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 |
13 |
use Config::IniFiles; |
11 |
dpavlin |
1 |
|
12 |
dpavlin |
10 |
$|=1; |
13 |
dpavlin |
9 |
|
14 |
dpavlin |
13 |
my $config_file = $0; |
15 |
|
|
$config_file =~ s/\.pl$/.conf/; |
16 |
|
|
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
17 |
|
|
|
18 |
dpavlin |
10 |
my $config; |
19 |
|
|
|
20 |
|
|
use index_DBI; # there is no other, right now ;-) |
21 |
|
|
|
22 |
|
|
my $index = new index_DBI(); # open index |
23 |
|
|
|
24 |
dpavlin |
1 |
my %opts; |
25 |
|
|
|
26 |
dpavlin |
7 |
# usage: |
27 |
|
|
# -d directory name |
28 |
|
|
# -m multiple directories |
29 |
|
|
# -q quiet |
30 |
|
|
# -s run swish |
31 |
dpavlin |
1 |
|
32 |
dpavlin |
7 |
getopts('d:m:qs', \%opts); |
33 |
|
|
|
34 |
dpavlin |
13 |
my $db_dir; |
35 |
dpavlin |
1 |
|
36 |
dpavlin |
3 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
37 |
dpavlin |
1 |
|
38 |
dpavlin |
3 |
#print Dumper($config->{indexer}); |
39 |
|
|
#print "-" x 70,"\n"; |
40 |
dpavlin |
1 |
|
41 |
dpavlin |
10 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
42 |
dpavlin |
3 |
|
43 |
dpavlin |
13 |
#my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
44 |
|
|
#my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
45 |
|
|
my $isis_codepage; |
46 |
|
|
my $index_codepage; |
47 |
dpavlin |
10 |
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
48 |
dpavlin |
13 |
my $xml_codepage; |
49 |
dpavlin |
10 |
|
50 |
dpavlin |
3 |
sub isis2xml { |
51 |
|
|
|
52 |
dpavlin |
10 |
use xmlify; |
53 |
|
|
|
54 |
dpavlin |
3 |
my $row = shift @_; |
55 |
dpavlin |
13 |
my $add_xml = shift @_; |
56 |
dpavlin |
3 |
|
57 |
|
|
my $xml; |
58 |
|
|
|
59 |
dpavlin |
10 |
use parse_format; |
60 |
dpavlin |
3 |
|
61 |
dpavlin |
13 |
my $html = ""; # html formatted display output |
62 |
dpavlin |
10 |
|
63 |
dpavlin |
13 |
my %field_usage; # counter for usage of each field |
64 |
|
|
|
65 |
dpavlin |
3 |
foreach my $field (keys %{$config->{indexer}}) { |
66 |
|
|
|
67 |
dpavlin |
13 |
$field_usage{$field}++; |
68 |
|
|
|
69 |
dpavlin |
10 |
my $swish_data = ""; |
70 |
dpavlin |
3 |
my $display_data = ""; |
71 |
|
|
|
72 |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
73 |
|
|
|
74 |
|
|
my $format = $x->{content}; |
75 |
dpavlin |
10 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
76 |
dpavlin |
9 |
$s = 0 if (lc($x->{type}) eq "display"); |
77 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
78 |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
79 |
|
|
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
80 |
dpavlin |
3 |
|
81 |
dpavlin |
20 |
# FIX: this is ugly, UGLY, cludge string is returned |
82 |
|
|
# in UTF8 encoding , but as if source charset |
83 |
|
|
# is ISO8859-1 and not some other. This break other |
84 |
|
|
# character encodings, so we convert it first |
85 |
|
|
# back to ISO8859-1 |
86 |
dpavlin |
10 |
$format = $cludge_codepage->convert($format); |
87 |
dpavlin |
20 |
|
88 |
dpavlin |
10 |
my ($swish,$display) = parse_format($format,$row); |
89 |
|
|
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
90 |
dpavlin |
3 |
|
91 |
dpavlin |
10 |
#print STDERR "swish: $swish<-- display: $display<--\n"; |
92 |
|
|
|
93 |
dpavlin |
20 |
# filter="name" ; filter this field through |
94 |
|
|
# filter/[name].pm |
95 |
|
|
my $filter; |
96 |
|
|
if ($x->{filter}) { |
97 |
|
|
$filter = "filter/".$x->{filter}.".pm"; |
98 |
|
|
require $filter; |
99 |
|
|
} |
100 |
|
|
# type="swish" ; field for swish |
101 |
|
|
if ($s && $swish) { |
102 |
|
|
if ($filter) { |
103 |
|
|
$swish_data .= join(" ",&filter($swish)); |
104 |
|
|
} else { |
105 |
|
|
$swish_data .= $swish if ($s && $swish); |
106 |
|
|
} |
107 |
|
|
} |
108 |
dpavlin |
17 |
|
109 |
dpavlin |
20 |
# type="display" ; field for display |
110 |
|
|
if ($d && $display) { |
111 |
|
|
if ($filter) { |
112 |
|
|
$display_data .= join(" ",&filter($display)); |
113 |
|
|
} else { |
114 |
|
|
$display_data .= $display if ($s && $display); |
115 |
|
|
} |
116 |
|
|
} |
117 |
|
|
|
118 |
|
|
# type="index" ; insert into index |
119 |
dpavlin |
17 |
if ($i && $display) { |
120 |
|
|
my $index_data = $index_codepage->convert($display) || $display; |
121 |
dpavlin |
20 |
if ($filter) { |
122 |
|
|
foreach my $d (&filter($index_data)) { |
123 |
|
|
$index->insert($field, $d, $db_dir); |
124 |
|
|
} |
125 |
|
|
} else { |
126 |
|
|
$index->insert($field, $index_data, $db_dir); |
127 |
|
|
} |
128 |
dpavlin |
17 |
} |
129 |
dpavlin |
3 |
} |
130 |
dpavlin |
9 |
|
131 |
dpavlin |
13 |
|
132 |
dpavlin |
10 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
133 |
|
|
if ($display_data) { |
134 |
dpavlin |
13 |
$display_data = $isis_codepage->convert($display_data) || die "Can't convert '$display_data' !"; |
135 |
|
|
# FIX: this is removed and replaced by html tag. |
136 |
|
|
#$xml .= xmlify($field."_display", $display_data); |
137 |
|
|
|
138 |
|
|
if ($field eq "headline") { |
139 |
|
|
$xml .= xmlify("headline", $display_data); |
140 |
|
|
} else { |
141 |
|
|
|
142 |
|
|
# find field name (signular, plural) |
143 |
|
|
my $field_name = ""; |
144 |
|
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
145 |
|
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
146 |
|
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
147 |
|
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
148 |
dpavlin |
20 |
} elsif ($config->{indexer}->{$field}->{name}) { |
149 |
|
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
150 |
dpavlin |
13 |
} else { |
151 |
dpavlin |
20 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
152 |
dpavlin |
13 |
} |
153 |
|
|
if ($field_name) { |
154 |
|
|
$html .= $xml_codepage->convert($field_name); |
155 |
|
|
} |
156 |
|
|
$html .= $display_data."###\n"; |
157 |
|
|
} |
158 |
dpavlin |
10 |
} |
159 |
|
|
if ($swish_data) { |
160 |
|
|
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
161 |
dpavlin |
20 |
# remove extra spaces |
162 |
|
|
$swish_data =~ s/ +/ /g; |
163 |
|
|
$swish_data =~ s/ +$//g; |
164 |
|
|
|
165 |
dpavlin |
10 |
$swish_data = $i->convert($swish_data); |
166 |
|
|
$xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); |
167 |
|
|
#$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; |
168 |
|
|
#$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); |
169 |
|
|
} |
170 |
|
|
|
171 |
dpavlin |
9 |
|
172 |
dpavlin |
3 |
} |
173 |
dpavlin |
13 |
|
174 |
|
|
# dump formatted output in <html> |
175 |
|
|
if ($html) { |
176 |
|
|
$xml .= xmlify("html",$html); |
177 |
|
|
} |
178 |
|
|
|
179 |
dpavlin |
3 |
if ($xml) { |
180 |
dpavlin |
10 |
#print STDERR "x: $xml\n"; |
181 |
dpavlin |
13 |
$xml .= $add_xml if ($add_xml); |
182 |
dpavlin |
10 |
return "<xml>\n$xml</xml>\n"; |
183 |
dpavlin |
3 |
} else { |
184 |
|
|
return; |
185 |
|
|
} |
186 |
|
|
} |
187 |
|
|
|
188 |
|
|
########################################################################## |
189 |
|
|
|
190 |
dpavlin |
13 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
191 |
dpavlin |
1 |
|
192 |
dpavlin |
13 |
foreach my $database ($cfg->Sections) { |
193 |
dpavlin |
1 |
|
194 |
dpavlin |
13 |
my $isis_db = $cfg -> val($database, 'isis_db'); |
195 |
|
|
my $type = $cfg -> val($database, 'type'); |
196 |
|
|
my $add_xml = $cfg -> val($database, 'xml'); |
197 |
dpavlin |
1 |
|
198 |
dpavlin |
13 |
# read configuration for this type |
199 |
|
|
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
200 |
|
|
$isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
201 |
|
|
$index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
202 |
|
|
$xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),'UTF8'); |
203 |
dpavlin |
1 |
|
204 |
dpavlin |
3 |
my $db = OpenIsis::open( $isis_db ); |
205 |
|
|
if (0) { |
206 |
|
|
# # FIX |
207 |
|
|
# if (! $db ) { |
208 |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
209 |
|
|
next ; |
210 |
|
|
} |
211 |
|
|
|
212 |
dpavlin |
1 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
213 |
|
|
|
214 |
dpavlin |
3 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
215 |
dpavlin |
1 |
|
216 |
dpavlin |
13 |
my $path = $database; # was $isis_db |
217 |
dpavlin |
10 |
|
218 |
dpavlin |
3 |
my $last_p = 0; |
219 |
|
|
|
220 |
dpavlin |
10 |
# { my $row_id = 4514; |
221 |
dpavlin |
3 |
# FIX |
222 |
dpavlin |
1 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
223 |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
224 |
dpavlin |
3 |
if ($row && $row->{mfn}) { |
225 |
dpavlin |
10 |
#print STDERR "mfn: ",$row->{mfn},"\n"; |
226 |
dpavlin |
3 |
# output current process indicator |
227 |
|
|
my $p = int($row->{mfn} * 100 / $max_rowid); |
228 |
|
|
if ($p != $last_p) { |
229 |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
230 |
|
|
$last_p = $p; |
231 |
|
|
} |
232 |
|
|
|
233 |
dpavlin |
13 |
if (my $xml = isis2xml($row,$add_xml)) { |
234 |
dpavlin |
10 |
#print STDERR "--ret-->$xml\n"; |
235 |
|
|
print "Path-Name: $path#".int($row->{mfn})."\n"; |
236 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
237 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
238 |
dpavlin |
3 |
} |
239 |
dpavlin |
1 |
} |
240 |
|
|
} |
241 |
dpavlin |
3 |
print STDERR "\n"; |
242 |
dpavlin |
1 |
} |
243 |
dpavlin |
3 |
|
244 |
dpavlin |
10 |
# call this to commit index |
245 |
|
|
$index->close; |
246 |
dpavlin |
3 |
|
247 |
|
|
1; |
248 |
|
|
__END__ |
249 |
|
|
########################################################################## |
250 |
|
|
|
251 |
|
|
=head1 NAME |
252 |
|
|
|
253 |
|
|
isis2xml.pl - read isis file and dump XML |
254 |
|
|
|
255 |
|
|
=head1 DESCRIPTION |
256 |
|
|
|
257 |
|
|
This command will read ISIS data file using OpenIsis perl module and |
258 |
|
|
create XML file for usage with I<SWISH-E> |
259 |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
260 |
|
|
from isis files (isis allready has something like that). Output of this |
261 |
|
|
script is tailor-made for SWISH-E. |
262 |
|
|
|
263 |
|
|
=head1 AUTHOR |
264 |
|
|
|
265 |
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
266 |
|
|
|
267 |
|
|
=head1 COPYRIGHT |
268 |
|
|
|
269 |
|
|
GNU Public License (GPL) v2 or later |
270 |
|
|
|
271 |
|
|
=head1 SEE ALSO |
272 |
|
|
|
273 |
|
|
SWISH-E web site at http://www.swish-e.org |
274 |
|
|
|
275 |
|
|
=cut |