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 |
|
|
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 |
=back |
=back |
93 |
|
|
94 |
=cut |
=cut |
106 |
my $marc_lint = 1; |
my $marc_lint = 1; |
107 |
my $marc_dump = 0; |
my $marc_dump = 0; |
108 |
|
|
109 |
|
my $parallel = 0; |
110 |
|
|
111 |
GetOptions( |
GetOptions( |
112 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
113 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
122 |
"marc-output=s" => \$marc_output, |
"marc-output=s" => \$marc_output, |
123 |
"marc-lint!" => \$marc_lint, |
"marc-lint!" => \$marc_lint, |
124 |
"marc-dump!" => \$marc_dump, |
"marc-dump!" => \$marc_dump, |
125 |
|
"parallel=i" => \$parallel, |
126 |
); |
); |
127 |
|
|
128 |
$config = LoadFile($config); |
$config = LoadFile($config); |
156 |
my @links; |
my @links; |
157 |
my $indexer; |
my $indexer; |
158 |
|
|
159 |
my $lint = new MARC::Lint if ($marc_lint); |
if ($parallel) { |
160 |
|
$log->info("Using $parallel processes for speedup"); |
161 |
|
Proc::Queue::size($parallel); |
162 |
|
} |
163 |
|
|
164 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
165 |
|
|
166 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
167 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
168 |
|
|
169 |
|
if ($parallel) { |
170 |
|
my $f=fork; |
171 |
|
if(defined ($f) and $f==0) { |
172 |
|
$log->info("Created processes $$ for speedup"); |
173 |
|
} else { |
174 |
|
next; |
175 |
|
} |
176 |
|
} |
177 |
|
|
178 |
if ($use_indexer) { |
if ($use_indexer) { |
179 |
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"); |
180 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
247 |
|
|
248 |
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)); |
249 |
|
|
250 |
my $lookup = new WebPAC::Lookup( |
my $lookup; |
251 |
lookup_file => $input->{lookup}, |
if ($input->{lookup}) { |
252 |
) if ($input->{lookup}); |
$lookup = new WebPAC::Lookup( |
253 |
|
lookup_file => $input->{lookup}, |
254 |
|
); |
255 |
|
delete( $input->{lookup} ); |
256 |
|
} |
257 |
|
|
258 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
259 |
|
|
263 |
|
|
264 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
265 |
module => $input_module, |
module => $input_module, |
266 |
code_page => $config->{webpac}->{webpac_encoding}, |
encoding => $config->{webpac}->{webpac_encoding}, |
267 |
limit => $limit || $input->{limit}, |
limit => $limit || $input->{limit}, |
268 |
offset => $offset, |
offset => $offset, |
269 |
lookup => $lookup, |
lookup_coderef => sub { |
270 |
|
my $rec = shift || return; |
271 |
|
$lookup->add( $rec ); |
272 |
|
}, |
273 |
recode => $input->{recode}, |
recode => $input->{recode}, |
274 |
stats => $stats, |
stats => $stats, |
275 |
|
modify_records => $input->{modify_records}, |
276 |
); |
); |
277 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
278 |
|
|
302 |
|
|
303 |
$log->info("Using $normalize_path for normalization..."); |
$log->info("Using $normalize_path for normalization..."); |
304 |
|
|
305 |
my $marc_fh; |
my $marc = new WebPAC::Output::MARC( |
306 |
if (my $path = $normalize->{output}) { |
path => $normalize->{output}, |
307 |
open($marc_fh, '>', $path) || |
lint => $marc_lint, |
308 |
$log->logdie("can't open MARC output $path: $!"); |
dump => $marc_dump, |
309 |
|
) if ($normalize->{output}); |
|
$log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\n"); |
|
|
} |
|
310 |
|
|
311 |
# reset position in database |
# reset position in database |
312 |
$input_db->seek(1); |
$input_db->seek(1); |
329 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
330 |
} |
} |
331 |
|
|
332 |
|
my $ds_config = dclone($db_config); |
333 |
|
|
334 |
|
# default values -> database key |
335 |
|
$ds_config->{_} = $database; |
336 |
|
|
337 |
|
# current mfn |
338 |
|
$ds_config->{_mfn} = $mfn; |
339 |
|
|
340 |
|
# attach current input |
341 |
|
$ds_config->{input} = $input; |
342 |
|
|
343 |
my $ds = WebPAC::Normalize::data_structure( |
my $ds = WebPAC::Normalize::data_structure( |
344 |
row => $row, |
row => $row, |
345 |
rules => $rules, |
rules => $rules, |
346 |
lookup => $lookup ? $lookup->lookup_hash : undef, |
lookup => $lookup ? $lookup->lookup_hash : undef, |
347 |
|
config => $ds_config, |
348 |
marc_encoding => 'utf-8', |
marc_encoding => 'utf-8', |
349 |
); |
); |
350 |
|
|
360 |
type => $config->{$use_indexer}->{type}, |
type => $config->{$use_indexer}->{type}, |
361 |
) if ($indexer && $ds); |
) if ($indexer && $ds); |
362 |
|
|
363 |
if ($marc_fh) { |
if ($marc) { |
364 |
my $marc = new MARC::Record; |
my $i = 0; |
|
$marc->encoding( 'utf-8' ); |
|
|
my @marc_fields = WebPAC::Normalize::_get_marc_fields(); |
|
|
if (! @marc_fields) { |
|
|
$log->warn("MARC record $mfn is empty, skipping"); |
|
|
} else { |
|
|
$marc->add_fields( @marc_fields ); |
|
|
|
|
|
# tweak leader |
|
|
if (my $new_l = WebPAC::Normalize::marc_leader()) { |
|
|
|
|
|
my $leader = $marc->leader; |
|
|
|
|
|
foreach my $o ( keys %$new_l ) { |
|
|
my $insert = $new_l->{$o}; |
|
|
$leader = substr($leader, 0, $o) . |
|
|
$insert . substr($leader, $o+length($insert)); |
|
|
} |
|
|
$marc->leader( $leader ); |
|
|
} |
|
|
|
|
|
if ($marc_lint) { |
|
|
$lint->check_record( $marc ); |
|
|
my $err = join( "\n", $lint->warnings ); |
|
|
$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); |
|
|
} |
|
|
|
|
|
if ($marc_dump) { |
|
|
$log->info("MARC record on MFN $mfn\tleader:", $marc->leader(), "\n", |
|
|
"Original imput row: ",dump($row), "\n", |
|
|
"Normalized MARC row: ",dump(@marc_fields), "\n", |
|
|
); |
|
|
} |
|
365 |
|
|
366 |
print $marc_fh $marc->as_usmarc; |
while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { |
367 |
|
$marc->add( |
368 |
|
id => $mfn . ( $i ? "/$i" : '' ), |
369 |
|
fields => $fields, |
370 |
|
leader => WebPAC::Normalize::marc_leader(), |
371 |
|
row => $row, |
372 |
|
); |
373 |
|
$i++; |
374 |
} |
} |
375 |
|
|
376 |
|
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1); |
377 |
} |
} |
378 |
|
|
379 |
$total_rows++; |
$total_rows++; |
382 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
383 |
|
|
384 |
# close MARC file |
# close MARC file |
385 |
close($marc_fh) if ($marc_fh); |
$marc->finish if ($marc); |
386 |
|
|
387 |
} |
} |
388 |
|
|
415 |
} |
} |
416 |
} |
} |
417 |
|
|
418 |
|
# end forked process |
419 |
|
if ($parallel) { |
420 |
|
$log->info("parallel process $$ finished"); |
421 |
|
exit(0); |
422 |
|
} |
423 |
|
|
424 |
|
} |
425 |
|
|
426 |
|
if ($parallel) { |
427 |
|
# wait all children to finish |
428 |
|
sleep(1) while wait != -1; |
429 |
|
$log->info("all parallel processes finished"); |
430 |
} |
} |
431 |
|
|
432 |
foreach my $link (@links) { |
foreach my $link (@links) { |
433 |
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
434 |
$indexer->add_link( %{ $link } ); |
$indexer->add_link( %{ $link } ); |
435 |
} |
} |
436 |
|
|