7 |
use lib './lib'; |
use lib './lib'; |
8 |
|
|
9 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
10 |
use WebPAC::Parser 0.04; |
use WebPAC::Parser 0.08; |
11 |
use WebPAC::Input 0.11; |
use WebPAC::Input 0.16; |
12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.14; |
13 |
use WebPAC::Normalize 0.11; |
use WebPAC::Normalize 0.22; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
15 |
use WebPAC::Validate 0.06; |
use WebPAC::Validate 0.06; |
16 |
use WebPAC::Output::MARC; |
use WebPAC::Output::MARC; |
67 |
|
|
68 |
turn on extra validation of imput records, see L<WebPAC::Validation> |
turn on extra validation of imput records, see L<WebPAC::Validation> |
69 |
|
|
|
=item --marc-normalize conf/normalize/mapping.pl |
|
|
|
|
|
This option specifies normalisation file for MARC creation |
|
|
|
|
|
=item --marc-output out/marc/test.marc |
|
|
|
|
|
Optional path to output file |
|
|
|
|
70 |
=item --marc-lint |
=item --marc-lint |
71 |
|
|
72 |
By default turned on if C<--marc-normalize> is used. You can disable lint |
By default turned on if normalisation file has C<marc*> directives. You can disable lint |
73 |
messages with C<--no-marc-lint>. |
messages with C<--no-marc-lint>. |
74 |
|
|
75 |
=item --marc-dump |
=item --marc-dump |
102 |
my $only_filter; |
my $only_filter; |
103 |
my $stats = 0; |
my $stats = 0; |
104 |
my $validate_path; |
my $validate_path; |
|
my ($marc_normalize, $marc_output); |
|
105 |
my $marc_lint = 1; |
my $marc_lint = 1; |
106 |
my $marc_dump = 0; |
my $marc_dump = 0; |
107 |
my $parallel = 0; |
my $parallel = 0; |
120 |
"debug+" => \$debug, |
"debug+" => \$debug, |
121 |
"stats" => \$stats, |
"stats" => \$stats, |
122 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
|
"marc-normalize=s" => \$marc_normalize, |
|
|
"marc-output=s" => \$marc_output, |
|
123 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
124 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
125 |
"parallel=i" => \$parallel, |
"parallel=i" => \$parallel, |
135 |
|
|
136 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
137 |
|
|
138 |
|
my $log_file = 'log'; |
139 |
|
|
140 |
|
if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) { |
141 |
|
$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'"); |
142 |
|
rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!"); |
143 |
|
} |
144 |
|
|
145 |
my $estcmd_fh; |
my $estcmd_fh; |
146 |
my $estcmd_path = './estcmd-merge.sh'; |
my $estcmd_path = './estcmd-merge.sh'; |
159 |
|
|
160 |
|
|
161 |
my $use_indexer = $config->use_indexer; |
my $use_indexer = $config->use_indexer; |
162 |
|
$stats ||= $validate; |
163 |
if ($stats) { |
if ($stats) { |
164 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("disabled indexing for stats collection"); |
165 |
$use_indexer = undef; |
$use_indexer = undef; |
166 |
} else { |
} else { |
167 |
$log->info("using $use_indexer indexing engine..."); |
$log->info("using $use_indexer indexing engine..."); |
168 |
} |
} |
169 |
|
|
|
# disable indexing when creating marc |
|
|
$use_indexer = undef if ($marc_normalize); |
|
|
|
|
170 |
# parse normalize files and create source files for lookup and normalization |
# parse normalize files and create source files for lookup and normalization |
171 |
|
|
172 |
my $parser = new WebPAC::Parser( config => $config ); |
my $parser = new WebPAC::Parser( config => $config ); |
181 |
Proc::Queue::size($parallel); |
Proc::Queue::size($parallel); |
182 |
} |
} |
183 |
|
|
184 |
|
sub create_ds_config { |
185 |
|
my ($db_config, $database, $input, $mfn) = @_; |
186 |
|
my $c = dclone( $db_config ); |
187 |
|
$c->{_} = $database || $log->logconfess("need database"); |
188 |
|
$c->{_mfn} = $mfn || $log->logconfess("need mfn"); |
189 |
|
$c->{input} = $input || $log->logconfess("need input"); |
190 |
|
return $c; |
191 |
|
} |
192 |
|
|
193 |
while (my ($database, $db_config) = each %{ $config->databases }) { |
while (my ($database, $db_config) = each %{ $config->databases }) { |
194 |
|
|
195 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
205 |
} |
} |
206 |
|
|
207 |
my $indexer; |
my $indexer; |
208 |
if ($use_indexer) { |
if ($use_indexer && $parser->have_rules( 'search', $database )) { |
209 |
|
|
210 |
my $cfg_name = $use_indexer; |
my $cfg_name = $use_indexer; |
211 |
$cfg_name =~ s/\-.*$//; |
$cfg_name =~ s/\-.*$//; |
278 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
279 |
$abs_path =~ s#/[^/]*$#/#; |
$abs_path =~ s#/[^/]*$#/#; |
280 |
|
|
281 |
my $db_path = $config->get('webpac')->{db_path} . '/' . $database; |
my $db_path = $config->webpac('db_path'); |
282 |
|
|
283 |
if ($clean) { |
if ($clean) { |
284 |
$log->info("creating new database '$database' in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
287 |
$log->info("working on database '$database' in $db_path"); |
$log->info("working on database '$database' in $db_path"); |
288 |
} |
} |
289 |
|
|
290 |
my $db = new WebPAC::Store( |
my $store = new WebPAC::Store( |
291 |
path => $db_path, |
path => $db_path, |
|
database => $database, |
|
292 |
debug => $debug, |
debug => $debug, |
293 |
); |
); |
294 |
|
|
308 |
|
|
309 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
310 |
|
|
311 |
next if ($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)); |
312 |
|
|
313 |
|
next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
314 |
|
|
315 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
316 |
|
|
320 |
|
|
321 |
my @lookups = $parser->have_lookup_create($database, $input); |
my @lookups = $parser->have_lookup_create($database, $input); |
322 |
|
|
323 |
$log->info("working on input '$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 $input_module", |
324 |
@lookups ? "lookup ".join(", ", @lookups) : "" |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
325 |
); |
); |
326 |
|
|
|
warn "lookups = ", dump( @lookups ); |
|
|
|
|
327 |
if ($stats) { |
if ($stats) { |
328 |
# disable modification of records if --stats is in use |
# disable modification of records if --stats is in use |
329 |
delete($input->{modify_records}); |
delete($input->{modify_records}); |
330 |
delete($input->{modify_file}); |
delete($input->{modify_file}); |
331 |
} |
} |
332 |
|
|
|
warn "parser->depends = ", dump( $parser->{depends} ); |
|
|
warn "depends on: ", dump( $parser->depends($database, $input->{name}) ); |
|
|
warn "lookup_create_rules = ", dump( $parser->lookup_create_rules($database, $input->{name}) ); |
|
|
warn "parser->_lookup_create = ", dump( $parser->{_lookup_create} ); |
|
|
|
|
|
my $lookup; |
|
|
|
|
333 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
334 |
module => $input_module, |
module => $input_module, |
335 |
encoding => $config->webpac('webpac_encoding'), |
encoding => $config->webpac('webpac_encoding'), |
336 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
337 |
offset => $offset, |
offset => $offset, |
|
lookup_coderef => sub { |
|
|
my $rec = shift || return; |
|
|
$lookup->add( $rec ); |
|
|
}, |
|
338 |
recode => $input->{recode}, |
recode => $input->{recode}, |
339 |
stats => $stats, |
stats => $stats, |
340 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
343 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
344 |
|
|
345 |
if (defined( $input->{lookup} )) { |
if (defined( $input->{lookup} )) { |
346 |
$log->warn("$database/", $input->{name}, " has depriciated lookup definition, removing it..."); |
$log->warn("$database/$input_name has depriciated lookup definition, removing it..."); |
347 |
delete( $input->{lookup} ); |
delete( $input->{lookup} ); |
348 |
} |
} |
349 |
|
|
350 |
|
my $lookup_coderef; |
351 |
|
|
352 |
|
if (@lookups) { |
353 |
|
|
354 |
|
my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input"); |
355 |
|
|
356 |
|
$lookup_coderef = sub { |
357 |
|
my $rec = shift || die "need rec!"; |
358 |
|
my $mfn = $rec->{'000'}->[0] || die "need mfn in 000"; |
359 |
|
|
360 |
|
WebPAC::Normalize::data_structure( |
361 |
|
row => $rec, |
362 |
|
rules => $rules, |
363 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
364 |
|
); |
365 |
|
|
366 |
|
#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup()); |
367 |
|
}; |
368 |
|
|
369 |
|
WebPAC::Normalize::_set_lookup( undef ); |
370 |
|
|
371 |
|
$log->debug("created lookup_coderef using:\n$rules"); |
372 |
|
|
373 |
|
}; |
374 |
|
|
375 |
|
my $lookup_jar; |
376 |
|
|
377 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
378 |
path => $input->{path}, |
path => $input->{path}, |
379 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
380 |
|
lookup_coderef => $lookup_coderef, |
381 |
|
lookup => $lookup_jar, |
382 |
%{ $input }, |
%{ $input }, |
383 |
|
load_row => sub { |
384 |
|
my $a = shift; |
385 |
|
return $store->load_row( |
386 |
|
database => $database, |
387 |
|
input => $input_name, |
388 |
|
id => $a->{id}, |
389 |
|
); |
390 |
|
}, |
391 |
|
save_row => sub { |
392 |
|
my $a = shift; |
393 |
|
return $store->save_row( |
394 |
|
database => $database, |
395 |
|
input => $input_name, |
396 |
|
id => $a->{id}, |
397 |
|
row => $a->{row}, |
398 |
|
); |
399 |
|
}, |
400 |
|
|
401 |
); |
); |
402 |
|
|
403 |
|
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
404 |
|
|
405 |
|
if (defined( $lookup_data->{$database}->{$input_name} )) { |
406 |
|
$log->debug("created following lookups: ", sub { dump( $lookup_data ) } ); |
407 |
|
|
408 |
|
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
409 |
|
$store->save_lookup( |
410 |
|
database => $database, |
411 |
|
input => $input_name, |
412 |
|
key => $key, |
413 |
|
data => $lookup_data->{$database}->{$input_name}->{$key}, |
414 |
|
); |
415 |
|
} |
416 |
|
} |
417 |
|
|
418 |
my $report_fh; |
my $report_fh; |
419 |
if ($stats || $validate) { |
if ($stats || $validate) { |
420 |
my $path = "out/report/" . $database . '-' . $input->{name} . '.txt'; |
my $path = "out/report/${database}-${input_name}.txt"; |
421 |
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
422 |
|
|
423 |
print $report_fh "Report for database '$database' input '$input->{name}' records ", |
print $report_fh "Report for database '$database' input '$input_name' records ", |
424 |
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
425 |
$log->info("Generating report file $path"); |
$log->info("Generating report file $path"); |
426 |
} |
} |
427 |
|
|
428 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my $marc; |
429 |
@{ $input->{normalize} } : ( $input->{normalize} ); |
if ($parser->have_rules( 'marc', $database, $input_name )) { |
430 |
|
$marc = new WebPAC::Output::MARC( |
431 |
if ($marc_normalize) { |
path => "out/marc/${database}-${input_name}.marc", |
|
@norm_array = ( { |
|
|
path => $marc_normalize, |
|
|
output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc', |
|
|
} ); |
|
|
} |
|
|
|
|
|
foreach my $normalize (@norm_array) { |
|
|
|
|
|
my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); |
|
|
|
|
|
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
|
|
|
|
|
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
|
|
|
|
|
$log->info("Using $normalize_path for normalization..."); |
|
|
|
|
|
my $marc = new WebPAC::Output::MARC( |
|
|
path => $normalize->{output}, |
|
432 |
lint => $marc_lint, |
lint => $marc_lint, |
433 |
dump => $marc_dump, |
dump => $marc_dump, |
434 |
) if ($normalize->{output}); |
); |
435 |
|
} |
|
# reset position in database |
|
|
$input_db->seek(1); |
|
|
|
|
|
# generate name of config key for indexer (strip everything after -) |
|
|
my $indexer_config = $use_indexer; |
|
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
|
|
|
|
|
foreach my $pos ( 0 ... $input_db->size ) { |
|
|
|
|
|
my $row = $input_db->fetch || next; |
|
|
|
|
|
my $mfn = $row->{'000'}->[0]; |
|
436 |
|
|
437 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
438 |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
$log->debug("parsed normalize rules:\n$rules"); |
|
$mfn = $pos; |
|
|
push @{ $row->{'000'} }, $pos; |
|
|
} |
|
439 |
|
|
440 |
|
# reset position in database |
441 |
|
$input_db->seek(1); |
442 |
|
|
443 |
if ($validate) { |
# generate name of config key for indexer (strip everything after -) |
444 |
if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) { |
my $indexer_config = $use_indexer; |
445 |
$log->error( "MFN $mfn validation error:\n", |
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
446 |
$validate->report_error( $errors ) |
|
447 |
|
my $lookup_hash; |
448 |
|
my $depends = $parser->depends($database,$input_name); |
449 |
|
|
450 |
|
if ($depends) { |
451 |
|
$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends); |
452 |
|
$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH'); |
453 |
|
|
454 |
|
foreach my $db (keys %$depends) { |
455 |
|
foreach my $i (keys %{$depends->{$db}}) { |
456 |
|
foreach my $k (keys %{$depends->{$db}->{$i}}) { |
457 |
|
my $t = time(); |
458 |
|
$log->debug("loading lookup $db/$i"); |
459 |
|
$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup( |
460 |
|
database => $db, |
461 |
|
input => $i, |
462 |
|
key => $k, |
463 |
); |
); |
464 |
|
$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t)); |
465 |
} |
} |
466 |
} |
} |
467 |
|
} |
468 |
|
|
469 |
my $ds_config = dclone($db_config); |
$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) }); |
470 |
|
} |
471 |
|
|
|
# default values -> database key |
|
|
$ds_config->{_} = $database; |
|
472 |
|
|
473 |
# current mfn |
foreach my $pos ( 0 ... $input_db->size ) { |
|
$ds_config->{_mfn} = $mfn; |
|
474 |
|
|
475 |
# attach current input |
my $row = $input_db->fetch || next; |
|
$ds_config->{input} = $input; |
|
476 |
|
|
477 |
my $ds = WebPAC::Normalize::data_structure( |
$total_rows++; |
|
row => $row, |
|
|
rules => $rules, |
|
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
|
|
config => $ds_config, |
|
|
marc_encoding => 'utf-8', |
|
|
); |
|
478 |
|
|
479 |
$db->save_ds( |
my $mfn = $row->{'000'}->[0]; |
|
id => $mfn, |
|
|
ds => $ds, |
|
|
prefix => $input->{name}, |
|
|
) if ($ds && !$stats); |
|
|
|
|
|
$indexer->add( |
|
|
id => $input->{name} . "/" . $mfn, |
|
|
ds => $ds, |
|
|
type => $config->get($indexer_config)->{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++; |
|
|
} |
|
|
|
|
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
|
|
} |
|
480 |
|
|
481 |
$total_rows++; |
if (! $mfn || $mfn !~ m#^\d+$#) { |
482 |
|
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
483 |
|
$mfn = $pos; |
484 |
|
push @{ $row->{'000'} }, $pos; |
485 |
} |
} |
486 |
|
|
487 |
|
|
488 |
if ($validate) { |
if ($validate) { |
489 |
my $errors = $validate->report; |
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
490 |
if ($errors) { |
$log->error( "MFN $mfn validation error:\n", |
491 |
$log->info("validation errors:\n$errors\n" ); |
$validate->report_error( $errors ) |
492 |
print $report_fh "$errors\n" if ($report_fh); |
); |
493 |
} |
} |
494 |
|
next; # validation doesn't create any output |
495 |
} |
} |
496 |
|
|
497 |
if ($stats) { |
my $ds = WebPAC::Normalize::data_structure( |
498 |
my $s = $input_db->stats; |
row => $row, |
499 |
$log->info("statistics of fields usage:\n$s"); |
rules => $rules, |
500 |
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
lookup => $lookup_hash, |
501 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
502 |
|
marc_encoding => 'utf-8', |
503 |
|
load_row_coderef => sub { |
504 |
|
my ($database,$input,$mfn) = @_; |
505 |
|
return $store->load_row( |
506 |
|
database => $database, |
507 |
|
input => $input, |
508 |
|
id => $mfn, |
509 |
|
); |
510 |
|
}, |
511 |
|
); |
512 |
|
|
513 |
|
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
514 |
|
|
515 |
|
$store->save_ds( |
516 |
|
database => $database, |
517 |
|
input => $input_name, |
518 |
|
id => $mfn, |
519 |
|
ds => $ds, |
520 |
|
) if ($ds && !$stats); |
521 |
|
|
522 |
|
$indexer->add( |
523 |
|
id => "${input_name}/${mfn}", |
524 |
|
ds => $ds, |
525 |
|
type => $config->get($indexer_config)->{type}, |
526 |
|
) if ($indexer && $ds); |
527 |
|
|
528 |
|
if ($marc) { |
529 |
|
my $i = 0; |
530 |
|
|
531 |
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
532 |
|
$marc->add( |
533 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
534 |
|
fields => $fields, |
535 |
|
leader => WebPAC::Normalize::marc_leader(), |
536 |
|
row => $row, |
537 |
|
); |
538 |
|
$i++; |
539 |
|
} |
540 |
|
|
541 |
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
542 |
} |
} |
543 |
|
} |
544 |
|
|
545 |
# close MARC file |
if ($validate) { |
546 |
$marc->finish if ($marc); |
my $errors = $validate->report; |
547 |
|
if ($errors) { |
548 |
|
$log->info("validation errors:\n$errors\n" ); |
549 |
|
print $report_fh "$errors\n" if ($report_fh); |
550 |
|
} |
551 |
|
} |
552 |
|
|
553 |
# close report |
if ($stats) { |
554 |
close($report_fh) if ($report_fh) |
my $s = $input_db->stats; |
555 |
|
$log->info("statistics of fields usage:\n$s"); |
556 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
557 |
} |
} |
558 |
|
|
559 |
|
# close MARC file |
560 |
|
$marc->finish if ($marc); |
561 |
|
|
562 |
|
# close report |
563 |
|
close($report_fh) if ($report_fh) |
564 |
|
|
565 |
} |
} |
566 |
|
|
567 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |