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 |
|
|
15 |
use lib './lib'; |
use lib './lib'; |
16 |
use WebPAC; |
use WebPAC; |
17 |
|
use WebPAC::jsFind; |
18 |
|
use WebPAC::Index; |
19 |
|
use WebPAC::Tree; |
20 |
|
|
21 |
|
# options which can be changed via command line |
22 |
|
# |
23 |
|
my $code_page = 'ISO-8859-2'; |
24 |
|
my ($limit_mfn, $start_mfn, $debug, $low_mem); |
25 |
|
my $index_path = './out/index'; |
26 |
|
|
27 |
|
my $result = GetOptions( |
28 |
|
"code_page=s" => \$code_page, |
29 |
|
"limit_mfn=i" => \$limit_mfn, |
30 |
|
"start_mfn=i" => \$start_mfn, |
31 |
|
"debug!" => \$debug, |
32 |
|
"low_mem!" => \$low_mem, |
33 |
|
); |
34 |
|
|
35 |
|
my $filter = { |
36 |
|
'CROVOC' => sub { |
37 |
|
my $tmp = shift || return; |
38 |
|
return undef unless ($tmp =~ s/\s*CROVOC.*$/ */); |
39 |
|
return $tmp; |
40 |
|
}, |
41 |
|
'CROVOC_tree' => sub { |
42 |
|
my $tmp = shift || return; |
43 |
|
$tmp =~ s/\s*CROVOC.*$/ */; |
44 |
|
$tmp =~ s/\s*EUROVOC.*//; |
45 |
|
return $tmp; |
46 |
|
}, |
47 |
|
}; |
48 |
|
|
49 |
|
# create WebPAC object |
50 |
|
# |
51 |
my $webpac = new WebPAC( |
my $webpac = new WebPAC( |
52 |
code_page => 'ISO-8859-2', |
code_page => $code_page, |
53 |
limit_mfn => 500, |
limit_mfn => $limit_mfn, |
54 |
|
start_mfn => $start_mfn, |
55 |
|
debug => $debug, |
56 |
|
low_mem => $low_mem, |
57 |
|
filter => $filter, |
58 |
|
) || die; |
59 |
|
|
60 |
|
my $log = $webpac->_get_logger() || die "can't get logger"; |
61 |
|
|
62 |
|
$log->debug("creating WebPAC::jsFind object"); |
63 |
|
|
64 |
|
my $index = new WebPAC::jsFind( |
65 |
|
index_path => $index_path, |
66 |
|
keys => 10, |
67 |
) || die; |
) || die; |
68 |
|
|
69 |
|
my $thes; |
70 |
|
|
71 |
$|=1; |
$|=1; |
72 |
|
|
|
print "reading database\n"; |
|
73 |
my $maxmfn = $webpac->open_isis( |
my $maxmfn = $webpac->open_isis( |
74 |
filename => shift @ARGV || '/data/hidra/THS/THS', |
filename => shift @ARGV || '/data/hidra/THS/THS', |
75 |
lookup => [ |
lookup => [ |
76 |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
{ 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' }, |
77 |
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
78 |
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
79 |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
80 |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
81 |
|
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
82 |
|
# tree structure |
83 |
|
{ 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' }, |
84 |
|
{ 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, |
85 |
|
{ 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' }, |
86 |
], |
], |
87 |
); |
); |
88 |
|
|
89 |
print "rows: $maxmfn\n\n"; |
$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); |
90 |
|
|
91 |
|
$log->info("rows: $maxmfn"); |
92 |
|
|
93 |
|
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
94 |
|
|
95 |
|
if(1) { # XXX |
96 |
|
|
97 |
while (my $rec = $webpac->fetch_rec) { |
while (my $rec = $webpac->fetch_rec) { |
98 |
|
|
99 |
print "-- ",$webpac->parse($rec,'v250^a / [v562^4] v562^a'),"\n"; |
my @ds = $webpac->data_structure($rec); |
100 |
|
|
101 |
print " ", |
if (0 && $log->is_debug) { |
102 |
$webpac->parse($rec,'eval{v901^a eq "Deskriptor"}v250^a / [v562^4] v562^a'), |
$log->debug("rec = ",Dump($rec)); |
103 |
$webpac->parse($rec,'eval{v901^a ne "Deskriptor"}250a [251]'), |
$log->debug("ds = ",Dump(\@ds)); |
104 |
"\n"; |
} |
105 |
|
|
106 |
|
next if (! @ds); |
107 |
|
|
108 |
|
my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); |
109 |
|
|
110 |
|
if ($filename) { |
111 |
|
$webpac->output_file( |
112 |
|
file => $filename, |
113 |
|
template => 'html.tt', |
114 |
|
data => \@ds, |
115 |
|
headline => $webpac->{'headline'}, |
116 |
|
); |
117 |
|
} else { |
118 |
|
print $webpac->output( |
119 |
|
template => 'text.tt', |
120 |
|
data => \@ds, |
121 |
|
headline => $webpac->{'headline'}, |
122 |
|
); |
123 |
|
} |
124 |
|
|
125 |
|
my $headline = $webpac->{'headline'}; |
126 |
|
|
127 |
|
my $f = $filename; |
128 |
|
$f =~ s!out/!!; |
129 |
|
|
130 |
|
# save into index |
131 |
|
foreach my $ds (@ds) { |
132 |
|
next if (! $ds->{'swish'}); |
133 |
|
|
134 |
|
# strip all non word characters from beginning or end |
135 |
|
# of word |
136 |
|
my $words = join(" ",@{$ds->{'swish'}}); |
137 |
|
$words =~ s/^\W+//; |
138 |
|
$words =~ s/\W*\s+\W*/ /g; |
139 |
|
$words =~ s/\W+$//; |
140 |
|
|
141 |
|
# first try to generate headline for this entry from index |
142 |
|
my $h = shift @{$ds->{'index'}}; |
143 |
|
# then, from display |
144 |
|
$h ||= shift @{$ds->{'display'}}; |
145 |
|
# and as last resport, fallback to headline |
146 |
|
$h ||= $headline; |
147 |
|
|
148 |
|
$index->insert( |
149 |
|
index_name => $ds->{'tag'}, |
150 |
|
#path => $f, |
151 |
|
path => $webpac->mfn, |
152 |
|
headline => $h, |
153 |
|
words => $words, |
154 |
|
); |
155 |
|
} |
156 |
|
|
157 |
|
# save into sorted index (thesaurus) |
158 |
|
foreach my $ds (@ds) { |
159 |
|
next if (! $ds->{'index'}); |
160 |
|
|
161 |
|
$thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} ); |
162 |
|
|
163 |
|
foreach my $h (@{$ds->{'index'}}) { |
164 |
|
$thes->{$ds->{'tag'}}->insert( |
165 |
|
headline => $h, |
166 |
|
mfn => $webpac->mfn, |
167 |
|
); |
168 |
|
} |
169 |
|
} |
170 |
|
|
171 |
my @t = $webpac->fill_in($rec,'v553^1;;v553^a'); |
# print Dump(\@ds); |
|
print " Uži pojam: ",join("\t\n",@t),"\n" if (@t); |
|
172 |
|
|
173 |
@t = $webpac->fill_in($rec,'eval{v901^a eq "Područje"}[a:v251::];;[d:[a:v251::]]'); |
} |
174 |
print " Područje: ",join("\t\n",@t),"\n" if (@t); |
|
175 |
|
foreach my $t (keys %{$thes}) { |
176 |
|
|
177 |
@t = $webpac->fill_in($rec,'eval{v901^a eq "Mikrotezaurus"}[a:v561^4:v251:];;[d:[a:v561^4:v251:]]'); |
my @e = $thes->{$t}->elements; |
178 |
print " Mikrotezaurus: ",join("\t\n",@t),"\n" if (@t); |
if (! @e) { |
179 |
|
$log->logwarn("no elements in sorted index $t?"); |
180 |
|
next; |
181 |
|
} |
182 |
|
|
183 |
|
my $file = "./out/bfilter/$t.txt"; |
184 |
|
$log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); |
185 |
|
|
186 |
|
$webpac->output_file( |
187 |
|
file => $file, |
188 |
|
template => 'index.tt', |
189 |
|
data => \@e, |
190 |
|
index_name => $t, |
191 |
|
); |
192 |
|
} |
193 |
|
|
194 |
@t = $webpac->fill_in($rec,'eval{v901^a eq "Deskriptor"}[a:v561^4:v562^4:v900];;[d:[a:v561^4:v562^4:v900]]'); |
if (0 && $log->is_debug) { |
195 |
print " Deskriptor: ",join("\t\n",@t),"\n" if (@t); |
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
196 |
|
$log->debug("data hash: ",Dump($webpac->{'data'})); |
197 |
|
foreach my $t (keys %{$thes}) { |
198 |
|
$log->debug("thesaurus $t hash: ",Dump($thes->{$t})); |
199 |
|
} |
200 |
} |
} |
201 |
|
|
202 |
print "## lookup ",Dumper($webpac->{'lookup'}); |
} # XXX if(0) |
203 |
print "## data ",Dumper($webpac->{'data'}); |
|
204 |
|
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
205 |
|
|
206 |
|
$log->info("creating tree"); |
207 |
|
|
208 |
|
# |
209 |
|
# define tree structure |
210 |
|
# |
211 |
|
|
212 |
|
my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup"); |
213 |
|
|
214 |
|
my @tree = ({ |
215 |
|
# level 0 |
216 |
|
code_arr => sub { sort keys %{$l} }, |
217 |
|
filter_code => sub { |
218 |
|
my $t = shift; |
219 |
|
return $t if ($t =~ s/root://); |
220 |
|
}, |
221 |
|
lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} }, |
222 |
|
lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, |
223 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, |
224 |
|
have_children => sub { return $l->{"a:".$_[0]."::"} }, |
225 |
|
have_children_at_level => sub { |
226 |
|
return unless (defined($l->{"code:".$_[1]})); |
227 |
|
my $code = shift @{$l->{"code:".$_[1]}}; |
228 |
|
print STDERR "## $_[1] -> $code\n"; |
229 |
|
return undef unless($code); |
230 |
|
return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"})); |
231 |
|
}, |
232 |
|
style => 'display: none', |
233 |
|
},{ |
234 |
|
# 1 |
235 |
|
code_arr => sub { @{$_[0]} }, |
236 |
|
filter_code => sub { shift }, # nop |
237 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
238 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
239 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
240 |
|
have_children => sub { return $l->{"a:".$_[1].":"} }, |
241 |
|
style => 'display: none', |
242 |
|
},{ |
243 |
|
# 2 |
244 |
|
code_arr => sub { @{$_[0]} }, |
245 |
|
filter_code => sub { shift }, |
246 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
247 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
248 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
249 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
250 |
|
#style => 'display: none', |
251 |
|
},{ |
252 |
|
# 3 uži pojam |
253 |
|
code_arr => sub { @{$_[0]} }, |
254 |
|
filter_code => sub { shift }, |
255 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
256 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
257 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
258 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
259 |
|
},{ |
260 |
|
# 4 |
261 |
|
code_arr => sub { @{$_[0]} }, |
262 |
|
filter_code => sub { shift }, |
263 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
264 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
265 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
266 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
267 |
|
},{ |
268 |
|
# 5 |
269 |
|
code_arr => sub { @{$_[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].":".$_[0]} }, |
275 |
|
},{ |
276 |
|
# 6 |
277 |
|
code_arr => sub { @{$_[0]} }, |
278 |
|
filter_code => sub { shift }, |
279 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
280 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
281 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
282 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
283 |
|
},{ |
284 |
|
# 7 |
285 |
|
code_arr => sub { @{$_[0]} }, |
286 |
|
filter_code => sub { shift }, |
287 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
288 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
289 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
290 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
291 |
|
},{ |
292 |
|
# 8 |
293 |
|
code_arr => sub { @{$_[0]} }, |
294 |
|
filter_code => sub { shift }, |
295 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
296 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
297 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
298 |
|
# have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
299 |
|
have_children => sub { 0 }, |
300 |
|
},{ |
301 |
|
# 9 - level which is never reached except explicitly |
302 |
|
code_arr => sub { @{$_[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 { 0 }, |
308 |
|
have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) }, |
309 |
|
},{ |
310 |
|
}); |
311 |
|
|
312 |
|
my $tree = new WebPAC::Tree( |
313 |
|
dir => './out', |
314 |
|
html => 'browse.html', |
315 |
|
template => './output_template/tree.tt', |
316 |
|
js => 'tree-ids.js', |
317 |
|
tree => \@tree, |
318 |
|
); |
319 |
|
|
320 |
|
$log->info("closing index"); |
321 |
|
$index->close; |
322 |
|
|
323 |
|
$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'})); |