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 0.11; |
use WebPAC::Normalize 0.11; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
20 |
use Time::HiRes qw/time/; |
use Time::HiRes qw/time/; |
21 |
use File::Slurp; |
use File::Slurp; |
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 |
|
|
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; |
189 |
my $indexer; |
|
190 |
|
if ($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 |
|
|
337 |
|
|
338 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
339 |
module => $input_module, |
module => $input_module, |
340 |
code_page => $config->{webpac}->{webpac_encoding}, |
encoding => $config->{webpac}->{webpac_encoding}, |
341 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
342 |
offset => $offset, |
offset => $offset, |
343 |
lookup => $lookup, |
lookup_coderef => sub { |
344 |
|
my $rec = shift || return; |
345 |
|
$lookup->add( $rec ); |
346 |
|
}, |
347 |
recode => $input->{recode}, |
recode => $input->{recode}, |
348 |
stats => $stats, |
stats => $stats, |
349 |
|
modify_records => $input->{modify_records}, |
350 |
); |
); |
351 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
352 |
|
|
403 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
404 |
} |
} |
405 |
|
|
406 |
|
my $ds_config = dclone($db_config); |
407 |
|
|
408 |
|
# default values -> database key |
409 |
|
$ds_config->{_} = $database; |
410 |
|
|
411 |
|
# current mfn |
412 |
|
$ds_config->{_mfn} = $mfn; |
413 |
|
|
414 |
|
# attach current input |
415 |
|
$ds_config->{input} = $input; |
416 |
|
|
417 |
my $ds = WebPAC::Normalize::data_structure( |
my $ds = WebPAC::Normalize::data_structure( |
418 |
row => $row, |
row => $row, |
419 |
rules => $rules, |
rules => $rules, |
420 |
lookup => $lookup ? $lookup->lookup_hash : undef, |
lookup => $lookup ? $lookup->lookup_hash : undef, |
421 |
|
config => $ds_config, |
422 |
marc_encoding => 'utf-8', |
marc_encoding => 'utf-8', |
423 |
); |
); |
424 |
|
|
471 |
) |
) |
472 |
); |
); |
473 |
|
|
474 |
# |
|
475 |
# add Hyper Estraier links to other databases |
# end forked process |
476 |
# |
if ($parallel) { |
477 |
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
$log->info("parallel process $$ finished"); |
478 |
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}]"); |
|
|
} |
|
|
} |
|
479 |
} |
} |
480 |
|
|
481 |
} |
} |
482 |
|
|
483 |
foreach my $link (@links) { |
if ($parallel) { |
484 |
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
# wait all children to finish |
485 |
$indexer->add_link( %{ $link } ); |
sleep(1) while wait != -1; |
486 |
|
$log->info("all parallel processes finished"); |
487 |
|
} |
488 |
|
|
489 |
|
# |
490 |
|
# handle links or merge after indexing |
491 |
|
# |
492 |
|
|
493 |
|
if ($merge) { |
494 |
|
print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/; |
495 |
|
close($estcmd_fh); |
496 |
|
chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!"); |
497 |
|
system $estcmd_path; |
498 |
|
} else { |
499 |
|
foreach my $link (@links) { |
500 |
|
$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE'); |
501 |
|
$link->(); |
502 |
|
} |
503 |
} |
} |