| 1 |
#!/usr/bin/perl -w |
| 2 |
|
| 3 |
use strict; |
| 4 |
|
| 5 |
use Cwd qw/abs_path/; |
| 6 |
use File::Temp qw/tempdir/; |
| 7 |
use Data::Dumper; |
| 8 |
use lib './lib'; |
| 9 |
|
| 10 |
use WebPAC::Common 0.02; |
| 11 |
use WebPAC::Lookup; |
| 12 |
use WebPAC::Input 0.03; |
| 13 |
use WebPAC::Store 0.03; |
| 14 |
use WebPAC::Normalize; |
| 15 |
use WebPAC::Output::TT; |
| 16 |
use WebPAC::Validate; |
| 17 |
use YAML qw/LoadFile/; |
| 18 |
use Getopt::Long; |
| 19 |
use File::Path; |
| 20 |
use Time::HiRes qw/time/; |
| 21 |
use File::Slurp; |
| 22 |
|
| 23 |
=head1 NAME |
| 24 |
|
| 25 |
run.pl - start WebPAC indexing |
| 26 |
|
| 27 |
B<this command will probably go away. Don't get used to it!> |
| 28 |
|
| 29 |
Options: |
| 30 |
|
| 31 |
=over 4 |
| 32 |
|
| 33 |
=item --offset 42 |
| 34 |
|
| 35 |
start loading (all) databases at offset 42 |
| 36 |
|
| 37 |
=item --limit 100 |
| 38 |
|
| 39 |
limit loading to 100 records |
| 40 |
|
| 41 |
=item --clean |
| 42 |
|
| 43 |
remove database and Hyper Estraier index before indexing |
| 44 |
|
| 45 |
=item --only=database_name/input_filter |
| 46 |
|
| 47 |
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 |
| 53 |
|
| 54 |
path to YAML configuration file |
| 55 |
|
| 56 |
=item --stats |
| 57 |
|
| 58 |
disable indexing and dump statistics about field and subfield |
| 59 |
usage for each input |
| 60 |
|
| 61 |
=item --validate path/to/validation_file |
| 62 |
|
| 63 |
turn on extra validation of imput records, see L<WebPAC::Validation> |
| 64 |
|
| 65 |
=back |
| 66 |
|
| 67 |
=cut |
| 68 |
|
| 69 |
my $offset; |
| 70 |
my $limit; |
| 71 |
|
| 72 |
my $clean = 0; |
| 73 |
my $config = 'conf/config.yml'; |
| 74 |
my $debug = 0; |
| 75 |
my $only_filter; |
| 76 |
my $stats = 0; |
| 77 |
my $validate_path; |
| 78 |
|
| 79 |
GetOptions( |
| 80 |
"limit=i" => \$limit, |
| 81 |
"offset=i" => \$offset, |
| 82 |
"clean" => \$clean, |
| 83 |
"one=s" => \$only_filter, |
| 84 |
"only=s" => \$only_filter, |
| 85 |
"config" => \$config, |
| 86 |
"debug" => \$debug, |
| 87 |
"stats" => \$stats, |
| 88 |
"validate=s" => \$validate_path, |
| 89 |
); |
| 90 |
|
| 91 |
$config = LoadFile($config); |
| 92 |
|
| 93 |
print "config = ",Dumper($config) if ($debug); |
| 94 |
|
| 95 |
die "no databases in config file!\n" unless ($config->{databases}); |
| 96 |
|
| 97 |
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 = $config->{use_indexer} || 'hyperestraier'; |
| 106 |
if ($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; |
| 114 |
my $start_t = time(); |
| 115 |
|
| 116 |
my @links; |
| 117 |
my $indexer; |
| 118 |
|
| 119 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
| 120 |
|
| 121 |
my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter); |
| 122 |
next if ($only_database && $database !~ m/$only_database/i); |
| 123 |
|
| 124 |
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 |
} else { |
| 144 |
$log->logdie("unknown use_indexer: $use_indexer"); |
| 145 |
} |
| 146 |
|
| 147 |
$log->logide("can't continue without valid indexer") unless ($indexer); |
| 148 |
} |
| 149 |
|
| 150 |
|
| 151 |
# |
| 152 |
# now WebPAC::Store |
| 153 |
# |
| 154 |
my $abs_path = abs_path($0); |
| 155 |
$abs_path =~ s#/[^/]*$#/#; |
| 156 |
|
| 157 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
| 158 |
|
| 159 |
if ($clean) { |
| 160 |
$log->info("creating new database '$database' in $db_path"); |
| 161 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
| 162 |
} else { |
| 163 |
$log->info("working on database '$database' in $db_path"); |
| 164 |
} |
| 165 |
|
| 166 |
my $db = new WebPAC::Store( |
| 167 |
path => $db_path, |
| 168 |
database => $database, |
| 169 |
debug => $debug, |
| 170 |
); |
| 171 |
|
| 172 |
|
| 173 |
# |
| 174 |
# now, iterate through input formats |
| 175 |
# |
| 176 |
|
| 177 |
my @inputs; |
| 178 |
if (ref($db_config->{input}) eq 'ARRAY') { |
| 179 |
@inputs = @{ $db_config->{input} }; |
| 180 |
} elsif ($db_config->{input}) { |
| 181 |
push @inputs, $db_config->{input}; |
| 182 |
} else { |
| 183 |
$log->info("database $database doesn't have inputs defined"); |
| 184 |
} |
| 185 |
|
| 186 |
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
| 187 |
|
| 188 |
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}); |
| 193 |
|
| 194 |
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( |
| 197 |
lookup_file => $input->{lookup}, |
| 198 |
) if ($input->{lookup}); |
| 199 |
|
| 200 |
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", |
| 203 |
$input->{lookup} ? "lookup '$input->{lookup}'" : "" |
| 204 |
); |
| 205 |
|
| 206 |
my $input_db = new WebPAC::Input( |
| 207 |
module => $input_module, |
| 208 |
code_page => $config->{webpac}->{webpac_encoding}, |
| 209 |
limit => $limit || $input->{limit}, |
| 210 |
offset => $offset, |
| 211 |
lookup => $lookup, |
| 212 |
recode => $input->{recode}, |
| 213 |
stats => $stats, |
| 214 |
); |
| 215 |
$log->logdie("can't create input using $input_module") unless ($input); |
| 216 |
|
| 217 |
my $maxmfn = $input_db->open( |
| 218 |
path => $input->{path}, |
| 219 |
code_page => $input->{encoding}, # database encoding |
| 220 |
%{ $input }, |
| 221 |
); |
| 222 |
|
| 223 |
my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? |
| 224 |
@{ $input->{normalize} } : ( $input->{normalize} ); |
| 225 |
|
| 226 |
foreach my $normalize (@norm_array) { |
| 227 |
|
| 228 |
my $rules; |
| 229 |
my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); |
| 230 |
|
| 231 |
$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); |
| 232 |
|
| 233 |
my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
| 234 |
|
| 235 |
$log->info("Using $normalize_path for normalization..."); |
| 236 |
|
| 237 |
# reset position in database |
| 238 |
$input_db->seek(1); |
| 239 |
|
| 240 |
foreach my $pos ( 0 ... $input_db->size ) { |
| 241 |
|
| 242 |
my $row = $input_db->fetch || next; |
| 243 |
|
| 244 |
my $mfn = $row->{'000'}->[0]; |
| 245 |
|
| 246 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
| 247 |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
| 248 |
$mfn = $pos; |
| 249 |
push @{ $row->{'000'} }, $pos; |
| 250 |
} |
| 251 |
|
| 252 |
|
| 253 |
if ($validate) { |
| 254 |
my @errors = $validate->validate_errors( $row ); |
| 255 |
$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); |
| 256 |
} |
| 257 |
|
| 258 |
|
| 259 |
my $ds = WebPAC::Normalize::data_structure( |
| 260 |
row => $row, |
| 261 |
rules => $rules, |
| 262 |
lookup => $lookup ? $lookup->lookup_hash : undef, |
| 263 |
); |
| 264 |
|
| 265 |
$db->save_ds( |
| 266 |
id => $mfn, |
| 267 |
ds => $ds, |
| 268 |
prefix => $input->{name}, |
| 269 |
) if ($ds && !$stats); |
| 270 |
|
| 271 |
$indexer->add( |
| 272 |
id => $input->{name} . "/" . $mfn, |
| 273 |
ds => $ds, |
| 274 |
type => $config->{$use_indexer}->{type}, |
| 275 |
) if ($indexer); |
| 276 |
|
| 277 |
$total_rows++; |
| 278 |
} |
| 279 |
|
| 280 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
| 281 |
|
| 282 |
}; |
| 283 |
|
| 284 |
} |
| 285 |
|
| 286 |
eval { $indexer->finish } if ($indexer && $indexer->can('finish')); |
| 287 |
|
| 288 |
my $dt = time() - $start_t; |
| 289 |
$log->info("$total_rows records ", $indexer ? "indexed " : "", |
| 290 |
sprintf("in %.2f sec [%.2f rec/sec]", |
| 291 |
$dt, ($total_rows / $dt) |
| 292 |
) |
| 293 |
); |
| 294 |
|
| 295 |
# |
| 296 |
# add Hyper Estraier links to other databases |
| 297 |
# |
| 298 |
if (ref($db_config->{links}) eq 'ARRAY') { |
| 299 |
foreach my $link (@{ $db_config->{links} }) { |
| 300 |
if ($use_indexer eq 'hyperestraier') { |
| 301 |
$log->info("saving link $database -> $link->{to} [$link->{credit}]"); |
| 302 |
push @links, { |
| 303 |
from => $database, |
| 304 |
to => $link->{to}, |
| 305 |
credit => $link->{credit}, |
| 306 |
}; |
| 307 |
} else { |
| 308 |
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
| 309 |
} |
| 310 |
} |
| 311 |
} |
| 312 |
|
| 313 |
} |
| 314 |
|
| 315 |
foreach my $link (@links) { |
| 316 |
$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); |
| 317 |
$indexer->add_link( %{ $link } ); |
| 318 |
} |
Parent Directory
| 
Revision Log