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::Lookup 0.03; |
11 |
use WebPAC::Input 0.03; |
use WebPAC::Input 0.07; |
12 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.03; |
13 |
use WebPAC::Normalize::XML; |
use WebPAC::Normalize 0.11; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
15 |
use WebPAC::Output::Estraier 0.05; |
use WebPAC::Validate; |
16 |
|
use WebPAC::Output::MARC; |
17 |
use YAML qw/LoadFile/; |
use YAML qw/LoadFile/; |
18 |
use LWP::Simple; |
use Getopt::Long; |
19 |
|
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 |
my $limit = shift @ARGV; |
use Proc::Queue size => 1; |
26 |
|
use POSIX ":sys_wait_h"; # imports WNOHANG |
27 |
|
|
28 |
my $config = LoadFile('conf/config.yml'); |
=head1 NAME |
29 |
|
|
30 |
print "config = ",Dumper($config); |
run.pl - start WebPAC indexing |
31 |
|
|
32 |
|
B<this command will probably go away. Don't get used to it!> |
33 |
|
|
34 |
|
Options: |
35 |
|
|
36 |
|
=over 4 |
37 |
|
|
38 |
|
=item --offset 42 |
39 |
|
|
40 |
|
start loading (all) databases at offset 42 |
41 |
|
|
42 |
|
=item --limit 100 |
43 |
|
|
44 |
|
limit loading to 100 records |
45 |
|
|
46 |
|
=item --clean |
47 |
|
|
48 |
|
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 |
58 |
|
|
59 |
|
path to YAML configuration file |
60 |
|
|
61 |
|
=item --stats |
62 |
|
|
63 |
|
disable indexing and dump statistics about field and subfield |
64 |
|
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-normalize conf/normalize/mapping.pl |
71 |
|
|
72 |
|
This option specifies normalisation file for MARC creation |
73 |
|
|
74 |
|
=item --marc-output out/marc/test.marc |
75 |
|
|
76 |
|
Optional path to output file |
77 |
|
|
78 |
|
=item --marc-lint |
79 |
|
|
80 |
|
By default turned on if C<--marc-normalize> is used. You can disable lint |
81 |
|
messages with C<--no-marc-lint>. |
82 |
|
|
83 |
|
=item --marc-dump |
84 |
|
|
85 |
|
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 |
|
=back |
97 |
|
|
98 |
|
=cut |
99 |
|
|
100 |
|
my $offset; |
101 |
|
my $limit; |
102 |
|
|
103 |
|
my $clean = 0; |
104 |
|
my $config = 'conf/config.yml'; |
105 |
|
my $debug = 0; |
106 |
|
my $only_filter; |
107 |
|
my $stats = 0; |
108 |
|
my $validate_path; |
109 |
|
my ($marc_normalize, $marc_output); |
110 |
|
my $marc_lint = 1; |
111 |
|
my $marc_dump = 0; |
112 |
|
my $parallel = 0; |
113 |
|
my $only_links = 0; |
114 |
|
|
115 |
|
GetOptions( |
116 |
|
"limit=i" => \$limit, |
117 |
|
"offset=i" => \$offset, |
118 |
|
"clean" => \$clean, |
119 |
|
"one=s" => \$only_filter, |
120 |
|
"only=s" => \$only_filter, |
121 |
|
"config" => \$config, |
122 |
|
"debug+" => \$debug, |
123 |
|
"stats" => \$stats, |
124 |
|
"validate=s" => \$validate_path, |
125 |
|
"marc-normalize=s" => \$marc_normalize, |
126 |
|
"marc-output=s" => \$marc_output, |
127 |
|
"marc-lint!" => \$marc_lint, |
128 |
|
"marc-dump!" => \$marc_dump, |
129 |
|
"parallel=i" => \$parallel, |
130 |
|
"only-links!" => \$only_links, |
131 |
|
); |
132 |
|
|
133 |
|
$config = LoadFile($config); |
134 |
|
|
135 |
|
print "config = ",dump($config) if ($debug); |
136 |
|
|
137 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->{databases}); |
138 |
|
|
139 |
|
my $log = _new WebPAC::Common()->_get_logger(); |
140 |
|
$log->info( "-" x 79 ); |
141 |
|
|
142 |
|
my $validate; |
143 |
|
$validate = new WebPAC::Validate( |
144 |
|
path => $validate_path, |
145 |
|
) if ($validate_path); |
146 |
|
|
147 |
|
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
148 |
|
if ($stats) { |
149 |
|
$log->debug("option --stats disables update of indexing engine..."); |
150 |
|
$use_indexer = undef; |
151 |
|
} else { |
152 |
|
$log->info("using $use_indexer indexing engine..."); |
153 |
|
} |
154 |
|
|
155 |
|
# disable indexing when creating marc |
156 |
|
$use_indexer = undef if ($marc_normalize); |
157 |
|
|
158 |
my $total_rows = 0; |
my $total_rows = 0; |
159 |
|
my $start_t = time(); |
160 |
|
|
161 |
|
my @links; |
162 |
|
|
163 |
|
if ($parallel) { |
164 |
|
$log->info("Using $parallel processes for speedup"); |
165 |
|
Proc::Queue::size($parallel); |
166 |
|
} |
167 |
|
|
168 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
169 |
|
|
170 |
my $log = _new WebPAC::Common()->_get_logger(); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
171 |
|
next if ($only_database && $database !~ m/$only_database/i); |
172 |
|
|
173 |
# |
if ($parallel) { |
174 |
# open Hyper Estraier database |
my $f=fork; |
175 |
# |
if(defined ($f) and $f==0) { |
176 |
|
$log->info("Created processes $$ for speedup"); |
177 |
|
} else { |
178 |
|
next; |
179 |
|
} |
180 |
|
} |
181 |
|
|
182 |
my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); |
my $indexer; |
183 |
$est_config->{database} = $database; |
if ($use_indexer) { |
184 |
|
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
185 |
|
$indexer_config->{database} = $database; |
186 |
|
$indexer_config->{clean} = $clean; |
187 |
|
$indexer_config->{label} = $db_config->{name}; |
188 |
|
|
189 |
|
if ($use_indexer eq 'hyperestraier') { |
190 |
|
|
191 |
|
# open Hyper Estraier database |
192 |
|
use WebPAC::Output::Estraier '0.10'; |
193 |
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
194 |
|
|
195 |
|
} elsif ($use_indexer eq 'kinosearch') { |
196 |
|
|
197 |
|
# open KinoSearch |
198 |
|
use WebPAC::Output::KinoSearch; |
199 |
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
200 |
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
201 |
|
|
202 |
$log->info("using HyperEstraier URL $est_config->{masterurl}"); |
} else { |
203 |
|
$log->logdie("unknown use_indexer: $use_indexer"); |
204 |
|
} |
205 |
|
|
206 |
|
$log->logide("can't continue without valid indexer") unless ($indexer); |
207 |
|
} |
208 |
|
|
209 |
|
|
210 |
|
# |
211 |
|
# store Hyper Estraier links to other databases |
212 |
|
# |
213 |
|
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
214 |
|
foreach my $link (@{ $db_config->{links} }) { |
215 |
|
if ($use_indexer eq 'hyperestraier') { |
216 |
|
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
217 |
|
push @links, sub { |
218 |
|
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
219 |
|
$indexer->add_link( |
220 |
|
from => $database, |
221 |
|
to => $link->{to}, |
222 |
|
credit => $link->{credit}, |
223 |
|
); |
224 |
|
}; |
225 |
|
} else { |
226 |
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
227 |
|
} |
228 |
|
} |
229 |
|
} |
230 |
|
next if ($only_links); |
231 |
|
|
|
my $est = new WebPAC::Output::Estraier( |
|
|
%{ $est_config }, |
|
|
); |
|
232 |
|
|
233 |
# |
# |
234 |
# now WebPAC::Store |
# now WebPAC::Store |
238 |
|
|
239 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
240 |
|
|
241 |
$log->info("working on $database in $db_path"); |
if ($clean) { |
242 |
|
$log->info("creating new database '$database' in $db_path"); |
243 |
|
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
244 |
|
} else { |
245 |
|
$log->info("working on database '$database' in $db_path"); |
246 |
|
} |
247 |
|
|
248 |
my $db = new WebPAC::Store( |
my $db = new WebPAC::Store( |
249 |
path => $db_path, |
path => $db_path, |
250 |
database => $database, |
database => $database, |
251 |
debug => 1, |
debug => $debug, |
252 |
); |
); |
253 |
|
|
254 |
|
|
269 |
|
|
270 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
271 |
|
|
272 |
|
next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
273 |
|
|
274 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
275 |
|
|
276 |
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)); |
277 |
|
|
278 |
my $lookup = new WebPAC::Lookup( |
my $lookup; |
279 |
lookup_file => $input->{lookup}, |
if ($input->{lookup}) { |
280 |
); |
$lookup = new WebPAC::Lookup( |
281 |
|
lookup_file => $input->{lookup}, |
282 |
|
); |
283 |
|
delete( $input->{lookup} ); |
284 |
|
} |
285 |
|
|
286 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
287 |
|
|
288 |
$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", |
289 |
|
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
290 |
|
); |
291 |
|
|
292 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
293 |
module => $input_module, |
module => $input_module, |
294 |
code_page => $config->{webpac}->{webpac_encoding}, |
encoding => $config->{webpac}->{webpac_encoding}, |
295 |
limit_mfn => $input->{limit}, |
limit => $limit || $input->{limit}, |
296 |
lookup => $lookup, |
offset => $offset, |
297 |
|
lookup_coderef => sub { |
298 |
|
my $rec = shift || return; |
299 |
|
$lookup->add( $rec ); |
300 |
|
}, |
301 |
|
recode => $input->{recode}, |
302 |
|
stats => $stats, |
303 |
|
modify_records => $input->{modify_records}, |
304 |
); |
); |
305 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
306 |
|
|
307 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
308 |
path => $input->{path}, |
path => $input->{path}, |
309 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
310 |
|
%{ $input }, |
311 |
); |
); |
312 |
|
|
313 |
my $n = new WebPAC::Normalize::XML( |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
314 |
# filter => { 'foo' => sub { shift } }, |
@{ $input->{normalize} } : ( $input->{normalize} ); |
|
db => $db, |
|
|
lookup_regex => $lookup->regex, |
|
|
lookup => $lookup, |
|
|
prefix => $input->{name}, |
|
|
); |
|
315 |
|
|
316 |
my $normalize_path = $input->{normalize}->{path}; |
if ($marc_normalize) { |
317 |
|
@norm_array = ( { |
318 |
if ($normalize_path =~ m/\.xml$/i) { |
path => $marc_normalize, |
319 |
$n->open( |
output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc', |
320 |
tag => $input->{normalize}->{tag}, |
} ); |
|
xml_file => $input->{normalize}->{path}, |
|
|
); |
|
|
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
|
|
$n->open_yaml( |
|
|
path => $normalize_path, |
|
|
tag => $input->{normalize}->{tag}, |
|
|
); |
|
321 |
} |
} |
322 |
|
|
323 |
for ( 0 ... $input_db->size ) { |
foreach my $normalize (@norm_array) { |
324 |
|
|
325 |
my $row = $input_db->fetch || next; |
my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); |
326 |
|
|
327 |
my $mfn = $row->{'000'}->[0] || die "can't find MFN"; |
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
328 |
|
|
329 |
my $ds = $n->data_structure($row); |
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
330 |
|
|
331 |
$est->add( |
$log->info("Using $normalize_path for normalization..."); |
|
id => $input->{name} . "#" . $mfn, |
|
|
ds => $ds, |
|
|
type => $config->{hyperestraier}->{type}, |
|
|
); |
|
332 |
|
|
333 |
$total_rows++; |
my $marc = new WebPAC::Output::MARC( |
334 |
} |
path => $normalize->{output}, |
335 |
|
lint => $marc_lint, |
336 |
|
dump => $marc_dump, |
337 |
|
) if ($normalize->{output}); |
338 |
|
|
339 |
}; |
# reset position in database |
340 |
|
$input_db->seek(1); |
341 |
|
|
342 |
$log->info("$total_rows records indexed"); |
foreach my $pos ( 0 ... $input_db->size ) { |
343 |
|
|
344 |
|
my $row = $input_db->fetch || next; |
345 |
|
|
346 |
|
my $mfn = $row->{'000'}->[0]; |
347 |
|
|
348 |
|
if (! $mfn || $mfn !~ m#^\d+$#) { |
349 |
|
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
350 |
|
$mfn = $pos; |
351 |
|
push @{ $row->{'000'} }, $pos; |
352 |
|
} |
353 |
|
|
354 |
|
|
355 |
|
if ($validate) { |
356 |
|
my @errors = $validate->validate_errors( $row ); |
357 |
|
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
358 |
|
} |
359 |
|
|
360 |
|
my $ds_config = dclone($db_config); |
361 |
|
|
362 |
|
# default values -> database key |
363 |
|
$ds_config->{_} = $database; |
364 |
|
|
365 |
|
# current mfn |
366 |
|
$ds_config->{_mfn} = $mfn; |
367 |
|
|
368 |
|
# attach current input |
369 |
|
$ds_config->{input} = $input; |
370 |
|
|
371 |
|
my $ds = WebPAC::Normalize::data_structure( |
372 |
|
row => $row, |
373 |
|
rules => $rules, |
374 |
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
375 |
|
config => $ds_config, |
376 |
|
marc_encoding => 'utf-8', |
377 |
|
); |
378 |
|
|
379 |
|
$db->save_ds( |
380 |
|
id => $mfn, |
381 |
|
ds => $ds, |
382 |
|
prefix => $input->{name}, |
383 |
|
) if ($ds && !$stats); |
384 |
|
|
385 |
|
$indexer->add( |
386 |
|
id => $input->{name} . "/" . $mfn, |
387 |
|
ds => $ds, |
388 |
|
type => $config->{$use_indexer}->{type}, |
389 |
|
) if ($indexer && $ds); |
390 |
|
|
391 |
|
if ($marc) { |
392 |
|
my $i = 0; |
393 |
|
|
394 |
|
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
395 |
|
$marc->add( |
396 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
397 |
|
fields => $fields, |
398 |
|
leader => WebPAC::Normalize::marc_leader(), |
399 |
|
row => $row, |
400 |
|
); |
401 |
|
$i++; |
402 |
|
} |
403 |
|
|
404 |
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
405 |
|
} |
406 |
|
|
407 |
|
$total_rows++; |
408 |
|
} |
409 |
|
|
410 |
|
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
411 |
|
|
412 |
|
# close MARC file |
413 |
|
$marc->finish if ($marc); |
414 |
|
|
|
# |
|
|
# add Hyper Estraier links to other databases |
|
|
# |
|
|
if (ref($db_config->{links}) eq 'ARRAY') { |
|
|
foreach my $link (@{ $db_config->{links} }) { |
|
|
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
|
|
$est->add_link( |
|
|
from => $database, |
|
|
to => $link->{to}, |
|
|
credit => $link->{credit}, |
|
|
); |
|
415 |
} |
} |
416 |
|
|
417 |
} |
} |
418 |
|
|
419 |
|
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
420 |
|
|
421 |
|
my $dt = time() - $start_t; |
422 |
|
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
423 |
|
sprintf("in %.2f sec [%.2f rec/sec]", |
424 |
|
$dt, ($total_rows / $dt) |
425 |
|
) |
426 |
|
); |
427 |
|
|
428 |
|
|
429 |
|
# end forked process |
430 |
|
if ($parallel) { |
431 |
|
$log->info("parallel process $$ finished"); |
432 |
|
exit(0); |
433 |
|
} |
434 |
|
|
435 |
|
} |
436 |
|
|
437 |
|
if ($parallel) { |
438 |
|
# wait all children to finish |
439 |
|
sleep(1) while wait != -1; |
440 |
|
$log->info("all parallel processes finished"); |
441 |
|
} |
442 |
|
|
443 |
|
|
444 |
|
foreach my $link (@links) { |
445 |
|
$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE'); |
446 |
|
$link->(); |
447 |
} |
} |
448 |
|
|