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 |
=item --marc-generate |
80 |
|
|
81 |
Generate MARC file. This will automatically be on if file contains C<marc*> directives. |
Generate MARC file. This will automatically be on if file contains C<marc*> directives. |
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; |
my $marc_generate = 1; |
121 |
my $marc_lint = 1; |
my $marc_lint = 1; |
122 |
my $marc_dump = 0; |
my $marc_dump = 0; |
133 |
"clean" => \$clean, |
"clean" => \$clean, |
134 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
135 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
136 |
"config" => \$config_path, |
"config=s" => \$config_path, |
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, |
"marc-generate!" => \$marc_generate, |
142 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
143 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
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 ); |
155 |
|
|
156 |
|
WebPAC::Normalize::_debug( $debug - 1 ) if $debug > 1; |
157 |
|
|
158 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
159 |
|
|
160 |
die "no databases in config file!\n" unless ($config->databases); |
die "no databases in config file!\n" unless ($config->databases); |
177 |
$log->info("created merge batch file $estcmd_path"); |
$log->info("created merge batch file $estcmd_path"); |
178 |
} |
} |
179 |
|
|
|
|
|
180 |
my $validate; |
my $validate; |
181 |
$validate = new WebPAC::Validate( |
$validate = new WebPAC::Validate( |
182 |
path => $validate_path, |
delimiters => $config->webpac('delimiters'), |
183 |
) if ($validate_path); |
) if ($validate_path || $validate_delimiters_path); |
|
|
|
184 |
|
|
185 |
my $use_indexer = $config->use_indexer; |
my $use_indexer = $config->use_indexer; |
186 |
$stats ||= $validate; |
$stats ||= $validate; |
187 |
if ($stats) { |
if ($stats) { |
188 |
$log->debug("disabled indexing for stats collection"); |
$log->debug("disabled indexing for stats collection"); |
189 |
$use_indexer = undef; |
$use_indexer = undef; |
190 |
} else { |
} elsif ( $use_indexer ) { |
191 |
$log->info("using $use_indexer indexing engine..."); |
$log->info("using $use_indexer indexing engine..."); |
192 |
} |
} |
193 |
|
|
194 |
# parse normalize files and create source files for lookup and normalization |
# parse normalize files and create source files for lookup and normalization |
195 |
|
|
196 |
my $parser = new WebPAC::Parser( config => $config ); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if $only_filter; |
197 |
|
|
198 |
|
my $parser = new WebPAC::Parser( |
199 |
|
config => $config, |
200 |
|
only_database => $only_database, |
201 |
|
only_input => $only_input, |
202 |
|
); |
203 |
|
|
204 |
my $total_rows = 0; |
my $total_rows = 0; |
205 |
my $start_t = time(); |
my $start_t = time(); |
220 |
return $c; |
return $c; |
221 |
} |
} |
222 |
|
|
223 |
while (my ($database, $db_config) = each %{ $config->databases }) { |
foreach my $database ( sort keys %{ $config->databases } ) { |
224 |
|
my $db_config = $config->databases->{$database}; |
225 |
|
|
|
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
|
226 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
227 |
|
|
228 |
if ($parallel) { |
if ($parallel) { |
251 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer eq 'hyperestraier') { |
252 |
|
|
253 |
# open Hyper Estraier database |
# open Hyper Estraier database |
254 |
use WebPAC::Output::Estraier '0.10'; |
require WebPAC::Output::Estraier; |
255 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
256 |
|
|
257 |
} elsif ($use_indexer eq 'hyperestraier-native') { |
} elsif ($use_indexer eq 'hyperestraier-native') { |
258 |
|
|
259 |
# open Hyper Estraier database |
# open Hyper Estraier database |
260 |
use WebPAC::Output::EstraierNative; |
require WebPAC::Output::EstraierNative; |
261 |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
262 |
|
|
263 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
264 |
|
|
265 |
# 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 } ); |
|
266 |
|
|
267 |
} else { |
} else { |
268 |
$log->logdie("unknown use_indexer: $use_indexer"); |
$log->logdie("unknown use_indexer: $use_indexer"); |
269 |
} |
} |
270 |
|
|
271 |
$log->logide("can't continue without valid indexer") unless ($indexer); |
$log->logdie("can't continue without valid indexer") unless ($indexer); |
272 |
} |
} |
273 |
|
|
274 |
|
|
302 |
# |
# |
303 |
# now WebPAC::Store |
# now WebPAC::Store |
304 |
# |
# |
305 |
my $abs_path = abs_path($0); |
my $store = new WebPAC::Store({ |
306 |
$abs_path =~ s#/[^/]*$#/#; |
debug => $debug, |
307 |
|
}); |
308 |
|
|
|
my $db_path = $config->webpac('db_path'); |
|
309 |
|
|
310 |
if ($clean) { |
# |
311 |
$log->info("creating new database '$database' in $db_path"); |
# prepare output |
312 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
# |
313 |
} else { |
my @outputs = force_array( $db_config->{output}, sub { |
314 |
$log->info("working on database '$database' in $db_path"); |
$log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" ); |
315 |
} |
} ); |
316 |
|
|
317 |
my $store = new WebPAC::Store( |
my @output_modules; |
318 |
path => $db_path, |
|
319 |
debug => $debug, |
foreach my $output ( @outputs ) { |
320 |
); |
|
321 |
|
#warn '## output = ',dump( $output ); |
322 |
|
|
323 |
|
my $module = $output->{module} || $log->logdie("need module in output section of $database"); |
324 |
|
$module = 'WebPAC::Output::' . $module unless $module =~ m/::/; |
325 |
|
|
326 |
|
$log->debug("loading output module $module"); |
327 |
|
eval "require $module"; |
328 |
|
|
329 |
|
# add database to arugemnts for output filter |
330 |
|
$output->{database} = $database; |
331 |
|
$output->{clean} = $clean; |
332 |
|
|
333 |
|
$log->debug("calling $module->new(",dump( $output ),")"); |
334 |
|
my $out = new $module->new( $output ); |
335 |
|
if ( $out->init ) { |
336 |
|
push @output_modules, $out; |
337 |
|
} else { |
338 |
|
$log->warn("SKIPPED $module"); |
339 |
|
} |
340 |
|
} |
341 |
|
|
342 |
|
|
343 |
# |
# |
344 |
# now, iterate through input formats |
# now, iterate through input formats |
345 |
# |
# |
346 |
|
|
347 |
my @inputs; |
|
348 |
if (ref($db_config->{input}) eq 'ARRAY') { |
my @inputs = force_array( $db_config->{input}, sub { |
|
@inputs = @{ $db_config->{input} }; |
|
|
} elsif ($db_config->{input}) { |
|
|
push @inputs, $db_config->{input}; |
|
|
} else { |
|
349 |
$log->info("database $database doesn't have inputs defined"); |
$log->info("database $database doesn't have inputs defined"); |
350 |
} |
} ); |
351 |
|
|
352 |
|
if ( -e 'out/debug' ) { # fixme flag? |
353 |
|
my $out; |
354 |
|
foreach my $i ( @inputs ) { |
355 |
|
warn dump( $i ); |
356 |
|
next unless defined $i->{normalize}; |
357 |
|
warn dump( $i->{normalize} ); |
358 |
|
foreach my $normalize ( @{ $i->{normalize} } ) { |
359 |
|
my $path = $normalize->{path}; |
360 |
|
$out .= qq/\n##\n## $path\n##\n\n/; |
361 |
|
$out .= read_file( $path ); |
362 |
|
} |
363 |
|
} |
364 |
|
my $all = "out/debug/all-normalize.pl"; |
365 |
|
write_file( $all, $out ); |
366 |
|
warn "### all normalize for this input saved to: $all"; |
367 |
|
}; |
368 |
|
|
369 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
370 |
|
|
374 |
|
|
375 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
376 |
|
|
377 |
die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs'))); |
# FIXME check if input module exists |
378 |
|
my $input_module = $input->{module}; |
379 |
|
|
380 |
my $input_module = $config->webpac('inputs')->{$type}; |
if ( ! $input_module ) { |
381 |
|
if ( grep(/$type/, $config->webpac('inputs')) ) { |
382 |
|
$input_module = $config->webpac('inputs')->{$type}; |
383 |
|
} else { |
384 |
|
$log->logdie("I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!" ); |
385 |
|
} |
386 |
|
} |
387 |
|
|
388 |
my @lookups = $parser->have_lookup_create($database, $input); |
my @lookups = $parser->have_lookup_create($database, $input); |
389 |
|
|
390 |
$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module", |
$log->info("working on $database/$input_name with $input_module on $input->{path}", |
391 |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
392 |
); |
); |
393 |
|
|
406 |
stats => $stats, |
stats => $stats, |
407 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
408 |
modify_file => $input->{modify_file}, |
modify_file => $input->{modify_file}, |
409 |
|
input_config => $input, |
410 |
); |
); |
411 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
412 |
|
|
491 |
print $report_fh "Report for database '$database' input '$input_name' records ", |
print $report_fh "Report for database '$database' input '$input_name' records ", |
492 |
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
493 |
$log->info("Generating report file $path"); |
$log->info("Generating report file $path"); |
494 |
|
|
495 |
|
if ( $validate ) { |
496 |
|
$validate->read_validate_file( $validate->fill_in( $validate_path, database => $database, input => $input_name ) ) if ( $validate_path ); |
497 |
|
$validate->read_validate_delimiters_file( $validate->fill_in( $validate_delimiters_path, database => $database, input => $input_name ) ) if ( $validate_delimiters_path ); |
498 |
|
} |
499 |
} |
} |
500 |
|
|
501 |
my $marc; |
my $marc; |
507 |
); |
); |
508 |
} |
} |
509 |
|
|
510 |
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
my $rules = $parser->normalize_rules($database,$input_name); |
511 |
|
$log->logwarn("no normalize rules for $database/$input_name") unless $rules; |
512 |
|
|
513 |
$log->debug("parsed normalize rules:\n$rules"); |
$log->debug("parsed normalize rules:\n$rules"); |
514 |
|
|
515 |
# reset position in database |
# reset position in database |
545 |
} |
} |
546 |
|
|
547 |
|
|
548 |
|
# setup input name for all output filters |
549 |
|
foreach my $out ( @output_modules ) { |
550 |
|
if ( $out->can('input') ) { |
551 |
|
$out->input( $input_name ); |
552 |
|
} else { |
553 |
|
$log->warn("output filter ",ref($out)," doesn't support input name"); |
554 |
|
} |
555 |
|
} |
556 |
|
|
557 |
|
|
558 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
559 |
|
|
560 |
my $row = $input_db->fetch || next; |
my $row = $input_db->fetch || next; |
563 |
|
|
564 |
my $mfn = $row->{'000'}->[0]; |
my $mfn = $row->{'000'}->[0]; |
565 |
|
|
566 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
if (! $mfn || $mfn !~ m{^\d+$}) { |
567 |
$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"); |
568 |
$mfn = $pos; |
$mfn = $pos; |
569 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
579 |
next; # validation doesn't create any output |
next; # validation doesn't create any output |
580 |
} |
} |
581 |
|
|
582 |
my $ds = WebPAC::Normalize::data_structure( |
if ($rules) { |
583 |
row => $row, |
|
584 |
rules => $rules, |
my $ds = WebPAC::Normalize::data_structure( |
585 |
lookup => $lookup_hash, |
row => $row, |
586 |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
rules => $rules, |
587 |
marc_encoding => 'utf-8', |
lookup => $lookup_hash, |
588 |
load_row_coderef => sub { |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
589 |
my ($database,$input,$mfn) = @_; |
marc_encoding => 'utf-8', |
590 |
return $store->load_row( |
load_row_coderef => sub { |
591 |
|
my ($database,$input,$mfn) = @_; |
592 |
|
#warn "### load_row($database,$input,$mfn) from data_structure\n"; |
593 |
|
return $store->load_row( |
594 |
|
database => $database, |
595 |
|
input => $input, |
596 |
|
id => $mfn, |
597 |
|
); |
598 |
|
}, |
599 |
|
); |
600 |
|
|
601 |
|
$log->debug("ds = ", sub { dump($ds) }); |
602 |
|
|
603 |
|
if ( $ds ) { |
604 |
|
|
605 |
|
$store->save_ds( |
606 |
database => $database, |
database => $database, |
607 |
input => $input, |
input => $input_name, |
608 |
id => $mfn, |
id => $mfn, |
609 |
); |
ds => $ds, |
610 |
}, |
) if !$stats; |
611 |
); |
|
612 |
|
$indexer->add( |
613 |
|
id => "${input_name}/${mfn}", |
614 |
|
ds => $ds, |
615 |
|
type => $config->get($indexer_config)->{type}, |
616 |
|
) if $indexer; |
617 |
|
|
618 |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
foreach my $out ( @output_modules ) { |
619 |
|
$out->add( $mfn, $ds ) if $out->can('add'); |
620 |
|
} |
621 |
|
|
622 |
$store->save_ds( |
} else { |
623 |
database => $database, |
$log->warn("record $pos didn't produce any output after normalization rules!") unless $marc; |
624 |
input => $input_name, |
} |
625 |
id => $mfn, |
} |
|
ds => $ds, |
|
|
) if ($ds && !$stats); |
|
|
|
|
|
$indexer->add( |
|
|
id => "${input_name}/${mfn}", |
|
|
ds => $ds, |
|
|
type => $config->get($indexer_config)->{type}, |
|
|
) if ($indexer && $ds); |
|
626 |
|
|
627 |
if ($marc) { |
if ($marc) { |
628 |
my $i = 0; |
my $i = 0; |
629 |
|
|
630 |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
while (my $fields = WebPAC::Normalize::MARC::_get_marc_fields( fetch_next => 1 ) ) { |
631 |
$marc->add( |
$marc->add( |
632 |
id => $mfn . ( $i ? "/$i" : '' ), |
id => $mfn . ( $i ? "/$i" : '' ), |
633 |
fields => $fields, |
fields => $fields, |
634 |
leader => WebPAC::Normalize::_get_marc_leader(), |
leader => WebPAC::Normalize::MARC::_get_marc_leader(), |
635 |
row => $row, |
row => $row, |
636 |
); |
); |
637 |
$i++; |
$i++; |
639 |
|
|
640 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
641 |
} |
} |
642 |
|
|
643 |
} |
} |
644 |
|
|
645 |
if ($validate) { |
if ($validate) { |
648 |
$log->info("validation errors:\n$errors\n" ); |
$log->info("validation errors:\n$errors\n" ); |
649 |
print $report_fh "$errors\n" if ($report_fh); |
print $report_fh "$errors\n" if ($report_fh); |
650 |
} |
} |
651 |
|
|
652 |
|
print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n"; |
653 |
|
|
654 |
|
# must be last thing that touches $validate for this input |
655 |
|
$validate->reset; |
656 |
} |
} |
657 |
|
|
658 |
if ($stats) { |
if ($stats) { |
665 |
$marc->finish if ($marc); |
$marc->finish if ($marc); |
666 |
|
|
667 |
# close report |
# close report |
668 |
close($report_fh) if ($report_fh) |
close($report_fh) if ($report_fh); |
|
|
|
669 |
} |
} |
670 |
|
|
671 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
672 |
|
|
673 |
|
foreach my $out ( @output_modules ) { |
674 |
|
$out->finish if $out->can('finish'); |
675 |
|
} |
676 |
|
|
677 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
678 |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
679 |
sprintf("in %.2f sec [%.2f rec/sec]", |
sprintf("in %.2f sec [%.2f rec/sec]", |
696 |
$log->info("all parallel processes finished"); |
$log->info("all parallel processes finished"); |
697 |
} |
} |
698 |
|
|
699 |
|
# save new delimiters if needed |
700 |
|
$validate->save_delimiters_templates if ( $validate_delimiters_path ); |
701 |
|
|
702 |
# |
# |
703 |
# handle links or merge after indexing |
# handle links or merge after indexing |
704 |
# |
# |