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 |
10 |
use Text::Iconv; |
10 |
dpavlin |
1 |
|
11 |
|
|
|
12 |
dpavlin |
10 |
$|=1; |
13 |
dpavlin |
9 |
|
14 |
dpavlin |
10 |
my $config; |
15 |
|
|
|
16 |
|
|
$config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
17 |
|
|
|
18 |
|
|
use index_DBI; # there is no other, right now ;-) |
19 |
|
|
|
20 |
|
|
my $index = new index_DBI(); # open index |
21 |
|
|
|
22 |
dpavlin |
1 |
my %opts; |
23 |
|
|
|
24 |
dpavlin |
7 |
# usage: |
25 |
|
|
# -d directory name |
26 |
|
|
# -m multiple directories |
27 |
|
|
# -q quiet |
28 |
|
|
# -s run swish |
29 |
dpavlin |
1 |
|
30 |
dpavlin |
7 |
getopts('d:m:qs', \%opts); |
31 |
|
|
|
32 |
dpavlin |
3 |
my $db_dir = $opts{d} || "ps"; # FIX |
33 |
dpavlin |
1 |
|
34 |
dpavlin |
3 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
35 |
dpavlin |
1 |
|
36 |
dpavlin |
3 |
#print Dumper($config->{indexer}); |
37 |
|
|
#print "-" x 70,"\n"; |
38 |
dpavlin |
1 |
|
39 |
dpavlin |
10 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
40 |
dpavlin |
3 |
|
41 |
dpavlin |
10 |
my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
42 |
|
|
my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
43 |
|
|
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
44 |
|
|
|
45 |
dpavlin |
3 |
sub isis2xml { |
46 |
|
|
|
47 |
dpavlin |
10 |
use xmlify; |
48 |
|
|
|
49 |
dpavlin |
3 |
my $row = shift @_; |
50 |
|
|
|
51 |
|
|
my $xml; |
52 |
dpavlin |
10 |
$xml .= xmlify('db_dir',$db_dir); # FIX remove? |
53 |
dpavlin |
3 |
|
54 |
dpavlin |
10 |
use parse_format; |
55 |
dpavlin |
3 |
|
56 |
dpavlin |
10 |
|
57 |
dpavlin |
3 |
foreach my $field (keys %{$config->{indexer}}) { |
58 |
|
|
|
59 |
dpavlin |
10 |
my $swish_data = ""; |
60 |
dpavlin |
3 |
my $display_data = ""; |
61 |
|
|
my $index_data = ""; |
62 |
|
|
|
63 |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
64 |
|
|
|
65 |
|
|
my $format = $x->{content}; |
66 |
dpavlin |
10 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
67 |
dpavlin |
9 |
$s = 0 if (lc($x->{type}) eq "display"); |
68 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
69 |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
70 |
|
|
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
71 |
dpavlin |
3 |
|
72 |
dpavlin |
10 |
$format = $cludge_codepage->convert($format); |
73 |
|
|
my ($swish,$display) = parse_format($format,$row); |
74 |
|
|
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
75 |
dpavlin |
3 |
|
76 |
dpavlin |
10 |
#print STDERR "swish: $swish<-- display: $display<--\n"; |
77 |
|
|
# FIX: this is ugly, UGLY, cludge: OpenIsis return |
78 |
|
|
# UTF8 encoding of strings, but as if source charset |
79 |
|
|
# is ISO8859-1 and not some other. This breaks our |
80 |
|
|
# isis character encoding, so we convert it first |
81 |
|
|
# back to ISO8859-1 (which can actually be different |
82 |
|
|
# encoding in isis) |
83 |
|
|
|
84 |
|
|
$swish_data .= $swish if ($s && $swish); |
85 |
|
|
$display_data .= $display if ($d && $display); |
86 |
|
|
$index_data .= $display if ($i && $display); |
87 |
dpavlin |
3 |
} |
88 |
dpavlin |
9 |
|
89 |
dpavlin |
10 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
90 |
|
|
if ($display_data) { |
91 |
|
|
$display_data = $isis_codepage->convert($display_data)."##" || $display_data; |
92 |
|
|
$xml .= xmlify($field."_display", $display_data); |
93 |
|
|
} |
94 |
|
|
if ($swish_data) { |
95 |
|
|
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
96 |
|
|
$swish_data = $i->convert($swish_data); |
97 |
|
|
$xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); |
98 |
|
|
#$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; |
99 |
|
|
#$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); |
100 |
|
|
} |
101 |
|
|
|
102 |
dpavlin |
9 |
# index |
103 |
|
|
if ($index_data && $index_data ne "") { |
104 |
dpavlin |
10 |
$index_data = $index_codepage->convert($index_data) || $index_data; |
105 |
|
|
$index->insert($field, $index_data, $db_dir); |
106 |
dpavlin |
9 |
} |
107 |
|
|
|
108 |
dpavlin |
3 |
} |
109 |
|
|
if ($xml) { |
110 |
dpavlin |
10 |
#print STDERR "x: $xml\n"; |
111 |
|
|
return "<xml>\n$xml</xml>\n"; |
112 |
dpavlin |
3 |
} else { |
113 |
|
|
return; |
114 |
|
|
} |
115 |
|
|
} |
116 |
|
|
|
117 |
|
|
########################################################################## |
118 |
|
|
|
119 |
dpavlin |
1 |
my $last_tell=0; |
120 |
|
|
|
121 |
|
|
my @isis_dirs = ( '.' ); # use dirname as database name |
122 |
|
|
|
123 |
|
|
if ($opts{m}) { |
124 |
|
|
@isis_dirs = split(/,/,$opts{m}); |
125 |
|
|
} |
126 |
|
|
|
127 |
|
|
my @isis_dbs; |
128 |
|
|
|
129 |
|
|
foreach (@isis_dirs) { |
130 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
131 |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
132 |
dpavlin |
1 |
} |
133 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
134 |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
135 |
dpavlin |
1 |
} |
136 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
137 |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
138 |
dpavlin |
1 |
} |
139 |
dpavlin |
3 |
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
140 |
|
|
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
141 |
dpavlin |
1 |
} |
142 |
|
|
} |
143 |
|
|
|
144 |
dpavlin |
3 |
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
145 |
|
|
|
146 |
|
|
my $db; |
147 |
|
|
|
148 |
dpavlin |
1 |
foreach my $isis_db (@isis_dbs) { |
149 |
|
|
|
150 |
|
|
|
151 |
dpavlin |
3 |
my $db = OpenIsis::open( $isis_db ); |
152 |
|
|
if (0) { |
153 |
|
|
# # FIX |
154 |
|
|
# if (! $db ) { |
155 |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
156 |
|
|
next ; |
157 |
|
|
} |
158 |
|
|
|
159 |
dpavlin |
1 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
160 |
|
|
|
161 |
dpavlin |
3 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
162 |
dpavlin |
1 |
|
163 |
dpavlin |
10 |
my $path = $isis_db; |
164 |
|
|
$path =~ s#$config->{isis_data}/*##g; |
165 |
|
|
|
166 |
dpavlin |
3 |
my $last_p = 0; |
167 |
|
|
|
168 |
dpavlin |
10 |
# { my $row_id = 4514; |
169 |
dpavlin |
3 |
# FIX |
170 |
dpavlin |
1 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
171 |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
172 |
dpavlin |
3 |
if ($row && $row->{mfn}) { |
173 |
dpavlin |
10 |
#print STDERR "mfn: ",$row->{mfn},"\n"; |
174 |
dpavlin |
3 |
# output current process indicator |
175 |
|
|
my $p = int($row->{mfn} * 100 / $max_rowid); |
176 |
|
|
if ($p != $last_p) { |
177 |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
178 |
|
|
$last_p = $p; |
179 |
|
|
} |
180 |
|
|
|
181 |
|
|
if (my $xml = isis2xml($row)) { |
182 |
dpavlin |
10 |
#print STDERR "--ret-->$xml\n"; |
183 |
|
|
print "Path-Name: $path#".int($row->{mfn})."\n"; |
184 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
185 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
186 |
dpavlin |
3 |
} |
187 |
dpavlin |
1 |
} |
188 |
|
|
} |
189 |
dpavlin |
3 |
print STDERR "\n"; |
190 |
dpavlin |
1 |
} |
191 |
dpavlin |
3 |
|
192 |
dpavlin |
10 |
# call this to commit index |
193 |
|
|
$index->close; |
194 |
dpavlin |
3 |
|
195 |
|
|
1; |
196 |
|
|
__END__ |
197 |
|
|
########################################################################## |
198 |
|
|
|
199 |
|
|
=head1 NAME |
200 |
|
|
|
201 |
|
|
isis2xml.pl - read isis file and dump XML |
202 |
|
|
|
203 |
|
|
=head1 DESCRIPTION |
204 |
|
|
|
205 |
|
|
This command will read ISIS data file using OpenIsis perl module and |
206 |
|
|
create XML file for usage with I<SWISH-E> |
207 |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
208 |
|
|
from isis files (isis allready has something like that). Output of this |
209 |
|
|
script is tailor-made for SWISH-E. |
210 |
|
|
|
211 |
|
|
=head1 AUTHOR |
212 |
|
|
|
213 |
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
214 |
|
|
|
215 |
|
|
=head1 COPYRIGHT |
216 |
|
|
|
217 |
|
|
GNU Public License (GPL) v2 or later |
218 |
|
|
|
219 |
|
|
=head1 SEE ALSO |
220 |
|
|
|
221 |
|
|
SWISH-E web site at http://www.swish-e.org |
222 |
|
|
|
223 |
|
|
=cut |