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;
|
---|
34 | use dbutil;
|
---|
35 | use plugin;
|
---|
36 | use util;
|
---|
37 | use FileUtils;
|
---|
38 |
|
---|
39 |
|
---|
40 | BEGIN {
|
---|
41 | # set autoflush on for STDERR and STDOUT so that mgpp
|
---|
42 | # doesn't get out of sync with plugins
|
---|
43 | STDOUT->autoflush(1);
|
---|
44 | STDERR->autoflush(1);
|
---|
45 | }
|
---|
46 |
|
---|
47 | END {
|
---|
48 | STDOUT->autoflush(0);
|
---|
49 | STDERR->autoflush(0);
|
---|
50 | }
|
---|
51 |
|
---|
52 | our $maxdocsize = 12000;
|
---|
53 |
|
---|
54 | # used to signify "gs2"(default) or "gs3"
|
---|
55 | our $gs_mode = "gs2";
|
---|
56 |
|
---|
57 | sub new {
|
---|
58 | my ($class, $site, $collection, $source_dir, $build_dir, $verbosity,
|
---|
59 | $maxdocs, $debug, $keepold, $incremental, $incremental_mode,
|
---|
60 | $remove_empty_classifications,
|
---|
61 | $outhandle, $no_text, $failhandle, $gli) = @_;
|
---|
62 |
|
---|
63 | $outhandle = *STDERR unless defined $outhandle;
|
---|
64 | $no_text = 0 unless defined $no_text;
|
---|
65 | $failhandle = *STDERR unless defined $failhandle;
|
---|
66 |
|
---|
67 | # create a builder object
|
---|
68 | my $self = bless {'site'=>$site, # will be undef for Greenstone 2
|
---|
69 | 'collection'=>$collection,
|
---|
70 | 'source_dir'=>$source_dir,
|
---|
71 | 'build_dir'=>$build_dir,
|
---|
72 | 'verbosity'=>$verbosity,
|
---|
73 | 'maxdocs'=>$maxdocs,
|
---|
74 | 'debug'=>$debug,
|
---|
75 | 'keepold'=>$keepold,
|
---|
76 | 'incremental'=>$incremental,
|
---|
77 | 'incremental_mode'=>$incremental_mode,
|
---|
78 | 'remove_empty_classifications'=>$remove_empty_classifications,
|
---|
79 | 'outhandle'=>$outhandle,
|
---|
80 | 'no_text'=>$no_text,
|
---|
81 | 'failhandle'=>$failhandle,
|
---|
82 | 'notbuilt'=>{}, # indexes not built
|
---|
83 | 'gli'=>$gli
|
---|
84 | }, $class;
|
---|
85 |
|
---|
86 | $self->{'gli'} = 0 unless defined $self->{'gli'};
|
---|
87 |
|
---|
88 | # Read in the collection configuration file.
|
---|
89 | if ((defined $site) && ($site ne "")) { # GS3
|
---|
90 | $gs_mode = "gs3";
|
---|
91 | }
|
---|
92 |
|
---|
93 | my $colcfgname = &colcfg::get_collect_cfg_name($outhandle, $gs_mode);
|
---|
94 | $self->{'collect_cfg'} = &colcfg::read_collection_cfg ($colcfgname, $gs_mode);
|
---|
95 |
|
---|
96 | if ($gs_mode eq "gs3") {
|
---|
97 | # read it in again to save the original form for later writing out
|
---|
98 | # of buildConfig.xml
|
---|
99 | # we use this preserve object because $self->{'collect_cfg'}->{'classify'} somewhat gets modified during the calling of &classify::load_classifiers.
|
---|
100 | $self->{'collect_cfg_preserve'} = &colcfg::read_collection_cfg ($colcfgname, $gs_mode);
|
---|
101 | }
|
---|
102 |
|
---|
103 | # get the database type for this collection from the collect.cfg file (may be undefined)
|
---|
104 | $self->{'infodbtype'} = $self->{'collect_cfg'}->{'infodbtype'} || &dbutil::get_default_infodb_type();
|
---|
105 |
|
---|
106 |
|
---|
107 | # load up any dontdb fields
|
---|
108 | $self->{'dontdb'} = {};
|
---|
109 | if (defined ($self->{'collect_cfg'}->{'dontgdbm'})) {
|
---|
110 | foreach my $dg (@{$self->{'collect_cfg'}->{'dontgdbm'}}) {
|
---|
111 | $self->{'dontdb'}->{$dg} = 1;
|
---|
112 | }
|
---|
113 | }
|
---|
114 |
|
---|
115 | $self->{'maxnumeric'} = 4;
|
---|
116 | return $self;
|
---|
117 | }
|
---|
118 |
|
---|
119 | # stuff has been moved here from new, so we can use subclass methods
|
---|
120 | sub init {
|
---|
121 | my $self = shift(@_);
|
---|
122 |
|
---|
123 | my $outhandle = $self->{'outhandle'};
|
---|
124 | my $failhandle = $self->{'failhandle'};
|
---|
125 |
|
---|
126 | $self->generate_index_list();
|
---|
127 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
128 | if (defined $indexes) {
|
---|
129 | # sort out subcollection indexes
|
---|
130 | if (defined $self->{'collect_cfg'}->{'indexsubcollections'}) {
|
---|
131 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
132 | foreach my $subcollection (@{$self->{'collect_cfg'}->{'indexsubcollections'}}) {
|
---|
133 | foreach my $index (@$indexes) {
|
---|
134 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$subcollection");
|
---|
135 | }
|
---|
136 | }
|
---|
137 | }
|
---|
138 |
|
---|
139 | # sort out language subindexes
|
---|
140 | if (defined $self->{'collect_cfg'}->{'languages'}) {
|
---|
141 | $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
142 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
143 | foreach my $language (@{$self->{'collect_cfg'}->{'languages'}}) {
|
---|
144 | foreach my $index (@$indexes) {
|
---|
145 | if (defined ($self->{'collect_cfg'}->{'indexsubcollections'})) {
|
---|
146 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$language");
|
---|
147 | }
|
---|
148 | else { # add in an empty subcollection field
|
---|
149 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index\:\:$language");
|
---|
150 | }
|
---|
151 | }
|
---|
152 | }
|
---|
153 | }
|
---|
154 | }
|
---|
155 |
|
---|
156 | if (defined($self->{'collect_cfg'}->{'indexes'})) {
|
---|
157 | # make sure that the same index isn't specified more than once
|
---|
158 | my %tmphash = ();
|
---|
159 | my @tmparray = @{$self->{'collect_cfg'}->{'indexes'}};
|
---|
160 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
161 | foreach my $i (@tmparray) {
|
---|
162 | if (!defined ($tmphash{$i})) {
|
---|
163 | push (@{$self->{'collect_cfg'}->{'indexes'}}, $i);
|
---|
164 | $tmphash{$i} = 1;
|
---|
165 | }
|
---|
166 | }
|
---|
167 | } else {
|
---|
168 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
169 | }
|
---|
170 |
|
---|
171 | # check incremental against whether builder can cope or not.
|
---|
172 | if ($self->{'incremental'} && !$self->is_incremental_capable()) {
|
---|
173 | print $outhandle "WARNING: The indexer used is not capable of incremental building. Reverting to -removeold\n";
|
---|
174 | $self->{'keepold'} = 0;
|
---|
175 | $self->{'incremental'} = 0;
|
---|
176 | $self->{'incremental_mode'} = "none";
|
---|
177 |
|
---|
178 | }
|
---|
179 |
|
---|
180 | # gs_version for plugins
|
---|
181 | my $gs_version = "2";
|
---|
182 | if ($gs_mode eq "gs3") {
|
---|
183 | $gs_version = "3";
|
---|
184 | }
|
---|
185 | # get the list of plugins for this collection
|
---|
186 | my $plugins = [];
|
---|
187 | if (defined $self->{'collect_cfg'}->{'plugin'}) {
|
---|
188 | $plugins = $self->{'collect_cfg'}->{'plugin'};
|
---|
189 | }
|
---|
190 |
|
---|
191 | # load all the plugins
|
---|
192 |
|
---|
193 | #build up the extra global options for the plugins
|
---|
194 | my @global_opts = ();
|
---|
195 | if (defined $self->{'collect_cfg'}->{'separate_cjk'} && $self->{'collect_cfg'}->{'separate_cjk'} =~ /^true$/i) {
|
---|
196 | push @global_opts, "-separate_cjk";
|
---|
197 | }
|
---|
198 | $self->{'pluginfo'} = &plugin::load_plugins ($plugins, $self->{'verbosity'}, $outhandle, $failhandle, \@global_opts, $self->{'incremental_mode'}, $gs_version);
|
---|
199 |
|
---|
200 | if (scalar(@{$self->{'pluginfo'}}) == 0) {
|
---|
201 | print $outhandle "No plugins were loaded.\n";
|
---|
202 | die "\n";
|
---|
203 | }
|
---|
204 |
|
---|
205 | # get the list of classifiers for this collection
|
---|
206 | my $classifiers = [];
|
---|
207 | if (defined $self->{'collect_cfg'}->{'classify'}) {
|
---|
208 | $classifiers = $self->{'collect_cfg'}->{'classify'};
|
---|
209 | }
|
---|
210 |
|
---|
211 | # load all the classifiers
|
---|
212 | $self->{'classifiers'} = &classify::load_classifiers ($classifiers, $self->{'build_dir'}, $outhandle);
|
---|
213 |
|
---|
214 | # load up the document processor for building
|
---|
215 | # if a buildproc class has been created for this collection, use it
|
---|
216 | # otherwise, use the default buildproc for the builder we are initialising
|
---|
217 | my $buildprocdir = undef;
|
---|
218 | my $buildproctype;
|
---|
219 |
|
---|
220 | my $collection = $self->{'collection'};
|
---|
221 | if (-e "$ENV{'GSDLCOLLECTDIR'}/custom/${collection}/perllib/custombuildproc.pm") {
|
---|
222 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/custom/${collection}/perllib";
|
---|
223 | $buildproctype = "custombuildproc";
|
---|
224 | } elsif (-e "$ENV{'GSDLCOLLECTDIR'}/perllib/custombuildproc.pm") {
|
---|
225 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/perllib";
|
---|
226 | $buildproctype = "custombuildproc";
|
---|
227 | } elsif (-e "$ENV{'GSDLCOLLECTDIR'}/perllib/${collection}buildproc.pm") {
|
---|
228 | $buildprocdir = "$ENV{'GSDLCOLLECTDIR'}/perllib";
|
---|
229 | $buildproctype = "${collection}buildproc";
|
---|
230 | } else {
|
---|
231 | $buildproctype = $self->default_buildproc();
|
---|
232 | }
|
---|
233 | if (defined $buildprocdir) {
|
---|
234 | require "$buildprocdir/$buildproctype.pm";
|
---|
235 | }
|
---|
236 | else {
|
---|
237 | require "$buildproctype.pm";
|
---|
238 | }
|
---|
239 |
|
---|
240 | eval("\$self->{'buildproc'} = new $buildproctype(\$self->{'collection'}, " .
|
---|
241 | "\$self->{'source_dir'}, \$self->{'build_dir'}, \$self->{'keepold'}, \$self->{'verbosity'}, \$self->{'outhandle'})");
|
---|
242 | die "$@" if $@;
|
---|
243 |
|
---|
244 | # We call set_infodbtype() now so the buildproc knows the infodbtype for all phases of the build
|
---|
245 | $self->{'buildproc'}->set_infodbtype($self->{'infodbtype'});
|
---|
246 |
|
---|
247 | $self->generate_index_options();
|
---|
248 |
|
---|
249 | if (!$self->{'debug'} && !$self->{'keepold'}) {
|
---|
250 | # remove any old builds
|
---|
251 | &FileUtils::removeFilesRecursive($self->{'build_dir'});
|
---|
252 | &FileUtils::makeAllDirectories($self->{'build_dir'});
|
---|
253 |
|
---|
254 | # make the text directory
|
---|
255 | my $textdir = "$self->{'build_dir'}/text";
|
---|
256 | &FileUtils::makeAllDirectories($textdir);
|
---|
257 | }
|
---|
258 |
|
---|
259 | if ($self->{'incremental'}) {
|
---|
260 | # some classes may need to do some additional initialisation
|
---|
261 | $self->init_for_incremental_build();
|
---|
262 | }
|
---|
263 |
|
---|
264 | }
|
---|
265 |
|
---|
266 | sub is_incremental_capable
|
---|
267 | {
|
---|
268 | # By default we return 'no' as the answer
|
---|
269 | # Safer to assume non-incremental to start with, and then override in
|
---|
270 | # inherited classes that are.
|
---|
271 |
|
---|
272 | return 0;
|
---|
273 | }
|
---|
274 |
|
---|
275 | # implement this in subclass if want to do additional initialisation for an
|
---|
276 | # incremental build
|
---|
277 | sub init_for_incremental_build {
|
---|
278 | my $self = shift (@_);
|
---|
279 | }
|
---|
280 |
|
---|
281 | sub deinit {
|
---|
282 | my $self = shift (@_);
|
---|
283 |
|
---|
284 | &plugin::deinit($self->{'pluginfo'},$self->{'buildproc'});
|
---|
285 | }
|
---|
286 |
|
---|
287 | sub generate_index_options {
|
---|
288 | my $self = shift (@_);
|
---|
289 |
|
---|
290 | my $separate_cjk = 0;
|
---|
291 |
|
---|
292 | my $indexoptions = $self->{'collect_cfg'}->{'indexoptions'};
|
---|
293 | if (defined($indexoptions)) {
|
---|
294 |
|
---|
295 | foreach my $option (@$indexoptions) {
|
---|
296 | if ($option =~ /separate_cjk/) {
|
---|
297 | $separate_cjk = 1;
|
---|
298 | }
|
---|
299 | }
|
---|
300 | }
|
---|
301 | # set this for building
|
---|
302 | $self->{'buildproc'}->set_separate_cjk($separate_cjk);
|
---|
303 | # record it for build.cfg
|
---|
304 | $self->{'separate_cjk'} = $separate_cjk;
|
---|
305 | }
|
---|
306 |
|
---|
307 | sub set_sections_index_document_metadata {
|
---|
308 | my $self = shift (@_);
|
---|
309 | my ($index) = @_;
|
---|
310 |
|
---|
311 | $self->{'buildproc'}->set_sections_index_document_metadata($index);
|
---|
312 | }
|
---|
313 |
|
---|
314 | sub set_maxnumeric {
|
---|
315 | my $self = shift (@_);
|
---|
316 | my ($maxnumeric) = @_;
|
---|
317 |
|
---|
318 | $self->{'maxnumeric'} = $maxnumeric;
|
---|
319 | }
|
---|
320 | sub set_strip_html {
|
---|
321 | my $self = shift (@_);
|
---|
322 | my ($strip) = @_;
|
---|
323 |
|
---|
324 | $self->{'strip_html'} = $strip;
|
---|
325 | $self->{'buildproc'}->set_strip_html($strip);
|
---|
326 | }
|
---|
327 |
|
---|
328 | sub set_store_metadata_coverage {
|
---|
329 | my $self = shift (@_);
|
---|
330 | my ($store_metadata_coverage) = @_;
|
---|
331 |
|
---|
332 | $self->{'buildproc'}->set_store_metadata_coverage($store_metadata_coverage);
|
---|
333 | }
|
---|
334 |
|
---|
335 | sub compress_text {
|
---|
336 | my $self = shift (@_);
|
---|
337 | my ($textindex) = @_;
|
---|
338 |
|
---|
339 | print STDERR "compress_text() should be implemented in subclass!!";
|
---|
340 | return;
|
---|
341 | }
|
---|
342 |
|
---|
343 |
|
---|
344 | sub build_indexes {
|
---|
345 | my $self = shift (@_);
|
---|
346 | my ($indexname) = @_;
|
---|
347 | my $outhandle = $self->{'outhandle'};
|
---|
348 |
|
---|
349 | $self->pre_build_indexes();
|
---|
350 |
|
---|
351 | my $indexes = [];
|
---|
352 | if (defined $indexname && $indexname =~ /\w/) {
|
---|
353 | push @$indexes, $indexname;
|
---|
354 | } else {
|
---|
355 | $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
356 | }
|
---|
357 |
|
---|
358 | # create the mapping between the index descriptions
|
---|
359 | # and their directory names (includes subcolls and langs)
|
---|
360 | $self->{'index_mapping'} = $self->create_index_mapping ($indexes);
|
---|
361 |
|
---|
362 | # build each of the indexes
|
---|
363 | foreach my $index (@$indexes) {
|
---|
364 | if ($self->want_built($index)) {
|
---|
365 | print $outhandle "\n*** building index $index in subdirectory " .
|
---|
366 | "$self->{'index_mapping'}->{$index}\n" if ($self->{'verbosity'} >= 1);
|
---|
367 | print STDERR "<Stage name='Index' source='$index'>\n" if $self->{'gli'};
|
---|
368 | $self->build_index($index);
|
---|
369 | } else {
|
---|
370 | print $outhandle "\n*** ignoring index $index\n" if ($self->{'verbosity'} >= 1);
|
---|
371 | }
|
---|
372 | }
|
---|
373 |
|
---|
374 | $self->post_build_indexes();
|
---|
375 |
|
---|
376 | }
|
---|
377 |
|
---|
378 | # implement this in subclass if want to do extra stuff at before building
|
---|
379 | # all the indexes
|
---|
380 | sub pre_build_indexes {
|
---|
381 | my $self = shift(@_);
|
---|
382 | my ($indexname) = @_; # optional parameter
|
---|
383 | }
|
---|
384 |
|
---|
385 | # implement this in subclass if want to do extra stuff at the end of building
|
---|
386 | # all the indexes
|
---|
387 | sub post_build_indexes {
|
---|
388 | my $self = shift(@_);
|
---|
389 | }
|
---|
390 |
|
---|
391 | sub build_index {
|
---|
392 | my $self = shift (@_);
|
---|
393 | my ($index) = @_;
|
---|
394 |
|
---|
395 | print STDERR "build_index should be implemented in subclass\n";
|
---|
396 | return;
|
---|
397 | }
|
---|
398 |
|
---|
399 |
|
---|
400 |
|
---|
401 | sub make_infodatabase {
|
---|
402 | my $self = shift (@_);
|
---|
403 | my $outhandle = $self->{'outhandle'};
|
---|
404 |
|
---|
405 | print STDERR "BuildDir: $self->{'build_dir'}\n";
|
---|
406 |
|
---|
407 | my $textdir = &FileUtils::filenameConcatenate($self->{'build_dir'}, "text");
|
---|
408 | my $assocdir = &FileUtils::filenameConcatenate($self->{'build_dir'}, "assoc");
|
---|
409 | &FileUtils::makeAllDirectories ($textdir);
|
---|
410 | &FileUtils::makeAllDirectories ($assocdir);
|
---|
411 |
|
---|
412 | # Get info database file path
|
---|
413 | my $infodb_type = $self->{'infodbtype'};
|
---|
414 | my $infodb_file_path = &dbutil::get_infodb_file_path($infodb_type, $self->{'collection'}, $textdir);
|
---|
415 |
|
---|
416 | print $outhandle "\n*** creating the info database and processing associated files\n"
|
---|
417 | if ($self->{'verbosity'} >= 1);
|
---|
418 | print STDERR "<Stage name='CreateInfoData'>\n" if $self->{'gli'};
|
---|
419 |
|
---|
420 | # init all the classifiers
|
---|
421 | &classify::init_classifiers ($self->{'classifiers'});
|
---|
422 |
|
---|
423 | my $reconstructed_docs = undef;
|
---|
424 | my $database_recs = undef;
|
---|
425 |
|
---|
426 | if ($self->{'incremental'}) {
|
---|
427 | $database_recs = {};
|
---|
428 |
|
---|
429 | &dbutil::read_infodb_file($infodb_type, $infodb_file_path, $database_recs);
|
---|
430 | }
|
---|
431 |
|
---|
432 |
|
---|
433 | # Important (for memory usage reasons) that we obtain the filehandle
|
---|
434 | # here for writing out to the database, rather than after
|
---|
435 | # $reconstructed_docs has been set up (assuming -incremental is on)
|
---|
436 | #
|
---|
437 | # This is because when we open a pipe to txt2db [using open()]
|
---|
438 | # this triggers a fork() followed by exec(). $reconstructed_docs
|
---|
439 | # can get very large, and so if we did the open() after this, it means
|
---|
440 | # the fork creates a clone of the *large* process image which (admittedly)
|
---|
441 | # is then quickly replaced in the execve() with the much smaller image for
|
---|
442 | # 'txt2db'. The trouble is, in that seismic second caused by
|
---|
443 | # the fork(), the system really does need to have all that memory available
|
---|
444 | # even though it isn't ultimately used. The result is an out of memory
|
---|
445 | # error.
|
---|
446 |
|
---|
447 | my ($infodb_handle);
|
---|
448 | if ($self->{'debug'}) {
|
---|
449 | $infodb_handle = *STDOUT;
|
---|
450 | }
|
---|
451 | else {
|
---|
452 | $infodb_handle = &dbutil::open_infodb_write_handle($infodb_type, $infodb_file_path);
|
---|
453 | if (!defined($infodb_handle))
|
---|
454 | {
|
---|
455 | print STDERR "<FatalError name='NoRunText2DB'/>\n</Stage>\n" if $self->{'gli'};
|
---|
456 | die "builder::make_infodatabase - couldn't open infodb write handle\n";
|
---|
457 | }
|
---|
458 | }
|
---|
459 |
|
---|
460 | if ($self->{'incremental'}) {
|
---|
461 | # reconstruct doc_obj metadata from database for all docs
|
---|
462 | $reconstructed_docs
|
---|
463 | = &classify::reconstruct_doc_objs_metadata($infodb_type,
|
---|
464 | $infodb_file_path,
|
---|
465 | $database_recs);
|
---|
466 | }
|
---|
467 |
|
---|
468 | # set up the document processor
|
---|
469 |
|
---|
470 | $self->{'buildproc'}->set_output_handle ($infodb_handle);
|
---|
471 | $self->{'buildproc'}->set_mode ('infodb');
|
---|
472 | $self->{'buildproc'}->set_assocdir ($assocdir);
|
---|
473 | $self->{'buildproc'}->set_dontdb ($self->{'dontdb'});
|
---|
474 | $self->{'buildproc'}->set_classifiers ($self->{'classifiers'});
|
---|
475 | $self->{'buildproc'}->set_indexing_text (0);
|
---|
476 | $self->{'buildproc'}->set_store_text(1);
|
---|
477 |
|
---|
478 | # make_infodatabase needs full reset even for incremental build
|
---|
479 | # as incremental works by reconstructing all docs from the database and
|
---|
480 | # then adding in the new ones
|
---|
481 | $self->{'buildproc'}->zero_reset();
|
---|
482 |
|
---|
483 | $self->{'buildproc'}->{'mdprefix_fields'} = {};
|
---|
484 |
|
---|
485 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
---|
486 | "", {}, {}, $self->{'buildproc'}, $self->{'maxdocs'},0, $self->{'gli'});
|
---|
487 |
|
---|
488 | if ($self->{'incremental'}) {
|
---|
489 | # create flat classify structure, ready for new docs to be added
|
---|
490 | foreach my $doc_obj ( @$reconstructed_docs ) {
|
---|
491 | if (! defined $self->{'buildproc'}->{'dont_process_reconstructed'}->{$doc_obj->get_OID()}) {
|
---|
492 | print $outhandle " Adding reconstructed ", $doc_obj->get_OID(), " into classify structures\n";
|
---|
493 | $self->{'buildproc'}->process($doc_obj,undef);
|
---|
494 | }
|
---|
495 | }
|
---|
496 | }
|
---|
497 | # this has changed to only output collection meta if its
|
---|
498 | # not in the config file
|
---|
499 | $self->output_collection_meta($infodb_handle);
|
---|
500 |
|
---|
501 | # output classification information
|
---|
502 | &classify::output_classify_info ($self->{'classifiers'}, $infodb_type, $infodb_handle,
|
---|
503 | $self->{'remove_empty_classifications'},
|
---|
504 | $self->{'gli'});
|
---|
505 |
|
---|
506 | # Output classifier reverse lookup, used in incremental deletion
|
---|
507 | ####&classify::print_reverse_lookup($infodb_handle);
|
---|
508 |
|
---|
509 | # output doclist
|
---|
510 | my @doc_list = $self->{'buildproc'}->get_doc_list();
|
---|
511 | my $browselist_infodb = { 'hastxt' => [ "0" ],
|
---|
512 | 'childtype' => [ "VList" ],
|
---|
513 | 'numleafdocs' => [ scalar(@doc_list) ],
|
---|
514 | 'thistype' => [ "Invisible" ],
|
---|
515 | 'contains' => [ join(";", @doc_list) ] };
|
---|
516 | &dbutil::write_infodb_entry($infodb_type, $infodb_handle, "browselist", $browselist_infodb);
|
---|
517 |
|
---|
518 | &dbutil::close_infodb_write_handle($infodb_type, $infodb_handle) if !$self->{'debug'};
|
---|
519 |
|
---|
520 | if ($infodb_type eq "gdbm-txtgz") {
|
---|
521 | my $gdb_infodb_file_path = &dbutil::get_infodb_file_path("gdbm", $self->{'collection'}, $textdir);
|
---|
522 | if (-e $gdb_infodb_file_path) {
|
---|
523 | &FileUtils::removeFiles($gdb_infodb_file_path);
|
---|
524 | }
|
---|
525 | }
|
---|
526 | print STDERR "</Stage>\n" if $self->{'gli'};
|
---|
527 | }
|
---|
528 |
|
---|
529 | sub make_auxiliary_files {
|
---|
530 | my $self = shift (@_);
|
---|
531 | my ($index);
|
---|
532 | my $build_cfg = {};
|
---|
533 | # subclasses may have already defined stuff in here
|
---|
534 | if (defined $self->{'build_cfg'}) {
|
---|
535 | $build_cfg = $self->{'build_cfg'};
|
---|
536 | }
|
---|
537 |
|
---|
538 | my $outhandle = $self->{'outhandle'};
|
---|
539 |
|
---|
540 | print $outhandle "\n*** creating auxiliary files \n" if ($self->{'verbosity'} >= 1);
|
---|
541 | print STDERR "<Stage name='CreatingAuxilary'>\n" if $self->{'gli'};
|
---|
542 |
|
---|
543 | # get the text directory
|
---|
544 | &FileUtils::makeAllDirectories ($self->{'build_dir'});
|
---|
545 |
|
---|
546 | # store the build date
|
---|
547 | $build_cfg->{'builddate'} = time;
|
---|
548 | $build_cfg->{'buildtype'} = $self->{'buildtype'};
|
---|
549 | $build_cfg->{'indexstem'} = &util::get_dirsep_tail($self->{'collection'});
|
---|
550 | $build_cfg->{'stemindexes'} = $self->{'stemindexes'};
|
---|
551 | if ($self->{'separate_cjk'}) {
|
---|
552 | $build_cfg->{'separate_cjk'} = "true";
|
---|
553 | }
|
---|
554 |
|
---|
555 | # store the number of documents and number of bytes
|
---|
556 | $build_cfg->{'numdocs'} = $self->{'buildproc'}->get_num_docs();
|
---|
557 | $build_cfg->{'numsections'} = $self->{'buildproc'}->get_num_sections();
|
---|
558 | $build_cfg->{'numbytes'} = $self->{'buildproc'}->get_num_bytes();
|
---|
559 |
|
---|
560 | # store the mapping between the index names and the directory names
|
---|
561 | # the index map is used to determine what indexes there are, so any that are not built should not be put into the map.
|
---|
562 | my @indexmap = ();
|
---|
563 | foreach my $index (@{$self->{'index_mapping'}->{'indexmaporder'}}) {
|
---|
564 | if (not defined ($self->{'notbuilt'}->{$index})) {
|
---|
565 | push (@indexmap, "$index\-\>$self->{'index_mapping'}->{'indexmap'}->{$index}");
|
---|
566 | }
|
---|
567 | }
|
---|
568 |
|
---|
569 | # store the number of indexes built to later determine whether search serviceracks get written out to buildConfig.xml
|
---|
570 | $build_cfg->{'num_indexes'} = scalar (@indexmap);
|
---|
571 |
|
---|
572 | $build_cfg->{'indexmap'} = \@indexmap if scalar (@indexmap);
|
---|
573 |
|
---|
574 | my @subcollectionmap = ();
|
---|
575 | foreach my $subcollection (@{$self->{'index_mapping'}->{'subcollectionmaporder'}}) {
|
---|
576 | push (@subcollectionmap, "$subcollection\-\>" .
|
---|
577 | $self->{'index_mapping'}->{'subcollectionmap'}->{$subcollection});
|
---|
578 | }
|
---|
579 | $build_cfg->{'subcollectionmap'} = \@subcollectionmap if scalar (@subcollectionmap);
|
---|
580 |
|
---|
581 | my @languagemap = ();
|
---|
582 | foreach my $language (@{$self->{'index_mapping'}->{'languagemaporder'}}) {
|
---|
583 | push (@languagemap, "$language\-\>" .
|
---|
584 | $self->{'index_mapping'}->{'languagemap'}->{$language});
|
---|
585 | }
|
---|
586 | $build_cfg->{'languagemap'} = \@languagemap if scalar (@languagemap);
|
---|
587 |
|
---|
588 | my @notbuilt = ();
|
---|
589 | foreach my $nb (keys %{$self->{'notbuilt'}}) {
|
---|
590 | push (@notbuilt, $nb);
|
---|
591 | }
|
---|
592 | $build_cfg->{'notbuilt'} = \@notbuilt if scalar (@notbuilt);
|
---|
593 |
|
---|
594 | $build_cfg->{'maxnumeric'} = $self->{'maxnumeric'};
|
---|
595 |
|
---|
596 | $build_cfg->{'infodbtype'} = $self->{'infodbtype'};
|
---|
597 |
|
---|
598 | # write out the earliestDatestamp information needed for OAI
|
---|
599 | my $archivedir = &FileUtils::filenameConcatenate($ENV{'GSDLCOLLECTDIR'}, "archives");
|
---|
600 | if(!-d $archivedir) {
|
---|
601 | $archivedir = &FileUtils::filenameConcatenate($ENV{'GSDLCOLLECTDIR'}, "export");
|
---|
602 | }
|
---|
603 | my $earliestDatestampFile = &FileUtils::filenameConcatenate($archivedir, "earliestDatestamp");
|
---|
604 | my $earliestDatestamp = 0;
|
---|
605 | if (open(FIN,"<$earliestDatestampFile")) {
|
---|
606 | {
|
---|
607 | # slurp in file as a single line
|
---|
608 | local $/ = undef;
|
---|
609 | $earliestDatestamp = <FIN>;
|
---|
610 | #&unicode::ensure_utf8(\$earliestDatestamp); # turn any high bytes that aren't valid utf-8 into utf-8.
|
---|
611 | }
|
---|
612 | close(FIN);
|
---|
613 | }
|
---|
614 | else {
|
---|
615 | print $outhandle "Warning: unable to read collection's earliestDatestamp from $earliestDatestampFile.\n";
|
---|
616 | print $outhandle "Setting value to 0.\n";
|
---|
617 | }
|
---|
618 | $build_cfg->{'earliestdatestamp'} = $earliestDatestamp;
|
---|
619 |
|
---|
620 | $self->build_cfg_extra($build_cfg);
|
---|
621 |
|
---|
622 | if ($gs_mode eq "gs2") {
|
---|
623 | &colcfg::write_build_cfg(&FileUtils::filenameConcatenate($self->{'build_dir'},"build.cfg"), $build_cfg);
|
---|
624 | }
|
---|
625 | if ($gs_mode eq "gs3") {
|
---|
626 |
|
---|
627 | &colcfg::write_build_cfg_xml(&FileUtils::filenameConcatenate($self->{'build_dir'}, "buildConfig.xml"), $build_cfg, $self->{'collect_cfg_preserve'});
|
---|
628 | }
|
---|
629 |
|
---|
630 | print STDERR "</Stage>\n" if $self->{'gli'};
|
---|
631 | }
|
---|
632 |
|
---|
633 | # implement this in subclass if want to add extra stuff to build.cfg
|
---|
634 | sub build_cfg_extra {
|
---|
635 | my $self = shift(@_);
|
---|
636 | my ($build_cfg) = @_;
|
---|
637 |
|
---|
638 | }
|
---|
639 |
|
---|
640 |
|
---|
641 | sub collect_specific {
|
---|
642 | my $self = shift (@_);
|
---|
643 | }
|
---|
644 |
|
---|
645 | sub want_built {
|
---|
646 | my $self = shift (@_);
|
---|
647 | my ($index) = @_;
|
---|
648 |
|
---|
649 | if (defined ($self->{'collect_cfg'}->{'dontbuild'})) {
|
---|
650 | foreach my $checkstr (@{$self->{'collect_cfg'}->{'dontbuild'}}) {
|
---|
651 | if ($index =~ /^$checkstr$/) {
|
---|
652 | $self->{'notbuilt'}->{$index} = 1;
|
---|
653 | return 0;
|
---|
654 | }
|
---|
655 | }
|
---|
656 | }
|
---|
657 |
|
---|
658 | return 1;
|
---|
659 | }
|
---|
660 |
|
---|
661 | sub create_index_mapping {
|
---|
662 | my $self = shift (@_);
|
---|
663 | my ($indexes) = @_;
|
---|
664 |
|
---|
665 | print STDERR "create_index_mapping should be implemented in subclass\n";
|
---|
666 | my %mapping = ();
|
---|
667 | return \%mapping;
|
---|
668 | }
|
---|
669 |
|
---|
670 | # returns a processed version of a field.
|
---|
671 | # if the field has only one component the processed
|
---|
672 | # version will contain the first character and next consonant
|
---|
673 | # of that componant - otherwise it will contain the first
|
---|
674 | # character of the first two components
|
---|
675 | # only uses letdig (\w) characters now
|
---|
676 | sub process_field {
|
---|
677 | my $self = shift (@_);
|
---|
678 | my ($field) = @_;
|
---|
679 |
|
---|
680 | return "" unless (defined ($field) && $field =~ /\S/);
|
---|
681 |
|
---|
682 | my ($a, $b);
|
---|
683 | my @components = split /,/, $field;
|
---|
684 | if (scalar @components >= 2) {
|
---|
685 | # pick the first letdig from the first two field names
|
---|
686 | ($a) = $components[0] =~ /^[^\w]*(\w)/;
|
---|
687 | ($b) = $components[1] =~ /^[^\w]*(\w)/;
|
---|
688 | } else {
|
---|
689 | # pick the first two letdig chars
|
---|
690 | ($a, $b) = $field =~ /^[^\w]*(\w)[^\w]*?(\w)/i;
|
---|
691 | }
|
---|
692 | # there may not have been any letdigs...
|
---|
693 | $a = 'a' unless defined $a;
|
---|
694 | $b = '0' unless defined $b;
|
---|
695 |
|
---|
696 | my $newfield = "$a$b";
|
---|
697 | if ($newfield =~ /^\d\d$/) {
|
---|
698 | # digits only - Greenstone runtime doesn't like this.
|
---|
699 | $newfield = "a$a";
|
---|
700 | }
|
---|
701 | return $newfield;
|
---|
702 |
|
---|
703 | }
|
---|
704 |
|
---|
705 | sub get_next_version {
|
---|
706 | my $self = shift (@_);
|
---|
707 | my ($nameref) = @_;
|
---|
708 | my $num=0;
|
---|
709 | if ($$nameref =~ /(\d\d)$/) {
|
---|
710 | $num = $1; $num ++;
|
---|
711 | $$nameref =~ s/\d\d$/$num/;
|
---|
712 | } elsif ($$nameref =~ /(\d)$/) {
|
---|
713 | $num = $1;
|
---|
714 | if ($num == 9) {$$nameref =~ s/\d$/10/;}
|
---|
715 | else {$num ++; $$nameref =~ s/\d$/$num/;}
|
---|
716 | } else {
|
---|
717 | $$nameref =~ s/.$/0/;
|
---|
718 | }
|
---|
719 | }
|
---|
720 |
|
---|
721 |
|
---|
722 |
|
---|
723 | sub get_collection_meta_sets
|
---|
724 | {
|
---|
725 | my $self = shift(@_);
|
---|
726 | my $collection_infodb = shift(@_);
|
---|
727 |
|
---|
728 | my $mdprefix_fields = $self->{'buildproc'}->{'mdprefix_fields'};
|
---|
729 | foreach my $prefix (keys %$mdprefix_fields)
|
---|
730 | {
|
---|
731 | push(@{$collection_infodb->{"metadataset"}}, $prefix);
|
---|
732 |
|
---|
733 | foreach my $field (keys %{$mdprefix_fields->{$prefix}})
|
---|
734 | {
|
---|
735 | push(@{$collection_infodb->{"metadatalist-$prefix"}}, $field);
|
---|
736 |
|
---|
737 | my $val = $mdprefix_fields->{$prefix}->{$field};
|
---|
738 | push(@{$collection_infodb->{"metadatafreq-$prefix-$field"}}, $val);
|
---|
739 | }
|
---|
740 | }
|
---|
741 | }
|
---|
742 |
|
---|
743 |
|
---|
744 | # default is to output the metadata sets (prefixes) used in collection
|
---|
745 | sub output_collection_meta
|
---|
746 | {
|
---|
747 | my $self = shift(@_);
|
---|
748 | my $infodb_handle = shift(@_);
|
---|
749 |
|
---|
750 | my %collection_infodb = ();
|
---|
751 | $self->get_collection_meta_sets(\%collection_infodb);
|
---|
752 | &dbutil::write_infodb_entry($self->{'infodbtype'}, $infodb_handle, "collection", \%collection_infodb);
|
---|
753 | }
|
---|
754 |
|
---|
755 | # sometimes we need to read in an existing build.cfg - for example,
|
---|
756 | # if doing each stage of building separately, or when doing incremental
|
---|
757 | # building
|
---|
758 | sub read_build_cfg {
|
---|
759 | my $self = shift(@_);
|
---|
760 |
|
---|
761 | my $buildconfigfilename;
|
---|
762 |
|
---|
763 | if ($gs_mode eq "gs2") {
|
---|
764 | $buildconfigfilename = "build.cfg";
|
---|
765 | } else {
|
---|
766 | $buildconfigfilename = "buildConfig.xml";
|
---|
767 | }
|
---|
768 |
|
---|
769 | my $buildconfigfile = &FileUtils::filenameConcatenate($self->{'build_dir'}, $buildconfigfilename);
|
---|
770 |
|
---|
771 | if (!-e $buildconfigfile) {
|
---|
772 | # try the index dir - but do we know where it is?? try here
|
---|
773 | $buildconfigfile = &FileUtils::filenameConcatenate($ENV{'GSDLCOLLECTDIR'}, "index", $buildconfigfilename);
|
---|
774 | if (!-e $buildconfigfile) {
|
---|
775 | #we cant find a config file - just ignore the field list
|
---|
776 | return undef;
|
---|
777 | }
|
---|
778 | }
|
---|
779 | return &colcfg::read_building_cfg( $buildconfigfile, $gs_mode);
|
---|
780 |
|
---|
781 | }
|
---|
782 |
|
---|
783 | sub print_stats {
|
---|
784 | my $self = shift (@_);
|
---|
785 |
|
---|
786 | my $outhandle = $self->{'outhandle'};
|
---|
787 | my $indexing_text = $self->{'buildproc'}->get_indexing_text();
|
---|
788 | my $index = $self->{'buildproc'}->get_index();
|
---|
789 | my $num_bytes = $self->{'buildproc'}->get_num_bytes();
|
---|
790 | my $num_processed_bytes = $self->{'buildproc'}->get_num_processed_bytes();
|
---|
791 |
|
---|
792 | if ($indexing_text) {
|
---|
793 | print $outhandle "Stats (Creating index $index)\n";
|
---|
794 | } else {
|
---|
795 | print $outhandle "Stats (Compressing text from $index)\n";
|
---|
796 | }
|
---|
797 | print $outhandle "Total bytes in collection: $num_bytes\n";
|
---|
798 | print $outhandle "Total bytes in $index: $num_processed_bytes\n";
|
---|
799 |
|
---|
800 | if ($num_processed_bytes < 50 && ($indexing_text || !$self->{'no_text'})) {
|
---|
801 |
|
---|
802 | if ($self->{'incremental'}) {
|
---|
803 | if ($num_processed_bytes == 0) {
|
---|
804 | if ($indexing_text) {
|
---|
805 | print $outhandle "No additional text was added to $index\n";
|
---|
806 | } elsif (!$self->{'no_text'}) {
|
---|
807 | print $outhandle "No additional text was compressed\n";
|
---|
808 | }
|
---|
809 | }
|
---|
810 | }
|
---|
811 | else {
|
---|
812 | print $outhandle "***************\n";
|
---|
813 | if ($indexing_text) {
|
---|
814 | print $outhandle "WARNING: There is very little or no text to process for $index\n";
|
---|
815 | } elsif (!$self->{'no_text'}) {
|
---|
816 | print $outhandle "WARNING: There is very little or no text to compress\n";
|
---|
817 | }
|
---|
818 | print $outhandle " Was this your intention?\n";
|
---|
819 | print $outhandle "***************\n";
|
---|
820 | }
|
---|
821 |
|
---|
822 | }
|
---|
823 |
|
---|
824 | }
|
---|
825 |
|
---|
826 |
|
---|
827 | 1;
|
---|
828 |
|
---|