7 |
=cut |
=cut |
8 |
|
|
9 |
use strict; |
use strict; |
10 |
use Data::Dumper; |
use locale; |
11 |
|
use YAML; |
12 |
use Carp; |
use Carp; |
13 |
|
use Getopt::Long; |
14 |
|
use Text::Unaccent 1.02; |
15 |
|
|
16 |
use lib './lib'; |
use lib './lib'; |
17 |
use WebPAC; |
use WebPAC; |
18 |
use WebPAC::jsFind; |
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/đĐ/dD/; |
56 |
|
$string = unac_entities($string); |
57 |
|
return $string; |
58 |
|
} |
59 |
|
sub unac_2 { |
60 |
|
my $string = shift || return; |
61 |
|
if (length($string) > 2) { |
62 |
|
my $pr = substr($string,0,2); |
63 |
|
$string = unac_string($code_page,substr($string,2)); |
64 |
|
$string =~ tr/đĐ/dD/; |
65 |
|
$string = $pr . $string; |
66 |
|
} |
67 |
|
$string = unac_entities($string); |
68 |
|
return lc($string); |
69 |
|
} |
70 |
|
|
71 |
|
sub unac_entities { |
72 |
|
my $ent = shift || return; |
73 |
|
|
74 |
|
$ent =~ s/&(\w)(acute|cedil|circ|grave|ring|slash|tilde|uml);/$1/gi; |
75 |
|
$ent =~ s/ð/d/g; |
76 |
|
$ent =~ s/&E[tT][hH];/D/g; |
77 |
|
$ent =~ s/&(\w\w)lig;/$1/gi; |
78 |
|
|
79 |
|
return $ent; |
80 |
|
} |
81 |
|
|
82 |
|
# create WebPAC object |
83 |
|
# |
84 |
my $webpac = new WebPAC( |
my $webpac = new WebPAC( |
85 |
code_page => 'ISO-8859-2', |
code_page => $code_page, |
86 |
limit_mfn => 100, |
limit_mfn => $limit_mfn, |
87 |
# debug => 1, |
start_mfn => $start_mfn, |
88 |
|
debug => $debug, |
89 |
|
low_mem => $low_mem, |
90 |
|
filter => $filter, |
91 |
) || die; |
) || die; |
92 |
|
|
93 |
my $log = $webpac->_get_logger(); |
my $log = $webpac->_get_logger() || die "can't get logger"; |
94 |
|
|
95 |
|
$log->debug("creating WebPAC::jsFind object"); |
96 |
|
|
97 |
my $index = new WebPAC::jsFind( |
my $index = new WebPAC::jsFind( |
98 |
index_path => './out/index', |
index_path => $index_path, |
99 |
|
keys => 10, |
100 |
) || die; |
) || die; |
101 |
|
|
102 |
|
my $thes; |
103 |
|
|
104 |
$|=1; |
$|=1; |
105 |
|
|
106 |
my $maxmfn = $webpac->open_isis( |
my $maxmfn = $webpac->open_isis( |
107 |
filename => shift @ARGV || '/data/hidra/THS/THS', |
filename => shift @ARGV || '/data/hidra/THS/THS', |
108 |
lookup => [ |
lookup => [ |
109 |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
{ 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' }, |
110 |
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
111 |
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
112 |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
113 |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
114 |
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
115 |
|
# tree structure |
116 |
|
{ 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' }, |
117 |
|
{ 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, |
118 |
|
{ 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' }, |
119 |
], |
], |
120 |
); |
); |
121 |
|
|
122 |
|
$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); |
123 |
|
|
124 |
$log->info("rows: $maxmfn"); |
$log->info("rows: $maxmfn"); |
125 |
|
|
126 |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
127 |
|
|
128 |
|
if(1) { # XXX |
129 |
|
|
130 |
while (my $rec = $webpac->fetch_rec) { |
while (my $rec = $webpac->fetch_rec) { |
131 |
|
|
132 |
my @ds = $webpac->data_structure($rec); |
my @ds = $webpac->data_structure($rec); |
133 |
|
|
134 |
if ($log->is_debug) { |
if (0 && $log->is_debug) { |
135 |
$log->debug("rec = ",Dumper($rec)); |
$log->debug("rec = ",Dump($rec)); |
136 |
$log->debug("ds = ",Dumper(\@ds)); |
$log->debug("ds = ",Dump(\@ds)); |
137 |
} |
} |
138 |
|
|
139 |
next if (! @ds); |
next if (! @ds); |
140 |
|
|
141 |
my $filename = $webpac->{'current_filename'}; |
my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); |
142 |
|
|
143 |
if ($filename) { |
if ($filename) { |
144 |
open(OUT,"> $filename") || $log->logdie("can't open output '$filename': $!"); |
$webpac->output_file( |
145 |
print OUT $webpac->output( |
file => $filename, |
146 |
template => 'html.tt', |
template => 'html.tt', |
147 |
data => \@ds, |
data => \@ds, |
148 |
|
headline => $webpac->{'headline'}, |
149 |
); |
); |
|
close(OUT); |
|
150 |
} else { |
} else { |
151 |
print $webpac->output( |
print $webpac->output( |
152 |
template => 'text.tt', |
template => 'text.tt', |
153 |
data => \@ds, |
data => \@ds, |
154 |
|
headline => $webpac->{'headline'}, |
155 |
); |
); |
156 |
} |
} |
157 |
|
|
158 |
my $headline; |
my $headline = $webpac->{'headline'}; |
|
foreach my $ds (@ds) { |
|
|
if ($ds->{'tag'} eq 'headline') { |
|
|
$headline = join(" ",@{$ds->{'display'}}); |
|
|
last; |
|
|
} |
|
|
} |
|
159 |
|
|
160 |
my $f = $filename; |
my $f = $filename; |
161 |
$f =~ s!out/!!; |
$f =~ s!out/!!; |
164 |
foreach my $ds (@ds) { |
foreach my $ds (@ds) { |
165 |
next if (! $ds->{'swish'}); |
next if (! $ds->{'swish'}); |
166 |
|
|
167 |
|
# strip all non word characters from beginning or end |
168 |
|
# of word |
169 |
|
my $words = join(" ",@{$ds->{'swish'}}); |
170 |
|
$words =~ s/^\W+//; |
171 |
|
$words =~ s/\W*\s+\W*/ /g; |
172 |
|
$words =~ s/\W+$//; |
173 |
|
|
174 |
|
# first try to generate headline for this entry from index |
175 |
|
my $h = $ds->{'index'}->[0]; |
176 |
|
# then, from display |
177 |
|
$h ||= $ds->{'display'}->[0]; |
178 |
|
# and as last resport, fallback to headline |
179 |
|
$h ||= $headline; |
180 |
|
|
181 |
$index->insert( |
$index->insert( |
182 |
index_name => $ds->{'tag'}, |
index_name => $ds->{'tag'}, |
183 |
path => $f, |
#path => $f, |
184 |
headline => $headline, |
path => $webpac->mfn, |
185 |
words => join(" ",@{$ds->{'swish'}}) |
headline => $h, |
186 |
|
words => unac($words), |
187 |
); |
); |
188 |
} |
} |
189 |
|
|
190 |
# print Dumper(\@ds); |
# save into sorted index (thesaurus) |
191 |
|
foreach my $ds (@ds) { |
192 |
|
next if (! $ds->{'index'}); |
193 |
|
|
194 |
|
$thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} ); |
195 |
|
|
196 |
|
foreach my $h (@{$ds->{'index'}}) { |
197 |
|
$thes->{$ds->{'tag'}}->insert( |
198 |
|
sort_by => unac_2($h), |
199 |
|
mfn => $webpac->mfn, |
200 |
|
headline => $h, |
201 |
|
); |
202 |
|
} |
203 |
|
} |
204 |
|
|
205 |
|
# print Dump(\@ds); |
206 |
|
|
207 |
} |
} |
208 |
|
|
209 |
if ($log->is_debug) { |
foreach my $t (keys %{$thes}) { |
210 |
$log->debug("lookup hash: ",Dumper($webpac->{'lookup'})); |
|
211 |
$log->debug("data hash: ",Dumper($webpac->{'data'})); |
my @e = $thes->{$t}->elements; |
212 |
|
if (! @e) { |
213 |
|
$log->logwarn("no elements in sorted index $t?"); |
214 |
|
next; |
215 |
|
} |
216 |
|
|
217 |
|
my $file = "./out/bfilter/$t.txt"; |
218 |
|
$log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); |
219 |
|
|
220 |
|
$webpac->output_file( |
221 |
|
file => $file, |
222 |
|
template => 'index.tt', |
223 |
|
data => \@e, |
224 |
|
index_name => $t, |
225 |
|
); |
226 |
} |
} |
227 |
|
|
228 |
|
if (0 && $log->is_debug) { |
229 |
|
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
230 |
|
$log->debug("data hash: ",Dump($webpac->{'data'})); |
231 |
|
foreach my $t (keys %{$thes}) { |
232 |
|
$log->debug("thesaurus $t hash: ",Dump($thes->{$t})); |
233 |
|
} |
234 |
|
} |
235 |
|
|
236 |
|
} # XXX if(0) |
237 |
|
|
238 |
|
#$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
239 |
|
|
240 |
|
$log->info("creating tree"); |
241 |
|
|
242 |
|
# |
243 |
|
# define tree structure |
244 |
|
# |
245 |
|
|
246 |
|
my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup"); |
247 |
|
|
248 |
|
my @tree = ({ |
249 |
|
# level 0 |
250 |
|
code_arr => sub { sort keys %{$l} }, |
251 |
|
filter_code => sub { |
252 |
|
my $t = shift; |
253 |
|
return $t if ($t =~ s/root://); |
254 |
|
}, |
255 |
|
lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} }, |
256 |
|
lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, |
257 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, |
258 |
|
have_children => sub { return $l->{"a:".$_[0]."::"} }, |
259 |
|
have_children_at_level => sub { |
260 |
|
return unless (defined($l->{"code:".$_[1]})); |
261 |
|
my $code = shift @{$l->{"code:".$_[1]}}; |
262 |
|
print STDERR "## $_[1] -> $code\n"; |
263 |
|
return undef unless($code); |
264 |
|
return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"})); |
265 |
|
}, |
266 |
|
style => 'display: none', |
267 |
|
},{ |
268 |
|
# 1 |
269 |
|
code_arr => sub { @{$_[0]} }, |
270 |
|
filter_code => sub { shift }, # nop |
271 |
|
lookup_v900 => sub { shift @{$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].":"} }, |
275 |
|
style => 'display: none', |
276 |
|
},{ |
277 |
|
# 2 |
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].":".$_[0]} }, |
284 |
|
#style => 'display: none', |
285 |
|
},{ |
286 |
|
# 3 uži pojam |
287 |
|
code_arr => sub { @{$_[0]} }, |
288 |
|
filter_code => sub { shift }, |
289 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
290 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
291 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
292 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
293 |
|
},{ |
294 |
|
# 4 |
295 |
|
code_arr => sub { @{$_[0]} }, |
296 |
|
filter_code => sub { shift }, |
297 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
298 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
299 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
300 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
301 |
|
},{ |
302 |
|
# 5 |
303 |
|
code_arr => sub { @{$_[0]} }, |
304 |
|
filter_code => sub { shift }, |
305 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
306 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
307 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
308 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
309 |
|
},{ |
310 |
|
# 6 |
311 |
|
code_arr => sub { @{$_[0]} }, |
312 |
|
filter_code => sub { shift }, |
313 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
314 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
315 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
316 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
317 |
|
},{ |
318 |
|
# 7 |
319 |
|
code_arr => sub { @{$_[0]} }, |
320 |
|
filter_code => sub { shift }, |
321 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
322 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
323 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
324 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
325 |
|
},{ |
326 |
|
# 8 |
327 |
|
code_arr => sub { @{$_[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 { return $l->{"a:".$_[1].":".$_[0]} }, |
333 |
|
have_children => sub { 0 }, |
334 |
|
},{ |
335 |
|
# 9 - level which is never reached except explicitly |
336 |
|
code_arr => sub { @{$_[0]} }, |
337 |
|
filter_code => sub { shift }, |
338 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
339 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
340 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
341 |
|
have_children => sub { 0 }, |
342 |
|
have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) }, |
343 |
|
},{ |
344 |
|
}); |
345 |
|
|
346 |
|
my $tree = new WebPAC::Tree( |
347 |
|
tree => \@tree, |
348 |
|
); |
349 |
|
|
350 |
|
$tree->output( |
351 |
|
dir => './out', |
352 |
|
html => 'browse.html', |
353 |
|
template => './output_template/tree.tt', |
354 |
|
js => 'tree-ids.js', |
355 |
|
); |
356 |
|
|
357 |
|
$tree->output( |
358 |
|
dir => './eurovoc', |
359 |
|
html => 'hijerarhija.html', |
360 |
|
template => './output_template/hijerarhija.tt', |
361 |
|
js => 'tree-ids.js', |
362 |
|
); |
363 |
|
|
364 |
|
|
365 |
|
$log->info("closing index"); |
366 |
$index->close; |
$index->close; |
367 |
|
|
368 |
|
$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'})); |