| 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::XML; |
| 15 |
use WebPAC::Normalize::Set; |
| 16 |
use WebPAC::Output::TT; |
| 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 |
| 46 |
|
| 47 |
reindex just single database (legacy name is --one) |
| 48 |
|
| 49 |
=item --config conf/config.yml |
| 50 |
|
| 51 |
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 |
dump statistics about used fields and subfields in each input |
| 61 |
|
| 62 |
=back |
| 63 |
|
| 64 |
=cut |
| 65 |
|
| 66 |
my $offset; |
| 67 |
my $limit; |
| 68 |
|
| 69 |
my $clean = 0; |
| 70 |
my $config = 'conf/config.yml'; |
| 71 |
my $debug = 0; |
| 72 |
my $only_db_name; |
| 73 |
my $force_set = 0; |
| 74 |
my $stats = 0; |
| 75 |
|
| 76 |
GetOptions( |
| 77 |
"limit=i" => \$limit, |
| 78 |
"offset=i" => \$offset, |
| 79 |
"clean" => \$clean, |
| 80 |
"one=s" => \$only_db_name, |
| 81 |
"only=s" => \$only_db_name, |
| 82 |
"config" => \$config, |
| 83 |
"debug" => \$debug, |
| 84 |
"force-set" => \$force_set, |
| 85 |
"stats" => \$stats, |
| 86 |
); |
| 87 |
|
| 88 |
$config = LoadFile($config); |
| 89 |
|
| 90 |
print "config = ",Dumper($config) if ($debug); |
| 91 |
|
| 92 |
die "no databases in config file!\n" unless ($config->{databases}); |
| 93 |
|
| 94 |
my $log = _new WebPAC::Common()->_get_logger(); |
| 95 |
|
| 96 |
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
| 97 |
$log->info("using $use_indexer indexing engine..."); |
| 98 |
|
| 99 |
my $total_rows = 0; |
| 100 |
my $start_t = time(); |
| 101 |
|
| 102 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
| 103 |
|
| 104 |
next if ($only_db_name && $database !~ m/$only_db_name/i); |
| 105 |
|
| 106 |
my $indexer; |
| 107 |
|
| 108 |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
| 109 |
$indexer_config->{database} = $database; |
| 110 |
$indexer_config->{clean} = $clean; |
| 111 |
$indexer_config->{label} = $db_config->{name}; |
| 112 |
|
| 113 |
if ($use_indexer eq 'hyperestraier') { |
| 114 |
|
| 115 |
# open Hyper Estraier database |
| 116 |
use WebPAC::Output::Estraier '0.10'; |
| 117 |
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
| 118 |
|
| 119 |
} elsif ($use_indexer eq 'kinosearch') { |
| 120 |
|
| 121 |
# open KinoSearch |
| 122 |
use WebPAC::Output::KinoSearch; |
| 123 |
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
| 124 |
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
| 125 |
|
| 126 |
} else { |
| 127 |
$log->logdie("unknown use_indexer: $use_indexer"); |
| 128 |
} |
| 129 |
|
| 130 |
$log->logide("can't continue without valid indexer") unless ($indexer); |
| 131 |
|
| 132 |
# |
| 133 |
# now WebPAC::Store |
| 134 |
# |
| 135 |
my $abs_path = abs_path($0); |
| 136 |
$abs_path =~ s#/[^/]*$#/#; |
| 137 |
|
| 138 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
| 139 |
|
| 140 |
if ($clean) { |
| 141 |
$log->info("creating new database $database in $db_path"); |
| 142 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
| 143 |
} else { |
| 144 |
$log->debug("working on $database in $db_path"); |
| 145 |
} |
| 146 |
|
| 147 |
my $db = new WebPAC::Store( |
| 148 |
path => $db_path, |
| 149 |
database => $database, |
| 150 |
debug => $debug, |
| 151 |
); |
| 152 |
|
| 153 |
|
| 154 |
# |
| 155 |
# now, iterate through input formats |
| 156 |
# |
| 157 |
|
| 158 |
my @inputs; |
| 159 |
if (ref($db_config->{input}) eq 'ARRAY') { |
| 160 |
@inputs = @{ $db_config->{input} }; |
| 161 |
} elsif ($db_config->{input}) { |
| 162 |
push @inputs, $db_config->{input}; |
| 163 |
} else { |
| 164 |
$log->info("database $database doesn't have inputs defined"); |
| 165 |
} |
| 166 |
|
| 167 |
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
| 168 |
|
| 169 |
foreach my $input (@inputs) { |
| 170 |
|
| 171 |
my $type = lc($input->{type}); |
| 172 |
|
| 173 |
die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); |
| 174 |
|
| 175 |
my $lookup = new WebPAC::Lookup( |
| 176 |
lookup_file => $input->{lookup}, |
| 177 |
); |
| 178 |
|
| 179 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
| 180 |
|
| 181 |
$log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'"); |
| 182 |
|
| 183 |
my $input_db = new WebPAC::Input( |
| 184 |
module => $input_module, |
| 185 |
code_page => $config->{webpac}->{webpac_encoding}, |
| 186 |
limit => $limit || $input->{limit}, |
| 187 |
offset => $offset, |
| 188 |
lookup => $lookup, |
| 189 |
recode => $input->{recode}, |
| 190 |
stats => $stats, |
| 191 |
); |
| 192 |
$log->logdie("can't create input using $input_module") unless ($input); |
| 193 |
|
| 194 |
my $maxmfn = $input_db->open( |
| 195 |
path => $input->{path}, |
| 196 |
code_page => $input->{encoding}, # database encoding |
| 197 |
); |
| 198 |
|
| 199 |
my $n = new WebPAC::Normalize::XML( |
| 200 |
# filter => { 'foo' => sub { shift } }, |
| 201 |
db => $db, |
| 202 |
lookup_regex => $lookup->regex, |
| 203 |
lookup => $lookup, |
| 204 |
prefix => $input->{name}, |
| 205 |
); |
| 206 |
|
| 207 |
my $rules; |
| 208 |
my $normalize_path = $input->{normalize}->{path}; |
| 209 |
|
| 210 |
if ($force_set) { |
| 211 |
my $new_norm_path = $normalize_path; |
| 212 |
$new_norm_path =~ s/\.xml$/.pl/; |
| 213 |
if (-e $new_norm_path) { |
| 214 |
$log->debug("--force-set replaced $normalize_path with $new_norm_path"); |
| 215 |
$normalize_path = $new_norm_path; |
| 216 |
} else { |
| 217 |
$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path"); |
| 218 |
} |
| 219 |
} |
| 220 |
|
| 221 |
if ($normalize_path =~ m/\.xml$/i) { |
| 222 |
$n->open( |
| 223 |
tag => $input->{normalize}->{tag}, |
| 224 |
xml_file => $normalize_path, |
| 225 |
); |
| 226 |
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
| 227 |
$n->open_yaml( |
| 228 |
path => $normalize_path, |
| 229 |
tag => $input->{normalize}->{tag}, |
| 230 |
); |
| 231 |
} elsif ($normalize_path =~ m/\.(?:pl)$/i) { |
| 232 |
$n = undef; |
| 233 |
$log->info("using WebPAC::Normalize::Set to process $normalize_path"); |
| 234 |
$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; |
| 235 |
} |
| 236 |
|
| 237 |
foreach my $pos ( 0 ... $input_db->size ) { |
| 238 |
|
| 239 |
my $row = $input_db->fetch || next; |
| 240 |
|
| 241 |
my $mfn = $row->{'000'}->[0]; |
| 242 |
|
| 243 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
| 244 |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
| 245 |
$mfn = $pos; |
| 246 |
push @{ $row->{'000'} }, $pos; |
| 247 |
} |
| 248 |
|
| 249 |
|
| 250 |
my $ds; |
| 251 |
if ($n) { |
| 252 |
$ds = $n->data_structure($row); |
| 253 |
} else { |
| 254 |
$ds = WebPAC::Normalize::Set::data_structure( |
| 255 |
row => $row, |
| 256 |
rules => $rules, |
| 257 |
lookup => $lookup->lookup_hash, |
| 258 |
); |
| 259 |
|
| 260 |
$db->save_ds( |
| 261 |
id => $mfn, |
| 262 |
ds => $ds, |
| 263 |
prefix => $input->{name}, |
| 264 |
) if ($ds); |
| 265 |
} |
| 266 |
|
| 267 |
$indexer->add( |
| 268 |
id => $input->{name} . "/" . $mfn, |
| 269 |
ds => $ds, |
| 270 |
type => $config->{$use_indexer}->{type}, |
| 271 |
); |
| 272 |
|
| 273 |
$total_rows++; |
| 274 |
} |
| 275 |
|
| 276 |
$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); |
| 277 |
|
| 278 |
}; |
| 279 |
|
| 280 |
eval { $indexer->finish } if ($indexer->can('finish')); |
| 281 |
|
| 282 |
my $dt = time() - $start_t; |
| 283 |
$log->info("$total_rows records indexed in " . |
| 284 |
sprintf("%.2f sec [%.2f rec/sec]", |
| 285 |
$dt, ($total_rows / $dt) |
| 286 |
) |
| 287 |
); |
| 288 |
|
| 289 |
# |
| 290 |
# add Hyper Estraier links to other databases |
| 291 |
# |
| 292 |
if (ref($db_config->{links}) eq 'ARRAY') { |
| 293 |
foreach my $link (@{ $db_config->{links} }) { |
| 294 |
if ($use_indexer eq 'hyperestraier') { |
| 295 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
| 296 |
$indexer->add_link( |
| 297 |
from => $database, |
| 298 |
to => $link->{to}, |
| 299 |
credit => $link->{credit}, |
| 300 |
); |
| 301 |
} else { |
| 302 |
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
| 303 |
} |
| 304 |
} |
| 305 |
} |
| 306 |
|
| 307 |
} |
| 308 |
|
Parent Directory
| 
Revision Log