/[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 549 - (hide annotations)
Tue Oct 26 20:53:03 2004 UTC (14 years, 11 months ago) by dpavlin
File MIME type: text/plain
File size: 9301 byte(s)
Magic regexp to unaccent html entities in input data.

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     return $string;
57     }
58     sub unac_2 {
59     my $string = shift || return;
60     if (length($string) > 2) {
61     my $pr = substr($string,0,2);
62     $string = unac_string($code_page,substr($string,2));
63     $string =~ tr/ðÐ/dD/;
64     $string = $pr . $string;
65     }
66 dpavlin 549 $string =~ s/&(\w)(acute|cedil|circ|grave|ring|slash|tilde|uml);/$1/gi;
67     $string =~ s/ð/d/g;
68     $string =~ s/&E[tT][hH];/D/g;
69     $string =~ s/&(\w\w)lig;/$1/gi;
70 dpavlin 546 return lc($string);
71     }
72    
73 dpavlin 437 # create WebPAC object
74     #
75 dpavlin 354 my $webpac = new WebPAC(
76 dpavlin 437 code_page => $code_page,
77     limit_mfn => $limit_mfn,
78     start_mfn => $start_mfn,
79     debug => $debug,
80     low_mem => $low_mem,
81 dpavlin 500 filter => $filter,
82 dpavlin 352 ) || die;
83    
84 dpavlin 421 my $log = $webpac->_get_logger() || die "can't get logger";
85 dpavlin 373
86 dpavlin 421 $log->debug("creating WebPAC::jsFind object");
87    
88 dpavlin 390 my $index = new WebPAC::jsFind(
89 dpavlin 437 index_path => $index_path,
90 dpavlin 413 keys => 10,
91 dpavlin 390 ) || die;
92    
93 dpavlin 410 my $thes;
94    
95 dpavlin 10 $|=1;
96 dpavlin 9
97 dpavlin 352 my $maxmfn = $webpac->open_isis(
98 dpavlin 357 filename => shift @ARGV || '/data/hidra/THS/THS',
99 dpavlin 352 lookup => [
100 dpavlin 501 { 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' },
101 dpavlin 352 # { 'eval' => '"v901^a" eq "Podruèje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
102     # { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
103     # { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
104     { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
105 dpavlin 374 { 'key' => '900_mfn:v900', 'val' => 'v000' },
106 dpavlin 419 # tree structure
107 dpavlin 471 { 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' },
108 dpavlin 419 { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' },
109 dpavlin 460 { 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' },
110 dpavlin 352 ],
111 dpavlin 54 );
112 dpavlin 3
113 dpavlin 421 $log->debug("isis file ",$webpac->{'isis_filename'}," opened");
114    
115 dpavlin 372 $log->info("rows: $maxmfn");
116 dpavlin 170
117 dpavlin 363 $webpac->open_import_xml(type => 'isis_hidra_ths');
118    
119 dpavlin 425 if(1) { # XXX
120 dpavlin 424
121 dpavlin 362 while (my $rec = $webpac->fetch_rec) {
122 dpavlin 358
123 dpavlin 370 my @ds = $webpac->data_structure($rec);
124 dpavlin 366
125 dpavlin 411 if (0 && $log->is_debug) {
126 dpavlin 454 $log->debug("rec = ",Dump($rec));
127     $log->debug("ds = ",Dump(\@ds));
128 dpavlin 372 }
129    
130 dpavlin 374 next if (! @ds);
131 dpavlin 366
132 dpavlin 421 my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
133 dpavlin 374
134     if ($filename) {
135 dpavlin 411 $webpac->output_file(
136     file => $filename,
137 dpavlin 374 template => 'html.tt',
138     data => \@ds,
139 dpavlin 398 headline => $webpac->{'headline'},
140 dpavlin 374 );
141     } else {
142     print $webpac->output(
143     template => 'text.tt',
144     data => \@ds,
145 dpavlin 398 headline => $webpac->{'headline'},
146 dpavlin 374 );
147     }
148    
149 dpavlin 398 my $headline = $webpac->{'headline'};
150 dpavlin 390
151 dpavlin 392 my $f = $filename;
152     $f =~ s!out/!!;
153    
154 dpavlin 390 # save into index
155     foreach my $ds (@ds) {
156     next if (! $ds->{'swish'});
157    
158 dpavlin 466 # strip all non word characters from beginning or end
159     # of word
160     my $words = join(" ",@{$ds->{'swish'}});
161     $words =~ s/^\W+//;
162     $words =~ s/\W*\s+\W*/ /g;
163     $words =~ s/\W+$//;
164    
165 dpavlin 504 # first try to generate headline for this entry from index
166 dpavlin 520 my $h = $ds->{'index'}->[0];
167 dpavlin 504 # then, from display
168 dpavlin 520 $h ||= $ds->{'display'}->[0];
169 dpavlin 504 # and as last resport, fallback to headline
170     $h ||= $headline;
171    
172 dpavlin 390 $index->insert(
173     index_name => $ds->{'tag'},
174 dpavlin 466 #path => $f,
175     path => $webpac->mfn,
176 dpavlin 504 headline => $h,
177 dpavlin 546 words => unac($words),
178 dpavlin 390 );
179     }
180    
181 dpavlin 410 # save into sorted index (thesaurus)
182     foreach my $ds (@ds) {
183     next if (! $ds->{'index'});
184    
185 dpavlin 511 $thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} );
186 dpavlin 410
187 dpavlin 415 foreach my $h (@{$ds->{'index'}}) {
188     $thes->{$ds->{'tag'}}->insert(
189 dpavlin 546 sort_by => unac_2($h),
190     mfn => $webpac->mfn,
191 dpavlin 415 headline => $h,
192     );
193     }
194 dpavlin 410 }
195    
196 dpavlin 454 # print Dump(\@ds);
197 dpavlin 390
198 dpavlin 3 }
199    
200 dpavlin 411 foreach my $t (keys %{$thes}) {
201    
202     my @e = $thes->{$t}->elements;
203     if (! @e) {
204     $log->logwarn("no elements in sorted index $t?");
205     next;
206     }
207    
208 dpavlin 415 my $file = "./out/bfilter/$t.txt";
209     $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
210 dpavlin 411
211     $webpac->output_file(
212 dpavlin 415 file => $file,
213 dpavlin 411 template => 'index.tt',
214     data => \@e,
215     index_name => $t,
216     );
217     }
218    
219     if (0 && $log->is_debug) {
220 dpavlin 454 $log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
221     $log->debug("data hash: ",Dump($webpac->{'data'}));
222 dpavlin 410 foreach my $t (keys %{$thes}) {
223 dpavlin 454 $log->debug("thesaurus $t hash: ",Dump($thes->{$t}));
224 dpavlin 410 }
225 dpavlin 372 }
226 dpavlin 390
227 dpavlin 424 } # XXX if(0)
228    
229 dpavlin 530 #$log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
230 dpavlin 420
231 dpavlin 441 $log->info("creating tree");
232    
233 dpavlin 419 #
234 dpavlin 441 # define tree structure
235 dpavlin 419 #
236    
237 dpavlin 424 my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup");
238 dpavlin 419
239 dpavlin 424 my @tree = ({
240     # level 0
241     code_arr => sub { sort keys %{$l} },
242     filter_code => sub {
243     my $t = shift;
244     return $t if ($t =~ s/root://);
245     },
246     lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} },
247     lookup_term => sub { shift @{$l->{"d:".$_[1]}} },
248     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} },
249 dpavlin 454 have_children => sub { return $l->{"a:".$_[0]."::"} },
250     have_children_at_level => sub {
251 dpavlin 471 return unless (defined($l->{"code:".$_[1]}));
252     my $code = shift @{$l->{"code:".$_[1]}};
253     print STDERR "## $_[1] -> $code\n";
254     return undef unless($code);
255     return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"}));
256 dpavlin 454 },
257 dpavlin 424 style => 'display: none',
258     },{
259     # 1
260 dpavlin 454 code_arr => sub { @{$_[0]} },
261 dpavlin 424 filter_code => sub { shift }, # nop
262     lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
263     lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
264     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
265 dpavlin 454 have_children => sub { return $l->{"a:".$_[1].":"} },
266 dpavlin 424 style => 'display: none',
267     },{
268     # 2
269 dpavlin 454 code_arr => sub { @{$_[0]} },
270 dpavlin 424 filter_code => sub { shift },
271 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
272 dpavlin 424 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
273     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
274 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
275 dpavlin 454 #style => 'display: none',
276 dpavlin 425 },{
277     # 3 u¾i pojam
278 dpavlin 454 code_arr => sub { @{$_[0]} },
279 dpavlin 425 filter_code => sub { shift },
280 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
281 dpavlin 426 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].":".$_[0]} },
284 dpavlin 426 },{
285     # 4
286 dpavlin 454 code_arr => sub { @{$_[0]} },
287 dpavlin 426 filter_code => sub { shift },
288 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
289 dpavlin 426 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
290     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
291 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
292 dpavlin 427 },{
293     # 5
294 dpavlin 454 code_arr => sub { @{$_[0]} },
295 dpavlin 427 filter_code => sub { shift },
296 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
297 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
298     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
299 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
300 dpavlin 427 },{
301     # 6
302 dpavlin 454 code_arr => sub { @{$_[0]} },
303 dpavlin 427 filter_code => sub { shift },
304 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
305 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
306     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
307 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
308 dpavlin 427 },{
309     # 7
310 dpavlin 454 code_arr => sub { @{$_[0]} },
311 dpavlin 427 filter_code => sub { shift },
312 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
313 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
314     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
315 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
316 dpavlin 427 },{
317     # 8
318 dpavlin 454 code_arr => sub { @{$_[0]} },
319 dpavlin 427 filter_code => sub { shift },
320 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
321 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
322     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
323 dpavlin 460 # have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
324 dpavlin 424 have_children => sub { 0 },
325 dpavlin 454 },{
326     # 9 - level which is never reached except explicitly
327 dpavlin 471 code_arr => sub { @{$_[0]} },
328 dpavlin 454 filter_code => sub { shift },
329 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
330 dpavlin 454 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
331     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
332     have_children => sub { 0 },
333 dpavlin 471 have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) },
334 dpavlin 454 },{
335 dpavlin 424 });
336 dpavlin 419
337 dpavlin 441 my $tree = new WebPAC::Tree(
338 dpavlin 530 tree => \@tree,
339     );
340    
341     $tree->output(
342     dir => './out',
343     html => 'browse.html',
344 dpavlin 441 template => './output_template/tree.tt',
345     js => 'tree-ids.js',
346     );
347 dpavlin 428
348 dpavlin 530 $tree->output(
349     dir => './eurovoc',
350     html => 'hijerarhija.html',
351     template => './output_template/hijerarhija.tt',
352     js => 'tree-ids.js',
353     );
354    
355    
356 dpavlin 428 $log->info("closing index");
357 dpavlin 390 $index->close;
358    
359 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