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::Lookup 0.03; |
11 |
use WebPAC::Input 0.03; |
use WebPAC::Input 0.11; |
12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.03; |
13 |
use WebPAC::Normalize; |
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; |
17 |
use YAML qw/LoadFile/; |
use YAML qw/LoadFile/; |
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; |
|
use MARC::Record 2.0; # need 2.0 for utf-8 encoding see marcpm.sf.net |
|
|
use MARC::Lint; |
|
22 |
use Data::Dump qw/dump/; |
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 |
|
|
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 |
|
|
84 |
|
|
85 |
Force dump or input and marc record for debugging. |
Force dump or input and marc record for debugging. |
86 |
|
|
87 |
|
=item --parallel 4 |
88 |
|
|
89 |
|
Run databases in parallel (aproximatly same as number of processors in |
90 |
|
machine if you want to use full load) |
91 |
|
|
92 |
|
=item --only-links |
93 |
|
|
94 |
|
Create just links |
95 |
|
|
96 |
|
=item --merge |
97 |
|
|
98 |
|
Create merged index of databases which have links |
99 |
|
|
100 |
=back |
=back |
101 |
|
|
102 |
=cut |
=cut |
113 |
my ($marc_normalize, $marc_output); |
my ($marc_normalize, $marc_output); |
114 |
my $marc_lint = 1; |
my $marc_lint = 1; |
115 |
my $marc_dump = 0; |
my $marc_dump = 0; |
116 |
|
my $parallel = 0; |
117 |
|
my $only_links = 0; |
118 |
|
my $merge = 0; |
119 |
|
|
120 |
|
my $log = _new WebPAC::Common()->_get_logger(); |
121 |
|
|
122 |
|
my $hostname = `hostname`; |
123 |
|
chomp($hostname); |
124 |
|
$hostname =~ s/\..+$//; |
125 |
|
if (-e "conf/$hostname.yml") { |
126 |
|
$config = "conf/$hostname.yml"; |
127 |
|
$log->info("using host configuration file: $config"); |
128 |
|
} |
129 |
|
|
130 |
GetOptions( |
GetOptions( |
131 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
141 |
"marc-output=s" => \$marc_output, |
"marc-output=s" => \$marc_output, |
142 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
143 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
144 |
|
"parallel=i" => \$parallel, |
145 |
|
"only-links!" => \$only_links, |
146 |
|
"merge" => \$merge, |
147 |
); |
); |
148 |
|
|
149 |
$config = LoadFile($config); |
$config = LoadFile($config); |
150 |
|
|
151 |
print "config = ",dump($config) if ($debug); |
#print "config = ",dump($config) if ($debug); |
152 |
|
|
153 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->{databases}); |
154 |
|
|
|
my $log = _new WebPAC::Common()->_get_logger(); |
|
155 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
156 |
|
|
157 |
|
|
158 |
|
my $estcmd_fh; |
159 |
|
my $estcmd_path = './estcmd-merge.sh'; |
160 |
|
if ($merge) { |
161 |
|
open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!"); |
162 |
|
print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/; |
163 |
|
print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/; |
164 |
|
$log->info("created merge batch file $estcmd_path"); |
165 |
|
} |
166 |
|
|
167 |
|
|
168 |
my $validate; |
my $validate; |
169 |
$validate = new WebPAC::Validate( |
$validate = new WebPAC::Validate( |
170 |
path => $validate_path, |
path => $validate_path, |
171 |
) if ($validate_path); |
) if ($validate_path); |
172 |
|
|
173 |
|
|
174 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
175 |
if ($stats) { |
if ($stats) { |
176 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("option --stats disables update of indexing engine..."); |
186 |
my $start_t = time(); |
my $start_t = time(); |
187 |
|
|
188 |
my @links; |
my @links; |
|
my $indexer; |
|
189 |
|
|
190 |
my $lint = new MARC::Lint if ($marc_lint); |
if ($parallel) { |
191 |
|
$log->info("Using $parallel processes for speedup"); |
192 |
|
Proc::Queue::size($parallel); |
193 |
|
} |
194 |
|
|
195 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
196 |
|
|
197 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
198 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
199 |
|
|
200 |
|
if ($parallel) { |
201 |
|
my $f=fork; |
202 |
|
if(defined ($f) and $f==0) { |
203 |
|
$log->info("Created processes $$ for speedup"); |
204 |
|
} else { |
205 |
|
next; |
206 |
|
} |
207 |
|
} |
208 |
|
|
209 |
|
my $indexer; |
210 |
if ($use_indexer) { |
if ($use_indexer) { |
211 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
|
212 |
|
my $cfg_name = $use_indexer; |
213 |
|
$cfg_name =~ s/\-.*$//; |
214 |
|
|
215 |
|
my $indexer_config = $config->{$cfg_name} || $log->logdie("can't find '$cfg_name' part in confguration"); |
216 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
217 |
$indexer_config->{clean} = $clean; |
$indexer_config->{clean} = $clean; |
218 |
$indexer_config->{label} = $db_config->{name}; |
$indexer_config->{label} = $db_config->{name}; |
219 |
|
|
220 |
|
# force clean if database has links |
221 |
|
$indexer_config->{clean} = 1 if ($db_config->{links}); |
222 |
|
|
223 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer eq 'hyperestraier') { |
224 |
|
|
225 |
# open Hyper Estraier database |
# open Hyper Estraier database |
226 |
use WebPAC::Output::Estraier '0.10'; |
use WebPAC::Output::Estraier '0.10'; |
227 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
228 |
|
|
229 |
|
} elsif ($use_indexer eq 'hyperestraier-native') { |
230 |
|
|
231 |
|
# open Hyper Estraier database |
232 |
|
use WebPAC::Output::EstraierNative; |
233 |
|
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
234 |
|
|
235 |
|
$use_indexer = 'hyperestraier'; |
236 |
|
|
237 |
} elsif ($use_indexer eq 'kinosearch') { |
} elsif ($use_indexer eq 'kinosearch') { |
238 |
|
|
239 |
# open KinoSearch |
# open KinoSearch |
250 |
|
|
251 |
|
|
252 |
# |
# |
253 |
|
# store Hyper Estraier links to other databases |
254 |
|
# |
255 |
|
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
256 |
|
foreach my $link (@{ $db_config->{links} }) { |
257 |
|
if ($use_indexer eq 'hyperestraier') { |
258 |
|
if ($merge) { |
259 |
|
print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/; |
260 |
|
} else { |
261 |
|
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
262 |
|
push @links, sub { |
263 |
|
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
264 |
|
$indexer->add_link( |
265 |
|
from => $database, |
266 |
|
to => $link->{to}, |
267 |
|
credit => $link->{credit}, |
268 |
|
); |
269 |
|
}; |
270 |
|
} |
271 |
|
} else { |
272 |
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
273 |
|
} |
274 |
|
} |
275 |
|
} |
276 |
|
next if ($only_links); |
277 |
|
|
278 |
|
|
279 |
|
# |
280 |
# now WebPAC::Store |
# now WebPAC::Store |
281 |
# |
# |
282 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
321 |
|
|
322 |
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(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); |
323 |
|
|
324 |
my $lookup = new WebPAC::Lookup( |
my $lookup; |
325 |
lookup_file => $input->{lookup}, |
if ($input->{lookup}) { |
326 |
) if ($input->{lookup}); |
$lookup = new WebPAC::Lookup( |
327 |
|
lookup_file => $input->{lookup}, |
328 |
|
); |
329 |
|
delete( $input->{lookup} ); |
330 |
|
} |
331 |
|
|
332 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
333 |
|
|
335 |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
336 |
); |
); |
337 |
|
|
338 |
|
if ($stats) { |
339 |
|
# disable modification of records if --stats is in use |
340 |
|
delete($input->{modify_records}); |
341 |
|
delete($input->{modify_file}); |
342 |
|
} |
343 |
|
|
344 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
345 |
module => $input_module, |
module => $input_module, |
346 |
code_page => $config->{webpac}->{webpac_encoding}, |
encoding => $config->{webpac}->{webpac_encoding}, |
347 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
348 |
offset => $offset, |
offset => $offset, |
349 |
lookup => $lookup, |
lookup_coderef => sub { |
350 |
|
my $rec = shift || return; |
351 |
|
$lookup->add( $rec ); |
352 |
|
}, |
353 |
recode => $input->{recode}, |
recode => $input->{recode}, |
354 |
stats => $stats, |
stats => $stats, |
355 |
|
modify_records => $input->{modify_records}, |
356 |
|
modify_file => $input->{modify_file}, |
357 |
); |
); |
358 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
359 |
|
|
363 |
%{ $input }, |
%{ $input }, |
364 |
); |
); |
365 |
|
|
366 |
|
my $report_fh; |
367 |
|
if ($stats || $validate) { |
368 |
|
my $path = "out/report/" . $database . '-' . $input->{name} . '.txt'; |
369 |
|
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
370 |
|
|
371 |
|
print $report_fh "Report for database '$database' input '$input->{name}' records ", |
372 |
|
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
373 |
|
$log->info("Generating report file $path"); |
374 |
|
} |
375 |
|
|
376 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
377 |
@{ $input->{normalize} } : ( $input->{normalize} ); |
@{ $input->{normalize} } : ( $input->{normalize} ); |
378 |
|
|
393 |
|
|
394 |
$log->info("Using $normalize_path for normalization..."); |
$log->info("Using $normalize_path for normalization..."); |
395 |
|
|
396 |
my $marc_fh; |
my $marc = new WebPAC::Output::MARC( |
397 |
if (my $path = $normalize->{output}) { |
path => $normalize->{output}, |
398 |
open($marc_fh, '>', $path) || |
lint => $marc_lint, |
399 |
$log->logdie("can't open MARC output $path: $!"); |
dump => $marc_dump, |
400 |
|
) if ($normalize->{output}); |
|
$log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\n"); |
|
|
} |
|
401 |
|
|
402 |
# reset position in database |
# reset position in database |
403 |
$input_db->seek(1); |
$input_db->seek(1); |
416 |
|
|
417 |
|
|
418 |
if ($validate) { |
if ($validate) { |
419 |
my @errors = $validate->validate_errors( $row ); |
if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) { |
420 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
$log->error( "MFN $mfn validation error:\n", |
421 |
|
$validate->report_error( $errors ) |
422 |
|
); |
423 |
|
} |
424 |
} |
} |
425 |
|
|
426 |
|
my $ds_config = dclone($db_config); |
427 |
|
|
428 |
|
# default values -> database key |
429 |
|
$ds_config->{_} = $database; |
430 |
|
|
431 |
|
# current mfn |
432 |
|
$ds_config->{_mfn} = $mfn; |
433 |
|
|
434 |
|
# attach current input |
435 |
|
$ds_config->{input} = $input; |
436 |
|
|
437 |
my $ds = WebPAC::Normalize::data_structure( |
my $ds = WebPAC::Normalize::data_structure( |
438 |
row => $row, |
row => $row, |
439 |
rules => $rules, |
rules => $rules, |
440 |
lookup => $lookup ? $lookup->lookup_hash : undef, |
lookup => $lookup ? $lookup->lookup_hash : undef, |
441 |
|
config => $ds_config, |
442 |
marc_encoding => 'utf-8', |
marc_encoding => 'utf-8', |
443 |
); |
); |
444 |
|
|
454 |
type => $config->{$use_indexer}->{type}, |
type => $config->{$use_indexer}->{type}, |
455 |
) if ($indexer && $ds); |
) if ($indexer && $ds); |
456 |
|
|
457 |
if ($marc_fh) { |
if ($marc) { |
458 |
my $marc = new MARC::Record; |
my $i = 0; |
|
$marc->encoding( 'utf-8' ); |
|
|
my @marc_fields = WebPAC::Normalize::_get_marc_fields(); |
|
|
if (! @marc_fields) { |
|
|
$log->warn("MARC record $mfn is empty, skipping"); |
|
|
} else { |
|
|
$marc->add_fields( @marc_fields ); |
|
|
|
|
|
# tweak leader |
|
|
if (my $new_l = WebPAC::Normalize::marc_leader()) { |
|
|
|
|
|
my $leader = $marc->leader; |
|
|
|
|
|
foreach my $o ( keys %$new_l ) { |
|
|
my $insert = $new_l->{$o}; |
|
|
$leader = substr($leader, 0, $o) . |
|
|
$insert . substr($leader, $o+length($insert)); |
|
|
} |
|
|
$marc->leader( $leader ); |
|
|
} |
|
|
|
|
|
if ($marc_lint) { |
|
|
$lint->check_record( $marc ); |
|
|
my $err = join( "\n", $lint->warnings ); |
|
|
$log->error("MARC lint detected warning on MFN $mfn\tleader:", $marc->leader(), "\n", |
|
|
">>>>> Original imput row:\n",dump($row), "\n", |
|
|
">>>>> Normalized MARC row:\n",dump(@marc_fields), "\n", |
|
|
">>>>> MARC lint warnings:\n",$err,"\n" |
|
|
) if ($err); |
|
|
} |
|
|
|
|
|
if ($marc_dump) { |
|
|
$log->info("MARC record on MFN $mfn\tleader:", $marc->leader(), "\n", |
|
|
">>>>> Original imput row:\n",dump($row), "\n", |
|
|
">>>>> Normalized MARC row:\n",dump(@marc_fields), "\n", |
|
|
); |
|
|
} |
|
459 |
|
|
460 |
print $marc_fh $marc->as_usmarc; |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
461 |
|
$marc->add( |
462 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
463 |
|
fields => $fields, |
464 |
|
leader => WebPAC::Normalize::marc_leader(), |
465 |
|
row => $row, |
466 |
|
); |
467 |
|
$i++; |
468 |
} |
} |
469 |
|
|
470 |
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
471 |
} |
} |
472 |
|
|
473 |
$total_rows++; |
$total_rows++; |
474 |
} |
} |
475 |
|
|
476 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
if ($validate) { |
477 |
|
my $errors = $validate->report; |
478 |
|
if ($errors) { |
479 |
|
$log->info("validation errors:\n$errors\n" ); |
480 |
|
print $report_fh "$errors\n" if ($report_fh); |
481 |
|
} |
482 |
|
} |
483 |
|
|
484 |
|
if ($stats) { |
485 |
|
my $s = $input_db->stats; |
486 |
|
$log->info("statistics of fields usage:\n$s"); |
487 |
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
488 |
|
} |
489 |
|
|
490 |
# close MARC file |
# close MARC file |
491 |
close($marc_fh) if ($marc_fh); |
$marc->finish if ($marc); |
492 |
|
|
493 |
|
# close report |
494 |
|
close($report_fh) if ($report_fh) |
495 |
} |
} |
496 |
|
|
497 |
} |
} |
505 |
) |
) |
506 |
); |
); |
507 |
|
|
508 |
# |
|
509 |
# add Hyper Estraier links to other databases |
# end forked process |
510 |
# |
if ($parallel) { |
511 |
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
$log->info("parallel process $$ finished"); |
512 |
foreach my $link (@{ $db_config->{links} }) { |
exit(0); |
|
if ($use_indexer eq 'hyperestraier') { |
|
|
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
|
|
push @links, { |
|
|
from => $database, |
|
|
to => $link->{to}, |
|
|
credit => $link->{credit}, |
|
|
}; |
|
|
} else { |
|
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
|
|
} |
|
|
} |
|
513 |
} |
} |
514 |
|
|
515 |
} |
} |
516 |
|
|
517 |
foreach my $link (@links) { |
if ($parallel) { |
518 |
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
# wait all children to finish |
519 |
$indexer->add_link( %{ $link } ); |
sleep(1) while wait != -1; |
520 |
|
$log->info("all parallel processes finished"); |
521 |
|
} |
522 |
|
|
523 |
|
# |
524 |
|
# handle links or merge after indexing |
525 |
|
# |
526 |
|
|
527 |
|
if ($merge) { |
528 |
|
print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/; |
529 |
|
close($estcmd_fh); |
530 |
|
chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!"); |
531 |
|
system $estcmd_path; |
532 |
|
} else { |
533 |
|
foreach my $link (@links) { |
534 |
|
$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE'); |
535 |
|
$link->(); |
536 |
|
} |
537 |
} |
} |