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; |
16 |
use WebPAC::Output::MARC; |
use WebPAC::Output::MARC; |
17 |
use WebPAC::Config; |
use WebPAC::Config; |
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/; |
use Pod::Usage qw/pod2usage/; |
25 |
|
use LWP::Simple qw//; |
26 |
|
|
|
use Proc::Queue size => 1; |
|
27 |
use POSIX ":sys_wait_h"; # imports WNOHANG |
use POSIX ":sys_wait_h"; # imports WNOHANG |
28 |
|
|
29 |
=head1 NAME |
=head1 NAME |
103 |
|
|
104 |
Create merged index of databases which have links |
Create merged index of databases which have links |
105 |
|
|
106 |
|
=item --mirror http://www.example.com |
107 |
|
|
108 |
|
Tries to download input path files from mirror URI |
109 |
|
|
110 |
=back |
=back |
111 |
|
|
112 |
=cut |
=cut |
127 |
my $parallel = 0; |
my $parallel = 0; |
128 |
my $only_links = 0; |
my $only_links = 0; |
129 |
my $merge = 0; |
my $merge = 0; |
130 |
|
my $mirror; |
131 |
my $help; |
my $help; |
132 |
|
|
133 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
138 |
"clean" => \$clean, |
"clean" => \$clean, |
139 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
140 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
141 |
"config" => \$config_path, |
"config=s" => \$config_path, |
142 |
"debug+" => \$debug, |
"debug+" => \$debug, |
143 |
"stats" => \$stats, |
"stats" => \$stats, |
144 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
149 |
"parallel=i" => \$parallel, |
"parallel=i" => \$parallel, |
150 |
"only-links!" => \$only_links, |
"only-links!" => \$only_links, |
151 |
"merge" => \$merge, |
"merge" => \$merge, |
152 |
|
"mirror=s" => \$mirror, |
153 |
"help" => \$help, |
"help" => \$help, |
154 |
); |
); |
155 |
|
|
159 |
|
|
160 |
my $config = new WebPAC::Config( path => $config_path ); |
my $config = new WebPAC::Config( path => $config_path ); |
161 |
|
|
162 |
|
WebPAC::Normalize::_debug( $debug - 1 ) if $debug > 1; |
163 |
|
|
164 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
165 |
|
|
166 |
die "no databases in config file!\n" unless ($config->databases); |
die "no databases in config file!\n" unless ($config->databases); |
193 |
if ($stats) { |
if ($stats) { |
194 |
$log->debug("disabled indexing for stats collection"); |
$log->debug("disabled indexing for stats collection"); |
195 |
$use_indexer = undef; |
$use_indexer = undef; |
196 |
} else { |
} elsif ( $use_indexer ) { |
197 |
$log->info("using $use_indexer indexing engine..."); |
$log->info("using $use_indexer indexing engine..."); |
198 |
} |
} |
199 |
|
|
200 |
# parse normalize files and create source files for lookup and normalization |
# parse normalize files and create source files for lookup and normalization |
201 |
|
|
202 |
my $parser = new WebPAC::Parser( config => $config ); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if $only_filter; |
203 |
|
|
204 |
|
my $parser = new WebPAC::Parser( |
205 |
|
config => $config, |
206 |
|
only_database => $only_database, |
207 |
|
only_input => $only_input, |
208 |
|
); |
209 |
|
|
210 |
my $total_rows = 0; |
my $total_rows = 0; |
211 |
my $start_t = time(); |
my $start_t = time(); |
213 |
my @links; |
my @links; |
214 |
|
|
215 |
if ($parallel) { |
if ($parallel) { |
216 |
|
eval 'use Proc::Queue size => 1;'; |
217 |
|
die $@ if $@; |
218 |
$log->info("Using $parallel processes for speedup"); |
$log->info("Using $parallel processes for speedup"); |
219 |
Proc::Queue::size($parallel); |
Proc::Queue::size($parallel); |
220 |
} |
} |
231 |
foreach my $database ( sort keys %{ $config->databases } ) { |
foreach my $database ( sort keys %{ $config->databases } ) { |
232 |
my $db_config = $config->databases->{$database}; |
my $db_config = $config->databases->{$database}; |
233 |
|
|
|
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
|
234 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
235 |
|
|
236 |
if ($parallel) { |
if ($parallel) { |
259 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer eq 'hyperestraier') { |
260 |
|
|
261 |
# open Hyper Estraier database |
# open Hyper Estraier database |
262 |
use WebPAC::Output::Estraier '0.10'; |
require WebPAC::Output::Estraier; |
263 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
264 |
|
|
265 |
} elsif ($use_indexer eq 'hyperestraier-native') { |
} elsif ($use_indexer eq 'hyperestraier-native') { |
266 |
|
|
267 |
# open Hyper Estraier database |
# open Hyper Estraier database |
268 |
use WebPAC::Output::EstraierNative; |
require WebPAC::Output::EstraierNative; |
269 |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
270 |
|
|
271 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
272 |
|
|
273 |
# 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 } ); |
|
274 |
|
|
275 |
} else { |
} else { |
276 |
$log->logdie("unknown use_indexer: $use_indexer"); |
$log->logdie("unknown use_indexer: $use_indexer"); |
277 |
} |
} |
278 |
|
|
279 |
$log->logide("can't continue without valid indexer") unless ($indexer); |
$log->logdie("can't continue without valid indexer") unless ($indexer); |
280 |
} |
} |
281 |
|
|
282 |
|
|
310 |
# |
# |
311 |
# now WebPAC::Store |
# now WebPAC::Store |
312 |
# |
# |
313 |
my $abs_path = abs_path($0); |
my $store = new WebPAC::Store({ |
314 |
$abs_path =~ s#/[^/]*$#/#; |
debug => $debug, |
315 |
|
}); |
316 |
|
|
|
my $db_path = $config->webpac('db_path'); |
|
317 |
|
|
318 |
if ($clean) { |
# |
319 |
$log->info("creating new database '$database' in $db_path"); |
# prepare output |
320 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
# |
321 |
} else { |
my @outputs = force_array( $db_config->{output}, sub { |
322 |
$log->info("working on database '$database' in $db_path"); |
$log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" ); |
323 |
} |
} ); |
324 |
|
|
325 |
my $store = new WebPAC::Store( |
my @output_modules; |
326 |
path => $db_path, |
|
327 |
debug => $debug, |
foreach my $output ( @outputs ) { |
328 |
); |
|
329 |
|
#warn '## output = ',dump( $output ); |
330 |
|
|
331 |
|
my $module = $output->{module} || $log->logdie("need module in output section of $database"); |
332 |
|
$module = 'WebPAC::Output::' . $module unless $module =~ m/::/; |
333 |
|
|
334 |
|
$log->debug("loading output module $module"); |
335 |
|
eval "require $module"; |
336 |
|
|
337 |
|
# add database to arugemnts for output filter |
338 |
|
$output->{database} = $database; |
339 |
|
$output->{clean} = $clean; |
340 |
|
|
341 |
|
$log->debug("calling $module->new(",dump( $output ),")"); |
342 |
|
my $out = new $module->new( $output ); |
343 |
|
if ( $out->init ) { |
344 |
|
push @output_modules, $out; |
345 |
|
} else { |
346 |
|
$log->warn("SKIPPED $module"); |
347 |
|
} |
348 |
|
} |
349 |
|
|
350 |
|
|
351 |
# |
# |
352 |
# now, iterate through input formats |
# now, iterate through input formats |
353 |
# |
# |
354 |
|
|
355 |
my @inputs; |
|
356 |
if (ref($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 { |
|
357 |
$log->info("database $database doesn't have inputs defined"); |
$log->info("database $database doesn't have inputs defined"); |
358 |
} |
} ); |
359 |
|
|
360 |
|
if ( -e 'out/debug' ) { # FIXME flag? |
361 |
|
my $out; |
362 |
|
foreach my $i ( @inputs ) { |
363 |
|
warn dump( $i ); |
364 |
|
next unless defined $i->{normalize}; |
365 |
|
warn dump( $i->{normalize} ); |
366 |
|
foreach my $normalize ( @{ $i->{normalize} } ) { |
367 |
|
my $path = $normalize->{path}; |
368 |
|
$out .= qq/\n##\n## $path\n##\n\n/; |
369 |
|
$out .= read_file( $path ); |
370 |
|
} |
371 |
|
} |
372 |
|
my $all = "out/debug/all-normalize.pl"; |
373 |
|
write_file( $all, $out ); |
374 |
|
warn "### all normalize for this input saved to: $all"; |
375 |
|
}; |
376 |
|
|
377 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
378 |
|
|
379 |
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)); |
380 |
|
|
381 |
|
if ( $input->{skip} ) { |
382 |
|
$log->info("skip $input_name"); |
383 |
|
next; |
384 |
|
} |
385 |
|
|
386 |
next if ($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)); |
387 |
|
|
388 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
389 |
|
|
390 |
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 |
391 |
|
my $input_module = $input->{module}; |
392 |
|
|
393 |
my $input_module = $config->webpac('inputs')->{$type}; |
if ( ! $input_module ) { |
394 |
|
if ( grep(/$type/, $config->webpac('inputs')) ) { |
395 |
|
$input_module = $config->webpac('inputs')->{$type}; |
396 |
|
} else { |
397 |
|
$log->logdie("I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!" ); |
398 |
|
} |
399 |
|
} |
400 |
|
|
401 |
my @lookups = $parser->have_lookup_create($database, $input); |
my @lookups = $parser->have_lookup_create($database, $input); |
402 |
|
|
403 |
$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module", |
$log->info("working on $database/$input_name with $input_module on $input->{path}", |
404 |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
405 |
); |
); |
406 |
|
|
410 |
delete($input->{modify_file}); |
delete($input->{modify_file}); |
411 |
} |
} |
412 |
|
|
413 |
|
if ( $mirror ) { |
414 |
|
my $path = $input->{path} || die "no input path in ",dump( $input ); |
415 |
|
|
416 |
|
my $base = $path; |
417 |
|
$base =~ s{/[^/]+$}{}; |
418 |
|
mkpath $base unless -e $base; |
419 |
|
|
420 |
|
my $rc = LWP::Simple::mirror( "$mirror/$path", $path ); |
421 |
|
if (LWP::Simple::is_error( $rc )) { |
422 |
|
die "can't mirror $mirror/$path -> $path [$rc]"; |
423 |
|
} else { |
424 |
|
$log->info( "mirror ", $path, " [$rc] ", -s $path, " bytes" ); |
425 |
|
} |
426 |
|
|
427 |
|
} |
428 |
|
|
429 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
430 |
module => $input_module, |
module => $input_module, |
|
encoding => $config->webpac('webpac_encoding'), |
|
431 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
432 |
offset => $offset, |
offset => $offset, |
433 |
recode => $input->{recode}, |
recode => $input->{recode}, |
434 |
stats => $stats, |
stats => $stats, |
435 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
436 |
modify_file => $input->{modify_file}, |
modify_file => $input->{modify_file}, |
437 |
|
input_config => $input, |
438 |
); |
); |
439 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
440 |
|
|
472 |
|
|
473 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
474 |
path => $input->{path}, |
path => $input->{path}, |
475 |
code_page => $input->{encoding}, # database encoding |
input_encoding => $input->{encoding}, # database encoding |
476 |
lookup_coderef => $lookup_coderef, |
lookup_coderef => $lookup_coderef, |
477 |
lookup => $lookup_jar, |
lookup => $lookup_jar, |
478 |
%{ $input }, |
%{ $input }, |
513 |
|
|
514 |
my $report_fh; |
my $report_fh; |
515 |
if ($stats || $validate) { |
if ($stats || $validate) { |
516 |
my $path = "out/report/${database}-${input_name}.txt"; |
my $out_report = 'out/report'; # FIXME move to config |
517 |
|
mkpath $out_report unless -e $out_report; |
518 |
|
my $path = "$out_report/${database}-${input_name}.txt"; |
519 |
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
520 |
|
|
521 |
print $report_fh "Report for database '$database' input '$input_name' records ", |
print $report_fh "Report for database '$database' input '$input_name' records ", |
530 |
|
|
531 |
my $marc; |
my $marc; |
532 |
if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { |
if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { |
533 |
|
|
534 |
|
my $out_marc = 'out/marc'; # FIXME move to config |
535 |
|
mkpath $out_marc unless -e $out_marc; |
536 |
|
|
537 |
$marc = new WebPAC::Output::MARC( |
$marc = new WebPAC::Output::MARC( |
538 |
path => "out/marc/${database}-${input_name}.marc", |
path => "$out_marc/${database}-${input_name}.marc", |
539 |
lint => $marc_lint, |
lint => $marc_lint, |
540 |
dump => $marc_dump, |
dump => $marc_dump, |
541 |
); |
); |
542 |
} |
} |
543 |
|
|
544 |
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); |
545 |
|
if ( ! $rules ) { |
546 |
|
$log->logwarn("no normalize rules for $database/$input_name", $input_db->input_module->can('normalize') ? " using normalize from input module" : ''); |
547 |
|
next; |
548 |
|
} |
549 |
|
|
550 |
$log->debug("parsed normalize rules:\n$rules"); |
$log->debug("parsed normalize rules:\n$rules"); |
551 |
|
|
552 |
# reset position in database |
# reset position in database |
582 |
} |
} |
583 |
|
|
584 |
|
|
585 |
|
# setup input name for all output filters |
586 |
|
foreach my $out ( @output_modules ) { |
587 |
|
if ( $out->can('input') ) { |
588 |
|
$out->input( $input_name ); |
589 |
|
} else { |
590 |
|
$log->warn("output filter ",ref($out)," doesn't support input name"); |
591 |
|
} |
592 |
|
} |
593 |
|
|
594 |
|
|
595 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
596 |
|
|
597 |
my $row = $input_db->fetch || next; |
my $row = $input_db->fetch || next; |
600 |
|
|
601 |
my $mfn = $row->{'000'}->[0]; |
my $mfn = $row->{'000'}->[0]; |
602 |
|
|
603 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
if (! $mfn || $mfn !~ m{^\d+$}) { |
604 |
$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"); |
605 |
$mfn = $pos; |
$mfn = $pos; |
606 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
607 |
} |
} |
608 |
|
|
609 |
|
foreach my $out ( @output_modules ) { |
610 |
|
$out->add_row( $mfn, $row ) if $out->can('add_row'); |
611 |
|
} |
612 |
|
|
613 |
if ($validate) { |
if ($validate) { |
614 |
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
619 |
next; # validation doesn't create any output |
next; # validation doesn't create any output |
620 |
} |
} |
621 |
|
|
622 |
my $ds = WebPAC::Normalize::data_structure( |
my $ds; |
|
row => $row, |
|
|
rules => $rules, |
|
|
lookup => $lookup_hash, |
|
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
|
|
marc_encoding => 'utf-8', |
|
|
load_row_coderef => sub { |
|
|
my ($database,$input,$mfn) = @_; |
|
|
return $store->load_row( |
|
|
database => $database, |
|
|
input => $input, |
|
|
id => $mfn, |
|
|
); |
|
|
}, |
|
|
); |
|
623 |
|
|
624 |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
if ($rules) { |
625 |
|
|
626 |
$store->save_ds( |
$ds = WebPAC::Normalize::data_structure( |
627 |
database => $database, |
row => $row, |
628 |
input => $input_name, |
rules => $rules, |
629 |
id => $mfn, |
lookup => $lookup_hash, |
630 |
ds => $ds, |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
631 |
) if ($ds && !$stats); |
marc_encoding => 'utf-8', |
632 |
|
load_row_coderef => sub { |
633 |
$indexer->add( |
my ($database,$input,$mfn) = @_; |
634 |
id => "${input_name}/${mfn}", |
#warn "### load_row($database,$input,$mfn) from data_structure\n"; |
635 |
ds => $ds, |
return $store->load_row( |
636 |
type => $config->get($indexer_config)->{type}, |
database => $database, |
637 |
) if ($indexer && $ds); |
input => $input, |
638 |
|
id => $mfn, |
639 |
|
); |
640 |
|
}, |
641 |
|
); |
642 |
|
|
643 |
|
} elsif ( $input_db->input_module->can('normalize') ) { |
644 |
|
$ds = $input_db->input_module->normalize( $mfn ); |
645 |
|
} |
646 |
|
|
647 |
|
if ( $ds ) { |
648 |
|
$log->debug("ds = ", sub { dump($ds) }); |
649 |
|
|
650 |
|
$store->save_ds( |
651 |
|
database => $database, |
652 |
|
input => $input_name, |
653 |
|
id => $mfn, |
654 |
|
ds => $ds, |
655 |
|
) if !$stats; |
656 |
|
|
657 |
|
$indexer->add( |
658 |
|
id => "${input_name}/${mfn}", |
659 |
|
ds => $ds, |
660 |
|
type => $config->get($indexer_config)->{type}, |
661 |
|
) if $indexer; |
662 |
|
|
663 |
|
foreach my $out ( @output_modules ) { |
664 |
|
$out->add( $mfn, $ds ) if $out->can('add'); |
665 |
|
} |
666 |
|
|
667 |
|
} else { |
668 |
|
$log->warn("record $pos didn't produce any output after normalization rules!") unless $marc; |
669 |
|
} |
670 |
if ($marc) { |
if ($marc) { |
671 |
my $i = 0; |
my $i = 0; |
672 |
|
|
673 |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
while (my $fields = WebPAC::Normalize::MARC::_get_marc_fields( fetch_next => 1 ) ) { |
674 |
$marc->add( |
$marc->add( |
675 |
id => $mfn . ( $i ? "/$i" : '' ), |
id => $mfn . ( $i ? "/$i" : '' ), |
676 |
fields => $fields, |
fields => $fields, |
677 |
leader => WebPAC::Normalize::_get_marc_leader(), |
leader => WebPAC::Normalize::MARC::_get_marc_leader(), |
678 |
row => $row, |
row => $row, |
679 |
); |
); |
680 |
$i++; |
$i++; |
682 |
|
|
683 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
684 |
} |
} |
685 |
|
|
686 |
} |
} |
687 |
|
|
688 |
if ($validate) { |
if ($validate) { |
711 |
close($report_fh) if ($report_fh); |
close($report_fh) if ($report_fh); |
712 |
} |
} |
713 |
|
|
714 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
$indexer->finish if $indexer && $indexer->can('finish'); |
715 |
|
|
716 |
|
foreach my $out ( @output_modules ) { |
717 |
|
$out->finish if $out->can('finish'); |
718 |
|
} |
719 |
|
|
720 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
721 |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |