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 1.02; # 1.01 won't compile on my platform, |
9 |
use Text::Iconv; |
10 |
use Config::IniFiles; |
11 |
use Encode; |
12 |
|
13 |
$|=1; |
14 |
|
15 |
my $config_file = $0; |
16 |
$config_file =~ s/\.pl$/.conf/; |
17 |
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
18 |
|
19 |
my $config; |
20 |
|
21 |
use index_DBI; # there is no other, right now ;-) |
22 |
my $index; |
23 |
|
24 |
my %opts; |
25 |
|
26 |
# usage: |
27 |
# -d directory name |
28 |
# -m multiple directories |
29 |
# -q quiet |
30 |
# -s run swish |
31 |
|
32 |
getopts('d:m:qs', \%opts); |
33 |
|
34 |
my $path; # this is name of database |
35 |
|
36 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
37 |
|
38 |
# this is encoding of all files on disk, including import_xml/*.xml file and |
39 |
# filter/*.pm files! It will be used to store strings in perl internally! |
40 |
my $codepage = 'ISO-8859-2'; |
41 |
|
42 |
my $utf2cp = Text::Iconv->new('UTF-8',$codepage); |
43 |
# this function will convert data from XML files to local encoding |
44 |
sub x { |
45 |
return $utf2cp->convert($_[0]); |
46 |
} |
47 |
|
48 |
# decode isis import codepage |
49 |
my $isis2cp; |
50 |
|
51 |
# outgoing xml must be in UTF-8 |
52 |
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
53 |
|
54 |
sub isis2xml { |
55 |
|
56 |
use xmlify; |
57 |
|
58 |
my $row = shift @_; |
59 |
my $add_xml = shift @_; |
60 |
|
61 |
my $xml; |
62 |
|
63 |
use parse_format; |
64 |
|
65 |
my $html = ""; # html formatted display output |
66 |
|
67 |
my %field_usage; # counter for usage of each field |
68 |
|
69 |
# sort subrouting using order="" attribute |
70 |
sub by_order { |
71 |
return 0 if (! $config->{indexer}->{$a}->{order}); |
72 |
return 0 if (! $config->{indexer}->{$b}->{order}); |
73 |
|
74 |
return $config->{indexer}->{$a}->{order} <=> |
75 |
$config->{indexer}->{$b}->{order} ; |
76 |
} |
77 |
|
78 |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
79 |
|
80 |
$field=x($field); |
81 |
|
82 |
$field_usage{$field}++; |
83 |
|
84 |
my $swish_data = ""; |
85 |
my $display_data = ""; |
86 |
my $line_delimiter; |
87 |
|
88 |
my ($swish,$display); |
89 |
|
90 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
91 |
|
92 |
my $format = x($x->{content}); |
93 |
my $delimiter = x($x->{delimiter}) || ' '; |
94 |
|
95 |
my $isis_i = 0; # isis repeatable offset |
96 |
|
97 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
98 |
$s = 0 if (lc($x->{type}) eq "display"); |
99 |
$d = 0 if (lc($x->{type}) eq "swish"); |
100 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
101 |
|
102 |
# what will separate last line from this one? |
103 |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
104 |
$line_delimiter = ' '; |
105 |
} elsif ($display_data) { |
106 |
$line_delimiter = '<br/>'; |
107 |
} |
108 |
|
109 |
# init vars so that we go into while... |
110 |
($swish,$display) = (1,1); |
111 |
|
112 |
while ($swish || $display) { |
113 |
($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp); |
114 |
|
115 |
# filter="name" ; filter this field through |
116 |
# filter/[name].pm |
117 |
my $filter = $x->{filter}; |
118 |
if ($filter) { |
119 |
require "filter/".$filter.".pm"; |
120 |
} |
121 |
# type="swish" ; field for swish |
122 |
if ($s && $swish) { |
123 |
if ($filter) { |
124 |
no strict 'refs'; |
125 |
$swish_data .= join(" ",&$filter($swish)); |
126 |
} else { |
127 |
$swish_data .= $swish; |
128 |
} |
129 |
} |
130 |
|
131 |
# type="display" ; field for display |
132 |
if ($d && $display) { |
133 |
if ($line_delimiter && $display_data) { |
134 |
$display_data .= $line_delimiter; |
135 |
undef $line_delimiter; |
136 |
} |
137 |
if ($filter) { |
138 |
no strict 'refs'; |
139 |
$display_data .= join($delimiter,&$filter($display)); |
140 |
} else { |
141 |
if ($display_data) { |
142 |
$display_data .= $delimiter.$display; |
143 |
} else { |
144 |
$display_data .= $display; |
145 |
} |
146 |
} |
147 |
} |
148 |
|
149 |
# type="index" ; insert into index |
150 |
if ($i && $display) { |
151 |
my $index_data = $display; |
152 |
if ($filter) { |
153 |
no strict 'refs'; |
154 |
foreach my $d (&$filter($index_data)) { |
155 |
$index->insert($field, $d, $path); |
156 |
} |
157 |
} else { |
158 |
$index->insert($field, $index_data, $path); |
159 |
} |
160 |
} |
161 |
} |
162 |
} |
163 |
|
164 |
|
165 |
if ($display_data) { |
166 |
|
167 |
if ($field eq "headline") { |
168 |
$xml .= xmlify("headline", $display_data); |
169 |
} else { |
170 |
|
171 |
# find field name (signular, plural) |
172 |
my $field_name = ""; |
173 |
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
174 |
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
175 |
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
176 |
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
177 |
} elsif ($config->{indexer}->{$field}->{name}) { |
178 |
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
179 |
} else { |
180 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
181 |
} |
182 |
if ($field_name) { |
183 |
$html .= x($field_name); |
184 |
} |
185 |
$html .= $display_data."###\n"; |
186 |
} |
187 |
} |
188 |
if ($swish_data) { |
189 |
# remove extra spaces |
190 |
$swish_data =~ s/ +/ /g; |
191 |
$swish_data =~ s/ +$//g; |
192 |
|
193 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
194 |
} |
195 |
|
196 |
|
197 |
} |
198 |
|
199 |
# dump formatted output in <html> |
200 |
if ($html) { |
201 |
$xml .= xmlify("html",$html); |
202 |
} |
203 |
|
204 |
if ($xml) { |
205 |
$xml .= $add_xml if ($add_xml); |
206 |
return "<xml>\n$xml</xml>\n"; |
207 |
} else { |
208 |
return; |
209 |
} |
210 |
} |
211 |
|
212 |
########################################################################## |
213 |
|
214 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
215 |
|
216 |
# open index |
217 |
$index = new index_DBI( |
218 |
$cfg->val('global', 'dbi_dbd'), |
219 |
$cfg->val('global', 'dbi_dsn'), |
220 |
$cfg->val('global', 'dbi_user'), |
221 |
$cfg->val('global', 'dbi_passwd') || '', |
222 |
); |
223 |
|
224 |
# delete [global] section to leave just databases sections |
225 |
$cfg->DeleteSection('global'); |
226 |
|
227 |
foreach my $database ($cfg->Sections) { |
228 |
|
229 |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
230 |
my $type = $cfg -> val($database, 'type') || die "$database doesn't have 'type' defined"; |
231 |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
232 |
|
233 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
234 |
|
235 |
$isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
236 |
|
237 |
my $db = OpenIsis::open( $isis_db ); |
238 |
if (0) { |
239 |
# # FIX |
240 |
# if (! $db ) { |
241 |
print STDERR "WARNING: can't open '$isis_db'\n"; |
242 |
next ; |
243 |
} |
244 |
|
245 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
246 |
|
247 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
248 |
|
249 |
my $path = $database; # was $isis_db |
250 |
|
251 |
my $last_p = 0; |
252 |
|
253 |
# { my $row_id = 4514; |
254 |
# FIX |
255 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
256 |
my $row = OpenIsis::read( $db, $row_id ); |
257 |
if ($row && $row->{mfn}) { |
258 |
# output current process indicator |
259 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
260 |
if ($p != $last_p) { |
261 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
262 |
$last_p = $p; |
263 |
} |
264 |
|
265 |
my $swishpath = $path."#".int($row->{mfn}); |
266 |
|
267 |
if (my $xml = isis2xml($row,$add_xml)) { |
268 |
$xml = $cp2utf->convert($xml); |
269 |
use bytes; # as opposed to chars |
270 |
print "Path-Name: $swishpath\n"; |
271 |
print "Content-Length: ".(length($xml)+1)."\n"; |
272 |
print "Document-Type: XML\n\n$xml\n"; |
273 |
} |
274 |
} |
275 |
} |
276 |
print STDERR "\n"; |
277 |
} |
278 |
|
279 |
# call this to commit index |
280 |
$index->close; |
281 |
|
282 |
1; |
283 |
__END__ |
284 |
########################################################################## |
285 |
|
286 |
=head1 NAME |
287 |
|
288 |
isis2xml.pl - read isis file and dump XML |
289 |
|
290 |
=head1 DESCRIPTION |
291 |
|
292 |
This command will read ISIS data file using OpenIsis perl module and |
293 |
create XML file for usage with I<SWISH-E> |
294 |
indexer. Dispite it's name, this script B<isn't general xml generator> |
295 |
from isis files (isis allready has something like that). Output of this |
296 |
script is tailor-made for SWISH-E. |
297 |
|
298 |
=head1 AUTHOR |
299 |
|
300 |
Dobrica Pavlinusic <dpavlin@rot13.org> |
301 |
|
302 |
=head1 COPYRIGHT |
303 |
|
304 |
GNU Public License (GPL) v2 or later |
305 |
|
306 |
=head1 SEE ALSO |
307 |
|
308 |
SWISH-E web site at http://www.swish-e.org |
309 |
|
310 |
=cut |