/[webpac]/trunk2/all2all.pl
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Annotation of /trunk2/all2all.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 550 - (hide annotations)
Tue Oct 26 21:44:57 2004 UTC (19 years, 5 months ago) by dpavlin
File MIME type: text/plain
File size: 9426 byte(s)
unaccent html entities from index. This fixes searching for words without
accented characters which are stored as entities in input data.
Unfortunately, searhing via copy/paste of accented result doesn't work in
Firefox (it produces UTF-8 encoded char which isn't handled by l2_unescape).

1 dpavlin 1 #!/usr/bin/perl -w
2    
3 dpavlin 348 =head1 NAME
4    
5 dpavlin 354 all2all.pl - basic script for all WebPAC needs
6 dpavlin 348
7     =cut
8    
9 dpavlin 1 use strict;
10 dpavlin 490 use locale;
11 dpavlin 454 use YAML;
12 dpavlin 352 use Carp;
13 dpavlin 437 use Getopt::Long;
14 dpavlin 546 use Text::Unaccent 1.02;
15 dpavlin 1
16 dpavlin 352 use lib './lib';
17 dpavlin 354 use WebPAC;
18 dpavlin 390 use WebPAC::jsFind;
19 dpavlin 410 use WebPAC::Index;
20 dpavlin 441 use WebPAC::Tree;
21 dpavlin 348
22 dpavlin 437 # options which can be changed via command line
23     #
24     my $code_page = 'ISO-8859-2';
25     my ($limit_mfn, $start_mfn, $debug, $low_mem);
26     my $index_path = './out/index';
27    
28     my $result = GetOptions(
29     "code_page=s" => \$code_page,
30     "limit_mfn=i" => \$limit_mfn,
31     "start_mfn=i" => \$start_mfn,
32     "debug!" => \$debug,
33     "low_mem!" => \$low_mem,
34     );
35    
36 dpavlin 500 my $filter = {
37     'CROVOC' => sub {
38     my $tmp = shift || return;
39 dpavlin 503 return undef unless ($tmp =~ s/\s*CROVOC.*$/ */);
40 dpavlin 500 return $tmp;
41     },
42 dpavlin 501 'CROVOC_tree' => sub {
43     my $tmp = shift || return;
44 dpavlin 503 $tmp =~ s/\s*CROVOC.*$/ */;
45     $tmp =~ s/\s*EUROVOC.*//;
46 dpavlin 501 return $tmp;
47     },
48 dpavlin 500 };
49    
50 dpavlin 546 ## remove accented characters
51     #
52     sub unac {
53     my $string = shift || return;
54     $string = unac_string($code_page,$string);
55     $string =~ tr/ðÐ/dD/;
56 dpavlin 550 $string = unac_entities($string);
57 dpavlin 546 return $string;
58     }
59     sub unac_2 {
60     my $string = shift || return;
61     if (length($string) > 2) {
62     my $pr = substr($string,0,2);
63     $string = unac_string($code_page,substr($string,2));
64     $string =~ tr/ðÐ/dD/;
65     $string = $pr . $string;
66     }
67 dpavlin 550 $string = unac_entities($string);
68 dpavlin 546 return lc($string);
69     }
70    
71 dpavlin 550 sub unac_entities {
72     my $ent = shift || return;
73    
74     $ent =~ s/&(\w)(acute|cedil|circ|grave|ring|slash|tilde|uml);/$1/gi;
75     $ent =~ s/ð/d/g;
76     $ent =~ s/&E[tT][hH];/D/g;
77     $ent =~ s/&(\w\w)lig;/$1/gi;
78    
79     return $ent;
80     }
81    
82 dpavlin 437 # create WebPAC object
83     #
84 dpavlin 354 my $webpac = new WebPAC(
85 dpavlin 437 code_page => $code_page,
86     limit_mfn => $limit_mfn,
87     start_mfn => $start_mfn,
88     debug => $debug,
89     low_mem => $low_mem,
90 dpavlin 500 filter => $filter,
91 dpavlin 352 ) || die;
92    
93 dpavlin 421 my $log = $webpac->_get_logger() || die "can't get logger";
94 dpavlin 373
95 dpavlin 421 $log->debug("creating WebPAC::jsFind object");
96    
97 dpavlin 390 my $index = new WebPAC::jsFind(
98 dpavlin 437 index_path => $index_path,
99 dpavlin 413 keys => 10,
100 dpavlin 390 ) || die;
101    
102 dpavlin 410 my $thes;
103    
104 dpavlin 10 $|=1;
105 dpavlin 9
106 dpavlin 352 my $maxmfn = $webpac->open_isis(
107 dpavlin 357 filename => shift @ARGV || '/data/hidra/THS/THS',
108 dpavlin 352 lookup => [
109 dpavlin 501 { 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' },
110 dpavlin 352 # { 'eval' => '"v901^a" eq "Podruèje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
111     # { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
112     # { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
113     { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
114 dpavlin 374 { 'key' => '900_mfn:v900', 'val' => 'v000' },
115 dpavlin 419 # tree structure
116 dpavlin 471 { 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' },
117 dpavlin 419 { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' },
118 dpavlin 460 { 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' },
119 dpavlin 352 ],
120 dpavlin 54 );
121 dpavlin 3
122 dpavlin 421 $log->debug("isis file ",$webpac->{'isis_filename'}," opened");
123    
124 dpavlin 372 $log->info("rows: $maxmfn");
125 dpavlin 170
126 dpavlin 363 $webpac->open_import_xml(type => 'isis_hidra_ths');
127    
128 dpavlin 425 if(1) { # XXX
129 dpavlin 424
130 dpavlin 362 while (my $rec = $webpac->fetch_rec) {
131 dpavlin 358
132 dpavlin 370 my @ds = $webpac->data_structure($rec);
133 dpavlin 366
134 dpavlin 411 if (0 && $log->is_debug) {
135 dpavlin 454 $log->debug("rec = ",Dump($rec));
136     $log->debug("ds = ",Dump(\@ds));
137 dpavlin 372 }
138    
139 dpavlin 374 next if (! @ds);
140 dpavlin 366
141 dpavlin 421 my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
142 dpavlin 374
143     if ($filename) {
144 dpavlin 411 $webpac->output_file(
145     file => $filename,
146 dpavlin 374 template => 'html.tt',
147     data => \@ds,
148 dpavlin 398 headline => $webpac->{'headline'},
149 dpavlin 374 );
150     } else {
151     print $webpac->output(
152     template => 'text.tt',
153     data => \@ds,
154 dpavlin 398 headline => $webpac->{'headline'},
155 dpavlin 374 );
156     }
157    
158 dpavlin 398 my $headline = $webpac->{'headline'};
159 dpavlin 390
160 dpavlin 392 my $f = $filename;
161     $f =~ s!out/!!;
162    
163 dpavlin 390 # save into index
164     foreach my $ds (@ds) {
165     next if (! $ds->{'swish'});
166    
167 dpavlin 466 # strip all non word characters from beginning or end
168     # of word
169     my $words = join(" ",@{$ds->{'swish'}});
170     $words =~ s/^\W+//;
171     $words =~ s/\W*\s+\W*/ /g;
172     $words =~ s/\W+$//;
173    
174 dpavlin 504 # first try to generate headline for this entry from index
175 dpavlin 520 my $h = $ds->{'index'}->[0];
176 dpavlin 504 # then, from display
177 dpavlin 520 $h ||= $ds->{'display'}->[0];
178 dpavlin 504 # and as last resport, fallback to headline
179     $h ||= $headline;
180    
181 dpavlin 390 $index->insert(
182     index_name => $ds->{'tag'},
183 dpavlin 466 #path => $f,
184     path => $webpac->mfn,
185 dpavlin 504 headline => $h,
186 dpavlin 546 words => unac($words),
187 dpavlin 390 );
188     }
189    
190 dpavlin 410 # save into sorted index (thesaurus)
191     foreach my $ds (@ds) {
192     next if (! $ds->{'index'});
193    
194 dpavlin 511 $thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} );
195 dpavlin 410
196 dpavlin 415 foreach my $h (@{$ds->{'index'}}) {
197     $thes->{$ds->{'tag'}}->insert(
198 dpavlin 546 sort_by => unac_2($h),
199     mfn => $webpac->mfn,
200 dpavlin 415 headline => $h,
201     );
202     }
203 dpavlin 410 }
204    
205 dpavlin 454 # print Dump(\@ds);
206 dpavlin 390
207 dpavlin 3 }
208    
209 dpavlin 411 foreach my $t (keys %{$thes}) {
210    
211     my @e = $thes->{$t}->elements;
212     if (! @e) {
213     $log->logwarn("no elements in sorted index $t?");
214     next;
215     }
216    
217 dpavlin 415 my $file = "./out/bfilter/$t.txt";
218     $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
219 dpavlin 411
220     $webpac->output_file(
221 dpavlin 415 file => $file,
222 dpavlin 411 template => 'index.tt',
223     data => \@e,
224     index_name => $t,
225     );
226     }
227    
228     if (0 && $log->is_debug) {
229 dpavlin 454 $log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
230     $log->debug("data hash: ",Dump($webpac->{'data'}));
231 dpavlin 410 foreach my $t (keys %{$thes}) {
232 dpavlin 454 $log->debug("thesaurus $t hash: ",Dump($thes->{$t}));
233 dpavlin 410 }
234 dpavlin 372 }
235 dpavlin 390
236 dpavlin 424 } # XXX if(0)
237    
238 dpavlin 530 #$log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
239 dpavlin 420
240 dpavlin 441 $log->info("creating tree");
241    
242 dpavlin 419 #
243 dpavlin 441 # define tree structure
244 dpavlin 419 #
245    
246 dpavlin 424 my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup");
247 dpavlin 419
248 dpavlin 424 my @tree = ({
249     # level 0
250     code_arr => sub { sort keys %{$l} },
251     filter_code => sub {
252     my $t = shift;
253     return $t if ($t =~ s/root://);
254     },
255     lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} },
256     lookup_term => sub { shift @{$l->{"d:".$_[1]}} },
257     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} },
258 dpavlin 454 have_children => sub { return $l->{"a:".$_[0]."::"} },
259     have_children_at_level => sub {
260 dpavlin 471 return unless (defined($l->{"code:".$_[1]}));
261     my $code = shift @{$l->{"code:".$_[1]}};
262     print STDERR "## $_[1] -> $code\n";
263     return undef unless($code);
264     return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"}));
265 dpavlin 454 },
266 dpavlin 424 style => 'display: none',
267     },{
268     # 1
269 dpavlin 454 code_arr => sub { @{$_[0]} },
270 dpavlin 424 filter_code => sub { shift }, # nop
271     lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
272     lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
273     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
274 dpavlin 454 have_children => sub { return $l->{"a:".$_[1].":"} },
275 dpavlin 424 style => 'display: none',
276     },{
277     # 2
278 dpavlin 454 code_arr => sub { @{$_[0]} },
279 dpavlin 424 filter_code => sub { shift },
280 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
281 dpavlin 424 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
282     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
283 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
284 dpavlin 454 #style => 'display: none',
285 dpavlin 425 },{
286     # 3 u¾i pojam
287 dpavlin 454 code_arr => sub { @{$_[0]} },
288 dpavlin 425 filter_code => sub { shift },
289 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
290 dpavlin 426 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
291     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
292 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
293 dpavlin 426 },{
294     # 4
295 dpavlin 454 code_arr => sub { @{$_[0]} },
296 dpavlin 426 filter_code => sub { shift },
297 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
298 dpavlin 426 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
299     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
300 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
301 dpavlin 427 },{
302     # 5
303 dpavlin 454 code_arr => sub { @{$_[0]} },
304 dpavlin 427 filter_code => sub { shift },
305 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
306 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
307     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
308 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
309 dpavlin 427 },{
310     # 6
311 dpavlin 454 code_arr => sub { @{$_[0]} },
312 dpavlin 427 filter_code => sub { shift },
313 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
314 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
315     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
316 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
317 dpavlin 427 },{
318     # 7
319 dpavlin 454 code_arr => sub { @{$_[0]} },
320 dpavlin 427 filter_code => sub { shift },
321 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
322 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
323     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
324 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
325 dpavlin 427 },{
326     # 8
327 dpavlin 454 code_arr => sub { @{$_[0]} },
328 dpavlin 427 filter_code => sub { shift },
329 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
330 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
331     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
332 dpavlin 460 # have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
333 dpavlin 424 have_children => sub { 0 },
334 dpavlin 454 },{
335     # 9 - level which is never reached except explicitly
336 dpavlin 471 code_arr => sub { @{$_[0]} },
337 dpavlin 454 filter_code => sub { shift },
338 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
339 dpavlin 454 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
340     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
341     have_children => sub { 0 },
342 dpavlin 471 have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) },
343 dpavlin 454 },{
344 dpavlin 424 });
345 dpavlin 419
346 dpavlin 441 my $tree = new WebPAC::Tree(
347 dpavlin 530 tree => \@tree,
348     );
349    
350     $tree->output(
351     dir => './out',
352     html => 'browse.html',
353 dpavlin 441 template => './output_template/tree.tt',
354     js => 'tree-ids.js',
355     );
356 dpavlin 428
357 dpavlin 530 $tree->output(
358     dir => './eurovoc',
359     html => 'hijerarhija.html',
360     template => './output_template/hijerarhija.tt',
361     js => 'tree-ids.js',
362     );
363    
364    
365 dpavlin 428 $log->info("closing index");
366 dpavlin 390 $index->close;
367    
368 dpavlin 422 $log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'}));

Properties

Name Value
cvs2svn:cvs-rev 1.64
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26