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; |
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/CROVOC.*$/ */); |
39 |
|
return $tmp; |
40 |
|
}, |
41 |
|
}; |
42 |
|
|
43 |
|
# create WebPAC object |
44 |
|
# |
45 |
my $webpac = new WebPAC( |
my $webpac = new WebPAC( |
46 |
code_page => 'ISO-8859-2', |
code_page => $code_page, |
47 |
limit_mfn => 100, |
limit_mfn => $limit_mfn, |
48 |
# debug => 1, |
start_mfn => $start_mfn, |
49 |
|
debug => $debug, |
50 |
|
low_mem => $low_mem, |
51 |
|
filter => $filter, |
52 |
) || die; |
) || die; |
53 |
|
|
54 |
my $log = $webpac->_get_logger(); |
my $log = $webpac->_get_logger() || die "can't get logger"; |
55 |
|
|
56 |
|
$log->debug("creating WebPAC::jsFind object"); |
57 |
|
|
58 |
my $index = new WebPAC::jsFind( |
my $index = new WebPAC::jsFind( |
59 |
index_path => './out/index', |
index_path => $index_path, |
60 |
|
keys => 10, |
61 |
) || die; |
) || die; |
62 |
|
|
63 |
|
my $thes; |
64 |
|
|
65 |
$|=1; |
$|=1; |
66 |
|
|
67 |
my $maxmfn = $webpac->open_isis( |
my $maxmfn = $webpac->open_isis( |
73 |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
74 |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
75 |
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
76 |
|
# tree structure |
77 |
|
{ 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' }, |
78 |
|
{ 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, |
79 |
|
{ 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' }, |
80 |
], |
], |
81 |
); |
); |
82 |
|
|
83 |
|
$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); |
84 |
|
|
85 |
$log->info("rows: $maxmfn"); |
$log->info("rows: $maxmfn"); |
86 |
|
|
87 |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
88 |
|
|
89 |
|
if(1) { # XXX |
90 |
|
|
91 |
while (my $rec = $webpac->fetch_rec) { |
while (my $rec = $webpac->fetch_rec) { |
92 |
|
|
93 |
my @ds = $webpac->data_structure($rec); |
my @ds = $webpac->data_structure($rec); |
94 |
|
|
95 |
if ($log->is_debug) { |
if (0 && $log->is_debug) { |
96 |
$log->debug("rec = ",Dumper($rec)); |
$log->debug("rec = ",Dump($rec)); |
97 |
$log->debug("ds = ",Dumper(\@ds)); |
$log->debug("ds = ",Dump(\@ds)); |
98 |
} |
} |
99 |
|
|
100 |
next if (! @ds); |
next if (! @ds); |
101 |
|
|
102 |
my $filename = $webpac->{'current_filename'}; |
my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); |
103 |
|
|
104 |
if ($filename) { |
if ($filename) { |
105 |
open(OUT,"> $filename") || $log->logdie("can't open output '$filename': $!"); |
$webpac->output_file( |
106 |
print OUT $webpac->output( |
file => $filename, |
107 |
template => 'html.tt', |
template => 'html.tt', |
108 |
data => \@ds, |
data => \@ds, |
109 |
|
headline => $webpac->{'headline'}, |
110 |
); |
); |
|
close(OUT); |
|
111 |
} else { |
} else { |
112 |
print $webpac->output( |
print $webpac->output( |
113 |
template => 'text.tt', |
template => 'text.tt', |
114 |
data => \@ds, |
data => \@ds, |
115 |
|
headline => $webpac->{'headline'}, |
116 |
); |
); |
117 |
} |
} |
118 |
|
|
119 |
my $headline; |
my $headline = $webpac->{'headline'}; |
120 |
foreach my $ds (@ds) { |
|
121 |
if ($ds->{'tag'} eq 'headline') { |
my $f = $filename; |
122 |
$headline = join(" ",@{$ds->{'display'}}); |
$f =~ s!out/!!; |
|
last; |
|
|
} |
|
|
} |
|
123 |
|
|
124 |
# save into index |
# save into index |
125 |
foreach my $ds (@ds) { |
foreach my $ds (@ds) { |
126 |
next if (! $ds->{'swish'}); |
next if (! $ds->{'swish'}); |
127 |
|
|
128 |
|
# strip all non word characters from beginning or end |
129 |
|
# of word |
130 |
|
my $words = join(" ",@{$ds->{'swish'}}); |
131 |
|
$words =~ s/^\W+//; |
132 |
|
$words =~ s/\W*\s+\W*/ /g; |
133 |
|
$words =~ s/\W+$//; |
134 |
|
|
135 |
$index->insert( |
$index->insert( |
136 |
index_name => $ds->{'tag'}, |
index_name => $ds->{'tag'}, |
137 |
path => $filename, |
#path => $f, |
138 |
|
path => $webpac->mfn, |
139 |
headline => $headline, |
headline => $headline, |
140 |
words => join(" ",@{$ds->{'swish'}}) |
words => $words, |
141 |
); |
); |
142 |
} |
} |
143 |
|
|
144 |
# print Dumper(\@ds); |
# save into sorted index (thesaurus) |
145 |
|
foreach my $ds (@ds) { |
146 |
|
next if (! $ds->{'index'}); |
147 |
|
|
148 |
|
$thes->{$ds->{'tag'}} ||= new WebPAC::Index; |
149 |
|
|
150 |
|
foreach my $h (@{$ds->{'index'}}) { |
151 |
|
$thes->{$ds->{'tag'}}->insert( |
152 |
|
headline => $h, |
153 |
|
mfn => $webpac->mfn, |
154 |
|
); |
155 |
|
} |
156 |
|
} |
157 |
|
|
158 |
|
# print Dump(\@ds); |
159 |
|
|
160 |
} |
} |
161 |
|
|
162 |
if ($log->is_debug) { |
foreach my $t (keys %{$thes}) { |
163 |
$log->debug("lookup hash: ",Dumper($webpac->{'lookup'})); |
|
164 |
$log->debug("data hash: ",Dumper($webpac->{'data'})); |
my @e = $thes->{$t}->elements; |
165 |
|
if (! @e) { |
166 |
|
$log->logwarn("no elements in sorted index $t?"); |
167 |
|
next; |
168 |
|
} |
169 |
|
|
170 |
|
my $file = "./out/bfilter/$t.txt"; |
171 |
|
$log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); |
172 |
|
|
173 |
|
$webpac->output_file( |
174 |
|
file => $file, |
175 |
|
template => 'index.tt', |
176 |
|
data => \@e, |
177 |
|
index_name => $t, |
178 |
|
); |
179 |
} |
} |
180 |
|
|
181 |
|
if (0 && $log->is_debug) { |
182 |
|
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
183 |
|
$log->debug("data hash: ",Dump($webpac->{'data'})); |
184 |
|
foreach my $t (keys %{$thes}) { |
185 |
|
$log->debug("thesaurus $t hash: ",Dump($thes->{$t})); |
186 |
|
} |
187 |
|
} |
188 |
|
|
189 |
|
} # XXX if(0) |
190 |
|
|
191 |
|
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
192 |
|
|
193 |
|
$log->info("creating tree"); |
194 |
|
|
195 |
|
# |
196 |
|
# define tree structure |
197 |
|
# |
198 |
|
|
199 |
|
my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup"); |
200 |
|
|
201 |
|
my @tree = ({ |
202 |
|
# level 0 |
203 |
|
code_arr => sub { sort keys %{$l} }, |
204 |
|
filter_code => sub { |
205 |
|
my $t = shift; |
206 |
|
return $t if ($t =~ s/root://); |
207 |
|
}, |
208 |
|
lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} }, |
209 |
|
lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, |
210 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, |
211 |
|
have_children => sub { return $l->{"a:".$_[0]."::"} }, |
212 |
|
have_children_at_level => sub { |
213 |
|
return unless (defined($l->{"code:".$_[1]})); |
214 |
|
my $code = shift @{$l->{"code:".$_[1]}}; |
215 |
|
print STDERR "## $_[1] -> $code\n"; |
216 |
|
return undef unless($code); |
217 |
|
return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"})); |
218 |
|
}, |
219 |
|
style => 'display: none', |
220 |
|
},{ |
221 |
|
# 1 |
222 |
|
code_arr => sub { @{$_[0]} }, |
223 |
|
filter_code => sub { shift }, # nop |
224 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
225 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
226 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
227 |
|
have_children => sub { return $l->{"a:".$_[1].":"} }, |
228 |
|
style => 'display: none', |
229 |
|
},{ |
230 |
|
# 2 |
231 |
|
code_arr => sub { @{$_[0]} }, |
232 |
|
filter_code => sub { shift }, |
233 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
234 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
235 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
236 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
237 |
|
#style => 'display: none', |
238 |
|
},{ |
239 |
|
# 3 uži pojam |
240 |
|
code_arr => sub { @{$_[0]} }, |
241 |
|
filter_code => sub { shift }, |
242 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
243 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
244 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
245 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
246 |
|
},{ |
247 |
|
# 4 |
248 |
|
code_arr => sub { @{$_[0]} }, |
249 |
|
filter_code => sub { shift }, |
250 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
251 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
252 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
253 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
254 |
|
},{ |
255 |
|
# 5 |
256 |
|
code_arr => sub { @{$_[0]} }, |
257 |
|
filter_code => sub { shift }, |
258 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
259 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
260 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
261 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
262 |
|
},{ |
263 |
|
# 6 |
264 |
|
code_arr => sub { @{$_[0]} }, |
265 |
|
filter_code => sub { shift }, |
266 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
267 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
268 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
269 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
270 |
|
},{ |
271 |
|
# 7 |
272 |
|
code_arr => sub { @{$_[0]} }, |
273 |
|
filter_code => sub { shift }, |
274 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
275 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
276 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
277 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
278 |
|
},{ |
279 |
|
# 8 |
280 |
|
code_arr => sub { @{$_[0]} }, |
281 |
|
filter_code => sub { shift }, |
282 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
283 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
284 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
285 |
|
# have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
286 |
|
have_children => sub { 0 }, |
287 |
|
},{ |
288 |
|
# 9 - level which is never reached except explicitly |
289 |
|
code_arr => sub { @{$_[0]} }, |
290 |
|
filter_code => sub { shift }, |
291 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
292 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
293 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
294 |
|
have_children => sub { 0 }, |
295 |
|
have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) }, |
296 |
|
},{ |
297 |
|
}); |
298 |
|
|
299 |
|
my $tree = new WebPAC::Tree( |
300 |
|
dir => './out', |
301 |
|
html => 'browse.html', |
302 |
|
template => './output_template/tree.tt', |
303 |
|
js => 'tree-ids.js', |
304 |
|
tree => \@tree, |
305 |
|
); |
306 |
|
|
307 |
|
$log->info("closing index"); |
308 |
$index->close; |
$index->close; |
309 |
|
|
310 |
|
$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'})); |