4 |
|
|
5 |
use Cwd qw/abs_path/; |
use Cwd qw/abs_path/; |
6 |
use File::Temp qw/tempdir/; |
use File::Temp qw/tempdir/; |
|
use Data::Dumper; |
|
7 |
use lib './lib'; |
use lib './lib'; |
8 |
|
|
9 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
10 |
use WebPAC::Lookup; |
use WebPAC::Parser 0.08; |
11 |
use WebPAC::Input 0.03; |
use WebPAC::Input 0.16; |
12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.14; |
13 |
use WebPAC::Normalize::XML; |
use WebPAC::Normalize 0.22; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
15 |
use WebPAC::Output::Estraier 0.08; |
use WebPAC::Validate 0.06; |
16 |
use YAML qw/LoadFile/; |
use WebPAC::Output::MARC; |
17 |
|
use WebPAC::Config; |
18 |
use Getopt::Long; |
use Getopt::Long; |
19 |
use File::Path; |
use File::Path; |
20 |
|
use Time::HiRes qw/time/; |
21 |
|
use File::Slurp; |
22 |
|
use Data::Dump qw/dump/; |
23 |
|
use Storable qw/dclone/; |
24 |
|
|
25 |
|
use Proc::Queue size => 1; |
26 |
|
use POSIX ":sys_wait_h"; # imports WNOHANG |
27 |
|
|
28 |
=head1 NAME |
=head1 NAME |
29 |
|
|
47 |
|
|
48 |
remove database and Hyper Estraier index before indexing |
remove database and Hyper Estraier index before indexing |
49 |
|
|
50 |
|
=item --only=database_name/input_filter |
51 |
|
|
52 |
|
reindex just single database (legacy name is --one) |
53 |
|
|
54 |
|
C</input_filter> is optional part which can be C<name> |
55 |
|
or C<type> from input |
56 |
|
|
57 |
=item --config conf/config.yml |
=item --config conf/config.yml |
58 |
|
|
59 |
path to YAML configuration file |
path to YAML configuration file |
60 |
|
|
61 |
|
=item --stats |
62 |
|
|
63 |
|
disable indexing, modify_* in configuration and dump statistics about field |
64 |
|
and subfield usage for each input |
65 |
|
|
66 |
|
=item --validate path/to/validation_file |
67 |
|
|
68 |
|
turn on extra validation of imput records, see L<WebPAC::Validation> |
69 |
|
|
70 |
|
=item --marc-lint |
71 |
|
|
72 |
|
By default turned on if normalisation file has C<marc*> directives. You can disable lint |
73 |
|
messages with C<--no-marc-lint>. |
74 |
|
|
75 |
|
=item --marc-dump |
76 |
|
|
77 |
|
Force dump or input and marc record for debugging. |
78 |
|
|
79 |
|
=item --parallel 4 |
80 |
|
|
81 |
|
Run databases in parallel (aproximatly same as number of processors in |
82 |
|
machine if you want to use full load) |
83 |
|
|
84 |
|
=item --only-links |
85 |
|
|
86 |
|
Create just links |
87 |
|
|
88 |
|
=item --merge |
89 |
|
|
90 |
|
Create merged index of databases which have links |
91 |
|
|
92 |
=back |
=back |
93 |
|
|
94 |
=cut |
=cut |
97 |
my $limit; |
my $limit; |
98 |
|
|
99 |
my $clean = 0; |
my $clean = 0; |
100 |
my $config = 'conf/config.yml'; |
my $config_path; |
101 |
my $debug = 0; |
my $debug = 0; |
102 |
|
my $only_filter; |
103 |
|
my $stats = 0; |
104 |
|
my $validate_path; |
105 |
|
my $marc_lint = 1; |
106 |
|
my $marc_dump = 0; |
107 |
|
my $parallel = 0; |
108 |
|
my $only_links = 0; |
109 |
|
my $merge = 0; |
110 |
|
|
111 |
|
my $log = _new WebPAC::Common()->_get_logger(); |
112 |
|
|
113 |
GetOptions( |
GetOptions( |
114 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
115 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
116 |
"clean" => \$clean, |
"clean" => \$clean, |
117 |
"config" => \$config, |
"one=s" => \$only_filter, |
118 |
"debug" => \$debug, |
"only=s" => \$only_filter, |
119 |
|
"config" => \$config_path, |
120 |
|
"debug+" => \$debug, |
121 |
|
"stats" => \$stats, |
122 |
|
"validate=s" => \$validate_path, |
123 |
|
"marc-lint!" => \$marc_lint, |
124 |
|
"marc-dump!" => \$marc_dump, |
125 |
|
"parallel=i" => \$parallel, |
126 |
|
"only-links!" => \$only_links, |
127 |
|
"merge" => \$merge, |
128 |
); |
); |
129 |
|
|
130 |
$config = LoadFile($config); |
my $config = new WebPAC::Config( path => $config_path ); |
131 |
|
|
132 |
|
#print "config = ",dump($config) if ($debug); |
133 |
|
|
134 |
|
die "no databases in config file!\n" unless ($config->databases); |
135 |
|
|
136 |
|
$log->info( "-" x 79 ); |
137 |
|
|
138 |
|
my $log_file = 'log'; |
139 |
|
|
140 |
|
if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) { |
141 |
|
$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'"); |
142 |
|
rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!"); |
143 |
|
} |
144 |
|
|
145 |
|
my $estcmd_fh; |
146 |
|
my $estcmd_path = './estcmd-merge.sh'; |
147 |
|
if ($merge) { |
148 |
|
open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!"); |
149 |
|
print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/; |
150 |
|
print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/; |
151 |
|
$log->info("created merge batch file $estcmd_path"); |
152 |
|
} |
153 |
|
|
154 |
|
|
155 |
|
my $validate; |
156 |
|
$validate = new WebPAC::Validate( |
157 |
|
path => $validate_path, |
158 |
|
) if ($validate_path); |
159 |
|
|
160 |
|
|
161 |
|
my $use_indexer = $config->use_indexer; |
162 |
|
$stats ||= $validate; |
163 |
|
if ($stats) { |
164 |
|
$log->debug("disabled indexing for stats collection"); |
165 |
|
$use_indexer = undef; |
166 |
|
} else { |
167 |
|
$log->info("using $use_indexer indexing engine..."); |
168 |
|
} |
169 |
|
|
170 |
print "config = ",Dumper($config) if ($debug); |
# parse normalize files and create source files for lookup and normalization |
171 |
|
|
172 |
die "no databases in config file!\n" unless ($config->{databases}); |
my $parser = new WebPAC::Parser( config => $config ); |
173 |
|
|
174 |
my $total_rows = 0; |
my $total_rows = 0; |
175 |
|
my $start_t = time(); |
176 |
|
|
177 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
my @links; |
178 |
|
|
179 |
|
if ($parallel) { |
180 |
|
$log->info("Using $parallel processes for speedup"); |
181 |
|
Proc::Queue::size($parallel); |
182 |
|
} |
183 |
|
|
184 |
|
sub create_ds_config { |
185 |
|
my ($db_config, $database, $input, $mfn) = @_; |
186 |
|
my $c = dclone( $db_config ); |
187 |
|
$c->{_} = $database || $log->logconfess("need database"); |
188 |
|
$c->{_mfn} = $mfn || $log->logconfess("need mfn"); |
189 |
|
$c->{input} = $input || $log->logconfess("need input"); |
190 |
|
return $c; |
191 |
|
} |
192 |
|
|
193 |
|
while (my ($database, $db_config) = each %{ $config->databases }) { |
194 |
|
|
195 |
|
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
196 |
|
next if ($only_database && $database !~ m/$only_database/i); |
197 |
|
|
198 |
|
if ($parallel) { |
199 |
|
my $f=fork; |
200 |
|
if(defined ($f) and $f==0) { |
201 |
|
$log->info("Created processes $$ for speedup"); |
202 |
|
} else { |
203 |
|
next; |
204 |
|
} |
205 |
|
} |
206 |
|
|
207 |
|
my $indexer; |
208 |
|
if ($use_indexer && $parser->have_rules( 'search', $database )) { |
209 |
|
|
210 |
|
my $cfg_name = $use_indexer; |
211 |
|
$cfg_name =~ s/\-.*$//; |
212 |
|
|
213 |
|
my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration"); |
214 |
|
$indexer_config->{database} = $database; |
215 |
|
$indexer_config->{clean} = $clean; |
216 |
|
$indexer_config->{label} = $db_config->{name}; |
217 |
|
|
218 |
|
# force clean if database has links |
219 |
|
$indexer_config->{clean} = 1 if ($db_config->{links}); |
220 |
|
|
221 |
|
if ($use_indexer eq 'hyperestraier') { |
222 |
|
|
223 |
|
# open Hyper Estraier database |
224 |
|
use WebPAC::Output::Estraier '0.10'; |
225 |
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
226 |
|
|
227 |
|
} elsif ($use_indexer eq 'hyperestraier-native') { |
228 |
|
|
229 |
|
# open Hyper Estraier database |
230 |
|
use WebPAC::Output::EstraierNative; |
231 |
|
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
232 |
|
|
233 |
|
} elsif ($use_indexer eq 'kinosearch') { |
234 |
|
|
235 |
|
# open KinoSearch |
236 |
|
use WebPAC::Output::KinoSearch; |
237 |
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
238 |
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
239 |
|
|
240 |
|
} else { |
241 |
|
$log->logdie("unknown use_indexer: $use_indexer"); |
242 |
|
} |
243 |
|
|
244 |
|
$log->logide("can't continue without valid indexer") unless ($indexer); |
245 |
|
} |
246 |
|
|
|
my $log = _new WebPAC::Common()->_get_logger(); |
|
247 |
|
|
248 |
# |
# |
249 |
# open Hyper Estraier database |
# store Hyper Estraier links to other databases |
250 |
# |
# |
251 |
|
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
252 |
|
foreach my $link (@{ $db_config->{links} }) { |
253 |
|
if ($use_indexer eq 'hyperestraier') { |
254 |
|
if ($merge) { |
255 |
|
print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/; |
256 |
|
} else { |
257 |
|
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
258 |
|
push @links, sub { |
259 |
|
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
260 |
|
$indexer->add_link( |
261 |
|
from => $database, |
262 |
|
to => $link->{to}, |
263 |
|
credit => $link->{credit}, |
264 |
|
); |
265 |
|
}; |
266 |
|
} |
267 |
|
} else { |
268 |
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
269 |
|
} |
270 |
|
} |
271 |
|
} |
272 |
|
next if ($only_links); |
273 |
|
|
|
my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); |
|
|
$est_config->{database} = $database; |
|
|
$est_config->{clean} = $clean; |
|
|
|
|
|
my $est = new WebPAC::Output::Estraier( %{ $est_config } ); |
|
274 |
|
|
275 |
# |
# |
276 |
# now WebPAC::Store |
# now WebPAC::Store |
278 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
279 |
$abs_path =~ s#/[^/]*$#/#; |
$abs_path =~ s#/[^/]*$#/#; |
280 |
|
|
281 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->webpac('db_path'); |
282 |
|
|
283 |
if ($clean) { |
if ($clean) { |
284 |
$log->info("creating new database $database in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
285 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
286 |
} else { |
} else { |
287 |
$log->info("working on $database in $db_path"); |
$log->info("working on database '$database' in $db_path"); |
288 |
} |
} |
289 |
|
|
290 |
my $db = new WebPAC::Store( |
my $store = new WebPAC::Store( |
291 |
path => $db_path, |
path => $db_path, |
|
database => $database, |
|
292 |
debug => $debug, |
debug => $debug, |
293 |
); |
); |
294 |
|
|
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 |
|
my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input)); |
312 |
|
|
313 |
|
next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
314 |
|
|
315 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
316 |
|
|
317 |
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'))); |
318 |
|
|
319 |
my $lookup = new WebPAC::Lookup( |
my $input_module = $config->webpac('inputs')->{$type}; |
|
lookup_file => $input->{lookup}, |
|
|
); |
|
320 |
|
|
321 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my @lookups = $parser->have_lookup_create($database, $input); |
322 |
|
|
323 |
$log->info("working on input $input->{path} [$input->{type}] using $input_module"); |
$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module", |
324 |
|
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
325 |
|
); |
326 |
|
|
327 |
|
if ($stats) { |
328 |
|
# disable modification of records if --stats is in use |
329 |
|
delete($input->{modify_records}); |
330 |
|
delete($input->{modify_file}); |
331 |
|
} |
332 |
|
|
333 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
334 |
module => $input_module, |
module => $input_module, |
335 |
code_page => $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 => $lookup, |
recode => $input->{recode}, |
339 |
|
stats => $stats, |
340 |
|
modify_records => $input->{modify_records}, |
341 |
|
modify_file => $input->{modify_file}, |
342 |
); |
); |
343 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
344 |
|
|
345 |
|
if (defined( $input->{lookup} )) { |
346 |
|
$log->warn("$database/$input_name has depriciated lookup definition, removing it..."); |
347 |
|
delete( $input->{lookup} ); |
348 |
|
} |
349 |
|
|
350 |
|
my $lookup_coderef; |
351 |
|
|
352 |
|
if (@lookups) { |
353 |
|
|
354 |
|
my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input"); |
355 |
|
|
356 |
|
$lookup_coderef = sub { |
357 |
|
my $rec = shift || die "need rec!"; |
358 |
|
my $mfn = $rec->{'000'}->[0] || die "need mfn in 000"; |
359 |
|
|
360 |
|
WebPAC::Normalize::data_structure( |
361 |
|
row => $rec, |
362 |
|
rules => $rules, |
363 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
364 |
|
); |
365 |
|
|
366 |
|
#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup()); |
367 |
|
}; |
368 |
|
|
369 |
|
WebPAC::Normalize::_set_lookup( undef ); |
370 |
|
|
371 |
|
$log->debug("created lookup_coderef using:\n$rules"); |
372 |
|
|
373 |
|
}; |
374 |
|
|
375 |
|
my $lookup_jar; |
376 |
|
|
377 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
378 |
path => $input->{path}, |
path => $input->{path}, |
379 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
380 |
); |
lookup_coderef => $lookup_coderef, |
381 |
|
lookup => $lookup_jar, |
382 |
|
%{ $input }, |
383 |
|
load_row => sub { |
384 |
|
my $a = shift; |
385 |
|
return $store->load_row( |
386 |
|
database => $database, |
387 |
|
input => $input_name, |
388 |
|
id => $a->{id}, |
389 |
|
); |
390 |
|
}, |
391 |
|
save_row => sub { |
392 |
|
my $a = shift; |
393 |
|
return $store->save_row( |
394 |
|
database => $database, |
395 |
|
input => $input_name, |
396 |
|
id => $a->{id}, |
397 |
|
row => $a->{row}, |
398 |
|
); |
399 |
|
}, |
400 |
|
|
|
my $n = new WebPAC::Normalize::XML( |
|
|
# filter => { 'foo' => sub { shift } }, |
|
|
db => $db, |
|
|
lookup_regex => $lookup->regex, |
|
|
lookup => $lookup, |
|
|
prefix => $input->{name}, |
|
401 |
); |
); |
402 |
|
|
403 |
my $normalize_path = $input->{normalize}->{path}; |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
404 |
|
|
405 |
if ($normalize_path =~ m/\.xml$/i) { |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
406 |
$n->open( |
$log->debug("created following lookups: ", sub { dump( $lookup_data ) } ); |
407 |
tag => $input->{normalize}->{tag}, |
|
408 |
xml_file => $input->{normalize}->{path}, |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
409 |
); |
$store->save_lookup( |
410 |
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
database => $database, |
411 |
$n->open_yaml( |
input => $input_name, |
412 |
path => $normalize_path, |
key => $key, |
413 |
tag => $input->{normalize}->{tag}, |
data => $lookup_data->{$database}->{$input_name}->{$key}, |
414 |
|
); |
415 |
|
} |
416 |
|
} |
417 |
|
|
418 |
|
my $report_fh; |
419 |
|
if ($stats || $validate) { |
420 |
|
my $path = "out/report/${database}-${input_name}.txt"; |
421 |
|
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
422 |
|
|
423 |
|
print $report_fh "Report for database '$database' input '$input_name' records ", |
424 |
|
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
425 |
|
$log->info("Generating report file $path"); |
426 |
|
} |
427 |
|
|
428 |
|
my $marc; |
429 |
|
if ($parser->have_rules( 'marc', $database, $input_name )) { |
430 |
|
$marc = new WebPAC::Output::MARC( |
431 |
|
path => "out/marc/${database}-${input_name}.marc", |
432 |
|
lint => $marc_lint, |
433 |
|
dump => $marc_dump, |
434 |
); |
); |
435 |
} |
} |
436 |
|
|
437 |
|
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
438 |
|
$log->debug("parsed normalize rules:\n$rules"); |
439 |
|
|
440 |
|
# reset position in database |
441 |
|
$input_db->seek(1); |
442 |
|
|
443 |
|
# generate name of config key for indexer (strip everything after -) |
444 |
|
my $indexer_config = $use_indexer; |
445 |
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
446 |
|
|
447 |
|
my $lookup_hash; |
448 |
|
my $depends = $parser->depends($database,$input_name); |
449 |
|
|
450 |
|
if ($depends) { |
451 |
|
$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends); |
452 |
|
$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH'); |
453 |
|
|
454 |
|
foreach my $db (keys %$depends) { |
455 |
|
foreach my $i (keys %{$depends->{$db}}) { |
456 |
|
foreach my $k (keys %{$depends->{$db}->{$i}}) { |
457 |
|
my $t = time(); |
458 |
|
$log->debug("loading lookup $db/$i"); |
459 |
|
$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup( |
460 |
|
database => $db, |
461 |
|
input => $i, |
462 |
|
key => $k, |
463 |
|
); |
464 |
|
$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t)); |
465 |
|
} |
466 |
|
} |
467 |
|
} |
468 |
|
|
469 |
|
$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) }); |
470 |
|
} |
471 |
|
|
472 |
|
|
473 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
474 |
|
|
475 |
my $row = $input_db->fetch || next; |
my $row = $input_db->fetch || next; |
476 |
|
|
477 |
|
$total_rows++; |
478 |
|
|
479 |
my $mfn = $row->{'000'}->[0]; |
my $mfn = $row->{'000'}->[0]; |
480 |
|
|
481 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
if (! $mfn || $mfn !~ m#^\d+$#) { |
484 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
485 |
} |
} |
486 |
|
|
|
my $ds = $n->data_structure($row); |
|
487 |
|
|
488 |
$est->add( |
if ($validate) { |
489 |
id => $input->{name} . "/" . $mfn, |
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
490 |
ds => $ds, |
$log->error( "MFN $mfn validation error:\n", |
491 |
type => $config->{hyperestraier}->{type}, |
$validate->report_error( $errors ) |
492 |
|
); |
493 |
|
} |
494 |
|
next; # validation doesn't create any output |
495 |
|
} |
496 |
|
|
497 |
|
my $ds = WebPAC::Normalize::data_structure( |
498 |
|
row => $row, |
499 |
|
rules => $rules, |
500 |
|
lookup => $lookup_hash, |
501 |
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
502 |
|
marc_encoding => 'utf-8', |
503 |
|
load_row_coderef => sub { |
504 |
|
my ($database,$input,$mfn) = @_; |
505 |
|
return $store->load_row( |
506 |
|
database => $database, |
507 |
|
input => $input, |
508 |
|
id => $mfn, |
509 |
|
); |
510 |
|
}, |
511 |
); |
); |
512 |
|
|
513 |
$total_rows++; |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
|
} |
|
514 |
|
|
515 |
}; |
$store->save_ds( |
516 |
|
database => $database, |
517 |
|
input => $input_name, |
518 |
|
id => $mfn, |
519 |
|
ds => $ds, |
520 |
|
) if ($ds && !$stats); |
521 |
|
|
522 |
$log->info("$total_rows records indexed"); |
$indexer->add( |
523 |
|
id => "${input_name}/${mfn}", |
524 |
|
ds => $ds, |
525 |
|
type => $config->get($indexer_config)->{type}, |
526 |
|
) if ($indexer && $ds); |
527 |
|
|
528 |
# |
if ($marc) { |
529 |
# add Hyper Estraier links to other databases |
my $i = 0; |
530 |
# |
|
531 |
if (ref($db_config->{links}) eq 'ARRAY') { |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
532 |
foreach my $link (@{ $db_config->{links} }) { |
$marc->add( |
533 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
id => $mfn . ( $i ? "/$i" : '' ), |
534 |
$est->add_link( |
fields => $fields, |
535 |
from => $database, |
leader => WebPAC::Normalize::marc_leader(), |
536 |
to => $link->{to}, |
row => $row, |
537 |
credit => $link->{credit}, |
); |
538 |
); |
$i++; |
539 |
|
} |
540 |
|
|
541 |
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
542 |
|
} |
543 |
|
} |
544 |
|
|
545 |
|
if ($validate) { |
546 |
|
my $errors = $validate->report; |
547 |
|
if ($errors) { |
548 |
|
$log->info("validation errors:\n$errors\n" ); |
549 |
|
print $report_fh "$errors\n" if ($report_fh); |
550 |
|
} |
551 |
} |
} |
552 |
|
|
553 |
|
if ($stats) { |
554 |
|
my $s = $input_db->stats; |
555 |
|
$log->info("statistics of fields usage:\n$s"); |
556 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
557 |
|
} |
558 |
|
|
559 |
|
# close MARC file |
560 |
|
$marc->finish if ($marc); |
561 |
|
|
562 |
|
# close report |
563 |
|
close($report_fh) if ($report_fh) |
564 |
|
|
565 |
|
} |
566 |
|
|
567 |
|
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
568 |
|
|
569 |
|
my $dt = time() - $start_t; |
570 |
|
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
571 |
|
sprintf("in %.2f sec [%.2f rec/sec]", |
572 |
|
$dt, ($total_rows / $dt) |
573 |
|
) |
574 |
|
); |
575 |
|
|
576 |
|
|
577 |
|
# end forked process |
578 |
|
if ($parallel) { |
579 |
|
$log->info("parallel process $$ finished"); |
580 |
|
exit(0); |
581 |
} |
} |
582 |
|
|
583 |
} |
} |
584 |
|
|
585 |
|
if ($parallel) { |
586 |
|
# wait all children to finish |
587 |
|
sleep(1) while wait != -1; |
588 |
|
$log->info("all parallel processes finished"); |
589 |
|
} |
590 |
|
|
591 |
|
# |
592 |
|
# handle links or merge after indexing |
593 |
|
# |
594 |
|
|
595 |
|
if ($merge) { |
596 |
|
print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/; |
597 |
|
close($estcmd_fh); |
598 |
|
chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!"); |
599 |
|
system $estcmd_path; |
600 |
|
} else { |
601 |
|
foreach my $link (@links) { |
602 |
|
$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE'); |
603 |
|
$link->(); |
604 |
|
} |
605 |
|
} |