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 Data::Dumper;
   8 use lib './lib';
   9
  10 use WebPAC::Common 0.02;
  11 use WebPAC::Lookup;
  12 use WebPAC::Input 0.03;
  13 use WebPAC::Store 0.03;
  14 use WebPAC::Normalize::XML;
  15 use WebPAC::Normalize::Set;
  16 use WebPAC::Output::TT;
  17 use YAML qw/LoadFile/;
  18 use Getopt::Long;
  19 use File::Path;
  20 use Time::HiRes qw/time/;
  21 use File::Slurp;
  22
  23 =head1 NAME
  24
  25 run.pl - start WebPAC indexing
  26
  27 B<this command will probably go away. Don't get used to it!>
  28
  29 Options:
  30
  31 =over 4
  32
  33 =item --offset 42
  34
  35 start loading (all) databases at offset 42
  36
  37 =item --limit 100
  38
  39 limit loading to 100 records
  40
  41 =item --clean
  42
  43 remove database and Hyper Estraier index before indexing
  44
  45 =item --only=database_name
  46
  47 reindex just single database (legacy name is --one)
  48
  49 =item --config conf/config.yml
  50
  51 path to YAML configuration file
  52
  53 =item --force-set
  54
  55 force conversion C<< normalize->path >> in C<config.yml> from
  56 C<.xml> to C<.pl>
  57
  58 =item --stats
  59
  60 disable indexing and dump statistics about field and subfield
  61 usage for each input
  62
  63 =back
  64
  65 =cut
  66
  67 my $offset;
  68 my $limit;
  69
  70 my $clean = 0;
  71 my $config = 'conf/config.yml';
  72 my $debug = 0;
  73 my $only_db_name;
  74 my $force_set = 0;
  75 my $stats = 0;
  76
  77 GetOptions(
  78         "limit=i" => \$limit,
  79         "offset=i" => \$offset,
  80         "clean" => \$clean,
  81         "one=s" => \$only_db_name,
  82         "only=s" => \$only_db_name,
  83         "config" => \$config,
  84         "debug" => \$debug,
  85         "force-set" => \$force_set,
  86         "stats" => \$stats,
  87 );
  88
  89 $config = LoadFile($config);
  90
  91 print "config = ",Dumper($config) if ($debug);
  92
  93 die "no databases in config file!\n" unless ($config->{databases});
  94
  95 my $log = _new WebPAC::Common()->_get_logger();
  96 $log->info( "-" x 79 );
  97
  98 my $use_indexer = $config->{use_indexer} || 'hyperestraier';
  99 if ($stats) {
 100         $log->debug("option --stats disables update of indexing engine...");
 101         $use_indexer = undef;
 102 } else {
 103         $log->info("using $use_indexer indexing engine...");
 104 }
 105
 106 my $total_rows = 0;
 107 my $start_t = time();
 108
 109 while (my ($database, $db_config) = each %{ $config->{databases} }) {
 110
 111         next if ($only_db_name && $database !~ m/$only_db_name/i);
 112
 113         my $indexer;
 114
 115         if ($use_indexer) {
 116                 my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
 117                 $indexer_config->{database} = $database;
 118                 $indexer_config->{clean} = $clean;
 119                 $indexer_config->{label} = $db_config->{name};
 120
 121                 if ($use_indexer eq 'hyperestraier') {
 122
 123                         # open Hyper Estraier database
 124                         use WebPAC::Output::Estraier '0.10';
 125                         $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
 126
 127                 } elsif ($use_indexer eq 'kinosearch') {
 128
 129                         # open KinoSearch
 130                         use WebPAC::Output::KinoSearch;
 131                         $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
 132                         $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
 133
 134                 } else {
 135                         $log->logdie("unknown use_indexer: $use_indexer");
 136                 }
 137
 138                 $log->logide("can't continue without valid indexer") unless ($indexer);
 139         }
 140
 141
 142         #
 143         # now WebPAC::Store
 144         #
 145         my $abs_path = abs_path($0);
 146         $abs_path =~ s#/[^/]*$#/#;
 147
 148         my $db_path = $config->{webpac}->{db_path} . '/' . $database;
 149
 150         if ($clean) {
 151                 $log->info("creating new database $database in $db_path");
 152                 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
 153         } else {
 154                 $log->debug("working on $database in $db_path");
 155         }
 156
 157         my $db = new WebPAC::Store(
 158                 path => $db_path,
 159                 database => $database,
 160                 debug => $debug,
 161         );
 162
 163
 164         #
 165         # now, iterate through input formats
 166         #
 167
 168         my @inputs;
 169         if (ref($db_config->{input}) eq 'ARRAY') {
 170                 @inputs = @{ $db_config->{input} };
 171         } elsif ($db_config->{input}) {
 172                 push @inputs, $db_config->{input};
 173         } else {
 174                 $log->info("database $database doesn't have inputs defined");
 175         }
 176
 177         my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
 178
 179         foreach my $input (@inputs) {
 180
 181                 my $type = lc($input->{type});
 182
 183                 die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
 184
 185                 my $lookup = new WebPAC::Lookup(
 186                         lookup_file => $input->{lookup},
 187                 );
 188
 189                 my $input_module = $config->{webpac}->{inputs}->{$type};
 190
 191                 $log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'");
 192
 193                 my $input_db = new WebPAC::Input(
 194                         module => $input_module,
 195                         code_page => $config->{webpac}->{webpac_encoding},
 196                         limit => $limit || $input->{limit},
 197                         offset => $offset,
 198                         lookup => $lookup,
 199                         recode => $input->{recode},
 200                         stats => $stats,
 201                 );
 202                 $log->logdie("can't create input using $input_module") unless ($input);
 203
 204                 my $maxmfn = $input_db->open(
 205                         path => $input->{path},
 206                         code_page => $input->{encoding},        # database encoding
 207                 );
 208
 209                 my $n = new WebPAC::Normalize::XML(
 210                 #       filter => { 'foo' => sub { shift } },
 211                         db => $db,
 212                         lookup_regex => $lookup->regex,
 213                         lookup => $lookup,
 214                         prefix => $input->{name},
 215                 );
 216
 217                 my $rules;
 218                 my $normalize_path = $input->{normalize}->{path};
 219
 220                 if ($force_set) {
 221                         my $new_norm_path = $normalize_path;
 222                         $new_norm_path =~ s/\.xml$/.pl/;
 223                         if (-e $new_norm_path) {
 224                                 $log->debug("--force-set replaced $normalize_path with $new_norm_path");
 225                                 $normalize_path = $new_norm_path;
 226                         } else {
 227                                 $log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
 228                         }
 229                 }
 230
 231                 if ($normalize_path =~ m/\.xml$/i) {
 232                         $n->open(
 233                                 tag => $input->{normalize}->{tag},
 234                                 xml_file => $normalize_path,
 235                         );
 236                 } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
 237                         $n->open_yaml(
 238                                 path => $normalize_path,
 239                                 tag => $input->{normalize}->{tag},
 240                         );
 241                 } elsif ($normalize_path =~ m/\.(?:pl)$/i) {
 242                         $n = undef;
 243                         $log->info("using WebPAC::Normalize::Set to process $normalize_path");
 244                         $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
 245                 }
 246
 247                 foreach my $pos ( 0 ... $input_db->size ) {
 248
 249                         my $row = $input_db->fetch || next;
 250
 251                         my $mfn = $row->{'000'}->[0];
 252
 253                         if (! $mfn || $mfn !~ m#^\d+$#) {
 254                                 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
 255                                 $mfn = $pos;
 256                                 push @{ $row->{'000'} }, $pos;
 257                         }
 258
 259
 260                         my $ds;
 261                         if ($n) {
 262                                 $ds = $n->data_structure($row);
 263                         } else {
 264                                 $ds = WebPAC::Normalize::Set::data_structure(
 265                                         row => $row,
 266                                         rules => $rules,
 267                                         lookup => $lookup->lookup_hash,
 268                                 );
 269
 270                                 $db->save_ds(
 271                                         id => $mfn,
 272                                         ds => $ds,
 273                                         prefix => $input->{name},
 274                                 ) if ($ds && !$stats);
 275                         }
 276
 277                         $indexer->add(
 278                                 id => $input->{name} . "/" . $mfn,
 279                                 ds => $ds,
 280                                 type => $config->{$use_indexer}->{type},
 281                         ) if ($indexer);
 282
 283                         $total_rows++;
 284                 }
 285
 286                 $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
 287
 288         };
 289
 290         eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
 291
 292         my $dt = time() - $start_t;
 293         $log->info("$total_rows records indexed in " .
 294                 sprintf("%.2f sec [%.2f rec/sec]",
 295                         $dt, ($total_rows / $dt)
 296                 )
 297         );
 298
 299         #
 300         # add Hyper Estraier links to other databases
 301         #
 302         if (ref($db_config->{links}) eq 'ARRAY') {
 303                 foreach my $link (@{ $db_config->{links} }) {
 304                         if ($use_indexer eq 'hyperestraier') {
 305                                 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
 306                                 $indexer->add_link(
 307                                         from => $database,
 308                                         to => $link->{to},
 309                                         credit => $link->{credit},
 310                                 );
 311                         } else {
 312                                 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
 313                         }
 314                 }
 315         }
 316
 317 }
 318