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 |
40 |
use Encode; |
12 |
dpavlin |
1 |
|
13 |
dpavlin |
10 |
$|=1; |
14 |
dpavlin |
9 |
|
15 |
dpavlin |
13 |
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 |
dpavlin |
10 |
my $config; |
20 |
|
|
|
21 |
|
|
use index_DBI; # there is no other, right now ;-) |
22 |
dpavlin |
50 |
my $index; |
23 |
dpavlin |
10 |
|
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 |
43 |
my $path; # this is name of database |
35 |
dpavlin |
1 |
|
36 |
dpavlin |
40 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
37 |
dpavlin |
1 |
|
38 |
dpavlin |
40 |
# 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 |
dpavlin |
1 |
|
42 |
dpavlin |
40 |
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 |
dpavlin |
3 |
|
48 |
dpavlin |
54 |
# decode isis/excel or other import codepage |
49 |
|
|
my $import2cp; |
50 |
dpavlin |
10 |
|
51 |
dpavlin |
40 |
# outgoing xml must be in UTF-8 |
52 |
|
|
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
53 |
dpavlin |
29 |
|
54 |
dpavlin |
54 |
# mapping between data type and tag which specify |
55 |
|
|
# format in XML file |
56 |
|
|
my %type2tag = ( |
57 |
|
|
'isis' => 'isis', |
58 |
|
|
'excel' => 'column' |
59 |
|
|
); |
60 |
dpavlin |
3 |
|
61 |
dpavlin |
54 |
sub data2xml { |
62 |
|
|
|
63 |
dpavlin |
10 |
use xmlify; |
64 |
|
|
|
65 |
dpavlin |
54 |
my $type = shift @_; |
66 |
dpavlin |
3 |
my $row = shift @_; |
67 |
dpavlin |
13 |
my $add_xml = shift @_; |
68 |
dpavlin |
3 |
|
69 |
|
|
my $xml; |
70 |
|
|
|
71 |
dpavlin |
10 |
use parse_format; |
72 |
dpavlin |
3 |
|
73 |
dpavlin |
13 |
my $html = ""; # html formatted display output |
74 |
dpavlin |
10 |
|
75 |
dpavlin |
13 |
my %field_usage; # counter for usage of each field |
76 |
|
|
|
77 |
dpavlin |
32 |
# sort subrouting using order="" attribute |
78 |
|
|
sub by_order { |
79 |
|
|
return 0 if (! $config->{indexer}->{$a}->{order}); |
80 |
|
|
return 0 if (! $config->{indexer}->{$b}->{order}); |
81 |
dpavlin |
29 |
|
82 |
dpavlin |
32 |
return $config->{indexer}->{$a}->{order} <=> |
83 |
|
|
$config->{indexer}->{$b}->{order} ; |
84 |
|
|
} |
85 |
dpavlin |
3 |
|
86 |
dpavlin |
32 |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
87 |
|
|
|
88 |
dpavlin |
40 |
$field=x($field); |
89 |
|
|
|
90 |
dpavlin |
13 |
$field_usage{$field}++; |
91 |
|
|
|
92 |
dpavlin |
10 |
my $swish_data = ""; |
93 |
dpavlin |
3 |
my $display_data = ""; |
94 |
dpavlin |
35 |
my $line_delimiter; |
95 |
dpavlin |
3 |
|
96 |
dpavlin |
34 |
my ($swish,$display); |
97 |
|
|
|
98 |
dpavlin |
54 |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
99 |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
100 |
dpavlin |
3 |
|
101 |
dpavlin |
40 |
my $format = x($x->{content}); |
102 |
|
|
my $delimiter = x($x->{delimiter}) || ' '; |
103 |
dpavlin |
3 |
|
104 |
dpavlin |
54 |
my $repeat_off = 0; # repeatable offset |
105 |
dpavlin |
29 |
|
106 |
|
|
my ($s,$d,$i) = (1,1,0); # swish, display default |
107 |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
108 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
109 |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
110 |
|
|
|
111 |
dpavlin |
35 |
# what will separate last line from this one? |
112 |
|
|
if ($display_data && $x->{append} && $x->{append} eq "1") { |
113 |
dpavlin |
34 |
$line_delimiter = ' '; |
114 |
dpavlin |
35 |
} elsif ($display_data) { |
115 |
dpavlin |
34 |
$line_delimiter = '<br/>'; |
116 |
|
|
} |
117 |
dpavlin |
29 |
|
118 |
dpavlin |
34 |
# init vars so that we go into while... |
119 |
|
|
($swish,$display) = (1,1); |
120 |
dpavlin |
29 |
|
121 |
dpavlin |
54 |
if ($swish || $display) { |
122 |
|
|
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
123 |
dpavlin |
29 |
# filter="name" ; filter this field through |
124 |
|
|
# filter/[name].pm |
125 |
|
|
my $filter = $x->{filter}; |
126 |
dpavlin |
20 |
if ($filter) { |
127 |
dpavlin |
29 |
require "filter/".$filter.".pm"; |
128 |
dpavlin |
20 |
} |
129 |
dpavlin |
29 |
# type="swish" ; field for swish |
130 |
|
|
if ($s && $swish) { |
131 |
|
|
if ($filter) { |
132 |
|
|
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 |
dpavlin |
40 |
my $index_data = $display; |
160 |
dpavlin |
29 |
if ($filter) { |
161 |
|
|
no strict 'refs'; |
162 |
|
|
foreach my $d (&$filter($index_data)) { |
163 |
dpavlin |
43 |
$index->insert($field, $d, $path); |
164 |
dpavlin |
29 |
} |
165 |
|
|
} else { |
166 |
dpavlin |
43 |
$index->insert($field, $index_data, $path); |
167 |
dpavlin |
20 |
} |
168 |
|
|
} |
169 |
dpavlin |
17 |
} |
170 |
dpavlin |
3 |
} |
171 |
dpavlin |
9 |
|
172 |
dpavlin |
13 |
|
173 |
dpavlin |
10 |
if ($display_data) { |
174 |
dpavlin |
29 |
|
175 |
dpavlin |
13 |
if ($field eq "headline") { |
176 |
|
|
$xml .= xmlify("headline", $display_data); |
177 |
|
|
} else { |
178 |
|
|
|
179 |
|
|
# find field name (signular, plural) |
180 |
|
|
my $field_name = ""; |
181 |
|
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
182 |
|
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
183 |
|
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
184 |
|
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
185 |
dpavlin |
20 |
} elsif ($config->{indexer}->{$field}->{name}) { |
186 |
|
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
187 |
dpavlin |
13 |
} else { |
188 |
dpavlin |
20 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
189 |
dpavlin |
13 |
} |
190 |
|
|
if ($field_name) { |
191 |
dpavlin |
40 |
$html .= x($field_name); |
192 |
dpavlin |
13 |
} |
193 |
|
|
$html .= $display_data."###\n"; |
194 |
|
|
} |
195 |
dpavlin |
10 |
} |
196 |
|
|
if ($swish_data) { |
197 |
dpavlin |
20 |
# remove extra spaces |
198 |
|
|
$swish_data =~ s/ +/ /g; |
199 |
|
|
$swish_data =~ s/ +$//g; |
200 |
|
|
|
201 |
dpavlin |
40 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
202 |
dpavlin |
10 |
} |
203 |
|
|
|
204 |
dpavlin |
9 |
|
205 |
dpavlin |
3 |
} |
206 |
dpavlin |
13 |
|
207 |
|
|
# dump formatted output in <html> |
208 |
|
|
if ($html) { |
209 |
|
|
$xml .= xmlify("html",$html); |
210 |
|
|
} |
211 |
|
|
|
212 |
dpavlin |
3 |
if ($xml) { |
213 |
dpavlin |
13 |
$xml .= $add_xml if ($add_xml); |
214 |
dpavlin |
10 |
return "<xml>\n$xml</xml>\n"; |
215 |
dpavlin |
3 |
} else { |
216 |
|
|
return; |
217 |
|
|
} |
218 |
|
|
} |
219 |
|
|
|
220 |
|
|
########################################################################## |
221 |
|
|
|
222 |
dpavlin |
54 |
# read configuration for this script |
223 |
dpavlin |
13 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
224 |
dpavlin |
1 |
|
225 |
dpavlin |
54 |
# read global.conf configuration |
226 |
|
|
my $cfg_global = new Config::IniFiles( -file => 'global.conf' ); |
227 |
|
|
|
228 |
dpavlin |
50 |
# open index |
229 |
|
|
$index = new index_DBI( |
230 |
dpavlin |
54 |
$cfg_global->val('global', 'dbi_dbd'), |
231 |
|
|
$cfg_global->val('global', 'dbi_dsn'), |
232 |
|
|
$cfg_global->val('global', 'dbi_user'), |
233 |
|
|
$cfg_global->val('global', 'dbi_passwd') || '', |
234 |
dpavlin |
50 |
); |
235 |
|
|
|
236 |
dpavlin |
13 |
foreach my $database ($cfg->Sections) { |
237 |
dpavlin |
1 |
|
238 |
dpavlin |
54 |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
239 |
dpavlin |
40 |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
240 |
dpavlin |
1 |
|
241 |
dpavlin |
54 |
print STDERR "reading ./import_xml/$type.xml\n"; |
242 |
dpavlin |
1 |
|
243 |
dpavlin |
54 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type} ], forcecontent => 1); |
244 |
dpavlin |
40 |
|
245 |
dpavlin |
54 |
# output current progress indicator |
246 |
|
|
my $last_p = 0; |
247 |
|
|
sub progress { |
248 |
|
|
#return if (! $opts{q}); # FIXME |
249 |
|
|
my $current = shift; |
250 |
|
|
my $total = shift; |
251 |
|
|
my $p = int($current * 100 / $total); |
252 |
|
|
if ($p != $last_p) { |
253 |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
254 |
|
|
$last_p = $p; |
255 |
|
|
} |
256 |
dpavlin |
3 |
} |
257 |
|
|
|
258 |
dpavlin |
54 |
# now read database |
259 |
|
|
print STDERR "using: $type...\n"; |
260 |
dpavlin |
1 |
|
261 |
dpavlin |
54 |
if ($type eq "isis") { |
262 |
|
|
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
263 |
dpavlin |
1 |
|
264 |
dpavlin |
54 |
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
265 |
|
|
my $db = OpenIsis::open( $isis_db ); |
266 |
dpavlin |
10 |
|
267 |
dpavlin |
54 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
268 |
dpavlin |
3 |
|
269 |
dpavlin |
54 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
270 |
|
|
|
271 |
|
|
my $path = $database; |
272 |
|
|
|
273 |
|
|
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
274 |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
275 |
|
|
if ($row && $row->{mfn}) { |
276 |
|
|
|
277 |
|
|
progress($row->{mfn}, $max_rowid); |
278 |
|
|
|
279 |
|
|
my $swishpath = $path."#".int($row->{mfn}); |
280 |
|
|
|
281 |
|
|
if (my $xml = data2xml($type,$row,$add_xml)) { |
282 |
|
|
$xml = $cp2utf->convert($xml); |
283 |
|
|
use bytes; # as opposed to chars |
284 |
|
|
print "Path-Name: $swishpath\n"; |
285 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
286 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
287 |
|
|
} |
288 |
dpavlin |
3 |
} |
289 |
dpavlin |
54 |
} |
290 |
|
|
print STDERR "\n"; |
291 |
dpavlin |
3 |
|
292 |
dpavlin |
54 |
} elsif ($type eq "excel") { |
293 |
|
|
use Spreadsheet::ParseExcel; |
294 |
|
|
use Spreadsheet::ParseExcel::Utility qw(int2col); |
295 |
|
|
|
296 |
|
|
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
297 |
|
|
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
298 |
dpavlin |
43 |
|
299 |
dpavlin |
54 |
my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; |
300 |
|
|
my $start_row = x($config->{start_row}) || die "no start_row in $type.xml"; |
301 |
|
|
|
302 |
|
|
my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; |
303 |
|
|
|
304 |
|
|
my $sheet_nr = 0; |
305 |
|
|
foreach my $oWks (@{$oBook->{Worksheet}}) { |
306 |
|
|
#print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n"; |
307 |
|
|
last if ($oWks->{Name} eq $sheet); |
308 |
|
|
$sheet_nr++; |
309 |
|
|
} |
310 |
|
|
|
311 |
|
|
my $oWorksheet = $oBook->{Worksheet}[$sheet_nr]; |
312 |
|
|
|
313 |
|
|
print STDERR "using sheet: ",$oWorksheet->{Name},"\n"; |
314 |
|
|
defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; |
315 |
|
|
my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; |
316 |
|
|
|
317 |
|
|
for(my $iR = $oWorksheet->{MinRow} ; defined $end_row && $iR <= $end_row ; $iR++) { |
318 |
|
|
my $row; |
319 |
|
|
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
320 |
|
|
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
321 |
|
|
if ($cell) { |
322 |
|
|
$row->{int2col($iC)} = $cell->Value; |
323 |
|
|
} |
324 |
|
|
} |
325 |
|
|
|
326 |
|
|
progress($iR, $end_row); |
327 |
|
|
|
328 |
|
|
# print "row[$iR/$end_row] "; |
329 |
|
|
# foreach (keys %{$row}) { |
330 |
|
|
# print "$_: ",$row->{$_},"\t"; |
331 |
|
|
# } |
332 |
|
|
# print "\n"; |
333 |
|
|
|
334 |
|
|
my $swishpath = $database."#".$iR; |
335 |
|
|
|
336 |
|
|
next if (! $row); |
337 |
|
|
|
338 |
|
|
if (my $xml = data2xml($type,$row,$add_xml)) { |
339 |
dpavlin |
44 |
$xml = $cp2utf->convert($xml); |
340 |
dpavlin |
35 |
use bytes; # as opposed to chars |
341 |
dpavlin |
43 |
print "Path-Name: $swishpath\n"; |
342 |
dpavlin |
10 |
print "Content-Length: ".(length($xml)+1)."\n"; |
343 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
344 |
dpavlin |
3 |
} |
345 |
dpavlin |
1 |
} |
346 |
|
|
} |
347 |
|
|
} |
348 |
dpavlin |
3 |
|
349 |
dpavlin |
10 |
# call this to commit index |
350 |
|
|
$index->close; |
351 |
dpavlin |
3 |
|
352 |
|
|
1; |
353 |
|
|
__END__ |
354 |
|
|
########################################################################## |
355 |
|
|
|
356 |
|
|
=head1 NAME |
357 |
|
|
|
358 |
|
|
isis2xml.pl - read isis file and dump XML |
359 |
|
|
|
360 |
|
|
=head1 DESCRIPTION |
361 |
|
|
|
362 |
|
|
This command will read ISIS data file using OpenIsis perl module and |
363 |
|
|
create XML file for usage with I<SWISH-E> |
364 |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
365 |
|
|
from isis files (isis allready has something like that). Output of this |
366 |
|
|
script is tailor-made for SWISH-E. |
367 |
|
|
|
368 |
|
|
=head1 AUTHOR |
369 |
|
|
|
370 |
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
371 |
|
|
|
372 |
|
|
=head1 COPYRIGHT |
373 |
|
|
|
374 |
|
|
GNU Public License (GPL) v2 or later |
375 |
|
|
|
376 |
|
|
=head1 SEE ALSO |
377 |
|
|
|
378 |
|
|
SWISH-E web site at http://www.swish-e.org |
379 |
|
|
|
380 |
|
|
=cut |