7 |
use lib './lib'; |
use lib './lib'; |
8 |
|
|
9 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
10 |
use WebPAC::Lookup 0.03; |
use WebPAC::Parser 0.04; |
11 |
use WebPAC::Input 0.11; |
use WebPAC::Input 0.13; |
12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.11; |
13 |
use WebPAC::Normalize 0.11; |
use WebPAC::Normalize 0.21; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
15 |
use WebPAC::Validate; |
use WebPAC::Validate 0.06; |
16 |
use WebPAC::Output::MARC; |
use WebPAC::Output::MARC; |
17 |
use YAML qw/LoadFile/; |
use WebPAC::Config; |
18 |
use Getopt::Long; |
use Getopt::Long; |
19 |
use File::Path; |
use File::Path; |
20 |
use Time::HiRes qw/time/; |
use Time::HiRes qw/time/; |
105 |
my $limit; |
my $limit; |
106 |
|
|
107 |
my $clean = 0; |
my $clean = 0; |
108 |
my $config = 'conf/config.yml'; |
my $config_path; |
109 |
my $debug = 0; |
my $debug = 0; |
110 |
my $only_filter; |
my $only_filter; |
111 |
my $stats = 0; |
my $stats = 0; |
119 |
|
|
120 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
121 |
|
|
|
my $hostname = `hostname`; |
|
|
chomp($hostname); |
|
|
$hostname =~ s/\..+$//; |
|
|
if (-e "conf/$hostname.yml") { |
|
|
$config = "conf/$hostname.yml"; |
|
|
$log->info("using host configuration file: $config"); |
|
|
} |
|
|
|
|
122 |
GetOptions( |
GetOptions( |
123 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
124 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
125 |
"clean" => \$clean, |
"clean" => \$clean, |
126 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
127 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
128 |
"config" => \$config, |
"config" => \$config_path, |
129 |
"debug+" => \$debug, |
"debug+" => \$debug, |
130 |
"stats" => \$stats, |
"stats" => \$stats, |
131 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
138 |
"merge" => \$merge, |
"merge" => \$merge, |
139 |
); |
); |
140 |
|
|
141 |
$config = LoadFile($config); |
my $config = new WebPAC::Config( path => $config_path ); |
142 |
|
|
143 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
144 |
|
|
145 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->databases); |
146 |
|
|
147 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
148 |
|
|
163 |
) if ($validate_path); |
) if ($validate_path); |
164 |
|
|
165 |
|
|
166 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->use_indexer; |
167 |
if ($stats) { |
if ($stats) { |
168 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("option --stats disables update of indexing engine..."); |
169 |
$use_indexer = undef; |
$use_indexer = undef; |
174 |
# disable indexing when creating marc |
# disable indexing when creating marc |
175 |
$use_indexer = undef if ($marc_normalize); |
$use_indexer = undef if ($marc_normalize); |
176 |
|
|
177 |
|
# parse normalize files and create source files for lookup and normalization |
178 |
|
|
179 |
|
my $parser = new WebPAC::Parser( config => $config ); |
180 |
|
|
181 |
my $total_rows = 0; |
my $total_rows = 0; |
182 |
my $start_t = time(); |
my $start_t = time(); |
183 |
|
|
188 |
Proc::Queue::size($parallel); |
Proc::Queue::size($parallel); |
189 |
} |
} |
190 |
|
|
191 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
sub create_ds_config { |
192 |
|
my ($db_config, $database, $input, $mfn) = @_; |
193 |
|
my $c = dclone( $db_config ); |
194 |
|
$c->{_} = $database || $log->logconfess("need database"); |
195 |
|
$c->{_mfn} = $mfn || $log->logconfess("need mfn"); |
196 |
|
$c->{input} = $input || $log->logconfess("need input"); |
197 |
|
return $c; |
198 |
|
} |
199 |
|
|
200 |
|
while (my ($database, $db_config) = each %{ $config->databases }) { |
201 |
|
|
202 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
203 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
217 |
my $cfg_name = $use_indexer; |
my $cfg_name = $use_indexer; |
218 |
$cfg_name =~ s/\-.*$//; |
$cfg_name =~ s/\-.*$//; |
219 |
|
|
220 |
my $indexer_config = $config->{$cfg_name} || $log->logdie("can't find '$cfg_name' part in confguration"); |
my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration"); |
221 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
222 |
$indexer_config->{clean} = $clean; |
$indexer_config->{clean} = $clean; |
223 |
$indexer_config->{label} = $db_config->{name}; |
$indexer_config->{label} = $db_config->{name}; |
237 |
use WebPAC::Output::EstraierNative; |
use WebPAC::Output::EstraierNative; |
238 |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
239 |
|
|
|
$use_indexer = 'hyperestraier'; |
|
|
|
|
240 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
241 |
|
|
242 |
# open KinoSearch |
# open KinoSearch |
285 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
286 |
$abs_path =~ s#/[^/]*$#/#; |
$abs_path =~ s#/[^/]*$#/#; |
287 |
|
|
288 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->webpac('db_path'); |
289 |
|
|
290 |
if ($clean) { |
if ($clean) { |
291 |
$log->info("creating new database '$database' in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
294 |
$log->info("working on database '$database' in $db_path"); |
$log->info("working on database '$database' in $db_path"); |
295 |
} |
} |
296 |
|
|
297 |
my $db = new WebPAC::Store( |
my $store = new WebPAC::Store( |
298 |
path => $db_path, |
path => $db_path, |
|
database => $database, |
|
299 |
debug => $debug, |
debug => $debug, |
300 |
); |
); |
301 |
|
|
313 |
$log->info("database $database doesn't have inputs defined"); |
$log->info("database $database doesn't have inputs defined"); |
314 |
} |
} |
315 |
|
|
|
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
|
|
|
|
316 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
317 |
|
|
318 |
next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input)); |
319 |
|
|
320 |
|
next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
321 |
|
|
322 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
323 |
|
|
324 |
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(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs'))); |
325 |
|
|
326 |
my $lookup; |
my $input_module = $config->webpac('inputs')->{$type}; |
|
if ($input->{lookup}) { |
|
|
$lookup = new WebPAC::Lookup( |
|
|
lookup_file => $input->{lookup}, |
|
|
); |
|
|
delete( $input->{lookup} ); |
|
|
} |
|
327 |
|
|
328 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my @lookups = $parser->have_lookup_create($database, $input); |
329 |
|
|
330 |
$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module", |
$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module", |
331 |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
332 |
); |
); |
333 |
|
|
334 |
if ($stats) { |
if ($stats) { |
339 |
|
|
340 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
341 |
module => $input_module, |
module => $input_module, |
342 |
encoding => $config->{webpac}->{webpac_encoding}, |
encoding => $config->webpac('webpac_encoding'), |
343 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
344 |
offset => $offset, |
offset => $offset, |
|
lookup_coderef => sub { |
|
|
my $rec = shift || return; |
|
|
$lookup->add( $rec ); |
|
|
}, |
|
345 |
recode => $input->{recode}, |
recode => $input->{recode}, |
346 |
stats => $stats, |
stats => $stats, |
347 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
349 |
); |
); |
350 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
351 |
|
|
352 |
|
if (defined( $input->{lookup} )) { |
353 |
|
$log->warn("$database/$input_name has depriciated lookup definition, removing it..."); |
354 |
|
delete( $input->{lookup} ); |
355 |
|
} |
356 |
|
|
357 |
|
my $lookup_coderef; |
358 |
|
|
359 |
|
if (@lookups) { |
360 |
|
|
361 |
|
my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input"); |
362 |
|
|
363 |
|
$lookup_coderef = sub { |
364 |
|
my $rec = shift || die "need rec!"; |
365 |
|
my $mfn = $rec->{'000'}->[0] || die "need mfn in 000"; |
366 |
|
|
367 |
|
WebPAC::Normalize::data_structure( |
368 |
|
row => $rec, |
369 |
|
rules => $rules, |
370 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
371 |
|
); |
372 |
|
|
373 |
|
#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup()); |
374 |
|
}; |
375 |
|
|
376 |
|
WebPAC::Normalize::_set_lookup( undef ); |
377 |
|
|
378 |
|
$log->debug("created lookup_coderef using:\n$rules"); |
379 |
|
|
380 |
|
}; |
381 |
|
|
382 |
|
my $lookup_jar; |
383 |
|
|
384 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
385 |
path => $input->{path}, |
path => $input->{path}, |
386 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
387 |
|
lookup_coderef => $lookup_coderef, |
388 |
|
lookup => $lookup_jar, |
389 |
%{ $input }, |
%{ $input }, |
390 |
); |
); |
391 |
|
|
392 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
|
@{ $input->{normalize} } : ( $input->{normalize} ); |
|
|
|
|
|
if ($marc_normalize) { |
|
|
@norm_array = ( { |
|
|
path => $marc_normalize, |
|
|
output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc', |
|
|
} ); |
|
|
} |
|
|
|
|
|
foreach my $normalize (@norm_array) { |
|
393 |
|
|
394 |
my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
395 |
|
$log->debug("created following lookups: ", dump( $lookup_data )); |
396 |
|
|
397 |
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
398 |
|
$store->save_lookup( |
399 |
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
database => $database, |
400 |
|
input => $input_name, |
401 |
|
key => $key, |
402 |
|
data => $lookup_data->{$database}->{$input_name}->{$key}, |
403 |
|
); |
404 |
|
} |
405 |
|
} |
406 |
|
|
407 |
$log->info("Using $normalize_path for normalization..."); |
my $report_fh; |
408 |
|
if ($stats || $validate) { |
409 |
|
my $path = "out/report/${database}-${input_name}.txt"; |
410 |
|
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
411 |
|
|
412 |
|
print $report_fh "Report for database '$database' input '$input_name' records ", |
413 |
|
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
414 |
|
$log->info("Generating report file $path"); |
415 |
|
} |
416 |
|
|
417 |
my $marc = new WebPAC::Output::MARC( |
my $marc; |
418 |
path => $normalize->{output}, |
if ($marc_normalize) { |
419 |
|
$marc = new WebPAC::Output::MARC( |
420 |
|
path => $marc_output || "out/marc/${database}-${input_name}.marc", |
421 |
lint => $marc_lint, |
lint => $marc_lint, |
422 |
dump => $marc_dump, |
dump => $marc_dump, |
423 |
) if ($normalize->{output}); |
); |
424 |
|
} |
|
# reset position in database |
|
|
$input_db->seek(1); |
|
|
|
|
|
foreach my $pos ( 0 ... $input_db->size ) { |
|
425 |
|
|
426 |
my $row = $input_db->fetch || next; |
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
427 |
|
$log->debug("parsed normalize rules:\n$rules"); |
428 |
|
|
429 |
my $mfn = $row->{'000'}->[0]; |
# reset position in database |
430 |
|
$input_db->seek(1); |
431 |
|
|
432 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
# generate name of config key for indexer (strip everything after -) |
433 |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
my $indexer_config = $use_indexer; |
434 |
$mfn = $pos; |
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
435 |
push @{ $row->{'000'} }, $pos; |
|
436 |
|
my $lookup_hash; |
437 |
|
my $depends = $parser->depends($database,$input_name); |
438 |
|
|
439 |
|
if ($depends) { |
440 |
|
$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends); |
441 |
|
$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH'); |
442 |
|
|
443 |
|
foreach my $db (keys %$depends) { |
444 |
|
foreach my $i (keys %{$depends->{$db}}) { |
445 |
|
foreach my $k (keys %{$depends->{$db}->{$i}}) { |
446 |
|
$log->debug("loading lookup $db/$i"); |
447 |
|
$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup( |
448 |
|
database => $db, |
449 |
|
input => $i, |
450 |
|
key => $k, |
451 |
|
); |
452 |
|
} |
453 |
} |
} |
454 |
|
} |
455 |
|
|
456 |
|
$log->debug("lookup_hash = ", dump( $lookup_hash )); |
457 |
|
} |
458 |
|
|
|
if ($validate) { |
|
|
my @errors = $validate->validate_errors( $row ); |
|
|
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
|
|
} |
|
459 |
|
|
460 |
my $ds_config = dclone($db_config); |
foreach my $pos ( 0 ... $input_db->size ) { |
461 |
|
|
462 |
# default values -> database key |
my $row = $input_db->fetch || next; |
|
$ds_config->{_} = $database; |
|
463 |
|
|
464 |
# current mfn |
my $mfn = $row->{'000'}->[0]; |
|
$ds_config->{_mfn} = $mfn; |
|
465 |
|
|
466 |
# attach current input |
if (! $mfn || $mfn !~ m#^\d+$#) { |
467 |
$ds_config->{input} = $input; |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
468 |
|
$mfn = $pos; |
469 |
|
push @{ $row->{'000'} }, $pos; |
470 |
|
} |
471 |
|
|
|
my $ds = WebPAC::Normalize::data_structure( |
|
|
row => $row, |
|
|
rules => $rules, |
|
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
|
|
config => $ds_config, |
|
|
marc_encoding => 'utf-8', |
|
|
); |
|
472 |
|
|
473 |
$db->save_ds( |
if ($validate) { |
474 |
id => $mfn, |
if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) { |
475 |
ds => $ds, |
$log->error( "MFN $mfn validation error:\n", |
476 |
prefix => $input->{name}, |
$validate->report_error( $errors ) |
477 |
) if ($ds && !$stats); |
); |
478 |
|
} |
479 |
$indexer->add( |
} |
|
id => $input->{name} . "/" . $mfn, |
|
|
ds => $ds, |
|
|
type => $config->{$use_indexer}->{type}, |
|
|
) if ($indexer && $ds); |
|
|
|
|
|
if ($marc) { |
|
|
my $i = 0; |
|
|
|
|
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
|
|
$marc->add( |
|
|
id => $mfn . ( $i ? "/$i" : '' ), |
|
|
fields => $fields, |
|
|
leader => WebPAC::Normalize::marc_leader(), |
|
|
row => $row, |
|
|
); |
|
|
$i++; |
|
|
} |
|
480 |
|
|
481 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
my $ds = WebPAC::Normalize::data_structure( |
482 |
|
row => $row, |
483 |
|
rules => $rules, |
484 |
|
lookup => $lookup_hash, |
485 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
486 |
|
marc_encoding => 'utf-8', |
487 |
|
load_ds_coderef => sub { |
488 |
|
my ($database,$input,$mfn) = @_; |
489 |
|
return $store->load_ds( |
490 |
|
database => $database, |
491 |
|
input => $input, |
492 |
|
id => $mfn, |
493 |
|
); |
494 |
|
}, |
495 |
|
); |
496 |
|
|
497 |
|
$store->save_ds( |
498 |
|
database => $database, |
499 |
|
input => $input_name, |
500 |
|
id => $mfn, |
501 |
|
ds => $ds, |
502 |
|
) if ($ds && !$stats); |
503 |
|
|
504 |
|
$indexer->add( |
505 |
|
id => "${input_name}/${mfn}", |
506 |
|
ds => $ds, |
507 |
|
type => $config->get($indexer_config)->{type}, |
508 |
|
) if ($indexer && $ds); |
509 |
|
|
510 |
|
if ($marc) { |
511 |
|
my $i = 0; |
512 |
|
|
513 |
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
514 |
|
$marc->add( |
515 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
516 |
|
fields => $fields, |
517 |
|
leader => WebPAC::Normalize::marc_leader(), |
518 |
|
row => $row, |
519 |
|
); |
520 |
|
$i++; |
521 |
} |
} |
522 |
|
|
523 |
$total_rows++; |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
524 |
} |
} |
525 |
|
|
526 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
$total_rows++; |
527 |
|
} |
528 |
|
|
529 |
# close MARC file |
if ($validate) { |
530 |
$marc->finish if ($marc); |
my $errors = $validate->report; |
531 |
|
if ($errors) { |
532 |
|
$log->info("validation errors:\n$errors\n" ); |
533 |
|
print $report_fh "$errors\n" if ($report_fh); |
534 |
|
} |
535 |
|
} |
536 |
|
|
537 |
|
if ($stats) { |
538 |
|
my $s = $input_db->stats; |
539 |
|
$log->info("statistics of fields usage:\n$s"); |
540 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
541 |
} |
} |
542 |
|
|
543 |
|
# close MARC file |
544 |
|
$marc->finish if ($marc); |
545 |
|
|
546 |
|
# close report |
547 |
|
close($report_fh) if ($report_fh) |
548 |
|
|
549 |
} |
} |
550 |
|
|
551 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |