11 |
use YAML; |
use YAML; |
12 |
use Carp; |
use Carp; |
13 |
use Getopt::Long; |
use Getopt::Long; |
14 |
|
use Text::Unaccent 1.02; |
15 |
|
|
16 |
use lib './lib'; |
use lib './lib'; |
17 |
use WebPAC; |
use WebPAC; |
33 |
"low_mem!" => \$low_mem, |
"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 |
|
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/đĐ/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 |
# create WebPAC object |
74 |
# |
# |
75 |
my $webpac = new WebPAC( |
my $webpac = new WebPAC( |
78 |
start_mfn => $start_mfn, |
start_mfn => $start_mfn, |
79 |
debug => $debug, |
debug => $debug, |
80 |
low_mem => $low_mem, |
low_mem => $low_mem, |
81 |
|
filter => $filter, |
82 |
) || die; |
) || die; |
83 |
|
|
84 |
my $log = $webpac->_get_logger() || die "can't get logger"; |
my $log = $webpac->_get_logger() || die "can't get logger"; |
97 |
my $maxmfn = $webpac->open_isis( |
my $maxmfn = $webpac->open_isis( |
98 |
filename => shift @ARGV || '/data/hidra/THS/THS', |
filename => shift @ARGV || '/data/hidra/THS/THS', |
99 |
lookup => [ |
lookup => [ |
100 |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
{ 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' }, |
101 |
# { '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' }, |
102 |
# { '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' }, |
103 |
# { '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' }, |
162 |
$words =~ s/\W*\s+\W*/ /g; |
$words =~ s/\W*\s+\W*/ /g; |
163 |
$words =~ s/\W+$//; |
$words =~ s/\W+$//; |
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( |
$index->insert( |
173 |
index_name => $ds->{'tag'}, |
index_name => $ds->{'tag'}, |
174 |
#path => $f, |
#path => $f, |
175 |
path => $webpac->mfn, |
path => $webpac->mfn, |
176 |
headline => $headline, |
headline => $h, |
177 |
words => $words, |
words => unac($words), |
178 |
); |
); |
179 |
} |
} |
180 |
|
|
182 |
foreach my $ds (@ds) { |
foreach my $ds (@ds) { |
183 |
next if (! $ds->{'index'}); |
next if (! $ds->{'index'}); |
184 |
|
|
185 |
$thes->{$ds->{'tag'}} ||= new WebPAC::Index; |
$thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} ); |
186 |
|
|
187 |
foreach my $h (@{$ds->{'index'}}) { |
foreach my $h (@{$ds->{'index'}}) { |
188 |
$thes->{$ds->{'tag'}}->insert( |
$thes->{$ds->{'tag'}}->insert( |
189 |
headline => $h, |
sort_by => unac_2($h), |
190 |
mfn => $webpac->mfn, |
mfn => $webpac->mfn, |
191 |
|
headline => $h, |
192 |
); |
); |
193 |
} |
} |
194 |
} |
} |
226 |
|
|
227 |
} # XXX if(0) |
} # XXX if(0) |
228 |
|
|
229 |
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
#$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
230 |
|
|
231 |
$log->info("creating tree"); |
$log->info("creating tree"); |
232 |
|
|
335 |
}); |
}); |
336 |
|
|
337 |
my $tree = new WebPAC::Tree( |
my $tree = new WebPAC::Tree( |
338 |
dir => './out', |
tree => \@tree, |
339 |
html => 'browse.html', |
); |
340 |
|
|
341 |
|
$tree->output( |
342 |
|
dir => './out', |
343 |
|
html => 'browse.html', |
344 |
template => './output_template/tree.tt', |
template => './output_template/tree.tt', |
345 |
js => 'tree-ids.js', |
js => 'tree-ids.js', |
|
tree => \@tree, |
|
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"); |
$log->info("closing index"); |
357 |
$index->close; |
$index->close; |
358 |
|
|