6 use File::Temp qw/tempdir/;
9 use WebPAC::Common 0.02;
10 use WebPAC::Lookup 0.03;
11 use WebPAC::Input 0.11;
12 use WebPAC::Store 0.03;
13 use WebPAC::Normalize 0.11;
14 use WebPAC::Output::TT;
15 use WebPAC::Validate 0.06;
16 use WebPAC::Output::MARC;
17 use YAML qw/LoadFile/;
20 use Time::HiRes qw/time/;
22 use Data::Dump qw/dump/;
23 use Storable qw/dclone/;
25 use Proc::Queue size => 1;
26 use POSIX ":sys_wait_h"; # imports WNOHANG
30 run.pl - start WebPAC indexing
32 B<this command will probably go away. Don't get used to it!>
40 start loading (all) databases at offset 42
44 limit loading to 100 records
48 remove database and Hyper Estraier index before indexing
50 =item --only=database_name/input_filter
52 reindex just single database (legacy name is --one)
54 C</input_filter> is optional part which can be C<name>
57 =item --config conf/config.yml
59 path to YAML configuration file
63 disable indexing, modify_* in configuration and dump statistics about field
64 and subfield usage for each input
66 =item --validate path/to/validation_file
68 turn on extra validation of imput records, see L<WebPAC::Validation>
70 =item --marc-normalize conf/normalize/mapping.pl
72 This option specifies normalisation file for MARC creation
74 =item --marc-output out/marc/test.marc
76 Optional path to output file
80 By default turned on if C<--marc-normalize> is used. You can disable lint
81 messages with C<--no-marc-lint>.
85 Force dump or input and marc record for debugging.
89 Run databases in parallel (aproximatly same as number of processors in
90 machine if you want to use full load)
98 Create merged index of databases which have links
108 my $config = 'conf/config.yml';
113 my ($marc_normalize, $marc_output);
120 my $log = _new WebPAC::Common()->_get_logger();
122 my $hostname = `hostname`;
124 $hostname =~ s/\..+$//;
125 if (-e "conf/$hostname.yml") {
126 $config = "conf/$hostname.yml";
127 $log->info("using host configuration file: $config");
131 "limit=i" => \$limit,
132 "offset=i" => \$offset,
134 "one=s" => \$only_filter,
135 "only=s" => \$only_filter,
136 "config" => \$config,
139 "validate=s" => \$validate_path,
140 "marc-normalize=s" => \$marc_normalize,
141 "marc-output=s" => \$marc_output,
142 "marc-lint!" => \$marc_lint,
143 "marc-dump!" => \$marc_dump,
144 "parallel=i" => \$parallel,
145 "only-links!" => \$only_links,
149 $config = LoadFile($config);
151 #print "config = ",dump($config) if ($debug);
153 die "no databases in config file!\n" unless ($config->{databases});
155 $log->info( "-" x 79 );
159 my $estcmd_path = './estcmd-merge.sh';
161 open($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");
169 $validate = new WebPAC::Validate(
170 path => $validate_path,
171 ) if ($validate_path);
174 my $use_indexer = $config->{use_indexer} || 'hyperestraier';
176 $log->debug("option --stats disables update of indexing engine...");
177 $use_indexer = undef;
179 $log->info("using $use_indexer indexing engine...");
182 # disable indexing when creating marc
183 $use_indexer = undef if ($marc_normalize);
186 my $start_t = time();
191 $log->info("Using $parallel processes for speedup");
192 Proc::Queue::size($parallel);
195 while (my ($database, $db_config) = each %{ $config->{databases} }) {
197 my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
198 next if ($only_database && $database !~ m/$only_database/i);
202 if(defined ($f) and $f==0) {
203 $log->info("Created processes $$ for speedup");
212 my $cfg_name = $use_indexer;
213 $cfg_name =~ s/\-.*$//;
215 my $indexer_config = $config->{$cfg_name} || $log->logdie("can't find '$cfg_name' part in confguration");
216 $indexer_config->{database} = $database;
217 $indexer_config->{clean} = $clean;
218 $indexer_config->{label} = $db_config->{name};
220 # force clean if database has links
221 $indexer_config->{clean} = 1 if ($db_config->{links});
223 if ($use_indexer eq 'hyperestraier') {
225 # open Hyper Estraier database
226 use WebPAC::Output::Estraier '0.10';
227 $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
229 } elsif ($use_indexer eq 'hyperestraier-native') {
231 # open Hyper Estraier database
232 use WebPAC::Output::EstraierNative;
233 $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
235 $use_indexer = 'hyperestraier';
237 } elsif ($use_indexer eq 'kinosearch') {
240 use WebPAC::Output::KinoSearch;
241 $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
242 $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
245 $log->logdie("unknown use_indexer: $use_indexer");
248 $log->logide("can't continue without valid indexer") unless ($indexer);
253 # store Hyper Estraier links to other databases
255 if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
256 foreach my $link (@{ $db_config->{links} }) {
257 if ($use_indexer eq 'hyperestraier') {
259 print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
261 $log->info("saving link $database -> $link->{to} [$link->{credit}]");
263 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
267 credit => $link->{credit},
272 $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
276 next if ($only_links);
282 my $abs_path = abs_path($0);
283 $abs_path =~ s#/[^/]*$#/#;
285 my $db_path = $config->{webpac}->{db_path} . '/' . $database;
288 $log->info("creating new database '$database' in $db_path");
289 rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
291 $log->info("working on database '$database' in $db_path");
294 my $db = new WebPAC::Store(
296 database => $database,
302 # now, iterate through input formats
306 if (ref($db_config->{input}) eq 'ARRAY') {
307 @inputs = @{ $db_config->{input} };
308 } elsif ($db_config->{input}) {
309 push @inputs, $db_config->{input};
311 $log->info("database $database doesn't have inputs defined");
314 my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
316 foreach my $input (@inputs) {
318 next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
320 my $type = lc($input->{type});
322 die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
325 if ($input->{lookup}) {
326 $lookup = new WebPAC::Lookup(
327 lookup_file => $input->{lookup},
329 delete( $input->{lookup} );
332 my $input_module = $config->{webpac}->{inputs}->{$type};
334 $log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
335 $input->{lookup} ? "lookup '$input->{lookup}'" : ""
339 # disable modification of records if --stats is in use
340 delete($input->{modify_records});
341 delete($input->{modify_file});
344 my $input_db = new WebPAC::Input(
345 module => $input_module,
346 encoding => $config->{webpac}->{webpac_encoding},
347 limit => $limit || $input->{limit},
349 lookup_coderef => sub {
350 my $rec = shift || return;
351 $lookup->add( $rec );
353 recode => $input->{recode},
355 modify_records => $input->{modify_records},
356 modify_file => $input->{modify_file},
358 $log->logdie("can't create input using $input_module") unless ($input);
360 my $maxmfn = $input_db->open(
361 path => $input->{path},
362 code_page => $input->{encoding}, # database encoding
366 my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
367 @{ $input->{normalize} } : ( $input->{normalize} );
369 if ($marc_normalize) {
371 path => $marc_normalize,
372 output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc',
376 foreach my $normalize (@norm_array) {
378 my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config");
380 $log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
382 my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
384 $log->info("Using $normalize_path for normalization...");
386 my $marc = new WebPAC::Output::MARC(
387 path => $normalize->{output},
390 ) if ($normalize->{output});
392 # reset position in database
395 foreach my $pos ( 0 ... $input_db->size ) {
397 my $row = $input_db->fetch || next;
399 my $mfn = $row->{'000'}->[0];
401 if (! $mfn || $mfn !~ m#^\d+$#) {
402 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
404 push @{ $row->{'000'} }, $pos;
409 if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) {
410 $log->error( "MFN $mfn validation error:\n",
416 my $ds_config = dclone($db_config);
418 # default values -> database key
419 $ds_config->{_} = $database;
422 $ds_config->{_mfn} = $mfn;
424 # attach current input
425 $ds_config->{input} = $input;
427 my $ds = WebPAC::Normalize::data_structure(
430 lookup => $lookup ? $lookup->lookup_hash : undef,
431 config => $ds_config,
432 marc_encoding => 'utf-8',
438 prefix => $input->{name},
439 ) if ($ds && !$stats);
442 id => $input->{name} . "/" . $mfn,
444 type => $config->{$use_indexer}->{type},
445 ) if ($indexer && $ds);
450 while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
452 id => $mfn . ( $i ? "/$i" : '' ),
454 leader => WebPAC::Normalize::marc_leader(),
460 $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
466 if ($validate && defined($validate->all_errors)) {
467 my $validate_errors = $validate->all_errors;
469 $log->info("validation errors:\n", dump( $validate_errors ) );
472 $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
475 $marc->finish if ($marc);
481 eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
483 my $dt = time() - $start_t;
484 $log->info("$total_rows records ", $indexer ? "indexed " : "",
485 sprintf("in %.2f sec [%.2f rec/sec]",
486 $dt, ($total_rows / $dt)
493 $log->info("parallel process $$ finished");
500 # wait all children to finish
501 sleep(1) while wait != -1;
502 $log->info("all parallel processes finished");
506 # handle links or merge after indexing
510 print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
512 chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
515 foreach my $link (@links) {
516 $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');