9 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
10 |
use WebPAC::Parser 0.08; |
use WebPAC::Parser 0.08; |
11 |
use WebPAC::Input 0.16; |
use WebPAC::Input 0.16; |
12 |
use WebPAC::Store 0.14; |
use WebPAC::Store 0.15; |
13 |
use WebPAC::Normalize 0.22; |
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; |
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 |
|
You can use special variables C<$database> and $C<$input> in this parametar |
72 |
|
like C<--validate 'conf/validate/$database-$input'> to construct filename |
73 |
|
|
74 |
|
=item --validate-delimiters path/to/validate_delimiters_file |
75 |
|
|
76 |
|
this option is used with C<--validate> to turn on extra validation of |
77 |
|
delimiters. If file is non existant, it will be created on first run. |
78 |
|
|
79 |
|
=item --marc-generate |
80 |
|
|
81 |
|
Generate MARC file. This will automatically be on if file contains C<marc*> directives. |
82 |
|
You can use this option as C<--no-marc-generate> to disable MARC generation. |
83 |
|
|
84 |
=item --marc-lint |
=item --marc-lint |
85 |
|
|
86 |
By default turned on if normalisation file has C<marc*> directives. You can disable lint |
By default turned on if normalisation file has C<marc*> directives. You can disable lint |
116 |
my $only_filter; |
my $only_filter; |
117 |
my $stats = 0; |
my $stats = 0; |
118 |
my $validate_path; |
my $validate_path; |
119 |
|
my $validate_delimiters_path; |
120 |
|
my $marc_generate = 1; |
121 |
my $marc_lint = 1; |
my $marc_lint = 1; |
122 |
my $marc_dump = 0; |
my $marc_dump = 0; |
123 |
my $parallel = 0; |
my $parallel = 0; |
137 |
"debug+" => \$debug, |
"debug+" => \$debug, |
138 |
"stats" => \$stats, |
"stats" => \$stats, |
139 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
140 |
|
"validate-delimiters=s" => \$validate_delimiters_path, |
141 |
|
"marc-generate!" => \$marc_generate, |
142 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
143 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
144 |
"parallel=i" => \$parallel, |
"parallel=i" => \$parallel, |
147 |
"help" => \$help, |
"help" => \$help, |
148 |
); |
); |
149 |
|
|
150 |
|
$marc_generate = 0 if ( $validate_delimiters_path ); |
151 |
|
|
152 |
pod2usage(-verbose => 2) if ($help); |
pod2usage(-verbose => 2) if ($help); |
153 |
|
|
154 |
my $config = new WebPAC::Config( path => $config_path ); |
my $config = new WebPAC::Config( path => $config_path ); |
175 |
$log->info("created merge batch file $estcmd_path"); |
$log->info("created merge batch file $estcmd_path"); |
176 |
} |
} |
177 |
|
|
|
|
|
178 |
my $validate; |
my $validate; |
179 |
$validate = new WebPAC::Validate( |
$validate = new WebPAC::Validate( |
180 |
path => $validate_path, |
delimiters => $config->webpac('delimiters'), |
181 |
) if ($validate_path); |
) if ($validate_path || $validate_delimiters_path); |
|
|
|
182 |
|
|
183 |
my $use_indexer = $config->use_indexer; |
my $use_indexer = $config->use_indexer; |
184 |
$stats ||= $validate; |
$stats ||= $validate; |
185 |
if ($stats) { |
if ($stats) { |
186 |
$log->debug("disabled indexing for stats collection"); |
$log->debug("disabled indexing for stats collection"); |
187 |
$use_indexer = undef; |
$use_indexer = undef; |
188 |
} else { |
} elsif ( $use_indexer ) { |
189 |
$log->info("using $use_indexer indexing engine..."); |
$log->info("using $use_indexer indexing engine..."); |
190 |
} |
} |
191 |
|
|
212 |
return $c; |
return $c; |
213 |
} |
} |
214 |
|
|
215 |
while (my ($database, $db_config) = each %{ $config->databases }) { |
foreach my $database ( sort keys %{ $config->databases } ) { |
216 |
|
my $db_config = $config->databases->{$database}; |
217 |
|
|
218 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
219 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
244 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer eq 'hyperestraier') { |
245 |
|
|
246 |
# open Hyper Estraier database |
# open Hyper Estraier database |
247 |
use WebPAC::Output::Estraier '0.10'; |
require WebPAC::Output::Estraier; |
248 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
249 |
|
|
250 |
} elsif ($use_indexer eq 'hyperestraier-native') { |
} elsif ($use_indexer eq 'hyperestraier-native') { |
251 |
|
|
252 |
# open Hyper Estraier database |
# open Hyper Estraier database |
253 |
use WebPAC::Output::EstraierNative; |
require WebPAC::Output::EstraierNative; |
254 |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
255 |
|
|
256 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
257 |
|
|
258 |
# open KinoSearch |
die "no longer supported"; |
|
use WebPAC::Output::KinoSearch; |
|
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
|
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
|
259 |
|
|
260 |
} else { |
} else { |
261 |
$log->logdie("unknown use_indexer: $use_indexer"); |
$log->logdie("unknown use_indexer: $use_indexer"); |
262 |
} |
} |
263 |
|
|
264 |
$log->logide("can't continue without valid indexer") unless ($indexer); |
$log->logdie("can't continue without valid indexer") unless ($indexer); |
265 |
} |
} |
266 |
|
|
267 |
|
|
295 |
# |
# |
296 |
# now WebPAC::Store |
# now WebPAC::Store |
297 |
# |
# |
298 |
my $abs_path = abs_path($0); |
my $store = new WebPAC::Store({ |
299 |
$abs_path =~ s#/[^/]*$#/#; |
debug => $debug, |
300 |
|
}); |
|
my $db_path = $config->webpac('db_path'); |
|
301 |
|
|
302 |
if ($clean) { |
# |
303 |
$log->info("creating new database '$database' in $db_path"); |
# prepare output |
304 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
# |
305 |
} else { |
my @outputs; |
306 |
$log->info("working on database '$database' in $db_path"); |
if (defined( $db_config->{output} )) { |
307 |
|
my $module = $db_config->{output}->{module} || $log->logdie("need module in output section of $database"); |
308 |
|
$module = 'WebPAC::Output::' . $module unless $module =~ m/::/; |
309 |
|
$log->debug("loading output module $module"); |
310 |
|
eval "require $module"; |
311 |
|
$db_config->{output}->{database} = $database; |
312 |
|
$log->debug("calling $module->new(",dump( $db_config->{output} ),")"); |
313 |
|
my $out = new $module->new( $db_config->{output} ); |
314 |
|
$out->init; |
315 |
|
push @outputs, $out; |
316 |
} |
} |
317 |
|
|
|
my $store = new WebPAC::Store( |
|
|
path => $db_path, |
|
|
debug => $debug, |
|
|
); |
|
|
|
|
318 |
|
|
319 |
# |
# |
320 |
# now, iterate through input formats |
# now, iterate through input formats |
362 |
stats => $stats, |
stats => $stats, |
363 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
364 |
modify_file => $input->{modify_file}, |
modify_file => $input->{modify_file}, |
365 |
|
input_config => $input, |
366 |
); |
); |
367 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
368 |
|
|
447 |
print $report_fh "Report for database '$database' input '$input_name' records ", |
print $report_fh "Report for database '$database' input '$input_name' records ", |
448 |
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
449 |
$log->info("Generating report file $path"); |
$log->info("Generating report file $path"); |
450 |
|
|
451 |
|
if ( $validate ) { |
452 |
|
$validate->read_validate_file( $validate->fill_in( $validate_path, database => $database, input => $input_name ) ) if ( $validate_path ); |
453 |
|
$validate->read_validate_delimiters_file( $validate->fill_in( $validate_delimiters_path, database => $database, input => $input_name ) ) if ( $validate_delimiters_path ); |
454 |
|
} |
455 |
} |
} |
456 |
|
|
457 |
my $marc; |
my $marc; |
458 |
if ($parser->have_rules( 'marc', $database, $input_name )) { |
if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { |
459 |
$marc = new WebPAC::Output::MARC( |
$marc = new WebPAC::Output::MARC( |
460 |
path => "out/marc/${database}-${input_name}.marc", |
path => "out/marc/${database}-${input_name}.marc", |
461 |
lint => $marc_lint, |
lint => $marc_lint, |
507 |
|
|
508 |
my $mfn = $row->{'000'}->[0]; |
my $mfn = $row->{'000'}->[0]; |
509 |
|
|
510 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
if (! $mfn || $mfn !~ m{^\d+$}) { |
511 |
$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"); |
512 |
$mfn = $pos; |
$mfn = $pos; |
513 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
561 |
$marc->add( |
$marc->add( |
562 |
id => $mfn . ( $i ? "/$i" : '' ), |
id => $mfn . ( $i ? "/$i" : '' ), |
563 |
fields => $fields, |
fields => $fields, |
564 |
leader => WebPAC::Normalize::marc_leader(), |
leader => WebPAC::Normalize::_get_marc_leader(), |
565 |
row => $row, |
row => $row, |
566 |
); |
); |
567 |
$i++; |
$i++; |
569 |
|
|
570 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
571 |
} |
} |
572 |
|
|
573 |
|
foreach my $out ( @outputs ) { |
574 |
|
$out->add( $mfn, $ds ) if $out->can('add'); |
575 |
|
} |
576 |
|
|
577 |
} |
} |
578 |
|
|
579 |
if ($validate) { |
if ($validate) { |
582 |
$log->info("validation errors:\n$errors\n" ); |
$log->info("validation errors:\n$errors\n" ); |
583 |
print $report_fh "$errors\n" if ($report_fh); |
print $report_fh "$errors\n" if ($report_fh); |
584 |
} |
} |
585 |
|
|
586 |
|
print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n"; |
587 |
|
|
588 |
|
# must be last thing that touches $validate for this input |
589 |
|
$validate->reset; |
590 |
} |
} |
591 |
|
|
592 |
if ($stats) { |
if ($stats) { |
599 |
$marc->finish if ($marc); |
$marc->finish if ($marc); |
600 |
|
|
601 |
# close report |
# close report |
602 |
close($report_fh) if ($report_fh) |
close($report_fh) if ($report_fh); |
|
|
|
603 |
} |
} |
604 |
|
|
605 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
606 |
|
|
607 |
|
foreach my $out ( @outputs ) { |
608 |
|
$out->finish if $out->can('finish'); |
609 |
|
} |
610 |
|
|
611 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
612 |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
613 |
sprintf("in %.2f sec [%.2f rec/sec]", |
sprintf("in %.2f sec [%.2f rec/sec]", |
630 |
$log->info("all parallel processes finished"); |
$log->info("all parallel processes finished"); |
631 |
} |
} |
632 |
|
|
633 |
|
# save new delimiters if needed |
634 |
|
$validate->save_delimiters_templates if ( $validate_delimiters_path ); |
635 |
|
|
636 |
# |
# |
637 |
# handle links or merge after indexing |
# handle links or merge after indexing |
638 |
# |
# |