/[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

Contents of /trunk2/all2all.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 549 - (show annotations)
Tue Oct 26 20:53:03 2004 UTC (19 years, 5 months ago) by dpavlin
File MIME type: text/plain
File size: 9301 byte(s)
Magic regexp to unaccent html entities in input data.

1 #!/usr/bin/perl -w
2
3 =head1 NAME
4
5 all2all.pl - basic script for all WebPAC needs
6
7 =cut
8
9 use strict;
10 use locale;
11 use YAML;
12 use Carp;
13 use Getopt::Long;
14 use Text::Unaccent 1.02;
15
16 use lib './lib';
17 use WebPAC;
18 use WebPAC::jsFind;
19 use WebPAC::Index;
20 use WebPAC::Tree;
21
22 # 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 my $filter = {
37 'CROVOC' => sub {
38 my $tmp = shift || return;
39 return undef unless ($tmp =~ s/\s*CROVOC.*$/ */);
40 return $tmp;
41 },
42 'CROVOC_tree' => sub {
43 my $tmp = shift || return;
44 $tmp =~ s/\s*CROVOC.*$/ */;
45 $tmp =~ s/\s*EUROVOC.*//;
46 return $tmp;
47 },
48 };
49
50 ## 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 $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;/$1/gi;
70 return lc($string);
71 }
72
73 # create WebPAC object
74 #
75 my $webpac = new WebPAC(
76 code_page => $code_page,
77 limit_mfn => $limit_mfn,
78 start_mfn => $start_mfn,
79 debug => $debug,
80 low_mem => $low_mem,
81 filter => $filter,
82 ) || die;
83
84 my $log = $webpac->_get_logger() || die "can't get logger";
85
86 $log->debug("creating WebPAC::jsFind object");
87
88 my $index = new WebPAC::jsFind(
89 index_path => $index_path,
90 keys => 10,
91 ) || die;
92
93 my $thes;
94
95 $|=1;
96
97 my $maxmfn = $webpac->open_isis(
98 filename => shift @ARGV || '/data/hidra/THS/THS',
99 lookup => [
100 { 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' },
101 # { 'eval' => '"v901^a" eq "Podruèje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
102 # { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
103 # { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
104 { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
105 { 'key' => '900_mfn:v900', 'val' => 'v000' },
106 # tree structure
107 { 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' },
108 { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' },
109 { 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' },
110 ],
111 );
112
113 $log->debug("isis file ",$webpac->{'isis_filename'}," opened");
114
115 $log->info("rows: $maxmfn");
116
117 $webpac->open_import_xml(type => 'isis_hidra_ths');
118
119 if(1) { # XXX
120
121 while (my $rec = $webpac->fetch_rec) {
122
123 my @ds = $webpac->data_structure($rec);
124
125 if (0 && $log->is_debug) {
126 $log->debug("rec = ",Dump($rec));
127 $log->debug("ds = ",Dump(\@ds));
128 }
129
130 next if (! @ds);
131
132 my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
133
134 if ($filename) {
135 $webpac->output_file(
136 file => $filename,
137 template => 'html.tt',
138 data => \@ds,
139 headline => $webpac->{'headline'},
140 );
141 } else {
142 print $webpac->output(
143 template => 'text.tt',
144 data => \@ds,
145 headline => $webpac->{'headline'},
146 );
147 }
148
149 my $headline = $webpac->{'headline'};
150
151 my $f = $filename;
152 $f =~ s!out/!!;
153
154 # save into index
155 foreach my $ds (@ds) {
156 next if (! $ds->{'swish'});
157
158 # 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 # first try to generate headline for this entry from index
166 my $h = $ds->{'index'}->[0];
167 # then, from display
168 $h ||= $ds->{'display'}->[0];
169 # and as last resport, fallback to headline
170 $h ||= $headline;
171
172 $index->insert(
173 index_name => $ds->{'tag'},
174 #path => $f,
175 path => $webpac->mfn,
176 headline => $h,
177 words => unac($words),
178 );
179 }
180
181 # save into sorted index (thesaurus)
182 foreach my $ds (@ds) {
183 next if (! $ds->{'index'});
184
185 $thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} );
186
187 foreach my $h (@{$ds->{'index'}}) {
188 $thes->{$ds->{'tag'}}->insert(
189 sort_by => unac_2($h),
190 mfn => $webpac->mfn,
191 headline => $h,
192 );
193 }
194 }
195
196 # print Dump(\@ds);
197
198 }
199
200 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 my $file = "./out/bfilter/$t.txt";
209 $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
210
211 $webpac->output_file(
212 file => $file,
213 template => 'index.tt',
214 data => \@e,
215 index_name => $t,
216 );
217 }
218
219 if (0 && $log->is_debug) {
220 $log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
221 $log->debug("data hash: ",Dump($webpac->{'data'}));
222 foreach my $t (keys %{$thes}) {
223 $log->debug("thesaurus $t hash: ",Dump($thes->{$t}));
224 }
225 }
226
227 } # XXX if(0)
228
229 #$log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
230
231 $log->info("creating tree");
232
233 #
234 # define tree structure
235 #
236
237 my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup");
238
239 my @tree = ({
240 # level 0
241 code_arr => sub { sort keys %{$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 have_children => sub { return $l->{"a:".$_[0]."::"} },
250 have_children_at_level => sub {
251 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 },
257 style => 'display: none',
258 },{
259 # 1
260 code_arr => sub { @{$_[0]} },
261 filter_code => sub { shift }, # nop
262 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
263 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
264 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
265 have_children => sub { return $l->{"a:".$_[1].":"} },
266 style => 'display: none',
267 },{
268 # 2
269 code_arr => sub { @{$_[0]} },
270 filter_code => sub { shift },
271 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
272 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
273 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
274 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
275 #style => 'display: none',
276 },{
277 # 3 u¾i pojam
278 code_arr => sub { @{$_[0]} },
279 filter_code => sub { shift },
280 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
281 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
282 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
283 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
284 },{
285 # 4
286 code_arr => sub { @{$_[0]} },
287 filter_code => sub { shift },
288 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
289 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
290 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
291 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
292 },{
293 # 5
294 code_arr => sub { @{$_[0]} },
295 filter_code => sub { shift },
296 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
297 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
298 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
299 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
300 },{
301 # 6
302 code_arr => sub { @{$_[0]} },
303 filter_code => sub { shift },
304 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
305 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
306 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
307 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
308 },{
309 # 7
310 code_arr => sub { @{$_[0]} },
311 filter_code => sub { shift },
312 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
313 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
314 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
315 have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
316 },{
317 # 8
318 code_arr => sub { @{$_[0]} },
319 filter_code => sub { shift },
320 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
321 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
322 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
323 # have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
324 have_children => sub { 0 },
325 },{
326 # 9 - level which is never reached except explicitly
327 code_arr => sub { @{$_[0]} },
328 filter_code => sub { shift },
329 lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
330 lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
331 lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
332 have_children => sub { 0 },
333 have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) },
334 },{
335 });
336
337 my $tree = new WebPAC::Tree(
338 tree => \@tree,
339 );
340
341 $tree->output(
342 dir => './out',
343 html => 'browse.html',
344 template => './output_template/tree.tt',
345 js => 'tree-ids.js',
346 );
347
348 $tree->output(
349 dir => './eurovoc',
350 html => 'hijerarhija.html',
351 template => './output_template/hijerarhija.tt',
352 js => 'tree-ids.js',
353 );
354
355
356 $log->info("closing index");
357 $index->close;
358
359 $log->info("elapsed time: ",$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