/[webpac]/branches/hidra/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 /branches/hidra/all2xml.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 136 - (hide annotations)
Wed Oct 29 22:46:49 2003 UTC (16 years, 1 month ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 15074 byte(s)
another fix for repeatable fields

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

Properties

Name Value
cvs2svn:cvs-rev 1.40
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26