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 1.02; # 1.01 won't compile on my platform, |
|
|
require Unicode::Map8; |
|
4 |
|
|
5 |
my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
all2all.pl - basic script for all WebPAC needs |
6 |
|
|
7 |
my %opts; |
=cut |
8 |
|
|
9 |
# usage: |
use strict; |
10 |
# -d directory name |
use Data::Dumper; |
11 |
# -m multiple directories |
use Carp; |
|
# -q quiet |
|
|
# -s run swish |
|
12 |
|
|
13 |
getopts('d:m:qs', \%opts); |
use lib './lib'; |
14 |
|
use WebPAC; |
15 |
|
use WebPAC::jsFind; |
16 |
|
use WebPAC::Index; |
17 |
|
|
18 |
my $db_dir = $opts{d} || "ps"; # FIX |
my $webpac = new WebPAC( |
19 |
|
code_page => 'ISO-8859-2', |
20 |
|
limit_mfn => 500, |
21 |
|
# debug => 1, |
22 |
|
# low_mem => 1, |
23 |
|
) || die; |
24 |
|
|
25 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
my $log = $webpac->_get_logger() || die "can't get logger"; |
26 |
|
|
27 |
#print Dumper($config->{indexer}); |
$log->debug("creating WebPAC::jsFind object"); |
|
#print "-" x 70,"\n"; |
|
28 |
|
|
29 |
# how to convert isis code page to UTF8? |
my $index = new WebPAC::jsFind( |
30 |
my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die; |
index_path => './out/index', |
31 |
|
keys => 10, |
32 |
|
) || die; |
33 |
|
|
34 |
sub isis2xml { |
my $thes; |
35 |
|
|
36 |
my $row = shift @_; |
$|=1; |
37 |
|
|
38 |
my $xml; |
my $maxmfn = $webpac->open_isis( |
39 |
$xml->{db_dir} = [ $db_dir ]; # FIX remove? |
filename => shift @ARGV || '/data/hidra/THS/THS', |
40 |
|
lookup => [ |
41 |
|
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
42 |
|
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
43 |
|
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
44 |
|
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
45 |
|
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
46 |
|
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
47 |
|
# tree structure |
48 |
|
{ 'eval' => 'length("v251") == 2', 'key' => 'root:v251', 'val' => 'v900' }, |
49 |
|
{ 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, |
50 |
|
], |
51 |
|
); |
52 |
|
|
53 |
sub isis_sf { |
$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); |
|
my $row = shift @_; |
|
|
my $isis_id = shift @_; |
|
|
my $subfield = shift @_; |
|
|
if ($row->{$isis_id}->[0]) { |
|
|
my $sf = OpenIsis::subfields($row->{$isis_id}->[0]); |
|
|
if (! defined $subfield || length($subfield) == 0) { |
|
|
# subfield list undef, empty or no defined subfields for this record |
|
|
my $all_sf = $row->{$isis_id}->[0]; |
|
|
$all_sf =~ s/\^./ /g; nuke definirions |
|
|
return $all_sf; |
|
|
} elsif ($sf->{$subfield}) { |
|
|
return $sf->{$subfield}; |
|
|
} |
|
|
} |
|
|
} |
|
54 |
|
|
55 |
foreach my $field (keys %{$config->{indexer}}) { |
$log->info("rows: $maxmfn"); |
56 |
|
|
57 |
my $display_data = ""; |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
|
my $index_data = ""; |
|
58 |
|
|
59 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
if(1) { # XXX |
60 |
|
|
61 |
my $display_tmp = ""; |
while (my $rec = $webpac->fetch_rec) { |
|
my $index_tmp = ""; |
|
|
|
|
|
my $format = $x->{content}; |
|
|
my $i = 1; # index only |
|
|
my $d = 1; # display only |
|
|
$i = 0 if (lc($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); |
|
62 |
|
|
63 |
# $display_data .= $display_tmp if ($display_tmp ne ""); |
my @ds = $webpac->data_structure($rec); |
|
# $index_data .= $index_tmp if ($index_tmp ne ""); |
|
|
$display_data .= $display_tmp; |
|
|
$index_data .= $index_tmp; |
|
64 |
|
|
65 |
} |
if (0 && $log->is_debug) { |
66 |
#print "--display:$display_data\n--index:$index_data\n"; |
$log->debug("rec = ",Dumper($rec)); |
67 |
#$xml->{$field."_display"} = $isis_map->tou($display_data)->utf8 if ($display_data); |
$log->debug("ds = ",Dumper(\@ds)); |
|
#$xml->{$field."_index"} = unac_string($config->{isis_codepage},$index_data) if ($index_data); |
|
|
$xml->{$field."_display" } = [ $isis_map->tou($display_data)->utf8 ] if ($display_data); |
|
|
$xml->{$field."_index"} = [ unac_string($config->{isis_codepage},$index_data)." jabuka" ] if ($index_data); |
|
|
|
|
68 |
} |
} |
|
if ($xml) { |
|
|
return XMLout($xml, rootname => 'xml', keeproot => 0, noattr => 0 ); |
|
|
} else { |
|
|
return; |
|
|
} |
|
|
} |
|
69 |
|
|
70 |
########################################################################## |
next if (! @ds); |
71 |
|
|
72 |
my $last_tell=0; |
my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); |
73 |
|
|
74 |
my @isis_dirs = ( '.' ); # use dirname as database name |
if ($filename) { |
75 |
|
$webpac->output_file( |
76 |
|
file => $filename, |
77 |
|
template => 'html.tt', |
78 |
|
data => \@ds, |
79 |
|
headline => $webpac->{'headline'}, |
80 |
|
); |
81 |
|
} else { |
82 |
|
print $webpac->output( |
83 |
|
template => 'text.tt', |
84 |
|
data => \@ds, |
85 |
|
headline => $webpac->{'headline'}, |
86 |
|
); |
87 |
|
} |
88 |
|
|
89 |
if ($opts{m}) { |
my $headline = $webpac->{'headline'}; |
|
@isis_dirs = split(/,/,$opts{m}); |
|
|
} |
|
90 |
|
|
91 |
my @isis_dbs; |
my $f = $filename; |
92 |
|
$f =~ s!out/!!; |
93 |
|
|
94 |
foreach (@isis_dirs) { |
# save into index |
95 |
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
foreach my $ds (@ds) { |
96 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
next if (! $ds->{'swish'}); |
97 |
} |
|
98 |
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
$index->insert( |
99 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
index_name => $ds->{'tag'}, |
100 |
} |
path => $f, |
101 |
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
headline => $headline, |
102 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
words => join(" ",@{$ds->{'swish'}}) |
103 |
|
); |
104 |
} |
} |
|
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
|
|
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
|
|
} |
|
|
} |
|
105 |
|
|
106 |
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) |
107 |
|
foreach my $ds (@ds) { |
108 |
|
next if (! $ds->{'index'}); |
109 |
|
|
110 |
|
$thes->{$ds->{'tag'}} ||= new WebPAC::Index; |
111 |
|
|
112 |
|
foreach my $h (@{$ds->{'index'}}) { |
113 |
|
$thes->{$ds->{'tag'}}->insert( |
114 |
|
path => $f, |
115 |
|
headline => $h, |
116 |
|
); |
117 |
|
} |
118 |
|
} |
119 |
|
|
120 |
my $db; |
# print Dumper(\@ds); |
121 |
|
|
122 |
foreach my $isis_db (@isis_dbs) { |
} |
123 |
|
|
124 |
|
foreach my $t (keys %{$thes}) { |
125 |
|
|
126 |
my $db = OpenIsis::open( $isis_db ); |
my @e = $thes->{$t}->elements; |
127 |
if (0) { |
if (! @e) { |
128 |
# # FIX |
$log->logwarn("no elements in sorted index $t?"); |
129 |
# if (! $db ) { |
next; |
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
|
|
next ; |
|
130 |
} |
} |
131 |
|
|
132 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
my $file = "./out/bfilter/$t.txt"; |
133 |
|
$log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); |
|
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
|
134 |
|
|
135 |
my $last_p = 0; |
$webpac->output_file( |
136 |
|
file => $file, |
137 |
# { my $row_id = 1; |
template => 'index.tt', |
138 |
# FIX |
data => \@e, |
139 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
index_name => $t, |
140 |
my $row = OpenIsis::read( $db, $row_id ); |
); |
141 |
if ($row && $row->{mfn}) { |
} |
|
|
|
|
# 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; |
|
|
} |
|
142 |
|
|
143 |
if (my $xml = isis2xml($row)) { |
if (0 && $log->is_debug) { |
144 |
my $path = $isis_db; |
$log->debug("lookup hash: ",Dumper($webpac->{'lookup'})); |
145 |
$path =~ s#$config->{isis_data}/*##g; |
$log->debug("data hash: ",Dumper($webpac->{'data'})); |
146 |
my $out = "Path-Name: $path#".$row->{mfn}."\n"; |
foreach my $t (keys %{$thes}) { |
147 |
$out .= "Content-Length: ".(length($xml)+1)."\n"; |
$log->debug("thesaurus $t hash: ",Dumper($thes->{$t})); |
|
$out .= "Document-Type: XML\n\n$xml\n"; |
|
|
print $out; |
|
|
} |
|
|
} |
|
148 |
} |
} |
|
print STDERR "\n"; |
|
149 |
} |
} |
150 |
|
|
151 |
|
} # XXX if(0) |
152 |
|
|
153 |
1; |
$log->info("lookup hash: ",Dumper($webpac->{'lookup'})); |
|
__END__ |
|
|
########################################################################## |
|
154 |
|
|
155 |
=head1 NAME |
# |
156 |
|
# tree dump implementation which shouldn't be here :-) |
157 |
|
# |
158 |
|
|
159 |
|
my $file = 'out/browse.html'; |
160 |
|
open(HTML, "> $file") || $log->logdie("can't open '$file'"); |
161 |
|
|
162 |
|
print HTML qq{ |
163 |
|
<html> |
164 |
|
<head> |
165 |
|
<title>Browse</title> |
166 |
|
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-2"/> |
167 |
|
<script type="text/javascript" src="bfilter/hide.js"></script> |
168 |
|
</head> |
169 |
|
<body> |
170 |
|
<ul> |
171 |
|
}; |
172 |
|
|
173 |
|
my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup"); |
174 |
|
|
175 |
|
my @tree = ({ |
176 |
|
# level 0 |
177 |
|
code_arr => sub { sort keys %{$l} }, |
178 |
|
filter_code => sub { |
179 |
|
my $t = shift; |
180 |
|
return $t if ($t =~ s/root://); |
181 |
|
}, |
182 |
|
lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} }, |
183 |
|
lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, |
184 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, |
185 |
|
have_children => sub { $l->{"a:".$_[0]."::" } }, |
186 |
|
child_code => sub { return $_[0] }, |
187 |
|
style => 'display: none', |
188 |
|
},{ |
189 |
|
# 1 |
190 |
|
code_arr => sub { @{$l->{"a:".$_[0]."::"}} }, |
191 |
|
filter_code => sub { shift }, # nop |
192 |
|
lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, |
193 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
194 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
195 |
|
have_children => sub { $l->{"a:".$_[1].":" } }, |
196 |
|
child_code => sub { return $_[1] }, |
197 |
|
style => 'display: none', |
198 |
|
},{ |
199 |
|
# 2 |
200 |
|
code_arr => sub { @{$l->{"a:".$_[0].":"}} }, |
201 |
|
filter_code => sub { shift }, |
202 |
|
lookup_v900 => sub { shift }, |
203 |
|
lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, |
204 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, |
205 |
|
have_children => sub { $l->{"a:".$_[2].":".$_[1]} }, |
206 |
|
child_code => sub { return "a:".$_[2].":".$_[1] }, |
207 |
|
style => 'display: none', |
208 |
|
},{ |
209 |
|
# 3 uži pojam |
210 |
|
code_arr => sub { @{$l->{$_[0]}} }, |
211 |
|
filter_code => sub { shift }, |
212 |
|
lookup_v900 => sub { shift }, |
213 |
|
lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, |
214 |
|
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, |
215 |
|
have_children => sub { 0 }, |
216 |
|
child_code => sub { 0 }, |
217 |
|
}); |
218 |
|
|
219 |
|
unroll(0,''); |
220 |
|
|
221 |
|
$log->debug("test filter: ",$tree[0]->{'filter_code'}->("root:99")); |
222 |
|
|
223 |
|
sub unroll { |
224 |
|
my ($level,$start_code) = @_; |
225 |
|
|
226 |
|
$log->logconfess("need level") unless (defined($level)); |
227 |
|
|
228 |
|
# all levels passed? |
229 |
|
return if (! defined($tree[$level])); |
230 |
|
|
231 |
|
$log->debug("unroll level $level, start code $start_code"); |
232 |
|
|
233 |
|
foreach my $code ($tree[$level]->{'code_arr'}->($start_code)) { |
234 |
|
|
235 |
|
if ($code = $tree[$level]->{'filter_code'}->($code)) { |
236 |
|
|
237 |
|
$log->debug("# $level filter passed code $code"); |
238 |
|
|
239 |
|
my $v900 = $tree[$level]->{'lookup_v900'}->($code) || $log->logdie("can't lookup_v900 '$code'"); |
240 |
|
$log->debug("# $level lookup_v900($code) = $v900"); |
241 |
|
my $term = $tree[$level]->{'lookup_term'}->($code,$v900) || $log->logdie("can't lookup_term '$v900'"); |
242 |
|
$log->debug("# $level lookup_term($code,$v900) = $term"); |
243 |
|
my $mfn = $tree[$level]->{'lookup_mfn'}->($code,$v900) || $log->logdie("can't lookup_mfn '$v900'"); |
244 |
|
$log->debug("# $level lookup_mfn($code,$v900) = $mfn"); |
245 |
|
|
246 |
isis2xml.pl - read isis file and dump XML |
$log->debug("$code -> $v900 : $term [$mfn]"); |
247 |
|
|
248 |
=head1 DESCRIPTION |
|
249 |
|
my $have_children = $tree[$level]->{'have_children'}->($code,$v900,$start_code); |
250 |
|
|
251 |
This command will read ISIS data file using OpenIsis perl module and |
my ($link_start,$link_end) = ('',''); |
|
create XML file for usage with I<SWISH-E> |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
|
|
from isis files (isis allready has something like that). Output of this |
|
|
script is tailor-made for SWISH-E. |
|
252 |
|
|
253 |
=head1 AUTHOR |
($link_start,$link_end) = (qq{<a href="#mfn$mfn" onClick="return hide_show('mfn$mfn');">},qq{</a>}) if ($have_children); |
254 |
|
|
255 |
|
my $mfn_link = "thes/$mfn.html"; |
256 |
|
if (-e "out/$mfn_link") { |
257 |
|
print HTML " " x $level . |
258 |
|
qq{<li>${link_start}${term}${link_end}}. |
259 |
|
qq{ <a href="$mfn_link">»</a></li>\n}; |
260 |
|
} else { |
261 |
|
$log->warn("file 'out/$mfn_link' doesn't exist, skipping"); |
262 |
|
} |
263 |
|
|
264 |
Dobrica Pavlinusic <dpavlin@rot13.org> |
unless ($have_children) { |
265 |
|
$log->warn("doesn't have_children($code,$v900,$start_code)"); |
266 |
|
next; |
267 |
|
} |
268 |
|
|
269 |
|
print HTML " " x $level . |
270 |
|
qq{<a name="mfn$mfn"></a>\n <ul id="mfn$mfn"}. |
271 |
|
($tree[$level]->{'style'} ? ' style="'.$tree[$level]->{'style'}.'"' : ''). |
272 |
|
qq{>\n}; |
273 |
|
|
274 |
|
unroll($level+1, $tree[$level]->{'child_code'}->($code,$v900,$start_code)); |
275 |
|
|
276 |
|
print HTML " " x $level . qq{</ul>\n}; |
277 |
|
|
278 |
=head1 COPYRIGHT |
} |
279 |
|
} |
280 |
|
} |
281 |
|
|
282 |
GNU Public License (GPL) v2 or later |
print HTML qq{ |
283 |
|
</ul> |
284 |
|
</body> |
285 |
|
</html> |
286 |
|
}; |
287 |
|
|
288 |
=head1 SEE ALSO |
close(HTML); |
289 |
|
|
290 |
SWISH-E web site at http://www.swish-e.org |
$index->close; |
291 |
|
|
292 |
=cut |
$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'})); |