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 0.11; |
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 0.11; |
16 |
use WebPAC::Output::MARC; |
use WebPAC::Output::MARC; |
17 |
use WebPAC::Config; |
use WebPAC::Config; |
18 |
use Getopt::Long; |
use Getopt::Long; |
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 --validate-delimiters path/to/validate_delimiters_file |
72 |
|
|
73 |
This option specifies normalisation file for MARC creation |
this option is used with C<--validate> to turn on extra validation of |
74 |
|
delimiters. If file is non existant, it will be created on first run. |
75 |
|
|
76 |
=item --marc-output out/marc/test.marc |
=item --marc-generate |
77 |
|
|
78 |
Optional path to output file |
Generate MARC file. This will automatically be on if file contains C<marc*> directives. |
79 |
|
You can use this option as C<--no-marc-generate> to disable MARC generation. |
80 |
|
|
81 |
=item --marc-lint |
=item --marc-lint |
82 |
|
|
83 |
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 |
84 |
messages with C<--no-marc-lint>. |
messages with C<--no-marc-lint>. |
85 |
|
|
86 |
=item --marc-dump |
=item --marc-dump |
113 |
my $only_filter; |
my $only_filter; |
114 |
my $stats = 0; |
my $stats = 0; |
115 |
my $validate_path; |
my $validate_path; |
116 |
my ($marc_normalize, $marc_output); |
my $validate_delimiters_path; |
117 |
|
my $marc_generate = 1; |
118 |
my $marc_lint = 1; |
my $marc_lint = 1; |
119 |
my $marc_dump = 0; |
my $marc_dump = 0; |
120 |
my $parallel = 0; |
my $parallel = 0; |
121 |
my $only_links = 0; |
my $only_links = 0; |
122 |
my $merge = 0; |
my $merge = 0; |
123 |
|
my $help; |
124 |
|
|
125 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
126 |
|
|
134 |
"debug+" => \$debug, |
"debug+" => \$debug, |
135 |
"stats" => \$stats, |
"stats" => \$stats, |
136 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
137 |
"marc-normalize=s" => \$marc_normalize, |
"validate-delimiters=s" => \$validate_delimiters_path, |
138 |
"marc-output=s" => \$marc_output, |
"marc-generate!" => \$marc_generate, |
139 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
140 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
141 |
"parallel=i" => \$parallel, |
"parallel=i" => \$parallel, |
142 |
"only-links!" => \$only_links, |
"only-links!" => \$only_links, |
143 |
"merge" => \$merge, |
"merge" => \$merge, |
144 |
|
"help" => \$help, |
145 |
); |
); |
146 |
|
|
147 |
|
pod2usage(-verbose => 2) if ($help); |
148 |
|
|
149 |
my $config = new WebPAC::Config( path => $config_path ); |
my $config = new WebPAC::Config( path => $config_path ); |
150 |
|
|
151 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
154 |
|
|
155 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
156 |
|
|
157 |
|
my $log_file = 'log'; |
158 |
|
|
159 |
|
if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) { |
160 |
|
$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'"); |
161 |
|
rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!"); |
162 |
|
} |
163 |
|
|
164 |
my $estcmd_fh; |
my $estcmd_fh; |
165 |
my $estcmd_path = './estcmd-merge.sh'; |
my $estcmd_path = './estcmd-merge.sh'; |
170 |
$log->info("created merge batch file $estcmd_path"); |
$log->info("created merge batch file $estcmd_path"); |
171 |
} |
} |
172 |
|
|
|
|
|
173 |
my $validate; |
my $validate; |
174 |
$validate = new WebPAC::Validate( |
$validate = new WebPAC::Validate( |
175 |
path => $validate_path, |
path => $validate_path, |
176 |
) if ($validate_path); |
delimiters => $config->webpac('delimiters'), |
177 |
|
delimiters_path => $validate_delimiters_path, |
178 |
|
) if ($validate_path || $validate_delimiters_path); |
179 |
|
|
180 |
my $use_indexer = $config->use_indexer; |
my $use_indexer = $config->use_indexer; |
181 |
|
$stats ||= $validate; |
182 |
if ($stats) { |
if ($stats) { |
183 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("disabled indexing for stats collection"); |
184 |
$use_indexer = undef; |
$use_indexer = undef; |
185 |
} else { |
} else { |
186 |
$log->info("using $use_indexer indexing engine..."); |
$log->info("using $use_indexer indexing engine..."); |
187 |
} |
} |
188 |
|
|
|
# disable indexing when creating marc |
|
|
$use_indexer = undef if ($marc_normalize); |
|
|
|
|
189 |
# parse normalize files and create source files for lookup and normalization |
# parse normalize files and create source files for lookup and normalization |
190 |
|
|
191 |
my $parser = new WebPAC::Parser( config => $config ); |
my $parser = new WebPAC::Parser( config => $config ); |
224 |
} |
} |
225 |
|
|
226 |
my $indexer; |
my $indexer; |
227 |
if ($use_indexer) { |
if ($use_indexer && $parser->have_rules( 'search', $database )) { |
228 |
|
|
229 |
my $cfg_name = $use_indexer; |
my $cfg_name = $use_indexer; |
230 |
$cfg_name =~ s/\-.*$//; |
$cfg_name =~ s/\-.*$//; |
366 |
delete( $input->{lookup} ); |
delete( $input->{lookup} ); |
367 |
} |
} |
368 |
|
|
|
my $lookup; |
|
369 |
my $lookup_coderef; |
my $lookup_coderef; |
370 |
|
|
371 |
if (@lookups) { |
if (@lookups) { |
379 |
WebPAC::Normalize::data_structure( |
WebPAC::Normalize::data_structure( |
380 |
row => $rec, |
row => $rec, |
381 |
rules => $rules, |
rules => $rules, |
|
lookup => $lookup, |
|
382 |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
383 |
); |
); |
384 |
|
|
385 |
warn "current lookup = ", dump($lookup) if ($lookup); |
#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup()); |
386 |
}; |
}; |
387 |
|
|
388 |
WebPAC::Normalize::_set_lookup( undef ); |
WebPAC::Normalize::_set_lookup( undef ); |
391 |
|
|
392 |
}; |
}; |
393 |
|
|
394 |
|
my $lookup_jar; |
395 |
|
|
396 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
397 |
path => $input->{path}, |
path => $input->{path}, |
398 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
399 |
lookup_coderef => $lookup_coderef, |
lookup_coderef => $lookup_coderef, |
400 |
|
lookup => $lookup_jar, |
401 |
%{ $input }, |
%{ $input }, |
402 |
|
load_row => sub { |
403 |
|
my $a = shift; |
404 |
|
return $store->load_row( |
405 |
|
database => $database, |
406 |
|
input => $input_name, |
407 |
|
id => $a->{id}, |
408 |
|
); |
409 |
|
}, |
410 |
|
save_row => sub { |
411 |
|
my $a = shift; |
412 |
|
return $store->save_row( |
413 |
|
database => $database, |
414 |
|
input => $input_name, |
415 |
|
id => $a->{id}, |
416 |
|
row => $a->{row}, |
417 |
|
); |
418 |
|
}, |
419 |
|
|
420 |
); |
); |
421 |
|
|
422 |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
423 |
|
|
424 |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
425 |
$log->debug("created following lookups: ", dump( $lookup_data )); |
$log->debug("created following lookups: ", sub { dump( $lookup_data ) } ); |
426 |
|
|
427 |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
428 |
$store->save_lookup( |
$store->save_lookup( |
444 |
$log->info("Generating report file $path"); |
$log->info("Generating report file $path"); |
445 |
} |
} |
446 |
|
|
447 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my $marc; |
448 |
@{ $input->{normalize} } : ( $input->{normalize} ); |
if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { |
449 |
|
$marc = new WebPAC::Output::MARC( |
450 |
if ($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}, |
|
451 |
lint => $marc_lint, |
lint => $marc_lint, |
452 |
dump => $marc_dump, |
dump => $marc_dump, |
453 |
) if ($normalize->{output}); |
); |
454 |
|
} |
455 |
|
|
456 |
# reset position in database |
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
457 |
$input_db->seek(1); |
$log->debug("parsed normalize rules:\n$rules"); |
458 |
|
|
459 |
# generate name of config key for indexer (strip everything after -) |
# reset position in database |
460 |
my $indexer_config = $use_indexer; |
$input_db->seek(1); |
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
|
461 |
|
|
462 |
foreach my $pos ( 0 ... $input_db->size ) { |
# generate name of config key for indexer (strip everything after -) |
463 |
|
my $indexer_config = $use_indexer; |
464 |
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
465 |
|
|
466 |
|
my $lookup_hash; |
467 |
|
my $depends = $parser->depends($database,$input_name); |
468 |
|
|
469 |
|
if ($depends) { |
470 |
|
$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends); |
471 |
|
$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH'); |
472 |
|
|
473 |
|
foreach my $db (keys %$depends) { |
474 |
|
foreach my $i (keys %{$depends->{$db}}) { |
475 |
|
foreach my $k (keys %{$depends->{$db}->{$i}}) { |
476 |
|
my $t = time(); |
477 |
|
$log->debug("loading lookup $db/$i"); |
478 |
|
$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup( |
479 |
|
database => $db, |
480 |
|
input => $i, |
481 |
|
key => $k, |
482 |
|
); |
483 |
|
$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t)); |
484 |
|
} |
485 |
|
} |
486 |
|
} |
487 |
|
|
488 |
my $row = $input_db->fetch || next; |
$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) }); |
489 |
|
} |
490 |
|
|
|
my $mfn = $row->{'000'}->[0]; |
|
491 |
|
|
492 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
foreach my $pos ( 0 ... $input_db->size ) { |
|
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
|
|
$mfn = $pos; |
|
|
push @{ $row->{'000'} }, $pos; |
|
|
} |
|
493 |
|
|
494 |
|
my $row = $input_db->fetch || next; |
495 |
|
|
496 |
if ($validate) { |
$total_rows++; |
|
if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) { |
|
|
$log->error( "MFN $mfn validation error:\n", |
|
|
$validate->report_error( $errors ) |
|
|
); |
|
|
} |
|
|
} |
|
497 |
|
|
498 |
my $ds = WebPAC::Normalize::data_structure( |
my $mfn = $row->{'000'}->[0]; |
|
row => $row, |
|
|
rules => $rules, |
|
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
|
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
|
|
marc_encoding => 'utf-8', |
|
|
); |
|
499 |
|
|
500 |
$store->save_ds( |
if (! $mfn || $mfn !~ m#^\d+$#) { |
501 |
database => $database, |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
502 |
input => $input_name, |
$mfn = $pos; |
503 |
id => $mfn, |
push @{ $row->{'000'} }, $pos; |
504 |
ds => $ds, |
} |
|
) if ($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++; |
|
|
} |
|
505 |
|
|
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
|
|
} |
|
506 |
|
|
507 |
$total_rows++; |
if ($validate) { |
508 |
|
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
509 |
|
$log->error( "MFN $mfn validation error:\n", |
510 |
|
$validate->report_error( $errors ) |
511 |
|
); |
512 |
|
} |
513 |
|
next; # validation doesn't create any output |
514 |
} |
} |
515 |
|
|
516 |
if ($validate) { |
my $ds = WebPAC::Normalize::data_structure( |
517 |
my $errors = $validate->report; |
row => $row, |
518 |
if ($errors) { |
rules => $rules, |
519 |
$log->info("validation errors:\n$errors\n" ); |
lookup => $lookup_hash, |
520 |
print $report_fh "$errors\n" if ($report_fh); |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
521 |
|
marc_encoding => 'utf-8', |
522 |
|
load_row_coderef => sub { |
523 |
|
my ($database,$input,$mfn) = @_; |
524 |
|
return $store->load_row( |
525 |
|
database => $database, |
526 |
|
input => $input, |
527 |
|
id => $mfn, |
528 |
|
); |
529 |
|
}, |
530 |
|
); |
531 |
|
|
532 |
|
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
533 |
|
|
534 |
|
$store->save_ds( |
535 |
|
database => $database, |
536 |
|
input => $input_name, |
537 |
|
id => $mfn, |
538 |
|
ds => $ds, |
539 |
|
) if ($ds && !$stats); |
540 |
|
|
541 |
|
$indexer->add( |
542 |
|
id => "${input_name}/${mfn}", |
543 |
|
ds => $ds, |
544 |
|
type => $config->get($indexer_config)->{type}, |
545 |
|
) if ($indexer && $ds); |
546 |
|
|
547 |
|
if ($marc) { |
548 |
|
my $i = 0; |
549 |
|
|
550 |
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
551 |
|
$marc->add( |
552 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
553 |
|
fields => $fields, |
554 |
|
leader => WebPAC::Normalize::_get_marc_leader(), |
555 |
|
row => $row, |
556 |
|
); |
557 |
|
$i++; |
558 |
} |
} |
559 |
|
|
560 |
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
561 |
} |
} |
562 |
|
} |
563 |
|
|
564 |
if ($stats) { |
if ($validate) { |
565 |
my $s = $input_db->stats; |
my $errors = $validate->report; |
566 |
$log->info("statistics of fields usage:\n$s"); |
if ($errors) { |
567 |
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
$log->info("validation errors:\n$errors\n" ); |
568 |
|
print $report_fh "$errors\n" if ($report_fh); |
569 |
} |
} |
570 |
|
|
571 |
# close MARC file |
print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1 ), "\n\n"; |
572 |
$marc->finish if ($marc); |
} |
573 |
|
|
574 |
# close report |
if ($stats) { |
575 |
close($report_fh) if ($report_fh) |
my $s = $input_db->stats; |
576 |
|
$log->info("statistics of fields usage:\n$s"); |
577 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
578 |
} |
} |
579 |
|
|
580 |
|
# close MARC file |
581 |
|
$marc->finish if ($marc); |
582 |
|
|
583 |
|
# close report |
584 |
|
close($report_fh) if ($report_fh); |
585 |
|
|
586 |
|
# reset validate stats |
587 |
|
$validate->reset if ( $validate ); |
588 |
} |
} |
589 |
|
|
590 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
611 |
$log->info("all parallel processes finished"); |
$log->info("all parallel processes finished"); |
612 |
} |
} |
613 |
|
|
614 |
|
# save new delimiters if needed |
615 |
|
$validate->save_delimiters_templates if ( $validate ); |
616 |
|
|
617 |
# |
# |
618 |
# handle links or merge after indexing |
# handle links or merge after indexing |
619 |
# |
# |