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; |
17 |
use YAML qw/LoadFile/; |
use YAML qw/LoadFile/; |
18 |
use Getopt::Long; |
use Getopt::Long; |
19 |
use File::Path; |
use File::Path; |
42 |
|
|
43 |
remove database and Hyper Estraier index before indexing |
remove database and Hyper Estraier index before indexing |
44 |
|
|
45 |
=item --only=database_name |
=item --only=database_name/input_filter |
46 |
|
|
47 |
reindex just single database (legacy name is --one) |
reindex just single database (legacy name is --one) |
48 |
|
|
49 |
|
C</input_filter> is optional part which can be C<name> |
50 |
|
or C<type> from input |
51 |
|
|
52 |
=item --config conf/config.yml |
=item --config conf/config.yml |
53 |
|
|
54 |
path to YAML configuration file |
path to YAML configuration file |
55 |
|
|
56 |
=item --force-set |
=item --stats |
57 |
|
|
58 |
force conversion C<< normalize->path >> in C<config.yml> from |
disable indexing and dump statistics about field and subfield |
59 |
C<.xml> to C<.pl> |
usage for each input |
60 |
|
|
61 |
=item --stats |
=item --validate path/to/validation_file |
62 |
|
|
63 |
dump statistics about used fields and subfields in each input |
turn on extra validation of imput records, see L<WebPAC::Validation> |
64 |
|
|
65 |
=back |
=back |
66 |
|
|
72 |
my $clean = 0; |
my $clean = 0; |
73 |
my $config = 'conf/config.yml'; |
my $config = 'conf/config.yml'; |
74 |
my $debug = 0; |
my $debug = 0; |
75 |
my $only_db_name; |
my $only_filter; |
|
my $force_set = 0; |
|
76 |
my $stats = 0; |
my $stats = 0; |
77 |
|
my $validate_path; |
78 |
|
|
79 |
GetOptions( |
GetOptions( |
80 |
"limit=i" => \$limit, |
"limit=i" => \$limit, |
81 |
"offset=i" => \$offset, |
"offset=i" => \$offset, |
82 |
"clean" => \$clean, |
"clean" => \$clean, |
83 |
"one=s" => \$only_db_name, |
"one=s" => \$only_filter, |
84 |
"only=s" => \$only_db_name, |
"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, |
89 |
); |
); |
90 |
|
|
91 |
$config = LoadFile($config); |
$config = LoadFile($config); |
95 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->{databases}); |
96 |
|
|
97 |
my $log = _new WebPAC::Common()->_get_logger(); |
my $log = _new WebPAC::Common()->_get_logger(); |
98 |
|
$log->info( "-" x 79 ); |
99 |
|
|
100 |
|
my $validate; |
101 |
|
$validate = new WebPAC::Validate( |
102 |
|
path => $validate_path, |
103 |
|
) if ($validate_path); |
104 |
|
|
105 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
106 |
$log->info("using $use_indexer indexing engine..."); |
if ($stats) { |
107 |
|
$log->debug("option --stats disables update of indexing engine..."); |
108 |
|
$use_indexer = undef; |
109 |
|
} else { |
110 |
|
$log->info("using $use_indexer indexing engine..."); |
111 |
|
} |
112 |
|
|
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 %{ $config->{databases} }) { |
120 |
|
|
121 |
next if ($only_db_name && $database !~ m/$only_db_name/i); |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
122 |
|
next if ($only_database && $database !~ m/$only_database/i); |
123 |
|
|
124 |
my $indexer; |
if ($use_indexer) { |
125 |
|
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
126 |
|
$indexer_config->{database} = $database; |
127 |
|
$indexer_config->{clean} = $clean; |
128 |
|
$indexer_config->{label} = $db_config->{name}; |
129 |
|
|
130 |
|
if ($use_indexer eq 'hyperestraier') { |
131 |
|
|
132 |
|
# open Hyper Estraier database |
133 |
|
use WebPAC::Output::Estraier '0.10'; |
134 |
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
135 |
|
|
136 |
|
} elsif ($use_indexer eq 'kinosearch') { |
137 |
|
|
138 |
|
# open KinoSearch |
139 |
|
use WebPAC::Output::KinoSearch; |
140 |
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
141 |
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
142 |
|
|
143 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
} else { |
144 |
$indexer_config->{database} = $database; |
$log->logdie("unknown use_indexer: $use_indexer"); |
145 |
$indexer_config->{clean} = $clean; |
} |
|
$indexer_config->{label} = $db_config->{name}; |
|
|
|
|
|
if ($use_indexer eq 'hyperestraier') { |
|
|
|
|
|
# open Hyper Estraier database |
|
|
use WebPAC::Output::Estraier '0.10'; |
|
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
|
|
|
|
|
} elsif ($use_indexer eq 'kinosearch') { |
|
|
|
|
|
# open KinoSearch |
|
|
use WebPAC::Output::KinoSearch; |
|
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
|
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
|
146 |
|
|
147 |
} else { |
$log->logide("can't continue without valid indexer") unless ($indexer); |
|
$log->logdie("unknown use_indexer: $use_indexer"); |
|
148 |
} |
} |
149 |
|
|
|
$log->logide("can't continue without valid indexer") unless ($indexer); |
|
150 |
|
|
151 |
# |
# |
152 |
# now WebPAC::Store |
# now WebPAC::Store |
157 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
158 |
|
|
159 |
if ($clean) { |
if ($clean) { |
160 |
$log->info("creating new database $database in $db_path"); |
$log->info("creating new database '$database' in $db_path"); |
161 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
162 |
} else { |
} else { |
163 |
$log->debug("working on $database in $db_path"); |
$log->info("working on database '$database' in $db_path"); |
164 |
} |
} |
165 |
|
|
166 |
my $db = new WebPAC::Store( |
my $db = new WebPAC::Store( |
187 |
|
|
188 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
189 |
|
|
190 |
|
next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i)); |
191 |
|
|
192 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
193 |
|
|
194 |
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)); |
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->{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", |
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 |
|
|
239 |
push @{ $row->{'000'} }, $pos; |
push @{ $row->{'000'} }, $pos; |
240 |
} |
} |
241 |
|
|
242 |
|
|
243 |
my $ds; |
if ($validate) { |
244 |
if ($n) { |
my @errors = $validate->validate_errors( $row ); |
245 |
$ds = $n->data_structure($row); |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
|
} else { |
|
|
$ds = WebPAC::Normalize::Set::data_structure( |
|
|
row => $row, |
|
|
rules => $rules, |
|
|
lookup => $lookup->lookup_hash, |
|
|
); |
|
|
|
|
|
$db->save_ds( |
|
|
id => $mfn, |
|
|
ds => $ds, |
|
|
prefix => $input->{name}, |
|
|
) if ($ds); |
|
246 |
} |
} |
247 |
|
|
248 |
|
|
249 |
|
my $ds = WebPAC::Normalize::data_structure( |
250 |
|
row => $row, |
251 |
|
rules => $rules, |
252 |
|
lookup => $lookup ? $lookup->lookup_hash : undef, |
253 |
|
); |
254 |
|
|
255 |
|
$db->save_ds( |
256 |
|
id => $mfn, |
257 |
|
ds => $ds, |
258 |
|
prefix => $input->{name}, |
259 |
|
) if ($ds && !$stats); |
260 |
|
|
261 |
$indexer->add( |
$indexer->add( |
262 |
id => $input->{name} . "/" . $mfn, |
id => $input->{name} . "/" . $mfn, |
263 |
ds => $ds, |
ds => $ds, |
264 |
type => $config->{$use_indexer}->{type}, |
type => $config->{$use_indexer}->{type}, |
265 |
); |
) if ($indexer); |
266 |
|
|
267 |
$total_rows++; |
$total_rows++; |
268 |
} |
} |
271 |
|
|
272 |
}; |
}; |
273 |
|
|
274 |
eval { $indexer->finish } if ($indexer->can('finish')); |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
275 |
|
|
276 |
my $dt = time() - $start_t; |
my $dt = time() - $start_t; |
277 |
$log->info("$total_rows records indexed in " . |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
278 |
sprintf("%.2f sec [%.2f rec/sec]", |
sprintf("in %.2f sec [%.2f rec/sec]", |
279 |
$dt, ($total_rows / $dt) |
$dt, ($total_rows / $dt) |
280 |
) |
) |
281 |
); |
); |
286 |
if (ref($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 |
|
} |