source: gsdl/trunk/perllib/lucenebuilder.pm@ 15073

Last change on this file since 15073 was 15003, checked in by davidb, 16 years ago

With the new ability to group collections within a top-level collection, import.pl and buildcol.pl commands can now be given collection names such as "mygroup/demo". Some of the building code -- mostly involved with generating the index files and GDBM files -- had to be updated to correctly handle such compound collection names

  • Property svn:keywords set to Author Date Id Revision
File size: 14.7 KB
Line 
1###########################################################################
2#
3# lucenebuilder.pm -- perl wrapper for building index with Lucene
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###########################################################################
27# /*
28# * @version 1.0 ?
29# * @version 2.0 Incremental building assistance added, including
30# * remove_document_from_database which implements the granddad's
31# * empty function to call the lucene_passes.pl and full_lucene_passes_exe
32# * so there is one place in the code that works out where the
33# * perl script is. John Rowe
34# *
35# * @author John Rowe, DL Consulting Ltd.
36# */
37###########################################################################
38
39package lucenebuilder;
40
41# Use same basic XML structure setup by mgppbuilder/mgppbuildproc
42
43use mgppbuilder;
44
45# use IncrementalBuildUtils;
46
47sub BEGIN {
48 @lucenebuilder::ISA = ('mgppbuilder');
49}
50
51# /**
52# * @author John Thompson, DL Consulting Ltd.
53# */
54sub new {
55 my $class = shift(@_);
56 my $self = new mgppbuilder (@_);
57 $self = bless $self, $class;
58
59 $self->{'buildtype'} = "lucene";
60
61 # Do we need to put exe on the end?
62 my $exe = &util::get_os_exe ();
63 my $scriptdir = "$ENV{'GSDLHOME'}/bin/script";
64
65 # So where is lucene_passes.pl anyway?
66 my $lucene_passes_script = &util::filename_cat($scriptdir, "lucene_passes.pl");
67
68 # So tack perl on the beginning to ensure execution
69 $self->{'full_lucene_passes'} = "$lucene_passes_script";
70 if ($exe eq ".exe")
71 {
72 $self->{'full_lucene_passes_exe'} = "perl$exe \"$lucene_passes_script\"";
73 }
74 else
75 {
76 $self->{'full_lucene_passes_exe'} = "perl -S \"$lucene_passes_script\"";
77 }
78
79 return $self;
80}
81# /** new() **/
82
83# lucene has none of these options
84sub generate_index_options {
85 my $self = shift (@_);
86
87 $self->{'casefold'} = 0;
88 $self->{'stem'} = 0;
89 $self->{'accentfold'} = 0;
90 $self->{'stemindexes'} = 0;
91}
92
93sub default_buildproc {
94 my $self = shift (@_);
95
96 return "lucenebuildproc";
97}
98
99# this writes a nice version of the text docs
100sub compress_text
101{
102 my $self = shift (@_);
103 # we don't do anything if we don't want compressed text
104 return if $self->{'no_text'};
105
106 my ($textindex) = @_;
107 my $outhandle = $self->{'outhandle'};
108 print STDERR "Saving the document text\n";
109 # the text directory
110 my $text_dir = &util::filename_cat($self->{'build_dir'}, "text");
111 my $build_dir = &util::filename_cat($self->{'build_dir'},"");
112 &util::mk_all_dir ($text_dir);
113
114 my $osextra = "";
115 if ($ENV{'GSDLOS'} =~ /^windows$/i)
116 {
117 $text_dir =~ s@/@\\@g;
118 }
119 else
120 {
121 if ($outhandle ne "STDERR")
122 {
123 # so lucene_passes doesn't print to stderr if we redirect output
124 $osextra .= " 2>/dev/null";
125 }
126 }
127
128 # get any os specific stuff
129 my $scriptdir = "$ENV{'GSDLHOME'}/bin/script";
130
131 # Find the perl script to call to run lucene
132 my $full_lucene_passes = $self->{'full_lucene_passes'};
133 my $full_lucene_passes_exe = $self->{'full_lucene_passes_exe'};
134
135 my $lucene_passes_sections = "Doc";
136
137 my ($handle);
138
139 if ($self->{'debug'})
140 {
141 $handle = STDOUT;
142 }
143 else
144 {
145 print STDERR "Full Path: $full_lucene_passes\n";
146 print STDERR "Executable: $full_lucene_passes_exe\n";
147 print STDERR "Sections: $lucene_passes_sections\n";
148 print STDERR "Build Dir: $build_dir\n";
149 print STDERR "Cmd: $full_lucene_passes_exe text $lucene_passes_sections \"$build_dir\" \"dummy\" $osextra\n";
150 if (!-e "$full_lucene_passes" ||
151 !open (PIPEOUT, "| $full_lucene_passes_exe text $lucene_passes_sections \"$build_dir\" \"dummy\" $osextra"))
152 {
153 print STDERR "<FatalError name='NoRunLucenePasses'/>\n</Stage>\n" if $self->{'gli'};
154 die "lucenebuilder::build_index - couldn't run $full_lucene_passes_exe\n";
155 }
156 $handle = lucenebuilder::PIPEOUT;
157 }
158
159 # stored text is always Doc and Sec levels
160 my $levels = { 'document' => 1, 'section' => 1 };
161 # always do gdbm at section level
162 my $gdbm_level = "section";
163
164 # set up the document processr
165 $self->{'buildproc'}->set_output_handle ($handle);
166 $self->{'buildproc'}->set_mode ('text');
167 $self->{'buildproc'}->set_index ($textindex);
168 $self->{'buildproc'}->set_indexing_text (0);
169 $self->{'buildproc'}->set_indexfieldmap ($self->{'indexfieldmap'});
170 $self->{'buildproc'}->set_levels ($levels);
171 $self->{'buildproc'}->set_gdbm_level ($gdbm_level);
172 $self->{'buildproc'}->reset();
173 &plugin::begin($self->{'pluginfo'}, $self->{'source_dir'},
174 $self->{'buildproc'}, $self->{'maxdocs'});
175 &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
176 "", {}, $self->{'buildproc'}, $self->{'maxdocs'}, 0, $self->{'gli'});
177 &plugin::end($self->{'pluginfo'});
178 close ($handle) unless $self->{'debug'};
179 close PIPEOUT;
180 $self->print_stats();
181
182 print STDERR "</Stage>\n" if $self->{'gli'};
183}
184
185sub build_indexes {
186 my $self = shift (@_);
187 my ($indexname) = @_;
188 my $outhandle = $self->{'outhandle'};
189
190 my $indexes = [];
191 if (defined $indexname && $indexname =~ /\w/) {
192 push @$indexes, $indexname;
193 } else {
194 $indexes = $self->{'collect_cfg'}->{'indexes'};
195 }
196
197 # create the mapping between the index descriptions
198 # and their directory names (includes subcolls and langs)
199 $self->{'index_mapping'} = $self->create_index_mapping ($indexes);
200
201 # build each of the indexes
202 foreach $index (@$indexes) {
203 if ($self->want_built($index)) {
204
205 my $idx = $self->{'index_mapping'}->{$index};
206 foreach my $level (keys %{$self->{'levels'}}) {
207 next if $level =~ /paragraph/; # we don't do para indexing
208 my ($pindex) = $level =~ /^(.)/;
209 # should probably check that new name with level
210 # is unique ... but currently (with doc sec and para)
211 # each has unique first letter.
212 $self->{'index_mapping'}->{$index} = $pindex.$idx;
213
214 my $llevel = $mgppbuilder::level_map{$level};
215 print $outhandle "\n*** building index $index at level $llevel in subdirectory " .
216 "$self->{'index_mapping'}->{$index}\n" if ($self->{'verbosity'} >= 1);
217 print STDERR "<Stage name='Index' source='$index' level=$llevel>\n" if $self->{'gli'};
218
219 $self->build_index($index,$llevel);
220 }
221 $self->{'index_mapping'}->{$index} = $idx;
222
223 } else {
224 print $outhandle "\n*** ignoring index $index\n" if ($self->{'verbosity'} >= 1);
225 }
226 }
227
228 #define the final field lists
229 $self->make_final_field_list();
230}
231
232
233sub build_index {
234 my $self = shift (@_);
235 my ($index,$llevel) = @_;
236 my $outhandle = $self->{'outhandle'};
237 my $build_dir = $self->{'build_dir'};
238
239 # get the full index directory path and make sure it exists
240 my $indexdir = $self->{'index_mapping'}->{$index};
241 &util::mk_all_dir (&util::filename_cat($build_dir, $indexdir));
242
243 # get any os specific stuff
244 my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
245 my $scriptdir = "$ENV{'GSDLHOME'}/bin/script";
246
247 # Find the perl script to call to run lucene
248 my $full_lucene_passes = $self->{'full_lucene_passes'};
249 my $full_lucene_passes_exe = $self->{'full_lucene_passes_exe'};
250
251 # define the section names for lucenepasses
252 # define the section names and possibly the doc name for lucenepasses
253 my $lucene_passes_sections = $llevel;
254
255 my $opt_create_index = ($self->{'keepold'}) ? "" : "-create";
256
257 my $osextra = "";
258 if ($ENV{'GSDLOS'} =~ /^windows$/i) {
259 $build_dir =~ s@/@\\@g;
260 } else {
261 if ($outhandle ne "STDERR") {
262 # so lucene_passes doesn't print to stderr if we redirect output
263 $osextra .= " 2>/dev/null";
264 }
265 }
266
267 # get the index expression if this index belongs
268 # to a subcollection
269 my $indexexparr = [];
270 my $langarr = [];
271
272 # there may be subcollection info, and language info.
273 my ($fields, $subcollection, $language) = split (":", $index);
274 my @subcollections = ();
275 @subcollections = split /,/, $subcollection if (defined $subcollection);
276
277 foreach $subcollection (@subcollections) {
278 if (defined ($self->{'collect_cfg'}->{'subcollection'}->{$subcollection})) {
279 push (@$indexexparr, $self->{'collect_cfg'}->{'subcollection'}->{$subcollection});
280 }
281 }
282
283 # add expressions for languages if this index belongs to
284 # a language subcollection - only put languages expressions for the
285 # ones we want in the index
286 my @languages = ();
287 my $language_metadata = "Language";
288 if (defined ($self->{'collect_cfg'}->{'language_metadata'})) {
289 $language_metadata = $self->{'collect_cfg'}->{'language_metadata'};
290 }
291 @languages = split /,/, $language if (defined $language);
292 foreach my $language (@languages) {
293 my $not=0;
294 if ($language =~ s/^\!//) {
295 $not = 1;
296 }
297 if($not) {
298 push (@$langarr, "!$language");
299 } else {
300 push (@$langarr, "$language");
301 }
302 }
303
304 # Build index dictionary. Uses verbatim stem method
305 print $outhandle "\n creating index dictionary (lucene_passes -I1)\n" if ($self->{'verbosity'} >= 1);
306 print STDERR "<Phase name='CreatingIndexDic'/>\n" if $self->{'gli'};
307 my ($handle);
308
309 if ($self->{'debug'}) {
310 $handle = STDOUT;
311 } else {
312 print STDERR "Cmd: $full_lucene_passes_exe $opt_create_index index $lucene_passes_sections \"$build_dir\" \"$indexdir\" $osextra\n";
313 if (!-e "$full_lucene_passes" ||
314 !open (PIPEOUT, "| $full_lucene_passes_exe $opt_create_index index $lucene_passes_sections \"$build_dir\" \"$indexdir\" $osextra")) {
315 print STDERR "<FatalError name='NoRunLucenePasses'/>\n</Stage>\n" if $self->{'gli'};
316 die "lucenebuilder::build_index - couldn't run $lucene_passes_exe\n";
317 }
318 $handle = lucenebuilder::PIPEOUT;
319 }
320
321 my $store_levels = $self->{'levels'};
322 my $gdbm_level = "section"; #always
323 my $dom_level = "";
324 foreach my $key (keys %$store_levels) {
325 if ($mgppbuilder::level_map{$key} eq $llevel) {
326 $dom_level = $key;
327 }
328 }
329 if ($dom_level eq "") {
330 print STDERR "Warning: unrecognized tag level $llevel\n";
331 $dom_level = "document";
332 }
333
334 my $local_levels = { $dom_level => 1 }; # work on one level at a time
335
336 # set up the document processr
337 $self->{'buildproc'}->set_output_handle ($handle);
338 $self->{'buildproc'}->set_mode ('text');
339 $self->{'buildproc'}->set_index ($index, $indexexparr);
340 $self->{'buildproc'}->set_index_languages ($language_metadata, $langarr) if (defined $language);
341 $self->{'buildproc'}->set_indexing_text (1);
342 $self->{'buildproc'}->set_indexfieldmap ($self->{'indexfieldmap'});
343 $self->{'buildproc'}->set_levels ($local_levels);
344 $self->{'buildproc'}->set_gdbm_level($gdbm_level);
345 $self->{'buildproc'}->reset();
346 &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
347 "", {}, $self->{'buildproc'}, $self->{'maxdocs'}, 0, $self->{'gli'});
348 close ($handle) unless $self->{'debug'};
349
350 $self->print_stats();
351
352 $self->{'buildproc'}->set_levels ($store_levels);
353 print STDERR "</Stage>\n" if $self->{'gli'};
354}
355
356# /** A modified version of the basebuilder.pm's function that generates the
357# * information database (GDBM) from the GA documents. We need to change this
358# * so that if we've been asked to do an incremental build we only add
359# * metadata to autohierarchy classifiers via the IncrementalBuildUtils
360# * module. All other classifiers and metadata will be ignored.
361# */
362sub make_infodatabase
363{
364 my $self = shift (@_);
365 my $outhandle = $self->{'outhandle'};
366
367 my $dbext = ".bdb";
368 $dbext = ".ldb" if &util::is_little_endian();
369
370 my $collect_tail = &util::get_dirsep_tail($self->{'collection'});
371 my $infodb_file = &util::filename_cat($self->{'build_dir'}, "text", $collect_tail . $dbext);
372
373 # If we aren't doing an incremental addition, then we just call the super-
374 # classes version
375 # Note: Incremental addition can only occur if a text/<collection>.ldb
376 # already exists. If it doesn't, let the super classes function be
377 # called once to generate it.
378 if (!$self->{'incremental_dlc'} || !(-e $infodb_file))
379 {
380 # basebuilder::make_infodatabase(@_);
381 # Note: this doesn't work as the direct reference means all the $self
382 # data is lost.
383 $self->basebuilder::make_infodatabase(@_);
384 return;
385 }
386
387 # Carry on with an incremental addition
388 print $outhandle "\n*** performing an incremental addition to the info database\n" if ($self->{'verbosity'} >= 1);
389 print STDERR "<Stage name='CreateInfoData'>\n" if $self->{'gli'};
390
391 # 1. Init all the classifiers
392 &classify::init_classifiers ($self->{'classifiers'});
393 # 2. Init the buildproc settings.
394 # Note: we still need this to process any associated files - but we
395 # don't expect to pipe anything to txt2db so we can do away with the
396 # complex output handle.
397 my $assocdir = &util::filename_cat($self->{'build_dir'}, "assoc");
398 &util::mk_all_dir ($assocdir);
399 $self->{'buildproc'}->set_mode ('incinfodb'); # Very Important
400 $self->{'buildproc'}->set_assocdir ($assocdir);
401 # 3. Read in all the metadata from the files in the archives directory using
402 # the GAPlug and using ourselves as the document processor!
403 &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'}, "", {}, $self->{'buildproc'}, $self->{'maxdocs'},0, $self->{'gli'});
404
405 print STDERR "</Stage>\n" if $self->{'gli'};
406}
407
408# /** Lucene specific document removal function. This works by calling lucene_passes.pl with
409# * -remove and the document id on the command line.
410# *
411# * @param oid is the document identifier to be removed.
412# *
413# * @author John Rowe, DL Consulting Ltd.
414# */
415sub remove_document_from_database
416{
417 my ($self, $oid) = @_;
418 # Find the perl script to call to run lucene
419 my $full_lucene_passes_exe = $self->{'full_lucene_passes_exe'};
420 # Call lucene_passes.pl with -remove and the document ID on the command line
421 `$full_lucene_passes_exe -remove "$oid"`;
422}
423# /** remove_document_from_database **/
424
425
4261;
427
428
Note: See TracBrowser for help on using the repository browser.