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 0.13; |
use WebPAC::Input 0.16; |
12 |
use WebPAC::Store 0.10; |
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; |
21 |
use File::Slurp; |
use File::Slurp; |
22 |
use Data::Dump qw/dump/; |
use Data::Dump qw/dump/; |
23 |
use Storable qw/dclone/; |
use Storable qw/dclone/; |
24 |
|
use Pod::Usage qw/pod2usage/; |
25 |
|
|
26 |
use Proc::Queue size => 1; |
use Proc::Queue size => 1; |
27 |
use POSIX ":sys_wait_h"; # imports WNOHANG |
use POSIX ":sys_wait_h"; # imports WNOHANG |
32 |
|
|
33 |
B<this command will probably go away. Don't get used to it!> |
B<this command will probably go away. Don't get used to it!> |
34 |
|
|
35 |
Options: |
=head1 OPTIONS |
36 |
|
|
37 |
=over 4 |
=over 4 |
38 |
|
|
68 |
|
|
69 |
turn on extra validation of imput records, see L<WebPAC::Validation> |
turn on extra validation of imput records, see L<WebPAC::Validation> |
70 |
|
|
71 |
=item --marc-normalize conf/normalize/mapping.pl |
=item --marc-generate |
72 |
|
|
73 |
This option specifies normalisation file for MARC creation |
Generate MARC file. This will automatically be on if file contains C<marc*> directives. |
74 |
|
You can use this option as C<--no-marc-generate> to disable MARC generation. |
|
=item --marc-output out/marc/test.marc |
|
|
|
|
|
Optional path to output file |
|
75 |
|
|
76 |
=item --marc-lint |
=item --marc-lint |
77 |
|
|
78 |
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 |
79 |
messages with C<--no-marc-lint>. |
messages with C<--no-marc-lint>. |
80 |
|
|
81 |
=item --marc-dump |
=item --marc-dump |
108 |
my $only_filter; |
my $only_filter; |
109 |
my $stats = 0; |
my $stats = 0; |
110 |
my $validate_path; |
my $validate_path; |
111 |
my ($marc_normalize, $marc_output); |
my $marc_generate = 1; |
112 |
my $marc_lint = 1; |
my $marc_lint = 1; |
113 |
my $marc_dump = 0; |
my $marc_dump = 0; |
114 |
my $parallel = 0; |
my $parallel = 0; |
115 |
my $only_links = 0; |
my $only_links = 0; |
116 |
my $merge = 0; |
my $merge = 0; |
117 |
|
my $help; |
118 |
|
|
119 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
120 |
|
|
128 |
"debug+" => \$debug, |
"debug+" => \$debug, |
129 |
"stats" => \$stats, |
"stats" => \$stats, |
130 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
131 |
"marc-normalize=s" => \$marc_normalize, |
"marc-generate!" => \$marc_generate, |
|
"marc-output=s" => \$marc_output, |
|
132 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
133 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
134 |
"parallel=i" => \$parallel, |
"parallel=i" => \$parallel, |
135 |
"only-links!" => \$only_links, |
"only-links!" => \$only_links, |
136 |
"merge" => \$merge, |
"merge" => \$merge, |
137 |
|
"help" => \$help, |
138 |
); |
); |
139 |
|
|
140 |
|
pod2usage(-verbose => 2) if ($help); |
141 |
|
|
142 |
my $config = new WebPAC::Config( path => $config_path ); |
my $config = new WebPAC::Config( path => $config_path ); |
143 |
|
|
144 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
147 |
|
|
148 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
149 |
|
|
150 |
|
my $log_file = 'log'; |
151 |
|
|
152 |
|
if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) { |
153 |
|
$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'"); |
154 |
|
rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!"); |
155 |
|
} |
156 |
|
|
157 |
my $estcmd_fh; |
my $estcmd_fh; |
158 |
my $estcmd_path = './estcmd-merge.sh'; |
my $estcmd_path = './estcmd-merge.sh'; |
171 |
|
|
172 |
|
|
173 |
my $use_indexer = $config->use_indexer; |
my $use_indexer = $config->use_indexer; |
174 |
|
$stats ||= $validate; |
175 |
if ($stats) { |
if ($stats) { |
176 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("disabled indexing for stats collection"); |
177 |
$use_indexer = undef; |
$use_indexer = undef; |
178 |
} else { |
} else { |
179 |
$log->info("using $use_indexer indexing engine..."); |
$log->info("using $use_indexer indexing engine..."); |
180 |
} |
} |
181 |
|
|
|
# disable indexing when creating marc |
|
|
$use_indexer = undef if ($marc_normalize); |
|
|
|
|
182 |
# parse normalize files and create source files for lookup and normalization |
# parse normalize files and create source files for lookup and normalization |
183 |
|
|
184 |
my $parser = new WebPAC::Parser( config => $config ); |
my $parser = new WebPAC::Parser( config => $config ); |
217 |
} |
} |
218 |
|
|
219 |
my $indexer; |
my $indexer; |
220 |
if ($use_indexer) { |
if ($use_indexer && $parser->have_rules( 'search', $database )) { |
221 |
|
|
222 |
my $cfg_name = $use_indexer; |
my $cfg_name = $use_indexer; |
223 |
$cfg_name =~ s/\-.*$//; |
$cfg_name =~ s/\-.*$//; |
290 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
291 |
$abs_path =~ s#/[^/]*$#/#; |
$abs_path =~ s#/[^/]*$#/#; |
292 |
|
|
293 |
my $db_path = $config->get('webpac')->{db_path} . '/' . $database; |
my $db_path = $config->webpac('db_path'); |
294 |
|
|
295 |
if ($clean) { |
if ($clean) { |
296 |
$log->info("creating new database '$database' in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
301 |
|
|
302 |
my $store = new WebPAC::Store( |
my $store = new WebPAC::Store( |
303 |
path => $db_path, |
path => $db_path, |
|
database => $database, |
|
304 |
debug => $debug, |
debug => $debug, |
305 |
); |
); |
306 |
|
|
359 |
delete( $input->{lookup} ); |
delete( $input->{lookup} ); |
360 |
} |
} |
361 |
|
|
|
my $lookup; |
|
362 |
my $lookup_coderef; |
my $lookup_coderef; |
363 |
|
|
364 |
if (@lookups) { |
if (@lookups) { |
372 |
WebPAC::Normalize::data_structure( |
WebPAC::Normalize::data_structure( |
373 |
row => $rec, |
row => $rec, |
374 |
rules => $rules, |
rules => $rules, |
|
lookup => $lookup, |
|
375 |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
376 |
); |
); |
377 |
|
|
378 |
warn "current lookup = ", dump($lookup) if ($lookup); |
#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup()); |
379 |
}; |
}; |
380 |
|
|
381 |
WebPAC::Normalize::_set_lookup( undef ); |
WebPAC::Normalize::_set_lookup( undef ); |
384 |
|
|
385 |
}; |
}; |
386 |
|
|
387 |
|
my $lookup_jar; |
388 |
|
|
389 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
390 |
path => $input->{path}, |
path => $input->{path}, |
391 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
392 |
lookup_coderef => $lookup_coderef, |
lookup_coderef => $lookup_coderef, |
393 |
|
lookup => $lookup_jar, |
394 |
%{ $input }, |
%{ $input }, |
395 |
|
load_row => sub { |
396 |
|
my $a = shift; |
397 |
|
return $store->load_row( |
398 |
|
database => $database, |
399 |
|
input => $input_name, |
400 |
|
id => $a->{id}, |
401 |
|
); |
402 |
|
}, |
403 |
|
save_row => sub { |
404 |
|
my $a = shift; |
405 |
|
return $store->save_row( |
406 |
|
database => $database, |
407 |
|
input => $input_name, |
408 |
|
id => $a->{id}, |
409 |
|
row => $a->{row}, |
410 |
|
); |
411 |
|
}, |
412 |
|
|
413 |
); |
); |
414 |
|
|
415 |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
416 |
|
|
417 |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
418 |
$log->debug("created following lookups: ", dump( $lookup_data )); |
$log->debug("created following lookups: ", sub { dump( $lookup_data ) } ); |
419 |
|
|
420 |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
421 |
$store->save_lookup( $database, $input_name, $key, $lookup_data->{$database}->{$input_name}->{$key} ); |
$store->save_lookup( |
422 |
|
database => $database, |
423 |
|
input => $input_name, |
424 |
|
key => $key, |
425 |
|
data => $lookup_data->{$database}->{$input_name}->{$key}, |
426 |
|
); |
427 |
} |
} |
428 |
} |
} |
429 |
|
|
437 |
$log->info("Generating report file $path"); |
$log->info("Generating report file $path"); |
438 |
} |
} |
439 |
|
|
440 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my $marc; |
441 |
@{ $input->{normalize} } : ( $input->{normalize} ); |
if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { |
442 |
|
$marc = new WebPAC::Output::MARC( |
443 |
if ($marc_normalize) { |
path => "out/marc/${database}-${input_name}.marc", |
444 |
@norm_array = ( { |
lint => $marc_lint, |
445 |
path => $marc_normalize, |
dump => $marc_dump, |
446 |
output => $marc_output || "out/marc/${database}-${input_name}.marc", |
); |
|
} ); |
|
447 |
} |
} |
448 |
|
|
449 |
foreach my $normalize (@norm_array) { |
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
450 |
|
$log->debug("parsed normalize rules:\n$rules"); |
451 |
|
|
452 |
my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); |
# reset position in database |
453 |
|
$input_db->seek(1); |
454 |
|
|
455 |
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
# generate name of config key for indexer (strip everything after -) |
456 |
|
my $indexer_config = $use_indexer; |
457 |
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
458 |
|
|
459 |
$log->info("Using $normalize_path for normalization..."); |
my $lookup_hash; |
460 |
|
my $depends = $parser->depends($database,$input_name); |
461 |
|
|
462 |
|
if ($depends) { |
463 |
|
$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends); |
464 |
|
$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH'); |
465 |
|
|
466 |
|
foreach my $db (keys %$depends) { |
467 |
|
foreach my $i (keys %{$depends->{$db}}) { |
468 |
|
foreach my $k (keys %{$depends->{$db}->{$i}}) { |
469 |
|
my $t = time(); |
470 |
|
$log->debug("loading lookup $db/$i"); |
471 |
|
$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup( |
472 |
|
database => $db, |
473 |
|
input => $i, |
474 |
|
key => $k, |
475 |
|
); |
476 |
|
$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t)); |
477 |
|
} |
478 |
|
} |
479 |
|
} |
480 |
|
|
481 |
my $marc = new WebPAC::Output::MARC( |
$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) }); |
482 |
path => $normalize->{output}, |
} |
|
lint => $marc_lint, |
|
|
dump => $marc_dump, |
|
|
) if ($normalize->{output}); |
|
483 |
|
|
|
# reset position in database |
|
|
$input_db->seek(1); |
|
484 |
|
|
485 |
# generate name of config key for indexer (strip everything after -) |
foreach my $pos ( 0 ... $input_db->size ) { |
|
my $indexer_config = $use_indexer; |
|
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
|
486 |
|
|
487 |
foreach my $pos ( 0 ... $input_db->size ) { |
my $row = $input_db->fetch || next; |
488 |
|
|
489 |
my $row = $input_db->fetch || next; |
$total_rows++; |
490 |
|
|
491 |
my $mfn = $row->{'000'}->[0]; |
my $mfn = $row->{'000'}->[0]; |
492 |
|
|
493 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
if (! $mfn || $mfn !~ m#^\d+$#) { |
494 |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
495 |
$mfn = $pos; |
$mfn = $pos; |
496 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
497 |
} |
} |
498 |
|
|
499 |
|
|
500 |
if ($validate) { |
if ($validate) { |
501 |
if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) { |
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
502 |
$log->error( "MFN $mfn validation error:\n", |
$log->error( "MFN $mfn validation error:\n", |
503 |
$validate->report_error( $errors ) |
$validate->report_error( $errors ) |
504 |
); |
); |
|
} |
|
505 |
} |
} |
506 |
|
next; # validation doesn't create any output |
507 |
|
} |
508 |
|
|
509 |
my $ds = WebPAC::Normalize::data_structure( |
my $ds = WebPAC::Normalize::data_structure( |
510 |
row => $row, |
row => $row, |
511 |
rules => $rules, |
rules => $rules, |
512 |
lookup => $lookup ? $lookup->lookup_hash : undef, |
lookup => $lookup_hash, |
513 |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
514 |
marc_encoding => 'utf-8', |
marc_encoding => 'utf-8', |
515 |
); |
load_row_coderef => sub { |
516 |
|
my ($database,$input,$mfn) = @_; |
517 |
$store->save_ds( |
return $store->load_row( |
518 |
id => $mfn, |
database => $database, |
519 |
ds => $ds, |
input => $input, |
520 |
prefix => $input_name, |
id => $mfn, |
521 |
) if ($ds && !$stats); |
); |
522 |
|
}, |
523 |
$indexer->add( |
); |
524 |
id => "${input_name}/${mfn}", |
|
525 |
ds => $ds, |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
526 |
type => $config->get($indexer_config)->{type}, |
|
527 |
) if ($indexer && $ds); |
$store->save_ds( |
528 |
|
database => $database, |
529 |
if ($marc) { |
input => $input_name, |
530 |
my $i = 0; |
id => $mfn, |
531 |
|
ds => $ds, |
532 |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
) if ($ds && !$stats); |
533 |
$marc->add( |
|
534 |
id => $mfn . ( $i ? "/$i" : '' ), |
$indexer->add( |
535 |
fields => $fields, |
id => "${input_name}/${mfn}", |
536 |
leader => WebPAC::Normalize::marc_leader(), |
ds => $ds, |
537 |
row => $row, |
type => $config->get($indexer_config)->{type}, |
538 |
); |
) if ($indexer && $ds); |
539 |
$i++; |
|
540 |
} |
if ($marc) { |
541 |
|
my $i = 0; |
542 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
|
543 |
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
544 |
|
$marc->add( |
545 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
546 |
|
fields => $fields, |
547 |
|
leader => WebPAC::Normalize::_get_marc_leader(), |
548 |
|
row => $row, |
549 |
|
); |
550 |
|
$i++; |
551 |
} |
} |
552 |
|
|
553 |
$total_rows++; |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
554 |
} |
} |
555 |
|
} |
556 |
|
|
557 |
if ($validate) { |
if ($validate) { |
558 |
my $errors = $validate->report; |
my $errors = $validate->report; |
559 |
if ($errors) { |
if ($errors) { |
560 |
$log->info("validation errors:\n$errors\n" ); |
$log->info("validation errors:\n$errors\n" ); |
561 |
print $report_fh "$errors\n" if ($report_fh); |
print $report_fh "$errors\n" if ($report_fh); |
|
} |
|
562 |
} |
} |
563 |
|
} |
564 |
|
|
565 |
if ($stats) { |
if ($stats) { |
566 |
my $s = $input_db->stats; |
my $s = $input_db->stats; |
567 |
$log->info("statistics of fields usage:\n$s"); |
$log->info("statistics of fields usage:\n$s"); |
568 |
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
569 |
} |
} |
570 |
|
|
571 |
# close MARC file |
# close MARC file |
572 |
$marc->finish if ($marc); |
$marc->finish if ($marc); |
573 |
|
|
574 |
# close report |
# close report |
575 |
close($report_fh) if ($report_fh) |
close($report_fh) if ($report_fh) |
|
} |
|
576 |
|
|
577 |
} |
} |
578 |
|
|