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; |
use WebPAC::Normalize 0.11; |
14 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
15 |
use WebPAC::Validate; |
use WebPAC::Validate; |
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 |
|
|
78 |
=item --marc-lint |
=item --marc-lint |
79 |
|
|
80 |
By default turned on if C<--marc-normalize> is used. You can disable lint |
By default turned on if C<--marc-normalize> is used. You can disable lint |
81 |
messages with C<--no-arc-lint>. |
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 |
=back |
97 |
|
|
108 |
my $validate_path; |
my $validate_path; |
109 |
my ($marc_normalize, $marc_output); |
my ($marc_normalize, $marc_output); |
110 |
my $marc_lint = 1; |
my $marc_lint = 1; |
111 |
|
my $marc_dump = 0; |
112 |
|
my $parallel = 0; |
113 |
|
my $only_links = 0; |
114 |
|
|
115 |
GetOptions( |
GetOptions( |
116 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
119 |
"one=s" => \$only_filter, |
"one=s" => \$only_filter, |
120 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
121 |
"config" => \$config, |
"config" => \$config, |
122 |
"debug" => \$debug, |
"debug+" => \$debug, |
123 |
"stats" => \$stats, |
"stats" => \$stats, |
124 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
125 |
"marc-normalize=s" => \$marc_normalize, |
"marc-normalize=s" => \$marc_normalize, |
126 |
"marc-output=s" => \$marc_output, |
"marc-output=s" => \$marc_output, |
127 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
128 |
|
"marc-dump!" => \$marc_dump, |
129 |
|
"parallel=i" => \$parallel, |
130 |
|
"only-links!" => \$only_links, |
131 |
); |
); |
132 |
|
|
133 |
$config = LoadFile($config); |
$config = LoadFile($config); |
159 |
my $start_t = time(); |
my $start_t = time(); |
160 |
|
|
161 |
my @links; |
my @links; |
|
my $indexer; |
|
162 |
|
|
163 |
my $lint = new MARC::Lint if ($marc_lint); |
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 %{ $config->{databases} }) { |
169 |
|
|
170 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
171 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
172 |
|
|
173 |
|
if ($parallel) { |
174 |
|
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 $indexer; |
183 |
if ($use_indexer) { |
if ($use_indexer) { |
184 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
185 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
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 |
|
|
232 |
|
|
233 |
|
# |
234 |
# now WebPAC::Store |
# now WebPAC::Store |
235 |
# |
# |
236 |
my $abs_path = abs_path($0); |
my $abs_path = abs_path($0); |
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 |
) if ($input->{lookup}); |
$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 |
|
|
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 => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
296 |
offset => $offset, |
offset => $offset, |
297 |
lookup => $lookup, |
lookup_coderef => sub { |
298 |
|
my $rec = shift || return; |
299 |
|
$lookup->add( $rec ); |
300 |
|
}, |
301 |
recode => $input->{recode}, |
recode => $input->{recode}, |
302 |
stats => $stats, |
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 |
|
|
330 |
|
|
331 |
$log->info("Using $normalize_path for normalization..."); |
$log->info("Using $normalize_path for normalization..."); |
332 |
|
|
333 |
my $marc_fh; |
my $marc = new WebPAC::Output::MARC( |
334 |
if (my $path = $normalize->{output}) { |
path => $normalize->{output}, |
335 |
open($marc_fh, '>', $path) || |
lint => $marc_lint, |
336 |
$log->logdie("can't open MARC output $path: $!"); |
dump => $marc_dump, |
337 |
|
) if ($normalize->{output}); |
|
$log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\n"); |
|
|
} |
|
338 |
|
|
339 |
# reset position in database |
# reset position in database |
340 |
$input_db->seek(1); |
$input_db->seek(1); |
357 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
$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( |
my $ds = WebPAC::Normalize::data_structure( |
372 |
row => $row, |
row => $row, |
373 |
rules => $rules, |
rules => $rules, |
374 |
lookup => $lookup ? $lookup->lookup_hash : undef, |
lookup => $lookup ? $lookup->lookup_hash : undef, |
375 |
|
config => $ds_config, |
376 |
marc_encoding => 'utf-8', |
marc_encoding => 'utf-8', |
377 |
); |
); |
378 |
|
|
388 |
type => $config->{$use_indexer}->{type}, |
type => $config->{$use_indexer}->{type}, |
389 |
) if ($indexer && $ds); |
) if ($indexer && $ds); |
390 |
|
|
391 |
if ($marc_fh) { |
if ($marc) { |
392 |
my $marc = new MARC::Record; |
my $i = 0; |
393 |
$marc->encoding( 'utf-8' ); |
|
394 |
my @marc_fields = WebPAC::Normalize::_get_marc_fields(); |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
395 |
if (! @marc_fields) { |
$marc->add( |
396 |
$log->warn("MARC record $mfn is empty, skipping"); |
id => $mfn . ( $i ? "/$i" : '' ), |
397 |
} else { |
fields => $fields, |
398 |
$marc->add_fields( @marc_fields ); |
leader => WebPAC::Normalize::marc_leader(), |
399 |
if ($marc_lint) { |
row => $row, |
400 |
$lint->check_record( $marc ); |
); |
401 |
my $err = join( "\n", $lint->warnings ); |
$i++; |
|
$log->error("MARC lint detected warning on MFN $mfn\n", |
|
|
"Original imput row: ",dump($row), "\n", |
|
|
"Normalized MARC row: ",dump(@marc_fields), "\n", |
|
|
"MARC lint warnings: ",$err,"\n" |
|
|
) if ($err); |
|
|
} |
|
|
print $marc_fh $marc->as_usmarc; |
|
402 |
} |
} |
403 |
|
|
404 |
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
405 |
} |
} |
406 |
|
|
407 |
$total_rows++; |
$total_rows++; |
410 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
411 |
|
|
412 |
# close MARC file |
# close MARC file |
413 |
close($marc_fh) if ($marc_fh); |
$marc->finish if ($marc); |
414 |
|
|
415 |
} |
} |
416 |
|
|
425 |
) |
) |
426 |
); |
); |
427 |
|
|
428 |
# |
|
429 |
# add Hyper Estraier links to other databases |
# end forked process |
430 |
# |
if ($parallel) { |
431 |
if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { |
$log->info("parallel process $$ finished"); |
432 |
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}]"); |
|
|
} |
|
|
} |
|
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) { |
foreach my $link (@links) { |
445 |
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE'); |
446 |
$indexer->add_link( %{ $link } ); |
$link->(); |
447 |
} |
} |
448 |
|
|