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