| 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 |
|
|---|
| 34 | $maxdocsize = 12000;
|
|---|
| 35 |
|
|---|
| 36 | %wanted_index_files = ('td'=>1,
|
|---|
| 37 | 't'=>1,
|
|---|
| 38 | 'idb'=>1,
|
|---|
| 39 | 'ib1'=>1,
|
|---|
| 40 | 'ib2'=>1,
|
|---|
| 41 | 'ib3'=>1,
|
|---|
| 42 | 'i'=>1,
|
|---|
| 43 | 'ip'=>1,
|
|---|
| 44 | 'tiw'=>1,
|
|---|
| 45 | 'wa'=>1);
|
|---|
| 46 |
|
|---|
| 47 |
|
|---|
| 48 | sub new {
|
|---|
| 49 | my ($class, $collection, $source_dir, $build_dir, $verbosity,
|
|---|
| 50 | $maxdocs, $debug, $keepold, $allclassifications) = @_;
|
|---|
| 51 |
|
|---|
| 52 | # create an mgbuilder object
|
|---|
| 53 | my $self = bless {'collection'=>$collection,
|
|---|
| 54 | 'source_dir'=>$source_dir,
|
|---|
| 55 | 'build_dir'=>$build_dir,
|
|---|
| 56 | 'verbosity'=>$verbosity,
|
|---|
| 57 | 'maxdocs'=>$maxdocs,
|
|---|
| 58 | 'debug'=>$debug,
|
|---|
| 59 | 'keepold'=>$keepold,
|
|---|
| 60 | 'allclassifications'=>$allclassifications,
|
|---|
| 61 | 'notbuilt'=>[] # indexes not built
|
|---|
| 62 | }, $class;
|
|---|
| 63 |
|
|---|
| 64 |
|
|---|
| 65 | # read in the collection configuration file
|
|---|
| 66 | my $colcfgname = "$ENV{'GSDLCOLLECTDIR'}/etc/collect.cfg";
|
|---|
| 67 | if (!-e $colcfgname) {
|
|---|
| 68 | die "mgbuilder::new - couldn't find collect.cfg for collection $collection\n";
|
|---|
| 69 | }
|
|---|
| 70 | $self->{'collect_cfg'} = &colcfg::read_collect_cfg ($colcfgname);
|
|---|
| 71 |
|
|---|
| 72 | # sort out subcollection indexes
|
|---|
| 73 | if (defined $self->{'collect_cfg'}->{'indexsubcollections'}) {
|
|---|
| 74 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 75 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 76 | foreach $subcollection (@{$self->{'collect_cfg'}->{'indexsubcollections'}}) {
|
|---|
| 77 | foreach $index (@$indexes) {
|
|---|
| 78 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$subcollection");
|
|---|
| 79 | }
|
|---|
| 80 | }
|
|---|
| 81 | }
|
|---|
| 82 |
|
|---|
| 83 | # sort out language subindexes
|
|---|
| 84 | if (defined $self->{'collect_cfg'}->{'languages'}) {
|
|---|
| 85 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 86 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 87 | foreach $language (@{$self->{'collect_cfg'}->{'languages'}}) {
|
|---|
| 88 | foreach $index (@$indexes) {
|
|---|
| 89 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$language");
|
|---|
| 90 | }
|
|---|
| 91 | }
|
|---|
| 92 | }
|
|---|
| 93 |
|
|---|
| 94 | # get the list of plugins for this collection
|
|---|
| 95 | my $plugins = [];
|
|---|
| 96 | if (defined $self->{'collect_cfg'}->{'plugin'}) {
|
|---|
| 97 | $plugins = $self->{'collect_cfg'}->{'plugin'};
|
|---|
| 98 | }
|
|---|
| 99 |
|
|---|
| 100 | # load all the plugins
|
|---|
| 101 | $self->{'pluginfo'} = &plugin::load_plugins ($plugins, $verbosity);
|
|---|
| 102 | if (scalar(@{$self->{'pluginfo'}}) == 0) {
|
|---|
| 103 | print STDERR "No plugins were loaded.\n";
|
|---|
| 104 | die "\n";
|
|---|
| 105 | }
|
|---|
| 106 |
|
|---|
| 107 | # get the list of classifiers for this collection
|
|---|
| 108 | my $classifiers = [];
|
|---|
| 109 | if (defined $self->{'collect_cfg'}->{'classify'}) {
|
|---|
| 110 | $classifiers = $self->{'collect_cfg'}->{'classify'};
|
|---|
| 111 | }
|
|---|
| 112 |
|
|---|
| 113 | # load all the classifiers
|
|---|
| 114 | $self->{'classifiers'} = &classify::load_classifiers ($classifiers);
|
|---|
| 115 |
|
|---|
| 116 | # load up any dontgdbm fields
|
|---|
| 117 | $self->{'dontgdbm'} = {};
|
|---|
| 118 | if (defined ($self->{'collect_cfg'}->{'dontgdbm'})) {
|
|---|
| 119 | foreach $dg (@{$self->{'collect_cfg'}->{'dontgdbm'}}) {
|
|---|
| 120 | $self->{'dontgdbm'}->{$dg} = 1;
|
|---|
| 121 | }
|
|---|
| 122 | }
|
|---|
| 123 |
|
|---|
| 124 | # load up the document processor for building
|
|---|
| 125 | # if a buildproc class has been created for this collection, use it
|
|---|
| 126 | # otherwise, use the mg buildproc
|
|---|
| 127 | my ($buildprocdir, $buildproctype);
|
|---|
| 128 | if (-e "$ENV{'GSDLCOLLECTDIR'}/perllib/${collection}buildproc.pm") {
|
|---|
| 129 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/perllib";
|
|---|
| 130 | $buildproctype = "${collection}buildproc";
|
|---|
| 131 | } else {
|
|---|
| 132 | $buildprocdir = "$ENV{'GSDLHOME'}/perllib";
|
|---|
| 133 | $buildproctype = "mgbuildproc";
|
|---|
| 134 | }
|
|---|
| 135 | require "$buildprocdir/$buildproctype.pm";
|
|---|
| 136 |
|
|---|
| 137 | eval("\$self->{'buildproc'} = new $buildproctype(\$collection, " .
|
|---|
| 138 | "\$source_dir, \$build_dir, \$verbosity)");
|
|---|
| 139 | die "$@" if $@;
|
|---|
| 140 |
|
|---|
| 141 |
|
|---|
| 142 | return $self;
|
|---|
| 143 | }
|
|---|
| 144 |
|
|---|
| 145 | sub init {
|
|---|
| 146 | my $self = shift (@_);
|
|---|
| 147 |
|
|---|
| 148 | if (!$self->{'debug'} && !$self->{'keepold'}) {
|
|---|
| 149 | # remove any old builds
|
|---|
| 150 | &util::rm_r($self->{'build_dir'});
|
|---|
| 151 | &util::mk_all_dir($self->{'build_dir'});
|
|---|
| 152 |
|
|---|
| 153 | # make the text directory
|
|---|
| 154 | my $textdir = "$self->{'build_dir'}/text";
|
|---|
| 155 | &util::mk_all_dir($textdir);
|
|---|
| 156 | }
|
|---|
| 157 | }
|
|---|
| 158 |
|
|---|
| 159 | sub compress_text {
|
|---|
| 160 | my $self = shift (@_);
|
|---|
| 161 | my ($textindex) = @_;
|
|---|
| 162 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 163 | my $exe = &util::get_os_exe ();
|
|---|
| 164 | my $mg_passes_exe = &util::filename_cat($exedir, "mg_passes$exe");
|
|---|
| 165 | my $mg_compression_dict_exe = &util::filename_cat($exedir, "mg_compression_dict$exe");
|
|---|
| 166 |
|
|---|
| 167 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, "text"));
|
|---|
| 168 | my $basefilename = "text/$self->{'collection'}";
|
|---|
| 169 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, $basefilename);
|
|---|
| 170 |
|
|---|
| 171 | my $osextra = "";
|
|---|
| 172 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
|---|
| 173 | $fulltextprefix =~ s/\//\\/g;
|
|---|
| 174 | } else {
|
|---|
| 175 | $osextra = " -d /";
|
|---|
| 176 | }
|
|---|
| 177 |
|
|---|
| 178 | print STDERR "\n*** creating the compressed text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 179 |
|
|---|
| 180 | # collect the statistics for the text
|
|---|
| 181 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
|---|
| 182 | print STDERR "\n collecting text statistics\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 183 |
|
|---|
| 184 | my ($handle);
|
|---|
| 185 | if ($self->{'debug'}) {
|
|---|
| 186 | $handle = STDOUT;
|
|---|
| 187 | } else {
|
|---|
| 188 | if (!-e "$mg_passes_exe" ||
|
|---|
| 189 | !open (PIPEOUT, "| $mg_passes_exe -f $fulltextprefix -b $maxdocsize -T1 $osextra")) {
|
|---|
| 190 | die "mgbuilder::compress_text - couldn't run $mg_passes_exe\n";
|
|---|
| 191 | }
|
|---|
| 192 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 193 | }
|
|---|
| 194 |
|
|---|
| 195 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| 196 | $self->{'buildproc'}->set_mode ('text');
|
|---|
| 197 | $self->{'buildproc'}->set_index ($textindex);
|
|---|
| 198 | $self->{'buildproc'}->set_indexing_text (0);
|
|---|
| 199 | $self->{'buildproc'}->reset();
|
|---|
| 200 | &plugin::begin($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 201 | $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 202 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 203 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 204 | &plugin::end($self->{'pluginfo'});
|
|---|
| 205 | close (PIPEOUT);
|
|---|
| 206 |
|
|---|
| 207 | close ($handle) unless $self->{'debug'};
|
|---|
| 208 |
|
|---|
| 209 | $self->print_stats();
|
|---|
| 210 |
|
|---|
| 211 | # create the compression dictionary
|
|---|
| 212 | # the compression dictionary is built by assuming the stats are from a seed
|
|---|
| 213 | # dictionary (-S), if a novel word is encountered it is spelled out (-H),
|
|---|
| 214 | # and the resulting dictionary must be less than 5 meg with the most frequent
|
|---|
| 215 | # words being put into the dictionary first (-2 -k 5120)
|
|---|
| 216 | if (!$self->{'debug'}) {
|
|---|
| 217 | print STDERR "\n creating the compression dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 218 | if (!-e "$mg_compression_dict_exe") {
|
|---|
| 219 | die "mgbuilder::compress_text - couldn't run $mg_compression_dict_exe\n";
|
|---|
| 220 | }
|
|---|
| 221 | system ("$mg_compression_dict_exe -f $fulltextprefix -S -H -2 -k 5120 $osextra");
|
|---|
| 222 |
|
|---|
| 223 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
|---|
| 224 | if (!-e "$mg_passes_exe" ||
|
|---|
| 225 | !open ($handle, "| $mg_passes_exe -f $fulltextprefix -b $maxdocsize -T2 $osextra")) {
|
|---|
| 226 | die "mgbuilder::compress_text - couldn't run $mg_passes_exe\n";
|
|---|
| 227 | }
|
|---|
| 228 | }
|
|---|
| 229 |
|
|---|
| 230 | $self->{'buildproc'}->reset();
|
|---|
| 231 | # compress the text
|
|---|
| 232 | print STDERR "\n compressing the text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 233 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 234 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 235 | close ($handle) unless $self->{'debug'};
|
|---|
| 236 |
|
|---|
| 237 | $self->print_stats();
|
|---|
| 238 | }
|
|---|
| 239 |
|
|---|
| 240 | sub want_built {
|
|---|
| 241 | my $self = shift (@_);
|
|---|
| 242 | my ($index) = @_;
|
|---|
| 243 |
|
|---|
| 244 | if (defined ($self->{'collect_cfg'}->{'dontbuild'})) {
|
|---|
| 245 | foreach $checkstr (@{$self->{'collect_cfg'}->{'dontbuild'}}) {
|
|---|
| 246 | if ($index =~ /^$checkstr$/) {
|
|---|
| 247 | push (@{$self->{'notbuilt'}}, $self->{'index_mapping'}->{$index});
|
|---|
| 248 | return 0;
|
|---|
| 249 | }
|
|---|
| 250 | }
|
|---|
| 251 | }
|
|---|
| 252 |
|
|---|
| 253 | return 1;
|
|---|
| 254 | }
|
|---|
| 255 |
|
|---|
| 256 | sub build_indexes {
|
|---|
| 257 | my $self = shift (@_);
|
|---|
| 258 | my ($indexname) = @_;
|
|---|
| 259 |
|
|---|
| 260 | my $indexes = [];
|
|---|
| 261 | if (defined $indexname && $indexname =~ /\w/) {
|
|---|
| 262 | push @$indexes, $indexname;
|
|---|
| 263 | } else {
|
|---|
| 264 | $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 265 | }
|
|---|
| 266 |
|
|---|
| 267 | # create the mapping between the index descriptions
|
|---|
| 268 | # and their directory names
|
|---|
| 269 | $self->{'index_mapping'} = $self->create_index_mapping ($indexes);
|
|---|
| 270 |
|
|---|
| 271 | # build each of the indexes
|
|---|
| 272 | foreach $index (@$indexes) {
|
|---|
| 273 | if ($self->want_built($index)) {
|
|---|
| 274 | print STDERR "\n*** building index $index in subdirectory " .
|
|---|
| 275 | "$self->{'index_mapping'}->{$index}\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 276 | $self->build_index($index);
|
|---|
| 277 | } else {
|
|---|
| 278 | print STDERR "\n*** ignoring index $index\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 279 | }
|
|---|
| 280 | }
|
|---|
| 281 | }
|
|---|
| 282 |
|
|---|
| 283 | # creates directory names for each of the index descriptions
|
|---|
| 284 | sub create_index_mapping {
|
|---|
| 285 | my $self = shift (@_);
|
|---|
| 286 | my ($indexes) = @_;
|
|---|
| 287 |
|
|---|
| 288 | my %mapping = ();
|
|---|
| 289 | $mapping{'indexmaporder'} = [];
|
|---|
| 290 | $mapping{'subcollectionmaporder'} = [];
|
|---|
| 291 | $mapping{'languagemaporder'} = [];
|
|---|
| 292 |
|
|---|
| 293 | # dirnames is used to check for collisions. Start this off
|
|---|
| 294 | # with the manditory directory names
|
|---|
| 295 | my %dirnames = ('text'=>'text',
|
|---|
| 296 | 'extra'=>'extra');
|
|---|
| 297 | my %pnames = ('index' => '', 'subcollection' => '', 'languages' => '');
|
|---|
| 298 |
|
|---|
| 299 | foreach $index (@$indexes) {
|
|---|
| 300 | my ($level, $gran, $subcollection, $languages) = split (":", $index);
|
|---|
| 301 |
|
|---|
| 302 | # the directory name starts with the first character of the index level
|
|---|
| 303 | my ($pindex) = $level =~ /^(.)/;
|
|---|
| 304 |
|
|---|
| 305 | # next comes a processed version of the index
|
|---|
| 306 | $pindex .= $self->process_field ($gran);
|
|---|
| 307 | $pindex = lc ($pindex);
|
|---|
| 308 |
|
|---|
| 309 | # next comes a processed version of the subcollection if there is one.
|
|---|
| 310 | my $psub = $self->process_field ($subcollection);
|
|---|
| 311 | $psub = lc ($psub);
|
|---|
| 312 |
|
|---|
| 313 | # next comes a processed version of the language if there is one.
|
|---|
| 314 | my $plang = $self->process_field ($languages);
|
|---|
| 315 | $plang = lc ($plang);
|
|---|
| 316 |
|
|---|
| 317 | my $dirname = $pindex . $psub . $plang;
|
|---|
| 318 |
|
|---|
| 319 | # check to be sure all index names are unique
|
|---|
| 320 | while (defined ($dirnames{$dirname})) {
|
|---|
| 321 | $dirname = $self->make_unique (\%pnames, $index, \$pindex, \$psub, \$plang);
|
|---|
| 322 | }
|
|---|
| 323 |
|
|---|
| 324 | # store the mapping orders as well as the maps
|
|---|
| 325 | if (!defined $mapping{'indexmap'}{"$level:$gran"}) {
|
|---|
| 326 | $mapping{'indexmap'}{"$level:$gran"} = $pindex;
|
|---|
| 327 | push (@{$mapping{'indexmaporder'}}, "$level:$gran");
|
|---|
| 328 | }
|
|---|
| 329 | if ($psub =~ /\w/ && !defined ($mapping{'subcollectionmap'}{$subcollection})) {
|
|---|
| 330 | $mapping{'subcollectionmap'}{$subcollection} = $psub;
|
|---|
| 331 | push (@{$mapping{'subcollectionmaporder'}}, $subcollection);
|
|---|
| 332 | }
|
|---|
| 333 | if ($plang =~ /\w/ && !defined ($mapping{'languagemap'}{$languages})) {
|
|---|
| 334 | $mapping{'languagemap'}{$languages} = $plang;
|
|---|
| 335 | push (@{$mapping{'languagemaporder'}}, $language);
|
|---|
| 336 | }
|
|---|
| 337 | $mapping{$index} = $dirname;
|
|---|
| 338 | $dirnames{$dirname} = $index;
|
|---|
| 339 | $pnames{'index'}{$pindex} = "$level:$gran";
|
|---|
| 340 | $pnames{'subcollection'}{$psub} = $subcollection;
|
|---|
| 341 | $pnames{'languages'}{$plang} = $languages;
|
|---|
| 342 | }
|
|---|
| 343 |
|
|---|
| 344 | return \%mapping;
|
|---|
| 345 | }
|
|---|
| 346 |
|
|---|
| 347 | # returns a processed version of a field.
|
|---|
| 348 | # if the field has only one component the processed
|
|---|
| 349 | # version will contain the first character and next consonant
|
|---|
| 350 | # of that componant - otherwise it will contain the first
|
|---|
| 351 | # character of the first two components
|
|---|
| 352 | sub process_field {
|
|---|
| 353 | my $self = shift (@_);
|
|---|
| 354 | my ($field) = @_;
|
|---|
| 355 |
|
|---|
| 356 | return "" unless (defined ($field) && $field =~ /\w/);
|
|---|
| 357 |
|
|---|
| 358 | my @components = split /,/, $field;
|
|---|
| 359 | if (scalar @components >= 2) {
|
|---|
| 360 | splice (@components, 2);
|
|---|
| 361 | map {s/^(.).*$/$1/;} @components;
|
|---|
| 362 | return join("", @components);
|
|---|
| 363 | } else {
|
|---|
| 364 | my ($a, $b) = $field =~ /^(.).*?([bcdfghjklmnpqrstvwxyz])/i;
|
|---|
| 365 | ($a, $b) = $field =~ /^(.)(.)/ unless defined $a && defined $b;
|
|---|
| 366 | return "$a$b";
|
|---|
| 367 | }
|
|---|
| 368 | }
|
|---|
| 369 |
|
|---|
| 370 | sub make_unique {
|
|---|
| 371 | my $self = shift (@_);
|
|---|
| 372 | my ($namehash, $index, $indexref, $subref, $langref) = @_;
|
|---|
| 373 | my ($level, $gran, $subcollection, $languages) = split (":", $index);
|
|---|
| 374 |
|
|---|
| 375 | if ($namehash->{'index'}->{$$indexref} ne "$level:$gran") {
|
|---|
| 376 | $self->get_next_version ($indexref);
|
|---|
| 377 | } elsif ($namehash->{'subcollection'}->{$$subref} ne $subcollection) {
|
|---|
| 378 | $self->get_next_version ($subref);
|
|---|
| 379 | } elsif ($namehash->{'languages'}->{$$langref} ne $languages) {
|
|---|
| 380 | $self->get_next_version ($langref);
|
|---|
| 381 | }
|
|---|
| 382 | return "$$indexref$$subref$$langref";
|
|---|
| 383 | }
|
|---|
| 384 |
|
|---|
| 385 | sub get_next_version {
|
|---|
| 386 | my $self = shift (@_);
|
|---|
| 387 | my ($nameref) = @_;
|
|---|
| 388 |
|
|---|
| 389 | if ($$nameref =~ /(\d\d)$/) {
|
|---|
| 390 | my $num = $1; $num ++;
|
|---|
| 391 | $$nameref =~ s/\d\d$/$num/;
|
|---|
| 392 | } elsif ($$nameref =~ /(\d)$/) {
|
|---|
| 393 | my $num = $1;
|
|---|
| 394 | if ($num == 9) {$$nameref =~ s/\d\d$/10/;}
|
|---|
| 395 | else {$num ++; $$nameref =~ s/\d$/$num/;}
|
|---|
| 396 | } else {
|
|---|
| 397 | $$nameref =~ s/.$/0/;
|
|---|
| 398 | }
|
|---|
| 399 | }
|
|---|
| 400 |
|
|---|
| 401 | sub build_index {
|
|---|
| 402 | my $self = shift (@_);
|
|---|
| 403 | my ($index) = @_;
|
|---|
| 404 |
|
|---|
| 405 | # get the full index directory path and make sure it exists
|
|---|
| 406 | my $indexdir = $self->{'index_mapping'}->{$index};
|
|---|
| 407 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, $indexdir));
|
|---|
| 408 | my $fullindexprefix = &util::filename_cat ($self->{'build_dir'}, $indexdir,
|
|---|
| 409 | $self->{'collection'});
|
|---|
| 410 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, "text",
|
|---|
| 411 | $self->{'collection'});
|
|---|
| 412 |
|
|---|
| 413 | # get any os specific stuff
|
|---|
| 414 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 415 | my $exe = &util::get_os_exe ();
|
|---|
| 416 | my $mg_passes_exe = &util::filename_cat($exedir, "mg_passes$exe");
|
|---|
| 417 | my $mg_perf_hash_build_exe =
|
|---|
| 418 | &util::filename_cat($exedir, "mg_perf_hash_build$exe");
|
|---|
| 419 | my $mg_weights_build_exe =
|
|---|
| 420 | &util::filename_cat ($exedir, "mg_weights_build$exe");
|
|---|
| 421 | my $mg_invf_dict_exe =
|
|---|
| 422 | &util::filename_cat ($exedir, "mg_invf_dict$exe");
|
|---|
| 423 | my $mg_stem_idx_exe =
|
|---|
| 424 | &util::filename_cat ($exedir, "mg_stem_idx$exe");
|
|---|
| 425 |
|
|---|
| 426 | my $osextra = "";
|
|---|
| 427 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
|---|
| 428 | $fullindexprefix =~ s/\//\\/g;
|
|---|
| 429 | } else {
|
|---|
| 430 | $osextra = " -d /";
|
|---|
| 431 | }
|
|---|
| 432 |
|
|---|
| 433 | # get the index level from the index description
|
|---|
| 434 | # the index will be level 2 unless we are building a
|
|---|
| 435 | # paragraph level index
|
|---|
| 436 | my $index_level = 2;
|
|---|
| 437 | $index_level = 3 if $index =~ /^paragraph/i;
|
|---|
| 438 |
|
|---|
| 439 | # get the index expression if this index belongs
|
|---|
| 440 | # to a subcollection
|
|---|
| 441 | my $indexexparr = [];
|
|---|
| 442 | my ($level, $fields, $subcollection) = split (":", $index);
|
|---|
| 443 | my @subcollections = ();
|
|---|
| 444 | @subcollections = split /,/, $subcollection if (defined $subcollection);
|
|---|
| 445 |
|
|---|
| 446 | foreach $subcollection (@subcollections) {
|
|---|
| 447 | if (defined ($self->{'collect_cfg'}->{'subcollection'}->{$subcollection})) {
|
|---|
| 448 | push (@$indexexparr, $self->{'collect_cfg'}->{'subcollection'}->{$subcollection});
|
|---|
| 449 | }
|
|---|
| 450 | }
|
|---|
| 451 |
|
|---|
| 452 | # add expressions for languages if this index belongs to
|
|---|
| 453 | # a language subcollection
|
|---|
| 454 | foreach $language (@{$self->{'collect_cfg'}->{'languages'}}) {
|
|---|
| 455 | if ($language =~ s/^\!//) {
|
|---|
| 456 | push (@$indexexparr, "!Language/$language/");
|
|---|
| 457 | } else {
|
|---|
| 458 | push (@$indexexparr, "Language/$language/");
|
|---|
| 459 | }
|
|---|
| 460 | }
|
|---|
| 461 |
|
|---|
| 462 | # Build index dictionary. Uses verbatim stem method
|
|---|
| 463 | print STDERR "\n creating index dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 464 | my ($handle);
|
|---|
| 465 | if ($self->{'debug'}) {
|
|---|
| 466 | $handle = STDOUT;
|
|---|
| 467 | } else {
|
|---|
| 468 | if (!-e "$mg_passes_exe" ||
|
|---|
| 469 | !open (PIPEOUT, "| $mg_passes_exe -f $fullindexprefix -b $maxdocsize " .
|
|---|
| 470 | "-$index_level -m 32 -s 0 -G -t 10 -N1 $osextra")) {
|
|---|
| 471 | die "mgbuilder::build_index - couldn't run $mg_passes_exe\n";
|
|---|
| 472 | }
|
|---|
| 473 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 474 | }
|
|---|
| 475 |
|
|---|
| 476 | # set up the document processor
|
|---|
| 477 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| 478 | $self->{'buildproc'}->set_mode ('text');
|
|---|
| 479 | $self->{'buildproc'}->set_index ($index, $indexexparr);
|
|---|
| 480 | $self->{'buildproc'}->set_indexing_text (1);
|
|---|
| 481 |
|
|---|
| 482 | $self->{'buildproc'}->reset();
|
|---|
| 483 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 484 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 485 | close ($handle) unless $self->{'debug'};
|
|---|
| 486 |
|
|---|
| 487 | $self->print_stats();
|
|---|
| 488 |
|
|---|
| 489 | if (!$self->{'debug'}) {
|
|---|
| 490 | # create the perfect hash function
|
|---|
| 491 | if (!-e "$mg_perf_hash_build_exe") {
|
|---|
| 492 | die "mgbuilder::build_index - couldn't run $mg_perf_hash_build_exe\n";
|
|---|
| 493 | }
|
|---|
| 494 | system ("$mg_perf_hash_build_exe -f $fullindexprefix $osextra");
|
|---|
| 495 |
|
|---|
| 496 | if (!-e "$mg_passes_exe" ||
|
|---|
| 497 | !open ($handle, "| $mg_passes_exe -f $fullindexprefix -b $maxdocsize " .
|
|---|
| 498 | "-$index_level -c 3 -G -t 10 -N2 $osextra")) {
|
|---|
| 499 | die "mgbuilder::build_index - couldn't run $mg_passes_exe\n";
|
|---|
| 500 | }
|
|---|
| 501 | }
|
|---|
| 502 |
|
|---|
| 503 | # invert the text
|
|---|
| 504 | print STDERR "\n inverting the text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 505 |
|
|---|
| 506 | $self->{'buildproc'}->reset();
|
|---|
| 507 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 508 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 509 |
|
|---|
| 510 | $self->print_stats ();
|
|---|
| 511 |
|
|---|
| 512 | if (!$self->{'debug'}) {
|
|---|
| 513 |
|
|---|
| 514 | close ($handle);
|
|---|
| 515 |
|
|---|
| 516 | # create the weights file
|
|---|
| 517 | print STDERR "\n create the weights file\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 518 | if (!-e "$mg_weights_build_exe") {
|
|---|
| 519 | die "mgbuilder::build_index - couldn't run $mg_weights_build_exe\n";
|
|---|
| 520 | }
|
|---|
| 521 | system ("$mg_weights_build_exe -f $fullindexprefix -t $fulltextprefix $osextra");
|
|---|
| 522 |
|
|---|
| 523 | # create 'on-disk' stemmed dictionary
|
|---|
| 524 | print STDERR "\n creating 'on-disk' stemmed dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 525 | if (!-e "$mg_invf_dict_exe") {
|
|---|
| 526 | die "mgbuilder::build_index - couldn't run $mg_invf_dict_exe\n";
|
|---|
| 527 | }
|
|---|
| 528 | system ("$mg_invf_dict_exe -f $fullindexprefix $osextra");
|
|---|
| 529 |
|
|---|
| 530 |
|
|---|
| 531 | # creates stem index files for the various stemming methods
|
|---|
| 532 | print STDERR "\n creating stem indexes\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 533 | if (!-e "$mg_stem_idx_exe") {
|
|---|
| 534 | die "mgbuilder::build_index - couldn't run $mg_stem_idx_exe\n";
|
|---|
| 535 | }
|
|---|
| 536 | system ("$mg_stem_idx_exe -b 4096 -s1 -f $fullindexprefix $osextra");
|
|---|
| 537 | system ("$mg_stem_idx_exe -b 4096 -s2 -f $fullindexprefix $osextra");
|
|---|
| 538 | system ("$mg_stem_idx_exe -b 4096 -s3 -f $fullindexprefix $osextra");
|
|---|
| 539 |
|
|---|
| 540 |
|
|---|
| 541 | # remove unwanted files
|
|---|
| 542 | my $tmpdir = &util::filename_cat ($self->{'build_dir'}, $indexdir);
|
|---|
| 543 | opendir (DIR, $tmpdir) || die
|
|---|
| 544 | "mgbuilder::build_index - couldn't read directory $tmpdir\n";
|
|---|
| 545 | foreach $file (readdir(DIR)) {
|
|---|
| 546 | next if $file =~ /^\./;
|
|---|
| 547 | my ($suffix) = $file =~ /\.([^\.]+)$/;
|
|---|
| 548 | if (defined $suffix && !defined $wanted_index_files{$suffix}) {
|
|---|
| 549 | # delete it!
|
|---|
| 550 | print STDERR "deleting $file\n" if $self->{'verbosity'} > 2;
|
|---|
| 551 | &util::rm (&util::filename_cat ($tmpdir, $file));
|
|---|
| 552 | }
|
|---|
| 553 | }
|
|---|
| 554 | closedir (DIR);
|
|---|
| 555 | }
|
|---|
| 556 | }
|
|---|
| 557 |
|
|---|
| 558 | sub make_infodatabase {
|
|---|
| 559 | my $self = shift (@_);
|
|---|
| 560 | my $textdir = &util::filename_cat($self->{'build_dir'}, "text");
|
|---|
| 561 | my $assocdir = &util::filename_cat($self->{'build_dir'}, "assoc");
|
|---|
| 562 | &util::mk_all_dir ($textdir);
|
|---|
| 563 | &util::mk_all_dir ($assocdir);
|
|---|
| 564 |
|
|---|
| 565 | # get db name
|
|---|
| 566 | my $dbext = ".bdb";
|
|---|
| 567 | $dbext = ".ldb" if &util::is_little_endian();
|
|---|
| 568 | my $fulldbname = &util::filename_cat ($textdir, "$self->{'collection'}$dbext");
|
|---|
| 569 | $fulldbname =~ s/\//\\/g if ($ENV{'GSDLOS'} =~ /^windows$/i);
|
|---|
| 570 |
|
|---|
| 571 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 572 | my $exe = &util::get_os_exe ();
|
|---|
| 573 | my $txt2db_exe = &util::filename_cat($exedir, "txt2db$exe");
|
|---|
| 574 |
|
|---|
| 575 | print STDERR "\n*** creating the info database and processing associated files\n"
|
|---|
| 576 | if ($self->{'verbosity'} >= 1);
|
|---|
| 577 |
|
|---|
| 578 | # init all the classifiers
|
|---|
| 579 | &classify::init_classifiers ($self->{'classifiers'});
|
|---|
| 580 |
|
|---|
| 581 | # set up the document processor
|
|---|
| 582 | my ($handle);
|
|---|
| 583 | if ($self->{'debug'}) {
|
|---|
| 584 | $handle = STDOUT;
|
|---|
| 585 | } else {
|
|---|
| 586 | if (!-e "$txt2db_exe" || !open (PIPEOUT, "| $txt2db_exe $fulldbname")) {
|
|---|
| 587 | die "mgbuilder::make_infodatabase - couldn't run $txt2db_exe\n";
|
|---|
| 588 | }
|
|---|
| 589 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 590 | }
|
|---|
| 591 |
|
|---|
| 592 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| 593 | $self->{'buildproc'}->set_mode ('infodb');
|
|---|
| 594 | $self->{'buildproc'}->set_assocdir ($assocdir);
|
|---|
| 595 | $self->{'buildproc'}->set_dontgdbm ($self->{'dontgdbm'});
|
|---|
| 596 | $self->{'buildproc'}->set_classifiers ($self->{'classifiers'});
|
|---|
| 597 | $self->{'buildproc'}->set_indexing_text (0);
|
|---|
| 598 | $self->{'buildproc'}->reset();
|
|---|
| 599 |
|
|---|
| 600 | if (defined $self->{'collect_cfg'}->{'collectionmeta'}) {
|
|---|
| 601 |
|
|---|
| 602 | if (!defined $self->{'index_mapping'}) {
|
|---|
| 603 | $self->{'index_mapping'} =
|
|---|
| 604 | $self->create_index_mapping ($self->{'collect_cfg'}->{'indexes'});
|
|---|
| 605 | }
|
|---|
| 606 |
|
|---|
| 607 | print $handle "[collection]\n";
|
|---|
| 608 |
|
|---|
| 609 | foreach $cmeta (keys (%{$self->{'collect_cfg'}->{'collectionmeta'}})) {
|
|---|
| 610 | if ($cmeta =~ s/^\.//) {
|
|---|
| 611 | if (defined $self->{'index_mapping'}->{$cmeta}) {
|
|---|
| 612 | print $handle "<$self->{'index_mapping'}->{$cmeta}>" .
|
|---|
| 613 | $self->{'collect_cfg'}->{'collectionmeta'}->{".$cmeta"} . "\n";
|
|---|
| 614 | } else {
|
|---|
| 615 | print STDERR "mgbuilder: warning bad collectionmeta option '$cmeta' - ignored\n";
|
|---|
| 616 | }
|
|---|
| 617 | } else {
|
|---|
| 618 | print $handle "<$cmeta>$self->{'collect_cfg'}->{'collectionmeta'}->{$cmeta}\n";
|
|---|
| 619 | }
|
|---|
| 620 | }
|
|---|
| 621 | print $handle "\n" . ('-' x 70) . "\n";
|
|---|
| 622 |
|
|---|
| 623 | }
|
|---|
| 624 |
|
|---|
| 625 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 626 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 627 |
|
|---|
| 628 | # output classification information
|
|---|
| 629 | &classify::output_classify_info ($self->{'classifiers'}, $handle,
|
|---|
| 630 | $self->{'allclassifications'});
|
|---|
| 631 |
|
|---|
| 632 | close ($handle) if !$self->{'debug'};
|
|---|
| 633 | }
|
|---|
| 634 |
|
|---|
| 635 | sub collect_specific {
|
|---|
| 636 | my $self = shift (@_);
|
|---|
| 637 | }
|
|---|
| 638 |
|
|---|
| 639 | sub make_auxiliary_files {
|
|---|
| 640 | my $self = shift (@_);
|
|---|
| 641 | my ($index);
|
|---|
| 642 | my %build_cfg = ();
|
|---|
| 643 |
|
|---|
| 644 | print STDERR "\n*** creating auxiliary files \n" if ($self->{'verbosity'} >= 1);
|
|---|
| 645 |
|
|---|
| 646 | # get the text directory
|
|---|
| 647 | &util::mk_all_dir ($self->{'build_dir'});
|
|---|
| 648 |
|
|---|
| 649 | # store the build date
|
|---|
| 650 | $build_cfg->{'builddate'} = time;
|
|---|
| 651 |
|
|---|
| 652 | # store the number of documents and number of bytes
|
|---|
| 653 | $build_cfg->{'numdocs'} = $self->{'buildproc'}->get_num_docs();
|
|---|
| 654 | $build_cfg->{'numbytes'} = $self->{'buildproc'}->get_num_bytes();
|
|---|
| 655 |
|
|---|
| 656 | # store the mapping between the index names and the directory names
|
|---|
| 657 | my @indexmap = ();
|
|---|
| 658 | foreach $index (@{$self->{'index_mapping'}->{'indexmaporder'}}) {
|
|---|
| 659 | push (@indexmap, "$index\-\>$self->{'index_mapping'}->{'indexmap'}->{$index}");
|
|---|
| 660 | }
|
|---|
| 661 | $build_cfg->{'indexmap'} = \@indexmap;
|
|---|
| 662 |
|
|---|
| 663 | my @subcollectionmap = ();
|
|---|
| 664 | foreach $subcollection (@{$self->{'index_mapping'}->{'subcollectionmaporder'}}) {
|
|---|
| 665 | push (@subcollectionmap, "$subcollection\-\>" .
|
|---|
| 666 | $self->{'index_mapping'}->{'subcollectionmap'}->{$subcollection});
|
|---|
| 667 | }
|
|---|
| 668 | $build_cfg->{'subcollectionmap'} = \@subcollectionmap if scalar (@subcollectionmap);
|
|---|
| 669 |
|
|---|
| 670 | my @languagemap = ();
|
|---|
| 671 | foreach $language (@{$self->{'index_mapping'}->{'languagemaporder'}}) {
|
|---|
| 672 | push (@languagemap, "$language\-\>" .
|
|---|
| 673 | $self->{'index_mapping'}->{'languagemap'}->{$language});
|
|---|
| 674 | }
|
|---|
| 675 | $build_cfg->{'languagemap'} = \@languagemap if scalar (@languagemap);
|
|---|
| 676 |
|
|---|
| 677 | $build_cfg->{'notbuilt'} = $self->{'notbuilt'} if scalar @{$self->{'notbuilt'}};
|
|---|
| 678 |
|
|---|
| 679 | # write out the build information
|
|---|
| 680 | &cfgread::write_cfg_file("$self->{'build_dir'}/build.cfg", $build_cfg,
|
|---|
| 681 | '^(builddate|numdocs|numbytes)$',
|
|---|
| 682 | '^(indexmap|subcollectionmap|languagemap|notbuilt)$');
|
|---|
| 683 |
|
|---|
| 684 | }
|
|---|
| 685 |
|
|---|
| 686 | sub deinit {
|
|---|
| 687 | my $self = shift (@_);
|
|---|
| 688 | }
|
|---|
| 689 |
|
|---|
| 690 | sub print_stats {
|
|---|
| 691 | my $self = shift (@_);
|
|---|
| 692 |
|
|---|
| 693 | my $indexing_text = $self->{'buildproc'}->get_indexing_text();
|
|---|
| 694 | my $index = $self->{'buildproc'}->get_index();
|
|---|
| 695 | my $num_bytes = $self->{'buildproc'}->get_num_bytes();
|
|---|
| 696 | my $num_processed_bytes = $self->{'buildproc'}->get_num_processed_bytes();
|
|---|
| 697 |
|
|---|
| 698 | if ($indexing_text) {
|
|---|
| 699 | print STDERR "Stats (Creating index $index)\n";
|
|---|
| 700 | } else {
|
|---|
| 701 | print STDERR "Stats (Compressing text from $index)\n";
|
|---|
| 702 | }
|
|---|
| 703 | print STDERR "Total bytes in collection: $num_bytes\n";
|
|---|
| 704 | print STDERR "Total bytes in $index: $num_processed_bytes\n";
|
|---|
| 705 |
|
|---|
| 706 | if ($num_processed_bytes < 50) {
|
|---|
| 707 | print STDERR "***************\n";
|
|---|
| 708 | print STDERR "WARNING: There is very little or no text to process for $index\n";
|
|---|
| 709 | if ($indexing_text) {
|
|---|
| 710 | print STDERR "This may cause an error while attempting to build the index\n";
|
|---|
| 711 | } else {
|
|---|
| 712 | print STDERR "This may cause an error while attempting to compress the text\n";
|
|---|
| 713 | }
|
|---|
| 714 | print STDERR "***************\n";
|
|---|
| 715 | }
|
|---|
| 716 | }
|
|---|
| 717 |
|
|---|
| 718 | 1;
|
|---|
| 719 |
|
|---|
| 720 |
|
|---|
