1 |
dpavlin |
74 |
#!/usr/bin/perl -w |
2 |
|
|
|
3 |
|
|
use strict; |
4 |
|
|
|
5 |
|
|
use Cwd qw/abs_path/; |
6 |
|
|
use File::Temp qw/tempdir/; |
7 |
|
|
use lib './lib'; |
8 |
|
|
|
9 |
dpavlin |
255 |
use WebPAC::Common 0.02; |
10 |
dpavlin |
755 |
use WebPAC::Parser 0.08; |
11 |
dpavlin |
774 |
use WebPAC::Input 0.16; |
12 |
dpavlin |
762 |
use WebPAC::Store 0.14; |
13 |
dpavlin |
736 |
use WebPAC::Normalize 0.22; |
14 |
dpavlin |
74 |
use WebPAC::Output::TT; |
15 |
dpavlin |
652 |
use WebPAC::Validate 0.06; |
16 |
dpavlin |
578 |
use WebPAC::Output::MARC; |
17 |
dpavlin |
684 |
use WebPAC::Config; |
18 |
dpavlin |
301 |
use Getopt::Long; |
19 |
|
|
use File::Path; |
20 |
dpavlin |
389 |
use Time::HiRes qw/time/; |
21 |
dpavlin |
492 |
use File::Slurp; |
22 |
dpavlin |
556 |
use Data::Dump qw/dump/; |
23 |
dpavlin |
595 |
use Storable qw/dclone/; |
24 |
dpavlin |
790 |
use Pod::Usage qw/pod2usage/; |
25 |
dpavlin |
74 |
|
26 |
dpavlin |
606 |
use Proc::Queue size => 1; |
27 |
|
|
use POSIX ":sys_wait_h"; # imports WNOHANG |
28 |
|
|
|
29 |
dpavlin |
301 |
=head1 NAME |
30 |
dpavlin |
76 |
|
31 |
dpavlin |
301 |
run.pl - start WebPAC indexing |
32 |
dpavlin |
141 |
|
33 |
dpavlin |
301 |
B<this command will probably go away. Don't get used to it!> |
34 |
dpavlin |
141 |
|
35 |
dpavlin |
790 |
=head1 OPTIONS |
36 |
dpavlin |
301 |
|
37 |
|
|
=over 4 |
38 |
|
|
|
39 |
|
|
=item --offset 42 |
40 |
|
|
|
41 |
|
|
start loading (all) databases at offset 42 |
42 |
|
|
|
43 |
|
|
=item --limit 100 |
44 |
|
|
|
45 |
|
|
limit loading to 100 records |
46 |
|
|
|
47 |
|
|
=item --clean |
48 |
|
|
|
49 |
|
|
remove database and Hyper Estraier index before indexing |
50 |
|
|
|
51 |
dpavlin |
510 |
=item --only=database_name/input_filter |
52 |
dpavlin |
335 |
|
53 |
dpavlin |
423 |
reindex just single database (legacy name is --one) |
54 |
dpavlin |
335 |
|
55 |
dpavlin |
510 |
C</input_filter> is optional part which can be C<name> |
56 |
|
|
or C<type> from input |
57 |
|
|
|
58 |
dpavlin |
301 |
=item --config conf/config.yml |
59 |
|
|
|
60 |
|
|
path to YAML configuration file |
61 |
|
|
|
62 |
dpavlin |
507 |
=item --stats |
63 |
|
|
|
64 |
dpavlin |
638 |
disable indexing, modify_* in configuration and dump statistics about field |
65 |
|
|
and subfield usage for each input |
66 |
dpavlin |
507 |
|
67 |
dpavlin |
516 |
=item --validate path/to/validation_file |
68 |
|
|
|
69 |
|
|
turn on extra validation of imput records, see L<WebPAC::Validation> |
70 |
|
|
|
71 |
dpavlin |
803 |
=item --marc-generate |
72 |
|
|
|
73 |
|
|
Generate MARC file. This will automatically be on if file contains C<marc*> directives. |
74 |
|
|
You can use this option as C<--no-marc-generate> to disable MARC generation. |
75 |
|
|
|
76 |
dpavlin |
556 |
=item --marc-lint |
77 |
|
|
|
78 |
dpavlin |
738 |
By default turned on if normalisation file has C<marc*> directives. You can disable lint |
79 |
dpavlin |
558 |
messages with C<--no-marc-lint>. |
80 |
dpavlin |
556 |
|
81 |
dpavlin |
559 |
=item --marc-dump |
82 |
|
|
|
83 |
|
|
Force dump or input and marc record for debugging. |
84 |
|
|
|
85 |
dpavlin |
606 |
=item --parallel 4 |
86 |
|
|
|
87 |
|
|
Run databases in parallel (aproximatly same as number of processors in |
88 |
|
|
machine if you want to use full load) |
89 |
|
|
|
90 |
dpavlin |
607 |
=item --only-links |
91 |
|
|
|
92 |
|
|
Create just links |
93 |
|
|
|
94 |
dpavlin |
608 |
=item --merge |
95 |
|
|
|
96 |
|
|
Create merged index of databases which have links |
97 |
|
|
|
98 |
dpavlin |
301 |
=back |
99 |
|
|
|
100 |
|
|
=cut |
101 |
|
|
|
102 |
|
|
my $offset; |
103 |
|
|
my $limit; |
104 |
|
|
|
105 |
|
|
my $clean = 0; |
106 |
dpavlin |
684 |
my $config_path; |
107 |
dpavlin |
301 |
my $debug = 0; |
108 |
dpavlin |
510 |
my $only_filter; |
109 |
dpavlin |
507 |
my $stats = 0; |
110 |
dpavlin |
516 |
my $validate_path; |
111 |
dpavlin |
803 |
my $marc_generate = 1; |
112 |
dpavlin |
556 |
my $marc_lint = 1; |
113 |
dpavlin |
559 |
my $marc_dump = 0; |
114 |
dpavlin |
606 |
my $parallel = 0; |
115 |
dpavlin |
607 |
my $only_links = 0; |
116 |
dpavlin |
608 |
my $merge = 0; |
117 |
dpavlin |
790 |
my $help; |
118 |
dpavlin |
606 |
|
119 |
dpavlin |
611 |
my $log = _new WebPAC::Common()->_get_logger(); |
120 |
|
|
|
121 |
dpavlin |
301 |
GetOptions( |
122 |
|
|
"limit=i" => \$limit, |
123 |
|
|
"offset=i" => \$offset, |
124 |
|
|
"clean" => \$clean, |
125 |
dpavlin |
510 |
"one=s" => \$only_filter, |
126 |
|
|
"only=s" => \$only_filter, |
127 |
dpavlin |
684 |
"config" => \$config_path, |
128 |
dpavlin |
560 |
"debug+" => \$debug, |
129 |
dpavlin |
507 |
"stats" => \$stats, |
130 |
dpavlin |
516 |
"validate=s" => \$validate_path, |
131 |
dpavlin |
803 |
"marc-generate!" => \$marc_generate, |
132 |
dpavlin |
556 |
"marc-lint!" => \$marc_lint, |
133 |
dpavlin |
559 |
"marc-dump!" => \$marc_dump, |
134 |
dpavlin |
606 |
"parallel=i" => \$parallel, |
135 |
dpavlin |
607 |
"only-links!" => \$only_links, |
136 |
dpavlin |
608 |
"merge" => \$merge, |
137 |
dpavlin |
790 |
"help" => \$help, |
138 |
dpavlin |
301 |
); |
139 |
|
|
|
140 |
dpavlin |
790 |
pod2usage(-verbose => 2) if ($help); |
141 |
|
|
|
142 |
dpavlin |
684 |
my $config = new WebPAC::Config( path => $config_path ); |
143 |
dpavlin |
301 |
|
144 |
dpavlin |
611 |
#print "config = ",dump($config) if ($debug); |
145 |
dpavlin |
301 |
|
146 |
dpavlin |
684 |
die "no databases in config file!\n" unless ($config->databases); |
147 |
dpavlin |
431 |
|
148 |
dpavlin |
509 |
$log->info( "-" x 79 ); |
149 |
dpavlin |
431 |
|
150 |
dpavlin |
775 |
my $log_file = 'log'; |
151 |
dpavlin |
608 |
|
152 |
dpavlin |
775 |
if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) { |
153 |
|
|
$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'"); |
154 |
|
|
rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!"); |
155 |
|
|
} |
156 |
|
|
|
157 |
dpavlin |
608 |
my $estcmd_fh; |
158 |
|
|
my $estcmd_path = './estcmd-merge.sh'; |
159 |
|
|
if ($merge) { |
160 |
|
|
open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!"); |
161 |
|
|
print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/; |
162 |
|
|
print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/; |
163 |
|
|
$log->info("created merge batch file $estcmd_path"); |
164 |
|
|
} |
165 |
|
|
|
166 |
|
|
|
167 |
dpavlin |
516 |
my $validate; |
168 |
|
|
$validate = new WebPAC::Validate( |
169 |
|
|
path => $validate_path, |
170 |
|
|
) if ($validate_path); |
171 |
|
|
|
172 |
dpavlin |
608 |
|
173 |
dpavlin |
684 |
my $use_indexer = $config->use_indexer; |
174 |
dpavlin |
769 |
$stats ||= $validate; |
175 |
dpavlin |
509 |
if ($stats) { |
176 |
dpavlin |
769 |
$log->debug("disabled indexing for stats collection"); |
177 |
dpavlin |
509 |
$use_indexer = undef; |
178 |
|
|
} else { |
179 |
|
|
$log->info("using $use_indexer indexing engine..."); |
180 |
|
|
} |
181 |
dpavlin |
141 |
|
182 |
dpavlin |
698 |
# parse normalize files and create source files for lookup and normalization |
183 |
|
|
|
184 |
|
|
my $parser = new WebPAC::Parser( config => $config ); |
185 |
|
|
|
186 |
dpavlin |
213 |
my $total_rows = 0; |
187 |
dpavlin |
389 |
my $start_t = time(); |
188 |
dpavlin |
213 |
|
189 |
dpavlin |
529 |
my @links; |
190 |
|
|
|
191 |
dpavlin |
606 |
if ($parallel) { |
192 |
|
|
$log->info("Using $parallel processes for speedup"); |
193 |
|
|
Proc::Queue::size($parallel); |
194 |
|
|
} |
195 |
|
|
|
196 |
dpavlin |
707 |
sub create_ds_config { |
197 |
|
|
my ($db_config, $database, $input, $mfn) = @_; |
198 |
|
|
my $c = dclone( $db_config ); |
199 |
|
|
$c->{_} = $database || $log->logconfess("need database"); |
200 |
|
|
$c->{_mfn} = $mfn || $log->logconfess("need mfn"); |
201 |
|
|
$c->{input} = $input || $log->logconfess("need input"); |
202 |
|
|
return $c; |
203 |
|
|
} |
204 |
|
|
|
205 |
dpavlin |
684 |
while (my ($database, $db_config) = each %{ $config->databases }) { |
206 |
dpavlin |
141 |
|
207 |
dpavlin |
512 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
208 |
dpavlin |
510 |
next if ($only_database && $database !~ m/$only_database/i); |
209 |
dpavlin |
335 |
|
210 |
dpavlin |
606 |
if ($parallel) { |
211 |
|
|
my $f=fork; |
212 |
|
|
if(defined ($f) and $f==0) { |
213 |
|
|
$log->info("Created processes $$ for speedup"); |
214 |
|
|
} else { |
215 |
|
|
next; |
216 |
|
|
} |
217 |
|
|
} |
218 |
|
|
|
219 |
dpavlin |
607 |
my $indexer; |
220 |
dpavlin |
755 |
if ($use_indexer && $parser->have_rules( 'search', $database )) { |
221 |
dpavlin |
627 |
|
222 |
|
|
my $cfg_name = $use_indexer; |
223 |
|
|
$cfg_name =~ s/\-.*$//; |
224 |
|
|
|
225 |
dpavlin |
684 |
my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration"); |
226 |
dpavlin |
509 |
$indexer_config->{database} = $database; |
227 |
|
|
$indexer_config->{clean} = $clean; |
228 |
|
|
$indexer_config->{label} = $db_config->{name}; |
229 |
dpavlin |
431 |
|
230 |
dpavlin |
608 |
# force clean if database has links |
231 |
|
|
$indexer_config->{clean} = 1 if ($db_config->{links}); |
232 |
|
|
|
233 |
dpavlin |
509 |
if ($use_indexer eq 'hyperestraier') { |
234 |
dpavlin |
255 |
|
235 |
dpavlin |
509 |
# open Hyper Estraier database |
236 |
|
|
use WebPAC::Output::Estraier '0.10'; |
237 |
|
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
238 |
|
|
|
239 |
dpavlin |
627 |
} elsif ($use_indexer eq 'hyperestraier-native') { |
240 |
|
|
|
241 |
|
|
# open Hyper Estraier database |
242 |
|
|
use WebPAC::Output::EstraierNative; |
243 |
|
|
$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); |
244 |
|
|
|
245 |
dpavlin |
509 |
} elsif ($use_indexer eq 'kinosearch') { |
246 |
dpavlin |
430 |
|
247 |
dpavlin |
509 |
# open KinoSearch |
248 |
|
|
use WebPAC::Output::KinoSearch; |
249 |
|
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
250 |
|
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
251 |
dpavlin |
431 |
|
252 |
dpavlin |
509 |
} else { |
253 |
|
|
$log->logdie("unknown use_indexer: $use_indexer"); |
254 |
|
|
} |
255 |
|
|
|
256 |
|
|
$log->logide("can't continue without valid indexer") unless ($indexer); |
257 |
dpavlin |
430 |
} |
258 |
|
|
|
259 |
|
|
|
260 |
dpavlin |
255 |
# |
261 |
dpavlin |
607 |
# store Hyper Estraier links to other databases |
262 |
|
|
# |
263 |
|
|
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
264 |
|
|
foreach my $link (@{ $db_config->{links} }) { |
265 |
|
|
if ($use_indexer eq 'hyperestraier') { |
266 |
dpavlin |
608 |
if ($merge) { |
267 |
|
|
print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/; |
268 |
|
|
} else { |
269 |
|
|
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
270 |
|
|
push @links, sub { |
271 |
|
|
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
272 |
|
|
$indexer->add_link( |
273 |
|
|
from => $database, |
274 |
|
|
to => $link->{to}, |
275 |
|
|
credit => $link->{credit}, |
276 |
|
|
); |
277 |
|
|
}; |
278 |
|
|
} |
279 |
dpavlin |
607 |
} else { |
280 |
|
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
281 |
|
|
} |
282 |
|
|
} |
283 |
|
|
} |
284 |
|
|
next if ($only_links); |
285 |
|
|
|
286 |
|
|
|
287 |
|
|
# |
288 |
dpavlin |
255 |
# now WebPAC::Store |
289 |
|
|
# |
290 |
dpavlin |
210 |
my $abs_path = abs_path($0); |
291 |
|
|
$abs_path =~ s#/[^/]*$#/#; |
292 |
dpavlin |
141 |
|
293 |
dpavlin |
713 |
my $db_path = $config->webpac('db_path'); |
294 |
dpavlin |
74 |
|
295 |
dpavlin |
301 |
if ($clean) { |
296 |
dpavlin |
511 |
$log->info("creating new database '$database' in $db_path"); |
297 |
dpavlin |
301 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
298 |
|
|
} else { |
299 |
dpavlin |
511 |
$log->info("working on database '$database' in $db_path"); |
300 |
dpavlin |
301 |
} |
301 |
dpavlin |
255 |
|
302 |
dpavlin |
710 |
my $store = new WebPAC::Store( |
303 |
dpavlin |
210 |
path => $db_path, |
304 |
dpavlin |
301 |
debug => $debug, |
305 |
dpavlin |
210 |
); |
306 |
dpavlin |
74 |
|
307 |
dpavlin |
233 |
|
308 |
dpavlin |
213 |
# |
309 |
|
|
# now, iterate through input formats |
310 |
|
|
# |
311 |
dpavlin |
74 |
|
312 |
dpavlin |
213 |
my @inputs; |
313 |
|
|
if (ref($db_config->{input}) eq 'ARRAY') { |
314 |
|
|
@inputs = @{ $db_config->{input} }; |
315 |
dpavlin |
255 |
} elsif ($db_config->{input}) { |
316 |
|
|
push @inputs, $db_config->{input}; |
317 |
dpavlin |
213 |
} else { |
318 |
dpavlin |
255 |
$log->info("database $database doesn't have inputs defined"); |
319 |
dpavlin |
213 |
} |
320 |
dpavlin |
74 |
|
321 |
dpavlin |
213 |
foreach my $input (@inputs) { |
322 |
dpavlin |
233 |
|
323 |
dpavlin |
710 |
my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input)); |
324 |
dpavlin |
510 |
|
325 |
dpavlin |
710 |
next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
326 |
|
|
|
327 |
dpavlin |
233 |
my $type = lc($input->{type}); |
328 |
|
|
|
329 |
dpavlin |
684 |
die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs'))); |
330 |
dpavlin |
233 |
|
331 |
dpavlin |
684 |
my $input_module = $config->webpac('inputs')->{$type}; |
332 |
dpavlin |
74 |
|
333 |
dpavlin |
706 |
my @lookups = $parser->have_lookup_create($database, $input); |
334 |
|
|
|
335 |
dpavlin |
710 |
$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module", |
336 |
dpavlin |
707 |
@lookups ? " creating lookups: ".join(", ", @lookups) : "" |
337 |
dpavlin |
523 |
); |
338 |
dpavlin |
286 |
|
339 |
dpavlin |
638 |
if ($stats) { |
340 |
|
|
# disable modification of records if --stats is in use |
341 |
|
|
delete($input->{modify_records}); |
342 |
|
|
delete($input->{modify_file}); |
343 |
|
|
} |
344 |
|
|
|
345 |
dpavlin |
287 |
my $input_db = new WebPAC::Input( |
346 |
|
|
module => $input_module, |
347 |
dpavlin |
684 |
encoding => $config->webpac('webpac_encoding'), |
348 |
dpavlin |
301 |
limit => $limit || $input->{limit}, |
349 |
|
|
offset => $offset, |
350 |
dpavlin |
416 |
recode => $input->{recode}, |
351 |
dpavlin |
507 |
stats => $stats, |
352 |
dpavlin |
600 |
modify_records => $input->{modify_records}, |
353 |
dpavlin |
636 |
modify_file => $input->{modify_file}, |
354 |
dpavlin |
287 |
); |
355 |
dpavlin |
286 |
$log->logdie("can't create input using $input_module") unless ($input); |
356 |
dpavlin |
113 |
|
357 |
dpavlin |
702 |
if (defined( $input->{lookup} )) { |
358 |
dpavlin |
710 |
$log->warn("$database/$input_name has depriciated lookup definition, removing it..."); |
359 |
dpavlin |
702 |
delete( $input->{lookup} ); |
360 |
|
|
} |
361 |
|
|
|
362 |
dpavlin |
707 |
my $lookup_coderef; |
363 |
|
|
|
364 |
|
|
if (@lookups) { |
365 |
|
|
|
366 |
|
|
my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input"); |
367 |
|
|
|
368 |
|
|
$lookup_coderef = sub { |
369 |
|
|
my $rec = shift || die "need rec!"; |
370 |
|
|
my $mfn = $rec->{'000'}->[0] || die "need mfn in 000"; |
371 |
|
|
|
372 |
|
|
WebPAC::Normalize::data_structure( |
373 |
|
|
row => $rec, |
374 |
dpavlin |
732 |
rules => $rules, |
375 |
dpavlin |
707 |
config => create_ds_config( $db_config, $database, $input, $mfn ), |
376 |
|
|
); |
377 |
|
|
|
378 |
dpavlin |
732 |
#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup()); |
379 |
dpavlin |
707 |
}; |
380 |
|
|
|
381 |
|
|
WebPAC::Normalize::_set_lookup( undef ); |
382 |
|
|
|
383 |
|
|
$log->debug("created lookup_coderef using:\n$rules"); |
384 |
|
|
|
385 |
|
|
}; |
386 |
|
|
|
387 |
dpavlin |
731 |
my $lookup_jar; |
388 |
|
|
|
389 |
dpavlin |
287 |
my $maxmfn = $input_db->open( |
390 |
dpavlin |
285 |
path => $input->{path}, |
391 |
dpavlin |
213 |
code_page => $input->{encoding}, # database encoding |
392 |
dpavlin |
707 |
lookup_coderef => $lookup_coderef, |
393 |
dpavlin |
731 |
lookup => $lookup_jar, |
394 |
dpavlin |
523 |
%{ $input }, |
395 |
dpavlin |
762 |
load_row => sub { |
396 |
|
|
my $a = shift; |
397 |
|
|
return $store->load_row( |
398 |
|
|
database => $database, |
399 |
|
|
input => $input_name, |
400 |
|
|
id => $a->{id}, |
401 |
|
|
); |
402 |
|
|
}, |
403 |
|
|
save_row => sub { |
404 |
|
|
my $a = shift; |
405 |
|
|
return $store->save_row( |
406 |
|
|
database => $database, |
407 |
|
|
input => $input_name, |
408 |
|
|
id => $a->{id}, |
409 |
|
|
row => $a->{row}, |
410 |
|
|
); |
411 |
|
|
}, |
412 |
|
|
|
413 |
dpavlin |
213 |
); |
414 |
dpavlin |
113 |
|
415 |
dpavlin |
710 |
my $lookup_data = WebPAC::Normalize::_get_lookup(); |
416 |
dpavlin |
707 |
|
417 |
dpavlin |
712 |
if (defined( $lookup_data->{$database}->{$input_name} )) { |
418 |
dpavlin |
765 |
$log->debug("created following lookups: ", sub { dump( $lookup_data ) } ); |
419 |
dpavlin |
710 |
|
420 |
dpavlin |
712 |
foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) { |
421 |
dpavlin |
713 |
$store->save_lookup( |
422 |
|
|
database => $database, |
423 |
|
|
input => $input_name, |
424 |
|
|
key => $key, |
425 |
|
|
data => $lookup_data->{$database}->{$input_name}->{$key}, |
426 |
|
|
); |
427 |
dpavlin |
712 |
} |
428 |
dpavlin |
710 |
} |
429 |
|
|
|
430 |
dpavlin |
671 |
my $report_fh; |
431 |
|
|
if ($stats || $validate) { |
432 |
dpavlin |
710 |
my $path = "out/report/${database}-${input_name}.txt"; |
433 |
dpavlin |
671 |
open($report_fh, '>', $path) || $log->logdie("can't open $path: $!"); |
434 |
|
|
|
435 |
dpavlin |
710 |
print $report_fh "Report for database '$database' input '$input_name' records ", |
436 |
dpavlin |
671 |
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n"; |
437 |
|
|
$log->info("Generating report file $path"); |
438 |
|
|
} |
439 |
|
|
|
440 |
dpavlin |
731 |
my $marc; |
441 |
dpavlin |
803 |
if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { |
442 |
dpavlin |
731 |
$marc = new WebPAC::Output::MARC( |
443 |
dpavlin |
738 |
path => "out/marc/${database}-${input_name}.marc", |
444 |
dpavlin |
731 |
lint => $marc_lint, |
445 |
|
|
dump => $marc_dump, |
446 |
|
|
); |
447 |
dpavlin |
552 |
} |
448 |
|
|
|
449 |
dpavlin |
731 |
my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name"); |
450 |
|
|
$log->debug("parsed normalize rules:\n$rules"); |
451 |
dpavlin |
492 |
|
452 |
dpavlin |
731 |
# reset position in database |
453 |
|
|
$input_db->seek(1); |
454 |
dpavlin |
269 |
|
455 |
dpavlin |
731 |
# generate name of config key for indexer (strip everything after -) |
456 |
|
|
my $indexer_config = $use_indexer; |
457 |
|
|
$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config); |
458 |
dpavlin |
210 |
|
459 |
dpavlin |
731 |
my $lookup_hash; |
460 |
|
|
my $depends = $parser->depends($database,$input_name); |
461 |
|
|
|
462 |
|
|
if ($depends) { |
463 |
|
|
$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends); |
464 |
|
|
$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH'); |
465 |
dpavlin |
213 |
|
466 |
dpavlin |
731 |
foreach my $db (keys %$depends) { |
467 |
|
|
foreach my $i (keys %{$depends->{$db}}) { |
468 |
|
|
foreach my $k (keys %{$depends->{$db}->{$i}}) { |
469 |
dpavlin |
762 |
my $t = time(); |
470 |
dpavlin |
731 |
$log->debug("loading lookup $db/$i"); |
471 |
|
|
$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup( |
472 |
|
|
database => $db, |
473 |
|
|
input => $i, |
474 |
|
|
key => $k, |
475 |
|
|
); |
476 |
dpavlin |
762 |
$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t)); |
477 |
dpavlin |
731 |
} |
478 |
|
|
} |
479 |
|
|
} |
480 |
dpavlin |
213 |
|
481 |
dpavlin |
762 |
$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) }); |
482 |
dpavlin |
731 |
} |
483 |
dpavlin |
540 |
|
484 |
dpavlin |
290 |
|
485 |
dpavlin |
731 |
foreach my $pos ( 0 ... $input_db->size ) { |
486 |
dpavlin |
674 |
|
487 |
dpavlin |
731 |
my $row = $input_db->fetch || next; |
488 |
dpavlin |
516 |
|
489 |
dpavlin |
769 |
$total_rows++; |
490 |
|
|
|
491 |
dpavlin |
731 |
my $mfn = $row->{'000'}->[0]; |
492 |
dpavlin |
516 |
|
493 |
dpavlin |
731 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
494 |
|
|
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
495 |
|
|
$mfn = $pos; |
496 |
|
|
push @{ $row->{'000'} }, $pos; |
497 |
|
|
} |
498 |
dpavlin |
213 |
|
499 |
dpavlin |
508 |
|
500 |
dpavlin |
731 |
if ($validate) { |
501 |
dpavlin |
774 |
if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) { |
502 |
dpavlin |
731 |
$log->error( "MFN $mfn validation error:\n", |
503 |
|
|
$validate->report_error( $errors ) |
504 |
|
|
); |
505 |
dpavlin |
539 |
} |
506 |
dpavlin |
769 |
next; # validation doesn't create any output |
507 |
dpavlin |
731 |
} |
508 |
dpavlin |
213 |
|
509 |
dpavlin |
731 |
my $ds = WebPAC::Normalize::data_structure( |
510 |
|
|
row => $row, |
511 |
|
|
rules => $rules, |
512 |
|
|
lookup => $lookup_hash, |
513 |
|
|
config => create_ds_config( $db_config, $database, $input, $mfn ), |
514 |
|
|
marc_encoding => 'utf-8', |
515 |
dpavlin |
736 |
load_row_coderef => sub { |
516 |
dpavlin |
731 |
my ($database,$input,$mfn) = @_; |
517 |
dpavlin |
736 |
return $store->load_row( |
518 |
dpavlin |
731 |
database => $database, |
519 |
|
|
input => $input, |
520 |
|
|
id => $mfn, |
521 |
|
|
); |
522 |
|
|
}, |
523 |
|
|
); |
524 |
dpavlin |
507 |
|
525 |
dpavlin |
765 |
$log->debug("ds = ", sub { dump($ds) }) if ($ds); |
526 |
dpavlin |
736 |
|
527 |
dpavlin |
731 |
$store->save_ds( |
528 |
|
|
database => $database, |
529 |
|
|
input => $input_name, |
530 |
|
|
id => $mfn, |
531 |
|
|
ds => $ds, |
532 |
|
|
) if ($ds && !$stats); |
533 |
dpavlin |
210 |
|
534 |
dpavlin |
731 |
$indexer->add( |
535 |
|
|
id => "${input_name}/${mfn}", |
536 |
|
|
ds => $ds, |
537 |
|
|
type => $config->get($indexer_config)->{type}, |
538 |
|
|
) if ($indexer && $ds); |
539 |
dpavlin |
539 |
|
540 |
dpavlin |
731 |
if ($marc) { |
541 |
|
|
my $i = 0; |
542 |
dpavlin |
559 |
|
543 |
dpavlin |
731 |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
544 |
|
|
$marc->add( |
545 |
|
|
id => $mfn . ( $i ? "/$i" : '' ), |
546 |
|
|
fields => $fields, |
547 |
dpavlin |
814 |
leader => WebPAC::Normalize::_get_marc_leader(), |
548 |
dpavlin |
731 |
row => $row, |
549 |
|
|
); |
550 |
|
|
$i++; |
551 |
dpavlin |
580 |
} |
552 |
|
|
|
553 |
dpavlin |
731 |
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
554 |
dpavlin |
539 |
} |
555 |
dpavlin |
731 |
} |
556 |
dpavlin |
657 |
|
557 |
dpavlin |
731 |
if ($validate) { |
558 |
|
|
my $errors = $validate->report; |
559 |
|
|
if ($errors) { |
560 |
|
|
$log->info("validation errors:\n$errors\n" ); |
561 |
|
|
print $report_fh "$errors\n" if ($report_fh); |
562 |
dpavlin |
671 |
} |
563 |
dpavlin |
731 |
} |
564 |
dpavlin |
539 |
|
565 |
dpavlin |
731 |
if ($stats) { |
566 |
|
|
my $s = $input_db->stats; |
567 |
|
|
$log->info("statistics of fields usage:\n$s"); |
568 |
|
|
print $report_fh "Statistics of fields usage:\n$s" if ($report_fh); |
569 |
dpavlin |
540 |
} |
570 |
|
|
|
571 |
dpavlin |
731 |
# close MARC file |
572 |
|
|
$marc->finish if ($marc); |
573 |
|
|
|
574 |
|
|
# close report |
575 |
|
|
close($report_fh) if ($report_fh) |
576 |
|
|
|
577 |
dpavlin |
539 |
} |
578 |
|
|
|
579 |
dpavlin |
509 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
580 |
dpavlin |
434 |
|
581 |
dpavlin |
389 |
my $dt = time() - $start_t; |
582 |
dpavlin |
511 |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
583 |
|
|
sprintf("in %.2f sec [%.2f rec/sec]", |
584 |
dpavlin |
389 |
$dt, ($total_rows / $dt) |
585 |
|
|
) |
586 |
|
|
); |
587 |
dpavlin |
255 |
|
588 |
|
|
|
589 |
dpavlin |
606 |
# end forked process |
590 |
|
|
if ($parallel) { |
591 |
|
|
$log->info("parallel process $$ finished"); |
592 |
|
|
exit(0); |
593 |
|
|
} |
594 |
|
|
|
595 |
dpavlin |
210 |
} |
596 |
|
|
|
597 |
dpavlin |
606 |
if ($parallel) { |
598 |
|
|
# wait all children to finish |
599 |
|
|
sleep(1) while wait != -1; |
600 |
|
|
$log->info("all parallel processes finished"); |
601 |
|
|
} |
602 |
|
|
|
603 |
dpavlin |
608 |
# |
604 |
|
|
# handle links or merge after indexing |
605 |
|
|
# |
606 |
dpavlin |
607 |
|
607 |
dpavlin |
608 |
if ($merge) { |
608 |
|
|
print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/; |
609 |
|
|
close($estcmd_fh); |
610 |
|
|
chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!"); |
611 |
|
|
system $estcmd_path; |
612 |
|
|
} else { |
613 |
|
|
foreach my $link (@links) { |
614 |
|
|
$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE'); |
615 |
|
|
$link->(); |
616 |
|
|
} |
617 |
dpavlin |
529 |
} |