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