1 |
dpavlin |
1 |
#!/usr/bin/perl -w |
2 |
|
|
|
3 |
dpavlin |
348 |
=head1 NAME |
4 |
|
|
|
5 |
dpavlin |
354 |
all2all.pl - basic script for all WebPAC needs |
6 |
dpavlin |
348 |
|
7 |
|
|
=cut |
8 |
|
|
|
9 |
dpavlin |
1 |
use strict; |
10 |
dpavlin |
454 |
use YAML; |
11 |
dpavlin |
352 |
use Carp; |
12 |
dpavlin |
437 |
use Getopt::Long; |
13 |
dpavlin |
1 |
|
14 |
dpavlin |
352 |
use lib './lib'; |
15 |
dpavlin |
354 |
use WebPAC; |
16 |
dpavlin |
390 |
use WebPAC::jsFind; |
17 |
dpavlin |
410 |
use WebPAC::Index; |
18 |
dpavlin |
441 |
use WebPAC::Tree; |
19 |
dpavlin |
348 |
|
20 |
dpavlin |
437 |
# options which can be changed via command line |
21 |
|
|
# |
22 |
|
|
my $code_page = 'ISO-8859-2'; |
23 |
|
|
my ($limit_mfn, $start_mfn, $debug, $low_mem); |
24 |
|
|
my $index_path = './out/index'; |
25 |
|
|
|
26 |
|
|
my $result = GetOptions( |
27 |
|
|
"code_page=s" => \$code_page, |
28 |
|
|
"limit_mfn=i" => \$limit_mfn, |
29 |
|
|
"start_mfn=i" => \$start_mfn, |
30 |
|
|
"debug!" => \$debug, |
31 |
|
|
"low_mem!" => \$low_mem, |
32 |
|
|
); |
33 |
|
|
|
34 |
|
|
# create WebPAC object |
35 |
|
|
# |
36 |
dpavlin |
354 |
my $webpac = new WebPAC( |
37 |
dpavlin |
437 |
code_page => $code_page, |
38 |
|
|
limit_mfn => $limit_mfn, |
39 |
|
|
start_mfn => $start_mfn, |
40 |
|
|
debug => $debug, |
41 |
|
|
low_mem => $low_mem, |
42 |
dpavlin |
352 |
) || die; |
43 |
|
|
|
44 |
dpavlin |
421 |
my $log = $webpac->_get_logger() || die "can't get logger"; |
45 |
dpavlin |
373 |
|
46 |
dpavlin |
421 |
$log->debug("creating WebPAC::jsFind object"); |
47 |
|
|
|
48 |
dpavlin |
390 |
my $index = new WebPAC::jsFind( |
49 |
dpavlin |
437 |
index_path => $index_path, |
50 |
dpavlin |
413 |
keys => 10, |
51 |
dpavlin |
390 |
) || die; |
52 |
|
|
|
53 |
dpavlin |
410 |
my $thes; |
54 |
|
|
|
55 |
dpavlin |
10 |
$|=1; |
56 |
dpavlin |
9 |
|
57 |
dpavlin |
352 |
my $maxmfn = $webpac->open_isis( |
58 |
dpavlin |
357 |
filename => shift @ARGV || '/data/hidra/THS/THS', |
59 |
dpavlin |
352 |
lookup => [ |
60 |
dpavlin |
348 |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
61 |
dpavlin |
352 |
# { 'eval' => '"v901^a" eq "Podruèje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
62 |
|
|
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
63 |
|
|
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
64 |
|
|
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
65 |
dpavlin |
374 |
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
66 |
dpavlin |
419 |
# tree structure |
67 |
dpavlin |
471 |
{ 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' }, |
68 |
dpavlin |
419 |
{ 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, |
69 |
dpavlin |
460 |
{ 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' }, |
70 |
dpavlin |
352 |
], |
71 |
dpavlin |
54 |
); |
72 |
dpavlin |
3 |
|
73 |
dpavlin |
421 |
$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); |
74 |
|
|
|
75 |
dpavlin |
372 |
$log->info("rows: $maxmfn"); |
76 |
dpavlin |
170 |
|
77 |
dpavlin |
363 |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
78 |
|
|
|
79 |
dpavlin |
425 |
if(1) { # XXX |
80 |
dpavlin |
424 |
|
81 |
dpavlin |
362 |
while (my $rec = $webpac->fetch_rec) { |
82 |
dpavlin |
358 |
|
83 |
dpavlin |
370 |
my @ds = $webpac->data_structure($rec); |
84 |
dpavlin |
366 |
|
85 |
dpavlin |
411 |
if (0 && $log->is_debug) { |
86 |
dpavlin |
454 |
$log->debug("rec = ",Dump($rec)); |
87 |
|
|
$log->debug("ds = ",Dump(\@ds)); |
88 |
dpavlin |
372 |
} |
89 |
|
|
|
90 |
dpavlin |
374 |
next if (! @ds); |
91 |
dpavlin |
366 |
|
92 |
dpavlin |
421 |
my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); |
93 |
dpavlin |
374 |
|
94 |
|
|
if ($filename) { |
95 |
dpavlin |
411 |
$webpac->output_file( |
96 |
|
|
file => $filename, |
97 |
dpavlin |
374 |
template => 'html.tt', |
98 |
|
|
data => \@ds, |
99 |
dpavlin |
398 |
headline => $webpac->{'headline'}, |
100 |
dpavlin |
374 |
); |
101 |
|
|
} else { |
102 |
|
|
print $webpac->output( |
103 |
|
|
template => 'text.tt', |
104 |
|
|
data => \@ds, |
105 |
dpavlin |
398 |
headline => $webpac->{'headline'}, |
106 |
dpavlin |
374 |
); |
107 |
|
|
} |
108 |
|
|
|
109 |
dpavlin |
398 |
my $headline = $webpac->{'headline'}; |
110 |
dpavlin |
390 |
|
111 |
dpavlin |
392 |
my $f = $filename; |
112 |
|
|
$f =~ s!out/!!; |
113 |
|
|
|
114 |
dpavlin |
390 |
# save into index |
115 |
|
|
foreach my $ds (@ds) { |
116 |
|
|
next if (! $ds->{'swish'}); |
117 |
|
|
|
118 |
dpavlin |
466 |
# strip all non word characters from beginning or end |
119 |
|
|
# of word |
120 |
|
|
my $words = join(" ",@{$ds->{'swish'}}); |
121 |
|
|
$words =~ s/^\W+//; |
122 |
|
|
$words =~ s/\W*\s+\W*/ /g; |
123 |
|
|
$words =~ s/\W+$//; |
124 |
|
|
|
125 |
dpavlin |
390 |
$index->insert( |
126 |
|
|
index_name => $ds->{'tag'}, |
127 |
dpavlin |
466 |
#path => $f, |
128 |
|
|
path => $webpac->mfn, |
129 |
dpavlin |
390 |
headline => $headline, |
130 |
dpavlin |
466 |
words => $words, |
131 |
dpavlin |
390 |
); |
132 |
|
|
} |
133 |
|
|
|
134 |
dpavlin |
410 |
# save into sorted index (thesaurus) |
135 |
|
|
foreach my $ds (@ds) { |
136 |
|
|
next if (! $ds->{'index'}); |
137 |
|
|
|
138 |
|
|
$thes->{$ds->{'tag'}} ||= new WebPAC::Index; |
139 |
|
|
|
140 |
dpavlin |
415 |
foreach my $h (@{$ds->{'index'}}) { |
141 |
|
|
$thes->{$ds->{'tag'}}->insert( |
142 |
|
|
headline => $h, |
143 |
dpavlin |
448 |
mfn => $webpac->mfn, |
144 |
dpavlin |
415 |
); |
145 |
|
|
} |
146 |
dpavlin |
410 |
} |
147 |
|
|
|
148 |
dpavlin |
454 |
# print Dump(\@ds); |
149 |
dpavlin |
390 |
|
150 |
dpavlin |
3 |
} |
151 |
|
|
|
152 |
dpavlin |
411 |
foreach my $t (keys %{$thes}) { |
153 |
|
|
|
154 |
|
|
my @e = $thes->{$t}->elements; |
155 |
|
|
if (! @e) { |
156 |
|
|
$log->logwarn("no elements in sorted index $t?"); |
157 |
|
|
next; |
158 |
|
|
} |
159 |
|
|
|
160 |
dpavlin |
415 |
my $file = "./out/bfilter/$t.txt"; |
161 |
|
|
$log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); |
162 |
dpavlin |
411 |
|
163 |
|
|
$webpac->output_file( |
164 |
dpavlin |
415 |
file => $file, |
165 |
dpavlin |
411 |
template => 'index.tt', |
166 |
|
|
data => \@e, |
167 |
|
|
index_name => $t, |
168 |
|
|
); |
169 |
|
|
} |
170 |
|
|
|
171 |
|
|
if (0 && $log->is_debug) { |
172 |
dpavlin |
454 |
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
173 |
|
|
$log->debug("data hash: ",Dump($webpac->{'data'})); |
174 |
dpavlin |
410 |
foreach my $t (keys %{$thes}) { |
175 |
dpavlin |
454 |
$log->debug("thesaurus $t hash: ",Dump($thes->{$t})); |
176 |
dpavlin |
410 |
} |
177 |
dpavlin |
372 |
} |
178 |
dpavlin |
390 |
|
179 |
dpavlin |
424 |
} # XXX if(0) |
180 |
|
|
|
181 |
dpavlin |
454 |
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
182 |
dpavlin |
420 |
|
183 |
dpavlin |
441 |
$log->info("creating tree"); |
184 |
|
|
|
185 |
dpavlin |
419 |
# |
186 |
dpavlin |
441 |
# define tree structure |
187 |
dpavlin |
419 |
# |
188 |
|
|
|
189 |
dpavlin |
424 |
my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup"); |
190 |
dpavlin |
419 |
|
191 |
dpavlin |
424 |
my @tree = ({ |
192 |
|
|
# level 0 |
193 |
|
|
code_arr => sub { sort keys %{$l} }, |
194 |
|
|
filter_code => sub { |
195 |
|
|
my $t = shift; |
196 |
|
|
return $t if ($t =~ s/root://); |
197 |
|
|
}, |
198 |
|
|
lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} }, |
199 |
|
|
lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, |
200 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, |
201 |
dpavlin |
454 |
have_children => sub { return $l->{"a:".$_[0]."::"} }, |
202 |
|
|
have_children_at_level => sub { |
203 |
dpavlin |
471 |
return unless (defined($l->{"code:".$_[1]})); |
204 |
|
|
my $code = shift @{$l->{"code:".$_[1]}}; |
205 |
|
|
print STDERR "## $_[1] -> $code\n"; |
206 |
|
|
return undef unless($code); |
207 |
|
|
return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"})); |
208 |
dpavlin |
454 |
}, |
209 |
dpavlin |
424 |
style => 'display: none', |
210 |
|
|
},{ |
211 |
|
|
# 1 |
212 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
213 |
dpavlin |
424 |
filter_code => sub { shift }, # nop |
214 |
|
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
215 |
|
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
216 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
217 |
dpavlin |
454 |
have_children => sub { return $l->{"a:".$_[1].":"} }, |
218 |
dpavlin |
424 |
style => 'display: none', |
219 |
|
|
},{ |
220 |
|
|
# 2 |
221 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
222 |
dpavlin |
424 |
filter_code => sub { shift }, |
223 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
224 |
dpavlin |
424 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
225 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
226 |
dpavlin |
460 |
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
227 |
dpavlin |
454 |
#style => 'display: none', |
228 |
dpavlin |
425 |
},{ |
229 |
|
|
# 3 u¾i pojam |
230 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
231 |
dpavlin |
425 |
filter_code => sub { shift }, |
232 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
233 |
dpavlin |
426 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
234 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
235 |
dpavlin |
460 |
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
236 |
dpavlin |
426 |
},{ |
237 |
|
|
# 4 |
238 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
239 |
dpavlin |
426 |
filter_code => sub { shift }, |
240 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
241 |
dpavlin |
426 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
242 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
243 |
dpavlin |
460 |
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
244 |
dpavlin |
427 |
},{ |
245 |
|
|
# 5 |
246 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
247 |
dpavlin |
427 |
filter_code => sub { shift }, |
248 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
249 |
dpavlin |
427 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
250 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
251 |
dpavlin |
460 |
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
252 |
dpavlin |
427 |
},{ |
253 |
|
|
# 6 |
254 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
255 |
dpavlin |
427 |
filter_code => sub { shift }, |
256 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
257 |
dpavlin |
427 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
258 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
259 |
dpavlin |
460 |
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
260 |
dpavlin |
427 |
},{ |
261 |
|
|
# 7 |
262 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
263 |
dpavlin |
427 |
filter_code => sub { shift }, |
264 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
265 |
dpavlin |
427 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
266 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
267 |
dpavlin |
460 |
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
268 |
dpavlin |
427 |
},{ |
269 |
|
|
# 8 |
270 |
dpavlin |
454 |
code_arr => sub { @{$_[0]} }, |
271 |
dpavlin |
427 |
filter_code => sub { shift }, |
272 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
273 |
dpavlin |
427 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
274 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
275 |
dpavlin |
460 |
# have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
276 |
dpavlin |
424 |
have_children => sub { 0 }, |
277 |
dpavlin |
454 |
},{ |
278 |
|
|
# 9 - level which is never reached except explicitly |
279 |
dpavlin |
471 |
code_arr => sub { @{$_[0]} }, |
280 |
dpavlin |
454 |
filter_code => sub { shift }, |
281 |
dpavlin |
460 |
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
282 |
dpavlin |
454 |
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
283 |
|
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
284 |
|
|
have_children => sub { 0 }, |
285 |
dpavlin |
471 |
have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) }, |
286 |
dpavlin |
454 |
},{ |
287 |
dpavlin |
424 |
}); |
288 |
dpavlin |
419 |
|
289 |
dpavlin |
441 |
my $tree = new WebPAC::Tree( |
290 |
|
|
dir => './out', |
291 |
|
|
html => 'browse.html', |
292 |
|
|
template => './output_template/tree.tt', |
293 |
|
|
js => 'tree-ids.js', |
294 |
|
|
tree => \@tree, |
295 |
|
|
); |
296 |
dpavlin |
428 |
|
297 |
|
|
$log->info("closing index"); |
298 |
dpavlin |
390 |
$index->close; |
299 |
|
|
|
300 |
dpavlin |
422 |
$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'})); |