Line # Revision Author
1 1 dpavlin #!/usr/bin/perl -w
2
3 348 dpavlin =head1 NAME
4
5 354 dpavlin all2all.pl - basic script for all WebPAC needs
6 348 dpavlin
7 =cut
8
9 1 dpavlin use strict;
10 490 dpavlin use locale;
11 454 dpavlin use YAML;
12 352 dpavlin use Carp;
13 437 dpavlin use Getopt::Long;
14 546 dpavlin use Text::Unaccent 1.02;
15 1 dpavlin
16 352 dpavlin use lib './lib';
17 354 dpavlin use WebPAC;
18 708 dpavlin #use WebPAC::jsFind;
19 use WebPAC::HyperEstraier;
20 410 dpavlin use WebPAC::Index;
21 441 dpavlin use WebPAC::Tree;
22 348 dpavlin
23 437 dpavlin # options which can be changed via command line
24 #
25 my $code_page = 'ISO-8859-2';
26 my ($limit_mfn, $start_mfn, $debug, $low_mem);
27 708 dpavlin #my $index_path = './out/index';
28 my $index_path = './out/casket';
29 437 dpavlin
30 my $result = GetOptions(
31 "code_page=s" => \$code_page,
32 "limit_mfn=i" => \$limit_mfn,
33 "start_mfn=i" => \$start_mfn,
34 "debug!" => \$debug,
35 "low_mem!" => \$low_mem,
36 );
37
38 500 dpavlin my $filter = {
39 'CROVOC' => sub {
40 my $tmp = shift || return;
41 566 dpavlin $tmp =~ s/\s*HIDRA.*//;
42 560 dpavlin return undef unless ($tmp =~ s/\s*CROVOC.*$/ #C#/);
43 # remove repeating stars
44 # FIXME this should be fixed at right place, not here!
45 $tmp =~ s/(\s*#C#)+/ #C#/g;
46 500 dpavlin return $tmp;
47 },
48 501 dpavlin 'CROVOC_tree' => sub {
49 my $tmp = shift || return;
50 561 dpavlin $tmp =~ s/\s*CROVOC.*$/ <img src="img\/crovoc.png" border="0">/;
51 503 dpavlin $tmp =~ s/\s*EUROVOC.*//;
52 566 dpavlin $tmp =~ s/\s*HIDRA.*//;
53 501 dpavlin return $tmp;
54 },
55 560 dpavlin # TT filter
56 'CROVOC_img' => sub {
57 my $tmp = shift;
58 $tmp =~ s/\s*#C#\s*/ <img src="..\/img\/crovoc.png" border="0">/gis;
59 562 dpavlin $tmp =~ s/"img\/crovoc.png"/"..\/img\/crovoc.png"/gis;
60 560 dpavlin return $tmp;
61 }
62 500 dpavlin };
63
64 546 dpavlin ## remove accented characters
65 #
66 sub unac {
67 my $string = shift || return;
68 $string = unac_string($code_page,$string);
69 $string =~ tr/ðÐ/dD/;
70 550 dpavlin $string = unac_entities($string);
71 546 dpavlin return $string;
72 }
73 sub unac_2 {
74 my $string = shift || return;
75 if (length($string) > 2) {
76 my $pr = substr($string,0,2);
77 $string = unac_string($code_page,substr($string,2));
78 $string =~ tr/ðÐ/dD/;
79 $string = $pr . $string;
80 }
81 550 dpavlin $string = unac_entities($string);
82 546 dpavlin return lc($string);
83 }
84
85 550 dpavlin sub unac_entities {
86 my $ent = shift || return;
87
88 $ent =~ s/&(\w)(acute|cedil|circ|grave|ring|slash|tilde|uml);/$1/gi;
89 $ent =~ s/&eth;/d/g;
90 $ent =~ s/&E[tT][hH];/D/g;
91 $ent =~ s/&(\w\w)lig;/$1/gi;
92
93 return $ent;
94 }
95
96 437 dpavlin # create WebPAC object
97 #
98 354 dpavlin my $webpac = new WebPAC(
99 437 dpavlin code_page => $code_page,
100 limit_mfn => $limit_mfn,
101 start_mfn => $start_mfn,
102 debug => $debug,
103 low_mem => $low_mem,
104 500 dpavlin filter => $filter,
105 352 dpavlin ) || die;
106
107 421 dpavlin my $log = $webpac->_get_logger() || die "can't get logger";
108 373 dpavlin
109 708 dpavlin #$log->debug("creating WebPAC::jsFind object");
110 #
111 #my $index = new WebPAC::jsFind(
112 # index_path => $index_path,
113 # keys => 62,
114 #) || die;
115 421 dpavlin
116 708 dpavlin $log->debug("creating HyperEstraier object");
117 my $index = new WebPAC::HyperEstraier( index_path => $index_path ) || die;
118 390 dpavlin
119 410 dpavlin my $thes;
120
121 10 dpavlin $|=1;
122 9 dpavlin
123 352 dpavlin my $maxmfn = $webpac->open_isis(
124 357 dpavlin filename => shift @ARGV || '/data/hidra/THS/THS',
125 352 dpavlin lookup => [
126 501 dpavlin { 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' },
127 352 dpavlin # { 'eval' => '"v901^a" eq "Podruèje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
128 # { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
129 # { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
130 { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
131 374 dpavlin { 'key' => '900_mfn:v900', 'val' => 'v000' },
132 419 dpavlin # tree structure
133 471 dpavlin { 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' },
134 419 dpavlin { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' },
135 460 dpavlin { 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' },
136 591 dpavlin { 'key' => 'crovoc:v900', 'val' => 'filter{CROVOC}v800' },
137 352 dpavlin ],
138 54 dpavlin );
139 3 dpavlin
140 421 dpavlin $log->debug("isis file ",$webpac->{'isis_filename'}," opened");
141
142 372 dpavlin $log->info("rows: $maxmfn");
143 170 dpavlin
144 363 dpavlin $webpac->open_import_xml(type => 'isis_hidra_ths');
145
146 425 dpavlin if(1) { # XXX
147 424 dpavlin
148 362 dpavlin while (my $rec = $webpac->fetch_rec) {
149 358 dpavlin
150 370 dpavlin my @ds = $webpac->data_structure($rec);
151 366 dpavlin
152 411 dpavlin if (0 && $log->is_debug) {
153 454 dpavlin $log->debug("rec = ",Dump($rec));
154 $log->debug("ds = ",Dump(\@ds));
155 372 dpavlin }
156
157 374 dpavlin next if (! @ds);
158 366 dpavlin
159 421 dpavlin my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
160 374 dpavlin
161 if ($filename) {
162 411 dpavlin $webpac->output_file(
163 file => $filename,
164 374 dpavlin template => 'html.tt',
165 data => \@ds,
166 398 dpavlin headline => $webpac->{'headline'},
167 374 dpavlin );
168 } else {
169 print $webpac->output(
170 template => 'text.tt',
171 data => \@ds,
172 398 dpavlin headline => $webpac->{'headline'},
173 374 dpavlin );
174 }
175
176 398 dpavlin my $headline = $webpac->{'headline'};
177 390 dpavlin
178 392 dpavlin my $f = $filename;
179 $f =~ s!out/!!;
180
181 390 dpavlin # save into index
182 foreach my $ds (@ds) {
183 next if (! $ds->{'swish'});
184
185 466 dpavlin # strip all non word characters from beginning or end
186 # of word
187 587 dpavlin my $words = unac(join(" ",@{$ds->{'swish'}}));
188 466 dpavlin $words =~ s/^\W+//;
189 $words =~ s/\W*\s+\W*/ /g;
190 $words =~ s/\W+$//;
191
192 504 dpavlin # first try to generate headline for this entry from index
193 520 dpavlin my $h = $ds->{'index'}->[0];
194 504 dpavlin # then, from display
195 520 dpavlin $h ||= $ds->{'display'}->[0];
196 504 dpavlin # and as last resport, fallback to headline
197 $h ||= $headline;
198
199 390 dpavlin $index->insert(
200 index_name => $ds->{'tag'},
201 466 dpavlin #path => $f,
202 path => $webpac->mfn,
203 504 dpavlin headline => $h,
204 587 dpavlin words => $words,
205 390 dpavlin );
206 }
207
208 410 dpavlin # save into sorted index (thesaurus)
209 foreach my $ds (@ds) {
210 next if (! $ds->{'index'});
211
212 511 dpavlin $thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} );
213 410 dpavlin
214 415 dpavlin foreach my $h (@{$ds->{'index'}}) {
215 $thes->{$ds->{'tag'}}->insert(
216 546 dpavlin sort_by => unac_2($h),
217 mfn => $webpac->mfn,
218 415 dpavlin headline => $h,
219 );
220 }
221 410 dpavlin }
222
223 454 dpavlin # print Dump(\@ds);
224 390 dpavlin
225 3 dpavlin }
226
227 411 dpavlin foreach my $t (keys %{$thes}) {
228
229 my @e = $thes->{$t}->elements;
230 if (! @e) {
231 $log->logwarn("no elements in sorted index $t?");
232 next;
233 }
234
235 415 dpavlin my $file = "./out/bfilter/$t.txt";
236 $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
237 411 dpavlin
238 $webpac->output_file(
239 415 dpavlin file => $file,
240 411 dpavlin template => 'index.tt',
241 data => \@e,
242 index_name => $t,
243 );
244 }
245
246 if (0 && $log->is_debug) {
247 454 dpavlin $log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
248 $log->debug("data hash: ",Dump($webpac->{'data'}));
249 410 dpavlin foreach my $t (keys %{$thes}) {
250 454 dpavlin $log->debug("thesaurus $t hash: ",Dump($thes->{$t}));
251 410 dpavlin }
252 372 dpavlin }
253 390 dpavlin
254 424 dpavlin } # XXX if(0)
255
256 530 dpavlin #$log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
257 420 dpavlin
258 441 dpavlin $log->info("creating tree");
259
260 419 dpavlin #
261 441 dpavlin # define tree structure
262 419 dpavlin #
263
264 424 dpavlin my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup");
265 419 dpavlin
266 424 dpavlin my @tree = ({
267 # level 0
268 code_arr => sub { sort keys %{$l} },
269 filter_code => sub {
270 my $t = shift;
271 return $t if ($t =~ s/root://);
272 },
273 lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} },
274 lookup_term => sub { shift @{$l->{"d:".$_[1]}} },
275 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} },
276 454 dpavlin have_children => sub { return $l->{"a:".$_[0]."::"} },
277 have_children_at_level => sub {
278 471 dpavlin return unless (defined($l->{"code:".$_[1]}));
279 my $code = shift @{$l->{"code:".$_[1]}};
280 print STDERR "## $_[1] -> $code\n";
281 return undef unless($code);
282 return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"}));
283 454 dpavlin },
284 424 dpavlin style => 'display: none',
285 579 dpavlin iframe => 1,
286 424 dpavlin },{
287 # 1
288 454 dpavlin code_arr => sub { @{$_[0]} },
289 424 dpavlin filter_code => sub { shift }, # nop
290 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
291 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
292 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
293 454 dpavlin have_children => sub { return $l->{"a:".$_[1].":"} },
294 424 dpavlin style => 'display: none',
295 },{
296 # 2
297 454 dpavlin code_arr => sub { @{$_[0]} },
298 424 dpavlin filter_code => sub { shift },
299 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
300 424 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
301 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
302 460 dpavlin have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
303 572 dpavlin style => 'display: none',
304 425 dpavlin },{
305 # 3 u¾i pojam
306 454 dpavlin code_arr => sub { @{$_[0]} },
307 425 dpavlin filter_code => sub { shift },
308 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
309 426 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
310 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
311 460 dpavlin have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
312 426 dpavlin },{
313 # 4
314 454 dpavlin code_arr => sub { @{$_[0]} },
315 426 dpavlin filter_code => sub { shift },
316 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
317 426 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
318 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
319 460 dpavlin have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
320 427 dpavlin },{
321 # 5
322 454 dpavlin code_arr => sub { @{$_[0]} },
323 427 dpavlin filter_code => sub { shift },
324 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
325 427 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
326 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
327 460 dpavlin have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
328 427 dpavlin },{
329 # 6
330 454 dpavlin code_arr => sub { @{$_[0]} },
331 427 dpavlin filter_code => sub { shift },
332 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
333 427 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
334 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
335 460 dpavlin have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
336 427 dpavlin },{
337 # 7
338 454 dpavlin code_arr => sub { @{$_[0]} },
339 427 dpavlin filter_code => sub { shift },
340 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
341 427 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
342 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
343 460 dpavlin have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
344 427 dpavlin },{
345 # 8
346 454 dpavlin code_arr => sub { @{$_[0]} },
347 427 dpavlin filter_code => sub { shift },
348 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
349 427 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
350 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
351 460 dpavlin # have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
352 424 dpavlin have_children => sub { 0 },
353 454 dpavlin },{
354 # 9 - level which is never reached except explicitly
355 471 dpavlin code_arr => sub { @{$_[0]} },
356 454 dpavlin filter_code => sub { shift },
357 460 dpavlin lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
358 454 dpavlin lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
359 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
360 have_children => sub { 0 },
361 471 dpavlin have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) },
362 454 dpavlin },{
363 424 dpavlin });
364 419 dpavlin
365 441 dpavlin my $tree = new WebPAC::Tree(
366 530 dpavlin tree => \@tree,
367 573 dpavlin detail_url => sub {
368 my $mfn = shift;
369 574 dpavlin my $url = "thes/${mfn}.html";
370 return $url if (-e "./out/$url");
371 573 dpavlin },
372 575 dpavlin nodes_dir => 'nodes',
373 530 dpavlin );
374
375 $tree->output(
376 dir => './out',
377 html => 'browse.html',
378 572 dpavlin template_dir => './output_template/',
379 template_tree => 'tree.tt',
380 template_node => 'node.tt',
381 441 dpavlin js => 'tree-ids.js',
382 );
383 428 dpavlin
384 530 dpavlin $tree->output(
385 dir => './eurovoc',
386 html => 'hijerarhija.html',
387 572 dpavlin template_dir => './output_template/',
388 template_tree => 'hijerarhija.tt',
389 template_node => 'hijerarhija_node.tt',
390 530 dpavlin js => 'tree-ids.js',
391 );
392
393
394 428 dpavlin $log->info("closing index");
395 390 dpavlin $index->close;
396
397 422 dpavlin $log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'}));