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; |
44 |
|
|
my $index_codepage; |
45 |
dpavlin |
10 |
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
46 |
dpavlin |
13 |
my $xml_codepage; |
47 |
dpavlin |
10 |
|
48 |
dpavlin |
29 |
my $XML_CHARSET = 'UTF8'; |
49 |
|
|
|
50 |
|
|
|
51 |
dpavlin |
3 |
sub isis2xml { |
52 |
|
|
|
53 |
dpavlin |
10 |
use xmlify; |
54 |
|
|
|
55 |
dpavlin |
3 |
my $row = shift @_; |
56 |
dpavlin |
13 |
my $add_xml = shift @_; |
57 |
dpavlin |
3 |
|
58 |
|
|
my $xml; |
59 |
|
|
|
60 |
dpavlin |
10 |
use parse_format; |
61 |
dpavlin |
3 |
|
62 |
dpavlin |
13 |
my $html = ""; # html formatted display output |
63 |
dpavlin |
10 |
|
64 |
dpavlin |
13 |
my %field_usage; # counter for usage of each field |
65 |
|
|
|
66 |
dpavlin |
32 |
# sort subrouting using order="" attribute |
67 |
|
|
sub by_order { |
68 |
|
|
return 0 if (! $config->{indexer}->{$a}->{order}); |
69 |
|
|
return 0 if (! $config->{indexer}->{$b}->{order}); |
70 |
dpavlin |
29 |
|
71 |
dpavlin |
32 |
return $config->{indexer}->{$a}->{order} <=> |
72 |
|
|
$config->{indexer}->{$b}->{order} ; |
73 |
|
|
} |
74 |
dpavlin |
3 |
|
75 |
dpavlin |
32 |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
76 |
|
|
|
77 |
dpavlin |
13 |
$field_usage{$field}++; |
78 |
|
|
|
79 |
dpavlin |
10 |
my $swish_data = ""; |
80 |
dpavlin |
3 |
my $display_data = ""; |
81 |
dpavlin |
35 |
my $line_delimiter; |
82 |
dpavlin |
3 |
|
83 |
dpavlin |
34 |
my ($swish,$display); |
84 |
|
|
|
85 |
dpavlin |
3 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
86 |
|
|
|
87 |
|
|
my $format = $x->{content}; |
88 |
dpavlin |
29 |
my $delimiter = $x->{delimiter} || ' '; |
89 |
dpavlin |
3 |
|
90 |
dpavlin |
20 |
# FIX: this is ugly, UGLY, cludge string is returned |
91 |
|
|
# in UTF8 encoding , but as if source charset |
92 |
|
|
# is ISO8859-1 and not some other. This break other |
93 |
|
|
# character encodings, so we convert it first |
94 |
|
|
# back to ISO8859-1 |
95 |
dpavlin |
29 |
$format = $xml_codepage->convert($format); |
96 |
|
|
$delimiter = $xml_codepage->convert($delimiter) if ($delimiter); |
97 |
dpavlin |
20 |
|
98 |
dpavlin |
29 |
my $isis_i = 0; # isis repeatable offset |
99 |
|
|
|
100 |
|
|
my ($s,$d,$i) = (1,1,0); # swish, display default |
101 |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
102 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
103 |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
104 |
|
|
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
105 |
|
|
|
106 |
dpavlin |
35 |
# what will separate last line from this one? |
107 |
|
|
if ($display_data && $x->{append} && $x->{append} eq "1") { |
108 |
dpavlin |
34 |
$line_delimiter = ' '; |
109 |
dpavlin |
35 |
} elsif ($display_data) { |
110 |
dpavlin |
34 |
$line_delimiter = '<br/>'; |
111 |
|
|
} |
112 |
dpavlin |
29 |
|
113 |
dpavlin |
34 |
# init vars so that we go into while... |
114 |
|
|
($swish,$display) = (1,1); |
115 |
dpavlin |
29 |
|
116 |
|
|
while ($swish || $display) { |
117 |
|
|
($swish,$display) = parse_format($format,$row,$isis_i++); |
118 |
dpavlin |
10 |
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
119 |
dpavlin |
3 |
|
120 |
dpavlin |
10 |
#print STDERR "swish: $swish<-- display: $display<--\n"; |
121 |
|
|
|
122 |
dpavlin |
29 |
# filter="name" ; filter this field through |
123 |
|
|
# filter/[name].pm |
124 |
|
|
my $filter = $x->{filter}; |
125 |
dpavlin |
20 |
if ($filter) { |
126 |
dpavlin |
29 |
require "filter/".$filter.".pm"; |
127 |
dpavlin |
20 |
} |
128 |
dpavlin |
29 |
# type="swish" ; field for swish |
129 |
|
|
if ($s && $swish) { |
130 |
|
|
if ($filter) { |
131 |
dpavlin |
32 |
#print STDERR "using filter '$filter'\n"; |
132 |
dpavlin |
29 |
no strict 'refs'; |
133 |
|
|
$swish_data .= join(" ",&$filter($swish)); |
134 |
|
|
} else { |
135 |
|
|
$swish_data .= $swish; |
136 |
|
|
} |
137 |
|
|
} |
138 |
dpavlin |
17 |
|
139 |
dpavlin |
29 |
# type="display" ; field for display |
140 |
|
|
if ($d && $display) { |
141 |
dpavlin |
35 |
if ($line_delimiter && $display_data) { |
142 |
|
|
$display_data .= $line_delimiter; |
143 |
|
|
undef $line_delimiter; |
144 |
|
|
} |
145 |
dpavlin |
29 |
if ($filter) { |
146 |
|
|
no strict 'refs'; |
147 |
|
|
$display_data .= join($delimiter,&$filter($display)); |
148 |
|
|
} else { |
149 |
|
|
if ($display_data) { |
150 |
|
|
$display_data .= $delimiter.$display; |
151 |
|
|
} else { |
152 |
|
|
$display_data .= $display; |
153 |
|
|
} |
154 |
|
|
} |
155 |
dpavlin |
20 |
} |
156 |
dpavlin |
29 |
|
157 |
|
|
# type="index" ; insert into index |
158 |
|
|
if ($i && $display) { |
159 |
|
|
my $index_data = $index_codepage->convert($display) || $display; |
160 |
|
|
if ($filter) { |
161 |
|
|
no strict 'refs'; |
162 |
|
|
foreach my $d (&$filter($index_data)) { |
163 |
|
|
$index->insert($field, $d, $db_dir); |
164 |
|
|
} |
165 |
|
|
} else { |
166 |
|
|
$index->insert($field, $index_data, $db_dir); |
167 |
dpavlin |
20 |
} |
168 |
|
|
} |
169 |
dpavlin |
17 |
} |
170 |
dpavlin |
3 |
} |
171 |
dpavlin |
9 |
|
172 |
dpavlin |
13 |
|
173 |
dpavlin |
10 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
174 |
|
|
if ($display_data) { |
175 |
dpavlin |
29 |
|
176 |
dpavlin |
35 |
# $display_data = $isis_codepage->convert($display_data) || die "Can't convert '$display_data' !"; |
177 |
dpavlin |
13 |
# FIX: this is removed and replaced by html tag. |
178 |
|
|
#$xml .= xmlify($field."_display", $display_data); |
179 |
|
|
|
180 |
|
|
if ($field eq "headline") { |
181 |
|
|
$xml .= xmlify("headline", $display_data); |
182 |
|
|
} else { |
183 |
|
|
|
184 |
|
|
# find field name (signular, plural) |
185 |
|
|
my $field_name = ""; |
186 |
|
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
187 |
|
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
188 |
|
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
189 |
|
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
190 |
dpavlin |
20 |
} elsif ($config->{indexer}->{$field}->{name}) { |
191 |
|
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
192 |
dpavlin |
13 |
} else { |
193 |
dpavlin |
20 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
194 |
dpavlin |
13 |
} |
195 |
|
|
if ($field_name) { |
196 |
|
|
$html .= $xml_codepage->convert($field_name); |
197 |
|
|
} |
198 |
|
|
$html .= $display_data."###\n"; |
199 |
|
|
} |
200 |
dpavlin |
10 |
} |
201 |
|
|
if ($swish_data) { |
202 |
|
|
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
203 |
dpavlin |
20 |
# remove extra spaces |
204 |
|
|
$swish_data =~ s/ +/ /g; |
205 |
|
|
$swish_data =~ s/ +$//g; |
206 |
|
|
|
207 |
dpavlin |
10 |
$swish_data = $i->convert($swish_data); |
208 |
|
|
$xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); |
209 |
|
|
#$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; |
210 |
|
|
#$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); |
211 |
|
|
} |
212 |
|
|
|
213 |
dpavlin |
9 |
|
214 |
dpavlin |
3 |
} |
215 |
dpavlin |
13 |
|
216 |
|
|
# dump formatted output in <html> |
217 |
|
|
if ($html) { |
218 |
|
|
$xml .= xmlify("html",$html); |
219 |
|
|
} |
220 |
|
|
|
221 |
dpavlin |
3 |
if ($xml) { |
222 |
dpavlin |
10 |
#print STDERR "x: $xml\n"; |
223 |
dpavlin |
13 |
$xml .= $add_xml if ($add_xml); |
224 |
dpavlin |
10 |
return "<xml>\n$xml</xml>\n"; |
225 |
dpavlin |
3 |
} else { |
226 |
|
|
return; |
227 |
|
|
} |
228 |
|
|
} |
229 |
|
|
|
230 |
|
|
########################################################################## |
231 |
|
|
|
232 |
dpavlin |
13 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
233 |
dpavlin |
1 |
|
234 |
dpavlin |
13 |
foreach my $database ($cfg->Sections) { |
235 |
dpavlin |
1 |
|
236 |
dpavlin |
13 |
my $isis_db = $cfg -> val($database, 'isis_db'); |
237 |
|
|
my $type = $cfg -> val($database, 'type'); |
238 |
|
|
my $add_xml = $cfg -> val($database, 'xml'); |
239 |
dpavlin |
1 |
|
240 |
dpavlin |
13 |
# read configuration for this type |
241 |
|
|
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
242 |
dpavlin |
29 |
$isis_codepage = Text::Iconv->new($config->{isis_codepage},$XML_CHARSET); |
243 |
dpavlin |
13 |
$index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
244 |
dpavlin |
29 |
$xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),$XML_CHARSET); |
245 |
dpavlin |
1 |
|
246 |
dpavlin |
3 |
my $db = OpenIsis::open( $isis_db ); |
247 |
|
|
if (0) { |
248 |
|
|
# # FIX |
249 |
|
|
# if (! $db ) { |
250 |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
251 |
|
|
next ; |
252 |
|
|
} |
253 |
|
|
|
254 |
dpavlin |
1 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
255 |
|
|
|
256 |
dpavlin |
3 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
257 |
dpavlin |
1 |
|
258 |
dpavlin |
13 |
my $path = $database; # was $isis_db |
259 |
dpavlin |
10 |
|
260 |
dpavlin |
3 |
my $last_p = 0; |
261 |
|
|
|
262 |
dpavlin |
10 |
# { my $row_id = 4514; |
263 |
dpavlin |
3 |
# FIX |
264 |
dpavlin |
1 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
265 |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
266 |
dpavlin |
3 |
if ($row && $row->{mfn}) { |
267 |
dpavlin |
10 |
#print STDERR "mfn: ",$row->{mfn},"\n"; |
268 |
dpavlin |
3 |
# output current process indicator |
269 |
|
|
my $p = int($row->{mfn} * 100 / $max_rowid); |
270 |
|
|
if ($p != $last_p) { |
271 |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
272 |
|
|
$last_p = $p; |
273 |
|
|
} |
274 |
|
|
|
275 |
dpavlin |
13 |
if (my $xml = isis2xml($row,$add_xml)) { |
276 |
dpavlin |
10 |
#print STDERR "--ret-->$xml\n"; |
277 |
dpavlin |
35 |
use bytes; # as opposed to chars |
278 |
dpavlin |
10 |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
279 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
280 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
281 |
dpavlin |
3 |
} |
282 |
dpavlin |
1 |
} |
283 |
|
|
} |
284 |
dpavlin |
3 |
print STDERR "\n"; |
285 |
dpavlin |
1 |
} |
286 |
dpavlin |
3 |
|
287 |
dpavlin |
10 |
# call this to commit index |
288 |
|
|
$index->close; |
289 |
dpavlin |
3 |
|
290 |
|
|
1; |
291 |
|
|
__END__ |
292 |
|
|
########################################################################## |
293 |
|
|
|
294 |
|
|
=head1 NAME |
295 |
|
|
|
296 |
|
|
isis2xml.pl - read isis file and dump XML |
297 |
|
|
|
298 |
|
|
=head1 DESCRIPTION |
299 |
|
|
|
300 |
|
|
This command will read ISIS data file using OpenIsis perl module and |
301 |
|
|
create XML file for usage with I<SWISH-E> |
302 |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
303 |
|
|
from isis files (isis allready has something like that). Output of this |
304 |
|
|
script is tailor-made for SWISH-E. |
305 |
|
|
|
306 |
|
|
=head1 AUTHOR |
307 |
|
|
|
308 |
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
309 |
|
|
|
310 |
|
|
=head1 COPYRIGHT |
311 |
|
|
|
312 |
|
|
GNU Public License (GPL) v2 or later |
313 |
|
|
|
314 |
|
|
=head1 SEE ALSO |
315 |
|
|
|
316 |
|
|
SWISH-E web site at http://www.swish-e.org |
317 |
|
|
|
318 |
|
|
=cut |