12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.03; |
13 |
use WebPAC::Normalize 0.11; |
use WebPAC::Normalize 0.11; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
15 |
use WebPAC::Validate; |
use WebPAC::Validate 0.06; |
16 |
use WebPAC::Output::MARC; |
use WebPAC::Output::MARC; |
17 |
use YAML qw/LoadFile/; |
use WebPAC::Config; |
18 |
use Getopt::Long; |
use Getopt::Long; |
19 |
use File::Path; |
use File::Path; |
20 |
use Time::HiRes qw/time/; |
use Time::HiRes qw/time/; |
60 |
|
|
61 |
=item --stats |
=item --stats |
62 |
|
|
63 |
disable indexing and dump statistics about field and subfield |
disable indexing, modify_* in configuration and dump statistics about field |
64 |
usage for each input |
and subfield usage for each input |
65 |
|
|
66 |
=item --validate path/to/validation_file |
=item --validate path/to/validation_file |
67 |
|
|
105 |
my $limit; |
my $limit; |
106 |
|
|
107 |
my $clean = 0; |
my $clean = 0; |
108 |
my $config = 'conf/config.yml'; |
my $config_path; |
109 |
my $debug = 0; |
my $debug = 0; |
110 |
my $only_filter; |
my $only_filter; |
111 |
my $stats = 0; |
my $stats = 0; |
119 |
|
|
120 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
121 |
|
|
|
my $hostname = `hostname`; |
|
|
chomp($hostname); |
|
|
$hostname =~ s/\..+$//; |
|
|
if (-e "conf/$hostname.yml") { |
|
|
$config = "conf/$hostname.yml"; |
|
|
$log->info("using host configuration file: $config"); |
|
|
} |
|
|
|
|
122 |
GetOptions( |
GetOptions( |
123 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
124 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
125 |
"clean" => \$clean, |
"clean" => \$clean, |
126 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
127 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
128 |
"config" => \$config, |
"config" => \$config_path, |
129 |
"debug+" => \$debug, |
"debug+" => \$debug, |
130 |
"stats" => \$stats, |
"stats" => \$stats, |
131 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
138 |
"merge" => \$merge, |
"merge" => \$merge, |
139 |
); |
); |
140 |
|
|
141 |
$config = LoadFile($config); |
my $config = new WebPAC::Config( path => $config_path ); |
142 |
|
|
143 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
144 |
|
|
145 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->databases); |
146 |
|
|
147 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
148 |
|
|
163 |
) if ($validate_path); |
) if ($validate_path); |
164 |
|
|
165 |
|
|
166 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->use_indexer; |
167 |
if ($stats) { |
if ($stats) { |
168 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("option --stats disables update of indexing engine..."); |
169 |
$use_indexer = undef; |
$use_indexer = undef; |
184 |
Proc::Queue::size($parallel); |
Proc::Queue::size($parallel); |
185 |
} |
} |
186 |
|
|
187 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->databases }) { |
188 |
|
|
189 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
190 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
200 |
|
|
201 |
my $indexer; |
my $indexer; |
202 |
if ($use_indexer) { |
if ($use_indexer) { |
203 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
|
204 |
|
my $cfg_name = $use_indexer; |
205 |
|
$cfg_name =~ s/\-.*$//; |
206 |
|
|
207 |
|
my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration"); |
208 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
209 |
$indexer_config->{clean} = $clean; |
$indexer_config->{clean} = $clean; |
210 |
$indexer_config->{label} = $db_config->{name}; |
$indexer_config->{label} = $db_config->{name}; |
218 |
use WebPAC::Output::Estraier '0.10'; |
use WebPAC::Output::Estraier '0.10'; |
219 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
220 |
|
|
221 |
|
} elsif ($use_indexer eq 'hyperestraier-native') { |
222 |
|
|
223 |
|
# open Hyper Estraier database |
224 |
|
use WebPAC::Output::EstraierNative; |
225 |
|
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
226 |
|
|
227 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
228 |
|
|
229 |
# open KinoSearch |
# open KinoSearch |
272 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
273 |
$abs_path =~ s#/[^/]*$#/#; |
$abs_path =~ s#/[^/]*$#/#; |
274 |
|
|
275 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->get('webpac')->{db_path} . '/' . $database; |
276 |
|
|
277 |
if ($clean) { |
if ($clean) { |
278 |
$log->info("creating new database '$database' in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
301 |
$log->info("database $database doesn't have inputs defined"); |
$log->info("database $database doesn't have inputs defined"); |
302 |
} |
} |
303 |
|
|
|
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
|
|
|
|
304 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
305 |
|
|
306 |
next if ($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)); |
307 |
|
|
308 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
309 |
|
|
310 |
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'))); |
311 |
|
|
312 |
my $lookup; |
my $lookup; |
313 |
if ($input->{lookup}) { |
if ($input->{lookup}) { |
317 |
delete( $input->{lookup} ); |
delete( $input->{lookup} ); |
318 |
} |
} |
319 |
|
|
320 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->webpac('inputs')->{$type}; |
321 |
|
|
322 |
$log->info("working on input '$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", |
323 |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
324 |
); |
); |
325 |
|
|
326 |
|
if ($stats) { |
327 |
|
# disable modification of records if --stats is in use |
328 |
|
delete($input->{modify_records}); |
329 |
|
delete($input->{modify_file}); |
330 |
|
} |
331 |
|
|
332 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
333 |
module => $input_module, |
module => $input_module, |
334 |
encoding => $config->{webpac}->{webpac_encoding}, |
encoding => $config->webpac('webpac_encoding'), |
335 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
336 |
offset => $offset, |
offset => $offset, |
337 |
lookup_coderef => sub { |
lookup_coderef => sub { |
341 |
recode => $input->{recode}, |
recode => $input->{recode}, |
342 |
stats => $stats, |
stats => $stats, |
343 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
344 |
|
modify_file => $input->{modify_file}, |
345 |
); |
); |
346 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
347 |
|
|
351 |
%{ $input }, |
%{ $input }, |
352 |
); |
); |
353 |
|
|
354 |
|
my $report_fh; |
355 |
|
if ($stats || $validate) { |
356 |
|
my $path = "out/report/" . $database . '-' . $input->{name} . '.txt'; |
357 |
|
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
358 |
|
|
359 |
|
print $report_fh "Report for database '$database' input '$input->{name}' records ", |
360 |
|
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
361 |
|
$log->info("Generating report file $path"); |
362 |
|
} |
363 |
|
|
364 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
365 |
@{ $input->{normalize} } : ( $input->{normalize} ); |
@{ $input->{normalize} } : ( $input->{normalize} ); |
366 |
|
|
390 |
# reset position in database |
# reset position in database |
391 |
$input_db->seek(1); |
$input_db->seek(1); |
392 |
|
|
393 |
|
# generate name of config key for indexer (strip everything after -) |
394 |
|
my $indexer_config = $use_indexer; |
395 |
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
396 |
|
|
397 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
398 |
|
|
399 |
my $row = $input_db->fetch || next; |
my $row = $input_db->fetch || next; |
408 |
|
|
409 |
|
|
410 |
if ($validate) { |
if ($validate) { |
411 |
my @errors = $validate->validate_errors( $row ); |
if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) { |
412 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
$log->error( "MFN $mfn validation error:\n", |
413 |
|
$validate->report_error( $errors ) |
414 |
|
); |
415 |
|
} |
416 |
} |
} |
417 |
|
|
418 |
my $ds_config = dclone($db_config); |
my $ds_config = dclone($db_config); |
443 |
$indexer->add( |
$indexer->add( |
444 |
id => $input->{name} . "/" . $mfn, |
id => $input->{name} . "/" . $mfn, |
445 |
ds => $ds, |
ds => $ds, |
446 |
type => $config->{$use_indexer}->{type}, |
type => $config->get($indexer_config)->{type}, |
447 |
) if ($indexer && $ds); |
) if ($indexer && $ds); |
448 |
|
|
449 |
if ($marc) { |
if ($marc) { |
465 |
$total_rows++; |
$total_rows++; |
466 |
} |
} |
467 |
|
|
468 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
if ($validate) { |
469 |
|
my $errors = $validate->report; |
470 |
|
if ($errors) { |
471 |
|
$log->info("validation errors:\n$errors\n" ); |
472 |
|
print $report_fh "$errors\n" if ($report_fh); |
473 |
|
} |
474 |
|
} |
475 |
|
|
476 |
|
if ($stats) { |
477 |
|
my $s = $input_db->stats; |
478 |
|
$log->info("statistics of fields usage:\n$s"); |
479 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
480 |
|
} |
481 |
|
|
482 |
# close MARC file |
# close MARC file |
483 |
$marc->finish if ($marc); |
$marc->finish if ($marc); |
484 |
|
|
485 |
|
# close report |
486 |
|
close($report_fh) if ($report_fh) |
487 |
} |
} |
488 |
|
|
489 |
} |
} |