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 0.11; |
use WebPAC::Validate 0.11; |
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) { |
263 |
|
|
264 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
265 |
|
|
266 |
# open KinoSearch |
die "no longer supported"; |
|
require WebPAC::Output::KinoSearch; |
|
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
|
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
|
267 |
|
|
268 |
} else { |
} else { |
269 |
$log->logdie("unknown use_indexer: $use_indexer"); |
$log->logdie("unknown use_indexer: $use_indexer"); |
303 |
# |
# |
304 |
# now WebPAC::Store |
# now WebPAC::Store |
305 |
# |
# |
306 |
my $abs_path = abs_path($0); |
my $store = new WebPAC::Store({ |
|
$abs_path =~ s#/[^/]*$#/#; # |
|
|
|
|
|
my $db_path = $config->webpac('db_path'); |
|
|
|
|
|
if ($clean) { |
|
|
$log->info("creating new database '$database' in $db_path"); |
|
|
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
|
|
} else { |
|
|
$log->info("working on database '$database' in $db_path"); |
|
|
} |
|
|
|
|
|
my $store = new WebPAC::Store( |
|
|
path => $db_path, |
|
307 |
debug => $debug, |
debug => $debug, |
308 |
); |
}); |
309 |
|
|
310 |
|
|
311 |
# |
# |
312 |
# prepare output |
# prepare output |
313 |
# |
# |
314 |
my @outputs; |
my @outputs = force_array( $db_config->{output}, sub { |
315 |
if (defined( $db_config->{output} )) { |
$log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" ); |
316 |
my $module = $db_config->{output}->{module} || $log->logdie("need module in output section of $database"); |
} ); |
317 |
|
|
318 |
|
my @output_modules; |
319 |
|
|
320 |
|
foreach my $output ( @outputs ) { |
321 |
|
|
322 |
|
#warn '## output = ',dump( $output ); |
323 |
|
|
324 |
|
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 |
my $out = new $module->new( $db_config->{output} ); |
|
330 |
$out->init; |
# add database to arugemnts for output filter |
331 |
push @outputs, $out; |
$output->{database} = $database; |
332 |
|
$output->{clean} = $clean; |
333 |
|
|
334 |
|
$log->debug("calling $module->new(",dump( $output ),")"); |
335 |
|
my $out = new $module->new( $output ); |
336 |
|
if ( $out->init ) { |
337 |
|
push @output_modules, $out; |
338 |
|
} else { |
339 |
|
$log->warn("SKIPPED $module"); |
340 |
|
} |
341 |
} |
} |
342 |
|
|
343 |
|
|
345 |
# now, iterate through input formats |
# now, iterate through input formats |
346 |
# |
# |
347 |
|
|
348 |
my @inputs; |
|
349 |
if (ref($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 { |
|
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 |
|
|
375 |
|
|
376 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
377 |
|
|
378 |
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 |
379 |
|
my $input_module = $input->{module}; |
380 |
|
|
381 |
my $input_module = $config->webpac('inputs')->{$type}; |
if ( ! $input_module ) { |
382 |
|
if ( grep(/$type/, $config->webpac('inputs')) ) { |
383 |
|
$input_module = $config->webpac('inputs')->{$type}; |
384 |
|
} else { |
385 |
|
$log->logdie("I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!" ); |
386 |
|
} |
387 |
|
} |
388 |
|
|
389 |
my @lookups = $parser->have_lookup_create($database, $input); |
my @lookups = $parser->have_lookup_create($database, $input); |
390 |
|
|
391 |
$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module", |
$log->info("working on $database/$input_name with $input_module on $input->{path}", |
392 |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
393 |
); |
); |
394 |
|
|
400 |
|
|
401 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
402 |
module => $input_module, |
module => $input_module, |
|
encoding => $config->webpac('webpac_encoding'), |
|
403 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
404 |
offset => $offset, |
offset => $offset, |
405 |
recode => $input->{recode}, |
recode => $input->{recode}, |
444 |
|
|
445 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
446 |
path => $input->{path}, |
path => $input->{path}, |
447 |
code_page => $input->{encoding}, # database encoding |
input_encoding => $input->{encoding}, # database encoding |
448 |
lookup_coderef => $lookup_coderef, |
lookup_coderef => $lookup_coderef, |
449 |
lookup => $lookup_jar, |
lookup => $lookup_jar, |
450 |
%{ $input }, |
%{ $input }, |
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; |
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 |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
$indexer->add( |
613 |
|
id => "${input_name}/${mfn}", |
614 |
|
ds => $ds, |
615 |
|
type => $config->get($indexer_config)->{type}, |
616 |
|
) if $indexer; |
617 |
|
|
618 |
$store->save_ds( |
foreach my $out ( @output_modules ) { |
619 |
database => $database, |
$out->add( $mfn, $ds ) if $out->can('add'); |
620 |
input => $input_name, |
} |
621 |
id => $mfn, |
|
622 |
ds => $ds, |
} else { |
623 |
) if ($ds && !$stats); |
$log->warn("record $pos didn't produce any output after normalization rules!") unless $marc; |
624 |
|
} |
625 |
$indexer->add( |
} |
|
id => "${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++; |
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 |
|
|
|
foreach my $out ( @outputs ) { |
|
|
$out->add( $mfn, $ds ) if $out->can('add'); |
|
|
} |
|
|
|
|
643 |
} |
} |
644 |
|
|
645 |
if ($validate) { |
if ($validate) { |
668 |
close($report_fh) if ($report_fh); |
close($report_fh) if ($report_fh); |
669 |
} |
} |
670 |
|
|
671 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
$indexer->finish if $indexer && $indexer->can('finish'); |
672 |
|
|
673 |
foreach my $out ( @outputs ) { |
foreach my $out ( @output_modules ) { |
674 |
$out->finish if $out->can('finish'); |
$out->finish if $out->can('finish'); |
675 |
} |
} |
676 |
|
|