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