7 |
use lib './lib'; |
use lib './lib'; |
8 |
|
|
9 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
10 |
|
use WebPAC::Parser 0.04; |
11 |
use WebPAC::Lookup 0.03; |
use WebPAC::Lookup 0.03; |
12 |
use WebPAC::Input 0.07; |
use WebPAC::Input 0.11; |
13 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.03; |
14 |
use WebPAC::Normalize 0.11; |
use WebPAC::Normalize 0.11; |
15 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
16 |
use WebPAC::Validate; |
use WebPAC::Validate 0.06; |
17 |
use WebPAC::Output::MARC; |
use WebPAC::Output::MARC; |
18 |
use YAML qw/LoadFile/; |
use WebPAC::Config; |
19 |
use Getopt::Long; |
use Getopt::Long; |
20 |
use File::Path; |
use File::Path; |
21 |
use Time::HiRes qw/time/; |
use Time::HiRes qw/time/; |
61 |
|
|
62 |
=item --stats |
=item --stats |
63 |
|
|
64 |
disable indexing and dump statistics about field and subfield |
disable indexing, modify_* in configuration and dump statistics about field |
65 |
usage for each input |
and subfield usage for each input |
66 |
|
|
67 |
=item --validate path/to/validation_file |
=item --validate path/to/validation_file |
68 |
|
|
106 |
my $limit; |
my $limit; |
107 |
|
|
108 |
my $clean = 0; |
my $clean = 0; |
109 |
my $config = 'conf/config.yml'; |
my $config_path; |
110 |
my $debug = 0; |
my $debug = 0; |
111 |
my $only_filter; |
my $only_filter; |
112 |
my $stats = 0; |
my $stats = 0; |
120 |
|
|
121 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
122 |
|
|
|
my $hostname = `hostname`; |
|
|
chomp($hostname); |
|
|
$hostname =~ s/\..+$//; |
|
|
if (-e "conf/$hostname.yml") { |
|
|
$config = "conf/$hostname.yml"; |
|
|
$log->info("using host configuration file: $config"); |
|
|
} |
|
|
|
|
123 |
GetOptions( |
GetOptions( |
124 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
125 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
126 |
"clean" => \$clean, |
"clean" => \$clean, |
127 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
128 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
129 |
"config" => \$config, |
"config" => \$config_path, |
130 |
"debug+" => \$debug, |
"debug+" => \$debug, |
131 |
"stats" => \$stats, |
"stats" => \$stats, |
132 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
139 |
"merge" => \$merge, |
"merge" => \$merge, |
140 |
); |
); |
141 |
|
|
142 |
$config = LoadFile($config); |
my $config = new WebPAC::Config( path => $config_path ); |
143 |
|
|
144 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
145 |
|
|
146 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->databases); |
147 |
|
|
148 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
149 |
|
|
164 |
) if ($validate_path); |
) if ($validate_path); |
165 |
|
|
166 |
|
|
167 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->use_indexer; |
168 |
if ($stats) { |
if ($stats) { |
169 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("option --stats disables update of indexing engine..."); |
170 |
$use_indexer = undef; |
$use_indexer = undef; |
175 |
# disable indexing when creating marc |
# disable indexing when creating marc |
176 |
$use_indexer = undef if ($marc_normalize); |
$use_indexer = undef if ($marc_normalize); |
177 |
|
|
178 |
|
# parse normalize files and create source files for lookup and normalization |
179 |
|
|
180 |
|
my $parser = new WebPAC::Parser( config => $config ); |
181 |
|
|
182 |
my $total_rows = 0; |
my $total_rows = 0; |
183 |
my $start_t = time(); |
my $start_t = time(); |
184 |
|
|
189 |
Proc::Queue::size($parallel); |
Proc::Queue::size($parallel); |
190 |
} |
} |
191 |
|
|
192 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->databases }) { |
193 |
|
|
194 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
195 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
205 |
|
|
206 |
my $indexer; |
my $indexer; |
207 |
if ($use_indexer) { |
if ($use_indexer) { |
208 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
|
209 |
|
my $cfg_name = $use_indexer; |
210 |
|
$cfg_name =~ s/\-.*$//; |
211 |
|
|
212 |
|
my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration"); |
213 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
214 |
$indexer_config->{clean} = $clean; |
$indexer_config->{clean} = $clean; |
215 |
$indexer_config->{label} = $db_config->{name}; |
$indexer_config->{label} = $db_config->{name}; |
223 |
use WebPAC::Output::Estraier '0.10'; |
use WebPAC::Output::Estraier '0.10'; |
224 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
225 |
|
|
226 |
|
} elsif ($use_indexer eq 'hyperestraier-native') { |
227 |
|
|
228 |
|
# open Hyper Estraier database |
229 |
|
use WebPAC::Output::EstraierNative; |
230 |
|
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
231 |
|
|
232 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
233 |
|
|
234 |
# open KinoSearch |
# open KinoSearch |
277 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
278 |
$abs_path =~ s#/[^/]*$#/#; |
$abs_path =~ s#/[^/]*$#/#; |
279 |
|
|
280 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->get('webpac')->{db_path} . '/' . $database; |
281 |
|
|
282 |
if ($clean) { |
if ($clean) { |
283 |
$log->info("creating new database '$database' in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
306 |
$log->info("database $database doesn't have inputs defined"); |
$log->info("database $database doesn't have inputs defined"); |
307 |
} |
} |
308 |
|
|
|
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
|
|
|
|
309 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
310 |
|
|
311 |
next if ($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)); |
312 |
|
|
313 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
314 |
|
|
315 |
die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); |
die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs'))); |
316 |
|
|
317 |
my $lookup; |
my $input_module = $config->webpac('inputs')->{$type}; |
|
if ($input->{lookup}) { |
|
|
$lookup = new WebPAC::Lookup( |
|
|
lookup_file => $input->{lookup}, |
|
|
); |
|
|
delete( $input->{lookup} ); |
|
|
} |
|
|
|
|
|
my $input_module = $config->{webpac}->{inputs}->{$type}; |
|
318 |
|
|
319 |
$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module", |
$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module", |
320 |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
321 |
); |
); |
322 |
|
|
323 |
|
if ($stats) { |
324 |
|
# disable modification of records if --stats is in use |
325 |
|
delete($input->{modify_records}); |
326 |
|
delete($input->{modify_file}); |
327 |
|
} |
328 |
|
|
329 |
|
warn "depends on: ", dump( $parser->depends($database, $input->{name}), $parser->{depends}, $parser->lookup_create_rules($database, $input->{name}), $parser->{_lookup_create} ); |
330 |
|
|
331 |
|
my $lookup; |
332 |
|
|
333 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
334 |
module => $input_module, |
module => $input_module, |
335 |
encoding => $config->{webpac}->{webpac_encoding}, |
encoding => $config->webpac('webpac_encoding'), |
336 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
337 |
offset => $offset, |
offset => $offset, |
338 |
lookup_coderef => sub { |
lookup_coderef => sub { |
342 |
recode => $input->{recode}, |
recode => $input->{recode}, |
343 |
stats => $stats, |
stats => $stats, |
344 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
345 |
|
modify_file => $input->{modify_file}, |
346 |
); |
); |
347 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
348 |
|
|
352 |
%{ $input }, |
%{ $input }, |
353 |
); |
); |
354 |
|
|
355 |
|
my $report_fh; |
356 |
|
if ($stats || $validate) { |
357 |
|
my $path = "out/report/" . $database . '-' . $input->{name} . '.txt'; |
358 |
|
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
359 |
|
|
360 |
|
print $report_fh "Report for database '$database' input '$input->{name}' records ", |
361 |
|
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
362 |
|
$log->info("Generating report file $path"); |
363 |
|
} |
364 |
|
|
365 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
366 |
@{ $input->{normalize} } : ( $input->{normalize} ); |
@{ $input->{normalize} } : ( $input->{normalize} ); |
367 |
|
|
391 |
# reset position in database |
# reset position in database |
392 |
$input_db->seek(1); |
$input_db->seek(1); |
393 |
|
|
394 |
|
# generate name of config key for indexer (strip everything after -) |
395 |
|
my $indexer_config = $use_indexer; |
396 |
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
397 |
|
|
398 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
399 |
|
|
400 |
my $row = $input_db->fetch || next; |
my $row = $input_db->fetch || next; |
409 |
|
|
410 |
|
|
411 |
if ($validate) { |
if ($validate) { |
412 |
my @errors = $validate->validate_errors( $row ); |
if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) { |
413 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
$log->error( "MFN $mfn validation error:\n", |
414 |
|
$validate->report_error( $errors ) |
415 |
|
); |
416 |
|
} |
417 |
} |
} |
418 |
|
|
419 |
my $ds_config = dclone($db_config); |
my $ds_config = dclone($db_config); |
444 |
$indexer->add( |
$indexer->add( |
445 |
id => $input->{name} . "/" . $mfn, |
id => $input->{name} . "/" . $mfn, |
446 |
ds => $ds, |
ds => $ds, |
447 |
type => $config->{$use_indexer}->{type}, |
type => $config->get($indexer_config)->{type}, |
448 |
) if ($indexer && $ds); |
) if ($indexer && $ds); |
449 |
|
|
450 |
if ($marc) { |
if ($marc) { |
466 |
$total_rows++; |
$total_rows++; |
467 |
} |
} |
468 |
|
|
469 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
if ($validate) { |
470 |
|
my $errors = $validate->report; |
471 |
|
if ($errors) { |
472 |
|
$log->info("validation errors:\n$errors\n" ); |
473 |
|
print $report_fh "$errors\n" if ($report_fh); |
474 |
|
} |
475 |
|
} |
476 |
|
|
477 |
|
if ($stats) { |
478 |
|
my $s = $input_db->stats; |
479 |
|
$log->info("statistics of fields usage:\n$s"); |
480 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
481 |
|
} |
482 |
|
|
483 |
# close MARC file |
# close MARC file |
484 |
$marc->finish if ($marc); |
$marc->finish if ($marc); |
485 |
|
|
486 |
|
# close report |
487 |
|
close($report_fh) if ($report_fh) |
488 |
} |
} |
489 |
|
|
490 |
} |
} |