1 |
dpavlin |
1 |
#!/usr/bin/perl -w |
2 |
|
|
|
3 |
|
|
use strict; |
4 |
|
|
use OpenIsis; |
5 |
|
|
use Getopt::Std; |
6 |
|
|
use Data::Dumper; |
7 |
|
|
use XML::Simple; |
8 |
dpavlin |
5 |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
9 |
dpavlin |
3 |
require Unicode::Map8; |
10 |
dpavlin |
1 |
|
11 |
dpavlin |
3 |
my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
12 |
dpavlin |
1 |
|
13 |
|
|
my %opts; |
14 |
|
|
|
15 |
dpavlin |
7 |
# usage: |
16 |
|
|
# -d directory name |
17 |
|
|
# -m multiple directories |
18 |
|
|
# -q quiet |
19 |
|
|
# -s run swish |
20 |
dpavlin |
1 |
|
21 |
dpavlin |
7 |
getopts('d:m:qs', \%opts); |
22 |
|
|
|
23 |
dpavlin |
3 |
my $db_dir = $opts{d} || "ps"; # FIX |
24 |
dpavlin |
1 |
|
25 |
dpavlin |
3 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
26 |
dpavlin |
1 |
|
27 |
dpavlin |
3 |
#print Dumper($config->{indexer}); |
28 |
|
|
#print "-" x 70,"\n"; |
29 |
dpavlin |
1 |
|
30 |
dpavlin |
3 |
# how to convert isis code page to UTF8? |
31 |
|
|
my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die; |
32 |
|
|
|
33 |
|
|
sub isis2xml { |
34 |
|
|
|
35 |
|
|
my $row = shift @_; |
36 |
|
|
|
37 |
|
|
my $xml; |
38 |
dpavlin |
7 |
$xml->{db_dir} = [ $db_dir ]; # FIX remove? |
39 |
dpavlin |
3 |
|
40 |
|
|
sub isis_sf { |
41 |
|
|
my $row = shift @_; |
42 |
|
|
my $isis_id = shift @_; |
43 |
|
|
my $subfield = shift @_; |
44 |
|
|
if ($row->{$isis_id}->[0]) { |
45 |
|
|
my $sf = OpenIsis::subfields($row->{$isis_id}->[0]); |
46 |
|
|
if (! defined $subfield || length($subfield) == 0) { |
47 |
|
|
# subfield list undef, empty or no defined subfields for this record |
48 |
dpavlin |
4 |
my $all_sf = $row->{$isis_id}->[0]; |
49 |
|
|
$all_sf =~ s/\^./ /g; nuke definirions |
50 |
|
|
return $all_sf; |
51 |
dpavlin |
3 |
} elsif ($sf->{$subfield}) { |
52 |
|
|
return $sf->{$subfield}; |
53 |
|
|
} |
54 |
|
|
} |
55 |
|
|
} |
56 |
|
|
|
57 |
|
|
foreach my $field (keys %{$config->{indexer}}) { |
58 |
|
|
|
59 |
|
|
my $display_data = ""; |
60 |
|
|
my $index_data = ""; |
61 |
|
|
|
62 |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
63 |
|
|
|
64 |
|
|
my $display_tmp = ""; |
65 |
|
|
my $index_tmp = ""; |
66 |
|
|
|
67 |
|
|
my $format = $x->{content}; |
68 |
|
|
my $i = 1; # index only |
69 |
|
|
my $d = 1; # display only |
70 |
|
|
$i = 0 if (lc($x->{type}) eq "display"); |
71 |
|
|
$d = 0 if (lc($x->{type}) eq "index"); |
72 |
|
|
#print "## i: $i d: $d ## $format ##"; |
73 |
|
|
# parse format |
74 |
|
|
my $prefix = ""; |
75 |
|
|
if ($format =~ s/^([^\d]+)//) { |
76 |
|
|
$prefix = $1; |
77 |
|
|
} |
78 |
|
|
while ($format) { |
79 |
|
|
if ($format =~ s/^(\d\d\d)(\w?)//) { |
80 |
|
|
my $isis_tmp = isis_sf($row,$1,$2); |
81 |
|
|
if ($isis_tmp) { |
82 |
|
|
# $display_tmp .= $prefix . "/$1/$2/".$isis_tmp if ($d); |
83 |
|
|
$display_tmp .= $prefix . $isis_tmp if ($d); |
84 |
|
|
$index_tmp .= $isis_tmp." " if ($i); |
85 |
|
|
#print " $isis_tmp <--\n"; |
86 |
|
|
} |
87 |
|
|
$prefix = ""; |
88 |
|
|
} elsif ($format =~ s/^([^\d]+)//) { |
89 |
|
|
$prefix = $1; |
90 |
|
|
} else { |
91 |
|
|
print STDERR "WARNING: unparsed format '$format'\n"; |
92 |
|
|
last; |
93 |
|
|
}; |
94 |
|
|
} |
95 |
|
|
# add suffix |
96 |
|
|
$display_tmp .= $prefix if ($display_tmp); |
97 |
|
|
|
98 |
|
|
# $display_data .= $display_tmp if ($display_tmp ne ""); |
99 |
|
|
# $index_data .= $index_tmp if ($index_tmp ne ""); |
100 |
|
|
$display_data .= $display_tmp; |
101 |
|
|
$index_data .= $index_tmp; |
102 |
|
|
|
103 |
|
|
} |
104 |
|
|
#print "--display:$display_data\n--index:$index_data\n"; |
105 |
dpavlin |
7 |
#$xml->{$field."_display"} = $isis_map->tou($display_data)->utf8 if ($display_data); |
106 |
|
|
#$xml->{$field."_index"} = unac_string($config->{isis_codepage},$index_data) if ($index_data); |
107 |
|
|
$xml->{$field."_display" } = [ $isis_map->tou($display_data)->utf8 ] if ($display_data); |
108 |
|
|
$xml->{$field."_index"} = [ unac_string($config->{isis_codepage},$index_data)." jabuka" ] if ($index_data); |
109 |
dpavlin |
3 |
|
110 |
|
|
} |
111 |
|
|
if ($xml) { |
112 |
dpavlin |
7 |
return XMLout($xml, rootname => 'xml', keeproot => 0, noattr => 0 ); |
113 |
dpavlin |
3 |
} else { |
114 |
|
|
return; |
115 |
|
|
} |
116 |
|
|
} |
117 |
|
|
|
118 |
|
|
########################################################################## |
119 |
|
|
|
120 |
dpavlin |
1 |
my $last_tell=0; |
121 |
|
|
|
122 |
|
|
my @isis_dirs = ( '.' ); # use dirname as database name |
123 |
|
|
|
124 |
|
|
if ($opts{m}) { |
125 |
|
|
@isis_dirs = split(/,/,$opts{m}); |
126 |
|
|
} |
127 |
|
|
|
128 |
|
|
my @isis_dbs; |
129 |
|
|
|
130 |
|
|
foreach (@isis_dirs) { |
131 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
132 |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
133 |
dpavlin |
1 |
} |
134 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
135 |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
136 |
dpavlin |
1 |
} |
137 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
138 |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
139 |
dpavlin |
1 |
} |
140 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
141 |
|
|
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
142 |
dpavlin |
1 |
} |
143 |
|
|
} |
144 |
|
|
|
145 |
dpavlin |
3 |
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
146 |
|
|
|
147 |
|
|
my $db; |
148 |
|
|
|
149 |
dpavlin |
1 |
foreach my $isis_db (@isis_dbs) { |
150 |
|
|
|
151 |
|
|
|
152 |
dpavlin |
3 |
my $db = OpenIsis::open( $isis_db ); |
153 |
|
|
if (0) { |
154 |
|
|
# # FIX |
155 |
|
|
# if (! $db ) { |
156 |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
157 |
|
|
next ; |
158 |
|
|
} |
159 |
|
|
|
160 |
dpavlin |
1 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
161 |
|
|
|
162 |
dpavlin |
3 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
163 |
dpavlin |
1 |
|
164 |
dpavlin |
3 |
my $last_p = 0; |
165 |
|
|
|
166 |
|
|
# { my $row_id = 1; |
167 |
|
|
# FIX |
168 |
dpavlin |
1 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
169 |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
170 |
dpavlin |
3 |
if ($row && $row->{mfn}) { |
171 |
dpavlin |
1 |
|
172 |
dpavlin |
3 |
# output current process indicator |
173 |
|
|
my $p = int($row->{mfn} * 100 / $max_rowid); |
174 |
|
|
if ($p != $last_p) { |
175 |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
176 |
|
|
$last_p = $p; |
177 |
|
|
} |
178 |
|
|
|
179 |
|
|
if (my $xml = isis2xml($row)) { |
180 |
dpavlin |
7 |
my $path = $isis_db; |
181 |
|
|
$path =~ s#$config->{isis_data}/*##g; |
182 |
|
|
my $out = "Path-Name: $path#".$row->{mfn}."\n"; |
183 |
|
|
$out .= "Content-Length: ".(length($xml)+1)."\n"; |
184 |
|
|
$out .= "Document-Type: XML\n\n$xml\n"; |
185 |
|
|
print $out; |
186 |
dpavlin |
3 |
} |
187 |
dpavlin |
1 |
} |
188 |
|
|
} |
189 |
dpavlin |
3 |
print STDERR "\n"; |
190 |
dpavlin |
1 |
} |
191 |
dpavlin |
3 |
|
192 |
|
|
|
193 |
|
|
1; |
194 |
|
|
__END__ |
195 |
|
|
########################################################################## |
196 |
|
|
|
197 |
|
|
=head1 NAME |
198 |
|
|
|
199 |
|
|
isis2xml.pl - read isis file and dump XML |
200 |
|
|
|
201 |
|
|
=head1 DESCRIPTION |
202 |
|
|
|
203 |
|
|
This command will read ISIS data file using OpenIsis perl module and |
204 |
|
|
create XML file for usage with I<SWISH-E> |
205 |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
206 |
|
|
from isis files (isis allready has something like that). Output of this |
207 |
|
|
script is tailor-made for SWISH-E. |
208 |
|
|
|
209 |
|
|
=head1 AUTHOR |
210 |
|
|
|
211 |
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
212 |
|
|
|
213 |
|
|
=head1 COPYRIGHT |
214 |
|
|
|
215 |
|
|
GNU Public License (GPL) v2 or later |
216 |
|
|
|
217 |
|
|
=head1 SEE ALSO |
218 |
|
|
|
219 |
|
|
SWISH-E web site at http://www.swish-e.org |
220 |
|
|
|
221 |
|
|
=cut |