/[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 135 - (show annotations)
Wed Oct 29 21:27:00 2003 UTC (20 years, 4 months ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 15165 byte(s)
fix repeatable fields in index data

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 my $va = $config->{indexer}->{$a}->{order} ||
86 $config->{indexer}->{$a};
87 my $vb = $config->{indexer}->{$b}->{order} ||
88 $config->{indexer}->{$b};
89
90 return $va <=> $vb;
91 }
92
93 foreach my $field (sort by_order keys %{$config->{indexer}}) {
94
95 $field=x($field);
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 # placeholder for all repeatable entries for index
128 my @index_data;
129 my $index_filter;
130
131 # while because of repeatable fields
132 while ($swish || $display) {
133 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
134 if ($repeat_off > 1000) {
135 print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
136 last;
137 }
138
139 # filter="name" ; filter this field through
140 # filter/[name].pm
141 my $filter = $x->{filter};
142 if ($filter) {
143 require "filter/".$filter.".pm";
144 }
145 # type="swish" ; field for swish
146 if ($s && $swish) {
147 if ($filter) {
148 no strict 'refs';
149 $swish_data .= join(" ",&$filter($swish));
150 } else {
151 $swish_data .= $swish;
152 }
153 }
154
155 # type="display" ; field for display
156 if ($d && $display) {
157 if ($line_delimiter && $display_data) {
158 $display_data .= $line_delimiter;
159 undef $line_delimiter;
160 }
161 if ($filter) {
162 no strict 'refs';
163 $display_data .= join($delimiter,&$filter($display));
164 } else {
165 if ($display_data) {
166 $display_data .= $delimiter.$display;
167 } else {
168 $display_data .= $display;
169 }
170 }
171 }
172
173 # type="index" ; insert into index
174 if ($i && $display) {
175 push @index_data, $display;
176 $index_filter = $filter if ($filter);
177 }
178 }
179
180 # fill data in index
181 if (@index_data) {
182 if ($index_filter) {
183 no strict 'refs';
184 foreach my $d (@index_data) {
185 $index->insert($field, &$index_filter($d), $path);
186 }
187 } else {
188 foreach my $d (@index_data) {
189 $index->insert($field, $d, $path);
190 }
191 }
192 }
193 }
194
195 # now try to parse variables from configuration file
196 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
197
198 my $delimiter = x($x->{delimiter}) || ' ';
199 my $val = $cfg->val($database, x($x->{content}));
200
201 my ($s,$d,$i) = (1,1,0); # swish, display default
202 $s = 0 if (lc($x->{type}) eq "display");
203 $d = 0 if (lc($x->{type}) eq "swish");
204 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
205
206 if ($val) {
207 $display_data .= $delimiter.$val if ($d);
208 $swish_data .= $val if ($s);
209 $index->insert($field, $val, $path) if ($i);
210 }
211
212 }
213
214
215 if ($display_data) {
216
217 if ($field eq "headline") {
218 $xml .= xmlify("headline", $display_data);
219 } else {
220
221 # find field name (signular, plural)
222 my $field_name = "";
223 if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
224 $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
225 } elsif ($config->{indexer}->{$field}->{name_plural}) {
226 $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
227 } elsif ($config->{indexer}->{$field}->{name}) {
228 $field_name = $config->{indexer}->{$field}->{name}."#-#";
229 } else {
230 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
231 }
232 if ($field_name) {
233 $html .= x($field_name);
234 }
235 $html .= $display_data."###\n";
236 }
237 }
238 if ($swish_data) {
239 # remove extra spaces
240 $swish_data =~ s/ +/ /g;
241 $swish_data =~ s/ +$//g;
242
243 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
244 }
245
246
247 }
248
249 # dump formatted output in <html>
250 if ($html) {
251 #$xml .= xmlify("html",$html);
252 $xml .= "<html><![CDATA[ $html ]]></html>";
253 }
254
255 if ($xml) {
256 $xml .= $add_xml if ($add_xml);
257 return "<xml>\n$xml</xml>\n";
258 } else {
259 return;
260 }
261 }
262
263 ##########################################################################
264
265 # read configuration for this script
266 my $cfg = new Config::IniFiles( -file => $config_file );
267
268 # read global.conf configuration
269 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
270
271 # open index
272 $index = new index_DBI(
273 $cfg_global->val('global', 'dbi_dbd'),
274 $cfg_global->val('global', 'dbi_dsn'),
275 $cfg_global->val('global', 'dbi_user'),
276 $cfg_global->val('global', 'dbi_passwd') || '',
277 );
278
279 my $show_progress = $cfg_global->val('global', 'show_progress');
280
281 foreach my $database ($cfg->Sections) {
282
283 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
284 my $add_xml = $cfg -> val($database, 'xml'); # optional
285
286 print STDERR "reading ./import_xml/$type.xml\n";
287
288 # extract just type basic
289 my $type_base = $type;
290 $type_base =~ s/_.+$//g;
291
292 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
293
294 # output current progress indicator
295 my $last_p = 0;
296 sub progress {
297 return if (! $show_progress);
298 my $current = shift;
299 my $total = shift || 1;
300 my $p = int($current * 100 / $total);
301 if ($p != $last_p) {
302 printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
303 $last_p = $p;
304 }
305 }
306
307 my $fake_dir = 1;
308 sub fakeprogress {
309 return if (! $show_progress);
310 my $current = shift @_;
311
312 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
313
314 $last_p += $fake_dir;
315 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
316 if ($last_p % 10 == 0) {
317 printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
318 }
319 }
320
321 # now read database
322 print STDERR "using: $type...\n";
323
324 if ($type_base eq "isis") {
325
326 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
327
328 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
329 my $db = OpenIsis::open( $isis_db );
330
331 # check if .txt database for OpenIsis is zero length,
332 # if so, erase it and re-open database
333 sub check_txt_db {
334 my $isis_db = shift || die "need isis database name";
335 my $reopen = 0;
336
337 if (-e $isis_db.".TXT") {
338 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
339 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
340 $reopen++;
341 }
342 if (-e $isis_db.".PTR") {
343 print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
344 unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
345 $reopen++;
346 }
347 return OpenIsis::open( $isis_db ) if ($reopen);
348 }
349
350 # EOF error
351 if ($db == -1) {
352 $db = check_txt_db($isis_db);
353 if ($db == -1) {
354 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
355 next;
356 }
357 }
358
359 # OpenIsis::ERR_BADF
360 if ($db == -4) {
361 print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
362 next;
363 # OpenIsis::ERR_IO
364 } elsif ($db == -5) {
365 print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
366 next;
367 } elsif ($db < 0) {
368 print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
369 next;
370 }
371
372 my $max_rowid = OpenIsis::maxRowid( $db );
373
374 # if 0 records, try to rease isis .txt database
375 if ($max_rowid == 0) {
376 # force removal of database
377 $db = check_txt_db($isis_db);
378 $max_rowid = OpenIsis::maxRowid( $db );
379 }
380
381 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
382
383 my $path = $database;
384
385 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
386 my $row = OpenIsis::read( $db, $row_id );
387 if ($row && $row->{mfn}) {
388
389 progress($row->{mfn}, $max_rowid);
390
391 my $swishpath = $path."#".int($row->{mfn});
392
393 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
394 $xml = $cp2utf->convert($xml);
395 use bytes; # as opposed to chars
396 print "Path-Name: $swishpath\n";
397 print "Content-Length: ".(length($xml)+1)."\n";
398 print "Document-Type: XML\n\n$xml\n";
399 }
400 }
401 }
402 # for this to work with current version of OpenIsis (0.9.0)
403 # you might need my patch from
404 # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
405 OpenIsis::close($db);
406 print STDERR "\n";
407
408 } elsif ($type_base eq "excel") {
409 use Spreadsheet::ParseExcel;
410 use Spreadsheet::ParseExcel::Utility qw(int2col);
411
412 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
413 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
414
415 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
416 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
417
418 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
419
420 my $sheet_nr = 0;
421 foreach my $oWks (@{$oBook->{Worksheet}}) {
422 #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
423 last if ($oWks->{Name} eq $sheet);
424 $sheet_nr++;
425 }
426
427 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
428
429 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
430 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
431 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
432
433 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
434 my $row;
435 for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
436 my $cell = $oWorksheet->{Cells}[$iR][$iC];
437 if ($cell) {
438 $row->{int2col($iC)} = $cell->Value;
439 }
440 }
441
442 progress($iR, $end_row);
443
444 # print "row[$iR/$end_row] ";
445 # foreach (keys %{$row}) {
446 # print "$_: ",$row->{$_},"\t";
447 # }
448 # print "\n";
449
450 my $swishpath = $database."#".$iR;
451
452 next if (! $row);
453
454 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
455 $xml = $cp2utf->convert($xml);
456 use bytes; # as opposed to chars
457 print "Path-Name: $swishpath\n";
458 print "Content-Length: ".(length($xml)+1)."\n";
459 print "Document-Type: XML\n\n$xml\n";
460 }
461 }
462 } elsif ($type_base eq "marc") {
463
464 use MARC;
465
466 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
467 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
468
469 # optional argument is format
470 my $format = x($config->{format}) || 'usmarc';
471
472 print STDERR "Reading MARC file '$marc_file'\n";
473
474 my $marc = new MARC;
475 my $nr = $marc->openmarc({
476 file=>$marc_file, format=>$format
477 }) || die "Can't open MARC file '$marc_file'";
478
479 my $i=0; # record nr.
480
481 my $rec;
482
483 while ($marc->nextmarc(1)) {
484
485 # XXX
486 fakeprogress($i++);
487
488 my $swishpath = $database."#".$i;
489
490 if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
491 $xml = $cp2utf->convert($xml);
492 use bytes; # as opposed to chars
493 print "Path-Name: $swishpath\n";
494 print "Content-Length: ".(length($xml)+1)."\n";
495 print "Document-Type: XML\n\n$xml\n";
496 }
497 }
498 } elsif ($type_base eq "feed") {
499
500 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
501 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
502
503 print STDERR "Reading feed from program '$prog'\n";
504
505 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
506
507 my $i=1; # record nr.
508
509 my $data;
510 my $line=1;
511
512 while (<FEED>) {
513 chomp;
514
515 if (/^$/) {
516 my $swishpath = $database."#".$i++;
517
518 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
519 $xml = $cp2utf->convert($xml);
520 use bytes; # as opposed to chars
521 print "Path-Name: $swishpath\n";
522 print "Content-Length: ".(length($xml)+1)."\n";
523 print "Document-Type: XML\n\n$xml\n";
524 }
525 $line = 1;
526 $data = {};
527 next;
528 }
529
530 $line = $1 if (s/^(\d+):\s*//);
531 $data->{$line++} = $_;
532
533 fakeprogress($i);
534
535 }
536 }
537 }
538
539 # call this to commit index
540 $index->close;
541
542 1;
543 __END__
544 ##########################################################################
545
546 =head1 NAME
547
548 all2xml.pl - read various file formats and dump XML for SWISH-E
549
550 =head1 DESCRIPTION
551
552 This command will read ISIS data file using OpenIsis perl module, MARC
553 records using MARC module and optionally Micro$oft Excel files to
554 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
555 this script B<isn't general xml generator> from isis files (isis allready
556 has something like that). Output of this script is tailor-made for SWISH-E.
557
558 =head1 BUGS
559
560 Documentation is really lacking. However, in true Open Source spirit, source
561 is best documentation. I even made considerable effort to comment parts
562 which are not intuitively clear, so...
563
564 =head1 AUTHOR
565
566 Dobrica Pavlinusic <dpavlin@rot13.org>
567
568 =head1 COPYRIGHT
569
570 GNU Public License (GPL) v2 or later
571
572 =head1 SEE ALSO
573
574 SWISH-E web site at http://www.swish-e.org
575
576 =cut

Properties

Name Value
cvs2svn:cvs-rev 1.39
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26