[14930] | 1 | ###########################################################################
|
---|
| 2 | #
|
---|
| 3 | # basebuilder.pm -- base class for collection builders
|
---|
| 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 basebuilder;
|
---|
| 27 |
|
---|
| 28 | use strict;
|
---|
| 29 | no strict 'refs'; # allow filehandles to be variables and viceversa
|
---|
| 30 |
|
---|
| 31 | use classify;
|
---|
| 32 | use cfgread;
|
---|
| 33 | use colcfg;
|
---|
[15709] | 34 | use dbutil;
|
---|
[14930] | 35 | use plugin;
|
---|
| 36 | use util;
|
---|
| 37 |
|
---|
[15709] | 38 |
|
---|
[14930] | 39 | BEGIN {
|
---|
| 40 | # set autoflush on for STDERR and STDOUT so that mgpp
|
---|
| 41 | # doesn't get out of sync with plugins
|
---|
| 42 | STDOUT->autoflush(1);
|
---|
| 43 | STDERR->autoflush(1);
|
---|
| 44 | }
|
---|
| 45 |
|
---|
| 46 | END {
|
---|
| 47 | STDOUT->autoflush(0);
|
---|
| 48 | STDERR->autoflush(0);
|
---|
| 49 | }
|
---|
| 50 |
|
---|
| 51 | our $maxdocsize = 12000;
|
---|
| 52 |
|
---|
| 53 | # used to signify "gs2"(default) or "gs3"
|
---|
[20095] | 54 | our $gs_mode = "gs2";
|
---|
[14930] | 55 |
|
---|
| 56 | sub new {
|
---|
| 57 | my ($class, $collection, $source_dir, $build_dir, $verbosity,
|
---|
[20647] | 58 | $maxdocs, $debug, $keepold, $incremental, $incremental_mode,
|
---|
[14930] | 59 | $remove_empty_classifications,
|
---|
[21785] | 60 | $outhandle, $no_text, $failhandle, $gli) = @_;
|
---|
[14930] | 61 |
|
---|
| 62 | $outhandle = *STDERR unless defined $outhandle;
|
---|
| 63 | $no_text = 0 unless defined $no_text;
|
---|
| 64 | $failhandle = *STDERR unless defined $failhandle;
|
---|
| 65 |
|
---|
| 66 | # create a builder object
|
---|
| 67 | my $self = bless {'collection'=>$collection,
|
---|
| 68 | 'source_dir'=>$source_dir,
|
---|
| 69 | 'build_dir'=>$build_dir,
|
---|
| 70 | 'verbosity'=>$verbosity,
|
---|
| 71 | 'maxdocs'=>$maxdocs,
|
---|
| 72 | 'debug'=>$debug,
|
---|
| 73 | 'keepold'=>$keepold,
|
---|
| 74 | 'incremental'=>$incremental,
|
---|
[20647] | 75 | 'incremental_mode'=>$incremental_mode,
|
---|
[14930] | 76 | 'remove_empty_classifications'=>$remove_empty_classifications,
|
---|
| 77 | 'outhandle'=>$outhandle,
|
---|
| 78 | 'no_text'=>$no_text,
|
---|
| 79 | 'failhandle'=>$failhandle,
|
---|
| 80 | 'notbuilt'=>{}, # indexes not built
|
---|
[21785] | 81 | 'gli'=>$gli
|
---|
[14930] | 82 | }, $class;
|
---|
| 83 |
|
---|
| 84 | $self->{'gli'} = 0 unless defined $self->{'gli'};
|
---|
| 85 |
|
---|
| 86 | # Read in the collection configuration file.
|
---|
| 87 | my ($colcfgname);
|
---|
| 88 | ($colcfgname, $gs_mode) = &colcfg::get_collect_cfg_name($outhandle);
|
---|
[20100] | 89 | $self->{'collect_cfg'} = &colcfg::read_collection_cfg ($colcfgname, $gs_mode);
|
---|
[14384] | 90 |
|
---|
[20100] | 91 | if ($gs_mode eq "gs3") {
|
---|
| 92 | # read it in again to save the original form for later writing out
|
---|
| 93 | # of buildConfig.xml
|
---|
| 94 | # we use this preserve object because $self->{'collect_cfg'}->{'classify'} somewhat gets modified during the calling of &classify::load_classifiers.
|
---|
| 95 | $self->{'collect_cfg_preserve'} = &colcfg::read_collection_cfg ($colcfgname, $gs_mode);
|
---|
[14930] | 96 | }
|
---|
[20100] | 97 |
|
---|
[15725] | 98 | # get the database type for this collection from the collect.cfg file (may be undefined)
|
---|
[15727] | 99 | $self->{'infodbtype'} = $self->{'collect_cfg'}->{'infodbtype'} || &dbutil::get_default_infodb_type();
|
---|
[15725] | 100 |
|
---|
[14384] | 101 |
|
---|
[15688] | 102 | # load up any dontdb fields
|
---|
| 103 | $self->{'dontdb'} = {};
|
---|
[14930] | 104 | if (defined ($self->{'collect_cfg'}->{'dontgdbm'})) {
|
---|
| 105 | foreach my $dg (@{$self->{'collect_cfg'}->{'dontgdbm'}}) {
|
---|
[15688] | 106 | $self->{'dontdb'}->{$dg} = 1;
|
---|
[14930] | 107 | }
|
---|
| 108 | }
|
---|
| 109 |
|
---|
| 110 | $self->{'maxnumeric'} = 4;
|
---|
| 111 | return $self;
|
---|
| 112 | }
|
---|
| 113 |
|
---|
| 114 | # stuff has been moved here from new, so we can use subclass methods
|
---|
| 115 | sub init {
|
---|
| 116 | my $self = shift(@_);
|
---|
| 117 |
|
---|
[20647] | 118 | my $outhandle = $self->{'outhandle'};
|
---|
| 119 | my $failhandle = $self->{'failhandle'};
|
---|
| 120 |
|
---|
[14930] | 121 | $self->generate_index_list();
|
---|
[19218] | 122 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
| 123 | if (defined $indexes) {
|
---|
| 124 | # sort out subcollection indexes
|
---|
| 125 | if (defined $self->{'collect_cfg'}->{'indexsubcollections'}) {
|
---|
| 126 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
| 127 | foreach my $subcollection (@{$self->{'collect_cfg'}->{'indexsubcollections'}}) {
|
---|
| 128 | foreach my $index (@$indexes) {
|
---|
| 129 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$subcollection");
|
---|
| 130 | }
|
---|
[14930] | 131 | }
|
---|
| 132 | }
|
---|
[19218] | 133 |
|
---|
| 134 | # sort out language subindexes
|
---|
| 135 | if (defined $self->{'collect_cfg'}->{'languages'}) {
|
---|
| 136 | $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
| 137 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
| 138 | foreach my $language (@{$self->{'collect_cfg'}->{'languages'}}) {
|
---|
| 139 | foreach my $index (@$indexes) {
|
---|
| 140 | if (defined ($self->{'collect_cfg'}->{'indexsubcollections'})) {
|
---|
| 141 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$language");
|
---|
| 142 | }
|
---|
| 143 | else { # add in an empty subcollection field
|
---|
| 144 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index\:\:$language");
|
---|
| 145 | }
|
---|
[14930] | 146 | }
|
---|
| 147 | }
|
---|
| 148 | }
|
---|
| 149 | }
|
---|
[19218] | 150 |
|
---|
[14930] | 151 | if (defined($self->{'collect_cfg'}->{'indexes'})) {
|
---|
| 152 | # make sure that the same index isn't specified more than once
|
---|
| 153 | my %tmphash = ();
|
---|
| 154 | my @tmparray = @{$self->{'collect_cfg'}->{'indexes'}};
|
---|
| 155 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
| 156 | foreach my $i (@tmparray) {
|
---|
| 157 | if (!defined ($tmphash{$i})) {
|
---|
| 158 | push (@{$self->{'collect_cfg'}->{'indexes'}}, $i);
|
---|
| 159 | $tmphash{$i} = 1;
|
---|
| 160 | }
|
---|
| 161 | }
|
---|
| 162 | } else {
|
---|
| 163 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
| 164 | }
|
---|
| 165 |
|
---|
[20647] | 166 | # check incremental against whether builder can cope or not.
|
---|
[20681] | 167 | if ($self->{'incremental'} && !$self->is_incremental_capable()) {
|
---|
[20647] | 168 | print $outhandle "WARNING: The indexer used is not capable of incremental building. Reverting to -removeold\n";
|
---|
| 169 | $self->{'keepold'} = 0;
|
---|
| 170 | $self->{'incremental'} = 0;
|
---|
| 171 | $self->{'incremental_mode'} = "none";
|
---|
| 172 |
|
---|
| 173 | }
|
---|
| 174 |
|
---|
| 175 |
|
---|
| 176 | # get the list of plugins for this collection
|
---|
| 177 | my $plugins = [];
|
---|
| 178 | if (defined $self->{'collect_cfg'}->{'plugin'}) {
|
---|
| 179 | $plugins = $self->{'collect_cfg'}->{'plugin'};
|
---|
| 180 | }
|
---|
| 181 |
|
---|
| 182 | # load all the plugins
|
---|
| 183 |
|
---|
| 184 | #build up the extra global options for the plugins
|
---|
| 185 | my @global_opts = ();
|
---|
| 186 | if (defined $self->{'collect_cfg'}->{'separate_cjk'} && $self->{'collect_cfg'}->{'separate_cjk'} =~ /^true$/i) {
|
---|
| 187 | push @global_opts, "-separate_cjk";
|
---|
| 188 | }
|
---|
| 189 | $self->{'pluginfo'} = &plugin::load_plugins ($plugins, $self->{'verbosity'}, $outhandle, $failhandle, \@global_opts, $self->{'incremental_mode'});
|
---|
| 190 |
|
---|
| 191 | if (scalar(@{$self->{'pluginfo'}}) == 0) {
|
---|
| 192 | print $outhandle "No plugins were loaded.\n";
|
---|
| 193 | die "\n";
|
---|
| 194 | }
|
---|
| 195 |
|
---|
| 196 | # get the list of classifiers for this collection
|
---|
| 197 | my $classifiers = [];
|
---|
| 198 | if (defined $self->{'collect_cfg'}->{'classify'}) {
|
---|
| 199 | $classifiers = $self->{'collect_cfg'}->{'classify'};
|
---|
| 200 | }
|
---|
| 201 |
|
---|
| 202 | # load all the classifiers
|
---|
| 203 | $self->{'classifiers'} = &classify::load_classifiers ($classifiers, $self->{'build_dir'}, $outhandle);
|
---|
| 204 |
|
---|
[14930] | 205 | # load up the document processor for building
|
---|
| 206 | # if a buildproc class has been created for this collection, use it
|
---|
[20647] | 207 | # otherwise, use the default buildproc for the builder we are initialising
|
---|
[24342] | 208 | my $buildprocdir = undef;
|
---|
| 209 | my $buildproctype;
|
---|
| 210 |
|
---|
[14930] | 211 | my $collection = $self->{'collection'};
|
---|
| 212 | if (-e "$ENV{'GSDLCOLLECTDIR'}/custom/${collection}/perllib/custombuildproc.pm") {
|
---|
| 213 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/custom/${collection}/perllib";
|
---|
| 214 | $buildproctype = "custombuildproc";
|
---|
| 215 | } elsif (-e "$ENV{'GSDLCOLLECTDIR'}/perllib/custombuildproc.pm") {
|
---|
| 216 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/perllib";
|
---|
| 217 | $buildproctype = "custombuildproc";
|
---|
| 218 | } elsif (-e "$ENV{'GSDLCOLLECTDIR'}/perllib/${collection}buildproc.pm") {
|
---|
| 219 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/perllib";
|
---|
| 220 | $buildproctype = "${collection}buildproc";
|
---|
| 221 | } else {
|
---|
| 222 | $buildproctype = $self->default_buildproc();
|
---|
| 223 | }
|
---|
[24342] | 224 | if (defined $buildprocdir) {
|
---|
| 225 | require "$buildprocdir/$buildproctype.pm";
|
---|
| 226 | }
|
---|
| 227 | else {
|
---|
| 228 | require "$buildproctype.pm";
|
---|
| 229 | }
|
---|
[14930] | 230 |
|
---|
| 231 | eval("\$self->{'buildproc'} = new $buildproctype(\$self->{'collection'}, " .
|
---|
| 232 | "\$self->{'source_dir'}, \$self->{'build_dir'}, \$self->{'keepold'}, \$self->{'verbosity'}, \$self->{'outhandle'})");
|
---|
| 233 | die "$@" if $@;
|
---|
| 234 |
|
---|
[21607] | 235 | # We call set_infodbtype() now so the buildproc knows the infodbtype for all phases of the build
|
---|
| 236 | $self->{'buildproc'}->set_infodbtype($self->{'infodbtype'});
|
---|
[17110] | 237 |
|
---|
| 238 | $self->generate_index_options();
|
---|
| 239 |
|
---|
[14930] | 240 | if (!$self->{'debug'} && !$self->{'keepold'}) {
|
---|
| 241 | # remove any old builds
|
---|
| 242 | &util::rm_r($self->{'build_dir'});
|
---|
| 243 | &util::mk_all_dir($self->{'build_dir'});
|
---|
| 244 |
|
---|
| 245 | # make the text directory
|
---|
| 246 | my $textdir = "$self->{'build_dir'}/text";
|
---|
| 247 | &util::mk_all_dir($textdir);
|
---|
| 248 | }
|
---|
[17573] | 249 |
|
---|
| 250 | if ($self->{'incremental'}) {
|
---|
| 251 | # some classes may need to do some additional initialisation
|
---|
| 252 | $self->init_for_incremental_build();
|
---|
| 253 | }
|
---|
[14930] | 254 |
|
---|
| 255 | }
|
---|
| 256 |
|
---|
[20647] | 257 | sub is_incremental_capable
|
---|
| 258 | {
|
---|
| 259 | # By default we return 'no' as the answer
|
---|
| 260 | # Safer to assume non-incremental to start with, and then override in
|
---|
| 261 | # inherited classes that are.
|
---|
| 262 |
|
---|
| 263 | return 0;
|
---|
| 264 | }
|
---|
| 265 |
|
---|
[17573] | 266 | # implement this in subclass if want to do additional initialisation for an
|
---|
| 267 | # incremental build
|
---|
| 268 | sub init_for_incremental_build {
|
---|
| 269 | my $self = shift (@_);
|
---|
| 270 | }
|
---|
| 271 |
|
---|
[14930] | 272 | sub deinit {
|
---|
| 273 | my $self = shift (@_);
|
---|
| 274 |
|
---|
| 275 | &plugin::deinit($self->{'pluginfo'},$self->{'buildproc'});
|
---|
| 276 | }
|
---|
| 277 |
|
---|
[17110] | 278 | sub generate_index_options {
|
---|
| 279 | my $self = shift (@_);
|
---|
| 280 |
|
---|
| 281 | my $separate_cjk = 0;
|
---|
| 282 |
|
---|
| 283 | if (defined($self->{'collect_cfg'}->{'indexoptions'})) {
|
---|
| 284 | foreach my $option (@{$self->{'collect_cfg'}->{'indexoptions'}}) {
|
---|
| 285 | if ($option =~ /separate_cjk/) {
|
---|
| 286 | $separate_cjk = 1;
|
---|
| 287 | }
|
---|
| 288 | }
|
---|
| 289 | }
|
---|
| 290 | # set this for building
|
---|
| 291 | $self->{'buildproc'}->set_separate_cjk($separate_cjk);
|
---|
| 292 | # record it for build.cfg
|
---|
| 293 | $self->{'separate_cjk'} = $separate_cjk;
|
---|
| 294 | }
|
---|
| 295 |
|
---|
[14930] | 296 | sub set_sections_index_document_metadata {
|
---|
| 297 | my $self = shift (@_);
|
---|
| 298 | my ($index) = @_;
|
---|
| 299 |
|
---|
| 300 | $self->{'buildproc'}->set_sections_index_document_metadata($index);
|
---|
| 301 | }
|
---|
| 302 |
|
---|
| 303 | sub set_maxnumeric {
|
---|
| 304 | my $self = shift (@_);
|
---|
| 305 | my ($maxnumeric) = @_;
|
---|
| 306 |
|
---|
| 307 | $self->{'maxnumeric'} = $maxnumeric;
|
---|
| 308 | }
|
---|
| 309 | sub set_strip_html {
|
---|
| 310 | my $self = shift (@_);
|
---|
| 311 | my ($strip) = @_;
|
---|
| 312 |
|
---|
| 313 | $self->{'strip_html'} = $strip;
|
---|
| 314 | $self->{'buildproc'}->set_strip_html($strip);
|
---|
| 315 | }
|
---|
| 316 |
|
---|
| 317 | sub compress_text {
|
---|
| 318 | my $self = shift (@_);
|
---|
| 319 | my ($textindex) = @_;
|
---|
| 320 |
|
---|
| 321 | print STDERR "compress_text() should be implemented in subclass!!";
|
---|
| 322 | return;
|
---|
| 323 | }
|
---|
| 324 |
|
---|
| 325 |
|
---|
| 326 | sub build_indexes {
|
---|
| 327 | my $self = shift (@_);
|
---|
| 328 | my ($indexname) = @_;
|
---|
| 329 | my $outhandle = $self->{'outhandle'};
|
---|
| 330 |
|
---|
| 331 | my $indexes = [];
|
---|
| 332 | if (defined $indexname && $indexname =~ /\w/) {
|
---|
| 333 | push @$indexes, $indexname;
|
---|
| 334 | } else {
|
---|
| 335 | $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
| 336 | }
|
---|
| 337 |
|
---|
| 338 | # create the mapping between the index descriptions
|
---|
| 339 | # and their directory names (includes subcolls and langs)
|
---|
| 340 | $self->{'index_mapping'} = $self->create_index_mapping ($indexes);
|
---|
| 341 |
|
---|
| 342 | # build each of the indexes
|
---|
| 343 | foreach my $index (@$indexes) {
|
---|
| 344 | if ($self->want_built($index)) {
|
---|
| 345 | print $outhandle "\n*** building index $index in subdirectory " .
|
---|
| 346 | "$self->{'index_mapping'}->{$index}\n" if ($self->{'verbosity'} >= 1);
|
---|
| 347 | print STDERR "<Stage name='Index' source='$index'>\n" if $self->{'gli'};
|
---|
| 348 | $self->build_index($index);
|
---|
| 349 | } else {
|
---|
| 350 | print $outhandle "\n*** ignoring index $index\n" if ($self->{'verbosity'} >= 1);
|
---|
| 351 | }
|
---|
| 352 | }
|
---|
| 353 |
|
---|
| 354 | $self->build_indexes_extra();
|
---|
| 355 |
|
---|
| 356 | }
|
---|
| 357 |
|
---|
[17573] | 358 | # implement this in subclass if want to do extra stuff at the end of building
|
---|
| 359 | # all the indexes
|
---|
[14930] | 360 | sub build_indexes_extra {
|
---|
| 361 | my $self = shift(@_);
|
---|
| 362 |
|
---|
| 363 | }
|
---|
| 364 |
|
---|
| 365 | sub build_index {
|
---|
| 366 | my $self = shift (@_);
|
---|
| 367 | my ($index) = @_;
|
---|
| 368 |
|
---|
| 369 | print STDERR "build_index should be implemented in subclass\n";
|
---|
| 370 | return;
|
---|
| 371 | }
|
---|
| 372 |
|
---|
| 373 |
|
---|
| 374 |
|
---|
| 375 | sub make_infodatabase {
|
---|
| 376 | my $self = shift (@_);
|
---|
| 377 | my $outhandle = $self->{'outhandle'};
|
---|
| 378 |
|
---|
| 379 | print STDERR "BuildDir: $self->{'build_dir'}\n";
|
---|
| 380 |
|
---|
| 381 | my $textdir = &util::filename_cat($self->{'build_dir'}, "text");
|
---|
| 382 | my $assocdir = &util::filename_cat($self->{'build_dir'}, "assoc");
|
---|
| 383 | &util::mk_all_dir ($textdir);
|
---|
| 384 | &util::mk_all_dir ($assocdir);
|
---|
| 385 |
|
---|
[15710] | 386 | # Get info database file path
|
---|
[20575] | 387 | my $infodb_type = $self->{'infodbtype'};
|
---|
| 388 | my $infodb_file_path = &dbutil::get_infodb_file_path($infodb_type, $self->{'collection'}, $textdir);
|
---|
[14930] | 389 |
|
---|
| 390 | print $outhandle "\n*** creating the info database and processing associated files\n"
|
---|
| 391 | if ($self->{'verbosity'} >= 1);
|
---|
| 392 | print STDERR "<Stage name='CreateInfoData'>\n" if $self->{'gli'};
|
---|
| 393 |
|
---|
| 394 | # init all the classifiers
|
---|
| 395 | &classify::init_classifiers ($self->{'classifiers'});
|
---|
| 396 |
|
---|
| 397 | my $reconstructed_docs = undef;
|
---|
[20575] | 398 | my $database_recs = undef;
|
---|
| 399 |
|
---|
[20686] | 400 | if ($self->{'incremental'}) {
|
---|
[20575] | 401 | $database_recs = {};
|
---|
| 402 |
|
---|
| 403 | &dbutil::read_infodb_file($infodb_type, $infodb_file_path, $database_recs);
|
---|
[14930] | 404 | }
|
---|
[20575] | 405 |
|
---|
[14930] | 406 |
|
---|
[20575] | 407 | # Important (for memory usage reasons) that we obtain the filehandle
|
---|
| 408 | # here for writing out to the database, rather than after
|
---|
[20686] | 409 | # $reconstructed_docs has been set up (assuming -incremental is on)
|
---|
[20575] | 410 | #
|
---|
| 411 | # This is because when we open a pipe to txt2db [using open()]
|
---|
| 412 | # this triggers a fork() followed by exec(). $reconstructed_docs
|
---|
| 413 | # can get very large, and so if we did the open() after this, it means
|
---|
| 414 | # the fork creates a clone of the *large* process image which (admittedly)
|
---|
| 415 | # is then quickly replaced in the execve() with the much smaller image for
|
---|
[20647] | 416 | # 'txt2db'. The trouble is, in that seismic second caused by
|
---|
[20575] | 417 | # the fork(), the system really does need to have all that memory available
|
---|
| 418 | # even though it isn't ultimately used. The result is an out of memory
|
---|
| 419 | # error.
|
---|
| 420 |
|
---|
[15700] | 421 | my ($infodb_handle);
|
---|
[14930] | 422 | if ($self->{'debug'}) {
|
---|
[15700] | 423 | $infodb_handle = *STDOUT;
|
---|
[15710] | 424 | }
|
---|
| 425 | else {
|
---|
[20575] | 426 | $infodb_handle = &dbutil::open_infodb_write_handle($infodb_type, $infodb_file_path);
|
---|
[15711] | 427 | if (!defined($infodb_handle))
|
---|
| 428 | {
|
---|
[14930] | 429 | print STDERR "<FatalError name='NoRunText2DB'/>\n</Stage>\n" if $self->{'gli'};
|
---|
[15711] | 430 | die "builder::make_infodatabase - couldn't open infodb write handle\n";
|
---|
[14930] | 431 | }
|
---|
| 432 | }
|
---|
[15725] | 433 |
|
---|
[20686] | 434 | if ($self->{'incremental'}) {
|
---|
[20575] | 435 | # reconstruct doc_obj metadata from database for all docs
|
---|
| 436 | $reconstructed_docs
|
---|
| 437 | = &classify::reconstruct_doc_objs_metadata($infodb_type,
|
---|
| 438 | $infodb_file_path,
|
---|
| 439 | $database_recs);
|
---|
| 440 | }
|
---|
| 441 |
|
---|
| 442 | # set up the document processor
|
---|
| 443 |
|
---|
[15700] | 444 | $self->{'buildproc'}->set_output_handle ($infodb_handle);
|
---|
[14930] | 445 | $self->{'buildproc'}->set_mode ('infodb');
|
---|
| 446 | $self->{'buildproc'}->set_assocdir ($assocdir);
|
---|
[15688] | 447 | $self->{'buildproc'}->set_dontdb ($self->{'dontdb'});
|
---|
[14930] | 448 | $self->{'buildproc'}->set_classifiers ($self->{'classifiers'});
|
---|
| 449 | $self->{'buildproc'}->set_indexing_text (0);
|
---|
| 450 | $self->{'buildproc'}->set_store_text(1);
|
---|
[16222] | 451 | $self->{'buildproc'}->set_store_metadata_coverage ($self->{'collect_cfg'}->{'store_metadata_coverage'});
|
---|
[14930] | 452 |
|
---|
| 453 | # make_infodatabase needs full reset even for incremental build
|
---|
[15688] | 454 | # as incremental works by reconstructing all docs from the database and
|
---|
[14930] | 455 | # then adding in the new ones
|
---|
| 456 | $self->{'buildproc'}->zero_reset();
|
---|
| 457 |
|
---|
[14934] | 458 | $self->{'buildproc'}->{'mdprefix_fields'} = {};
|
---|
[23120] | 459 |
|
---|
| 460 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
---|
| 461 | "", {}, {}, $self->{'buildproc'}, $self->{'maxdocs'},0, $self->{'gli'});
|
---|
[14934] | 462 |
|
---|
[20686] | 463 | if ($self->{'incremental'}) {
|
---|
[14930] | 464 | # create flat classify structure, ready for new docs to be added
|
---|
[23120] | 465 | foreach my $doc_obj ( @$reconstructed_docs ) {
|
---|
[23159] | 466 | if (! defined $self->{'buildproc'}->{'dont_process_reconstructed'}->{$doc_obj->get_OID()}) {
|
---|
[23120] | 467 | print $outhandle " Adding reconstructed ", $doc_obj->get_OID(), " into classify structures\n";
|
---|
| 468 | $self->{'buildproc'}->process($doc_obj,undef);
|
---|
| 469 | }
|
---|
[14930] | 470 | }
|
---|
| 471 | }
|
---|
| 472 | # this has changed to only output collection meta if its
|
---|
| 473 | # not in the config file
|
---|
[15700] | 474 | $self->output_collection_meta($infodb_handle);
|
---|
[14930] | 475 |
|
---|
| 476 | # output classification information
|
---|
[20575] | 477 | &classify::output_classify_info ($self->{'classifiers'}, $infodb_type, $infodb_handle,
|
---|
[14930] | 478 | $self->{'remove_empty_classifications'},
|
---|
| 479 | $self->{'gli'});
|
---|
| 480 |
|
---|
| 481 | # Output classifier reverse lookup, used in incremental deletion
|
---|
[18469] | 482 | ####&classify::print_reverse_lookup($infodb_handle);
|
---|
[14930] | 483 |
|
---|
[15700] | 484 | # output doclist
|
---|
| 485 | my @doc_list = $self->{'buildproc'}->get_doc_list();
|
---|
[15725] | 486 | my $browselist_infodb = { 'hastxt' => [ "0" ],
|
---|
| 487 | 'childtype' => [ "VList" ],
|
---|
| 488 | 'numleafdocs' => [ scalar(@doc_list) ],
|
---|
| 489 | 'thistype' => [ "Invisible" ],
|
---|
| 490 | 'contains' => [ join(";", @doc_list) ] };
|
---|
[20575] | 491 | &dbutil::write_infodb_entry($infodb_type, $infodb_handle, "browselist", $browselist_infodb);
|
---|
[14930] | 492 |
|
---|
[20575] | 493 | &dbutil::close_infodb_write_handle($infodb_type, $infodb_handle) if !$self->{'debug'};
|
---|
[23172] | 494 |
|
---|
| 495 | if ($infodb_type eq "gdbm-txtgz") {
|
---|
| 496 | my $gdb_infodb_file_path = &dbutil::get_infodb_file_path("gdbm", $self->{'collection'}, $textdir);
|
---|
| 497 | if (-e $gdb_infodb_file_path) {
|
---|
| 498 | &util::rm($gdb_infodb_file_path);
|
---|
| 499 | }
|
---|
| 500 | }
|
---|
[14930] | 501 | print STDERR "</Stage>\n" if $self->{'gli'};
|
---|
| 502 | }
|
---|
| 503 |
|
---|
| 504 | sub make_auxiliary_files {
|
---|
| 505 | my $self = shift (@_);
|
---|
| 506 | my ($index);
|
---|
| 507 | my $build_cfg = {};
|
---|
| 508 | # subclasses may have already defined stuff in here
|
---|
| 509 | if (defined $self->{'build_cfg'}) {
|
---|
| 510 | $build_cfg = $self->{'build_cfg'};
|
---|
| 511 | }
|
---|
| 512 |
|
---|
| 513 | my $outhandle = $self->{'outhandle'};
|
---|
| 514 |
|
---|
| 515 | print $outhandle "\n*** creating auxiliary files \n" if ($self->{'verbosity'} >= 1);
|
---|
| 516 | print STDERR "<Stage name='CreatingAuxilary'>\n" if $self->{'gli'};
|
---|
| 517 |
|
---|
| 518 | # get the text directory
|
---|
| 519 | &util::mk_all_dir ($self->{'build_dir'});
|
---|
| 520 |
|
---|
| 521 | # store the build date
|
---|
| 522 | $build_cfg->{'builddate'} = time;
|
---|
| 523 | $build_cfg->{'buildtype'} = $self->{'buildtype'};
|
---|
[15003] | 524 | $build_cfg->{'indexstem'} = &util::get_dirsep_tail($self->{'collection'});
|
---|
[14930] | 525 | $build_cfg->{'stemindexes'} = $self->{'stemindexes'};
|
---|
[17110] | 526 | if ($self->{'separate_cjk'}) {
|
---|
| 527 | $build_cfg->{'separate_cjk'} = "true";
|
---|
| 528 | }
|
---|
[14930] | 529 |
|
---|
| 530 | # store the number of documents and number of bytes
|
---|
| 531 | $build_cfg->{'numdocs'} = $self->{'buildproc'}->get_num_docs();
|
---|
| 532 | $build_cfg->{'numsections'} = $self->{'buildproc'}->get_num_sections();
|
---|
| 533 | $build_cfg->{'numbytes'} = $self->{'buildproc'}->get_num_bytes();
|
---|
[18441] | 534 |
|
---|
[14930] | 535 | # store the mapping between the index names and the directory names
|
---|
| 536 | # the index map is used to determine what indexes there are, so any that are not built should not be put into the map.
|
---|
| 537 | my @indexmap = ();
|
---|
| 538 | foreach my $index (@{$self->{'index_mapping'}->{'indexmaporder'}}) {
|
---|
| 539 | if (not defined ($self->{'notbuilt'}->{$index})) {
|
---|
| 540 | push (@indexmap, "$index\-\>$self->{'index_mapping'}->{'indexmap'}->{$index}");
|
---|
| 541 | }
|
---|
| 542 | }
|
---|
| 543 | $build_cfg->{'indexmap'} = \@indexmap if scalar (@indexmap);
|
---|
| 544 |
|
---|
| 545 | my @subcollectionmap = ();
|
---|
| 546 | foreach my $subcollection (@{$self->{'index_mapping'}->{'subcollectionmaporder'}}) {
|
---|
| 547 | push (@subcollectionmap, "$subcollection\-\>" .
|
---|
| 548 | $self->{'index_mapping'}->{'subcollectionmap'}->{$subcollection});
|
---|
| 549 | }
|
---|
| 550 | $build_cfg->{'subcollectionmap'} = \@subcollectionmap if scalar (@subcollectionmap);
|
---|
| 551 |
|
---|
| 552 | my @languagemap = ();
|
---|
| 553 | foreach my $language (@{$self->{'index_mapping'}->{'languagemaporder'}}) {
|
---|
| 554 | push (@languagemap, "$language\-\>" .
|
---|
| 555 | $self->{'index_mapping'}->{'languagemap'}->{$language});
|
---|
| 556 | }
|
---|
| 557 | $build_cfg->{'languagemap'} = \@languagemap if scalar (@languagemap);
|
---|
| 558 |
|
---|
| 559 | my @notbuilt = ();
|
---|
| 560 | foreach my $nb (keys %{$self->{'notbuilt'}}) {
|
---|
| 561 | push (@notbuilt, $nb);
|
---|
| 562 | }
|
---|
| 563 | $build_cfg->{'notbuilt'} = \@notbuilt if scalar (@notbuilt);
|
---|
| 564 |
|
---|
| 565 | $build_cfg->{'maxnumeric'} = $self->{'maxnumeric'};
|
---|
| 566 |
|
---|
[15728] | 567 | $build_cfg->{'infodbtype'} = $self->{'infodbtype'};
|
---|
[23939] | 568 |
|
---|
| 569 | # write out the earliestDatestamp information needed for OAI
|
---|
| 570 | my $archivedir = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, "archives");
|
---|
| 571 | if(!-d $archivedir) {
|
---|
| 572 | $archivedir = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, "export");
|
---|
[23946] | 573 | }
|
---|
| 574 | my $earliestDatestampFile = &util::filename_cat ($archivedir, "earliestDatestamp");
|
---|
| 575 | my $earliestDatestamp = 0;
|
---|
| 576 | if (open(FIN,"<$earliestDatestampFile")) {
|
---|
[23939] | 577 | {
|
---|
| 578 | # slurp in file as a single line
|
---|
| 579 | local $/ = undef;
|
---|
| 580 | $earliestDatestamp = <FIN>;
|
---|
| 581 | #&unicode::ensure_utf8(\$earliestDatestamp); # turn any high bytes that aren't valid utf-8 into utf-8.
|
---|
| 582 | }
|
---|
[23946] | 583 | close(FIN);
|
---|
[23939] | 584 | }
|
---|
[23946] | 585 | else {
|
---|
| 586 | print $outhandle "Warning: unable to read collection's earliestDatestamp from $earliestDatestampFile.\n";
|
---|
| 587 | print $outhandle "Setting value to 0.\n";
|
---|
| 588 | }
|
---|
[24070] | 589 | $build_cfg->{'earliestdatestamp'} = $earliestDatestamp;
|
---|
[23946] | 590 |
|
---|
[14930] | 591 | $self->build_cfg_extra($build_cfg);
|
---|
| 592 |
|
---|
| 593 | if ($gs_mode eq "gs2") {
|
---|
[20100] | 594 | &colcfg::write_build_cfg(&util::filename_cat($self->{'build_dir'},"build.cfg"), $build_cfg);
|
---|
[14930] | 595 | }
|
---|
[14384] | 596 | if ($gs_mode eq "gs3") {
|
---|
[14930] | 597 |
|
---|
[21785] | 598 | &colcfg::write_build_cfg_xml(&util::filename_cat($self->{'build_dir'}, "buildConfig.xml"), $build_cfg, $self->{'collect_cfg_preserve'});
|
---|
[14930] | 599 | }
|
---|
| 600 |
|
---|
| 601 | print STDERR "</Stage>\n" if $self->{'gli'};
|
---|
| 602 | }
|
---|
| 603 |
|
---|
[17573] | 604 | # implement this in subclass if want to add extra stuff to build.cfg
|
---|
| 605 | sub build_cfg_extra {
|
---|
| 606 | my $self = shift(@_);
|
---|
| 607 | my ($build_cfg) = @_;
|
---|
| 608 |
|
---|
| 609 | }
|
---|
| 610 |
|
---|
| 611 |
|
---|
[14930] | 612 | sub collect_specific {
|
---|
| 613 | my $self = shift (@_);
|
---|
| 614 | }
|
---|
| 615 |
|
---|
| 616 | sub want_built {
|
---|
| 617 | my $self = shift (@_);
|
---|
| 618 | my ($index) = @_;
|
---|
| 619 |
|
---|
| 620 | if (defined ($self->{'collect_cfg'}->{'dontbuild'})) {
|
---|
| 621 | foreach my $checkstr (@{$self->{'collect_cfg'}->{'dontbuild'}}) {
|
---|
| 622 | if ($index =~ /^$checkstr$/) {
|
---|
| 623 | $self->{'notbuilt'}->{$index} = 1;
|
---|
| 624 | return 0;
|
---|
| 625 | }
|
---|
| 626 | }
|
---|
| 627 | }
|
---|
| 628 |
|
---|
| 629 | return 1;
|
---|
| 630 | }
|
---|
| 631 |
|
---|
| 632 | sub create_index_mapping {
|
---|
| 633 | my $self = shift (@_);
|
---|
| 634 | my ($indexes) = @_;
|
---|
| 635 |
|
---|
| 636 | print STDERR "create_index_mapping should be implemented in subclass\n";
|
---|
| 637 | my %mapping = ();
|
---|
| 638 | return \%mapping;
|
---|
| 639 | }
|
---|
| 640 |
|
---|
| 641 | # returns a processed version of a field.
|
---|
| 642 | # if the field has only one component the processed
|
---|
| 643 | # version will contain the first character and next consonant
|
---|
| 644 | # of that componant - otherwise it will contain the first
|
---|
| 645 | # character of the first two components
|
---|
| 646 | # only uses letdig (\w) characters now
|
---|
| 647 | sub process_field {
|
---|
| 648 | my $self = shift (@_);
|
---|
| 649 | my ($field) = @_;
|
---|
| 650 |
|
---|
| 651 | return "" unless (defined ($field) && $field =~ /\S/);
|
---|
| 652 |
|
---|
| 653 | my ($a, $b);
|
---|
| 654 | my @components = split /,/, $field;
|
---|
| 655 | if (scalar @components >= 2) {
|
---|
| 656 | # pick the first letdig from the first two field names
|
---|
| 657 | ($a) = $components[0] =~ /^[^\w]*(\w)/;
|
---|
| 658 | ($b) = $components[1] =~ /^[^\w]*(\w)/;
|
---|
| 659 | } else {
|
---|
| 660 | # pick the first two letdig chars
|
---|
| 661 | ($a, $b) = $field =~ /^[^\w]*(\w)[^\w]*?(\w)/i;
|
---|
| 662 | }
|
---|
| 663 | # there may not have been any letdigs...
|
---|
| 664 | $a = 'a' unless defined $a;
|
---|
| 665 | $b = '0' unless defined $b;
|
---|
| 666 |
|
---|
[22264] | 667 | my $newfield = "$a$b";
|
---|
| 668 | if ($newfield =~ /^\d\d$/) {
|
---|
| 669 | # digits only - Greenstone runtime doesn't like this.
|
---|
| 670 | $newfield = "a$a";
|
---|
| 671 | }
|
---|
| 672 | return $newfield;
|
---|
| 673 |
|
---|
[14930] | 674 | }
|
---|
| 675 |
|
---|
| 676 | sub get_next_version {
|
---|
| 677 | my $self = shift (@_);
|
---|
| 678 | my ($nameref) = @_;
|
---|
| 679 | my $num=0;
|
---|
| 680 | if ($$nameref =~ /(\d\d)$/) {
|
---|
| 681 | $num = $1; $num ++;
|
---|
| 682 | $$nameref =~ s/\d\d$/$num/;
|
---|
| 683 | } elsif ($$nameref =~ /(\d)$/) {
|
---|
| 684 | $num = $1;
|
---|
| 685 | if ($num == 9) {$$nameref =~ s/\d$/10/;}
|
---|
| 686 | else {$num ++; $$nameref =~ s/\d$/$num/;}
|
---|
| 687 | } else {
|
---|
| 688 | $$nameref =~ s/.$/0/;
|
---|
| 689 | }
|
---|
| 690 | }
|
---|
| 691 |
|
---|
| 692 |
|
---|
[14934] | 693 |
|
---|
[15709] | 694 | sub get_collection_meta_sets
|
---|
| 695 | {
|
---|
[14930] | 696 | my $self = shift(@_);
|
---|
[15709] | 697 | my $collection_infodb = shift(@_);
|
---|
[14930] | 698 |
|
---|
[14934] | 699 | my $mdprefix_fields = $self->{'buildproc'}->{'mdprefix_fields'};
|
---|
| 700 | foreach my $prefix (keys %$mdprefix_fields)
|
---|
| 701 | {
|
---|
[15709] | 702 | push(@{$collection_infodb->{"metadataset"}}, $prefix);
|
---|
[14934] | 703 |
|
---|
| 704 | foreach my $field (keys %{$mdprefix_fields->{$prefix}})
|
---|
| 705 | {
|
---|
[15709] | 706 | push(@{$collection_infodb->{"metadatalist-$prefix"}}, $field);
|
---|
| 707 |
|
---|
[14934] | 708 | my $val = $mdprefix_fields->{$prefix}->{$field};
|
---|
[15709] | 709 | push(@{$collection_infodb->{"metadatafreq-$prefix-$field"}}, $val);
|
---|
[14934] | 710 | }
|
---|
| 711 | }
|
---|
[15709] | 712 | }
|
---|
[14934] | 713 |
|
---|
| 714 |
|
---|
| 715 | # default is to output the metadata sets (prefixes) used in collection
|
---|
[15709] | 716 | sub output_collection_meta
|
---|
| 717 | {
|
---|
[14934] | 718 | my $self = shift(@_);
|
---|
[15709] | 719 | my $infodb_handle = shift(@_);
|
---|
[14934] | 720 |
|
---|
[15709] | 721 | my %collection_infodb = ();
|
---|
| 722 | $self->get_collection_meta_sets(\%collection_infodb);
|
---|
[15725] | 723 | &dbutil::write_infodb_entry($self->{'infodbtype'}, $infodb_handle, "collection", \%collection_infodb);
|
---|
[15709] | 724 | }
|
---|
[14934] | 725 |
|
---|
[17573] | 726 | # sometimes we need to read in an existing build.cfg - for example,
|
---|
| 727 | # if doing each stage of building separately, or when doing incremental
|
---|
| 728 | # building
|
---|
| 729 | sub read_build_cfg {
|
---|
| 730 | my $self = shift(@_);
|
---|
[14934] | 731 |
|
---|
[20095] | 732 | my $buildconfigfilename;
|
---|
[17573] | 733 |
|
---|
[20095] | 734 | if ($gs_mode eq "gs2") {
|
---|
| 735 | $buildconfigfilename = "build.cfg";
|
---|
| 736 | } else {
|
---|
| 737 | $buildconfigfilename = "buildConfig.xml";
|
---|
| 738 | }
|
---|
| 739 |
|
---|
| 740 | my $buildconfigfile = &util::filename_cat($self->{'build_dir'}, $buildconfigfilename);
|
---|
| 741 |
|
---|
[17573] | 742 | if (!-e $buildconfigfile) {
|
---|
| 743 | # try the index dir - but do we know where it is?? try here
|
---|
[20095] | 744 | $buildconfigfile = &util::filename_cat($ENV{'GSDLCOLLECTDIR'}, "index", $buildconfigfilename);
|
---|
[17573] | 745 | if (!-e $buildconfigfile) {
|
---|
| 746 | #we cant find a config file - just ignore the field list
|
---|
| 747 | return undef;
|
---|
| 748 | }
|
---|
| 749 | }
|
---|
[20100] | 750 | return &colcfg::read_building_cfg( $buildconfigfile, $gs_mode);
|
---|
[20095] | 751 |
|
---|
[17573] | 752 | }
|
---|
| 753 |
|
---|
[14930] | 754 | sub print_stats {
|
---|
| 755 | my $self = shift (@_);
|
---|
| 756 |
|
---|
| 757 | my $outhandle = $self->{'outhandle'};
|
---|
| 758 | my $indexing_text = $self->{'buildproc'}->get_indexing_text();
|
---|
| 759 | my $index = $self->{'buildproc'}->get_index();
|
---|
| 760 | my $num_bytes = $self->{'buildproc'}->get_num_bytes();
|
---|
| 761 | my $num_processed_bytes = $self->{'buildproc'}->get_num_processed_bytes();
|
---|
| 762 |
|
---|
| 763 | if ($indexing_text) {
|
---|
| 764 | print $outhandle "Stats (Creating index $index)\n";
|
---|
| 765 | } else {
|
---|
| 766 | print $outhandle "Stats (Compressing text from $index)\n";
|
---|
| 767 | }
|
---|
| 768 | print $outhandle "Total bytes in collection: $num_bytes\n";
|
---|
| 769 | print $outhandle "Total bytes in $index: $num_processed_bytes\n";
|
---|
| 770 |
|
---|
| 771 | if ($num_processed_bytes < 50 && ($indexing_text || !$self->{'no_text'})) {
|
---|
| 772 |
|
---|
[20686] | 773 | if ($self->{'incremental'}) {
|
---|
[14930] | 774 | if ($num_processed_bytes == 0) {
|
---|
| 775 | if ($indexing_text) {
|
---|
| 776 | print $outhandle "No additional text was added to $index\n";
|
---|
| 777 | } elsif (!$self->{'no_text'}) {
|
---|
| 778 | print $outhandle "No additional text was compressed\n";
|
---|
| 779 | }
|
---|
| 780 | }
|
---|
| 781 | }
|
---|
| 782 | else {
|
---|
| 783 | print $outhandle "***************\n";
|
---|
| 784 | if ($indexing_text) {
|
---|
| 785 | print $outhandle "WARNING: There is very little or no text to process for $index\n";
|
---|
| 786 | } elsif (!$self->{'no_text'}) {
|
---|
| 787 | print $outhandle "WARNING: There is very little or no text to compress\n";
|
---|
| 788 | }
|
---|
| 789 | print $outhandle " Was this your intention?\n";
|
---|
| 790 | print $outhandle "***************\n";
|
---|
| 791 | }
|
---|
| 792 |
|
---|
| 793 | }
|
---|
| 794 |
|
---|
| 795 | }
|
---|
| 796 |
|
---|
| 797 |
|
---|
| 798 | 1;
|
---|
| 799 |
|
---|