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::Output::TT;
  16 use YAML qw/LoadFile/;
  17 use Getopt::Long;
  18 use File::Path;
  19 use Time::HiRes qw/time/;
  20
  21 =head1 NAME
  22
  23 run.pl - start WebPAC indexing
  24
  25 B<this command will probably go away. Don't get used to it!>
  26
  27 Options:
  28
  29 =over 4
  30
  31 =item --offset 42
  32
  33 start loading (all) databases at offset 42
  34
  35 =item --limit 100
  36
  37 limit loading to 100 records
  38
  39 =item --clean
  40
  41 remove database and Hyper Estraier index before indexing
  42
  43 =item --only=database_name
  44
  45 reindex just single database (legacy name is --one)
  46
  47 =item --config conf/config.yml
  48
  49 path to YAML configuration file
  50
  51 =back
  52
  53 =cut
  54
  55 my $offset;
  56 my $limit;
  57
  58 my $clean = 0;
  59 my $config = 'conf/config.yml';
  60 my $debug = 0;
  61 my $only_db_name;
  62
  63 GetOptions(
  64         "limit=i" => \$limit,
  65         "offset=i" => \$offset,
  66         "clean" => \$clean,
  67         "one=s" => \$only_db_name,
  68         "only=s" => \$only_db_name,
  69         "config" => \$config,
  70         "debug" => \$debug,
  71 );
  72
  73 $config = LoadFile($config);
  74
  75 print "config = ",Dumper($config) if ($debug);
  76
  77 die "no databases in config file!\n" unless ($config->{databases});
  78 my $use_indexer = $config->{use_indexer} || 'hyperestraier';
  79
  80 my $total_rows = 0;
  81 my $start_t = time();
  82
  83 while (my ($database, $db_config) = each %{ $config->{databases} }) {
  84
  85         next if ($only_db_name && $database !~ m/$only_db_name/i);
  86
  87         my $log = _new WebPAC::Common()->_get_logger();
  88
  89         my $indexer;
  90
  91         if ($use_indexer eq 'hyperestraier') {
  92
  93                 #
  94                 # open Hyper Estraier database
  95                 #
  96
  97                 use WebPAC::Output::Estraier '0.10';
  98                 my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration");
  99                 $est_config->{database} = $database;
 100                 $est_config->{clean} = $clean;
 101                 $est_config->{label} = $db_config->{name};
 102
 103                 $indexer = new WebPAC::Output::Estraier( %{ $est_config } );
 104         } else {
 105                 $log->logdie("unknown use_indexer: $use_indexer");
 106         }
 107
 108         $log->logide("can't continue without valid indexer") unless ($indexer);
 109
 110         #
 111         # now WebPAC::Store
 112         #
 113         my $abs_path = abs_path($0);
 114         $abs_path =~ s#/[^/]*$#/#;
 115
 116         my $db_path = $config->{webpac}->{db_path} . '/' . $database;
 117
 118         if ($clean) {
 119                 $log->info("creating new database $database in $db_path");
 120                 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
 121         } else {
 122                 $log->debug("working on $database in $db_path");
 123         }
 124
 125         my $db = new WebPAC::Store(
 126                 path => $db_path,
 127                 database => $database,
 128                 debug => $debug,
 129         );
 130
 131
 132         #
 133         # now, iterate through input formats
 134         #
 135
 136         my @inputs;
 137         if (ref($db_config->{input}) eq 'ARRAY') {
 138                 @inputs = @{ $db_config->{input} };
 139         } elsif ($db_config->{input}) {
 140                 push @inputs, $db_config->{input};
 141         } else {
 142                 $log->info("database $database doesn't have inputs defined");
 143         }
 144
 145         my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
 146
 147         foreach my $input (@inputs) {
 148
 149                 my $type = lc($input->{type});
 150
 151                 die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
 152
 153                 my $lookup = new WebPAC::Lookup(
 154                         lookup_file => $input->{lookup},
 155                 );
 156
 157                 my $input_module = $config->{webpac}->{inputs}->{$type};
 158
 159                 $log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'");
 160
 161                 my $input_db = new WebPAC::Input(
 162                         module => $input_module,
 163                         code_page => $config->{webpac}->{webpac_encoding},
 164                         limit => $limit || $input->{limit},
 165                         offset => $offset,
 166                         lookup => $lookup,
 167                         recode => $input->{recode},
 168                 );
 169                 $log->logdie("can't create input using $input_module") unless ($input);
 170
 171                 my $maxmfn = $input_db->open(
 172                         path => $input->{path},
 173                         code_page => $input->{encoding},        # database encoding
 174                 );
 175
 176                 my $n = new WebPAC::Normalize::XML(
 177                 #       filter => { 'foo' => sub { shift } },
 178                         db => $db,
 179                         lookup_regex => $lookup->regex,
 180                         lookup => $lookup,
 181                         prefix => $input->{name},
 182                 );
 183
 184                 my $normalize_path = $input->{normalize}->{path};
 185
 186                 if ($normalize_path =~ m/\.xml$/i) {
 187                         $n->open(
 188                                 tag => $input->{normalize}->{tag},
 189                                 xml_file => $input->{normalize}->{path},
 190                         );
 191                 } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
 192                         $n->open_yaml(
 193                                 path => $normalize_path,
 194                                 tag => $input->{normalize}->{tag},
 195                         );
 196                 }
 197
 198                 foreach my $pos ( 0 ... $input_db->size ) {
 199
 200                         my $row = $input_db->fetch || next;
 201
 202                         my $mfn = $row->{'000'}->[0];
 203
 204                         if (! $mfn || $mfn !~ m#^\d+$#) {
 205                                 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
 206                                 $mfn = $pos;
 207                                 push @{ $row->{'000'} }, $pos;
 208                         }
 209
 210                         my $ds = $n->data_structure($row);
 211
 212                         $indexer->add(
 213                                 id => $input->{name} . "/" . $mfn,
 214                                 ds => $ds,
 215                                 type => $config->{$use_indexer}->{type},
 216                         );
 217
 218                         $total_rows++;
 219                 }
 220
 221         };
 222
 223         my $dt = time() - $start_t;
 224         $log->info("$total_rows records indexed in " .
 225                 sprintf("%.2f sec [%.2f rec/sec]",
 226                         $dt, ($total_rows / $dt)
 227                 )
 228         );
 229
 230         if ($use_indexer eq 'hyperestraier') {
 231                 #
 232                 # add Hyper Estraier links to other databases
 233                 #
 234                 if (ref($db_config->{links}) eq 'ARRAY') {
 235                         foreach my $link (@{ $db_config->{links} }) {
 236                                 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
 237                                 $indexer->add_link(
 238                                         from => $database,
 239                                         to => $link->{to},
 240                                         credit => $link->{credit},
 241                                 );
 242                         }
 243                 }
 244         } else {
 245                 $log->warn("links not implemented for $use_indexer");
 246         }
 247
 248 }
 249