| 4 |
|
|
| 5 |
use Cwd qw/abs_path/; |
use Cwd qw/abs_path/; |
| 6 |
use File::Temp qw/tempdir/; |
use File::Temp qw/tempdir/; |
|
use Data::Dumper; |
|
| 7 |
use lib './lib'; |
use lib './lib'; |
| 8 |
|
|
| 9 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
| 10 |
use WebPAC::Lookup; |
use WebPAC::Lookup; |
| 11 |
use WebPAC::Input 0.03; |
use WebPAC::Input 0.03; |
| 12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.03; |
| 13 |
use WebPAC::Normalize::XML; |
use WebPAC::Normalize; |
| 14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
| 15 |
use WebPAC::Output::Estraier 0.05; |
use WebPAC::Validate; |
| 16 |
use YAML qw/LoadFile/; |
use YAML qw/LoadFile/; |
| 17 |
use LWP::Simple; |
use Getopt::Long; |
| 18 |
|
use File::Path; |
| 19 |
|
use Time::HiRes qw/time/; |
| 20 |
|
use File::Slurp; |
| 21 |
|
use MARC::Record 2.0; # need 2.0 for utf-8 encoding see marcpm.sf.net |
| 22 |
|
use MARC::Lint; |
| 23 |
|
use Data::Dump qw/dump/; |
| 24 |
|
|
| 25 |
my $limit = shift @ARGV; |
=head1 NAME |
| 26 |
|
|
| 27 |
my $config = LoadFile('conf/config.yml'); |
run.pl - start WebPAC indexing |
| 28 |
|
|
| 29 |
print "config = ",Dumper($config); |
B<this command will probably go away. Don't get used to it!> |
| 30 |
|
|
| 31 |
|
Options: |
| 32 |
|
|
| 33 |
|
=over 4 |
| 34 |
|
|
| 35 |
|
=item --offset 42 |
| 36 |
|
|
| 37 |
|
start loading (all) databases at offset 42 |
| 38 |
|
|
| 39 |
|
=item --limit 100 |
| 40 |
|
|
| 41 |
|
limit loading to 100 records |
| 42 |
|
|
| 43 |
|
=item --clean |
| 44 |
|
|
| 45 |
|
remove database and Hyper Estraier index before indexing |
| 46 |
|
|
| 47 |
|
=item --only=database_name/input_filter |
| 48 |
|
|
| 49 |
|
reindex just single database (legacy name is --one) |
| 50 |
|
|
| 51 |
|
C</input_filter> is optional part which can be C<name> |
| 52 |
|
or C<type> from input |
| 53 |
|
|
| 54 |
|
=item --config conf/config.yml |
| 55 |
|
|
| 56 |
|
path to YAML configuration file |
| 57 |
|
|
| 58 |
|
=item --stats |
| 59 |
|
|
| 60 |
|
disable indexing and dump statistics about field and subfield |
| 61 |
|
usage for each input |
| 62 |
|
|
| 63 |
|
=item --validate path/to/validation_file |
| 64 |
|
|
| 65 |
|
turn on extra validation of imput records, see L<WebPAC::Validation> |
| 66 |
|
|
| 67 |
|
=item --marc-normalize conf/normalize/mapping.pl |
| 68 |
|
|
| 69 |
|
This option specifies normalisation file for MARC creation |
| 70 |
|
|
| 71 |
|
=item --marc-output out/marc/test.marc |
| 72 |
|
|
| 73 |
|
Optional path to output file |
| 74 |
|
|
| 75 |
|
=item --marc-lint |
| 76 |
|
|
| 77 |
|
By default turned on if C<--marc-normalize> is used. You can disable lint |
| 78 |
|
messages with C<--no-marc-lint>. |
| 79 |
|
|
| 80 |
|
=item --marc-dump |
| 81 |
|
|
| 82 |
|
Force dump or input and marc record for debugging. |
| 83 |
|
|
| 84 |
|
=back |
| 85 |
|
|
| 86 |
|
=cut |
| 87 |
|
|
| 88 |
|
my $offset; |
| 89 |
|
my $limit; |
| 90 |
|
|
| 91 |
|
my $clean = 0; |
| 92 |
|
my $config = 'conf/config.yml'; |
| 93 |
|
my $debug = 0; |
| 94 |
|
my $only_filter; |
| 95 |
|
my $stats = 0; |
| 96 |
|
my $validate_path; |
| 97 |
|
my ($marc_normalize, $marc_output); |
| 98 |
|
my $marc_lint = 1; |
| 99 |
|
my $marc_dump = 0; |
| 100 |
|
|
| 101 |
|
GetOptions( |
| 102 |
|
"limit=i" => \$limit, |
| 103 |
|
"offset=i" => \$offset, |
| 104 |
|
"clean" => \$clean, |
| 105 |
|
"one=s" => \$only_filter, |
| 106 |
|
"only=s" => \$only_filter, |
| 107 |
|
"config" => \$config, |
| 108 |
|
"debug+" => \$debug, |
| 109 |
|
"stats" => \$stats, |
| 110 |
|
"validate=s" => \$validate_path, |
| 111 |
|
"marc-normalize=s" => \$marc_normalize, |
| 112 |
|
"marc-output=s" => \$marc_output, |
| 113 |
|
"marc-lint!" => \$marc_lint, |
| 114 |
|
"marc-dump!" => \$marc_dump, |
| 115 |
|
); |
| 116 |
|
|
| 117 |
|
$config = LoadFile($config); |
| 118 |
|
|
| 119 |
|
print "config = ",dump($config) if ($debug); |
| 120 |
|
|
| 121 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->{databases}); |
| 122 |
|
|
| 123 |
|
my $log = _new WebPAC::Common()->_get_logger(); |
| 124 |
|
$log->info( "-" x 79 ); |
| 125 |
|
|
| 126 |
|
my $validate; |
| 127 |
|
$validate = new WebPAC::Validate( |
| 128 |
|
path => $validate_path, |
| 129 |
|
) if ($validate_path); |
| 130 |
|
|
| 131 |
|
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
| 132 |
|
if ($stats) { |
| 133 |
|
$log->debug("option --stats disables update of indexing engine..."); |
| 134 |
|
$use_indexer = undef; |
| 135 |
|
} else { |
| 136 |
|
$log->info("using $use_indexer indexing engine..."); |
| 137 |
|
} |
| 138 |
|
|
| 139 |
|
# disable indexing when creating marc |
| 140 |
|
$use_indexer = undef if ($marc_normalize); |
| 141 |
|
|
| 142 |
my $total_rows = 0; |
my $total_rows = 0; |
| 143 |
|
my $start_t = time(); |
| 144 |
|
|
| 145 |
|
my @links; |
| 146 |
|
my $indexer; |
| 147 |
|
|
| 148 |
|
my $lint = new MARC::Lint if ($marc_lint); |
| 149 |
|
|
| 150 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
| 151 |
|
|
| 152 |
my $log = _new WebPAC::Common()->_get_logger(); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
| 153 |
|
next if ($only_database && $database !~ m/$only_database/i); |
| 154 |
|
|
| 155 |
# |
if ($use_indexer) { |
| 156 |
# open Hyper Estraier database |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
| 157 |
# |
$indexer_config->{database} = $database; |
| 158 |
|
$indexer_config->{clean} = $clean; |
| 159 |
|
$indexer_config->{label} = $db_config->{name}; |
| 160 |
|
|
| 161 |
|
if ($use_indexer eq 'hyperestraier') { |
| 162 |
|
|
| 163 |
|
# open Hyper Estraier database |
| 164 |
|
use WebPAC::Output::Estraier '0.10'; |
| 165 |
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
| 166 |
|
|
| 167 |
|
} elsif ($use_indexer eq 'kinosearch') { |
| 168 |
|
|
| 169 |
|
# open KinoSearch |
| 170 |
|
use WebPAC::Output::KinoSearch; |
| 171 |
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
| 172 |
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
| 173 |
|
|
| 174 |
my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); |
} else { |
| 175 |
$est_config->{database} = $database; |
$log->logdie("unknown use_indexer: $use_indexer"); |
| 176 |
|
} |
| 177 |
|
|
| 178 |
$log->info("using HyperEstraier URL $est_config->{masterurl}"); |
$log->logide("can't continue without valid indexer") unless ($indexer); |
| 179 |
|
} |
| 180 |
|
|
|
my $est = new WebPAC::Output::Estraier( |
|
|
%{ $est_config }, |
|
|
); |
|
| 181 |
|
|
| 182 |
# |
# |
| 183 |
# now WebPAC::Store |
# now WebPAC::Store |
| 187 |
|
|
| 188 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
| 189 |
|
|
| 190 |
$log->info("working on $database in $db_path"); |
if ($clean) { |
| 191 |
|
$log->info("creating new database '$database' in $db_path"); |
| 192 |
|
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
| 193 |
|
} else { |
| 194 |
|
$log->info("working on database '$database' in $db_path"); |
| 195 |
|
} |
| 196 |
|
|
| 197 |
my $db = new WebPAC::Store( |
my $db = new WebPAC::Store( |
| 198 |
path => $db_path, |
path => $db_path, |
| 199 |
database => $database, |
database => $database, |
| 200 |
debug => 1, |
debug => $debug, |
| 201 |
); |
); |
| 202 |
|
|
| 203 |
|
|
| 218 |
|
|
| 219 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
| 220 |
|
|
| 221 |
|
next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
| 222 |
|
|
| 223 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
| 224 |
|
|
| 225 |
die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); |
die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); |
| 226 |
|
|
| 227 |
my $lookup = new WebPAC::Lookup( |
my $lookup = new WebPAC::Lookup( |
| 228 |
lookup_file => $input->{lookup}, |
lookup_file => $input->{lookup}, |
| 229 |
); |
) if ($input->{lookup}); |
| 230 |
|
|
| 231 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
| 232 |
|
|
| 233 |
$log->info("working on input $input->{path} [$input->{type}] using $input_module"); |
$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module", |
| 234 |
|
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
| 235 |
|
); |
| 236 |
|
|
| 237 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
| 238 |
module => $input_module, |
module => $input_module, |
| 239 |
code_page => $config->{webpac}->{webpac_encoding}, |
code_page => $config->{webpac}->{webpac_encoding}, |
| 240 |
limit_mfn => $input->{limit}, |
limit => $limit || $input->{limit}, |
| 241 |
|
offset => $offset, |
| 242 |
lookup => $lookup, |
lookup => $lookup, |
| 243 |
|
recode => $input->{recode}, |
| 244 |
|
stats => $stats, |
| 245 |
); |
); |
| 246 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
| 247 |
|
|
| 248 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
| 249 |
path => $input->{path}, |
path => $input->{path}, |
| 250 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
| 251 |
|
%{ $input }, |
| 252 |
); |
); |
| 253 |
|
|
| 254 |
my $n = new WebPAC::Normalize::XML( |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
| 255 |
# filter => { 'foo' => sub { shift } }, |
@{ $input->{normalize} } : ( $input->{normalize} ); |
|
db => $db, |
|
|
lookup_regex => $lookup->regex, |
|
|
lookup => $lookup, |
|
|
prefix => $input->{name}, |
|
|
); |
|
|
|
|
|
my $normalize_path = $input->{normalize}->{path}; |
|
| 256 |
|
|
| 257 |
if ($normalize_path =~ m/\.xml$/i) { |
if ($marc_normalize) { |
| 258 |
$n->open( |
@norm_array = ( { |
| 259 |
tag => $input->{normalize}->{tag}, |
path => $marc_normalize, |
| 260 |
xml_file => $input->{normalize}->{path}, |
output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc', |
| 261 |
); |
} ); |
|
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
|
|
$n->open_yaml( |
|
|
path => $normalize_path, |
|
|
tag => $input->{normalize}->{tag}, |
|
|
); |
|
| 262 |
} |
} |
| 263 |
|
|
| 264 |
for ( 0 ... $input_db->size ) { |
foreach my $normalize (@norm_array) { |
| 265 |
|
|
| 266 |
my $row = $input_db->fetch || next; |
my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); |
| 267 |
|
|
| 268 |
my $mfn = $row->{'000'}->[0] || die "can't find MFN"; |
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
| 269 |
|
|
| 270 |
my $ds = $n->data_structure($row); |
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
| 271 |
|
|
| 272 |
$est->add( |
$log->info("Using $normalize_path for normalization..."); |
| 273 |
id => $input->{name} . "#" . $mfn, |
|
| 274 |
ds => $ds, |
my $marc_fh; |
| 275 |
type => $config->{hyperestraier}->{type}, |
if (my $path = $normalize->{output}) { |
| 276 |
); |
open($marc_fh, '>', $path) || |
| 277 |
|
$log->logdie("can't open MARC output $path: $!"); |
| 278 |
|
|
| 279 |
|
$log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\n"); |
| 280 |
|
} |
| 281 |
|
|
| 282 |
|
# reset position in database |
| 283 |
|
$input_db->seek(1); |
| 284 |
|
|
| 285 |
|
foreach my $pos ( 0 ... $input_db->size ) { |
| 286 |
|
|
| 287 |
|
my $row = $input_db->fetch || next; |
| 288 |
|
|
| 289 |
|
my $mfn = $row->{'000'}->[0]; |
| 290 |
|
|
| 291 |
|
if (! $mfn || $mfn !~ m#^\d+$#) { |
| 292 |
|
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
| 293 |
|
$mfn = $pos; |
| 294 |
|
push @{ $row->{'000'} }, $pos; |
| 295 |
|
} |
| 296 |
|
|
| 297 |
|
|
| 298 |
|
if ($validate) { |
| 299 |
|
my @errors = $validate->validate_errors( $row ); |
| 300 |
|
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
| 301 |
|
} |
| 302 |
|
|
| 303 |
|
my $ds = WebPAC::Normalize::data_structure( |
| 304 |
|
row => $row, |
| 305 |
|
rules => $rules, |
| 306 |
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
| 307 |
|
marc_encoding => 'utf-8', |
| 308 |
|
); |
| 309 |
|
|
| 310 |
|
$db->save_ds( |
| 311 |
|
id => $mfn, |
| 312 |
|
ds => $ds, |
| 313 |
|
prefix => $input->{name}, |
| 314 |
|
) if ($ds && !$stats); |
| 315 |
|
|
| 316 |
|
$indexer->add( |
| 317 |
|
id => $input->{name} . "/" . $mfn, |
| 318 |
|
ds => $ds, |
| 319 |
|
type => $config->{$use_indexer}->{type}, |
| 320 |
|
) if ($indexer && $ds); |
| 321 |
|
|
| 322 |
|
if ($marc_fh) { |
| 323 |
|
my $marc = new MARC::Record; |
| 324 |
|
$marc->encoding( 'utf-8' ); |
| 325 |
|
my @marc_fields = WebPAC::Normalize::_get_marc_fields(); |
| 326 |
|
if (! @marc_fields) { |
| 327 |
|
$log->warn("MARC record $mfn is empty, skipping"); |
| 328 |
|
} else { |
| 329 |
|
$marc->add_fields( @marc_fields ); |
| 330 |
|
|
| 331 |
|
# tweak leader |
| 332 |
|
if (my $new_l = WebPAC::Normalize::marc_leader()) { |
| 333 |
|
|
| 334 |
|
my $leader = $marc->leader; |
| 335 |
|
|
| 336 |
|
foreach my $o ( keys %$new_l ) { |
| 337 |
|
my $insert = $new_l->{$o}; |
| 338 |
|
$leader = substr($leader, 0, $o) . |
| 339 |
|
$insert . substr($leader, $o+length($insert)); |
| 340 |
|
} |
| 341 |
|
$marc->leader( $leader ); |
| 342 |
|
} |
| 343 |
|
|
| 344 |
|
if ($marc_lint) { |
| 345 |
|
$lint->check_record( $marc ); |
| 346 |
|
my $err = join( "\n", $lint->warnings ); |
| 347 |
|
$log->error("MARC lint detected warning on MFN $mfn\n", |
| 348 |
|
"Original imput row: ",dump($row), "\n", |
| 349 |
|
"Normalized MARC row: ",dump(@marc_fields), "\n", |
| 350 |
|
"MARC lint warnings: ",$err,"\n" |
| 351 |
|
) if ($err); |
| 352 |
|
} |
| 353 |
|
|
| 354 |
|
if ($marc_dump) { |
| 355 |
|
$log->info("MARC record on MFN $mfn\tleader:", $marc->leader(), "\n", |
| 356 |
|
"Original imput row: ",dump($row), "\n", |
| 357 |
|
"Normalized MARC row: ",dump(@marc_fields), "\n", |
| 358 |
|
); |
| 359 |
|
} |
| 360 |
|
|
| 361 |
|
print $marc_fh $marc->as_usmarc; |
| 362 |
|
} |
| 363 |
|
} |
| 364 |
|
|
| 365 |
|
$total_rows++; |
| 366 |
|
} |
| 367 |
|
|
| 368 |
|
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
| 369 |
|
|
| 370 |
|
# close MARC file |
| 371 |
|
close($marc_fh) if ($marc_fh); |
| 372 |
|
|
|
$total_rows++; |
|
| 373 |
} |
} |
| 374 |
|
|
| 375 |
}; |
} |
| 376 |
|
|
| 377 |
$log->info("$total_rows records indexed"); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
| 378 |
|
|
| 379 |
|
my $dt = time() - $start_t; |
| 380 |
|
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
| 381 |
|
sprintf("in %.2f sec [%.2f rec/sec]", |
| 382 |
|
$dt, ($total_rows / $dt) |
| 383 |
|
) |
| 384 |
|
); |
| 385 |
|
|
| 386 |
# |
# |
| 387 |
# add Hyper Estraier links to other databases |
# add Hyper Estraier links to other databases |
| 388 |
# |
# |
| 389 |
if (ref($db_config->{links}) eq 'ARRAY') { |
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
| 390 |
foreach my $link (@{ $db_config->{links} }) { |
foreach my $link (@{ $db_config->{links} }) { |
| 391 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
if ($use_indexer eq 'hyperestraier') { |
| 392 |
$est->add_link( |
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
| 393 |
from => $database, |
push @links, { |
| 394 |
to => $link->{to}, |
from => $database, |
| 395 |
credit => $link->{credit}, |
to => $link->{to}, |
| 396 |
); |
credit => $link->{credit}, |
| 397 |
|
}; |
| 398 |
|
} else { |
| 399 |
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
| 400 |
|
} |
| 401 |
} |
} |
| 402 |
} |
} |
| 403 |
|
|
| 404 |
} |
} |
| 405 |
|
|
| 406 |
|
foreach my $link (@links) { |
| 407 |
|
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
| 408 |
|
$indexer->add_link( %{ $link } ); |
| 409 |
|
} |
Parent Directory
| 
Revision Log
| 
Patch