| 1 | ###########################################################################
|
|---|
| 2 | #
|
|---|
| 3 | # mgbuilder.pm -- MGBuilder object
|
|---|
| 4 | # A component of the Greenstone digital library software
|
|---|
| 5 | # from the New Zealand Digital Library Project at the
|
|---|
| 6 | # University of Waikato, New Zealand.
|
|---|
| 7 | #
|
|---|
| 8 | # Copyright (C) 1999 New Zealand Digital Library Project
|
|---|
| 9 | #
|
|---|
| 10 | # This program is free software; you can redistribute it and/or modify
|
|---|
| 11 | # it under the terms of the GNU General Public License as published by
|
|---|
| 12 | # the Free Software Foundation; either version 2 of the License, or
|
|---|
| 13 | # (at your option) any later version.
|
|---|
| 14 | #
|
|---|
| 15 | # This program is distributed in the hope that it will be useful,
|
|---|
| 16 | # but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|---|
| 17 | # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|---|
| 18 | # GNU General Public License for more details.
|
|---|
| 19 | #
|
|---|
| 20 | # You should have received a copy of the GNU General Public License
|
|---|
| 21 | # along with this program; if not, write to the Free Software
|
|---|
| 22 | # Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
|
|---|
| 23 | #
|
|---|
| 24 | ###########################################################################
|
|---|
| 25 |
|
|---|
| 26 | package mgbuilder;
|
|---|
| 27 |
|
|---|
| 28 | use classify;
|
|---|
| 29 | use cfgread;
|
|---|
| 30 | use colcfg;
|
|---|
| 31 | use plugin;
|
|---|
| 32 | use util;
|
|---|
| 33 | use FileHandle;
|
|---|
| 34 |
|
|---|
| 35 | BEGIN {
|
|---|
| 36 | # set autoflush on for STDERR and STDOUT so that mg
|
|---|
| 37 | # doesn't get out of sync with plugins
|
|---|
| 38 | STDOUT->autoflush(1);
|
|---|
| 39 | STDERR->autoflush(1);
|
|---|
| 40 | }
|
|---|
| 41 |
|
|---|
| 42 | END {
|
|---|
| 43 | STDOUT->autoflush(0);
|
|---|
| 44 | STDERR->autoflush(0);
|
|---|
| 45 | }
|
|---|
| 46 |
|
|---|
| 47 | my $maxdocsize = 12000;
|
|---|
| 48 |
|
|---|
| 49 | my %wanted_index_files = ('td'=>1,
|
|---|
| 50 | 't'=>1,
|
|---|
| 51 | 'idb'=>1,
|
|---|
| 52 | 'ib1'=>1,
|
|---|
| 53 | 'ib2'=>1,
|
|---|
| 54 | 'ib3'=>1,
|
|---|
| 55 | 'i'=>1,
|
|---|
| 56 | 'ip'=>1,
|
|---|
| 57 | 'tiw'=>1,
|
|---|
| 58 | 'wa'=>1);
|
|---|
| 59 |
|
|---|
| 60 |
|
|---|
| 61 | sub new {
|
|---|
| 62 | my ($class, $collection, $source_dir, $build_dir, $verbosity,
|
|---|
| 63 | $maxdocs, $debug, $keepold, $remove_empty_classifications,
|
|---|
| 64 | $outhandle, $no_text, $failhandle, $gli) = @_;
|
|---|
| 65 |
|
|---|
| 66 | $outhandle = STDERR unless defined $outhandle;
|
|---|
| 67 | $no_text = 0 unless defined $no_text;
|
|---|
| 68 | $failhandle = STDERR unless defined $failhandle;
|
|---|
| 69 |
|
|---|
| 70 | # create an mgbuilder object
|
|---|
| 71 | my $self = bless {'collection'=>$collection,
|
|---|
| 72 | 'source_dir'=>$source_dir,
|
|---|
| 73 | 'build_dir'=>$build_dir,
|
|---|
| 74 | 'verbosity'=>$verbosity,
|
|---|
| 75 | 'maxdocs'=>$maxdocs,
|
|---|
| 76 | 'debug'=>$debug,
|
|---|
| 77 | 'keepold'=>$keepold,
|
|---|
| 78 | 'remove_empty_classifications'=>$remove_empty_classifications,
|
|---|
| 79 | 'outhandle'=>$outhandle,
|
|---|
| 80 | 'no_text'=>$no_text,
|
|---|
| 81 | 'failhandle'=>$failhandle,
|
|---|
| 82 | 'notbuilt'=>{}, # indexes not built
|
|---|
| 83 | 'gli'=>$gli
|
|---|
| 84 | }, $class;
|
|---|
| 85 |
|
|---|
| 86 | $self->{'gli'} = 0 unless defined $self->{'gli'};
|
|---|
| 87 |
|
|---|
| 88 | # read in the collection configuration file
|
|---|
| 89 | my $colcfgname = "$ENV{'GSDLCOLLECTDIR'}/etc/collect.cfg";
|
|---|
| 90 | if (!-e $colcfgname) {
|
|---|
| 91 | die "mgbuilder::new - couldn't find collect.cfg for collection $collection\n";
|
|---|
| 92 | }
|
|---|
| 93 | $self->{'collect_cfg'} = &colcfg::read_collect_cfg ($colcfgname);
|
|---|
| 94 |
|
|---|
| 95 | if (!defined($self->{'collect_cfg'}->{'indexes'})) {
|
|---|
| 96 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 97 | }
|
|---|
| 98 |
|
|---|
| 99 | # sort out subcollection indexes
|
|---|
| 100 | if (defined $self->{'collect_cfg'}->{'indexsubcollections'}) {
|
|---|
| 101 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 102 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 103 | foreach my $subcollection (@{$self->{'collect_cfg'}->{'indexsubcollections'}}) {
|
|---|
| 104 | foreach my $index (@$indexes) {
|
|---|
| 105 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$subcollection");
|
|---|
| 106 | }
|
|---|
| 107 | }
|
|---|
| 108 | }
|
|---|
| 109 |
|
|---|
| 110 | # sort out language subindexes
|
|---|
| 111 | if (defined $self->{'collect_cfg'}->{'languages'}) {
|
|---|
| 112 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 113 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 114 | foreach my $language (@{$self->{'collect_cfg'}->{'languages'}}) {
|
|---|
| 115 | foreach my $index (@$indexes) {
|
|---|
| 116 | if (defined ($self->{'collect_cfg'}->{'indexsubcollections'})) {
|
|---|
| 117 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$language");
|
|---|
| 118 | }
|
|---|
| 119 | else { # add in an empty subcollection field
|
|---|
| 120 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index\:\:$language");
|
|---|
| 121 | }
|
|---|
| 122 | }
|
|---|
| 123 | }
|
|---|
| 124 | }
|
|---|
| 125 |
|
|---|
| 126 | if (defined($self->{'collect_cfg'}->{'indexes'})) {
|
|---|
| 127 | # make sure that the same index isn't specified more than once
|
|---|
| 128 | my %tmphash = ();
|
|---|
| 129 | my @tmparray = @{$self->{'collect_cfg'}->{'indexes'}};
|
|---|
| 130 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 131 | foreach my $i (@tmparray) {
|
|---|
| 132 | if (!defined ($tmphash{$i})) {
|
|---|
| 133 | push (@{$self->{'collect_cfg'}->{'indexes'}}, $i);
|
|---|
| 134 | $tmphash{$i} = 1;
|
|---|
| 135 | }
|
|---|
| 136 | }
|
|---|
| 137 | } else {
|
|---|
| 138 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 139 | }
|
|---|
| 140 |
|
|---|
| 141 | if (scalar(@{$self->{'collect_cfg'}->{'indexes'}}) == 0) {
|
|---|
| 142 | # no indexes have been specified so we'll build a "dummy:text" index
|
|---|
| 143 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "dummy:text");
|
|---|
| 144 | }
|
|---|
| 145 |
|
|---|
| 146 | # get the list of plugins for this collection
|
|---|
| 147 | my $plugins = [];
|
|---|
| 148 | if (defined $self->{'collect_cfg'}->{'plugin'}) {
|
|---|
| 149 | $plugins = $self->{'collect_cfg'}->{'plugin'};
|
|---|
| 150 | }
|
|---|
| 151 |
|
|---|
| 152 | # load all the plugins
|
|---|
| 153 |
|
|---|
| 154 | #build up the extra global options for the plugins
|
|---|
| 155 | my @global_opts = ();
|
|---|
| 156 | if (defined $self->{'collect_cfg'}->{'separate_cjk'} && $self->{'collect_cfg'}->{'separate_cjk'} =~ /^true$/i) {
|
|---|
| 157 | push @global_opts, "-separate_cjk";
|
|---|
| 158 | }
|
|---|
| 159 | $self->{'pluginfo'} = &plugin::load_plugins ($plugins, $verbosity, $outhandle, $failhandle, \@global_opts);
|
|---|
| 160 |
|
|---|
| 161 | if (scalar(@{$self->{'pluginfo'}}) == 0) {
|
|---|
| 162 | print $outhandle "No plugins were loaded.\n";
|
|---|
| 163 | die "\n";
|
|---|
| 164 | }
|
|---|
| 165 |
|
|---|
| 166 | # get the list of classifiers for this collection
|
|---|
| 167 | my $classifiers = [];
|
|---|
| 168 | if (defined $self->{'collect_cfg'}->{'classify'}) {
|
|---|
| 169 | $classifiers = $self->{'collect_cfg'}->{'classify'};
|
|---|
| 170 | }
|
|---|
| 171 |
|
|---|
| 172 | # load all the classifiers
|
|---|
| 173 | $self->{'classifiers'} = &classify::load_classifiers ($classifiers, $build_dir, $outhandle);
|
|---|
| 174 |
|
|---|
| 175 | # load up any dontgdbm fields
|
|---|
| 176 | $self->{'dontgdbm'} = {};
|
|---|
| 177 | if (defined ($self->{'collect_cfg'}->{'dontgdbm'})) {
|
|---|
| 178 | foreach my $dg (@{$self->{'collect_cfg'}->{'dontgdbm'}}) {
|
|---|
| 179 | $self->{'dontgdbm'}->{$dg} = 1;
|
|---|
| 180 | }
|
|---|
| 181 | }
|
|---|
| 182 |
|
|---|
| 183 | # load up the document processor for building
|
|---|
| 184 | # if a buildproc class has been created for this collection, use it
|
|---|
| 185 | # otherwise, use the mg buildproc
|
|---|
| 186 | my ($buildprocdir, $buildproctype);
|
|---|
| 187 | if (-e "$ENV{'GSDLCOLLECTDIR'}/perllib/${collection}buildproc.pm") {
|
|---|
| 188 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/perllib";
|
|---|
| 189 | $buildproctype = "${collection}buildproc";
|
|---|
| 190 | } else {
|
|---|
| 191 | $buildprocdir = "$ENV{'GSDLHOME'}/perllib";
|
|---|
| 192 | $buildproctype = "mgbuildproc";
|
|---|
| 193 | }
|
|---|
| 194 |
|
|---|
| 195 | require "$buildprocdir/$buildproctype.pm";
|
|---|
| 196 |
|
|---|
| 197 | eval("\$self->{'buildproc'} = new $buildproctype(\$collection, " .
|
|---|
| 198 | "\$source_dir, \$build_dir, \$keepold, \$verbosity, \$outhandle)");
|
|---|
| 199 | die "$@" if $@;
|
|---|
| 200 |
|
|---|
| 201 | return $self;
|
|---|
| 202 | }
|
|---|
| 203 |
|
|---|
| 204 | sub init {
|
|---|
| 205 | my $self = shift (@_);
|
|---|
| 206 |
|
|---|
| 207 | if (!$self->{'debug'} && !$self->{'keepold'}) {
|
|---|
| 208 | # remove any old builds
|
|---|
| 209 | &util::rm_r($self->{'build_dir'});
|
|---|
| 210 | &util::mk_all_dir($self->{'build_dir'});
|
|---|
| 211 |
|
|---|
| 212 | # make the text directory
|
|---|
| 213 | my $textdir = "$self->{'build_dir'}/text";
|
|---|
| 214 | &util::mk_all_dir($textdir);
|
|---|
| 215 | }
|
|---|
| 216 | }
|
|---|
| 217 |
|
|---|
| 218 | sub compress_text {
|
|---|
| 219 | my $self = shift (@_);
|
|---|
| 220 | my ($textindex) = @_;
|
|---|
| 221 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 222 | my $exe = &util::get_os_exe ();
|
|---|
| 223 | my $mg_passes_exe = &util::filename_cat($exedir, "mg_passes$exe");
|
|---|
| 224 | my $mg_compression_dict_exe = &util::filename_cat($exedir, "mg_compression_dict$exe");
|
|---|
| 225 | my $outhandle = $self->{'outhandle'};
|
|---|
| 226 |
|
|---|
| 227 | my $maxnumeric = 4;
|
|---|
| 228 | if (defined($self->{'collect_cfg'}->{'maxnumeric'}) &&
|
|---|
| 229 | $self->{'collect_cfg'}->{'maxnumeric'} =~ /^\d+$/) {
|
|---|
| 230 | $maxnumeric = $self->{'collect_cfg'}->{'maxnumeric'};
|
|---|
| 231 | }
|
|---|
| 232 |
|
|---|
| 233 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, "text"));
|
|---|
| 234 | my $basefilename = "text/$self->{'collection'}";
|
|---|
| 235 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, $basefilename);
|
|---|
| 236 |
|
|---|
| 237 | my $osextra = "";
|
|---|
| 238 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
|---|
| 239 | $fulltextprefix =~ s@/@\\@g;
|
|---|
| 240 | } else {
|
|---|
| 241 | $osextra = " -d /";
|
|---|
| 242 | }
|
|---|
| 243 |
|
|---|
| 244 | print $outhandle "\n*** creating the compressed text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 245 | print STDERR "<Stage name='CompressText'>\n" if $self->{'gli'};
|
|---|
| 246 |
|
|---|
| 247 | # collect the statistics for the text
|
|---|
| 248 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
|---|
| 249 | print $outhandle "\n collecting text statistics\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 250 | print STDERR "<Phase name='CollectTextStats'/>\n" if $self->{'gli'};
|
|---|
| 251 |
|
|---|
| 252 | my ($handle);
|
|---|
| 253 | if ($self->{'debug'}) {
|
|---|
| 254 | $handle = STDOUT;
|
|---|
| 255 | } else {
|
|---|
| 256 | if (!-e "$mg_passes_exe" ||
|
|---|
| 257 | !open (PIPEOUT, "| mg_passes$exe -f \"$fulltextprefix\" -b $maxdocsize -T1 -M $maxnumeric $osextra")) {
|
|---|
| 258 | print STDERR "<FatalError name='NoRunMGPasses'>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 259 | die "mgbuilder::compress_text - couldn't run $mg_passes_exe\n";
|
|---|
| 260 | }
|
|---|
| 261 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 262 | }
|
|---|
| 263 |
|
|---|
| 264 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| 265 | $self->{'buildproc'}->set_mode ('text');
|
|---|
| 266 | $self->{'buildproc'}->set_index ($textindex);
|
|---|
| 267 | $self->{'buildproc'}->set_indexing_text (0);
|
|---|
| 268 |
|
|---|
| 269 |
|
|---|
| 270 | if ($self->{'no_text'}) {
|
|---|
| 271 | $self->{'buildproc'}->set_store_text(0);
|
|---|
| 272 | } else {
|
|---|
| 273 | $self->{'buildproc'}->set_store_text(1);
|
|---|
| 274 | }
|
|---|
| 275 | $self->{'buildproc'}->reset();
|
|---|
| 276 |
|
|---|
| 277 | &plugin::begin($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 278 | $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 279 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 280 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'}, 0, $self->{'gli'});
|
|---|
| 281 | &plugin::end($self->{'pluginfo'});
|
|---|
| 282 |
|
|---|
| 283 |
|
|---|
| 284 | close ($handle) unless $self->{'debug'};
|
|---|
| 285 |
|
|---|
| 286 | $self->print_stats();
|
|---|
| 287 |
|
|---|
| 288 | # create the compression dictionary
|
|---|
| 289 | # the compression dictionary is built by assuming the stats are from a seed
|
|---|
| 290 | # dictionary (-S), if a novel word is encountered it is spelled out (-H),
|
|---|
| 291 | # and the resulting dictionary must be less than 5 meg with the most frequent
|
|---|
| 292 | # words being put into the dictionary first (-2 -k 5120)
|
|---|
| 293 | if (!$self->{'debug'}) {
|
|---|
| 294 | print $outhandle "\n creating the compression dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 295 | print STDERR "<Phase name='CreatingCompress'/>\n" if $self->{'gli'};
|
|---|
| 296 | if (!-e "$mg_compression_dict_exe") {
|
|---|
| 297 | die "mgbuilder::compress_text - couldn't run $mg_compression_dict_exe\n";
|
|---|
| 298 | }
|
|---|
| 299 | system ("mg_compression_dict$exe -f \"$fulltextprefix\" -S -H -2 -k 5120 $osextra");
|
|---|
| 300 |
|
|---|
| 301 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
|---|
| 302 | if (!-e "$mg_passes_exe" ||
|
|---|
| 303 | !open ($handle, "| mg_passes$exe -f \"$fulltextprefix\" -b $maxdocsize -T2 -M $maxnumeric $osextra")) {
|
|---|
| 304 | print STDERR "<FatalError name='NoRunMGPasses'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 305 | die "mgbuilder::compress_text - couldn't run $mg_passes_exe\n";
|
|---|
| 306 | }
|
|---|
| 307 | }
|
|---|
| 308 | else {
|
|---|
| 309 | print STDERR "<Phase name='SkipCreatingComp'/>\n" if $self->{'gli'};
|
|---|
| 310 | }
|
|---|
| 311 |
|
|---|
| 312 | $self->{'buildproc'}->reset();
|
|---|
| 313 | # compress the text
|
|---|
| 314 | print $outhandle "\n compressing the text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 315 | print STDERR "<Phase name='CompressingText'/>\n" if $self->{'gli'};
|
|---|
| 316 |
|
|---|
| 317 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 318 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'}, 0, $self->{'gli'});
|
|---|
| 319 |
|
|---|
| 320 | close ($handle) unless $self->{'debug'};
|
|---|
| 321 |
|
|---|
| 322 | $self->print_stats();
|
|---|
| 323 | print STDERR "</Stage>\n" if $self->{'gli'};
|
|---|
| 324 | }
|
|---|
| 325 |
|
|---|
| 326 | sub want_built {
|
|---|
| 327 | my $self = shift (@_);
|
|---|
| 328 | my ($index) = @_;
|
|---|
| 329 |
|
|---|
| 330 | if (defined ($self->{'collect_cfg'}->{'dontbuild'})) {
|
|---|
| 331 | foreach my $checkstr (@{$self->{'collect_cfg'}->{'dontbuild'}}) {
|
|---|
| 332 | if ($index =~ /^$checkstr$/) {
|
|---|
| 333 | #push (@{$self->{'notbuilt'}}, $self->{'index_mapping'}->{$index});
|
|---|
| 334 | $self->{'notbuilt'}->{$index} = 1;
|
|---|
| 335 | return 0;
|
|---|
| 336 | }
|
|---|
| 337 | }
|
|---|
| 338 | }
|
|---|
| 339 |
|
|---|
| 340 | return 1;
|
|---|
| 341 | }
|
|---|
| 342 |
|
|---|
| 343 | sub build_indexes {
|
|---|
| 344 | my $self = shift (@_);
|
|---|
| 345 | my ($indexname) = @_;
|
|---|
| 346 | my $outhandle = $self->{'outhandle'};
|
|---|
| 347 | my $indexes = [];
|
|---|
| 348 | if (defined $indexname && $indexname =~ /\w/) {
|
|---|
| 349 | push @$indexes, $indexname;
|
|---|
| 350 | } else {
|
|---|
| 351 | $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 352 | }
|
|---|
| 353 |
|
|---|
| 354 | # create the mapping between the index descriptions
|
|---|
| 355 | # and their directory names
|
|---|
| 356 | $self->{'index_mapping'} = $self->create_index_mapping ($indexes);
|
|---|
| 357 |
|
|---|
| 358 | # build each of the indexes
|
|---|
| 359 | foreach my $index (@$indexes) {
|
|---|
| 360 | if ($self->want_built($index)) {
|
|---|
| 361 | print $outhandle "\n*** building index $index in subdirectory " .
|
|---|
| 362 | "$self->{'index_mapping'}->{$index}\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 363 | print STDERR "<Stage name='Index' source='$index'>\n" if $self->{'gli'};
|
|---|
| 364 | $self->build_index($index);
|
|---|
| 365 | } else {
|
|---|
| 366 | print $outhandle "\n*** ignoring index $index\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 367 | }
|
|---|
| 368 | }
|
|---|
| 369 | }
|
|---|
| 370 |
|
|---|
| 371 | # creates directory names for each of the index descriptions
|
|---|
| 372 | sub create_index_mapping {
|
|---|
| 373 | my $self = shift (@_);
|
|---|
| 374 | my ($indexes) = @_;
|
|---|
| 375 |
|
|---|
| 376 | my %mapping = ();
|
|---|
| 377 | $mapping{'indexmaporder'} = [];
|
|---|
| 378 | $mapping{'subcollectionmaporder'} = [];
|
|---|
| 379 | $mapping{'languagemaporder'} = [];
|
|---|
| 380 |
|
|---|
| 381 | # dirnames is used to check for collisions. Start this off
|
|---|
| 382 | # with the manditory directory names
|
|---|
| 383 | my %dirnames = ('text'=>'text',
|
|---|
| 384 | 'extra'=>'extra');
|
|---|
| 385 | my %pnames = ('index' => {}, 'subcollection' => {}, 'languages' => {});
|
|---|
| 386 | foreach my $index (@$indexes) {
|
|---|
| 387 | my ($level, $gran, $subcollection, $languages) = split (":", $index);
|
|---|
| 388 |
|
|---|
| 389 | # the directory name starts with the first character of the index level
|
|---|
| 390 | my ($pindex) = $level =~ /^(.)/;
|
|---|
| 391 |
|
|---|
| 392 | # next comes a processed version of the index
|
|---|
| 393 | $pindex .= $self->process_field ($gran);
|
|---|
| 394 | $pindex = lc ($pindex);
|
|---|
| 395 |
|
|---|
| 396 | # next comes a processed version of the subcollection if there is one.
|
|---|
| 397 | my $psub = $self->process_field ($subcollection);
|
|---|
| 398 | $psub = lc ($psub);
|
|---|
| 399 |
|
|---|
| 400 | # next comes a processed version of the language if there is one.
|
|---|
| 401 | my $plang = $self->process_field ($languages);
|
|---|
| 402 | $plang = lc ($plang);
|
|---|
| 403 |
|
|---|
| 404 | my $dirname = $pindex . $psub . $plang;
|
|---|
| 405 |
|
|---|
| 406 | # check to be sure all index names are unique
|
|---|
| 407 | while (defined ($dirnames{$dirname})) {
|
|---|
| 408 | $dirname = $self->make_unique (\%pnames, $index, \$pindex, \$psub, \$plang);
|
|---|
| 409 | }
|
|---|
| 410 | $mapping{$index} = $dirname;
|
|---|
| 411 |
|
|---|
| 412 | # store the mapping orders as well as the maps
|
|---|
| 413 | # also put index, subcollection and language fields into the mapping thing -
|
|---|
| 414 | # (the full index name (eg document:text:subcol:lang) is not used on
|
|---|
| 415 | # the query page) -these are used for collectionmeta later on
|
|---|
| 416 | if (!defined $mapping{'indexmap'}{"$level:$gran"}) {
|
|---|
| 417 | $mapping{'indexmap'}{"$level:$gran"} = $pindex;
|
|---|
| 418 | push (@{$mapping{'indexmaporder'}}, "$level:$gran");
|
|---|
| 419 | if (!defined $mapping{"$level:$gran"}) {
|
|---|
| 420 | $mapping{"$level:$gran"} = $pindex;
|
|---|
| 421 | }
|
|---|
| 422 | }
|
|---|
| 423 | if ($psub =~ /\w/ && !defined ($mapping{'subcollectionmap'}{$subcollection})) {
|
|---|
| 424 | $mapping{'subcollectionmap'}{$subcollection} = $psub;
|
|---|
| 425 | push (@{$mapping{'subcollectionmaporder'}}, $subcollection);
|
|---|
| 426 | $mapping{$subcollection} = $psub;
|
|---|
| 427 | }
|
|---|
| 428 | if ($plang =~ /\w/ && !defined ($mapping{'languagemap'}{$languages})) {
|
|---|
| 429 | $mapping{'languagemap'}{$languages} = $plang;
|
|---|
| 430 | push (@{$mapping{'languagemaporder'}}, $languages);
|
|---|
| 431 | $mapping{$languages} = $plang;
|
|---|
| 432 | }
|
|---|
| 433 | $dirnames{$dirname} = $index;
|
|---|
| 434 | $pnames{'index'}->{$pindex} = "$level:$gran";
|
|---|
| 435 | $pnames{'subcollection'}->{$psub} = $subcollection;
|
|---|
| 436 | $pnames{'languages'}->{$plang} = $languages;
|
|---|
| 437 | }
|
|---|
| 438 |
|
|---|
| 439 | return \%mapping;
|
|---|
| 440 | }
|
|---|
| 441 |
|
|---|
| 442 | # returns a processed version of a field.
|
|---|
| 443 | # if the field has only one component the processed
|
|---|
| 444 | # version will contain the first character and next consonant
|
|---|
| 445 | # of that componant - otherwise it will contain the first
|
|---|
| 446 | # character of the first two components
|
|---|
| 447 | sub process_field {
|
|---|
| 448 | my $self = shift (@_);
|
|---|
| 449 | my ($field) = @_;
|
|---|
| 450 |
|
|---|
| 451 | return "" unless (defined ($field) && $field =~ /\w/);
|
|---|
| 452 |
|
|---|
| 453 | my @components = split /,/, $field;
|
|---|
| 454 | if (scalar @components >= 2) {
|
|---|
| 455 | splice (@components, 2);
|
|---|
| 456 | map {s/^(.).*$/$1/;} @components;
|
|---|
| 457 | return join("", @components);
|
|---|
| 458 | } else {
|
|---|
| 459 | my ($a, $b) = $field =~ /^(.).*?([bcdfghjklmnpqrstvwxyz])/i;
|
|---|
| 460 | ($a, $b) = $field =~ /^(.)(.)/ unless defined $a && defined $b;
|
|---|
| 461 | return "$a$b";
|
|---|
| 462 | }
|
|---|
| 463 | }
|
|---|
| 464 |
|
|---|
| 465 | sub make_unique {
|
|---|
| 466 | my $self = shift (@_);
|
|---|
| 467 | my ($namehash, $index, $indexref, $subref, $langref) = @_;
|
|---|
| 468 | my ($level, $gran, $subcollection, $languages) = split (":", $index);
|
|---|
| 469 |
|
|---|
| 470 | if ($namehash->{'index'}->{$$indexref} ne "$level:$gran") {
|
|---|
| 471 | $self->get_next_version ($indexref);
|
|---|
| 472 | } elsif ($namehash->{'subcollection'}->{$$subref} ne $subcollection) {
|
|---|
| 473 | $self->get_next_version ($subref);
|
|---|
| 474 | } elsif ($namehash->{'languages'}->{$$langref} ne $languages) {
|
|---|
| 475 | $self->get_next_version ($langref);
|
|---|
| 476 | }
|
|---|
| 477 | return "$$indexref$$subref$$langref";
|
|---|
| 478 | }
|
|---|
| 479 |
|
|---|
| 480 | sub get_next_version {
|
|---|
| 481 | my $self = shift (@_);
|
|---|
| 482 | my ($nameref) = @_;
|
|---|
| 483 | if ($$nameref =~ /(\d\d)$/) {
|
|---|
| 484 | my $num = $1; $num ++;
|
|---|
| 485 | $$nameref =~ s/\d\d$/$num/;
|
|---|
| 486 | } elsif ($$nameref =~ /(\d)$/) {
|
|---|
| 487 | my $num = $1;
|
|---|
| 488 | if ($num == 9) {$$nameref =~ s/\d$/10/;}
|
|---|
| 489 | else {$num ++; $$nameref =~ s/\d$/$num/;}
|
|---|
| 490 | } else {
|
|---|
| 491 | $$nameref =~ s/.$/0/;
|
|---|
| 492 | }
|
|---|
| 493 | }
|
|---|
| 494 |
|
|---|
| 495 | sub build_index {
|
|---|
| 496 | my $self = shift (@_);
|
|---|
| 497 | my ($index) = @_;
|
|---|
| 498 | my $outhandle = $self->{'outhandle'};
|
|---|
| 499 |
|
|---|
| 500 | # get the full index directory path and make sure it exists
|
|---|
| 501 | my $indexdir = $self->{'index_mapping'}->{$index};
|
|---|
| 502 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, $indexdir));
|
|---|
| 503 | my $fullindexprefix = &util::filename_cat ($self->{'build_dir'}, $indexdir,
|
|---|
| 504 | $self->{'collection'});
|
|---|
| 505 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, "text",
|
|---|
| 506 | $self->{'collection'});
|
|---|
| 507 |
|
|---|
| 508 | # get any os specific stuff
|
|---|
| 509 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 510 | my $exe = &util::get_os_exe ();
|
|---|
| 511 | my $mg_passes_exe = &util::filename_cat($exedir, "mg_passes$exe");
|
|---|
| 512 | my $mg_perf_hash_build_exe =
|
|---|
| 513 | &util::filename_cat($exedir, "mg_perf_hash_build$exe");
|
|---|
| 514 | my $mg_weights_build_exe =
|
|---|
| 515 | &util::filename_cat ($exedir, "mg_weights_build$exe");
|
|---|
| 516 | my $mg_invf_dict_exe =
|
|---|
| 517 | &util::filename_cat ($exedir, "mg_invf_dict$exe");
|
|---|
| 518 | my $mg_stem_idx_exe =
|
|---|
| 519 | &util::filename_cat ($exedir, "mg_stem_idx$exe");
|
|---|
| 520 |
|
|---|
| 521 | my $maxnumeric = 4;
|
|---|
| 522 | if (defined($self->{'collect_cfg'}->{'maxnumeric'}) &&
|
|---|
| 523 | $self->{'collect_cfg'}->{'maxnumeric'} =~ /^\d+$/) {
|
|---|
| 524 | $maxnumeric = $self->{'collect_cfg'}->{'maxnumeric'};
|
|---|
| 525 | }
|
|---|
| 526 |
|
|---|
| 527 | my $osextra = "";
|
|---|
| 528 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
|---|
| 529 | $fullindexprefix =~ s@/@\\@g;
|
|---|
| 530 | } else {
|
|---|
| 531 | $osextra = " -d /";
|
|---|
| 532 | if ($outhandle ne "STDERR") {
|
|---|
| 533 | # so mg_passes doesn't print to stderr if we redirect output
|
|---|
| 534 | $osextra .= " 2>/dev/null";
|
|---|
| 535 | }
|
|---|
| 536 | }
|
|---|
| 537 |
|
|---|
| 538 | # get the index level from the index description
|
|---|
| 539 | # the index will be level 2 unless we are building a
|
|---|
| 540 | # paragraph level index
|
|---|
| 541 | my $index_level = 2;
|
|---|
| 542 | $index_level = 3 if $index =~ /^paragraph/i;
|
|---|
| 543 |
|
|---|
| 544 | # get the index expression if this index belongs
|
|---|
| 545 | # to a subcollection
|
|---|
| 546 | my $indexexparr = [];
|
|---|
| 547 | my $langarr = [];
|
|---|
| 548 | # there may be subcollection info, and language info.
|
|---|
| 549 | my ($level, $fields, $subcollection, $language) = split (":", $index);
|
|---|
| 550 | my @subcollections = ();
|
|---|
| 551 | @subcollections = split /,/, $subcollection if (defined $subcollection);
|
|---|
| 552 |
|
|---|
| 553 | foreach my $subcollection (@subcollections) {
|
|---|
| 554 | if (defined ($self->{'collect_cfg'}->{'subcollection'}->{$subcollection})) {
|
|---|
| 555 | push (@$indexexparr, $self->{'collect_cfg'}->{'subcollection'}->{$subcollection});
|
|---|
| 556 | }
|
|---|
| 557 | }
|
|---|
| 558 |
|
|---|
| 559 | # add expressions for languages if this index belongs to
|
|---|
| 560 | # a language subcollection - only put languages expressions for the
|
|---|
| 561 | # ones we want in the index
|
|---|
| 562 |
|
|---|
| 563 | my @languages = ();
|
|---|
| 564 | my $language_metadata = "Language";
|
|---|
| 565 | if (defined ($self->{'collect_cfg'}->{'language_metadata'})) {
|
|---|
| 566 | $language_metadata = $self->{'collect_cfg'}->{'language_metadata'};
|
|---|
| 567 | }
|
|---|
| 568 | @languages = split /,/, $language if (defined $language);
|
|---|
| 569 | foreach my $language (@languages) {
|
|---|
| 570 | my $not=0;
|
|---|
| 571 | if ($language =~ s/^\!//) {
|
|---|
| 572 | $not = 1;
|
|---|
| 573 | }
|
|---|
| 574 | if($not) {
|
|---|
| 575 | push (@$langarr, "!$language");
|
|---|
| 576 | } else {
|
|---|
| 577 | push (@$langarr, "$language");
|
|---|
| 578 | }
|
|---|
| 579 | }
|
|---|
| 580 |
|
|---|
| 581 | # Build index dictionary. Uses verbatim stem method
|
|---|
| 582 | print $outhandle "\n creating index dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 583 | print STDERR "<Phase name='CreatingIndexDic'/>\n" if $self->{'gli'};
|
|---|
| 584 | my ($handle);
|
|---|
| 585 | if ($self->{'debug'}) {
|
|---|
| 586 | $handle = STDOUT;
|
|---|
| 587 | } else {
|
|---|
| 588 | if (!-e "$mg_passes_exe" ||
|
|---|
| 589 | !open (PIPEOUT, "| mg_passes$exe -f \"$fullindexprefix\" -b $maxdocsize " .
|
|---|
| 590 | "-$index_level -m 32 -s 0 -G -t 10 -N1 -M $maxnumeric $osextra")) {
|
|---|
| 591 | print STDERR "<FatalError name='NoRunMGPasses'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 592 | die "mgbuilder::build_index - couldn't run $mg_passes_exe\n";
|
|---|
| 593 | }
|
|---|
| 594 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 595 | }
|
|---|
| 596 |
|
|---|
| 597 | # set up the document processor
|
|---|
| 598 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| 599 | $self->{'buildproc'}->set_mode ('text');
|
|---|
| 600 | $self->{'buildproc'}->set_index ($index, $indexexparr);
|
|---|
| 601 | $self->{'buildproc'}->set_index_languages ($language_metadata, $langarr) if (defined $language);
|
|---|
| 602 | $self->{'buildproc'}->set_indexing_text (1);
|
|---|
| 603 | $self->{'buildproc'}->set_store_text(1);
|
|---|
| 604 |
|
|---|
| 605 | $self->{'buildproc'}->reset();
|
|---|
| 606 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 607 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'},0, $self->{'gli'});
|
|---|
| 608 | close ($handle) unless $self->{'debug'};
|
|---|
| 609 |
|
|---|
| 610 | $self->print_stats();
|
|---|
| 611 |
|
|---|
| 612 | # now we check to see if the required files have been produced - if not we quit building this index so the whole process doesn't crap out.
|
|---|
| 613 | # we check on the .id file - index dictionary
|
|---|
| 614 | my $dict_file = "$fullindexprefix.id";
|
|---|
| 615 | if (!-e $dict_file) {
|
|---|
| 616 | print $outhandle "mgbuilder::build_index - Couldn't create index $index\n";
|
|---|
| 617 | $self->{'notbuilt'}->{$index}=1;
|
|---|
| 618 | return;
|
|---|
| 619 | }
|
|---|
| 620 | if (!$self->{'debug'}) {
|
|---|
| 621 | # create the perfect hash function
|
|---|
| 622 | if (!-e "$mg_perf_hash_build_exe") {
|
|---|
| 623 | print STDERR "<FatalError name='NoRunMGHash'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 624 | die "mgbuilder::build_index - couldn't run $mg_perf_hash_build_exe\n";
|
|---|
| 625 | }
|
|---|
| 626 | system ("mg_perf_hash_build$exe -f \"$fullindexprefix\" $osextra");
|
|---|
| 627 |
|
|---|
| 628 | if (!-e "$mg_passes_exe" ||
|
|---|
| 629 | !open ($handle, "| mg_passes$exe -f \"$fullindexprefix\" -b $maxdocsize " .
|
|---|
| 630 | "-$index_level -c 3 -G -t 10 -N2 -M $maxnumeric $osextra")) {
|
|---|
| 631 | print STDERR "<FatalError name='NoRunMGPasses'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 632 | die "mgbuilder::build_index - couldn't run $mg_passes_exe\n";
|
|---|
| 633 | }
|
|---|
| 634 | }
|
|---|
| 635 |
|
|---|
| 636 | # invert the text
|
|---|
| 637 | print $outhandle "\n inverting the text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 638 | print STDERR "<Phase name='InvertingText'/>\n" if $self->{'gli'};
|
|---|
| 639 | $self->{'buildproc'}->reset();
|
|---|
| 640 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 641 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'},0, $self->{'gli'});
|
|---|
| 642 |
|
|---|
| 643 |
|
|---|
| 644 | $self->print_stats ();
|
|---|
| 645 |
|
|---|
| 646 | if (!$self->{'debug'}) {
|
|---|
| 647 |
|
|---|
| 648 | close ($handle);
|
|---|
| 649 |
|
|---|
| 650 | # create the weights file
|
|---|
| 651 | print $outhandle "\n create the weights file\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 652 | print STDERR "<Phase name='CreateTheWeights'/>\n" if $self->{'gli'};
|
|---|
| 653 | if (!-e "$mg_weights_build_exe") {
|
|---|
| 654 | print STDERR "<FatalError name='NoRunMGWeights'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 655 | die "mgbuilder::build_index - couldn't run $mg_weights_build_exe\n";
|
|---|
| 656 | }
|
|---|
| 657 | system ("mg_weights_build$exe -f \"$fullindexprefix\" -t \"$fulltextprefix\" $osextra");
|
|---|
| 658 |
|
|---|
| 659 | # create 'on-disk' stemmed dictionary
|
|---|
| 660 | print $outhandle "\n creating 'on-disk' stemmed dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 661 | print STDERR "<Phase name='CreateStemmedDic'/>\n" if $self->{'gli'};
|
|---|
| 662 | if (!-e "$mg_invf_dict_exe") {
|
|---|
| 663 | print STDERR "<FatalError name='NoRunMGInvf'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 664 | die "mgbuilder::build_index - couldn't run $mg_invf_dict_exe\n";
|
|---|
| 665 | }
|
|---|
| 666 | system ("mg_invf_dict$exe -f \"$fullindexprefix\" $osextra");
|
|---|
| 667 |
|
|---|
| 668 |
|
|---|
| 669 | # creates stem index files for the various stemming methods
|
|---|
| 670 | print $outhandle "\n creating stem indexes\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 671 | print STDERR "<Phase name='CreatingStemIndx'/>\n" if $self->{'gli'};
|
|---|
| 672 | if (!-e "$mg_stem_idx_exe") {
|
|---|
| 673 | print STDERR "<FatalError name='NoRunMGStem'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 674 | die "mgbuilder::build_index - couldn't run $mg_stem_idx_exe\n";
|
|---|
| 675 | }
|
|---|
| 676 | system ("mg_stem_idx$exe -b 4096 -s1 -f \"$fullindexprefix\" $osextra");
|
|---|
| 677 | system ("mg_stem_idx$exe -b 4096 -s2 -f \"$fullindexprefix\" $osextra");
|
|---|
| 678 | system ("mg_stem_idx$exe -b 4096 -s3 -f \"$fullindexprefix\" $osextra");
|
|---|
| 679 |
|
|---|
| 680 | # remove unwanted files
|
|---|
| 681 | my $tmpdir = &util::filename_cat ($self->{'build_dir'}, $indexdir);
|
|---|
| 682 | opendir (DIR, $tmpdir) || die
|
|---|
| 683 | "mgbuilder::build_index - couldn't read directory $tmpdir\n";
|
|---|
| 684 | foreach my $file (readdir(DIR)) {
|
|---|
| 685 | next if $file =~ /^\./;
|
|---|
| 686 | my ($suffix) = $file =~ /\.([^\.]+)$/;
|
|---|
| 687 | if (defined $suffix && !defined $wanted_index_files{$suffix}) {
|
|---|
| 688 | # delete it!
|
|---|
| 689 | print $outhandle "deleting $file\n" if $self->{'verbosity'} > 2;
|
|---|
| 690 | &util::rm (&util::filename_cat ($tmpdir, $file));
|
|---|
| 691 | }
|
|---|
| 692 | }
|
|---|
| 693 | closedir (DIR);
|
|---|
| 694 | }
|
|---|
| 695 | print STDERR "</Stage>\n" if $self->{'gli'};
|
|---|
| 696 | }
|
|---|
| 697 |
|
|---|
| 698 | sub make_infodatabase {
|
|---|
| 699 | my $self = shift (@_);
|
|---|
| 700 | my $outhandle = $self->{'outhandle'};
|
|---|
| 701 |
|
|---|
| 702 | my $textdir = &util::filename_cat($self->{'build_dir'}, "text");
|
|---|
| 703 | my $assocdir = &util::filename_cat($self->{'build_dir'}, "assoc");
|
|---|
| 704 | &util::mk_all_dir ($textdir);
|
|---|
| 705 | &util::mk_all_dir ($assocdir);
|
|---|
| 706 |
|
|---|
| 707 | # get db name
|
|---|
| 708 | my $dbext = ".bdb";
|
|---|
| 709 | $dbext = ".ldb" if &util::is_little_endian();
|
|---|
| 710 | my $fulldbname = &util::filename_cat ($textdir, "$self->{'collection'}$dbext");
|
|---|
| 711 | $fulldbname =~ s/\//\\/g if ($ENV{'GSDLOS'} =~ /^windows$/i);
|
|---|
| 712 |
|
|---|
| 713 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 714 | my $exe = &util::get_os_exe ();
|
|---|
| 715 | my $txt2db_exe = &util::filename_cat($exedir, "txt2db$exe");
|
|---|
| 716 |
|
|---|
| 717 | print $outhandle "\n*** creating the info database and processing associated files\n"
|
|---|
| 718 | if ($self->{'verbosity'} >= 1);
|
|---|
| 719 | print STDERR "<Stage name='CreateInfoData'>\n" if $self->{'gli'};
|
|---|
| 720 |
|
|---|
| 721 | # init all the classifiers
|
|---|
| 722 | &classify::init_classifiers ($self->{'classifiers'});
|
|---|
| 723 |
|
|---|
| 724 | # set up the document processor
|
|---|
| 725 | my ($handle);
|
|---|
| 726 | if ($self->{'debug'}) {
|
|---|
| 727 | $handle = STDOUT;
|
|---|
| 728 | } else {
|
|---|
| 729 | if (!-e "$txt2db_exe" || !open (PIPEOUT, "| txt2db$exe \"$fulldbname\"")) {
|
|---|
| 730 | print STDERR "<FatalError name='NoRunText2DB'/>\n</Stage>\n" if $self->{'gli'};
|
|---|
| 731 | die "mgbuilder::make_infodatabase - couldn't run $txt2db_exe\n";
|
|---|
| 732 | }
|
|---|
| 733 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 734 | }
|
|---|
| 735 |
|
|---|
| 736 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| 737 | $self->{'buildproc'}->set_mode ('infodb');
|
|---|
| 738 | $self->{'buildproc'}->set_assocdir ($assocdir);
|
|---|
| 739 | $self->{'buildproc'}->set_dontgdbm ($self->{'dontgdbm'});
|
|---|
| 740 | $self->{'buildproc'}->set_classifiers ($self->{'classifiers'});
|
|---|
| 741 | $self->{'buildproc'}->set_indexing_text (0);
|
|---|
| 742 | $self->{'buildproc'}->set_store_text(1);
|
|---|
| 743 | $self->{'buildproc'}->reset();
|
|---|
| 744 |
|
|---|
| 745 | if (defined $self->{'collect_cfg'}->{'collectionmeta'}) {
|
|---|
| 746 |
|
|---|
| 747 | if (!defined $self->{'index_mapping'}) {
|
|---|
| 748 | $self->{'index_mapping'} =
|
|---|
| 749 | $self->create_index_mapping ($self->{'collect_cfg'}->{'indexes'});
|
|---|
| 750 | }
|
|---|
| 751 |
|
|---|
| 752 | print $handle "[collection]\n";
|
|---|
| 753 |
|
|---|
| 754 | foreach my $cmeta (keys (%{$self->{'collect_cfg'}->{'collectionmeta'}})) {
|
|---|
| 755 | my $defaultfound=0;
|
|---|
| 756 | my $first=1;
|
|---|
| 757 | my $metadata_entry = "";
|
|---|
| 758 | my $default="";
|
|---|
| 759 | my $cmetamap = "";
|
|---|
| 760 | if ($cmeta =~ s/^\.//) {
|
|---|
| 761 | if (defined $self->{'index_mapping'}->{$cmeta}) {
|
|---|
| 762 | $cmetamap = $self->{'index_mapping'}->{$cmeta};
|
|---|
| 763 | $cmeta = ".$cmeta";
|
|---|
| 764 | }
|
|---|
| 765 | else {
|
|---|
| 766 | print $outhandle "mgbuilder: warning bad collectionmeta option '$cmeta' - ignored\n";
|
|---|
| 767 | next; #ignore this one
|
|---|
| 768 | }
|
|---|
| 769 | }
|
|---|
| 770 | else {
|
|---|
| 771 | $cmetamap = $cmeta; # just using the same name
|
|---|
| 772 | }
|
|---|
| 773 | #iterate through the languages
|
|---|
| 774 | foreach my $lang (keys (%{$self->{'collect_cfg'}->{'collectionmeta'}->{$cmeta}})) {
|
|---|
| 775 | if ($first) {
|
|---|
| 776 | $first=0;
|
|---|
| 777 | #set the default default to the first entry
|
|---|
| 778 | $default=$self->{'collect_cfg'}->{'collectionmeta'}->{$cmeta}->{$lang};
|
|---|
| 779 | }
|
|---|
| 780 | if ($lang =~ /default/) {
|
|---|
| 781 | $defaultfound=1;
|
|---|
| 782 | #the default entry goes first
|
|---|
| 783 | $metadata_entry = "<$cmetamap>" .
|
|---|
| 784 | $self->{'collect_cfg'}->{'collectionmeta'}->{$cmeta}->{'default'} . "\n" . $metadata_entry;
|
|---|
| 785 | }
|
|---|
| 786 | else {
|
|---|
| 787 | my ($l) = $lang =~ /^\[l=(\w*)\]$/;
|
|---|
| 788 | if ($l) {
|
|---|
| 789 | $metadata_entry .= "<$cmetamap:$l>" .
|
|---|
| 790 | $self->{'collect_cfg'}->{'collectionmeta'}->{$cmeta}->{$lang} . "\n";
|
|---|
| 791 |
|
|---|
| 792 | # Use the English value as the default if no default is specified
|
|---|
| 793 | if ($l =~ /en/i) {
|
|---|
| 794 | $default=$self->{'collect_cfg'}->{'collectionmeta'}->{$cmeta}->{$lang};
|
|---|
| 795 | }
|
|---|
| 796 | }
|
|---|
| 797 | }
|
|---|
| 798 | }
|
|---|
| 799 | #if we haven't found a default, put one in
|
|---|
| 800 | if (!$defaultfound) {
|
|---|
| 801 | $metadata_entry = "<$cmetamap>$default\n" . $metadata_entry;
|
|---|
| 802 | }
|
|---|
| 803 | #write the entry to the file
|
|---|
| 804 | print $handle $metadata_entry;
|
|---|
| 805 |
|
|---|
| 806 | }
|
|---|
| 807 |
|
|---|
| 808 | print $handle "\n" . ('-' x 70) . "\n";
|
|---|
| 809 | }
|
|---|
| 810 |
|
|---|
| 811 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 812 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'},0, $self->{'gli'});
|
|---|
| 813 |
|
|---|
| 814 | # output classification information
|
|---|
| 815 | &classify::output_classify_info ($self->{'classifiers'}, $handle,
|
|---|
| 816 | $self->{'remove_empty_classifications'},
|
|---|
| 817 | $self->{'gli'});
|
|---|
| 818 |
|
|---|
| 819 |
|
|---|
| 820 | #output doclist
|
|---|
| 821 | my @doclist = $self->{'buildproc'}->get_doc_list();
|
|---|
| 822 | my $docs = join (";",@doclist);
|
|---|
| 823 | print $handle "[browselist]\n";
|
|---|
| 824 | print $handle "<hastxt>0\n";
|
|---|
| 825 | print $handle "<childtype>VList\n";
|
|---|
| 826 | print $handle "<numleafdocs>" . ($#doclist+1) . "\n";
|
|---|
| 827 | print $handle "<thistype>Invisible\n";
|
|---|
| 828 | print $handle "<contains>$docs";
|
|---|
| 829 | print $handle "\n" . ('-' x 70) . "\n";
|
|---|
| 830 |
|
|---|
| 831 | close ($handle) if !$self->{'debug'};
|
|---|
| 832 |
|
|---|
| 833 | print STDERR "</Stage>\n" if $self->{'gli'};
|
|---|
| 834 | }
|
|---|
| 835 |
|
|---|
| 836 | sub collect_specific {
|
|---|
| 837 | my $self = shift (@_);
|
|---|
| 838 | }
|
|---|
| 839 |
|
|---|
| 840 | sub make_auxiliary_files {
|
|---|
| 841 | my $self = shift (@_);
|
|---|
| 842 | my ($index);
|
|---|
| 843 | my $build_cfg = {};
|
|---|
| 844 | my $outhandle = $self->{'outhandle'};
|
|---|
| 845 |
|
|---|
| 846 | print $outhandle "\n*** creating auxiliary files \n" if ($self->{'verbosity'} >= 1);
|
|---|
| 847 | print STDERR "<Stage name='CreatingAuxilary'>\n" if $self->{'gli'};
|
|---|
| 848 |
|
|---|
| 849 | # get the text directory
|
|---|
| 850 | &util::mk_all_dir ($self->{'build_dir'});
|
|---|
| 851 |
|
|---|
| 852 | # store the build date
|
|---|
| 853 | $build_cfg->{'builddate'} = time;
|
|---|
| 854 | $build_cfg->{'indexstem'} = $self->{'collection'};
|
|---|
| 855 | # store the number of documents and number of bytes
|
|---|
| 856 | $build_cfg->{'numdocs'} = $self->{'buildproc'}->get_num_docs();
|
|---|
| 857 | $build_cfg->{'numbytes'} = $self->{'buildproc'}->get_num_bytes();
|
|---|
| 858 |
|
|---|
| 859 | # get additional stats from mg
|
|---|
| 860 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 861 | my $exe = &util::get_os_exe ();
|
|---|
| 862 | my $mgstat_exe = &util::filename_cat($exedir, "mgstat$exe");
|
|---|
| 863 | my $input_file = &util::filename_cat ("text", $self->{'collection'});
|
|---|
| 864 | if (!-e "$mgstat_exe" || !open (PIPEIN, "mgstat$exe -d \"$self->{'build_dir'}\" -f \"$input_file\" |")) {
|
|---|
| 865 | print $outhandle "Warning: Couldn't open pipe to $mgstat_exe to get additional stats\n";
|
|---|
| 866 | } else {
|
|---|
| 867 | my $line = "";
|
|---|
| 868 | while (defined ($line = <PIPEIN>)) {
|
|---|
| 869 | if ($line =~ /^Words in collection \[dict\]\s+:\s+(\d+)/) {
|
|---|
| 870 | ($build_cfg->{'numwords'}) = $1;
|
|---|
| 871 | } elsif ($line =~ /^Documents\s+:\s+(\d+)/) {
|
|---|
| 872 | ($build_cfg->{'numsections'}) = $1;
|
|---|
| 873 | }
|
|---|
| 874 | }
|
|---|
| 875 | close PIPEIN;
|
|---|
| 876 | }
|
|---|
| 877 |
|
|---|
| 878 | # store the mapping between the index names and the directory names
|
|---|
| 879 | # the index map is used to determine what indexes there are, so any that are not built should not be put into the map.
|
|---|
| 880 | my @indexmap = ();
|
|---|
| 881 | foreach my $index (@{$self->{'index_mapping'}->{'indexmaporder'}}) {
|
|---|
| 882 | if (not defined ($self->{'notbuilt'}->{$index})) {
|
|---|
| 883 | push (@indexmap, "$index\-\>$self->{'index_mapping'}->{'indexmap'}->{$index}");
|
|---|
| 884 | }
|
|---|
| 885 | }
|
|---|
| 886 | $build_cfg->{'indexmap'} = \@indexmap;
|
|---|
| 887 |
|
|---|
| 888 | my @subcollectionmap = ();
|
|---|
| 889 | foreach my $subcollection (@{$self->{'index_mapping'}->{'subcollectionmaporder'}}) {
|
|---|
| 890 | push (@subcollectionmap, "$subcollection\-\>" .
|
|---|
| 891 | $self->{'index_mapping'}->{'subcollectionmap'}->{$subcollection});
|
|---|
| 892 | }
|
|---|
| 893 | $build_cfg->{'subcollectionmap'} = \@subcollectionmap if scalar (@subcollectionmap);
|
|---|
| 894 |
|
|---|
| 895 | my @languagemap = ();
|
|---|
| 896 | foreach my $language (@{$self->{'index_mapping'}->{'languagemaporder'}}) {
|
|---|
| 897 | push (@languagemap, "$language\-\>" .
|
|---|
| 898 | $self->{'index_mapping'}->{'languagemap'}->{$language});
|
|---|
| 899 | }
|
|---|
| 900 | $build_cfg->{'languagemap'} = \@languagemap if scalar (@languagemap);
|
|---|
| 901 |
|
|---|
| 902 | #$build_cfg->{'notbuilt'} = $self->{'notbuilt'} if scalar @{$self->{'notbuilt'}};
|
|---|
| 903 | my @notbuilt = ();
|
|---|
| 904 | foreach my $nb (keys %{$self->{'notbuilt'}}) {
|
|---|
| 905 | push (@notbuilt, $nb);
|
|---|
| 906 | }
|
|---|
| 907 | $build_cfg->{'notbuilt'} = \@notbuilt if scalar (@notbuilt);
|
|---|
| 908 | $build_cfg->{'maxnumeric'} = 4;
|
|---|
| 909 | if (defined($self->{'collect_cfg'}->{'maxnumeric'}) &&
|
|---|
| 910 | $self->{'collect_cfg'}->{'maxnumeric'} =~ /^\d+$/) {
|
|---|
| 911 | $build_cfg->{'maxnumeric'} = $self->{'collect_cfg'}->{'maxnumeric'};
|
|---|
| 912 | }
|
|---|
| 913 |
|
|---|
| 914 | # write out the build information
|
|---|
| 915 | &cfgread::write_cfg_file("$self->{'build_dir'}/build.cfg", $build_cfg,
|
|---|
| 916 | '^(builddate|numdocs|numbytes|numwords|numsections|maxnumeric|indexstem)$',
|
|---|
| 917 | '^(indexmap|subcollectionmap|languagemap|notbuilt)$');
|
|---|
| 918 |
|
|---|
| 919 | print STDERR "</Stage>\n" if $self->{'gli'};
|
|---|
| 920 | }
|
|---|
| 921 |
|
|---|
| 922 | sub deinit {
|
|---|
| 923 | my $self = shift (@_);
|
|---|
| 924 |
|
|---|
| 925 | &plugin::deinit($self->{'pluginfo'},$self->{'buildproc'});
|
|---|
| 926 | }
|
|---|
| 927 |
|
|---|
| 928 | sub print_stats {
|
|---|
| 929 | my $self = shift (@_);
|
|---|
| 930 |
|
|---|
| 931 | my $outhandle = $self->{'outhandle'};
|
|---|
| 932 | my $indexing_text = $self->{'buildproc'}->get_indexing_text();
|
|---|
| 933 | my $index = $self->{'buildproc'}->get_index();
|
|---|
| 934 | my $num_bytes = $self->{'buildproc'}->get_num_bytes();
|
|---|
| 935 | my $num_processed_bytes = $self->{'buildproc'}->get_num_processed_bytes();
|
|---|
| 936 |
|
|---|
| 937 | if ($indexing_text) {
|
|---|
| 938 | print $outhandle "Stats (Creating index $index)\n";
|
|---|
| 939 | } else {
|
|---|
| 940 | print $outhandle "Stats (Compressing text from $index)\n";
|
|---|
| 941 | }
|
|---|
| 942 | print $outhandle "Total bytes in collection: $num_bytes\n";
|
|---|
| 943 | print $outhandle "Total bytes in $index: $num_processed_bytes\n";
|
|---|
| 944 |
|
|---|
| 945 | if ($num_processed_bytes < 50 && ($indexing_text || !$self->{'no_text'})) {
|
|---|
| 946 | print $outhandle "***************\n";
|
|---|
| 947 | if ($indexing_text) {
|
|---|
| 948 | print $outhandle "WARNING: There is very little or no text to process for $index\n";
|
|---|
| 949 | } elsif (!$self->{'no_text'}) {
|
|---|
| 950 | print $outhandle "WARNING: There is very little or no text to compress\n";
|
|---|
| 951 | }
|
|---|
| 952 | print $outhandle " Was this your intention?\n";
|
|---|
| 953 | print $outhandle "***************\n";
|
|---|
| 954 | print STDERR "<Warning name='LittleOrNoText'>\n" if $self->{'gli'};
|
|---|
| 955 | }
|
|---|
| 956 | }
|
|---|
| 957 |
|
|---|
| 958 | 1;
|
|---|
| 959 |
|
|---|
| 960 |
|
|---|
| 961 |
|
|---|
