/[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 708 - (hide annotations)
Wed Jul 13 23:38:04 2005 UTC (14 years, 3 months ago) by dpavlin
File MIME type: text/plain
File size: 10353 byte(s)
added HyperEstraier back-end indexer

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 708 #use WebPAC::jsFind;
19     use WebPAC::HyperEstraier;
20 dpavlin 410 use WebPAC::Index;
21 dpavlin 441 use WebPAC::Tree;
22 dpavlin 348
23 dpavlin 437 # 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 dpavlin 708 #my $index_path = './out/index';
28     my $index_path = './out/casket';
29 dpavlin 437
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 dpavlin 500 my $filter = {
39     'CROVOC' => sub {
40     my $tmp = shift || return;
41 dpavlin 566 $tmp =~ s/\s*HIDRA.*//;
42 dpavlin 560 return undef unless ($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 dpavlin 500 return $tmp;
47     },
48 dpavlin 501 'CROVOC_tree' => sub {
49     my $tmp = shift || return;
50 dpavlin 561 $tmp =~ s/\s*CROVOC.*$/ <img src="img\/crovoc.png" border="0">/;
51 dpavlin 503 $tmp =~ s/\s*EUROVOC.*//;
52 dpavlin 566 $tmp =~ s/\s*HIDRA.*//;
53 dpavlin 501 return $tmp;
54     },
55 dpavlin 560 # 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 dpavlin 562 $tmp =~ s/"img\/crovoc.png"/"..\/img\/crovoc.png"/gis;
60 dpavlin 560 return $tmp;
61     }
62 dpavlin 500 };
63    
64 dpavlin 546 ## 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 dpavlin 550 $string = unac_entities($string);
71 dpavlin 546 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 dpavlin 550 $string = unac_entities($string);
82 dpavlin 546 return lc($string);
83     }
84    
85 dpavlin 550 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 dpavlin 437 # create WebPAC object
97     #
98 dpavlin 354 my $webpac = new WebPAC(
99 dpavlin 437 code_page => $code_page,
100     limit_mfn => $limit_mfn,
101     start_mfn => $start_mfn,
102     debug => $debug,
103     low_mem => $low_mem,
104 dpavlin 500 filter => $filter,
105 dpavlin 352 ) || die;
106    
107 dpavlin 421 my $log = $webpac->_get_logger() || die "can't get logger";
108 dpavlin 373
109 dpavlin 708 #$log->debug("creating WebPAC::jsFind object");
110     #
111     #my $index = new WebPAC::jsFind(
112     # index_path => $index_path,
113     # keys => 62,
114     #) || die;
115 dpavlin 421
116 dpavlin 708 $log->debug("creating HyperEstraier object");
117     my $index = new WebPAC::HyperEstraier( index_path => $index_path ) || die;
118 dpavlin 390
119 dpavlin 410 my $thes;
120    
121 dpavlin 10 $|=1;
122 dpavlin 9
123 dpavlin 352 my $maxmfn = $webpac->open_isis(
124 dpavlin 357 filename => shift @ARGV || '/data/hidra/THS/THS',
125 dpavlin 352 lookup => [
126 dpavlin 501 { 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' },
127 dpavlin 352 # { 'eval' => '"v901^a" eq "Podruèje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
128     # { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
129     # { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
130     { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
131 dpavlin 374 { 'key' => '900_mfn:v900', 'val' => 'v000' },
132 dpavlin 419 # tree structure
133 dpavlin 471 { 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' },
134 dpavlin 419 { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' },
135 dpavlin 460 { 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' },
136 dpavlin 591 { 'key' => 'crovoc:v900', 'val' => 'filter{CROVOC}v800' },
137 dpavlin 352 ],
138 dpavlin 54 );
139 dpavlin 3
140 dpavlin 421 $log->debug("isis file ",$webpac->{'isis_filename'}," opened");
141    
142 dpavlin 372 $log->info("rows: $maxmfn");
143 dpavlin 170
144 dpavlin 363 $webpac->open_import_xml(type => 'isis_hidra_ths');
145    
146 dpavlin 425 if(1) { # XXX
147 dpavlin 424
148 dpavlin 362 while (my $rec = $webpac->fetch_rec) {
149 dpavlin 358
150 dpavlin 370 my @ds = $webpac->data_structure($rec);
151 dpavlin 366
152 dpavlin 411 if (0 && $log->is_debug) {
153 dpavlin 454 $log->debug("rec = ",Dump($rec));
154     $log->debug("ds = ",Dump(\@ds));
155 dpavlin 372 }
156    
157 dpavlin 374 next if (! @ds);
158 dpavlin 366
159 dpavlin 421 my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
160 dpavlin 374
161     if ($filename) {
162 dpavlin 411 $webpac->output_file(
163     file => $filename,
164 dpavlin 374 template => 'html.tt',
165     data => \@ds,
166 dpavlin 398 headline => $webpac->{'headline'},
167 dpavlin 374 );
168     } else {
169     print $webpac->output(
170     template => 'text.tt',
171     data => \@ds,
172 dpavlin 398 headline => $webpac->{'headline'},
173 dpavlin 374 );
174     }
175    
176 dpavlin 398 my $headline = $webpac->{'headline'};
177 dpavlin 390
178 dpavlin 392 my $f = $filename;
179     $f =~ s!out/!!;
180    
181 dpavlin 390 # save into index
182     foreach my $ds (@ds) {
183     next if (! $ds->{'swish'});
184    
185 dpavlin 466 # strip all non word characters from beginning or end
186     # of word
187 dpavlin 587 my $words = unac(join(" ",@{$ds->{'swish'}}));
188 dpavlin 466 $words =~ s/^\W+//;
189     $words =~ s/\W*\s+\W*/ /g;
190     $words =~ s/\W+$//;
191    
192 dpavlin 504 # first try to generate headline for this entry from index
193 dpavlin 520 my $h = $ds->{'index'}->[0];
194 dpavlin 504 # then, from display
195 dpavlin 520 $h ||= $ds->{'display'}->[0];
196 dpavlin 504 # and as last resport, fallback to headline
197     $h ||= $headline;
198    
199 dpavlin 390 $index->insert(
200     index_name => $ds->{'tag'},
201 dpavlin 466 #path => $f,
202     path => $webpac->mfn,
203 dpavlin 504 headline => $h,
204 dpavlin 587 words => $words,
205 dpavlin 390 );
206     }
207    
208 dpavlin 410 # save into sorted index (thesaurus)
209     foreach my $ds (@ds) {
210     next if (! $ds->{'index'});
211    
212 dpavlin 511 $thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} );
213 dpavlin 410
214 dpavlin 415 foreach my $h (@{$ds->{'index'}}) {
215     $thes->{$ds->{'tag'}}->insert(
216 dpavlin 546 sort_by => unac_2($h),
217     mfn => $webpac->mfn,
218 dpavlin 415 headline => $h,
219     );
220     }
221 dpavlin 410 }
222    
223 dpavlin 454 # print Dump(\@ds);
224 dpavlin 390
225 dpavlin 3 }
226    
227 dpavlin 411 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 dpavlin 415 my $file = "./out/bfilter/$t.txt";
236     $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
237 dpavlin 411
238     $webpac->output_file(
239 dpavlin 415 file => $file,
240 dpavlin 411 template => 'index.tt',
241     data => \@e,
242     index_name => $t,
243     );
244     }
245    
246     if (0 && $log->is_debug) {
247 dpavlin 454 $log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
248     $log->debug("data hash: ",Dump($webpac->{'data'}));
249 dpavlin 410 foreach my $t (keys %{$thes}) {
250 dpavlin 454 $log->debug("thesaurus $t hash: ",Dump($thes->{$t}));
251 dpavlin 410 }
252 dpavlin 372 }
253 dpavlin 390
254 dpavlin 424 } # XXX if(0)
255    
256 dpavlin 530 #$log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
257 dpavlin 420
258 dpavlin 441 $log->info("creating tree");
259    
260 dpavlin 419 #
261 dpavlin 441 # define tree structure
262 dpavlin 419 #
263    
264 dpavlin 424 my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup");
265 dpavlin 419
266 dpavlin 424 my @tree = ({
267     # level 0
268     code_arr => sub { sort keys %{$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 dpavlin 454 have_children => sub { return $l->{"a:".$_[0]."::"} },
277     have_children_at_level => sub {
278 dpavlin 471 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 dpavlin 454 },
284 dpavlin 424 style => 'display: none',
285 dpavlin 579 iframe => 1,
286 dpavlin 424 },{
287     # 1
288 dpavlin 454 code_arr => sub { @{$_[0]} },
289 dpavlin 424 filter_code => sub { shift }, # nop
290     lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
291     lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
292     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
293 dpavlin 454 have_children => sub { return $l->{"a:".$_[1].":"} },
294 dpavlin 424 style => 'display: none',
295     },{
296     # 2
297 dpavlin 454 code_arr => sub { @{$_[0]} },
298 dpavlin 424 filter_code => sub { shift },
299 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
300 dpavlin 424 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
301     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
302 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
303 dpavlin 572 style => 'display: none',
304 dpavlin 425 },{
305     # 3 u¾i pojam
306 dpavlin 454 code_arr => sub { @{$_[0]} },
307 dpavlin 425 filter_code => sub { shift },
308 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
309 dpavlin 426 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
310     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
311 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
312 dpavlin 426 },{
313     # 4
314 dpavlin 454 code_arr => sub { @{$_[0]} },
315 dpavlin 426 filter_code => sub { shift },
316 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
317 dpavlin 426 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
318     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
319 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
320 dpavlin 427 },{
321     # 5
322 dpavlin 454 code_arr => sub { @{$_[0]} },
323 dpavlin 427 filter_code => sub { shift },
324 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
325 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
326     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
327 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
328 dpavlin 427 },{
329     # 6
330 dpavlin 454 code_arr => sub { @{$_[0]} },
331 dpavlin 427 filter_code => sub { shift },
332 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
333 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
334     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
335 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
336 dpavlin 427 },{
337     # 7
338 dpavlin 454 code_arr => sub { @{$_[0]} },
339 dpavlin 427 filter_code => sub { shift },
340 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
341 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
342     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
343 dpavlin 460 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
344 dpavlin 427 },{
345     # 8
346 dpavlin 454 code_arr => sub { @{$_[0]} },
347 dpavlin 427 filter_code => sub { shift },
348 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
349 dpavlin 427 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
350     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
351 dpavlin 460 # have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
352 dpavlin 424 have_children => sub { 0 },
353 dpavlin 454 },{
354     # 9 - level which is never reached except explicitly
355 dpavlin 471 code_arr => sub { @{$_[0]} },
356 dpavlin 454 filter_code => sub { shift },
357 dpavlin 460 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
358 dpavlin 454 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
359     lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
360     have_children => sub { 0 },
361 dpavlin 471 have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) },
362 dpavlin 454 },{
363 dpavlin 424 });
364 dpavlin 419
365 dpavlin 441 my $tree = new WebPAC::Tree(
366 dpavlin 530 tree => \@tree,
367 dpavlin 573 detail_url => sub {
368     my $mfn = shift;
369 dpavlin 574 my $url = "thes/${mfn}.html";
370     return $url if (-e "./out/$url");
371 dpavlin 573 },
372 dpavlin 575 nodes_dir => 'nodes',
373 dpavlin 530 );
374    
375     $tree->output(
376     dir => './out',
377     html => 'browse.html',
378 dpavlin 572 template_dir => './output_template/',
379     template_tree => 'tree.tt',
380     template_node => 'node.tt',
381 dpavlin 441 js => 'tree-ids.js',
382     );
383 dpavlin 428
384 dpavlin 530 $tree->output(
385     dir => './eurovoc',
386     html => 'hijerarhija.html',
387 dpavlin 572 template_dir => './output_template/',
388     template_tree => 'hijerarhija.tt',
389     template_node => 'hijerarhija_node.tt',
390 dpavlin 530 js => 'tree-ids.js',
391     );
392    
393    
394 dpavlin 428 $log->info("closing index");
395 dpavlin 390 $index->close;
396    
397 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