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::XML; |
15 |
|
use WebPAC::Normalize::Set; |
16 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
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; |
22 |
|
|
23 |
=head1 NAME |
=head1 NAME |
24 |
|
|
50 |
|
|
51 |
path to YAML configuration file |
path to YAML configuration file |
52 |
|
|
53 |
|
=item --force-set |
54 |
|
|
55 |
|
force conversion C<< normalize->path >> in C<config.yml> from |
56 |
|
C<.xml> to C<.pl> |
57 |
|
|
58 |
|
=item --stats |
59 |
|
|
60 |
|
disable indexing and dump statistics about field and subfield |
61 |
|
usage for each input |
62 |
|
|
63 |
=back |
=back |
64 |
|
|
65 |
=cut |
=cut |
71 |
my $config = 'conf/config.yml'; |
my $config = 'conf/config.yml'; |
72 |
my $debug = 0; |
my $debug = 0; |
73 |
my $only_db_name; |
my $only_db_name; |
74 |
|
my $force_set = 0; |
75 |
|
my $stats = 0; |
76 |
|
|
77 |
GetOptions( |
GetOptions( |
78 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
82 |
"only=s" => \$only_db_name, |
"only=s" => \$only_db_name, |
83 |
"config" => \$config, |
"config" => \$config, |
84 |
"debug" => \$debug, |
"debug" => \$debug, |
85 |
|
"force-set" => \$force_set, |
86 |
|
"stats" => \$stats, |
87 |
); |
); |
88 |
|
|
89 |
$config = LoadFile($config); |
$config = LoadFile($config); |
91 |
print "config = ",Dumper($config) if ($debug); |
print "config = ",Dumper($config) if ($debug); |
92 |
|
|
93 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->{databases}); |
94 |
|
|
95 |
|
my $log = _new WebPAC::Common()->_get_logger(); |
96 |
|
$log->info( "-" x 79 ); |
97 |
|
|
98 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
99 |
|
if ($stats) { |
100 |
|
$log->debug("option --stats disables update of indexing engine..."); |
101 |
|
$use_indexer = undef; |
102 |
|
} else { |
103 |
|
$log->info("using $use_indexer indexing engine..."); |
104 |
|
} |
105 |
|
|
106 |
my $total_rows = 0; |
my $total_rows = 0; |
107 |
my $start_t = time(); |
my $start_t = time(); |
110 |
|
|
111 |
next if ($only_db_name && $database !~ m/$only_db_name/i); |
next if ($only_db_name && $database !~ m/$only_db_name/i); |
112 |
|
|
|
my $log = _new WebPAC::Common()->_get_logger(); |
|
|
|
|
113 |
my $indexer; |
my $indexer; |
114 |
|
|
115 |
if ($use_indexer eq 'hyperestraier') { |
if ($use_indexer) { |
116 |
|
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
117 |
|
$indexer_config->{database} = $database; |
118 |
|
$indexer_config->{clean} = $clean; |
119 |
|
$indexer_config->{label} = $db_config->{name}; |
120 |
|
|
121 |
|
if ($use_indexer eq 'hyperestraier') { |
122 |
|
|
123 |
|
# open Hyper Estraier database |
124 |
|
use WebPAC::Output::Estraier '0.10'; |
125 |
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
126 |
|
|
127 |
|
} elsif ($use_indexer eq 'kinosearch') { |
128 |
|
|
129 |
|
# open KinoSearch |
130 |
|
use WebPAC::Output::KinoSearch; |
131 |
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
132 |
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
133 |
|
|
134 |
# |
} else { |
135 |
# open Hyper Estraier database |
$log->logdie("unknown use_indexer: $use_indexer"); |
136 |
# |
} |
|
|
|
|
use WebPAC::Output::Estraier '0.10'; |
|
|
my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); |
|
|
$est_config->{database} = $database; |
|
|
$est_config->{clean} = $clean; |
|
|
$est_config->{label} = $db_config->{name}; |
|
137 |
|
|
138 |
$indexer = new WebPAC::Output::Estraier( %{ $est_config } ); |
$log->logide("can't continue without valid indexer") unless ($indexer); |
|
} else { |
|
|
$log->logdie("unknown use_indexer: $use_indexer"); |
|
139 |
} |
} |
140 |
|
|
|
$log->logide("can't continue without valid indexer") unless ($indexer); |
|
141 |
|
|
142 |
# |
# |
143 |
# now WebPAC::Store |
# now WebPAC::Store |
197 |
offset => $offset, |
offset => $offset, |
198 |
lookup => $lookup, |
lookup => $lookup, |
199 |
recode => $input->{recode}, |
recode => $input->{recode}, |
200 |
|
stats => $stats, |
201 |
); |
); |
202 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
203 |
|
|
214 |
prefix => $input->{name}, |
prefix => $input->{name}, |
215 |
); |
); |
216 |
|
|
217 |
|
my $rules; |
218 |
my $normalize_path = $input->{normalize}->{path}; |
my $normalize_path = $input->{normalize}->{path}; |
219 |
|
|
220 |
|
if ($force_set) { |
221 |
|
my $new_norm_path = $normalize_path; |
222 |
|
$new_norm_path =~ s/\.xml$/.pl/; |
223 |
|
if (-e $new_norm_path) { |
224 |
|
$log->debug("--force-set replaced $normalize_path with $new_norm_path"); |
225 |
|
$normalize_path = $new_norm_path; |
226 |
|
} else { |
227 |
|
$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path"); |
228 |
|
} |
229 |
|
} |
230 |
|
|
231 |
if ($normalize_path =~ m/\.xml$/i) { |
if ($normalize_path =~ m/\.xml$/i) { |
232 |
$n->open( |
$n->open( |
233 |
tag => $input->{normalize}->{tag}, |
tag => $input->{normalize}->{tag}, |
234 |
xml_file => $input->{normalize}->{path}, |
xml_file => $normalize_path, |
235 |
); |
); |
236 |
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
237 |
$n->open_yaml( |
$n->open_yaml( |
238 |
path => $normalize_path, |
path => $normalize_path, |
239 |
tag => $input->{normalize}->{tag}, |
tag => $input->{normalize}->{tag}, |
240 |
); |
); |
241 |
|
} elsif ($normalize_path =~ m/\.(?:pl)$/i) { |
242 |
|
$n = undef; |
243 |
|
$log->info("using WebPAC::Normalize::Set to process $normalize_path"); |
244 |
|
$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
245 |
} |
} |
246 |
|
|
247 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
256 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
257 |
} |
} |
258 |
|
|
259 |
my $ds = $n->data_structure($row); |
|
260 |
|
my $ds; |
261 |
|
if ($n) { |
262 |
|
$ds = $n->data_structure($row); |
263 |
|
} else { |
264 |
|
$ds = WebPAC::Normalize::Set::data_structure( |
265 |
|
row => $row, |
266 |
|
rules => $rules, |
267 |
|
lookup => $lookup->lookup_hash, |
268 |
|
); |
269 |
|
|
270 |
|
$db->save_ds( |
271 |
|
id => $mfn, |
272 |
|
ds => $ds, |
273 |
|
prefix => $input->{name}, |
274 |
|
) if ($ds && !$stats); |
275 |
|
} |
276 |
|
|
277 |
$indexer->add( |
$indexer->add( |
278 |
id => $input->{name} . "/" . $mfn, |
id => $input->{name} . "/" . $mfn, |
279 |
ds => $ds, |
ds => $ds, |
280 |
type => $config->{$use_indexer}->{type}, |
type => $config->{$use_indexer}->{type}, |
281 |
); |
) if ($indexer); |
282 |
|
|
283 |
$total_rows++; |
$total_rows++; |
284 |
} |
} |
285 |
|
|
286 |
|
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
287 |
|
|
288 |
}; |
}; |
289 |
|
|
290 |
|
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
291 |
|
|
292 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
293 |
$log->info("$total_rows records indexed in " . |
$log->info("$total_rows records indexed in " . |
294 |
sprintf("%.2f sec [%.2f rec/sec]", |
sprintf("%.2f sec [%.2f rec/sec]", |
296 |
) |
) |
297 |
); |
); |
298 |
|
|
299 |
if ($use_indexer eq 'hyperestraier') { |
# |
300 |
# |
# add Hyper Estraier links to other databases |
301 |
# add Hyper Estraier links to other databases |
# |
302 |
# |
if (ref($db_config->{links}) eq 'ARRAY') { |
303 |
if (ref($db_config->{links}) eq 'ARRAY') { |
foreach my $link (@{ $db_config->{links} }) { |
304 |
foreach my $link (@{ $db_config->{links} }) { |
if ($use_indexer eq 'hyperestraier') { |
305 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
306 |
$indexer->add_link( |
$indexer->add_link( |
307 |
from => $database, |
from => $database, |
308 |
to => $link->{to}, |
to => $link->{to}, |
309 |
credit => $link->{credit}, |
credit => $link->{credit}, |
310 |
); |
); |
311 |
|
} else { |
312 |
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
313 |
} |
} |
314 |
} |
} |
|
} else { |
|
|
$log->warn("links not implemented for $use_indexer"); |
|
315 |
} |
} |
316 |
|
|
317 |
} |
} |