/[webpac]/trunk/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

Annotation of /trunk/all2xml.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 144 - (hide annotations)
Sun Nov 16 11:55:18 2003 UTC (20 years, 4 months ago) by dpavlin
File MIME type: text/plain
File size: 15286 byte(s)
fixed filters (again)

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 dpavlin 58 #use index_DBI; # default DBI module for index
22     use index_DBI_cache; # faster DBI module using memory cache
23 dpavlin 50 my $index;
24 dpavlin 10
25 dpavlin 1 my %opts;
26    
27 dpavlin 7 # usage:
28     # -d directory name
29     # -m multiple directories
30     # -q quiet
31     # -s run swish
32 dpavlin 1
33 dpavlin 7 getopts('d:m:qs', \%opts);
34    
35 dpavlin 43 my $path; # this is name of database
36 dpavlin 1
37 dpavlin 57 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
38 dpavlin 1
39 dpavlin 40 # 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 dpavlin 1
43 dpavlin 40 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 dpavlin 3
49 dpavlin 54 # decode isis/excel or other import codepage
50     my $import2cp;
51 dpavlin 10
52 dpavlin 40 # outgoing xml must be in UTF-8
53     my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
54 dpavlin 29
55 dpavlin 54 # mapping between data type and tag which specify
56     # format in XML file
57     my %type2tag = (
58     'isis' => 'isis',
59 dpavlin 62 'excel' => 'column',
60     'marc' => 'marc',
61 dpavlin 67 'feed' => 'feed'
62 dpavlin 54 );
63 dpavlin 3
64 dpavlin 54 sub data2xml {
65    
66 dpavlin 10 use xmlify;
67    
68 dpavlin 54 my $type = shift @_;
69 dpavlin 3 my $row = shift @_;
70 dpavlin 13 my $add_xml = shift @_;
71 dpavlin 59 # needed to read values from configuration file
72     my $cfg = shift @_;
73     my $database = shift @_;
74 dpavlin 3
75     my $xml;
76    
77 dpavlin 10 use parse_format;
78 dpavlin 3
79 dpavlin 13 my $html = ""; # html formatted display output
80 dpavlin 10
81 dpavlin 13 my %field_usage; # counter for usage of each field
82    
83 dpavlin 32 # sort subrouting using order="" attribute
84     sub by_order {
85 dpavlin 98 my $va = $config->{indexer}->{$a}->{order} ||
86     $config->{indexer}->{$a};
87     my $vb = $config->{indexer}->{$b}->{order} ||
88     $config->{indexer}->{$b};
89 dpavlin 29
90 dpavlin 98 return $va <=> $vb;
91 dpavlin 32 }
92 dpavlin 3
93 dpavlin 32 foreach my $field (sort by_order keys %{$config->{indexer}}) {
94    
95 dpavlin 40 $field=x($field);
96 dpavlin 13 $field_usage{$field}++;
97    
98 dpavlin 10 my $swish_data = "";
99 dpavlin 3 my $display_data = "";
100 dpavlin 35 my $line_delimiter;
101 dpavlin 3
102 dpavlin 34 my ($swish,$display);
103    
104 dpavlin 54 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
105     foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
106 dpavlin 3
107 dpavlin 40 my $format = x($x->{content});
108     my $delimiter = x($x->{delimiter}) || ' ';
109 dpavlin 3
110 dpavlin 54 my $repeat_off = 0; # repeatable offset
111 dpavlin 29
112     my ($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 dpavlin 35 # what will separate last line from this one?
118     if ($display_data && $x->{append} && $x->{append} eq "1") {
119 dpavlin 34 $line_delimiter = ' ';
120 dpavlin 35 } elsif ($display_data) {
121 dpavlin 34 $line_delimiter = '<br/>';
122     }
123 dpavlin 29
124 dpavlin 34 # init vars so that we go into while...
125     ($swish,$display) = (1,1);
126 dpavlin 29
127 dpavlin 97 # placeholder for all repeatable entries for index
128     my @index_data;
129     my $index_filter;
130    
131 dpavlin 90 # while because of repeatable fields
132     while ($swish || $display) {
133 dpavlin 54 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
134 dpavlin 90 if ($repeat_off > 1000) {
135     print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
136     last;
137     }
138    
139 dpavlin 29 # filter="name" ; filter this field through
140     # filter/[name].pm
141     my $filter = $x->{filter};
142 dpavlin 20 if ($filter) {
143 dpavlin 29 require "filter/".$filter.".pm";
144 dpavlin 20 }
145 dpavlin 29 # type="swish" ; field for swish
146     if ($s && $swish) {
147     if ($filter) {
148     no strict 'refs';
149     $swish_data .= join(" ",&$filter($swish));
150     } else {
151     $swish_data .= $swish;
152     }
153     }
154 dpavlin 17
155 dpavlin 29 # type="display" ; field for display
156     if ($d && $display) {
157 dpavlin 35 if ($line_delimiter && $display_data) {
158     $display_data .= $line_delimiter;
159     undef $line_delimiter;
160     }
161 dpavlin 29 if ($filter) {
162     no strict 'refs';
163 dpavlin 138 if ($display_data) {
164 dpavlin 144 $display_data .= $delimiter.join($delimiter,&$filter($display));
165 dpavlin 138 } else {
166 dpavlin 144 $display_data = join($delimiter,&$filter($display));
167 dpavlin 138 }
168 dpavlin 29 } else {
169 dpavlin 138 if ($display_data) {
170     $display_data .= $delimiter.$display;
171     } else {
172     $display_data = $display;
173     }
174 dpavlin 29 }
175 dpavlin 20 }
176 dpavlin 29
177     # type="index" ; insert into index
178     if ($i && $display) {
179 dpavlin 97 push @index_data, $display;
180     $index_filter = $filter if ($filter);
181     }
182     }
183    
184     # fill data in index
185     if (@index_data) {
186     if ($index_filter) {
187     no strict 'refs';
188 dpavlin 135 foreach my $d (@index_data) {
189     $index->insert($field, &$index_filter($d), $path);
190 dpavlin 20 }
191 dpavlin 97 } else {
192     foreach my $d (@index_data) {
193     $index->insert($field, $d, $path);
194     }
195 dpavlin 20 }
196 dpavlin 17 }
197 dpavlin 3 }
198 dpavlin 9
199 dpavlin 59 # now try to parse variables from configuration file
200     foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
201 dpavlin 13
202 dpavlin 62 my $delimiter = x($x->{delimiter}) || ' ';
203 dpavlin 59 my $val = $cfg->val($database, x($x->{content}));
204    
205     my ($s,$d,$i) = (1,1,0); # swish, display default
206     $s = 0 if (lc($x->{type}) eq "display");
207     $d = 0 if (lc($x->{type}) eq "swish");
208     ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
209    
210     if ($val) {
211 dpavlin 62 $display_data .= $delimiter.$val if ($d);
212 dpavlin 59 $swish_data .= $val if ($s);
213     $index->insert($field, $val, $path) if ($i);
214     }
215    
216     }
217    
218    
219 dpavlin 10 if ($display_data) {
220 dpavlin 29
221 dpavlin 13 if ($field eq "headline") {
222     $xml .= xmlify("headline", $display_data);
223     } else {
224    
225     # find field name (signular, plural)
226     my $field_name = "";
227     if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
228     $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
229     } elsif ($config->{indexer}->{$field}->{name_plural}) {
230     $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
231 dpavlin 20 } elsif ($config->{indexer}->{$field}->{name}) {
232     $field_name = $config->{indexer}->{$field}->{name}."#-#";
233 dpavlin 13 } else {
234 dpavlin 20 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
235 dpavlin 13 }
236     if ($field_name) {
237 dpavlin 40 $html .= x($field_name);
238 dpavlin 13 }
239     $html .= $display_data."###\n";
240     }
241 dpavlin 10 }
242     if ($swish_data) {
243 dpavlin 20 # remove extra spaces
244     $swish_data =~ s/ +/ /g;
245     $swish_data =~ s/ +$//g;
246    
247 dpavlin 40 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
248 dpavlin 10 }
249    
250 dpavlin 9
251 dpavlin 3 }
252 dpavlin 13
253     # dump formatted output in <html>
254     if ($html) {
255 dpavlin 81 #$xml .= xmlify("html",$html);
256     $xml .= "<html><![CDATA[ $html ]]></html>";
257 dpavlin 13 }
258    
259 dpavlin 3 if ($xml) {
260 dpavlin 13 $xml .= $add_xml if ($add_xml);
261 dpavlin 10 return "<xml>\n$xml</xml>\n";
262 dpavlin 3 } else {
263     return;
264     }
265     }
266    
267     ##########################################################################
268    
269 dpavlin 54 # read configuration for this script
270 dpavlin 13 my $cfg = new Config::IniFiles( -file => $config_file );
271 dpavlin 1
272 dpavlin 54 # read global.conf configuration
273     my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
274    
275 dpavlin 50 # open index
276     $index = new index_DBI(
277 dpavlin 54 $cfg_global->val('global', 'dbi_dbd'),
278     $cfg_global->val('global', 'dbi_dsn'),
279     $cfg_global->val('global', 'dbi_user'),
280     $cfg_global->val('global', 'dbi_passwd') || '',
281 dpavlin 50 );
282    
283 dpavlin 97 my $show_progress = $cfg_global->val('global', 'show_progress');
284    
285 dpavlin 13 foreach my $database ($cfg->Sections) {
286 dpavlin 1
287 dpavlin 54 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
288 dpavlin 40 my $add_xml = $cfg -> val($database, 'xml'); # optional
289 dpavlin 1
290 dpavlin 54 print STDERR "reading ./import_xml/$type.xml\n";
291 dpavlin 1
292 dpavlin 58 # extract just type basic
293     my $type_base = $type;
294     $type_base =~ s/_.+$//g;
295 dpavlin 40
296 dpavlin 59 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
297 dpavlin 58
298 dpavlin 54 # output current progress indicator
299     my $last_p = 0;
300     sub progress {
301 dpavlin 101 return if (! $show_progress);
302 dpavlin 54 my $current = shift;
303 dpavlin 57 my $total = shift || 1;
304 dpavlin 54 my $p = int($current * 100 / $total);
305     if ($p != $last_p) {
306     printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
307     $last_p = $p;
308     }
309 dpavlin 3 }
310    
311 dpavlin 67 my $fake_dir = 1;
312     sub fakeprogress {
313 dpavlin 104 return if (! $show_progress);
314 dpavlin 67 my $current = shift @_;
315    
316     my @ind = ('-','\\','|','/','-','\\','|','/', '-');
317    
318     $last_p += $fake_dir;
319     $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
320     if ($last_p % 10 == 0) {
321     printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
322     }
323     }
324    
325 dpavlin 54 # now read database
326     print STDERR "using: $type...\n";
327 dpavlin 1
328 dpavlin 58 if ($type_base eq "isis") {
329    
330 dpavlin 54 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
331 dpavlin 1
332 dpavlin 54 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
333     my $db = OpenIsis::open( $isis_db );
334 dpavlin 10
335 dpavlin 106 # check if .txt database for OpenIsis is zero length,
336     # if so, erase it and re-open database
337     sub check_txt_db {
338     my $isis_db = shift || die "need isis database name";
339 dpavlin 109 my $reopen = 0;
340    
341 dpavlin 106 if (-e $isis_db.".TXT") {
342 dpavlin 109 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
343 dpavlin 108 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
344 dpavlin 109 $reopen++;
345 dpavlin 106 }
346 dpavlin 109 if (-e $isis_db.".PTR") {
347     print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
348     unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
349     $reopen++;
350     }
351     return OpenIsis::open( $isis_db ) if ($reopen);
352 dpavlin 106 }
353    
354     # EOF error
355     if ($db == -1) {
356     $db = check_txt_db($isis_db);
357 dpavlin 108 if ($db == -1) {
358 dpavlin 106 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
359     next;
360     }
361     }
362    
363 dpavlin 101 # OpenIsis::ERR_BADF
364     if ($db == -4) {
365     print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
366     next;
367     # OpenIsis::ERR_IO
368     } elsif ($db == -5) {
369     print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
370     next;
371     } elsif ($db < 0) {
372     print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
373     next;
374     }
375    
376 dpavlin 54 my $max_rowid = OpenIsis::maxRowid( $db );
377 dpavlin 3
378 dpavlin 106 # if 0 records, try to rease isis .txt database
379     if ($max_rowid == 0) {
380     # force removal of database
381     $db = check_txt_db($isis_db);
382     $max_rowid = OpenIsis::maxRowid( $db );
383     }
384    
385 dpavlin 54 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
386    
387     my $path = $database;
388    
389     for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
390     my $row = OpenIsis::read( $db, $row_id );
391     if ($row && $row->{mfn}) {
392    
393     progress($row->{mfn}, $max_rowid);
394    
395     my $swishpath = $path."#".int($row->{mfn});
396    
397 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
398 dpavlin 54 $xml = $cp2utf->convert($xml);
399     use bytes; # as opposed to chars
400     print "Path-Name: $swishpath\n";
401     print "Content-Length: ".(length($xml)+1)."\n";
402     print "Document-Type: XML\n\n$xml\n";
403     }
404 dpavlin 3 }
405 dpavlin 54 }
406 dpavlin 108 # for this to work with current version of OpenIsis (0.9.0)
407     # you might need my patch from
408     # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
409     OpenIsis::close($db);
410 dpavlin 54 print STDERR "\n";
411 dpavlin 3
412 dpavlin 58 } elsif ($type_base eq "excel") {
413 dpavlin 54 use Spreadsheet::ParseExcel;
414     use Spreadsheet::ParseExcel::Utility qw(int2col);
415    
416     $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
417     my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
418 dpavlin 43
419 dpavlin 54 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
420 dpavlin 56 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
421 dpavlin 54
422     my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
423    
424     my $sheet_nr = 0;
425     foreach my $oWks (@{$oBook->{Worksheet}}) {
426     #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
427     last if ($oWks->{Name} eq $sheet);
428     $sheet_nr++;
429     }
430    
431     my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
432    
433     print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
434     defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
435     my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
436    
437 dpavlin 56 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
438 dpavlin 54 my $row;
439     for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
440     my $cell = $oWorksheet->{Cells}[$iR][$iC];
441     if ($cell) {
442     $row->{int2col($iC)} = $cell->Value;
443     }
444     }
445    
446     progress($iR, $end_row);
447    
448     # print "row[$iR/$end_row] ";
449     # foreach (keys %{$row}) {
450     # print "$_: ",$row->{$_},"\t";
451     # }
452     # print "\n";
453    
454     my $swishpath = $database."#".$iR;
455    
456     next if (! $row);
457    
458 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
459 dpavlin 44 $xml = $cp2utf->convert($xml);
460 dpavlin 35 use bytes; # as opposed to chars
461 dpavlin 43 print "Path-Name: $swishpath\n";
462 dpavlin 10 print "Content-Length: ".(length($xml)+1)."\n";
463     print "Document-Type: XML\n\n$xml\n";
464 dpavlin 3 }
465 dpavlin 1 }
466 dpavlin 62 } elsif ($type_base eq "marc") {
467 dpavlin 67
468 dpavlin 62 use MARC;
469    
470     $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
471     my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
472    
473     # optional argument is format
474     my $format = x($config->{format}) || 'usmarc';
475    
476     print STDERR "Reading MARC file '$marc_file'\n";
477    
478     my $marc = new MARC;
479     my $nr = $marc->openmarc({
480     file=>$marc_file, format=>$format
481     }) || die "Can't open MARC file '$marc_file'";
482    
483     my $i=0; # record nr.
484    
485     my $rec;
486    
487     while ($marc->nextmarc(1)) {
488    
489     # XXX
490 dpavlin 67 fakeprogress($i++);
491 dpavlin 62
492     my $swishpath = $database."#".$i;
493    
494     if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
495     $xml = $cp2utf->convert($xml);
496     use bytes; # as opposed to chars
497     print "Path-Name: $swishpath\n";
498     print "Content-Length: ".(length($xml)+1)."\n";
499     print "Document-Type: XML\n\n$xml\n";
500     }
501     }
502 dpavlin 67 } elsif ($type_base eq "feed") {
503    
504     $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
505     my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
506    
507     print STDERR "Reading feed from program '$prog'\n";
508    
509     open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
510    
511 dpavlin 74 my $i=1; # record nr.
512 dpavlin 67
513     my $data;
514     my $line=1;
515    
516     while (<FEED>) {
517     chomp;
518    
519     if (/^$/) {
520 dpavlin 74 my $swishpath = $database."#".$i++;
521 dpavlin 67
522     if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
523     $xml = $cp2utf->convert($xml);
524     use bytes; # as opposed to chars
525     print "Path-Name: $swishpath\n";
526     print "Content-Length: ".(length($xml)+1)."\n";
527     print "Document-Type: XML\n\n$xml\n";
528     }
529     $line = 1;
530     $data = {};
531     next;
532     }
533    
534 dpavlin 74 $line = $1 if (s/^(\d+):\s*//);
535 dpavlin 67 $data->{$line++} = $_;
536    
537 dpavlin 74 fakeprogress($i);
538 dpavlin 67
539     }
540 dpavlin 1 }
541     }
542 dpavlin 3
543 dpavlin 10 # call this to commit index
544     $index->close;
545 dpavlin 3
546     1;
547     __END__
548     ##########################################################################
549    
550     =head1 NAME
551    
552 dpavlin 81 all2xml.pl - read various file formats and dump XML for SWISH-E
553 dpavlin 3
554     =head1 DESCRIPTION
555    
556 dpavlin 81 This command will read ISIS data file using OpenIsis perl module, MARC
557     records using MARC module and optionally Micro$oft Excel files to
558     create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
559     this script B<isn't general xml generator> from isis files (isis allready
560     has something like that). Output of this script is tailor-made for SWISH-E.
561 dpavlin 3
562 dpavlin 81 =head1 BUGS
563    
564     Documentation is really lacking. However, in true Open Source spirit, source
565     is best documentation. I even made considerable effort to comment parts
566     which are not intuitively clear, so...
567    
568 dpavlin 3 =head1 AUTHOR
569    
570     Dobrica Pavlinusic <dpavlin@rot13.org>
571    
572     =head1 COPYRIGHT
573    
574     GNU Public License (GPL) v2 or later
575    
576     =head1 SEE ALSO
577    
578     SWISH-E web site at http://www.swish-e.org
579    
580     =cut

Properties

Name Value
cvs2svn:cvs-rev 1.43
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26