| [537] | 1 | ###########################################################################
|
|---|
| [4] | 2 | #
|
|---|
| [537] | 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 | ###########################################################################
|
|---|
| [4] | 25 |
|
|---|
| 26 | package mgbuilder;
|
|---|
| 27 |
|
|---|
| [215] | 28 | use classify;
|
|---|
| [4] | 29 | use cfgread;
|
|---|
| 30 | use colcfg;
|
|---|
| 31 | use plugin;
|
|---|
| 32 | use util;
|
|---|
| [1304] | 33 | use FileHandle;
|
|---|
| [4] | 34 |
|
|---|
| [1304] | 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 |
|
|---|
| [4] | 47 | $maxdocsize = 12000;
|
|---|
| 48 |
|
|---|
| 49 | %wanted_index_files = ('td'=>1,
|
|---|
| 50 | 't'=>1,
|
|---|
| 51 | 'idb'=>1,
|
|---|
| 52 | 'ib1'=>1,
|
|---|
| 53 | 'ib2'=>1,
|
|---|
| 54 | 'ib3'=>1,
|
|---|
| 55 | 'i'=>1,
|
|---|
| 56 | 'ip'=>1,
|
|---|
| 57 | 'tiw'=>1,
|
|---|
| 58 | 'wa'=>1);
|
|---|
| 59 |
|
|---|
| 60 |
|
|---|
| 61 | sub new {
|
|---|
| [784] | 62 | my ($class, $collection, $source_dir, $build_dir, $verbosity,
|
|---|
| [1424] | 63 | $maxdocs, $debug, $keepold, $allclassifications, $outhandle) = @_;
|
|---|
| [4] | 64 |
|
|---|
| [1424] | 65 | $outhandle = STDERR unless defined $outhandle;
|
|---|
| 66 |
|
|---|
| [4] | 67 | # create an mgbuilder object
|
|---|
| 68 | my $self = bless {'collection'=>$collection,
|
|---|
| 69 | 'source_dir'=>$source_dir,
|
|---|
| 70 | 'build_dir'=>$build_dir,
|
|---|
| [215] | 71 | 'verbosity'=>$verbosity,
|
|---|
| [315] | 72 | 'maxdocs'=>$maxdocs,
|
|---|
| [782] | 73 | 'debug'=>$debug,
|
|---|
| [784] | 74 | 'keepold'=>$keepold,
|
|---|
| [486] | 75 | 'allclassifications'=>$allclassifications,
|
|---|
| [1424] | 76 | 'outhandle'=>$outhandle,
|
|---|
| [486] | 77 | 'notbuilt'=>[] # indexes not built
|
|---|
| [315] | 78 | }, $class;
|
|---|
| [4] | 79 |
|
|---|
| 80 |
|
|---|
| 81 | # read in the collection configuration file
|
|---|
| [215] | 82 | my $colcfgname = "$ENV{'GSDLCOLLECTDIR'}/etc/collect.cfg";
|
|---|
| 83 | if (!-e $colcfgname) {
|
|---|
| [4] | 84 | die "mgbuilder::new - couldn't find collect.cfg for collection $collection\n";
|
|---|
| 85 | }
|
|---|
| [215] | 86 | $self->{'collect_cfg'} = &colcfg::read_collect_cfg ($colcfgname);
|
|---|
| [4] | 87 |
|
|---|
| [69] | 88 | # sort out subcollection indexes
|
|---|
| 89 | if (defined $self->{'collect_cfg'}->{'indexsubcollections'}) {
|
|---|
| 90 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 91 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 92 | foreach $subcollection (@{$self->{'collect_cfg'}->{'indexsubcollections'}}) {
|
|---|
| 93 | foreach $index (@$indexes) {
|
|---|
| 94 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$subcollection");
|
|---|
| 95 | }
|
|---|
| 96 | }
|
|---|
| 97 | }
|
|---|
| 98 |
|
|---|
| [139] | 99 | # sort out language subindexes
|
|---|
| 100 | if (defined $self->{'collect_cfg'}->{'languages'}) {
|
|---|
| 101 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 102 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 103 | foreach $language (@{$self->{'collect_cfg'}->{'languages'}}) {
|
|---|
| 104 | foreach $index (@$indexes) {
|
|---|
| 105 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$language");
|
|---|
| 106 | }
|
|---|
| 107 | }
|
|---|
| 108 | }
|
|---|
| 109 |
|
|---|
| [1799] | 110 | # make sure that the same index isn't specified more than once
|
|---|
| 111 | my %tmphash = ();
|
|---|
| 112 | my @tmparray = @{$self->{'collect_cfg'}->{'indexes'}};
|
|---|
| 113 | $self->{'collect_cfg'}->{'indexes'} = [];
|
|---|
| 114 | foreach my $i (@tmparray) {
|
|---|
| 115 | if (!defined ($tmphash{$i})) {
|
|---|
| 116 | push (@{$self->{'collect_cfg'}->{'indexes'}}, $i);
|
|---|
| 117 | $tmphash{$i} = 1;
|
|---|
| 118 | }
|
|---|
| 119 | }
|
|---|
| 120 |
|
|---|
| [4] | 121 | # get the list of plugins for this collection
|
|---|
| [810] | 122 | my $plugins = [];
|
|---|
| 123 | if (defined $self->{'collect_cfg'}->{'plugin'}) {
|
|---|
| 124 | $plugins = $self->{'collect_cfg'}->{'plugin'};
|
|---|
| [4] | 125 | }
|
|---|
| 126 |
|
|---|
| 127 | # load all the plugins
|
|---|
| [1424] | 128 | $self->{'pluginfo'} = &plugin::load_plugins ($plugins, $verbosity, $outhandle);
|
|---|
| [4] | 129 | if (scalar(@{$self->{'pluginfo'}}) == 0) {
|
|---|
| [1424] | 130 | print $outhandle "No plugins were loaded.\n";
|
|---|
| [4] | 131 | die "\n";
|
|---|
| 132 | }
|
|---|
| 133 |
|
|---|
| [810] | 134 | # get the list of classifiers for this collection
|
|---|
| 135 | my $classifiers = [];
|
|---|
| 136 | if (defined $self->{'collect_cfg'}->{'classify'}) {
|
|---|
| 137 | $classifiers = $self->{'collect_cfg'}->{'classify'};
|
|---|
| 138 | }
|
|---|
| 139 |
|
|---|
| [215] | 140 | # load all the classifiers
|
|---|
| [1803] | 141 | $self->{'classifiers'} = &classify::load_classifiers ($classifiers, $build_dir, $outhandle);
|
|---|
| [215] | 142 |
|
|---|
| [780] | 143 | # load up any dontgdbm fields
|
|---|
| 144 | $self->{'dontgdbm'} = {};
|
|---|
| 145 | if (defined ($self->{'collect_cfg'}->{'dontgdbm'})) {
|
|---|
| 146 | foreach $dg (@{$self->{'collect_cfg'}->{'dontgdbm'}}) {
|
|---|
| 147 | $self->{'dontgdbm'}->{$dg} = 1;
|
|---|
| 148 | }
|
|---|
| 149 | }
|
|---|
| 150 |
|
|---|
| [4] | 151 | # load up the document processor for building
|
|---|
| 152 | # if a buildproc class has been created for this collection, use it
|
|---|
| 153 | # otherwise, use the mg buildproc
|
|---|
| 154 | my ($buildprocdir, $buildproctype);
|
|---|
| [134] | 155 | if (-e "$ENV{'GSDLCOLLECTDIR'}/perllib/${collection}buildproc.pm") {
|
|---|
| 156 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/perllib";
|
|---|
| [4] | 157 | $buildproctype = "${collection}buildproc";
|
|---|
| 158 | } else {
|
|---|
| [16] | 159 | $buildprocdir = "$ENV{'GSDLHOME'}/perllib";
|
|---|
| [4] | 160 | $buildproctype = "mgbuildproc";
|
|---|
| 161 | }
|
|---|
| 162 | require "$buildprocdir/$buildproctype.pm";
|
|---|
| 163 |
|
|---|
| 164 | eval("\$self->{'buildproc'} = new $buildproctype(\$collection, " .
|
|---|
| [1424] | 165 | "\$source_dir, \$build_dir, \$verbosity, \$outhandle)");
|
|---|
| [4] | 166 | die "$@" if $@;
|
|---|
| 167 |
|
|---|
| 168 | return $self;
|
|---|
| 169 | }
|
|---|
| 170 |
|
|---|
| 171 | sub init {
|
|---|
| 172 | my $self = shift (@_);
|
|---|
| 173 |
|
|---|
| [784] | 174 | if (!$self->{'debug'} && !$self->{'keepold'}) {
|
|---|
| [782] | 175 | # remove any old builds
|
|---|
| 176 | &util::rm_r($self->{'build_dir'});
|
|---|
| 177 | &util::mk_all_dir($self->{'build_dir'});
|
|---|
| [4] | 178 |
|
|---|
| [782] | 179 | # make the text directory
|
|---|
| 180 | my $textdir = "$self->{'build_dir'}/text";
|
|---|
| 181 | &util::mk_all_dir($textdir);
|
|---|
| 182 | }
|
|---|
| [4] | 183 | }
|
|---|
| 184 |
|
|---|
| 185 | sub compress_text {
|
|---|
| 186 | my $self = shift (@_);
|
|---|
| [134] | 187 | my ($textindex) = @_;
|
|---|
| [4] | 188 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 189 | my $exe = &util::get_os_exe ();
|
|---|
| [486] | 190 | my $mg_passes_exe = &util::filename_cat($exedir, "mg_passes$exe");
|
|---|
| 191 | my $mg_compression_dict_exe = &util::filename_cat($exedir, "mg_compression_dict$exe");
|
|---|
| [1424] | 192 | my $outhandle = $self->{'outhandle'};
|
|---|
| [4] | 193 |
|
|---|
| 194 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, "text"));
|
|---|
| 195 | my $basefilename = "text/$self->{'collection'}";
|
|---|
| 196 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, $basefilename);
|
|---|
| 197 |
|
|---|
| 198 | my $osextra = "";
|
|---|
| 199 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
|---|
| 200 | $fulltextprefix =~ s/\//\\/g;
|
|---|
| 201 | } else {
|
|---|
| 202 | $osextra = " -d /";
|
|---|
| 203 | }
|
|---|
| 204 |
|
|---|
| [1424] | 205 | print $outhandle "\n*** creating the compressed text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [4] | 206 |
|
|---|
| 207 | # collect the statistics for the text
|
|---|
| 208 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
|---|
| [1424] | 209 | print $outhandle "\n collecting text statistics\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [782] | 210 |
|
|---|
| 211 | my ($handle);
|
|---|
| 212 | if ($self->{'debug'}) {
|
|---|
| 213 | $handle = STDOUT;
|
|---|
| 214 | } else {
|
|---|
| 215 | if (!-e "$mg_passes_exe" ||
|
|---|
| [1679] | 216 | # !open (PIPEOUT, "| \"$mg_passes_exe\" -f \"$fulltextprefix\" -b $maxdocsize -T1 $osextra")) {
|
|---|
| 217 | !open (PIPEOUT, "| mg_passes$exe -f \"$fulltextprefix\" -b $maxdocsize -T1 $osextra")) {
|
|---|
| [782] | 218 | die "mgbuilder::compress_text - couldn't run $mg_passes_exe\n";
|
|---|
| 219 | }
|
|---|
| 220 | $handle = mgbuilder::PIPEOUT;
|
|---|
| [4] | 221 | }
|
|---|
| [782] | 222 |
|
|---|
| 223 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| 224 | $self->{'buildproc'}->set_mode ('text');
|
|---|
| 225 | $self->{'buildproc'}->set_index ($textindex);
|
|---|
| 226 | $self->{'buildproc'}->set_indexing_text (0);
|
|---|
| [4] | 227 | $self->{'buildproc'}->reset();
|
|---|
| [835] | 228 | &plugin::begin($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 229 | $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 230 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| 231 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| 232 | &plugin::end($self->{'pluginfo'});
|
|---|
| [4] | 233 |
|
|---|
| [782] | 234 | close ($handle) unless $self->{'debug'};
|
|---|
| 235 |
|
|---|
| [1251] | 236 | $self->print_stats();
|
|---|
| 237 |
|
|---|
| [4] | 238 | # create the compression dictionary
|
|---|
| 239 | # the compression dictionary is built by assuming the stats are from a seed
|
|---|
| 240 | # dictionary (-S), if a novel word is encountered it is spelled out (-H),
|
|---|
| 241 | # and the resulting dictionary must be less than 5 meg with the most frequent
|
|---|
| 242 | # words being put into the dictionary first (-2 -k 5120)
|
|---|
| [782] | 243 | if (!$self->{'debug'}) {
|
|---|
| [1424] | 244 | print $outhandle "\n creating the compression dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [782] | 245 | if (!-e "$mg_compression_dict_exe") {
|
|---|
| 246 | die "mgbuilder::compress_text - couldn't run $mg_compression_dict_exe\n";
|
|---|
| 247 | }
|
|---|
| [1679] | 248 | # system ("\"$mg_compression_dict_exe\" -f \"$fulltextprefix\" -S -H -2 -k 5120 $osextra");
|
|---|
| 249 | system ("mg_compression_dict$exe -f \"$fulltextprefix\" -S -H -2 -k 5120 $osextra");
|
|---|
| [782] | 250 |
|
|---|
| 251 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
|---|
| [1072] | 252 | if (!-e "$mg_passes_exe" ||
|
|---|
| [1679] | 253 | # !open ($handle, "| \"$mg_passes_exe\" -f \"$fulltextprefix\" -b $maxdocsize -T2 $osextra")) {
|
|---|
| 254 | !open ($handle, "| mg_passes$exe -f \"$fulltextprefix\" -b $maxdocsize -T2 $osextra")) {
|
|---|
| [1072] | 255 | die "mgbuilder::compress_text - couldn't run $mg_passes_exe\n";
|
|---|
| [782] | 256 | }
|
|---|
| [4] | 257 | }
|
|---|
| 258 |
|
|---|
| [782] | 259 | $self->{'buildproc'}->reset();
|
|---|
| [4] | 260 | # compress the text
|
|---|
| [1424] | 261 | print $outhandle "\n compressing the text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [4] | 262 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| [315] | 263 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| [782] | 264 | close ($handle) unless $self->{'debug'};
|
|---|
| [1251] | 265 |
|
|---|
| 266 | $self->print_stats();
|
|---|
| [4] | 267 | }
|
|---|
| 268 |
|
|---|
| [486] | 269 | sub want_built {
|
|---|
| 270 | my $self = shift (@_);
|
|---|
| 271 | my ($index) = @_;
|
|---|
| 272 |
|
|---|
| 273 | if (defined ($self->{'collect_cfg'}->{'dontbuild'})) {
|
|---|
| 274 | foreach $checkstr (@{$self->{'collect_cfg'}->{'dontbuild'}}) {
|
|---|
| 275 | if ($index =~ /^$checkstr$/) {
|
|---|
| 276 | push (@{$self->{'notbuilt'}}, $self->{'index_mapping'}->{$index});
|
|---|
| 277 | return 0;
|
|---|
| 278 | }
|
|---|
| 279 | }
|
|---|
| 280 | }
|
|---|
| 281 |
|
|---|
| 282 | return 1;
|
|---|
| 283 | }
|
|---|
| 284 |
|
|---|
| [4] | 285 | sub build_indexes {
|
|---|
| 286 | my $self = shift (@_);
|
|---|
| [782] | 287 | my ($indexname) = @_;
|
|---|
| [1424] | 288 | my $outhandle = $self->{'outhandle'};
|
|---|
| [4] | 289 |
|
|---|
| [782] | 290 | my $indexes = [];
|
|---|
| 291 | if (defined $indexname && $indexname =~ /\w/) {
|
|---|
| 292 | push @$indexes, $indexname;
|
|---|
| 293 | } else {
|
|---|
| 294 | $indexes = $self->{'collect_cfg'}->{'indexes'};
|
|---|
| 295 | }
|
|---|
| 296 |
|
|---|
| [4] | 297 | # create the mapping between the index descriptions
|
|---|
| 298 | # and their directory names
|
|---|
| 299 | $self->{'index_mapping'} = $self->create_index_mapping ($indexes);
|
|---|
| 300 |
|
|---|
| 301 | # build each of the indexes
|
|---|
| 302 | foreach $index (@$indexes) {
|
|---|
| [486] | 303 | if ($self->want_built($index)) {
|
|---|
| [1424] | 304 | print $outhandle "\n*** building index $index in subdirectory " .
|
|---|
| [486] | 305 | "$self->{'index_mapping'}->{$index}\n" if ($self->{'verbosity'} >= 1);
|
|---|
| 306 | $self->build_index($index);
|
|---|
| 307 | } else {
|
|---|
| [1424] | 308 | print $outhandle "\n*** ignoring index $index\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [486] | 309 | }
|
|---|
| [4] | 310 | }
|
|---|
| 311 | }
|
|---|
| 312 |
|
|---|
| 313 | # creates directory names for each of the index descriptions
|
|---|
| 314 | sub create_index_mapping {
|
|---|
| 315 | my $self = shift (@_);
|
|---|
| 316 | my ($indexes) = @_;
|
|---|
| 317 |
|
|---|
| 318 | my %mapping = ();
|
|---|
| [290] | 319 | $mapping{'indexmaporder'} = [];
|
|---|
| 320 | $mapping{'subcollectionmaporder'} = [];
|
|---|
| 321 | $mapping{'languagemaporder'} = [];
|
|---|
| [4] | 322 |
|
|---|
| 323 | # dirnames is used to check for collisions. Start this off
|
|---|
| 324 | # with the manditory directory names
|
|---|
| 325 | my %dirnames = ('text'=>'text',
|
|---|
| 326 | 'extra'=>'extra');
|
|---|
| [139] | 327 | my %pnames = ('index' => '', 'subcollection' => '', 'languages' => '');
|
|---|
| 328 |
|
|---|
| [4] | 329 | foreach $index (@$indexes) {
|
|---|
| [139] | 330 | my ($level, $gran, $subcollection, $languages) = split (":", $index);
|
|---|
| [4] | 331 |
|
|---|
| [139] | 332 | # the directory name starts with the first character of the index level
|
|---|
| 333 | my ($pindex) = $level =~ /^(.)/;
|
|---|
| [4] | 334 |
|
|---|
| [139] | 335 | # next comes a processed version of the index
|
|---|
| 336 | $pindex .= $self->process_field ($gran);
|
|---|
| 337 | $pindex = lc ($pindex);
|
|---|
| 338 |
|
|---|
| [69] | 339 | # next comes a processed version of the subcollection if there is one.
|
|---|
| [139] | 340 | my $psub = $self->process_field ($subcollection);
|
|---|
| 341 | $psub = lc ($psub);
|
|---|
| [69] | 342 |
|
|---|
| [139] | 343 | # next comes a processed version of the language if there is one.
|
|---|
| 344 | my $plang = $self->process_field ($languages);
|
|---|
| 345 | $plang = lc ($plang);
|
|---|
| [4] | 346 |
|
|---|
| [139] | 347 | my $dirname = $pindex . $psub . $plang;
|
|---|
| 348 |
|
|---|
| 349 | # check to be sure all index names are unique
|
|---|
| 350 | while (defined ($dirnames{$dirname})) {
|
|---|
| 351 | $dirname = $self->make_unique (\%pnames, $index, \$pindex, \$psub, \$plang);
|
|---|
| [4] | 352 | }
|
|---|
| [139] | 353 |
|
|---|
| [290] | 354 | # store the mapping orders as well as the maps
|
|---|
| 355 | if (!defined $mapping{'indexmap'}{"$level:$gran"}) {
|
|---|
| 356 | $mapping{'indexmap'}{"$level:$gran"} = $pindex;
|
|---|
| 357 | push (@{$mapping{'indexmaporder'}}, "$level:$gran");
|
|---|
| 358 | }
|
|---|
| 359 | if ($psub =~ /\w/ && !defined ($mapping{'subcollectionmap'}{$subcollection})) {
|
|---|
| 360 | $mapping{'subcollectionmap'}{$subcollection} = $psub;
|
|---|
| 361 | push (@{$mapping{'subcollectionmaporder'}}, $subcollection);
|
|---|
| 362 | }
|
|---|
| 363 | if ($plang =~ /\w/ && !defined ($mapping{'languagemap'}{$languages})) {
|
|---|
| 364 | $mapping{'languagemap'}{$languages} = $plang;
|
|---|
| 365 | push (@{$mapping{'languagemaporder'}}, $language);
|
|---|
| 366 | }
|
|---|
| [4] | 367 | $mapping{$index} = $dirname;
|
|---|
| 368 | $dirnames{$dirname} = $index;
|
|---|
| [139] | 369 | $pnames{'index'}{$pindex} = "$level:$gran";
|
|---|
| 370 | $pnames{'subcollection'}{$psub} = $subcollection;
|
|---|
| 371 | $pnames{'languages'}{$plang} = $languages;
|
|---|
| [4] | 372 | }
|
|---|
| 373 |
|
|---|
| 374 | return \%mapping;
|
|---|
| 375 | }
|
|---|
| 376 |
|
|---|
| [139] | 377 | # returns a processed version of a field.
|
|---|
| 378 | # if the field has only one component the processed
|
|---|
| 379 | # version will contain the first character and next consonant
|
|---|
| 380 | # of that componant - otherwise it will contain the first
|
|---|
| 381 | # character of the first two components
|
|---|
| 382 | sub process_field {
|
|---|
| 383 | my $self = shift (@_);
|
|---|
| 384 | my ($field) = @_;
|
|---|
| 385 |
|
|---|
| 386 | return "" unless (defined ($field) && $field =~ /\w/);
|
|---|
| [4] | 387 |
|
|---|
| [139] | 388 | my @components = split /,/, $field;
|
|---|
| 389 | if (scalar @components >= 2) {
|
|---|
| 390 | splice (@components, 2);
|
|---|
| 391 | map {s/^(.).*$/$1/;} @components;
|
|---|
| 392 | return join("", @components);
|
|---|
| 393 | } else {
|
|---|
| 394 | my ($a, $b) = $field =~ /^(.).*?([bcdfghjklmnpqrstvwxyz])/i;
|
|---|
| 395 | ($a, $b) = $field =~ /^(.)(.)/ unless defined $a && defined $b;
|
|---|
| 396 | return "$a$b";
|
|---|
| 397 | }
|
|---|
| 398 | }
|
|---|
| 399 |
|
|---|
| 400 | sub make_unique {
|
|---|
| 401 | my $self = shift (@_);
|
|---|
| 402 | my ($namehash, $index, $indexref, $subref, $langref) = @_;
|
|---|
| 403 | my ($level, $gran, $subcollection, $languages) = split (":", $index);
|
|---|
| 404 |
|
|---|
| 405 | if ($namehash->{'index'}->{$$indexref} ne "$level:$gran") {
|
|---|
| 406 | $self->get_next_version ($indexref);
|
|---|
| 407 | } elsif ($namehash->{'subcollection'}->{$$subref} ne $subcollection) {
|
|---|
| 408 | $self->get_next_version ($subref);
|
|---|
| 409 | } elsif ($namehash->{'languages'}->{$$langref} ne $languages) {
|
|---|
| 410 | $self->get_next_version ($langref);
|
|---|
| 411 | }
|
|---|
| 412 | return "$$indexref$$subref$$langref";
|
|---|
| 413 | }
|
|---|
| 414 |
|
|---|
| 415 | sub get_next_version {
|
|---|
| 416 | my $self = shift (@_);
|
|---|
| 417 | my ($nameref) = @_;
|
|---|
| 418 |
|
|---|
| 419 | if ($$nameref =~ /(\d\d)$/) {
|
|---|
| 420 | my $num = $1; $num ++;
|
|---|
| 421 | $$nameref =~ s/\d\d$/$num/;
|
|---|
| 422 | } elsif ($$nameref =~ /(\d)$/) {
|
|---|
| 423 | my $num = $1;
|
|---|
| 424 | if ($num == 9) {$$nameref =~ s/\d\d$/10/;}
|
|---|
| 425 | else {$num ++; $$nameref =~ s/\d$/$num/;}
|
|---|
| 426 | } else {
|
|---|
| 427 | $$nameref =~ s/.$/0/;
|
|---|
| 428 | }
|
|---|
| 429 | }
|
|---|
| 430 |
|
|---|
| [4] | 431 | sub build_index {
|
|---|
| 432 | my $self = shift (@_);
|
|---|
| 433 | my ($index) = @_;
|
|---|
| [1424] | 434 | my $outhandle = $self->{'outhandle'};
|
|---|
| [4] | 435 |
|
|---|
| 436 | # get the full index directory path and make sure it exists
|
|---|
| 437 | my $indexdir = $self->{'index_mapping'}->{$index};
|
|---|
| 438 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, $indexdir));
|
|---|
| 439 | my $fullindexprefix = &util::filename_cat ($self->{'build_dir'}, $indexdir,
|
|---|
| 440 | $self->{'collection'});
|
|---|
| 441 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, "text",
|
|---|
| 442 | $self->{'collection'});
|
|---|
| 443 |
|
|---|
| 444 | # get any os specific stuff
|
|---|
| 445 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 446 | my $exe = &util::get_os_exe ();
|
|---|
| [486] | 447 | my $mg_passes_exe = &util::filename_cat($exedir, "mg_passes$exe");
|
|---|
| 448 | my $mg_perf_hash_build_exe =
|
|---|
| 449 | &util::filename_cat($exedir, "mg_perf_hash_build$exe");
|
|---|
| 450 | my $mg_weights_build_exe =
|
|---|
| 451 | &util::filename_cat ($exedir, "mg_weights_build$exe");
|
|---|
| 452 | my $mg_invf_dict_exe =
|
|---|
| 453 | &util::filename_cat ($exedir, "mg_invf_dict$exe");
|
|---|
| 454 | my $mg_stem_idx_exe =
|
|---|
| 455 | &util::filename_cat ($exedir, "mg_stem_idx$exe");
|
|---|
| 456 |
|
|---|
| [4] | 457 | my $osextra = "";
|
|---|
| 458 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
|---|
| 459 | $fullindexprefix =~ s/\//\\/g;
|
|---|
| 460 | } else {
|
|---|
| 461 | $osextra = " -d /";
|
|---|
| 462 | }
|
|---|
| 463 |
|
|---|
| 464 | # get the index level from the index description
|
|---|
| 465 | # the index will be level 2 unless we are building a
|
|---|
| 466 | # paragraph level index
|
|---|
| 467 | my $index_level = 2;
|
|---|
| 468 | $index_level = 3 if $index =~ /^paragraph/i;
|
|---|
| 469 |
|
|---|
| [69] | 470 | # get the index expression if this index belongs
|
|---|
| 471 | # to a subcollection
|
|---|
| 472 | my $indexexparr = [];
|
|---|
| 473 | my ($level, $fields, $subcollection) = split (":", $index);
|
|---|
| [85] | 474 | my @subcollections = ();
|
|---|
| 475 | @subcollections = split /,/, $subcollection if (defined $subcollection);
|
|---|
| [69] | 476 |
|
|---|
| 477 | foreach $subcollection (@subcollections) {
|
|---|
| 478 | if (defined ($self->{'collect_cfg'}->{'subcollection'}->{$subcollection})) {
|
|---|
| 479 | push (@$indexexparr, $self->{'collect_cfg'}->{'subcollection'}->{$subcollection});
|
|---|
| 480 | }
|
|---|
| 481 | }
|
|---|
| 482 |
|
|---|
| [139] | 483 | # add expressions for languages if this index belongs to
|
|---|
| 484 | # a language subcollection
|
|---|
| 485 | foreach $language (@{$self->{'collect_cfg'}->{'languages'}}) {
|
|---|
| 486 | if ($language =~ s/^\!//) {
|
|---|
| 487 | push (@$indexexparr, "!Language/$language/");
|
|---|
| 488 | } else {
|
|---|
| 489 | push (@$indexexparr, "Language/$language/");
|
|---|
| 490 | }
|
|---|
| 491 | }
|
|---|
| [782] | 492 |
|
|---|
| 493 | # Build index dictionary. Uses verbatim stem method
|
|---|
| [1424] | 494 | print $outhandle "\n creating index dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [782] | 495 | my ($handle);
|
|---|
| 496 | if ($self->{'debug'}) {
|
|---|
| 497 | $handle = STDOUT;
|
|---|
| 498 | } else {
|
|---|
| 499 | if (!-e "$mg_passes_exe" ||
|
|---|
| [1679] | 500 | # !open (PIPEOUT, "| \"$mg_passes_exe\" -f \"$fullindexprefix\" -b $maxdocsize " .
|
|---|
| 501 | !open (PIPEOUT, "| mg_passes$exe -f \"$fullindexprefix\" -b $maxdocsize " .
|
|---|
| [782] | 502 | "-$index_level -m 32 -s 0 -G -t 10 -N1 $osextra")) {
|
|---|
| 503 | die "mgbuilder::build_index - couldn't run $mg_passes_exe\n";
|
|---|
| 504 | }
|
|---|
| 505 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 506 | }
|
|---|
| 507 |
|
|---|
| [4] | 508 | # set up the document processor
|
|---|
| [782] | 509 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| [4] | 510 | $self->{'buildproc'}->set_mode ('text');
|
|---|
| [69] | 511 | $self->{'buildproc'}->set_index ($index, $indexexparr);
|
|---|
| [292] | 512 | $self->{'buildproc'}->set_indexing_text (1);
|
|---|
| [4] | 513 |
|
|---|
| 514 | $self->{'buildproc'}->reset();
|
|---|
| 515 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| [315] | 516 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| [782] | 517 | close ($handle) unless $self->{'debug'};
|
|---|
| [4] | 518 |
|
|---|
| [1251] | 519 | $self->print_stats();
|
|---|
| 520 |
|
|---|
| [782] | 521 | if (!$self->{'debug'}) {
|
|---|
| 522 | # create the perfect hash function
|
|---|
| 523 | if (!-e "$mg_perf_hash_build_exe") {
|
|---|
| 524 | die "mgbuilder::build_index - couldn't run $mg_perf_hash_build_exe\n";
|
|---|
| 525 | }
|
|---|
| [1679] | 526 | # system ("\"$mg_perf_hash_build_exe\" -f \"$fullindexprefix\" $osextra");
|
|---|
| 527 | system ("mg_perf_hash_build$exe -f \"$fullindexprefix\" $osextra");
|
|---|
| [782] | 528 |
|
|---|
| 529 | if (!-e "$mg_passes_exe" ||
|
|---|
| [1679] | 530 | # !open ($handle, "| \"$mg_passes_exe\" -f \"$fullindexprefix\" -b $maxdocsize " .
|
|---|
| 531 | !open ($handle, "| mg_passes$exe -f \"$fullindexprefix\" -b $maxdocsize " .
|
|---|
| [782] | 532 | "-$index_level -c 3 -G -t 10 -N2 $osextra")) {
|
|---|
| 533 | die "mgbuilder::build_index - couldn't run $mg_passes_exe\n";
|
|---|
| 534 | }
|
|---|
| [4] | 535 | }
|
|---|
| [782] | 536 |
|
|---|
| [4] | 537 | # invert the text
|
|---|
| [1424] | 538 | print $outhandle "\n inverting the text\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [782] | 539 |
|
|---|
| [4] | 540 | $self->{'buildproc'}->reset();
|
|---|
| 541 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| [315] | 542 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| [1304] | 543 |
|
|---|
| [1251] | 544 | $self->print_stats ();
|
|---|
| 545 |
|
|---|
| [782] | 546 | if (!$self->{'debug'}) {
|
|---|
| [4] | 547 |
|
|---|
| [782] | 548 | close ($handle);
|
|---|
| 549 |
|
|---|
| 550 | # create the weights file
|
|---|
| [1424] | 551 | print $outhandle "\n create the weights file\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [782] | 552 | if (!-e "$mg_weights_build_exe") {
|
|---|
| 553 | die "mgbuilder::build_index - couldn't run $mg_weights_build_exe\n";
|
|---|
| 554 | }
|
|---|
| [1679] | 555 | # system ("\"$mg_weights_build_exe\" -f \"$fullindexprefix\" -t \"$fulltextprefix\" $osextra");
|
|---|
| 556 | system ("mg_weights_build$exe -f \"$fullindexprefix\" -t \"$fulltextprefix\" $osextra");
|
|---|
| [4] | 557 |
|
|---|
| [782] | 558 | # create 'on-disk' stemmed dictionary
|
|---|
| [1424] | 559 | print $outhandle "\n creating 'on-disk' stemmed dictionary\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [782] | 560 | if (!-e "$mg_invf_dict_exe") {
|
|---|
| 561 | die "mgbuilder::build_index - couldn't run $mg_invf_dict_exe\n";
|
|---|
| 562 | }
|
|---|
| [1679] | 563 | # system ("\"$mg_invf_dict_exe\" -f \"$fullindexprefix\" $osextra");
|
|---|
| 564 | system ("mg_invf_dict$exe -f \"$fullindexprefix\" $osextra");
|
|---|
| [4] | 565 |
|
|---|
| 566 |
|
|---|
| [782] | 567 | # creates stem index files for the various stemming methods
|
|---|
| [1424] | 568 | print $outhandle "\n creating stem indexes\n" if ($self->{'verbosity'} >= 1);
|
|---|
| [782] | 569 | if (!-e "$mg_stem_idx_exe") {
|
|---|
| 570 | die "mgbuilder::build_index - couldn't run $mg_stem_idx_exe\n";
|
|---|
| 571 | }
|
|---|
| [1679] | 572 | # system ("\"$mg_stem_idx_exe\" -b 4096 -s1 -f \"$fullindexprefix\" $osextra");
|
|---|
| 573 | system ("mg_stem_idx$exe -b 4096 -s1 -f \"$fullindexprefix\" $osextra");
|
|---|
| 574 | # system ("\"$mg_stem_idx_exe\" -b 4096 -s2 -f \"$fullindexprefix\" $osextra");
|
|---|
| 575 | system ("mg_stem_idx$exe -b 4096 -s2 -f \"$fullindexprefix\" $osextra");
|
|---|
| 576 | # system ("\"$mg_stem_idx_exe\" -b 4096 -s3 -f \"$fullindexprefix\" $osextra");
|
|---|
| 577 | system ("mg_stem_idx$exe -b 4096 -s3 -f \"$fullindexprefix\" $osextra");
|
|---|
| [4] | 578 |
|
|---|
| 579 |
|
|---|
| [782] | 580 | # remove unwanted files
|
|---|
| 581 | my $tmpdir = &util::filename_cat ($self->{'build_dir'}, $indexdir);
|
|---|
| 582 | opendir (DIR, $tmpdir) || die
|
|---|
| 583 | "mgbuilder::build_index - couldn't read directory $tmpdir\n";
|
|---|
| 584 | foreach $file (readdir(DIR)) {
|
|---|
| 585 | next if $file =~ /^\./;
|
|---|
| 586 | my ($suffix) = $file =~ /\.([^\.]+)$/;
|
|---|
| 587 | if (defined $suffix && !defined $wanted_index_files{$suffix}) {
|
|---|
| 588 | # delete it!
|
|---|
| [1424] | 589 | print $outhandle "deleting $file\n" if $self->{'verbosity'} > 2;
|
|---|
| [782] | 590 | &util::rm (&util::filename_cat ($tmpdir, $file));
|
|---|
| 591 | }
|
|---|
| [4] | 592 | }
|
|---|
| [782] | 593 | closedir (DIR);
|
|---|
| [4] | 594 | }
|
|---|
| 595 | }
|
|---|
| 596 |
|
|---|
| 597 | sub make_infodatabase {
|
|---|
| 598 | my $self = shift (@_);
|
|---|
| [1424] | 599 | my $outhandle = $self->{'outhandle'};
|
|---|
| 600 |
|
|---|
| [4] | 601 | my $textdir = &util::filename_cat($self->{'build_dir'}, "text");
|
|---|
| [810] | 602 | my $assocdir = &util::filename_cat($self->{'build_dir'}, "assoc");
|
|---|
| [4] | 603 | &util::mk_all_dir ($textdir);
|
|---|
| [810] | 604 | &util::mk_all_dir ($assocdir);
|
|---|
| [4] | 605 |
|
|---|
| [85] | 606 | # get db name
|
|---|
| 607 | my $dbext = ".bdb";
|
|---|
| 608 | $dbext = ".ldb" if &util::is_little_endian();
|
|---|
| 609 | my $fulldbname = &util::filename_cat ($textdir, "$self->{'collection'}$dbext");
|
|---|
| [4] | 610 | $fulldbname =~ s/\//\\/g if ($ENV{'GSDLOS'} =~ /^windows$/i);
|
|---|
| [85] | 611 |
|
|---|
| [4] | 612 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 613 | my $exe = &util::get_os_exe ();
|
|---|
| [486] | 614 | my $txt2db_exe = &util::filename_cat($exedir, "txt2db$exe");
|
|---|
| [4] | 615 |
|
|---|
| [1424] | 616 | print $outhandle "\n*** creating the info database and processing associated files\n"
|
|---|
| [810] | 617 | if ($self->{'verbosity'} >= 1);
|
|---|
| [4] | 618 |
|
|---|
| [215] | 619 | # init all the classifiers
|
|---|
| [315] | 620 | &classify::init_classifiers ($self->{'classifiers'});
|
|---|
| [215] | 621 |
|
|---|
| [4] | 622 | # set up the document processor
|
|---|
| [782] | 623 | my ($handle);
|
|---|
| 624 | if ($self->{'debug'}) {
|
|---|
| 625 | $handle = STDOUT;
|
|---|
| 626 | } else {
|
|---|
| [1679] | 627 | # if (!-e "$txt2db_exe" || !open (PIPEOUT, "| \"$txt2db_exe\" \"$fulldbname\"")) {
|
|---|
| 628 | if (!-e "$txt2db_exe" || !open (PIPEOUT, "| txt2db$exe \"$fulldbname\"")) {
|
|---|
| [782] | 629 | die "mgbuilder::make_infodatabase - couldn't run $txt2db_exe\n";
|
|---|
| 630 | }
|
|---|
| 631 | $handle = mgbuilder::PIPEOUT;
|
|---|
| 632 | }
|
|---|
| 633 |
|
|---|
| 634 | $self->{'buildproc'}->set_output_handle ($handle);
|
|---|
| [315] | 635 | $self->{'buildproc'}->set_mode ('infodb');
|
|---|
| [810] | 636 | $self->{'buildproc'}->set_assocdir ($assocdir);
|
|---|
| [780] | 637 | $self->{'buildproc'}->set_dontgdbm ($self->{'dontgdbm'});
|
|---|
| [315] | 638 | $self->{'buildproc'}->set_classifiers ($self->{'classifiers'});
|
|---|
| [292] | 639 | $self->{'buildproc'}->set_indexing_text (0);
|
|---|
| [4] | 640 | $self->{'buildproc'}->reset();
|
|---|
| [246] | 641 |
|
|---|
| [461] | 642 | if (defined $self->{'collect_cfg'}->{'collectionmeta'}) {
|
|---|
| [782] | 643 |
|
|---|
| [461] | 644 | if (!defined $self->{'index_mapping'}) {
|
|---|
| 645 | $self->{'index_mapping'} =
|
|---|
| 646 | $self->create_index_mapping ($self->{'collect_cfg'}->{'indexes'});
|
|---|
| 647 | }
|
|---|
| 648 |
|
|---|
| [782] | 649 | print $handle "[collection]\n";
|
|---|
| 650 |
|
|---|
| [461] | 651 | foreach $cmeta (keys (%{$self->{'collect_cfg'}->{'collectionmeta'}})) {
|
|---|
| 652 | if ($cmeta =~ s/^\.//) {
|
|---|
| 653 | if (defined $self->{'index_mapping'}->{$cmeta}) {
|
|---|
| [782] | 654 | print $handle "<$self->{'index_mapping'}->{$cmeta}>" .
|
|---|
| [461] | 655 | $self->{'collect_cfg'}->{'collectionmeta'}->{".$cmeta"} . "\n";
|
|---|
| 656 | } else {
|
|---|
| [1424] | 657 | print $outhandle "mgbuilder: warning bad collectionmeta option '$cmeta' - ignored\n";
|
|---|
| [461] | 658 | }
|
|---|
| 659 | } else {
|
|---|
| [782] | 660 | print $handle "<$cmeta>$self->{'collect_cfg'}->{'collectionmeta'}->{$cmeta}\n";
|
|---|
| [461] | 661 | }
|
|---|
| 662 | }
|
|---|
| [782] | 663 | print $handle "\n" . ('-' x 70) . "\n";
|
|---|
| [461] | 664 |
|
|---|
| 665 | }
|
|---|
| 666 |
|
|---|
| [4] | 667 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
|---|
| [315] | 668 | "", {}, $self->{'buildproc'}, $self->{'maxdocs'});
|
|---|
| [215] | 669 |
|
|---|
| 670 | # output classification information
|
|---|
| [782] | 671 | &classify::output_classify_info ($self->{'classifiers'}, $handle,
|
|---|
| [315] | 672 | $self->{'allclassifications'});
|
|---|
| [215] | 673 |
|
|---|
| [782] | 674 | close ($handle) if !$self->{'debug'};
|
|---|
| [4] | 675 | }
|
|---|
| 676 |
|
|---|
| [626] | 677 | sub collect_specific {
|
|---|
| 678 | my $self = shift (@_);
|
|---|
| 679 | }
|
|---|
| 680 |
|
|---|
| [4] | 681 | sub make_auxiliary_files {
|
|---|
| 682 | my $self = shift (@_);
|
|---|
| 683 | my ($index);
|
|---|
| 684 | my %build_cfg = ();
|
|---|
| [1424] | 685 | my $outhandle = $self->{'outhandle'};
|
|---|
| [4] | 686 |
|
|---|
| [1424] | 687 | print $outhandle "\n*** creating auxiliary files \n" if ($self->{'verbosity'} >= 1);
|
|---|
| [4] | 688 |
|
|---|
| 689 | # get the text directory
|
|---|
| 690 | &util::mk_all_dir ($self->{'build_dir'});
|
|---|
| 691 |
|
|---|
| 692 | # store the build date
|
|---|
| 693 | $build_cfg->{'builddate'} = time;
|
|---|
| 694 |
|
|---|
| 695 | # store the number of documents and number of bytes
|
|---|
| 696 | $build_cfg->{'numdocs'} = $self->{'buildproc'}->get_num_docs();
|
|---|
| 697 | $build_cfg->{'numbytes'} = $self->{'buildproc'}->get_num_bytes();
|
|---|
| 698 |
|
|---|
| [1252] | 699 | # get additional stats from mg
|
|---|
| 700 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
|---|
| 701 | my $exe = &util::get_os_exe ();
|
|---|
| 702 | my $mgstat_exe = &util::filename_cat($exedir, "mgstat$exe");
|
|---|
| 703 | my $input_file = &util::filename_cat ("text", $self->{'collection'});
|
|---|
| [1679] | 704 | # if (!-e "$mgstat_exe" || !open (PIPEIN, "\"$mgstat_exe\" -d \"$self->{'build_dir'}\" -f \"$input_file\" |")) {
|
|---|
| 705 | if (!-e "$mgstat_exe" || !open (PIPEIN, "mgstat$exe -d \"$self->{'build_dir'}\" -f \"$input_file\" |")) {
|
|---|
| [1424] | 706 | print $outhandle "Warning: Couldn't open pipe to $mgstat_exe to get additional stats\n";
|
|---|
| [1252] | 707 | } else {
|
|---|
| 708 | my $line = "";
|
|---|
| 709 | while (defined ($line = <PIPEIN>)) {
|
|---|
| 710 | if ($line =~ /^Words in collection \[dict\]\s+:\s+(\d+)/) {
|
|---|
| 711 | ($build_cfg->{'numwords'}) = $1;
|
|---|
| 712 | } elsif ($line =~ /^Documents\s+:\s+(\d+)/) {
|
|---|
| 713 | ($build_cfg->{'numsections'}) = $1;
|
|---|
| 714 | }
|
|---|
| 715 | }
|
|---|
| 716 | close PIPEIN;
|
|---|
| 717 | }
|
|---|
| 718 |
|
|---|
| [4] | 719 | # store the mapping between the index names and the directory names
|
|---|
| 720 | my @indexmap = ();
|
|---|
| [290] | 721 | foreach $index (@{$self->{'index_mapping'}->{'indexmaporder'}}) {
|
|---|
| [139] | 722 | push (@indexmap, "$index\-\>$self->{'index_mapping'}->{'indexmap'}->{$index}");
|
|---|
| [4] | 723 | }
|
|---|
| 724 | $build_cfg->{'indexmap'} = \@indexmap;
|
|---|
| 725 |
|
|---|
| [139] | 726 | my @subcollectionmap = ();
|
|---|
| [290] | 727 | foreach $subcollection (@{$self->{'index_mapping'}->{'subcollectionmaporder'}}) {
|
|---|
| [378] | 728 | push (@subcollectionmap, "$subcollection\-\>" .
|
|---|
| 729 | $self->{'index_mapping'}->{'subcollectionmap'}->{$subcollection});
|
|---|
| [139] | 730 | }
|
|---|
| 731 | $build_cfg->{'subcollectionmap'} = \@subcollectionmap if scalar (@subcollectionmap);
|
|---|
| 732 |
|
|---|
| 733 | my @languagemap = ();
|
|---|
| [290] | 734 | foreach $language (@{$self->{'index_mapping'}->{'languagemaporder'}}) {
|
|---|
| [378] | 735 | push (@languagemap, "$language\-\>" .
|
|---|
| 736 | $self->{'index_mapping'}->{'languagemap'}->{$language});
|
|---|
| [139] | 737 | }
|
|---|
| 738 | $build_cfg->{'languagemap'} = \@languagemap if scalar (@languagemap);
|
|---|
| 739 |
|
|---|
| [1246] | 740 | $build_cfg->{'notbuilt'} = $self->{'notbuilt'} if scalar @{$self->{'notbuilt'}};
|
|---|
| [486] | 741 |
|
|---|
| [4] | 742 | # write out the build information
|
|---|
| 743 | &cfgread::write_cfg_file("$self->{'build_dir'}/build.cfg", $build_cfg,
|
|---|
| [1252] | 744 | '^(builddate|numdocs|numbytes|numwords|numsections)$',
|
|---|
| [626] | 745 | '^(indexmap|subcollectionmap|languagemap|notbuilt)$');
|
|---|
| [4] | 746 |
|
|---|
| 747 | }
|
|---|
| 748 |
|
|---|
| 749 | sub deinit {
|
|---|
| 750 | my $self = shift (@_);
|
|---|
| 751 | }
|
|---|
| 752 |
|
|---|
| [1251] | 753 | sub print_stats {
|
|---|
| 754 | my $self = shift (@_);
|
|---|
| [4] | 755 |
|
|---|
| [1424] | 756 | my $outhandle = $self->{'outhandle'};
|
|---|
| [1251] | 757 | my $indexing_text = $self->{'buildproc'}->get_indexing_text();
|
|---|
| 758 | my $index = $self->{'buildproc'}->get_index();
|
|---|
| 759 | my $num_bytes = $self->{'buildproc'}->get_num_bytes();
|
|---|
| 760 | my $num_processed_bytes = $self->{'buildproc'}->get_num_processed_bytes();
|
|---|
| 761 |
|
|---|
| 762 | if ($indexing_text) {
|
|---|
| [1424] | 763 | print $outhandle "Stats (Creating index $index)\n";
|
|---|
| [1251] | 764 | } else {
|
|---|
| [1424] | 765 | print $outhandle "Stats (Compressing text from $index)\n";
|
|---|
| [1251] | 766 | }
|
|---|
| [1424] | 767 | print $outhandle "Total bytes in collection: $num_bytes\n";
|
|---|
| 768 | print $outhandle "Total bytes in $index: $num_processed_bytes\n";
|
|---|
| [1251] | 769 |
|
|---|
| 770 | if ($num_processed_bytes < 50) {
|
|---|
| [1424] | 771 | print $outhandle "***************\n";
|
|---|
| 772 | print $outhandle "WARNING: There is very little or no text to process for $index\n";
|
|---|
| [1251] | 773 | if ($indexing_text) {
|
|---|
| [1424] | 774 | print $outhandle "This may cause an error while attempting to build the index\n";
|
|---|
| [1251] | 775 | } else {
|
|---|
| [1424] | 776 | print $outhandle "This may cause an error while attempting to compress the text\n";
|
|---|
| [1251] | 777 | }
|
|---|
| [1424] | 778 | print $outhandle "***************\n";
|
|---|
| [1251] | 779 | }
|
|---|
| 780 | }
|
|---|
| 781 |
|
|---|
| [4] | 782 | 1;
|
|---|
| 783 |
|
|---|
| 784 |
|
|---|
