1 |
#!/usr/bin/perl -w |
#!/usr/bin/perl -w |
2 |
|
|
3 |
use strict; |
=head1 NAME |
|
use OpenIsis; |
|
|
use Getopt::Std; |
|
|
use Data::Dumper; |
|
|
use XML::Simple; |
|
|
use Text::Unaccent; |
|
|
require Unicode::Map8; |
|
|
|
|
|
my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
|
|
|
|
|
my %opts; |
|
|
|
|
|
getopts('d:m:q', \%opts); |
|
|
|
|
|
my $db_dir = $opts{d} || "ps"; # FIX |
|
|
|
|
|
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
|
|
|
|
|
#print Dumper($config->{indexer}); |
|
|
#print "-" x 70,"\n"; |
|
|
|
|
|
# how to convert isis code page to UTF8? |
|
|
my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die; |
|
|
|
|
|
sub isis2xml { |
|
4 |
|
|
5 |
my $row = shift @_; |
all2all.pl - basic script for all WebPAC needs |
6 |
|
|
7 |
my $xml; |
=cut |
8 |
|
|
9 |
sub isis_sf { |
use strict; |
10 |
my $row = shift @_; |
use YAML; |
11 |
my $isis_id = shift @_; |
use Carp; |
12 |
my $subfield = shift @_; |
use Getopt::Long; |
13 |
if ($row->{$isis_id}->[0]) { |
|
14 |
my $sf = OpenIsis::subfields($row->{$isis_id}->[0]); |
use lib './lib'; |
15 |
if (! defined $subfield || length($subfield) == 0) { |
use WebPAC; |
16 |
# subfield list undef, empty or no defined subfields for this record |
use WebPAC::jsFind; |
17 |
return $row->{$isis_id}->[0]; |
use WebPAC::Index; |
18 |
} elsif ($sf->{$subfield}) { |
use WebPAC::Tree; |
19 |
return $sf->{$subfield}; |
|
20 |
} |
# 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 |
|
my $webpac = new WebPAC( |
37 |
|
code_page => $code_page, |
38 |
|
limit_mfn => $limit_mfn, |
39 |
|
start_mfn => $start_mfn, |
40 |
|
debug => $debug, |
41 |
|
low_mem => $low_mem, |
42 |
|
) || die; |
43 |
|
|
44 |
|
my $log = $webpac->_get_logger() || die "can't get logger"; |
45 |
|
|
46 |
|
$log->debug("creating WebPAC::jsFind object"); |
47 |
|
|
48 |
|
my $index = new WebPAC::jsFind( |
49 |
|
index_path => $index_path, |
50 |
|
keys => 10, |
51 |
|
) || die; |
52 |
|
|
53 |
|
my $thes; |
54 |
|
|
55 |
|
$|=1; |
56 |
|
|
57 |
|
my $maxmfn = $webpac->open_isis( |
58 |
|
filename => shift @ARGV || '/data/hidra/THS/THS', |
59 |
|
lookup => [ |
60 |
|
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
61 |
|
# { 'eval' => '"v901^a" eq "Područje"', '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 |
|
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
66 |
|
# tree structure |
67 |
|
{ 'eval' => 'length("v251") == 2', 'key' => 'root:v251', 'val' => 'v900' }, |
68 |
|
{ 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, |
69 |
|
{ 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' }, |
70 |
|
], |
71 |
|
); |
72 |
|
|
73 |
|
$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); |
74 |
|
|
75 |
|
$log->info("rows: $maxmfn"); |
76 |
|
|
77 |
|
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
78 |
|
|
79 |
|
if(1) { # XXX |
80 |
|
|
81 |
|
while (my $rec = $webpac->fetch_rec) { |
82 |
|
|
83 |
|
my @ds = $webpac->data_structure($rec); |
84 |
|
|
85 |
|
if (0 && $log->is_debug) { |
86 |
|
$log->debug("rec = ",Dump($rec)); |
87 |
|
$log->debug("ds = ",Dump(\@ds)); |
88 |
} |
} |
89 |
|
|
90 |
foreach my $field (keys %{$config->{indexer}}) { |
next if (! @ds); |
|
|
|
|
my $display_data = ""; |
|
|
my $index_data = ""; |
|
91 |
|
|
92 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); |
93 |
|
|
94 |
my $display_tmp = ""; |
if ($filename) { |
95 |
my $index_tmp = ""; |
$webpac->output_file( |
96 |
|
file => $filename, |
97 |
my $format = $x->{content}; |
template => 'html.tt', |
98 |
my $i = 1; # index only |
data => \@ds, |
99 |
my $d = 1; # display only |
headline => $webpac->{'headline'}, |
100 |
$i = 0 if (lc($x->{type}) eq "display"); |
); |
|
$d = 0 if (lc($x->{type}) eq "index"); |
|
|
#print "## i: $i d: $d ## $format ##"; |
|
|
# parse format |
|
|
my $prefix = ""; |
|
|
if ($format =~ s/^([^\d]+)//) { |
|
|
$prefix = $1; |
|
|
} |
|
|
while ($format) { |
|
|
if ($format =~ s/^(\d\d\d)(\w?)//) { |
|
|
my $isis_tmp = isis_sf($row,$1,$2); |
|
|
if ($isis_tmp) { |
|
|
# $display_tmp .= $prefix . "/$1/$2/".$isis_tmp if ($d); |
|
|
$display_tmp .= $prefix . $isis_tmp if ($d); |
|
|
$index_tmp .= $isis_tmp." " if ($i); |
|
|
#print " $isis_tmp <--\n"; |
|
|
} |
|
|
$prefix = ""; |
|
|
} elsif ($format =~ s/^([^\d]+)//) { |
|
|
$prefix = $1; |
|
|
} else { |
|
|
print STDERR "WARNING: unparsed format '$format'\n"; |
|
|
last; |
|
|
}; |
|
|
} |
|
|
# add suffix |
|
|
$display_tmp .= $prefix if ($display_tmp); |
|
|
|
|
|
# $display_data .= $display_tmp if ($display_tmp ne ""); |
|
|
# $index_data .= $index_tmp if ($index_tmp ne ""); |
|
|
$display_data .= $display_tmp; |
|
|
$index_data .= $index_tmp; |
|
|
|
|
|
} |
|
|
#print "--display:$display_data\n--index:$index_data\n"; |
|
|
$xml->{$field}->{display} .= $isis_map->tou($display_data)->utf8 if ($display_data); |
|
|
$xml->{$field}->{index} .= unac_string($config->{isis_codepage},$index_data) if ($index_data); |
|
|
|
|
|
} |
|
|
if ($xml) { |
|
|
return XMLout($xml, rootname => 'xml', noattr => 1 ); |
|
101 |
} else { |
} else { |
102 |
return; |
print $webpac->output( |
103 |
|
template => 'text.tt', |
104 |
|
data => \@ds, |
105 |
|
headline => $webpac->{'headline'}, |
106 |
|
); |
107 |
} |
} |
|
} |
|
|
|
|
|
########################################################################## |
|
|
|
|
|
my $last_tell=0; |
|
108 |
|
|
109 |
my @isis_dirs = ( '.' ); # use dirname as database name |
my $headline = $webpac->{'headline'}; |
|
|
|
|
if ($opts{m}) { |
|
|
@isis_dirs = split(/,/,$opts{m}); |
|
|
} |
|
110 |
|
|
111 |
my @isis_dbs; |
my $f = $filename; |
112 |
|
$f =~ s!out/!!; |
113 |
|
|
114 |
foreach (@isis_dirs) { |
# save into index |
115 |
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
foreach my $ds (@ds) { |
116 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
next if (! $ds->{'swish'}); |
117 |
|
|
118 |
|
$index->insert( |
119 |
|
index_name => $ds->{'tag'}, |
120 |
|
path => $f, |
121 |
|
headline => $headline, |
122 |
|
words => join(" ",@{$ds->{'swish'}}) |
123 |
|
); |
124 |
} |
} |
|
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
|
|
} |
|
|
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
|
|
} |
|
|
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
|
|
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
|
|
} |
|
|
} |
|
125 |
|
|
126 |
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
# save into sorted index (thesaurus) |
127 |
|
foreach my $ds (@ds) { |
128 |
|
next if (! $ds->{'index'}); |
129 |
|
|
130 |
|
$thes->{$ds->{'tag'}} ||= new WebPAC::Index; |
131 |
|
|
132 |
|
foreach my $h (@{$ds->{'index'}}) { |
133 |
|
$thes->{$ds->{'tag'}}->insert( |
134 |
|
headline => $h, |
135 |
|
mfn => $webpac->mfn, |
136 |
|
); |
137 |
|
} |
138 |
|
} |
139 |
|
|
140 |
my $db; |
# print Dump(\@ds); |
141 |
|
|
142 |
foreach my $isis_db (@isis_dbs) { |
} |
143 |
|
|
144 |
|
foreach my $t (keys %{$thes}) { |
145 |
|
|
146 |
my $db = OpenIsis::open( $isis_db ); |
my @e = $thes->{$t}->elements; |
147 |
if (0) { |
if (! @e) { |
148 |
# # FIX |
$log->logwarn("no elements in sorted index $t?"); |
149 |
# if (! $db ) { |
next; |
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
|
|
next ; |
|
150 |
} |
} |
151 |
|
|
152 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
my $file = "./out/bfilter/$t.txt"; |
153 |
|
$log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); |
|
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
|
|
|
|
|
my $last_p = 0; |
|
154 |
|
|
155 |
# { my $row_id = 1; |
$webpac->output_file( |
156 |
# FIX |
file => $file, |
157 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
template => 'index.tt', |
158 |
my $row = OpenIsis::read( $db, $row_id ); |
data => \@e, |
159 |
if ($row && $row->{mfn}) { |
index_name => $t, |
160 |
|
); |
161 |
# output current process indicator |
} |
|
my $p = int($row->{mfn} * 100 / $max_rowid); |
|
|
if ($p != $last_p) { |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
|
|
$last_p = $p; |
|
|
} |
|
162 |
|
|
163 |
if (my $xml = isis2xml($row)) { |
if (0 && $log->is_debug) { |
164 |
print "Path-Name: $isis_db#".$row->{mfn}."\n"; |
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
165 |
print "Content-Length: ".(length($xml)+1)."\n"; |
$log->debug("data hash: ",Dump($webpac->{'data'})); |
166 |
print "Document-Type: XML\n\n$xml\n"; |
foreach my $t (keys %{$thes}) { |
167 |
} |
$log->debug("thesaurus $t hash: ",Dump($thes->{$t})); |
|
} |
|
168 |
} |
} |
|
print STDERR "\n"; |
|
169 |
} |
} |
170 |
|
|
171 |
|
} # XXX if(0) |
172 |
|
|
173 |
1; |
$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); |
|
__END__ |
|
|
########################################################################## |
|
|
|
|
|
=head1 NAME |
|
|
|
|
|
isis2xml.pl - read isis file and dump XML |
|
|
|
|
|
=head1 DESCRIPTION |
|
174 |
|
|
175 |
This command will read ISIS data file using OpenIsis perl module and |
$log->info("creating tree"); |
|
create XML file for usage with I<SWISH-E> |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
|
|
from isis files (isis allready has something like that). Output of this |
|
|
script is tailor-made for SWISH-E. |
|
176 |
|
|
177 |
=head1 AUTHOR |
# |
178 |
|
# define tree structure |
179 |
|
# |
180 |
|
|
181 |
|
my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup"); |
182 |
|
|
183 |
|
my @tree = ({ |
184 |
|
# level 0 |
185 |
|
code_arr => sub { sort keys %{$l} }, |
186 |
|
filter_code => sub { |
187 |
|
my $t = shift; |
188 |
|
return $t if ($t =~ s/root://); |
189 |
|
}, |
190 |
|
lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} }, |
191 |
|
lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, |
192 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, |
193 |
|
have_children => sub { return $l->{"a:".$_[0]."::"} }, |
194 |
|
have_children_at_level => sub { |
195 |
|
my $l2 = shift; |
196 |
|
my $l1 = substr($l2,0,1); |
197 |
|
return(9, "a:$l1:$l2:" ) if ($l->{"a:$l1:$l2"}); |
198 |
|
}, |
199 |
|
style => 'display: none', |
200 |
|
},{ |
201 |
|
# 1 |
202 |
|
code_arr => sub { @{$_[0]} }, |
203 |
|
filter_code => sub { shift }, # nop |
204 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
205 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
206 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
207 |
|
have_children => sub { return $l->{"a:".$_[1].":"} }, |
208 |
|
style => 'display: none', |
209 |
|
},{ |
210 |
|
# 2 |
211 |
|
code_arr => sub { @{$_[0]} }, |
212 |
|
filter_code => sub { shift }, |
213 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
214 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
215 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
216 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
217 |
|
#style => 'display: none', |
218 |
|
},{ |
219 |
|
# 3 uži pojam |
220 |
|
code_arr => sub { @{$_[0]} }, |
221 |
|
filter_code => sub { shift }, |
222 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
223 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
224 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
225 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
226 |
|
},{ |
227 |
|
# 4 |
228 |
|
code_arr => sub { @{$_[0]} }, |
229 |
|
filter_code => sub { shift }, |
230 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
231 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
232 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
233 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
234 |
|
},{ |
235 |
|
# 5 |
236 |
|
code_arr => sub { @{$_[0]} }, |
237 |
|
filter_code => sub { shift }, |
238 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
239 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
240 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
241 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
242 |
|
},{ |
243 |
|
# 6 |
244 |
|
code_arr => sub { @{$_[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].":".$_[0]} }, |
250 |
|
},{ |
251 |
|
# 7 |
252 |
|
code_arr => sub { @{$_[0]} }, |
253 |
|
filter_code => sub { shift }, |
254 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
255 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
256 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
257 |
|
have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
258 |
|
},{ |
259 |
|
# 8 |
260 |
|
code_arr => sub { @{$_[0]} }, |
261 |
|
filter_code => sub { shift }, |
262 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
263 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
264 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
265 |
|
# have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, |
266 |
|
have_children => sub { 0 }, |
267 |
|
},{ |
268 |
|
# 9 - level which is never reached except explicitly |
269 |
|
code_arr => sub { @{$l->{$_[0]}} }, |
270 |
|
filter_code => sub { shift }, |
271 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
272 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
273 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
274 |
|
have_children => sub { 0 }, |
275 |
|
have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,"a:".$_[1].":".$_[0]) }, |
276 |
|
},{ |
277 |
|
}); |
278 |
|
|
279 |
|
my $tree = new WebPAC::Tree( |
280 |
|
dir => './out', |
281 |
|
html => 'browse.html', |
282 |
|
template => './output_template/tree.tt', |
283 |
|
js => 'tree-ids.js', |
284 |
|
tree => \@tree, |
285 |
|
); |
286 |
|
|
287 |
Dobrica Pavlinusic <dpavlin@rot13.org> |
$log->info("closing index"); |
288 |
|
$index->close; |
289 |
|
|
290 |
=head1 COPYRIGHT |
$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'})); |
|
|
|
|
GNU Public License (GPL) v2 or later |
|
|
|
|
|
=head1 SEE ALSO |
|
|
|
|
|
SWISH-E web site at http://www.swish-e.org |
|
|
|
|
|
=cut |
|