7 |
use lib './lib'; |
use lib './lib'; |
8 |
|
|
9 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
10 |
use WebPAC::Lookup 0.03; |
use WebPAC::Parser 0.08; |
11 |
use WebPAC::Input 0.07; |
use WebPAC::Input 0.16; |
12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.15; |
13 |
use WebPAC::Normalize 0.11; |
use WebPAC::Normalize 0.22; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
15 |
use WebPAC::Validate; |
use WebPAC::Validate 0.11; |
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/; |
21 |
use File::Slurp; |
use File::Slurp; |
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/; |
25 |
|
|
26 |
use Proc::Queue size => 1; |
use Proc::Queue size => 1; |
27 |
use POSIX ":sys_wait_h"; # imports WNOHANG |
use POSIX ":sys_wait_h"; # imports WNOHANG |
32 |
|
|
33 |
B<this command will probably go away. Don't get used to it!> |
B<this command will probably go away. Don't get used to it!> |
34 |
|
|
35 |
Options: |
=head1 OPTIONS |
36 |
|
|
37 |
=over 4 |
=over 4 |
38 |
|
|
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 |
|
|
69 |
turn on extra validation of imput records, see L<WebPAC::Validation> |
turn on extra validation of imput records, see L<WebPAC::Validation> |
70 |
|
|
71 |
=item --marc-normalize conf/normalize/mapping.pl |
You can use special variables C<$database> and $C<$input> in this parametar |
72 |
|
like C<--validate 'conf/validate/$database-$input'> to construct filename |
73 |
|
|
74 |
This option specifies normalisation file for MARC creation |
=item --validate-delimiters path/to/validate_delimiters_file |
75 |
|
|
76 |
=item --marc-output out/marc/test.marc |
this option is used with C<--validate> to turn on extra validation of |
77 |
|
delimiters. If file is non existant, it will be created on first run. |
78 |
|
|
79 |
Optional path to output file |
=item --marc-generate |
80 |
|
|
81 |
|
Generate MARC file. This will automatically be on if file contains C<marc*> directives. |
82 |
|
You can use this option as C<--no-marc-generate> to disable MARC generation. |
83 |
|
|
84 |
=item --marc-lint |
=item --marc-lint |
85 |
|
|
86 |
By default turned on if C<--marc-normalize> is used. You can disable lint |
By default turned on if normalisation file has C<marc*> directives. You can disable lint |
87 |
messages with C<--no-marc-lint>. |
messages with C<--no-marc-lint>. |
88 |
|
|
89 |
=item --marc-dump |
=item --marc-dump |
111 |
my $limit; |
my $limit; |
112 |
|
|
113 |
my $clean = 0; |
my $clean = 0; |
114 |
my $config = 'conf/config.yml'; |
my $config_path; |
115 |
my $debug = 0; |
my $debug = 0; |
116 |
my $only_filter; |
my $only_filter; |
117 |
my $stats = 0; |
my $stats = 0; |
118 |
my $validate_path; |
my $validate_path; |
119 |
my ($marc_normalize, $marc_output); |
my $validate_delimiters_path; |
120 |
|
my $marc_generate = 1; |
121 |
my $marc_lint = 1; |
my $marc_lint = 1; |
122 |
my $marc_dump = 0; |
my $marc_dump = 0; |
123 |
my $parallel = 0; |
my $parallel = 0; |
124 |
my $only_links = 0; |
my $only_links = 0; |
125 |
my $merge = 0; |
my $merge = 0; |
126 |
|
my $help; |
127 |
|
|
128 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
129 |
|
|
|
my $hostname = `hostname`; |
|
|
chomp($hostname); |
|
|
$hostname =~ s/\..+$//; |
|
|
if (-e "conf/$hostname.yml") { |
|
|
$config = "conf/$hostname.yml"; |
|
|
$log->info("using host configuration file: $config"); |
|
|
} |
|
|
|
|
130 |
GetOptions( |
GetOptions( |
131 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
132 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
133 |
"clean" => \$clean, |
"clean" => \$clean, |
134 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
135 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
136 |
"config" => \$config, |
"config" => \$config_path, |
137 |
"debug+" => \$debug, |
"debug+" => \$debug, |
138 |
"stats" => \$stats, |
"stats" => \$stats, |
139 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
140 |
"marc-normalize=s" => \$marc_normalize, |
"validate-delimiters=s" => \$validate_delimiters_path, |
141 |
"marc-output=s" => \$marc_output, |
"marc-generate!" => \$marc_generate, |
142 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
143 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
144 |
"parallel=i" => \$parallel, |
"parallel=i" => \$parallel, |
145 |
"only-links!" => \$only_links, |
"only-links!" => \$only_links, |
146 |
"merge" => \$merge, |
"merge" => \$merge, |
147 |
|
"help" => \$help, |
148 |
); |
); |
149 |
|
|
150 |
$config = LoadFile($config); |
$marc_generate = 0 if ( $validate_delimiters_path ); |
151 |
|
|
152 |
|
pod2usage(-verbose => 2) if ($help); |
153 |
|
|
154 |
|
my $config = new WebPAC::Config( path => $config_path ); |
155 |
|
|
156 |
#print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
157 |
|
|
158 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->databases); |
159 |
|
|
160 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
161 |
|
|
162 |
|
my $log_file = 'log'; |
163 |
|
|
164 |
|
if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) { |
165 |
|
$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'"); |
166 |
|
rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!"); |
167 |
|
} |
168 |
|
|
169 |
my $estcmd_fh; |
my $estcmd_fh; |
170 |
my $estcmd_path = './estcmd-merge.sh'; |
my $estcmd_path = './estcmd-merge.sh'; |
175 |
$log->info("created merge batch file $estcmd_path"); |
$log->info("created merge batch file $estcmd_path"); |
176 |
} |
} |
177 |
|
|
|
|
|
178 |
my $validate; |
my $validate; |
179 |
$validate = new WebPAC::Validate( |
$validate = new WebPAC::Validate( |
180 |
path => $validate_path, |
delimiters => $config->webpac('delimiters'), |
181 |
) if ($validate_path); |
) if ($validate_path || $validate_delimiters_path); |
|
|
|
182 |
|
|
183 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->use_indexer; |
184 |
|
$stats ||= $validate; |
185 |
if ($stats) { |
if ($stats) { |
186 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("disabled indexing for stats collection"); |
187 |
$use_indexer = undef; |
$use_indexer = undef; |
188 |
} else { |
} elsif ( $use_indexer ) { |
189 |
$log->info("using $use_indexer indexing engine..."); |
$log->info("using $use_indexer indexing engine..."); |
190 |
} |
} |
191 |
|
|
192 |
# disable indexing when creating marc |
# parse normalize files and create source files for lookup and normalization |
193 |
$use_indexer = undef if ($marc_normalize); |
|
194 |
|
my $parser = new WebPAC::Parser( config => $config ); |
195 |
|
|
196 |
my $total_rows = 0; |
my $total_rows = 0; |
197 |
my $start_t = time(); |
my $start_t = time(); |
203 |
Proc::Queue::size($parallel); |
Proc::Queue::size($parallel); |
204 |
} |
} |
205 |
|
|
206 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
sub create_ds_config { |
207 |
|
my ($db_config, $database, $input, $mfn) = @_; |
208 |
|
my $c = dclone( $db_config ); |
209 |
|
$c->{_} = $database || $log->logconfess("need database"); |
210 |
|
$c->{_mfn} = $mfn || $log->logconfess("need mfn"); |
211 |
|
$c->{input} = $input || $log->logconfess("need input"); |
212 |
|
return $c; |
213 |
|
} |
214 |
|
|
215 |
|
foreach my $database ( sort keys %{ $config->databases } ) { |
216 |
|
my $db_config = $config->databases->{$database}; |
217 |
|
|
218 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
219 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
228 |
} |
} |
229 |
|
|
230 |
my $indexer; |
my $indexer; |
231 |
if ($use_indexer) { |
if ($use_indexer && $parser->have_rules( 'search', $database )) { |
232 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
|
233 |
|
my $cfg_name = $use_indexer; |
234 |
|
$cfg_name =~ s/\-.*$//; |
235 |
|
|
236 |
|
my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration"); |
237 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
238 |
$indexer_config->{clean} = $clean; |
$indexer_config->{clean} = $clean; |
239 |
$indexer_config->{label} = $db_config->{name}; |
$indexer_config->{label} = $db_config->{name}; |
244 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer eq 'hyperestraier') { |
245 |
|
|
246 |
# open Hyper Estraier database |
# open Hyper Estraier database |
247 |
use WebPAC::Output::Estraier '0.10'; |
require WebPAC::Output::Estraier; |
248 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
249 |
|
|
250 |
|
} elsif ($use_indexer eq 'hyperestraier-native') { |
251 |
|
|
252 |
|
# open Hyper Estraier database |
253 |
|
require WebPAC::Output::EstraierNative; |
254 |
|
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
255 |
|
|
256 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
257 |
|
|
258 |
# 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 } ); |
|
259 |
|
|
260 |
} else { |
} else { |
261 |
$log->logdie("unknown use_indexer: $use_indexer"); |
$log->logdie("unknown use_indexer: $use_indexer"); |
262 |
} |
} |
263 |
|
|
264 |
$log->logide("can't continue without valid indexer") unless ($indexer); |
$log->logdie("can't continue without valid indexer") unless ($indexer); |
265 |
} |
} |
266 |
|
|
267 |
|
|
295 |
# |
# |
296 |
# now WebPAC::Store |
# now WebPAC::Store |
297 |
# |
# |
298 |
my $abs_path = abs_path($0); |
my $store = new WebPAC::Store({ |
299 |
$abs_path =~ s#/[^/]*$#/#; |
debug => $debug, |
300 |
|
}); |
|
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
|
301 |
|
|
302 |
if ($clean) { |
sub iterate_over { |
303 |
$log->info("creating new database '$database' in $db_path"); |
my ( $what, $error ) = @_; |
304 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
my @result; |
305 |
} else { |
if ( ref( $what ) eq 'ARRAY' ) { |
306 |
$log->info("working on database '$database' in $db_path"); |
@result = @{ $what }; |
307 |
|
} elsif ($db_config->{input}) { |
308 |
|
@result = ( $what ); |
309 |
|
} else { |
310 |
|
$error->() if ref($error) eq 'CODE'; |
311 |
|
} |
312 |
|
return @result; |
313 |
} |
} |
314 |
|
|
315 |
my $db = new WebPAC::Store( |
# |
316 |
path => $db_path, |
# prepare output |
317 |
database => $database, |
# |
318 |
debug => $debug, |
my @outputs = iterate_over( $db_config->{output}, sub { |
319 |
); |
$log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" ); |
320 |
|
} ); |
321 |
|
|
322 |
|
my @output_modules; |
323 |
|
|
324 |
|
foreach my $output ( @outputs ) { |
325 |
|
|
326 |
|
warn '## output = ',dump( $output ); |
327 |
|
|
328 |
|
my $module = $output->{module} || $log->logdie("need module in output section of $database"); |
329 |
|
$module = 'WebPAC::Output::' . $module unless $module =~ m/::/; |
330 |
|
|
331 |
|
$log->debug("loading output module $module"); |
332 |
|
eval "require $module"; |
333 |
|
|
334 |
|
$output->{database} = $database; |
335 |
|
|
336 |
|
$log->debug("calling $module->new(",dump( $output ),")"); |
337 |
|
my $out = new $module->new( $output ); |
338 |
|
$out->init; |
339 |
|
|
340 |
|
push @output_modules, $out; |
341 |
|
} |
342 |
|
|
343 |
|
|
344 |
# |
# |
345 |
# now, iterate through input formats |
# now, iterate through input formats |
346 |
# |
# |
347 |
|
|
|
my @inputs; |
|
|
if (ref($db_config->{input}) eq 'ARRAY') { |
|
|
@inputs = @{ $db_config->{input} }; |
|
|
} elsif ($db_config->{input}) { |
|
|
push @inputs, $db_config->{input}; |
|
|
} else { |
|
|
$log->info("database $database doesn't have inputs defined"); |
|
|
} |
|
348 |
|
|
349 |
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
my @inputs = iterate_over( $db_config->{input}, sub { |
350 |
|
$log->info("database $database doesn't have inputs defined"); |
351 |
|
} ); |
352 |
|
|
353 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
354 |
|
|
355 |
next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input)); |
356 |
|
|
357 |
|
next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
358 |
|
|
359 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
360 |
|
|
361 |
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'))); |
362 |
|
|
363 |
my $lookup; |
my $input_module = $config->webpac('inputs')->{$type}; |
|
if ($input->{lookup}) { |
|
|
$lookup = new WebPAC::Lookup( |
|
|
lookup_file => $input->{lookup}, |
|
|
); |
|
|
delete( $input->{lookup} ); |
|
|
} |
|
364 |
|
|
365 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my @lookups = $parser->have_lookup_create($database, $input); |
366 |
|
|
367 |
$log->info("working on input '$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", |
368 |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
369 |
); |
); |
370 |
|
|
371 |
|
if ($stats) { |
372 |
|
# disable modification of records if --stats is in use |
373 |
|
delete($input->{modify_records}); |
374 |
|
delete($input->{modify_file}); |
375 |
|
} |
376 |
|
|
377 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
378 |
module => $input_module, |
module => $input_module, |
379 |
encoding => $config->{webpac}->{webpac_encoding}, |
encoding => $config->webpac('webpac_encoding'), |
380 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
381 |
offset => $offset, |
offset => $offset, |
|
lookup_coderef => sub { |
|
|
my $rec = shift || return; |
|
|
$lookup->add( $rec ); |
|
|
}, |
|
382 |
recode => $input->{recode}, |
recode => $input->{recode}, |
383 |
stats => $stats, |
stats => $stats, |
384 |
modify_records => $input->{modify_records}, |
modify_records => $input->{modify_records}, |
385 |
|
modify_file => $input->{modify_file}, |
386 |
|
input_config => $input, |
387 |
); |
); |
388 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
389 |
|
|
390 |
|
if (defined( $input->{lookup} )) { |
391 |
|
$log->warn("$database/$input_name has depriciated lookup definition, removing it..."); |
392 |
|
delete( $input->{lookup} ); |
393 |
|
} |
394 |
|
|
395 |
|
my $lookup_coderef; |
396 |
|
|
397 |
|
if (@lookups) { |
398 |
|
|
399 |
|
my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input"); |
400 |
|
|
401 |
|
$lookup_coderef = sub { |
402 |
|
my $rec = shift || die "need rec!"; |
403 |
|
my $mfn = $rec->{'000'}->[0] || die "need mfn in 000"; |
404 |
|
|
405 |
|
WebPAC::Normalize::data_structure( |
406 |
|
row => $rec, |
407 |
|
rules => $rules, |
408 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
409 |
|
); |
410 |
|
|
411 |
|
#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup()); |
412 |
|
}; |
413 |
|
|
414 |
|
WebPAC::Normalize::_set_lookup( undef ); |
415 |
|
|
416 |
|
$log->debug("created lookup_coderef using:\n$rules"); |
417 |
|
|
418 |
|
}; |
419 |
|
|
420 |
|
my $lookup_jar; |
421 |
|
|
422 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
423 |
path => $input->{path}, |
path => $input->{path}, |
424 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
425 |
|
lookup_coderef => $lookup_coderef, |
426 |
|
lookup => $lookup_jar, |
427 |
%{ $input }, |
%{ $input }, |
428 |
|
load_row => sub { |
429 |
|
my $a = shift; |
430 |
|
return $store->load_row( |
431 |
|
database => $database, |
432 |
|
input => $input_name, |
433 |
|
id => $a->{id}, |
434 |
|
); |
435 |
|
}, |
436 |
|
save_row => sub { |
437 |
|
my $a = shift; |
438 |
|
return $store->save_row( |
439 |
|
database => $database, |
440 |
|
input => $input_name, |
441 |
|
id => $a->{id}, |
442 |
|
row => $a->{row}, |
443 |
|
); |
444 |
|
}, |
445 |
|
|
446 |
); |
); |
447 |
|
|
448 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
|
@{ $input->{normalize} } : ( $input->{normalize} ); |
|
449 |
|
|
450 |
if ($marc_normalize) { |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
451 |
@norm_array = ( { |
$log->debug("created following lookups: ", sub { dump( $lookup_data ) } ); |
452 |
path => $marc_normalize, |
|
453 |
output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc', |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
454 |
} ); |
$store->save_lookup( |
455 |
|
database => $database, |
456 |
|
input => $input_name, |
457 |
|
key => $key, |
458 |
|
data => $lookup_data->{$database}->{$input_name}->{$key}, |
459 |
|
); |
460 |
|
} |
461 |
} |
} |
462 |
|
|
463 |
foreach my $normalize (@norm_array) { |
my $report_fh; |
464 |
|
if ($stats || $validate) { |
465 |
|
my $path = "out/report/${database}-${input_name}.txt"; |
466 |
|
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
467 |
|
|
468 |
|
print $report_fh "Report for database '$database' input '$input_name' records ", |
469 |
|
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
470 |
|
$log->info("Generating report file $path"); |
471 |
|
|
472 |
|
if ( $validate ) { |
473 |
|
$validate->read_validate_file( $validate->fill_in( $validate_path, database => $database, input => $input_name ) ) if ( $validate_path ); |
474 |
|
$validate->read_validate_delimiters_file( $validate->fill_in( $validate_delimiters_path, database => $database, input => $input_name ) ) if ( $validate_delimiters_path ); |
475 |
|
} |
476 |
|
} |
477 |
|
|
478 |
my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); |
my $marc; |
479 |
|
if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { |
480 |
|
$marc = new WebPAC::Output::MARC( |
481 |
|
path => "out/marc/${database}-${input_name}.marc", |
482 |
|
lint => $marc_lint, |
483 |
|
dump => $marc_dump, |
484 |
|
); |
485 |
|
} |
486 |
|
|
487 |
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
488 |
|
$log->debug("parsed normalize rules:\n$rules"); |
489 |
|
|
490 |
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
# reset position in database |
491 |
|
$input_db->seek(1); |
492 |
|
|
493 |
$log->info("Using $normalize_path for normalization..."); |
# generate name of config key for indexer (strip everything after -) |
494 |
|
my $indexer_config = $use_indexer; |
495 |
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
496 |
|
|
497 |
|
my $lookup_hash; |
498 |
|
my $depends = $parser->depends($database,$input_name); |
499 |
|
|
500 |
|
if ($depends) { |
501 |
|
$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends); |
502 |
|
$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH'); |
503 |
|
|
504 |
|
foreach my $db (keys %$depends) { |
505 |
|
foreach my $i (keys %{$depends->{$db}}) { |
506 |
|
foreach my $k (keys %{$depends->{$db}->{$i}}) { |
507 |
|
my $t = time(); |
508 |
|
$log->debug("loading lookup $db/$i"); |
509 |
|
$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup( |
510 |
|
database => $db, |
511 |
|
input => $i, |
512 |
|
key => $k, |
513 |
|
); |
514 |
|
$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t)); |
515 |
|
} |
516 |
|
} |
517 |
|
} |
518 |
|
|
519 |
my $marc = new WebPAC::Output::MARC( |
$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) }); |
520 |
path => $normalize->{output}, |
} |
|
lint => $marc_lint, |
|
|
dump => $marc_dump, |
|
|
) if ($normalize->{output}); |
|
521 |
|
|
|
# reset position in database |
|
|
$input_db->seek(1); |
|
522 |
|
|
523 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
524 |
|
|
525 |
my $row = $input_db->fetch || next; |
my $row = $input_db->fetch || next; |
526 |
|
|
527 |
my $mfn = $row->{'000'}->[0]; |
$total_rows++; |
528 |
|
|
529 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
my $mfn = $row->{'000'}->[0]; |
530 |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
|
531 |
$mfn = $pos; |
if (! $mfn || $mfn !~ m{^\d+$}) { |
532 |
push @{ $row->{'000'} }, $pos; |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
533 |
} |
$mfn = $pos; |
534 |
|
push @{ $row->{'000'} }, $pos; |
535 |
|
} |
536 |
|
|
537 |
|
|
538 |
if ($validate) { |
if ($validate) { |
539 |
my @errors = $validate->validate_errors( $row ); |
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
540 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
$log->error( "MFN $mfn validation error:\n", |
541 |
|
$validate->report_error( $errors ) |
542 |
|
); |
543 |
} |
} |
544 |
|
next; # validation doesn't create any output |
545 |
|
} |
546 |
|
|
547 |
my $ds_config = dclone($db_config); |
my $ds = WebPAC::Normalize::data_structure( |
548 |
|
row => $row, |
549 |
# default values -> database key |
rules => $rules, |
550 |
$ds_config->{_} = $database; |
lookup => $lookup_hash, |
551 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
552 |
|
marc_encoding => 'utf-8', |
553 |
|
load_row_coderef => sub { |
554 |
|
my ($database,$input,$mfn) = @_; |
555 |
|
return $store->load_row( |
556 |
|
database => $database, |
557 |
|
input => $input, |
558 |
|
id => $mfn, |
559 |
|
); |
560 |
|
}, |
561 |
|
); |
562 |
|
|
563 |
# current mfn |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
|
$ds_config->{_mfn} = $mfn; |
|
564 |
|
|
565 |
# attach current input |
$store->save_ds( |
566 |
$ds_config->{input} = $input; |
database => $database, |
567 |
|
input => $input_name, |
568 |
|
id => $mfn, |
569 |
|
ds => $ds, |
570 |
|
) if ($ds && !$stats); |
571 |
|
|
572 |
|
$indexer->add( |
573 |
|
id => "${input_name}/${mfn}", |
574 |
|
ds => $ds, |
575 |
|
type => $config->get($indexer_config)->{type}, |
576 |
|
) if ($indexer && $ds); |
577 |
|
|
578 |
|
if ($marc) { |
579 |
|
my $i = 0; |
580 |
|
|
581 |
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
582 |
|
$marc->add( |
583 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
584 |
|
fields => $fields, |
585 |
|
leader => WebPAC::Normalize::_get_marc_leader(), |
586 |
|
row => $row, |
587 |
|
); |
588 |
|
$i++; |
589 |
|
} |
590 |
|
|
591 |
my $ds = WebPAC::Normalize::data_structure( |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
592 |
row => $row, |
} |
|
rules => $rules, |
|
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
|
|
config => $ds_config, |
|
|
marc_encoding => 'utf-8', |
|
|
); |
|
593 |
|
|
594 |
$db->save_ds( |
foreach my $out ( @output_modules ) { |
595 |
id => $mfn, |
$out->add( $mfn, $ds ) if $out->can('add'); |
596 |
ds => $ds, |
} |
|
prefix => $input->{name}, |
|
|
) if ($ds && !$stats); |
|
|
|
|
|
$indexer->add( |
|
|
id => $input->{name} . "/" . $mfn, |
|
|
ds => $ds, |
|
|
type => $config->{$use_indexer}->{type}, |
|
|
) if ($indexer && $ds); |
|
|
|
|
|
if ($marc) { |
|
|
my $i = 0; |
|
|
|
|
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
|
|
$marc->add( |
|
|
id => $mfn . ( $i ? "/$i" : '' ), |
|
|
fields => $fields, |
|
|
leader => WebPAC::Normalize::marc_leader(), |
|
|
row => $row, |
|
|
); |
|
|
$i++; |
|
|
} |
|
597 |
|
|
598 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
} |
|
} |
|
599 |
|
|
600 |
$total_rows++; |
if ($validate) { |
601 |
|
my $errors = $validate->report; |
602 |
|
if ($errors) { |
603 |
|
$log->info("validation errors:\n$errors\n" ); |
604 |
|
print $report_fh "$errors\n" if ($report_fh); |
605 |
} |
} |
606 |
|
|
607 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n"; |
608 |
|
|
609 |
# close MARC file |
# must be last thing that touches $validate for this input |
610 |
$marc->finish if ($marc); |
$validate->reset; |
611 |
|
} |
612 |
|
|
613 |
|
if ($stats) { |
614 |
|
my $s = $input_db->stats; |
615 |
|
$log->info("statistics of fields usage:\n$s"); |
616 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
617 |
} |
} |
618 |
|
|
619 |
|
# close MARC file |
620 |
|
$marc->finish if ($marc); |
621 |
|
|
622 |
|
# close report |
623 |
|
close($report_fh) if ($report_fh); |
624 |
} |
} |
625 |
|
|
626 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
627 |
|
|
628 |
|
foreach my $out ( @output_modules ) { |
629 |
|
$out->finish if $out->can('finish'); |
630 |
|
} |
631 |
|
|
632 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
633 |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
634 |
sprintf("in %.2f sec [%.2f rec/sec]", |
sprintf("in %.2f sec [%.2f rec/sec]", |
651 |
$log->info("all parallel processes finished"); |
$log->info("all parallel processes finished"); |
652 |
} |
} |
653 |
|
|
654 |
|
# save new delimiters if needed |
655 |
|
$validate->save_delimiters_templates if ( $validate_delimiters_path ); |
656 |
|
|
657 |
# |
# |
658 |
# handle links or merge after indexing |
# handle links or merge after indexing |
659 |
# |
# |