23 |
use Storable qw/dclone/; |
use Storable qw/dclone/; |
24 |
use Pod::Usage qw/pod2usage/; |
use Pod::Usage qw/pod2usage/; |
25 |
|
|
|
use Proc::Queue size => 1; |
|
26 |
use POSIX ":sys_wait_h"; # imports WNOHANG |
use POSIX ":sys_wait_h"; # imports WNOHANG |
27 |
|
|
28 |
=head1 NAME |
=head1 NAME |
132 |
"clean" => \$clean, |
"clean" => \$clean, |
133 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
134 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
135 |
"config" => \$config_path, |
"config=s" => \$config_path, |
136 |
"debug+" => \$debug, |
"debug+" => \$debug, |
137 |
"stats" => \$stats, |
"stats" => \$stats, |
138 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
152 |
|
|
153 |
my $config = new WebPAC::Config( path => $config_path ); |
my $config = new WebPAC::Config( path => $config_path ); |
154 |
|
|
155 |
|
WebPAC::Normalize::_debug( $debug - 1 ) if $debug > 1; |
156 |
|
|
157 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
158 |
|
|
159 |
die "no databases in config file!\n" unless ($config->databases); |
die "no databases in config file!\n" unless ($config->databases); |
192 |
|
|
193 |
# parse normalize files and create source files for lookup and normalization |
# parse normalize files and create source files for lookup and normalization |
194 |
|
|
195 |
my $parser = new WebPAC::Parser( config => $config ); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if $only_filter; |
196 |
|
|
197 |
|
my $parser = new WebPAC::Parser( |
198 |
|
config => $config, |
199 |
|
only_database => $only_database, |
200 |
|
only_input => $only_input, |
201 |
|
); |
202 |
|
|
203 |
my $total_rows = 0; |
my $total_rows = 0; |
204 |
my $start_t = time(); |
my $start_t = time(); |
206 |
my @links; |
my @links; |
207 |
|
|
208 |
if ($parallel) { |
if ($parallel) { |
209 |
|
eval 'use Proc::Queue size => 1;'; |
210 |
|
die $@ if $@; |
211 |
$log->info("Using $parallel processes for speedup"); |
$log->info("Using $parallel processes for speedup"); |
212 |
Proc::Queue::size($parallel); |
Proc::Queue::size($parallel); |
213 |
} |
} |
224 |
foreach my $database ( sort keys %{ $config->databases } ) { |
foreach my $database ( sort keys %{ $config->databases } ) { |
225 |
my $db_config = $config->databases->{$database}; |
my $db_config = $config->databases->{$database}; |
226 |
|
|
|
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
|
227 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
228 |
|
|
229 |
if ($parallel) { |
if ($parallel) { |
307 |
debug => $debug, |
debug => $debug, |
308 |
}); |
}); |
309 |
|
|
|
sub iterate_over { |
|
|
my ( $what, $error ) = @_; |
|
|
my @result; |
|
|
if ( ref( $what ) eq 'ARRAY' ) { |
|
|
@result = @{ $what }; |
|
|
} elsif ($db_config->{input}) { |
|
|
@result = ( $what ); |
|
|
} else { |
|
|
$error->() if ref($error) eq 'CODE'; |
|
|
} |
|
|
return @result; |
|
|
} |
|
310 |
|
|
311 |
# |
# |
312 |
# prepare output |
# prepare output |
313 |
# |
# |
314 |
my @outputs = iterate_over( $db_config->{output}, sub { |
my @outputs = force_array( $db_config->{output}, sub { |
315 |
$log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" ); |
$log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" ); |
316 |
} ); |
} ); |
317 |
|
|
319 |
|
|
320 |
foreach my $output ( @outputs ) { |
foreach my $output ( @outputs ) { |
321 |
|
|
322 |
warn '## output = ',dump( $output ); |
#warn '## output = ',dump( $output ); |
323 |
|
|
324 |
my $module = $output->{module} || $log->logdie("need module in output section of $database"); |
my $module = $output->{module} || $log->logdie("need module in output section of $database"); |
325 |
$module = 'WebPAC::Output::' . $module unless $module =~ m/::/; |
$module = 'WebPAC::Output::' . $module unless $module =~ m/::/; |
326 |
|
|
327 |
$log->debug("loading output module $module"); |
$log->debug("loading output module $module"); |
328 |
eval "require $module"; |
eval "require $module"; |
329 |
|
|
330 |
|
# add database to arugemnts for output filter |
331 |
$output->{database} = $database; |
$output->{database} = $database; |
332 |
|
$output->{clean} = $clean; |
333 |
|
|
334 |
$log->debug("calling $module->new(",dump( $output ),")"); |
$log->debug("calling $module->new(",dump( $output ),")"); |
335 |
my $out = new $module->new( $output ); |
my $out = new $module->new( $output ); |
336 |
$out->init; |
if ( $out->init ) { |
337 |
|
push @output_modules, $out; |
338 |
push @output_modules, $out; |
} else { |
339 |
|
$log->warn("SKIPPED $module"); |
340 |
|
} |
341 |
} |
} |
342 |
|
|
343 |
|
|
346 |
# |
# |
347 |
|
|
348 |
|
|
349 |
my @inputs = iterate_over( $db_config->{input}, sub { |
my @inputs = force_array( $db_config->{input}, sub { |
350 |
$log->info("database $database doesn't have inputs defined"); |
$log->info("database $database doesn't have inputs defined"); |
351 |
} ); |
} ); |
352 |
|
|
353 |
|
if ( -e 'out/debug' ) { # FIXME flag? |
354 |
|
my $out; |
355 |
|
foreach my $i ( @inputs ) { |
356 |
|
warn dump( $i ); |
357 |
|
next unless defined $i->{normalize}; |
358 |
|
warn dump( $i->{normalize} ); |
359 |
|
foreach my $normalize ( @{ $i->{normalize} } ) { |
360 |
|
my $path = $normalize->{path}; |
361 |
|
$out .= qq/\n##\n## $path\n##\n\n/; |
362 |
|
$out .= read_file( $path ); |
363 |
|
} |
364 |
|
} |
365 |
|
my $all = "out/debug/all-normalize.pl"; |
366 |
|
write_file( $all, $out ); |
367 |
|
warn "### all normalize for this input saved to: $all"; |
368 |
|
}; |
369 |
|
|
370 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
371 |
|
|
372 |
my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input)); |
my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input)); |
373 |
|
|
374 |
|
if ( $input->{skip} ) { |
375 |
|
$log->info("skip $input_name"); |
376 |
|
next; |
377 |
|
} |
378 |
|
|
379 |
next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
380 |
|
|
381 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
382 |
|
|
383 |
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 |
384 |
|
my $input_module = $input->{module}; |
385 |
|
|
386 |
my $input_module = $config->webpac('inputs')->{$type}; |
if ( ! $input_module ) { |
387 |
|
if ( grep(/$type/, $config->webpac('inputs')) ) { |
388 |
|
$input_module = $config->webpac('inputs')->{$type}; |
389 |
|
} else { |
390 |
|
$log->logdie("I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!" ); |
391 |
|
} |
392 |
|
} |
393 |
|
|
394 |
my @lookups = $parser->have_lookup_create($database, $input); |
my @lookups = $parser->have_lookup_create($database, $input); |
395 |
|
|
396 |
$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module", |
$log->info("working on $database/$input_name with $input_module on $input->{path}", |
397 |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
398 |
); |
); |
399 |
|
|
405 |
|
|
406 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
407 |
module => $input_module, |
module => $input_module, |
|
encoding => $config->webpac('webpac_encoding'), |
|
408 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
409 |
offset => $offset, |
offset => $offset, |
410 |
recode => $input->{recode}, |
recode => $input->{recode}, |
449 |
|
|
450 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
451 |
path => $input->{path}, |
path => $input->{path}, |
452 |
code_page => $input->{encoding}, # database encoding |
input_encoding => $input->{encoding}, # database encoding |
453 |
lookup_coderef => $lookup_coderef, |
lookup_coderef => $lookup_coderef, |
454 |
lookup => $lookup_jar, |
lookup => $lookup_jar, |
455 |
%{ $input }, |
%{ $input }, |
512 |
); |
); |
513 |
} |
} |
514 |
|
|
515 |
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); |
516 |
|
$log->logwarn("no normalize rules for $database/$input_name") unless $rules; |
517 |
|
|
518 |
$log->debug("parsed normalize rules:\n$rules"); |
$log->debug("parsed normalize rules:\n$rules"); |
519 |
|
|
520 |
# reset position in database |
# reset position in database |
550 |
} |
} |
551 |
|
|
552 |
|
|
553 |
|
# setup input name for all output filters |
554 |
|
foreach my $out ( @output_modules ) { |
555 |
|
if ( $out->can('input') ) { |
556 |
|
$out->input( $input_name ); |
557 |
|
} else { |
558 |
|
$log->warn("output filter ",ref($out)," doesn't support input name"); |
559 |
|
} |
560 |
|
} |
561 |
|
|
562 |
|
|
563 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
564 |
|
|
565 |
my $row = $input_db->fetch || next; |
my $row = $input_db->fetch || next; |
584 |
next; # validation doesn't create any output |
next; # validation doesn't create any output |
585 |
} |
} |
586 |
|
|
587 |
my $ds = WebPAC::Normalize::data_structure( |
if ($rules) { |
588 |
row => $row, |
|
589 |
rules => $rules, |
my $ds = WebPAC::Normalize::data_structure( |
590 |
lookup => $lookup_hash, |
row => $row, |
591 |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
rules => $rules, |
592 |
marc_encoding => 'utf-8', |
lookup => $lookup_hash, |
593 |
load_row_coderef => sub { |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
594 |
my ($database,$input,$mfn) = @_; |
marc_encoding => 'utf-8', |
595 |
return $store->load_row( |
load_row_coderef => sub { |
596 |
|
my ($database,$input,$mfn) = @_; |
597 |
|
#warn "### load_row($database,$input,$mfn) from data_structure\n"; |
598 |
|
return $store->load_row( |
599 |
|
database => $database, |
600 |
|
input => $input, |
601 |
|
id => $mfn, |
602 |
|
); |
603 |
|
}, |
604 |
|
); |
605 |
|
|
606 |
|
$log->debug("ds = ", sub { dump($ds) }); |
607 |
|
|
608 |
|
if ( $ds ) { |
609 |
|
|
610 |
|
$store->save_ds( |
611 |
database => $database, |
database => $database, |
612 |
input => $input, |
input => $input_name, |
613 |
id => $mfn, |
id => $mfn, |
614 |
); |
ds => $ds, |
615 |
}, |
) if !$stats; |
|
); |
|
616 |
|
|
617 |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
$indexer->add( |
618 |
|
id => "${input_name}/${mfn}", |
619 |
|
ds => $ds, |
620 |
|
type => $config->get($indexer_config)->{type}, |
621 |
|
) if $indexer; |
622 |
|
|
623 |
|
foreach my $out ( @output_modules ) { |
624 |
|
$out->add( $mfn, $ds ) if $out->can('add'); |
625 |
|
} |
626 |
|
|
627 |
$store->save_ds( |
} else { |
628 |
database => $database, |
$log->warn("record $pos didn't produce any output after normalization rules!") unless $marc; |
629 |
input => $input_name, |
} |
630 |
id => $mfn, |
} |
|
ds => $ds, |
|
|
) if ($ds && !$stats); |
|
|
|
|
|
$indexer->add( |
|
|
id => "${input_name}/${mfn}", |
|
|
ds => $ds, |
|
|
type => $config->get($indexer_config)->{type}, |
|
|
) if ($indexer && $ds); |
|
631 |
|
|
632 |
if ($marc) { |
if ($marc) { |
633 |
my $i = 0; |
my $i = 0; |
634 |
|
|
635 |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
while (my $fields = WebPAC::Normalize::MARC::_get_marc_fields( fetch_next => 1 ) ) { |
636 |
$marc->add( |
$marc->add( |
637 |
id => $mfn . ( $i ? "/$i" : '' ), |
id => $mfn . ( $i ? "/$i" : '' ), |
638 |
fields => $fields, |
fields => $fields, |
639 |
leader => WebPAC::Normalize::_get_marc_leader(), |
leader => WebPAC::Normalize::MARC::_get_marc_leader(), |
640 |
row => $row, |
row => $row, |
641 |
); |
); |
642 |
$i++; |
$i++; |
645 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
646 |
} |
} |
647 |
|
|
|
foreach my $out ( @output_modules ) { |
|
|
$out->add( $mfn, $ds ) if $out->can('add'); |
|
|
} |
|
|
|
|
648 |
} |
} |
649 |
|
|
650 |
if ($validate) { |
if ($validate) { |
673 |
close($report_fh) if ($report_fh); |
close($report_fh) if ($report_fh); |
674 |
} |
} |
675 |
|
|
676 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
$indexer->finish if $indexer && $indexer->can('finish'); |
677 |
|
|
678 |
foreach my $out ( @output_modules ) { |
foreach my $out ( @output_modules ) { |
679 |
$out->finish if $out->can('finish'); |
$out->finish if $out->can('finish'); |