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 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; |
21 |
use Time::HiRes qw/time/; |
use Time::HiRes qw/time/; |
22 |
|
use File::Slurp; |
23 |
|
|
24 |
=head1 NAME |
=head1 NAME |
25 |
|
|
43 |
|
|
44 |
remove database and Hyper Estraier index before indexing |
remove database and Hyper Estraier index before indexing |
45 |
|
|
46 |
=item --only=database_name |
=item --only=database_name/input_filter |
47 |
|
|
48 |
reindex just single database (legacy name is --one) |
reindex just single database (legacy name is --one) |
49 |
|
|
50 |
|
C</input_filter> is optional part which can be C<name> |
51 |
|
or C<type> from input |
52 |
|
|
53 |
=item --config conf/config.yml |
=item --config conf/config.yml |
54 |
|
|
55 |
path to YAML configuration file |
path to YAML configuration file |
56 |
|
|
57 |
|
=item --force-set |
58 |
|
|
59 |
|
force conversion C<< normalize->path >> in C<config.yml> from |
60 |
|
C<.xml> to C<.pl> |
61 |
|
|
62 |
|
=item --stats |
63 |
|
|
64 |
|
disable indexing and dump statistics about field and subfield |
65 |
|
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 |
78 |
my $clean = 0; |
my $clean = 0; |
79 |
my $config = 'conf/config.yml'; |
my $config = 'conf/config.yml'; |
80 |
my $debug = 0; |
my $debug = 0; |
81 |
my $only_db_name; |
my $only_filter; |
82 |
|
my $force_set = 0; |
83 |
|
my $stats = 0; |
84 |
|
my $validate_path; |
85 |
|
|
86 |
GetOptions( |
GetOptions( |
87 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
88 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
89 |
"clean" => \$clean, |
"clean" => \$clean, |
90 |
"one=s" => \$only_db_name, |
"one=s" => \$only_filter, |
91 |
"only=s" => \$only_db_name, |
"only=s" => \$only_filter, |
92 |
"config" => \$config, |
"config" => \$config, |
93 |
"debug" => \$debug, |
"debug" => \$debug, |
94 |
|
"force-set" => \$force_set, |
95 |
|
"stats" => \$stats, |
96 |
|
"validate=s" => \$validate_path, |
97 |
); |
); |
98 |
|
|
99 |
$config = LoadFile($config); |
$config = LoadFile($config); |
103 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->{databases}); |
104 |
|
|
105 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
106 |
|
$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 |
$log->info("using $use_indexer indexing engine..."); |
if ($stats) { |
115 |
|
$log->debug("option --stats disables update of indexing engine..."); |
116 |
|
$use_indexer = undef; |
117 |
|
} else { |
118 |
|
$log->info("using $use_indexer indexing engine..."); |
119 |
|
} |
120 |
|
|
121 |
my $total_rows = 0; |
my $total_rows = 0; |
122 |
my $start_t = time(); |
my $start_t = time(); |
123 |
|
|
124 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
125 |
|
|
126 |
next if ($only_db_name && $database !~ m/$only_db_name/i); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
127 |
|
next if ($only_database && $database !~ m/$only_database/i); |
128 |
|
|
129 |
my $indexer; |
my $indexer; |
130 |
|
|
131 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
if ($use_indexer) { |
132 |
$indexer_config->{database} = $database; |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
133 |
$indexer_config->{clean} = $clean; |
$indexer_config->{database} = $database; |
134 |
$indexer_config->{label} = $db_config->{name}; |
$indexer_config->{clean} = $clean; |
135 |
|
$indexer_config->{label} = $db_config->{name}; |
136 |
if ($use_indexer eq 'hyperestraier') { |
|
137 |
|
if ($use_indexer eq 'hyperestraier') { |
138 |
# open Hyper Estraier database |
|
139 |
use WebPAC::Output::Estraier '0.10'; |
# open Hyper Estraier database |
140 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
use WebPAC::Output::Estraier '0.10'; |
141 |
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
142 |
} elsif ($use_indexer eq 'kinosearch') { |
|
143 |
|
} elsif ($use_indexer eq 'kinosearch') { |
144 |
# open KinoSearch |
|
145 |
use WebPAC::Output::KinoSearch; |
# open KinoSearch |
146 |
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
use WebPAC::Output::KinoSearch; |
147 |
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
148 |
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
149 |
|
|
150 |
} else { |
} else { |
151 |
$log->logdie("unknown use_indexer: $use_indexer"); |
$log->logdie("unknown use_indexer: $use_indexer"); |
152 |
|
} |
153 |
|
|
154 |
|
$log->logide("can't continue without valid indexer") unless ($indexer); |
155 |
} |
} |
156 |
|
|
|
$log->logide("can't continue without valid indexer") unless ($indexer); |
|
157 |
|
|
158 |
# |
# |
159 |
# now WebPAC::Store |
# now WebPAC::Store |
164 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
165 |
|
|
166 |
if ($clean) { |
if ($clean) { |
167 |
$log->info("creating new database $database in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
168 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
169 |
} else { |
} else { |
170 |
$log->debug("working on $database in $db_path"); |
$log->info("working on database '$database' in $db_path"); |
171 |
} |
} |
172 |
|
|
173 |
my $db = new WebPAC::Store( |
my $db = new WebPAC::Store( |
194 |
|
|
195 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
196 |
|
|
197 |
|
next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
198 |
|
|
199 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
200 |
|
|
201 |
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)); |
202 |
|
|
203 |
my $lookup = new WebPAC::Lookup( |
my $lookup = new WebPAC::Lookup( |
204 |
lookup_file => $input->{lookup}, |
lookup_file => $input->{lookup}, |
205 |
); |
) if ($input->{lookup}); |
206 |
|
|
207 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
208 |
|
|
209 |
$log->info("working on input '$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", |
210 |
|
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
211 |
|
); |
212 |
|
|
213 |
my $input_db = new WebPAC::Input( |
my $input_db = new WebPAC::Input( |
214 |
module => $input_module, |
module => $input_module, |
217 |
offset => $offset, |
offset => $offset, |
218 |
lookup => $lookup, |
lookup => $lookup, |
219 |
recode => $input->{recode}, |
recode => $input->{recode}, |
220 |
|
stats => $stats, |
221 |
); |
); |
222 |
$log->logdie("can't create input using $input_module") unless ($input); |
$log->logdie("can't create input using $input_module") unless ($input); |
223 |
|
|
224 |
my $maxmfn = $input_db->open( |
my $maxmfn = $input_db->open( |
225 |
path => $input->{path}, |
path => $input->{path}, |
226 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
227 |
|
%{ $input }, |
228 |
); |
); |
229 |
|
|
230 |
my $n = new WebPAC::Normalize::XML( |
my $n = new WebPAC::Normalize::XML( |
231 |
# filter => { 'foo' => sub { shift } }, |
# filter => { 'foo' => sub { shift } }, |
232 |
db => $db, |
db => $db, |
233 |
lookup_regex => $lookup->regex, |
lookup_regex => $lookup ? $lookup->regex : undef, |
234 |
lookup => $lookup, |
lookup => $lookup, |
235 |
prefix => $input->{name}, |
prefix => $input->{name}, |
236 |
); |
); |
237 |
|
|
238 |
|
my $rules; |
239 |
my $normalize_path = $input->{normalize}->{path}; |
my $normalize_path = $input->{normalize}->{path}; |
240 |
|
|
241 |
|
if ($force_set) { |
242 |
|
my $new_norm_path = $normalize_path; |
243 |
|
$new_norm_path =~ s/\.xml$/.pl/; |
244 |
|
if (-e $new_norm_path) { |
245 |
|
$log->debug("--force-set replaced $normalize_path with $new_norm_path"); |
246 |
|
$normalize_path = $new_norm_path; |
247 |
|
} else { |
248 |
|
$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path"); |
249 |
|
} |
250 |
|
} |
251 |
|
|
252 |
if ($normalize_path =~ m/\.xml$/i) { |
if ($normalize_path =~ m/\.xml$/i) { |
253 |
$n->open( |
$n->open( |
254 |
tag => $input->{normalize}->{tag}, |
tag => $input->{normalize}->{tag}, |
255 |
xml_file => $input->{normalize}->{path}, |
xml_file => $normalize_path, |
256 |
); |
); |
257 |
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
258 |
$n->open_yaml( |
$n->open_yaml( |
259 |
path => $normalize_path, |
path => $normalize_path, |
260 |
tag => $input->{normalize}->{tag}, |
tag => $input->{normalize}->{tag}, |
261 |
); |
); |
262 |
|
} elsif ($normalize_path =~ m/\.(?:pl)$/i) { |
263 |
|
$n = undef; |
264 |
|
$log->info("using WebPAC::Normalize::Set to process $normalize_path"); |
265 |
|
$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
266 |
} |
} |
267 |
|
|
268 |
foreach my $pos ( 0 ... $input_db->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
277 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
278 |
} |
} |
279 |
|
|
280 |
my $ds = $n->data_structure($row); |
|
281 |
|
if ($validate) { |
282 |
|
my @errors = $validate->validate_errors( $row ); |
283 |
|
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
284 |
|
} |
285 |
|
|
286 |
|
|
287 |
|
my $ds; |
288 |
|
if ($n) { |
289 |
|
$ds = $n->data_structure($row); |
290 |
|
} else { |
291 |
|
$ds = WebPAC::Normalize::Set::data_structure( |
292 |
|
row => $row, |
293 |
|
rules => $rules, |
294 |
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
295 |
|
); |
296 |
|
|
297 |
|
$db->save_ds( |
298 |
|
id => $mfn, |
299 |
|
ds => $ds, |
300 |
|
prefix => $input->{name}, |
301 |
|
) if ($ds && !$stats); |
302 |
|
} |
303 |
|
|
304 |
$indexer->add( |
$indexer->add( |
305 |
id => $input->{name} . "/" . $mfn, |
id => $input->{name} . "/" . $mfn, |
306 |
ds => $ds, |
ds => $ds, |
307 |
type => $config->{$use_indexer}->{type}, |
type => $config->{$use_indexer}->{type}, |
308 |
); |
) if ($indexer); |
309 |
|
|
310 |
$total_rows++; |
$total_rows++; |
311 |
} |
} |
312 |
|
|
313 |
|
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
314 |
|
|
315 |
}; |
}; |
316 |
|
|
317 |
|
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
318 |
|
|
319 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
320 |
$log->info("$total_rows records indexed in " . |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
321 |
sprintf("%.2f sec [%.2f rec/sec]", |
sprintf("in %.2f sec [%.2f rec/sec]", |
322 |
$dt, ($total_rows / $dt) |
$dt, ($total_rows / $dt) |
323 |
) |
) |
324 |
); |
); |