14 |
use WebPAC::Normalize::XML; |
use WebPAC::Normalize::XML; |
15 |
use WebPAC::Normalize::Set; |
use WebPAC::Normalize::Set; |
16 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
17 |
|
use WebPAC::Validate; |
18 |
use YAML qw/LoadFile/; |
use YAML qw/LoadFile/; |
19 |
use Getopt::Long; |
use Getopt::Long; |
20 |
use File::Path; |
use File::Path; |
64 |
disable indexing and dump statistics about field and subfield |
disable indexing and dump statistics about field and subfield |
65 |
usage for each input |
usage for each input |
66 |
|
|
67 |
|
=item --validate path/to/validation_file |
68 |
|
|
69 |
|
turn on extra validation of imput records, see L<WebPAC::Validation> |
70 |
|
|
71 |
=back |
=back |
72 |
|
|
73 |
=cut |
=cut |
81 |
my $only_filter; |
my $only_filter; |
82 |
my $force_set = 0; |
my $force_set = 0; |
83 |
my $stats = 0; |
my $stats = 0; |
84 |
|
my $validate_path; |
85 |
|
|
86 |
GetOptions( |
GetOptions( |
87 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
93 |
"debug" => \$debug, |
"debug" => \$debug, |
94 |
"force-set" => \$force_set, |
"force-set" => \$force_set, |
95 |
"stats" => \$stats, |
"stats" => \$stats, |
96 |
|
"validate=s" => \$validate_path, |
97 |
); |
); |
98 |
|
|
99 |
$config = LoadFile($config); |
$config = LoadFile($config); |
105 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
106 |
$log->info( "-" x 79 ); |
$log->info( "-" x 79 ); |
107 |
|
|
108 |
|
my $validate; |
109 |
|
$validate = new WebPAC::Validate( |
110 |
|
path => $validate_path, |
111 |
|
) if ($validate_path); |
112 |
|
|
113 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
114 |
if ($stats) { |
if ($stats) { |
115 |
$log->debug("option --stats disables update of indexing engine..."); |
$log->debug("option --stats disables update of indexing engine..."); |
121 |
my $total_rows = 0; |
my $total_rows = 0; |
122 |
my $start_t = time(); |
my $start_t = time(); |
123 |
|
|
124 |
|
my @links; |
125 |
|
my $indexer; |
126 |
|
|
127 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
128 |
|
|
129 |
my ($only_database,$only_input) = split(m#/#, $only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
130 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
131 |
|
|
|
my $indexer; |
|
|
|
|
132 |
if ($use_indexer) { |
if ($use_indexer) { |
133 |
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"); |
134 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
165 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
166 |
|
|
167 |
if ($clean) { |
if ($clean) { |
168 |
$log->info("creating new database $database in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
169 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
170 |
} else { |
} else { |
171 |
$log->debug("working on $database in $db_path"); |
$log->info("working on database '$database' in $db_path"); |
172 |
} |
} |
173 |
|
|
174 |
my $db = new WebPAC::Store( |
my $db = new WebPAC::Store( |
195 |
|
|
196 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
197 |
|
|
198 |
next if ($only_input && $input->{name} =~ m#$only_input#i || $input->{type} =~ m#$only_input#i); |
next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
199 |
|
|
200 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
201 |
|
|
203 |
|
|
204 |
my $lookup = new WebPAC::Lookup( |
my $lookup = new WebPAC::Lookup( |
205 |
lookup_file => $input->{lookup}, |
lookup_file => $input->{lookup}, |
206 |
); |
) if ($input->{lookup}); |
207 |
|
|
208 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
209 |
|
|
210 |
$log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'"); |
$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module", |
211 |
|
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
212 |
|
); |
213 |
|
|
214 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
215 |
module => $input_module, |
module => $input_module, |
225 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
226 |
path => $input->{path}, |
path => $input->{path}, |
227 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
228 |
|
%{ $input }, |
229 |
); |
); |
230 |
|
|
231 |
my $n = new WebPAC::Normalize::XML( |
my $n = new WebPAC::Normalize::XML( |
232 |
# filter => { 'foo' => sub { shift } }, |
# filter => { 'foo' => sub { shift } }, |
233 |
db => $db, |
db => $db, |
234 |
lookup_regex => $lookup->regex, |
lookup_regex => $lookup ? $lookup->regex : undef, |
235 |
lookup => $lookup, |
lookup => $lookup, |
236 |
prefix => $input->{name}, |
prefix => $input->{name}, |
237 |
); |
); |
278 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
279 |
} |
} |
280 |
|
|
281 |
|
|
282 |
|
if ($validate) { |
283 |
|
my @errors = $validate->validate_errors( $row ); |
284 |
|
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
285 |
|
} |
286 |
|
|
287 |
|
|
288 |
my $ds; |
my $ds; |
289 |
if ($n) { |
if ($n) { |
292 |
$ds = WebPAC::Normalize::Set::data_structure( |
$ds = WebPAC::Normalize::Set::data_structure( |
293 |
row => $row, |
row => $row, |
294 |
rules => $rules, |
rules => $rules, |
295 |
lookup => $lookup->lookup_hash, |
lookup => $lookup ? $lookup->lookup_hash : undef, |
296 |
); |
); |
297 |
|
|
298 |
$db->save_ds( |
$db->save_ds( |
318 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
319 |
|
|
320 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
321 |
$log->info("$total_rows records indexed in " . |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
322 |
sprintf("%.2f sec [%.2f rec/sec]", |
sprintf("in %.2f sec [%.2f rec/sec]", |
323 |
$dt, ($total_rows / $dt) |
$dt, ($total_rows / $dt) |
324 |
) |
) |
325 |
); |
); |
330 |
if (ref($db_config->{links}) eq 'ARRAY') { |
if (ref($db_config->{links}) eq 'ARRAY') { |
331 |
foreach my $link (@{ $db_config->{links} }) { |
foreach my $link (@{ $db_config->{links} }) { |
332 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer eq 'hyperestraier') { |
333 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
334 |
$indexer->add_link( |
push @links, { |
335 |
from => $database, |
from => $database, |
336 |
to => $link->{to}, |
to => $link->{to}, |
337 |
credit => $link->{credit}, |
credit => $link->{credit}, |
338 |
); |
}; |
339 |
} else { |
} else { |
340 |
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
341 |
} |
} |
344 |
|
|
345 |
} |
} |
346 |
|
|
347 |
|
foreach my $link (@links) { |
348 |
|
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
349 |
|
$indexer->add_link( %{ $link } ); |
350 |
|
} |