11 |
use WebPAC::Lookup; |
use WebPAC::Lookup; |
12 |
use WebPAC::Input 0.03; |
use WebPAC::Input 0.03; |
13 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.03; |
14 |
use WebPAC::Normalize::XML; |
use WebPAC::Normalize; |
|
use WebPAC::Normalize::Set; |
|
15 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
16 |
use WebPAC::Validate; |
use WebPAC::Validate; |
17 |
use YAML qw/LoadFile/; |
use YAML qw/LoadFile/; |
53 |
|
|
54 |
path to YAML configuration file |
path to YAML configuration file |
55 |
|
|
|
=item --force-set |
|
|
|
|
|
force conversion C<< normalize->path >> in C<config.yml> from |
|
|
C<.xml> to C<.pl> |
|
|
|
|
56 |
=item --stats |
=item --stats |
57 |
|
|
58 |
disable indexing and dump statistics about field and subfield |
disable indexing and dump statistics about field and subfield |
73 |
my $config = 'conf/config.yml'; |
my $config = 'conf/config.yml'; |
74 |
my $debug = 0; |
my $debug = 0; |
75 |
my $only_filter; |
my $only_filter; |
|
my $force_set = 0; |
|
76 |
my $stats = 0; |
my $stats = 0; |
77 |
my $validate_path; |
my $validate_path; |
78 |
|
|
84 |
"only=s" => \$only_filter, |
"only=s" => \$only_filter, |
85 |
"config" => \$config, |
"config" => \$config, |
86 |
"debug" => \$debug, |
"debug" => \$debug, |
|
"force-set" => \$force_set, |
|
87 |
"stats" => \$stats, |
"stats" => \$stats, |
88 |
"validate=s" => \$validate_path, |
"validate=s" => \$validate_path, |
89 |
); |
); |
113 |
my $total_rows = 0; |
my $total_rows = 0; |
114 |
my $start_t = time(); |
my $start_t = time(); |
115 |
|
|
116 |
|
my @links; |
117 |
|
my $indexer; |
118 |
|
|
119 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
120 |
|
|
121 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
122 |
next if ($only_database && $database !~ m/$only_database/i); |
next if ($only_database && $database !~ m/$only_database/i); |
123 |
|
|
|
my $indexer; |
|
|
|
|
124 |
if ($use_indexer) { |
if ($use_indexer) { |
125 |
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"); |
126 |
$indexer_config->{database} = $database; |
$indexer_config->{database} = $database; |
195 |
|
|
196 |
my $lookup = new WebPAC::Lookup( |
my $lookup = new WebPAC::Lookup( |
197 |
lookup_file => $input->{lookup}, |
lookup_file => $input->{lookup}, |
198 |
); |
) if ($input->{lookup}); |
199 |
|
|
200 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
201 |
|
|
202 |
$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module lookup '$input->{lookup}'"); |
$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module", |
203 |
|
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
204 |
|
); |
205 |
|
|
206 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
207 |
module => $input_module, |
module => $input_module, |
217 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
218 |
path => $input->{path}, |
path => $input->{path}, |
219 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
220 |
); |
%{ $input }, |
|
|
|
|
my $n = new WebPAC::Normalize::XML( |
|
|
# filter => { 'foo' => sub { shift } }, |
|
|
db => $db, |
|
|
lookup_regex => $lookup->regex, |
|
|
lookup => $lookup, |
|
|
prefix => $input->{name}, |
|
221 |
); |
); |
222 |
|
|
223 |
my $rules; |
my $rules; |
224 |
my $normalize_path = $input->{normalize}->{path}; |
my $normalize_path = $input->{normalize}->{path}; |
225 |
|
|
226 |
if ($force_set) { |
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
|
my $new_norm_path = $normalize_path; |
|
|
$new_norm_path =~ s/\.xml$/.pl/; |
|
|
if (-e $new_norm_path) { |
|
|
$log->debug("--force-set replaced $normalize_path with $new_norm_path"); |
|
|
$normalize_path = $new_norm_path; |
|
|
} else { |
|
|
$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path"); |
|
|
} |
|
|
} |
|
227 |
|
|
228 |
if ($normalize_path =~ m/\.xml$/i) { |
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
|
$n->open( |
|
|
tag => $input->{normalize}->{tag}, |
|
|
xml_file => $normalize_path, |
|
|
); |
|
|
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
|
|
$n->open_yaml( |
|
|
path => $normalize_path, |
|
|
tag => $input->{normalize}->{tag}, |
|
|
); |
|
|
} elsif ($normalize_path =~ m/\.(?:pl)$/i) { |
|
|
$n = undef; |
|
|
$log->info("using WebPAC::Normalize::Set to process $normalize_path"); |
|
|
$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
|
|
} |
|
229 |
|
|
230 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
231 |
|
|
246 |
} |
} |
247 |
|
|
248 |
|
|
249 |
my $ds; |
my $ds = WebPAC::Normalize::data_structure( |
250 |
if ($n) { |
row => $row, |
251 |
$ds = $n->data_structure($row); |
rules => $rules, |
252 |
} else { |
lookup => $lookup ? $lookup->lookup_hash : undef, |
253 |
$ds = WebPAC::Normalize::Set::data_structure( |
); |
254 |
row => $row, |
|
255 |
rules => $rules, |
$db->save_ds( |
256 |
lookup => $lookup->lookup_hash, |
id => $mfn, |
257 |
); |
ds => $ds, |
258 |
|
prefix => $input->{name}, |
259 |
$db->save_ds( |
) if ($ds && !$stats); |
|
id => $mfn, |
|
|
ds => $ds, |
|
|
prefix => $input->{name}, |
|
|
) if ($ds && !$stats); |
|
|
} |
|
260 |
|
|
261 |
$indexer->add( |
$indexer->add( |
262 |
id => $input->{name} . "/" . $mfn, |
id => $input->{name} . "/" . $mfn, |
286 |
if (ref($db_config->{links}) eq 'ARRAY') { |
if (ref($db_config->{links}) eq 'ARRAY') { |
287 |
foreach my $link (@{ $db_config->{links} }) { |
foreach my $link (@{ $db_config->{links} }) { |
288 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer eq 'hyperestraier') { |
289 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
290 |
$indexer->add_link( |
push @links, { |
291 |
from => $database, |
from => $database, |
292 |
to => $link->{to}, |
to => $link->{to}, |
293 |
credit => $link->{credit}, |
credit => $link->{credit}, |
294 |
); |
}; |
295 |
} else { |
} else { |
296 |
$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}]"); |
297 |
} |
} |
300 |
|
|
301 |
} |
} |
302 |
|
|
303 |
|
foreach my $link (@links) { |
304 |
|
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
305 |
|
$indexer->add_link( %{ $link } ); |
306 |
|
} |