/[webpac]/branches/ffzg/all2xml.pl
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Contents of /branches/ffzg/all2xml.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 90 - (show annotations)
Sun Jul 13 13:22:50 2003 UTC (20 years, 8 months ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 13300 byte(s)
repeatable fields (broken when other input formats where introduced) work
again

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; # default DBI module for index
22 use index_DBI_cache; # faster DBI module using memory cache
23 my $index;
24
25 my %opts;
26
27 # usage:
28 # -d directory name
29 # -m multiple directories
30 # -q quiet
31 # -s run swish
32
33 getopts('d:m:qs', \%opts);
34
35 my $path; # this is name of database
36
37 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
38
39 # this is encoding of all files on disk, including import_xml/*.xml file and
40 # filter/*.pm files! It will be used to store strings in perl internally!
41 my $codepage = 'ISO-8859-2';
42
43 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
44 # this function will convert data from XML files to local encoding
45 sub x {
46 return $utf2cp->convert($_[0]);
47 }
48
49 # decode isis/excel or other import codepage
50 my $import2cp;
51
52 # outgoing xml must be in UTF-8
53 my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
54
55 # mapping between data type and tag which specify
56 # format in XML file
57 my %type2tag = (
58 'isis' => 'isis',
59 'excel' => 'column',
60 'marc' => 'marc',
61 'feed' => 'feed'
62 );
63
64 sub data2xml {
65
66 use xmlify;
67
68 my $type = shift @_;
69 my $row = shift @_;
70 my $add_xml = shift @_;
71 # needed to read values from configuration file
72 my $cfg = shift @_;
73 my $database = shift @_;
74
75 my $xml;
76
77 use parse_format;
78
79 my $html = ""; # html formatted display output
80
81 my %field_usage; # counter for usage of each field
82
83 # sort subrouting using order="" attribute
84 sub by_order {
85 return 0 if (! $config->{indexer}->{$a}->{order});
86 return 0 if (! $config->{indexer}->{$b}->{order});
87
88 return $config->{indexer}->{$a}->{order} <=>
89 $config->{indexer}->{$b}->{order} ;
90 }
91
92 foreach my $field (sort by_order keys %{$config->{indexer}}) {
93
94 $field=x($field);
95
96 $field_usage{$field}++;
97
98 my $swish_data = "";
99 my $display_data = "";
100 my $line_delimiter;
101
102 my ($swish,$display);
103
104 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
105 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
106
107 my $format = x($x->{content});
108 my $delimiter = x($x->{delimiter}) || ' ';
109
110 my $repeat_off = 0; # repeatable offset
111
112 my ($s,$d,$i) = (1,1,0); # swish, display default
113 $s = 0 if (lc($x->{type}) eq "display");
114 $d = 0 if (lc($x->{type}) eq "swish");
115 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
116
117 # what will separate last line from this one?
118 if ($display_data && $x->{append} && $x->{append} eq "1") {
119 $line_delimiter = ' ';
120 } elsif ($display_data) {
121 $line_delimiter = '<br/>';
122 }
123
124 # init vars so that we go into while...
125 ($swish,$display) = (1,1);
126
127 # while because of repeatable fields
128 while ($swish || $display) {
129 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
130 if ($repeat_off > 1000) {
131 print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
132 last;
133 }
134
135 # filter="name" ; filter this field through
136 # filter/[name].pm
137 my $filter = $x->{filter};
138 if ($filter) {
139 require "filter/".$filter.".pm";
140 }
141 # type="swish" ; field for swish
142 if ($s && $swish) {
143 if ($filter) {
144 no strict 'refs';
145 $swish_data .= join(" ",&$filter($swish));
146 } else {
147 $swish_data .= $swish;
148 }
149 }
150
151 # type="display" ; field for display
152 if ($d && $display) {
153 if ($line_delimiter && $display_data) {
154 $display_data .= $line_delimiter;
155 undef $line_delimiter;
156 }
157 if ($filter) {
158 no strict 'refs';
159 $display_data .= join($delimiter,&$filter($display));
160 } else {
161 if ($display_data) {
162 $display_data .= $delimiter.$display;
163 } else {
164 $display_data .= $display;
165 }
166 }
167 }
168
169 # type="index" ; insert into index
170 if ($i && $display) {
171 my $index_data = $display;
172 if ($filter) {
173 no strict 'refs';
174 foreach my $d (&$filter($index_data)) {
175 $index->insert($field, $d, $path);
176 }
177 } else {
178 $index->insert($field, $index_data, $path);
179 }
180 }
181 }
182 }
183
184 # now try to parse variables from configuration file
185 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
186
187 my $delimiter = x($x->{delimiter}) || ' ';
188 my $val = $cfg->val($database, x($x->{content}));
189
190 my ($s,$d,$i) = (1,1,0); # swish, display default
191 $s = 0 if (lc($x->{type}) eq "display");
192 $d = 0 if (lc($x->{type}) eq "swish");
193 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
194
195 if ($val) {
196 $display_data .= $delimiter.$val if ($d);
197 $swish_data .= $val if ($s);
198 $index->insert($field, $val, $path) if ($i);
199 }
200
201 }
202
203
204 if ($display_data) {
205
206 if ($field eq "headline") {
207 $xml .= xmlify("headline", $display_data);
208 } else {
209
210 # find field name (signular, plural)
211 my $field_name = "";
212 if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
213 $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
214 } elsif ($config->{indexer}->{$field}->{name_plural}) {
215 $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
216 } elsif ($config->{indexer}->{$field}->{name}) {
217 $field_name = $config->{indexer}->{$field}->{name}."#-#";
218 } else {
219 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
220 }
221 if ($field_name) {
222 $html .= x($field_name);
223 }
224 $html .= $display_data."###\n";
225 }
226 }
227 if ($swish_data) {
228 # remove extra spaces
229 $swish_data =~ s/ +/ /g;
230 $swish_data =~ s/ +$//g;
231
232 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
233 }
234
235
236 }
237
238 # dump formatted output in <html>
239 if ($html) {
240 #$xml .= xmlify("html",$html);
241 $xml .= "<html><![CDATA[ $html ]]></html>";
242 }
243
244 if ($xml) {
245 $xml .= $add_xml if ($add_xml);
246 return "<xml>\n$xml</xml>\n";
247 } else {
248 return;
249 }
250 }
251
252 ##########################################################################
253
254 # read configuration for this script
255 my $cfg = new Config::IniFiles( -file => $config_file );
256
257 # read global.conf configuration
258 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
259
260 # open index
261 $index = new index_DBI(
262 $cfg_global->val('global', 'dbi_dbd'),
263 $cfg_global->val('global', 'dbi_dsn'),
264 $cfg_global->val('global', 'dbi_user'),
265 $cfg_global->val('global', 'dbi_passwd') || '',
266 );
267
268 foreach my $database ($cfg->Sections) {
269
270 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
271 my $add_xml = $cfg -> val($database, 'xml'); # optional
272
273 print STDERR "reading ./import_xml/$type.xml\n";
274
275 # extract just type basic
276 my $type_base = $type;
277 $type_base =~ s/_.+$//g;
278
279 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
280
281 # output current progress indicator
282 my $last_p = 0;
283 sub progress {
284 #return if (! $opts{q}); # FIXME
285 my $current = shift;
286 my $total = shift || 1;
287 my $p = int($current * 100 / $total);
288 if ($p != $last_p) {
289 printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
290 $last_p = $p;
291 }
292 }
293
294 my $fake_dir = 1;
295 sub fakeprogress {
296 my $current = shift @_;
297
298 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
299
300 $last_p += $fake_dir;
301 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
302 if ($last_p % 10 == 0) {
303 printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
304 }
305 }
306
307 # now read database
308 print STDERR "using: $type...\n";
309
310 if ($type_base eq "isis") {
311
312 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
313
314 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
315 my $db = OpenIsis::open( $isis_db );
316
317 my $max_rowid = OpenIsis::maxRowid( $db );
318
319 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
320
321 my $path = $database;
322
323 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
324 my $row = OpenIsis::read( $db, $row_id );
325 if ($row && $row->{mfn}) {
326
327 progress($row->{mfn}, $max_rowid);
328
329 my $swishpath = $path."#".int($row->{mfn});
330
331 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
332 $xml = $cp2utf->convert($xml);
333 use bytes; # as opposed to chars
334 print "Path-Name: $swishpath\n";
335 print "Content-Length: ".(length($xml)+1)."\n";
336 print "Document-Type: XML\n\n$xml\n";
337 }
338 }
339 }
340 print STDERR "\n";
341
342 } elsif ($type_base eq "excel") {
343 use Spreadsheet::ParseExcel;
344 use Spreadsheet::ParseExcel::Utility qw(int2col);
345
346 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
347 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
348
349 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
350 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
351
352 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
353
354 my $sheet_nr = 0;
355 foreach my $oWks (@{$oBook->{Worksheet}}) {
356 #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
357 last if ($oWks->{Name} eq $sheet);
358 $sheet_nr++;
359 }
360
361 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
362
363 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
364 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
365 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
366
367 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
368 my $row;
369 for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
370 my $cell = $oWorksheet->{Cells}[$iR][$iC];
371 if ($cell) {
372 $row->{int2col($iC)} = $cell->Value;
373 }
374 }
375
376 progress($iR, $end_row);
377
378 # print "row[$iR/$end_row] ";
379 # foreach (keys %{$row}) {
380 # print "$_: ",$row->{$_},"\t";
381 # }
382 # print "\n";
383
384 my $swishpath = $database."#".$iR;
385
386 next if (! $row);
387
388 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
389 $xml = $cp2utf->convert($xml);
390 use bytes; # as opposed to chars
391 print "Path-Name: $swishpath\n";
392 print "Content-Length: ".(length($xml)+1)."\n";
393 print "Document-Type: XML\n\n$xml\n";
394 }
395 }
396 } elsif ($type_base eq "marc") {
397
398 use MARC;
399
400 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
401 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
402
403 # optional argument is format
404 my $format = x($config->{format}) || 'usmarc';
405
406 print STDERR "Reading MARC file '$marc_file'\n";
407
408 my $marc = new MARC;
409 my $nr = $marc->openmarc({
410 file=>$marc_file, format=>$format
411 }) || die "Can't open MARC file '$marc_file'";
412
413 my $i=0; # record nr.
414
415 my $rec;
416
417 while ($marc->nextmarc(1)) {
418
419 # XXX
420 fakeprogress($i++);
421
422 my $swishpath = $database."#".$i;
423
424 if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
425 $xml = $cp2utf->convert($xml);
426 use bytes; # as opposed to chars
427 print "Path-Name: $swishpath\n";
428 print "Content-Length: ".(length($xml)+1)."\n";
429 print "Document-Type: XML\n\n$xml\n";
430 }
431 }
432 } elsif ($type_base eq "feed") {
433
434 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
435 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
436
437 print STDERR "Reading feed from program '$prog'\n";
438
439 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
440
441 my $i=1; # record nr.
442
443 my $data;
444 my $line=1;
445
446 while (<FEED>) {
447 chomp;
448
449 if (/^$/) {
450 my $swishpath = $database."#".$i++;
451
452 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
453 $xml = $cp2utf->convert($xml);
454 use bytes; # as opposed to chars
455 print "Path-Name: $swishpath\n";
456 print "Content-Length: ".(length($xml)+1)."\n";
457 print "Document-Type: XML\n\n$xml\n";
458 }
459 $line = 1;
460 $data = {};
461 next;
462 }
463
464 $line = $1 if (s/^(\d+):\s*//);
465 $data->{$line++} = $_;
466
467 fakeprogress($i);
468
469 }
470 }
471 }
472
473 # call this to commit index
474 $index->close;
475
476 1;
477 __END__
478 ##########################################################################
479
480 =head1 NAME
481
482 all2xml.pl - read various file formats and dump XML for SWISH-E
483
484 =head1 DESCRIPTION
485
486 This command will read ISIS data file using OpenIsis perl module, MARC
487 records using MARC module and optionally Micro$oft Excel files to
488 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
489 this script B<isn't general xml generator> from isis files (isis allready
490 has something like that). Output of this script is tailor-made for SWISH-E.
491
492 =head1 BUGS
493
494 Documentation is really lacking. However, in true Open Source spirit, source
495 is best documentation. I even made considerable effort to comment parts
496 which are not intuitively clear, so...
497
498 =head1 AUTHOR
499
500 Dobrica Pavlinusic <dpavlin@rot13.org>
501
502 =head1 COPYRIGHT
503
504 GNU Public License (GPL) v2 or later
505
506 =head1 SEE ALSO
507
508 SWISH-E web site at http://www.swish-e.org
509
510 =cut

Properties

Name Value
cvs2svn:cvs-rev 1.30
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26