1 | # MGBuilder object
|
---|
2 | #
|
---|
3 |
|
---|
4 | package mgbuilder;
|
---|
5 |
|
---|
6 | use cfgread;
|
---|
7 | use colcfg;
|
---|
8 | use plugin;
|
---|
9 | use util;
|
---|
10 |
|
---|
11 | $maxdocsize = 12000;
|
---|
12 |
|
---|
13 | %wanted_index_files = ('td'=>1,
|
---|
14 | 't'=>1,
|
---|
15 | 'idb'=>1,
|
---|
16 | 'ib1'=>1,
|
---|
17 | 'ib2'=>1,
|
---|
18 | 'ib3'=>1,
|
---|
19 | 'i'=>1,
|
---|
20 | 'ip'=>1,
|
---|
21 | 'tiw'=>1,
|
---|
22 | 'wa'=>1);
|
---|
23 |
|
---|
24 |
|
---|
25 | sub new {
|
---|
26 | my ($class, $collection, $source_dir, $build_dir, $verbosity) = @_;
|
---|
27 |
|
---|
28 | # create an mgbuilder object
|
---|
29 | my $self = bless {'collection'=>$collection,
|
---|
30 | 'source_dir'=>$source_dir,
|
---|
31 | 'build_dir'=>$build_dir,
|
---|
32 | 'verbosity'=>$verbosity}, $class;
|
---|
33 |
|
---|
34 |
|
---|
35 | # read in the collection configuration file
|
---|
36 | if (!-e "$ENV{'GSDLHOME'}/collect/$collection/collect.cfg") {
|
---|
37 | die "mgbuilder::new - couldn't find collect.cfg for collection $collection\n";
|
---|
38 | }
|
---|
39 | $self->{'collect_cfg'} = &colcfg::read_collect_cfg ("$ENV{'GSDLHOME'}/collect/" .
|
---|
40 | "$collection/collect.cfg");
|
---|
41 |
|
---|
42 |
|
---|
43 | # sort out subcollection indexes
|
---|
44 | if (defined $self->{'collect_cfg'}->{'indexsubcollections'}) {
|
---|
45 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
46 | $self->{'collect_cfg'}->{'indexes'} = [];
|
---|
47 | foreach $subcollection (@{$self->{'collect_cfg'}->{'indexsubcollections'}}) {
|
---|
48 | foreach $index (@$indexes) {
|
---|
49 | push (@{$self->{'collect_cfg'}->{'indexes'}}, "$index:$subcollection");
|
---|
50 | }
|
---|
51 | }
|
---|
52 | }
|
---|
53 |
|
---|
54 | # get the list of plugins for this collection
|
---|
55 | my @plugins = (); # some good choice of plugins .... ????
|
---|
56 | if (defined $self->{'collect_cfg'}->{'plugins'}) {
|
---|
57 | @plugins = @{$self->{'collect_cfg'}->{'plugins'}};
|
---|
58 | }
|
---|
59 |
|
---|
60 |
|
---|
61 | # load all the plugins
|
---|
62 | $self->{'pluginfo'} = &plugin::load_plugins ($collection, \@plugins);
|
---|
63 | if (scalar(@{$self->{'pluginfo'}}) == 0) {
|
---|
64 | print STDERR "No plugins were loaded.\n";
|
---|
65 | die "\n";
|
---|
66 | }
|
---|
67 |
|
---|
68 |
|
---|
69 | # load up the document processor for building
|
---|
70 | # if a buildproc class has been created for this collection, use it
|
---|
71 | # otherwise, use the mg buildproc
|
---|
72 | my ($buildprocdir, $buildproctype);
|
---|
73 | if (-e "$ENV{'GSDLHOME'}/collect/$collection/perllib/${collection}buildproc.pm") {
|
---|
74 | $buildprocdir = "$ENV{'GSDLHOME'}/collect/$collection/perllib";
|
---|
75 | $buildproctype = "${collection}buildproc";
|
---|
76 | } else {
|
---|
77 | $buildprocdir = "$ENV{'GSDLHOME'}/perllib";
|
---|
78 | $buildproctype = "mgbuildproc";
|
---|
79 | }
|
---|
80 | require "$buildprocdir/$buildproctype.pm";
|
---|
81 |
|
---|
82 | eval("\$self->{'buildproc'} = new $buildproctype(\$collection, " .
|
---|
83 | "\$source_dir, \$build_dir, \$verbosity)");
|
---|
84 | die "$@" if $@;
|
---|
85 |
|
---|
86 |
|
---|
87 | return $self;
|
---|
88 | }
|
---|
89 |
|
---|
90 | sub init {
|
---|
91 | my $self = shift (@_);
|
---|
92 |
|
---|
93 | # remove any old builds
|
---|
94 | &util::rm_r($self->{'build_dir'});
|
---|
95 | &util::mk_all_dir($self->{'build_dir'});
|
---|
96 |
|
---|
97 | # make the text directory
|
---|
98 | my $textdir = "$self->{'build_dir'}/text";
|
---|
99 | &util::mk_all_dir($textdir);
|
---|
100 | }
|
---|
101 |
|
---|
102 | sub compress_text {
|
---|
103 | my $self = shift (@_);
|
---|
104 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
---|
105 | my $exe = &util::get_os_exe ();
|
---|
106 |
|
---|
107 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, "text"));
|
---|
108 | my $basefilename = "text/$self->{'collection'}";
|
---|
109 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, $basefilename);
|
---|
110 |
|
---|
111 | my $osextra = "";
|
---|
112 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
---|
113 | $fulltextprefix =~ s/\//\\/g;
|
---|
114 | } else {
|
---|
115 | $osextra = " -d /";
|
---|
116 | }
|
---|
117 |
|
---|
118 | print STDERR "\n*** creating the compressed text\n" if ($self->{'verbosity'} >= 1);
|
---|
119 |
|
---|
120 | # set up the document processor
|
---|
121 | $self->{'buildproc'}->set_output_handle ('mgbuilder::PIPEOUT');
|
---|
122 | $self->{'buildproc'}->set_mode ('text');
|
---|
123 | $self->{'buildproc'}->set_index ('section:text');
|
---|
124 |
|
---|
125 | # collect the statistics for the text
|
---|
126 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
---|
127 | print STDERR "\n collecting text statistics\n" if ($self->{'verbosity'} >= 1);
|
---|
128 | if (!-e "$exedir/mg_passes$exe" || !open (PIPEOUT,
|
---|
129 | "| $exedir/mg_passes$exe -f $fulltextprefix -b $maxdocsize -T1 $osextra")) {
|
---|
130 | die "mgbuilder::compress_text - couldn't run $exedir/mg_passes$exe\n";
|
---|
131 | }
|
---|
132 | $self->{'buildproc'}->reset();
|
---|
133 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
---|
134 | "", {}, $self->{'buildproc'});
|
---|
135 | close (PIPEOUT);
|
---|
136 |
|
---|
137 | # create the compression dictionary
|
---|
138 | # the compression dictionary is built by assuming the stats are from a seed
|
---|
139 | # dictionary (-S), if a novel word is encountered it is spelled out (-H),
|
---|
140 | # and the resulting dictionary must be less than 5 meg with the most frequent
|
---|
141 | # words being put into the dictionary first (-2 -k 5120)
|
---|
142 | print STDERR "\n creating the compression dictionary\n" if ($self->{'verbosity'} >= 1);
|
---|
143 | if (!-e "$exedir/mg_compression_dict$exe") {
|
---|
144 | die "mgbuilder::compress_text - couldn't run $exedir/mg_compression_dict$exe\n";
|
---|
145 | }
|
---|
146 | system ("$exedir/mg_compression_dict$exe -f $fulltextprefix -S -H -2 -k 5120 $osextra");
|
---|
147 |
|
---|
148 | # compress the text
|
---|
149 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
---|
150 | print STDERR "\n compressing the text\n" if ($self->{'verbosity'} >= 1);
|
---|
151 | if (!-e "$exedir/mg_passes$exe" || !open (PIPEOUT,
|
---|
152 | "| $exedir/mg_passes$exe -f $fulltextprefix -b $maxdocsize -T2 $osextra")) {
|
---|
153 | die "mgbuilder::compress_text - couldn't run $exedir/mg_passes$exe\n";
|
---|
154 | }
|
---|
155 | $self->{'buildproc'}->reset();
|
---|
156 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
---|
157 | "", {}, $self->{'buildproc'});
|
---|
158 | close (PIPEOUT);
|
---|
159 | }
|
---|
160 |
|
---|
161 | sub build_indexes {
|
---|
162 | my $self = shift (@_);
|
---|
163 | my $indexes = $self->{'collect_cfg'}->{'indexes'};
|
---|
164 |
|
---|
165 | # create the mapping between the index descriptions
|
---|
166 | # and their directory names
|
---|
167 | $self->{'index_mapping'} = $self->create_index_mapping ($indexes);
|
---|
168 |
|
---|
169 | # build each of the indexes
|
---|
170 | foreach $index (@$indexes) {
|
---|
171 | print STDERR "\n*** building index $index in subdirectory " .
|
---|
172 | "$self->{'index_mapping'}->{$index}\n" if ($self->{'verbosity'} >= 1);
|
---|
173 | $self->build_index($index);
|
---|
174 | }
|
---|
175 | }
|
---|
176 |
|
---|
177 | # creates directory names for each of the index descriptions
|
---|
178 | sub create_index_mapping {
|
---|
179 | my $self = shift (@_);
|
---|
180 | my ($indexes) = @_;
|
---|
181 |
|
---|
182 | my %mapping = ();
|
---|
183 |
|
---|
184 | # dirnames is used to check for collisions. Start this off
|
---|
185 | # with the manditory directory names
|
---|
186 | my %dirnames = ('text'=>'text',
|
---|
187 | 'extra'=>'extra');
|
---|
188 |
|
---|
189 | foreach $index (@$indexes) {
|
---|
190 | my ($level, $fields, $subcollection) = split (":", $index);
|
---|
191 | my @fields = split (/,/, $fields);
|
---|
192 | splice (@fields, 2); # just want first two fields
|
---|
193 |
|
---|
194 | # the directory names starts with the first character of the index level
|
---|
195 | my ($dirname) = $level =~ /^(.)/;
|
---|
196 |
|
---|
197 | # next comes a processed version of the first two fields in the index
|
---|
198 | # the processed version contains the first character and the next
|
---|
199 | # consonant
|
---|
200 | map {s/^(.).*?([bcdfghjklmnpqrstvwxyz]).*$/$1$2/i;} @fields;
|
---|
201 | $dirname .= join("", @fields);
|
---|
202 |
|
---|
203 | # next comes a processed version of the subcollection if there is one.
|
---|
204 | # the processed version contains the first character and the next
|
---|
205 | # consonant if there's only one field, other wise the first character
|
---|
206 | # of the first two fields
|
---|
207 | if (defined ($subcollection) && $subcollection =~ /\w/) {
|
---|
208 | @fields = split /,/, $subcollection;
|
---|
209 | if (scalar @fields >= 2) {
|
---|
210 | splice (@fields, 2);
|
---|
211 | map {s/^(.).*$/$1/i;} @fields;
|
---|
212 | $dirname .= join("", @fields);
|
---|
213 | } else {
|
---|
214 | $subcollection =~ s/^(.).*?([bcdfghjklmnpqrstvwxyz]).*$/$1$2/i;
|
---|
215 | $dirname .= $subcollection;
|
---|
216 | }
|
---|
217 | }
|
---|
218 |
|
---|
219 | # convert the directory name to lowercase
|
---|
220 | $dirname = lc ($dirname);
|
---|
221 |
|
---|
222 | # add a number to make this directory name unique
|
---|
223 | if (defined $dirnames{$dirname}) {
|
---|
224 | my $num = 1;
|
---|
225 | while (defined $dirnames{"$dirname$num"}) {
|
---|
226 | $num++;
|
---|
227 | }
|
---|
228 | $dirname .= $num;
|
---|
229 | }
|
---|
230 |
|
---|
231 | $mapping{$index} = $dirname;
|
---|
232 | $dirnames{$dirname} = $index;
|
---|
233 | }
|
---|
234 |
|
---|
235 | return \%mapping;
|
---|
236 | }
|
---|
237 |
|
---|
238 |
|
---|
239 | sub build_index {
|
---|
240 | my $self = shift (@_);
|
---|
241 | my ($index) = @_;
|
---|
242 |
|
---|
243 | # get the full index directory path and make sure it exists
|
---|
244 | my $indexdir = $self->{'index_mapping'}->{$index};
|
---|
245 | &util::mk_all_dir (&util::filename_cat($self->{'build_dir'}, $indexdir));
|
---|
246 | my $fullindexprefix = &util::filename_cat ($self->{'build_dir'}, $indexdir,
|
---|
247 | $self->{'collection'});
|
---|
248 | my $fulltextprefix = &util::filename_cat ($self->{'build_dir'}, "text",
|
---|
249 | $self->{'collection'});
|
---|
250 |
|
---|
251 | # get any os specific stuff
|
---|
252 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
---|
253 | my $exe = &util::get_os_exe ();
|
---|
254 | my $osextra = "";
|
---|
255 | if ($ENV{'GSDLOS'} =~ /^windows$/i) {
|
---|
256 | $fullindexprefix =~ s/\//\\/g;
|
---|
257 | } else {
|
---|
258 | $osextra = " -d /";
|
---|
259 | }
|
---|
260 |
|
---|
261 | # get the index level from the index description
|
---|
262 | # the index will be level 2 unless we are building a
|
---|
263 | # paragraph level index
|
---|
264 | my $index_level = 2;
|
---|
265 | $index_level = 3 if $index =~ /^paragraph/i;
|
---|
266 |
|
---|
267 | # get the index expression if this index belongs
|
---|
268 | # to a subcollection
|
---|
269 | my $indexexparr = [];
|
---|
270 | my ($level, $fields, $subcollection) = split (":", $index);
|
---|
271 | my (@subcollections) = split /,/, $subcollection;
|
---|
272 |
|
---|
273 | foreach $subcollection (@subcollections) {
|
---|
274 | if (defined ($self->{'collect_cfg'}->{'subcollection'}->{$subcollection})) {
|
---|
275 | push (@$indexexparr, $self->{'collect_cfg'}->{'subcollection'}->{$subcollection});
|
---|
276 | }
|
---|
277 | }
|
---|
278 |
|
---|
279 | # set up the document processor
|
---|
280 | $self->{'buildproc'}->set_output_handle ('mgbuilder::PIPEOUT');
|
---|
281 | $self->{'buildproc'}->set_mode ('text');
|
---|
282 | $self->{'buildproc'}->set_index ($index, $indexexparr);
|
---|
283 |
|
---|
284 |
|
---|
285 | # Build index dictionary. Uses verbatim stem method
|
---|
286 | print STDERR "\n creating index dictionary\n" if ($self->{'verbosity'} >= 1);
|
---|
287 | if (!-e "$exedir/mg_passes$exe" || !open (PIPEOUT,
|
---|
288 | "| $exedir/mg_passes$exe -f $fullindexprefix -b $maxdocsize " .
|
---|
289 | "-$index_level -m 32 -s 0 -G -t 10 -N1 $osextra")) {
|
---|
290 | die "mgbuilder::build_index - couldn't run $exedir/mg_passes$exe\n";
|
---|
291 | }
|
---|
292 | $self->{'buildproc'}->reset();
|
---|
293 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
---|
294 | "", {}, $self->{'buildproc'});
|
---|
295 | close (PIPEOUT);
|
---|
296 |
|
---|
297 | # create the perfect hash function
|
---|
298 | if (!-e "$exedir/mg_perf_hash_build$exe") {
|
---|
299 | die "mgbuilder::build_index - couldn't run $exedir/mg_perf_hash_build$exe\n";
|
---|
300 | }
|
---|
301 | system ("$exedir/mg_perf_hash_build$exe -f $fullindexprefix $osextra");
|
---|
302 |
|
---|
303 | # invert the text
|
---|
304 | print STDERR "\n inverting the text\n" if ($self->{'verbosity'} >= 1);
|
---|
305 | if (!-e "$exedir/mg_passes$exe" || !open (PIPEOUT,
|
---|
306 | "| $exedir/mg_passes$exe -f $fullindexprefix -b $maxdocsize " .
|
---|
307 | "-$index_level -c 3 -G -t 10 -N2 $osextra")) {
|
---|
308 | die "mgbuilder::build_index - couldn't run $exedir/mg_passes$exe\n";
|
---|
309 | }
|
---|
310 | $self->{'buildproc'}->reset();
|
---|
311 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
---|
312 | "", {}, $self->{'buildproc'});
|
---|
313 | close (PIPEOUT);
|
---|
314 |
|
---|
315 | # create the weights file
|
---|
316 | print STDERR "\n create the weights file\n" if ($self->{'verbosity'} >= 1);
|
---|
317 | if (!-e "$exedir/mg_weights_build$exe") {
|
---|
318 | die "mgbuilder::build_index - couldn't run $exedir/mg_weights_build$exe\n";
|
---|
319 | }
|
---|
320 | system ("$exedir/mg_weights_build$exe -f $fullindexprefix -t $fulltextprefix $osextra");
|
---|
321 |
|
---|
322 | # create 'on-disk' stemmed dictionary
|
---|
323 | print STDERR "\n creating 'on-disk' stemmed dictionary\n" if ($self->{'verbosity'} >= 1);
|
---|
324 | if (!-e "$exedir/mg_invf_dict$exe") {
|
---|
325 | die "mgbuilder::build_index - couldn't run $exedir/mg_invf_dict$exe\n";
|
---|
326 | }
|
---|
327 | system ("$exedir/mg_invf_dict$exe -f $fullindexprefix $osextra");
|
---|
328 |
|
---|
329 |
|
---|
330 | # creates stem index files for the various stemming methods
|
---|
331 | print STDERR "\n creating stem indexes\n" if ($self->{'verbosity'} >= 1);
|
---|
332 | if (!-e "$exedir/mg_stem_idx$exe") {
|
---|
333 | die "mgbuilder::build_index - couldn't run $exedir/mg_stem_idx$exe\n";
|
---|
334 | }
|
---|
335 | system ("$exedir/mg_stem_idx$exe -b 4096 -s1 -f $fullindexprefix $osextra");
|
---|
336 | system ("$exedir/mg_stem_idx$exe -b 4096 -s2 -f $fullindexprefix $osextra");
|
---|
337 | system ("$exedir/mg_stem_idx$exe -b 4096 -s3 -f $fullindexprefix $osextra");
|
---|
338 |
|
---|
339 |
|
---|
340 | # remove unwanted files
|
---|
341 | my $tmpdir = &util::filename_cat ($self->{'build_dir'}, $indexdir);
|
---|
342 | opendir (DIR, $tmpdir) || die
|
---|
343 | "mgbuilder::build_index - couldn't read directory $tmpdir\n";
|
---|
344 | foreach $file (readdir(DIR)) {
|
---|
345 | next if $file =~ /^\./;
|
---|
346 | my ($suffix) = $file =~ /\.([^\.]+)$/;
|
---|
347 | if (defined $suffix && !defined $wanted_index_files{$suffix}) {
|
---|
348 | # delete it!
|
---|
349 | # print STDERR "deleting $file\n";
|
---|
350 | &util::rm (&util::filename_cat ($tmpdir, $file));
|
---|
351 | }
|
---|
352 | }
|
---|
353 | closedir (DIR);
|
---|
354 | }
|
---|
355 |
|
---|
356 | sub make_infodatabase {
|
---|
357 | my $self = shift (@_);
|
---|
358 | my $textdir = &util::filename_cat($self->{'build_dir'}, "text");
|
---|
359 | &util::mk_all_dir ($textdir);
|
---|
360 |
|
---|
361 | # assume little-endian for now :-)
|
---|
362 | my $fulldbname = &util::filename_cat ($textdir, "$self->{'collection'}.ldb");
|
---|
363 | $fulldbname =~ s/\//\\/g if ($ENV{'GSDLOS'} =~ /^windows$/i);
|
---|
364 | my $exedir = "$ENV{'GSDLHOME'}/bin/$ENV{'GSDLOS'}";
|
---|
365 | my $exe = &util::get_os_exe ();
|
---|
366 |
|
---|
367 | print STDERR "\n*** creating the info database\n" if ($self->{'verbosity'} >= 1);
|
---|
368 |
|
---|
369 | # set up the document processor
|
---|
370 | $self->{'buildproc'}->set_output_handle ('mgbuilder::PIPEOUT');
|
---|
371 | $self->{'buildproc'}->set_mode ('infodb');
|
---|
372 |
|
---|
373 | # collect the statistics for the text
|
---|
374 | # -b $maxdocsize sets the maximum document size to be 12 meg
|
---|
375 | if (!-e "$exedir/txt2db$exe" || !open (PIPEOUT,
|
---|
376 | "| $exedir/txt2db$exe $fulldbname")) {
|
---|
377 | die "mgbuilder::make_infodatabase - couldn't run $exedir/txt2db$exe\n";
|
---|
378 | }
|
---|
379 | $self->{'buildproc'}->reset();
|
---|
380 | &plugin::read ($self->{'pluginfo'}, $self->{'source_dir'},
|
---|
381 | "", {}, $self->{'buildproc'});
|
---|
382 | close (PIPEOUT);
|
---|
383 | }
|
---|
384 |
|
---|
385 | sub make_auxiliary_files {
|
---|
386 | my $self = shift (@_);
|
---|
387 | my ($index);
|
---|
388 | my %build_cfg = ();
|
---|
389 |
|
---|
390 | print STDERR "\n*** creating auxiliary files \n" if ($self->{'verbosity'} >= 1);
|
---|
391 |
|
---|
392 | # get the text directory
|
---|
393 | &util::mk_all_dir ($self->{'build_dir'});
|
---|
394 |
|
---|
395 | # store the build date
|
---|
396 | $build_cfg->{'builddate'} = time;
|
---|
397 |
|
---|
398 | # store the number of documents and number of bytes
|
---|
399 | $build_cfg->{'numdocs'} = $self->{'buildproc'}->get_num_docs();
|
---|
400 | $build_cfg->{'numbytes'} = $self->{'buildproc'}->get_num_bytes();
|
---|
401 |
|
---|
402 | # store the mapping between the index names and the directory names
|
---|
403 | my @indexmap = ();
|
---|
404 | foreach $index (@{$self->{'collect_cfg'}->{'indexes'}}) {
|
---|
405 | push (@indexmap, "$index\-\>$self->{'index_mapping'}->{$index}");
|
---|
406 | }
|
---|
407 | $build_cfg->{'indexmap'} = \@indexmap;
|
---|
408 |
|
---|
409 | # write out the build information
|
---|
410 | &cfgread::write_cfg_file("$self->{'build_dir'}/build.cfg", $build_cfg,
|
---|
411 | '^(builddate|numdocs|numbytes)$', '^(indexmap)$');
|
---|
412 |
|
---|
413 | }
|
---|
414 |
|
---|
415 | sub deinit {
|
---|
416 | my $self = shift (@_);
|
---|
417 | }
|
---|
418 |
|
---|
419 |
|
---|
420 | 1;
|
---|
421 |
|
---|
422 |
|
---|