -)
 449                 my $indexer_config = $use_indexer;
 450                 $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
 451
 452                 my $lookup_hash;
 453                 my $depends = $parser->depends($database,$input_name);
 454
 455                 if ($depends) {
 456                         $log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
 457                         $log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
 458
 459                         foreach my $db (keys %$depends) {
 460                                 foreach my $i (keys %{$depends->{$db}}) {
 461                                         foreach my $k (keys %{$depends->{$db}->{$i}}) {
 462                                                 my $t = time();
 463                                                 $log->debug("loading lookup $db/$i");
 464                                                 $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
 465                                                         database => $db,
 466                                                         input => $i,
 467                                                         key => $k,
 468                                                 );
 469                                                 $log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
 470                                         }
 471                                 }
 472                         }
 473
 474                         $log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
 475                 }
 476
 477
 478                 foreach my $pos ( 0 ... $input_db->size ) {
 479
 480                         my $row = $input_db->fetch || next;
 481
 482                         $total_rows++;
 483
 484                         my $mfn = $row->{'000'}->[0];
 485
 486                         if (! $mfn || $mfn !~ m#^\d+$#) {
 487                                 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
 488                                 $mfn = $pos;
 489                                 push @{ $row->{'000'} }, $pos;
 490                         }
 491
 492
 493                         if ($validate) {
 494                                 if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
 495                                         $log->error( "MFN $mfn validation error:\n",
 496                                                 $validate->report_error( $errors )
 497                                         );
 498                                 }
 499                                 next;   # validation doesn't create any output
 500                         }
 501
 502                         my $ds = WebPAC::Normalize::data_structure(
 503                                 row => $row,
 504                                 rules => $rules,
 505                                 lookup => $lookup_hash,
 506                                 config => create_ds_config( $db_config, $database, $input, $mfn ),
 507                                 marc_encoding => 'utf-8',
 508                                 load_row_coderef => sub {
 509                                         my ($database,$input,$mfn) = @_;
 510                                         return $store->load_row(
 511                                                 database => $database,
 512                                                 input => $input,
 513                                                 id => $mfn,
 514                                         );
 515                                 },
 516                         );
 517
 518                         $log->debug("ds = ", sub { dump($ds) }) if ($ds);
 519
 520                         $store->save_ds(
 521                                 database => $database,
 522                                 input => $input_name,
 523                                 id => $mfn,
 524                                 ds => $ds,
 525                         ) if ($ds && !$stats);
 526
 527                         $indexer->add(
 528                                 id => "${input_name}/${mfn}",
 529                                 ds => $ds,
 530                                 type => $config->get($indexer_config)->{type},
 531                         ) if ($indexer && $ds);
 532
 533                         if ($marc) {
 534                                 my $i = 0;
 535
 536                                 while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
 537                                         $marc->add(
 538                                                 id => $mfn . ( $i ? "/$i" : '' ),
 539                                                 fields => $fields,
 540                                                 leader => WebPAC::Normalize::marc_leader(),
 541                                                 row => $row,
 542                                         );
 543                                         $i++;
 544                                 }
 545
 546                                 $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
 547                         }
 548                 }
 549
 550                 if ($validate) {
 551                         my $errors = $validate->report;
 552                         if ($errors) {
 553                                 $log->info("validation errors:\n$errors\n" );
 554                                 print $report_fh "$errors\n" if ($report_fh);
 555                         }
 556                 }
 557
 558                 if ($stats) {
 559                         my $s = $input_db->stats;
 560                         $log->info("statistics of fields usage:\n$s");
 561                         print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
 562                 }
 563
 564                 # close MARC file
 565                 $marc->finish if ($marc);
 566
 567                 # close report
 568                 close($report_fh) if ($report_fh)
 569
 570         }
 571
 572         eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
 573
 574         my $dt = time() - $start_t;
 575         $log->info("$total_rows records ", $indexer ? "indexed " : "",
 576                 sprintf("in %.2f sec [%.2f rec/sec]",
 577                         $dt, ($total_rows / $dt)
 578                 )
 579         );
 580
 581
 582         # end forked process
 583         if ($parallel) {
 584                 $log->info("parallel process $$ finished");
 585                 exit(0);
 586         }
 587
 588 }
 589
 590 if ($parallel) {
 591         # wait all children to finish
 592         sleep(1) while wait != -1;
 593         $log->info("all parallel processes finished");
 594 }
 595
 596 #
 597 # handle links or merge after indexing
 598 #
 599
 600 if ($merge) {
 601         print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
 602         close($estcmd_fh);
 603         chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
 604         system $estcmd_path;
 605 } else {
 606         foreach my $link (@links) {
 607                 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
 608                 $link->();
 609         }
 610 }