/[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

Annotation of /branches/ffzg/all2xml.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 199 - (hide annotations)
Wed Jan 7 12:29:11 2004 UTC (15 years, 8 months ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 25019 byte(s)
fixed filter delimiter bug

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 177 #use GDBM_File;
13     use Fcntl; # for O_RDWR
14     use TDB_File;
15 dpavlin 1
16 dpavlin 10 $|=1;
17 dpavlin 9
18 dpavlin 13 my $config_file = $0;
19     $config_file =~ s/\.pl$/.conf/;
20     die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
21    
22 dpavlin 10 my $config;
23    
24 dpavlin 58 #use index_DBI; # default DBI module for index
25     use index_DBI_cache; # faster DBI module using memory cache
26 dpavlin 50 my $index;
27 dpavlin 10
28 dpavlin 1 my %opts;
29    
30 dpavlin 7 # usage:
31     # -d directory name
32     # -m multiple directories
33     # -q quiet
34     # -s run swish
35 dpavlin 1
36 dpavlin 7 getopts('d:m:qs', \%opts);
37    
38 dpavlin 43 my $path; # this is name of database
39 dpavlin 1
40 dpavlin 57 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
41 dpavlin 1
42 dpavlin 40 # this is encoding of all files on disk, including import_xml/*.xml file and
43     # filter/*.pm files! It will be used to store strings in perl internally!
44     my $codepage = 'ISO-8859-2';
45 dpavlin 1
46 dpavlin 40 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
47     # this function will convert data from XML files to local encoding
48     sub x {
49     return $utf2cp->convert($_[0]);
50     }
51 dpavlin 3
52 dpavlin 54 # decode isis/excel or other import codepage
53     my $import2cp;
54 dpavlin 10
55 dpavlin 40 # outgoing xml must be in UTF-8
56     my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
57 dpavlin 29
58 dpavlin 54 # mapping between data type and tag which specify
59     # format in XML file
60     my %type2tag = (
61     'isis' => 'isis',
62 dpavlin 62 'excel' => 'column',
63     'marc' => 'marc',
64 dpavlin 67 'feed' => 'feed'
65 dpavlin 54 );
66 dpavlin 3
67 dpavlin 170 my $cache; # for cacheing
68    
69 dpavlin 177 # lookup hash (tied to file)
70     my %lhash;
71     # this option will cache all lookup entries in memory.
72     # if you are tight on memory, turn this off
73     my $use_lhash_cache = 1;
74    
75 dpavlin 195 my $last_field_name; # cache to prevent repeated fields
76    
77 dpavlin 54 sub data2xml {
78    
79 dpavlin 10 use xmlify;
80    
81 dpavlin 54 my $type = shift @_;
82 dpavlin 3 my $row = shift @_;
83 dpavlin 13 my $add_xml = shift @_;
84 dpavlin 59 # needed to read values from configuration file
85     my $cfg = shift @_;
86     my $database = shift @_;
87 dpavlin 3
88     my $xml;
89    
90 dpavlin 10 use parse_format;
91 dpavlin 3
92 dpavlin 13 my $html = ""; # html formatted display output
93 dpavlin 10
94 dpavlin 13 my %field_usage; # counter for usage of each field
95    
96 dpavlin 32 # sort subrouting using order="" attribute
97     sub by_order {
98 dpavlin 98 my $va = $config->{indexer}->{$a}->{order} ||
99     $config->{indexer}->{$a};
100     my $vb = $config->{indexer}->{$b}->{order} ||
101     $config->{indexer}->{$b};
102 dpavlin 29
103 dpavlin 98 return $va <=> $vb;
104 dpavlin 32 }
105 dpavlin 3
106 dpavlin 170 my @sorted_tags;
107 dpavlin 180 if ($cache->{tags_by_order}) {
108     @sorted_tags = @{$cache->{tags_by_order}};
109 dpavlin 170 } else {
110     @sorted_tags = sort by_order keys %{$config->{indexer}};
111 dpavlin 180 $cache->{tags_by_order} = \@sorted_tags;
112 dpavlin 170 }
113 dpavlin 32
114 dpavlin 177 # lookup key
115     my $lookup_key;
116    
117 dpavlin 178 # cache for field in pages
118 dpavlin 180 delete $cache->{display_data};
119     delete $cache->{swish_data};
120     delete $cache->{swish_exact_data};
121 dpavlin 182 delete $cache->{index_data};
122 dpavlin 188 delete $cache->{index_delimiter};
123 dpavlin 178 my @page_fields; # names of fields
124    
125    
126     # subs used to produce output
127    
128     sub get_field_name($$$) {
129     my ($config,$field,$field_usage) = @_;
130    
131     # find field name (signular, plural)
132     my $field_name = "";
133     if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) {
134     $field_name = $config->{indexer}->{$field}->{name_singular};
135     } elsif ($config->{indexer}->{$field}->{name_plural}) {
136     $field_name = $config->{indexer}->{$field}->{name_plural};
137     } elsif ($config->{indexer}->{$field}->{name}) {
138     $field_name = $config->{indexer}->{$field}->{name};
139     } else {
140     print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
141     }
142     if ($field_name) {
143 dpavlin 195 if (! $last_field_name) {
144     $last_field_name = x($field_name);
145     return $last_field_name;
146     } elsif ($field_name ne $last_field_name) {
147     $last_field_name = x($field_name);
148     return $last_field_name;
149     }
150 dpavlin 178 }
151     }
152    
153    
154     # begin real work: go field by field
155 dpavlin 170 foreach my $field (@sorted_tags) {
156    
157 dpavlin 40 $field=x($field);
158 dpavlin 13 $field_usage{$field}++;
159    
160 dpavlin 10 my $swish_data = "";
161 dpavlin 163 my $swish_exact_data = "";
162 dpavlin 3 my $display_data = "";
163 dpavlin 182 my @index_data;
164 dpavlin 35 my $line_delimiter;
165 dpavlin 3
166 dpavlin 34 my ($swish,$display);
167    
168 dpavlin 54 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
169 dpavlin 178
170     # is this field page-by-page?
171     my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
172     push @page_fields,$field if ($iterate_by_page);
173     my %page_max = ();
174     # default line_delimiter if using
175     my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
176 dpavlin 188 $cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter};
177 dpavlin 178
178 dpavlin 195 my $format_name = $config->{indexer}->{$field}->{format_name};
179     my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter};
180     if ($format_name && $format_delimiter) {
181     $cache->{format}->{$field}->{format_name} = $format_name;
182     $cache->{format}->{$field}->{format_delimiter} = $format_delimiter;
183     }
184    
185 dpavlin 54 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
186 dpavlin 3
187 dpavlin 40 my $format = x($x->{content});
188     my $delimiter = x($x->{delimiter}) || ' ';
189 dpavlin 3
190 dpavlin 178 my $repeat_off = 0; # init repeatable offset
191 dpavlin 29
192 dpavlin 177 # swish, swish_exact, display, index, index_lookup
193     # swish and display defaults
194     my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
195 dpavlin 29 $s = 0 if (lc($x->{type}) eq "display");
196     $d = 0 if (lc($x->{type}) eq "swish");
197 dpavlin 188 ($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index");
198 dpavlin 181 ($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact");
199 dpavlin 182 ($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/);
200 dpavlin 29
201 dpavlin 35 # what will separate last line from this one?
202 dpavlin 182 if ($display_data && $x->{append}) {
203 dpavlin 34 $line_delimiter = ' ';
204 dpavlin 35 } elsif ($display_data) {
205 dpavlin 34 $line_delimiter = '<br/>';
206     }
207 dpavlin 29
208 dpavlin 34 # init vars so that we go into while...
209     ($swish,$display) = (1,1);
210 dpavlin 29
211 dpavlin 97 # placeholder for all repeatable entries for index
212    
213 dpavlin 195 sub mkformat($$) {
214 dpavlin 153 my $x = shift || die "mkformat needs tag reference";
215     my $data = shift || return;
216     my $format_name = x($x->{format_name}) || return $data;
217     my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
218     my $format_delimiter = x($x->{format_delimiter});
219     my @data;
220     if ($format_delimiter) {
221     @data = split(/$format_delimiter/,$data);
222     } else {
223     push @data,$data;
224     }
225    
226     if ($fmt) {
227     my $nr = scalar $fmt =~ s/%s/%s/g;
228     if (($#data+1) == $nr) {
229     return sprintf($fmt,@data);
230     } else {
231     print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
232     return $data;
233     }
234     } else {
235     print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
236     }
237     }
238    
239 dpavlin 90 # while because of repeatable fields
240     while ($swish || $display) {
241 dpavlin 178 my $page = $repeat_off;
242     $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
243 dpavlin 54 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
244 dpavlin 90 if ($repeat_off > 1000) {
245     print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
246     last;
247     }
248 dpavlin 177
249     # is this field is lookup?
250     if ($display && $x->{lookup}) {
251 dpavlin 178 my $null = "<!-- null -->";
252 dpavlin 177 if ($use_lhash_cache) {
253     if (!defined($cache->{lhash}->{$display})) {
254     my $new_display = $lhash{$display};
255 dpavlin 178 if (defined($new_display)) {
256 dpavlin 177 #print STDERR "lookup cache store '$display' = '$new_display'\n";
257     $display = $new_display;
258     $cache->{lhash}->{$display} = $new_display;
259     } else {
260     print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
261     $display = "";
262 dpavlin 178 $cache->{lhash}->{$display} = $null;
263 dpavlin 177 }
264     } else {
265     $display = $cache->{lhash}->{$display};
266     }
267     } else {
268 dpavlin 178 $display = $lhash{$display} || $null;
269 dpavlin 177 }
270     }
271    
272 dpavlin 29 # filter="name" ; filter this field through
273     # filter/[name].pm
274     my $filter = $x->{filter};
275 dpavlin 170 if ($filter && !$cache->{filter_loaded}->{$filter}) {
276 dpavlin 29 require "filter/".$filter.".pm";
277 dpavlin 170 $cache->{filter_loaded}->{$filter}++;
278 dpavlin 20 }
279 dpavlin 29 # type="swish" ; field for swish
280 dpavlin 163 if ($swish) {
281     if ($filter && ($s || $se)) {
282 dpavlin 29 no strict 'refs';
283 dpavlin 163 my $tmp = join(" ",&$filter($swish)) if ($s || $se);
284     $swish_data .= $tmp if ($s);
285 dpavlin 197 $swish_exact_data .= "xxbxx $tmp xxexx " if ($se && $tmp ne "");
286 dpavlin 163
287 dpavlin 29 } else {
288 dpavlin 163 $swish_data .= $swish if ($s);
289 dpavlin 197 $swish_exact_data .= "xxbxx $swish xxexx " if ($se && $swish ne "");
290 dpavlin 29 }
291     }
292 dpavlin 17
293 dpavlin 29 # type="display" ; field for display
294     if ($d && $display) {
295 dpavlin 35 if ($line_delimiter && $display_data) {
296     $display_data .= $line_delimiter;
297     }
298 dpavlin 29 if ($filter) {
299     no strict 'refs';
300 dpavlin 138 if ($display_data) {
301 dpavlin 199 $display_data .= $delimiter.mkformat($x,join($delimiter,&$filter($display)));
302 dpavlin 138 } else {
303 dpavlin 199 $display_data = mkformat($x,join($delimiter,&$filter($display)));
304 dpavlin 138 }
305 dpavlin 29 } else {
306 dpavlin 138 if ($display_data) {
307 dpavlin 153 $display_data .= $delimiter.mkformat($x,$display);
308 dpavlin 138 } else {
309 dpavlin 153 $display_data = mkformat($x,$display);
310 dpavlin 138 }
311 dpavlin 29 }
312 dpavlin 20 }
313 dpavlin 29
314     # type="index" ; insert into index
315 dpavlin 182 my $idisplay;
316 dpavlin 29 if ($i && $display) {
317 dpavlin 182 $idisplay = $display;
318 dpavlin 177 if ($filter) {
319     no strict 'refs';
320 dpavlin 182 $idisplay = &$filter($idisplay);
321 dpavlin 177 }
322 dpavlin 182 push @index_data, $idisplay if (! $iterate_by_page);
323 dpavlin 97 }
324    
325 dpavlin 177 # store fields in lookup
326     if ($il && $display) {
327     if (lc($x->{type}) eq "lookup_key") {
328     if ($lookup_key) {
329     print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
330     } else {
331     $lookup_key = $display;
332     }
333     } elsif (lc($x->{type}) eq "lookup_val") {
334     if ($lookup_key) {
335     $lhash{$lookup_key} = $display;
336     } else {
337     print STDERR "WARNING: no lookup_key defined for '$display'?";
338     }
339 dpavlin 20 }
340 dpavlin 178
341 dpavlin 20 }
342 dpavlin 178
343     # store data for page-by-page repeatable fields
344     if ($iterate_by_page) {
345     sub iterate_fld($$$$$$) {
346     my ($cache,$what,$field,$page,$data,$append) = @_;
347     return if (!$data);
348 dpavlin 182
349     my $ldel = $page_line_delimiter;
350     $ldel = " " if ($append);
351     #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel);
352 dpavlin 178 if (! $cache->{$what}->{$field}->[$page]) {
353     $cache->{$what}->{$field}->[$page] = $data;
354     } else {
355 dpavlin 182 $cache->{$what}->{$field}->[$page] .= $ldel.$data;
356 dpavlin 178 }
357     }
358    
359     if ($display_data) {
360     iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append});
361     }
362     $display_data = "";
363     if ($swish_data) {
364     iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append});
365     $swish_data = "";
366     }
367     if ($swish_exact_data) {
368     iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append});
369     $swish_exact_data = "";
370     }
371 dpavlin 182
372     if ($idisplay) {
373     my $ldel=$page_line_delimiter;
374     my @index_data;
375     if ($cache->{index_data}->{$field}->[$page]) {
376    
377     @index_data = @{$cache->{index_data}->{$field}->[$page]};
378     }
379     if ($x->{append}) {
380 dpavlin 195 if (@index_data) {
381     $index_data[$#index_data] .= $idisplay;
382     } else {
383     push @index_data, $idisplay;
384     }
385 dpavlin 182 } else {
386     push @index_data, $idisplay;
387     }
388     $idisplay = "";
389     @{$cache->{index_data}->{$field}->[$page]} = @index_data;
390     }
391 dpavlin 178 }
392 dpavlin 17 }
393 dpavlin 177
394 dpavlin 182 if (! $iterate_by_page) {
395 dpavlin 188 my $idel = $x->{index_delimiter};
396 dpavlin 182 # fill data in index
397 dpavlin 188 foreach my $tmp (@index_data) {
398     my $i = $d = $tmp;
399     if ($idel && $tmp =~ m/$idel/) {
400     ($i,$d) = split(/$idel/,$tmp);
401     }
402     $index->insert($field, $i, $d, $path);
403 dpavlin 182 }
404     @index_data = ();
405 dpavlin 177 }
406 dpavlin 3 }
407 dpavlin 9
408 dpavlin 59 # now try to parse variables from configuration file
409     foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
410 dpavlin 13
411 dpavlin 62 my $delimiter = x($x->{delimiter}) || ' ';
412 dpavlin 59 my $val = $cfg->val($database, x($x->{content}));
413    
414     my ($s,$d,$i) = (1,1,0); # swish, display default
415     $s = 0 if (lc($x->{type}) eq "display");
416     $d = 0 if (lc($x->{type}) eq "swish");
417 dpavlin 163 # no support for swish exact in config.
418     # IMHO, it's useless
419 dpavlin 59 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
420    
421     if ($val) {
422 dpavlin 62 $display_data .= $delimiter.$val if ($d);
423 dpavlin 59 $swish_data .= $val if ($s);
424     $index->insert($field, $val, $path) if ($i);
425     }
426    
427 dpavlin 178 if ($iterate_by_page) {
428     # FIXME data from config tag will appear just
429     # on first page!!!
430     my $page = 0;
431     if ($display_data) {
432     $cache->{display_data}->{$field}->[$page] = $display_data;
433     $display_data = "";
434     }
435     if ($swish_data) {
436     $cache->{swish_data}->{$field}->[$page] = $swish_data;
437     $swish_data = "";
438     }
439     if ($swish_exact_data) {
440     $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
441     $swish_exact_data = "";
442     }
443     }
444 dpavlin 59 }
445    
446 dpavlin 178 # save data page-by-page
447     foreach my $field (@page_fields) {
448     my $nr_pages = $page_max{$field} || next;
449     #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
450     #print STDERR Dumper($cache->{display_data});
451     for (my $page=0; $page <= $nr_pages; $page++) {
452 dpavlin 195 my $display_data;
453     if ($cache->{format}->{$field}) {
454     $display_data=mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
455     } else {
456     $display_data = $cache->{display_data}->{$field}->[$page];
457     }
458 dpavlin 178 if ($display_data) { # default
459     if ($field eq "headline") {
460     $xml .= xmlify("headline", $display_data);
461     } else {
462 dpavlin 29
463 dpavlin 178 # fallback to empty field name if needed
464     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
465     $html .= "#-#".$display_data."###\n";
466     }
467     }
468    
469     my $swish_data = $cache->{swish_data}->{$field}->[$page];
470     if ($swish_data) {
471     # remove extra spaces
472     $swish_data =~ s/ +/ /g;
473     $swish_data =~ s/ +$//g;
474 dpavlin 13
475 dpavlin 178 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
476 dpavlin 13 }
477 dpavlin 178
478     my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
479     if ($swish_exact_data) {
480     $swish_exact_data =~ s/ +/ /g;
481     $swish_exact_data =~ s/ +$//g;
482    
483     # add delimiters before and after word.
484     # That is required to produce exact match
485 dpavlin 197 $xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data));
486 dpavlin 13 }
487 dpavlin 182
488 dpavlin 188 my $idel = $cache->{index_delimiter}->{$field};
489     foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) {
490     my $i = $tmp;
491     my $d = $tmp;
492     if ($idel && $tmp =~ m/$idel/) {
493     ($i,$d) = split(/$idel/,$tmp);
494     }
495     $index->insert($field, $i, $d, $path);
496     #print STDERR "index [$idel] $field: $i --> $d [$path]\n";
497 dpavlin 182 }
498 dpavlin 13 }
499 dpavlin 178
500 dpavlin 10 }
501 dpavlin 178
502     if (! $iterate_by_page) {
503     if ($display_data) {
504     if ($field eq "headline") {
505     $xml .= xmlify("headline", $display_data);
506     } else {
507 dpavlin 20
508 dpavlin 178 # fallback to empty field name if needed
509     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
510     $html .= "#-#".$display_data."###\n";
511     }
512     }
513     if ($swish_data) {
514     # remove extra spaces
515     $swish_data =~ s/ +/ /g;
516     $swish_data =~ s/ +$//g;
517 dpavlin 10
518 dpavlin 178 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
519     }
520 dpavlin 9
521 dpavlin 178 if ($swish_exact_data) {
522     $swish_exact_data =~ s/ +/ /g;
523     $swish_exact_data =~ s/ +$//g;
524    
525     # add delimiters before and after word.
526     # That is required to produce exact match
527 dpavlin 197 $xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data));
528 dpavlin 178 }
529 dpavlin 163 }
530 dpavlin 3 }
531 dpavlin 13
532     # dump formatted output in <html>
533     if ($html) {
534 dpavlin 81 #$xml .= xmlify("html",$html);
535     $xml .= "<html><![CDATA[ $html ]]></html>";
536 dpavlin 13 }
537    
538 dpavlin 3 if ($xml) {
539 dpavlin 13 $xml .= $add_xml if ($add_xml);
540 dpavlin 10 return "<xml>\n$xml</xml>\n";
541 dpavlin 3 } else {
542     return;
543     }
544     }
545    
546     ##########################################################################
547    
548 dpavlin 54 # read configuration for this script
549 dpavlin 13 my $cfg = new Config::IniFiles( -file => $config_file );
550 dpavlin 1
551 dpavlin 54 # read global.conf configuration
552     my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
553    
554 dpavlin 50 # open index
555     $index = new index_DBI(
556 dpavlin 54 $cfg_global->val('global', 'dbi_dbd'),
557     $cfg_global->val('global', 'dbi_dsn'),
558     $cfg_global->val('global', 'dbi_user'),
559     $cfg_global->val('global', 'dbi_passwd') || '',
560 dpavlin 50 );
561    
562 dpavlin 97 my $show_progress = $cfg_global->val('global', 'show_progress');
563    
564 dpavlin 164 my $unac_filter = $cfg_global->val('global', 'unac_filter');
565     if ($unac_filter) {
566     require $unac_filter;
567     }
568    
569 dpavlin 13 foreach my $database ($cfg->Sections) {
570 dpavlin 1
571 dpavlin 54 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
572 dpavlin 40 my $add_xml = $cfg -> val($database, 'xml'); # optional
573 dpavlin 1
574 dpavlin 177 # create new lookup file
575     my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
576     if ($lookup_file) {
577     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
578     tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
579     print STDERR "creating lookup file '$lookup_file'\n";
580     }
581    
582     # open existing lookup file
583     $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
584     if ($lookup_file) {
585     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
586     tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
587     print STDERR "opening lookup file '$lookup_file'\n";
588     }
589    
590 dpavlin 54 print STDERR "reading ./import_xml/$type.xml\n";
591 dpavlin 1
592 dpavlin 58 # extract just type basic
593     my $type_base = $type;
594     $type_base =~ s/_.+$//g;
595 dpavlin 40
596 dpavlin 153 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
597 dpavlin 58
598 dpavlin 54 # output current progress indicator
599     my $last_p = 0;
600     sub progress {
601 dpavlin 101 return if (! $show_progress);
602 dpavlin 54 my $current = shift;
603 dpavlin 57 my $total = shift || 1;
604 dpavlin 54 my $p = int($current * 100 / $total);
605     if ($p != $last_p) {
606     printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
607     $last_p = $p;
608     }
609 dpavlin 3 }
610    
611 dpavlin 67 my $fake_dir = 1;
612     sub fakeprogress {
613 dpavlin 104 return if (! $show_progress);
614 dpavlin 67 my $current = shift @_;
615    
616     my @ind = ('-','\\','|','/','-','\\','|','/', '-');
617    
618     $last_p += $fake_dir;
619     $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
620     if ($last_p % 10 == 0) {
621     printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
622     }
623     }
624    
625 dpavlin 54 # now read database
626     print STDERR "using: $type...\n";
627 dpavlin 1
628 dpavlin 180 # erase cache for tags by order in this database
629     delete $cache->{tags_by_order};
630    
631 dpavlin 58 if ($type_base eq "isis") {
632    
633 dpavlin 54 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
634 dpavlin 1
635 dpavlin 54 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
636     my $db = OpenIsis::open( $isis_db );
637 dpavlin 10
638 dpavlin 106 # check if .txt database for OpenIsis is zero length,
639     # if so, erase it and re-open database
640     sub check_txt_db {
641     my $isis_db = shift || die "need isis database name";
642 dpavlin 109 my $reopen = 0;
643    
644 dpavlin 106 if (-e $isis_db.".TXT") {
645 dpavlin 109 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
646 dpavlin 108 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
647 dpavlin 109 $reopen++;
648 dpavlin 106 }
649 dpavlin 109 if (-e $isis_db.".PTR") {
650     print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
651     unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
652     $reopen++;
653     }
654     return OpenIsis::open( $isis_db ) if ($reopen);
655 dpavlin 106 }
656    
657     # EOF error
658     if ($db == -1) {
659     $db = check_txt_db($isis_db);
660 dpavlin 108 if ($db == -1) {
661 dpavlin 106 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
662     next;
663     }
664     }
665    
666 dpavlin 101 # OpenIsis::ERR_BADF
667     if ($db == -4) {
668     print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
669     next;
670     # OpenIsis::ERR_IO
671     } elsif ($db == -5) {
672     print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
673     next;
674     } elsif ($db < 0) {
675     print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
676     next;
677     }
678    
679 dpavlin 54 my $max_rowid = OpenIsis::maxRowid( $db );
680 dpavlin 3
681 dpavlin 106 # if 0 records, try to rease isis .txt database
682     if ($max_rowid == 0) {
683     # force removal of database
684     $db = check_txt_db($isis_db);
685     $max_rowid = OpenIsis::maxRowid( $db );
686     }
687    
688 dpavlin 54 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
689    
690     my $path = $database;
691    
692     for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
693     my $row = OpenIsis::read( $db, $row_id );
694     if ($row && $row->{mfn}) {
695    
696     progress($row->{mfn}, $max_rowid);
697    
698     my $swishpath = $path."#".int($row->{mfn});
699    
700 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
701 dpavlin 54 $xml = $cp2utf->convert($xml);
702     use bytes; # as opposed to chars
703     print "Path-Name: $swishpath\n";
704     print "Content-Length: ".(length($xml)+1)."\n";
705     print "Document-Type: XML\n\n$xml\n";
706     }
707 dpavlin 3 }
708 dpavlin 54 }
709 dpavlin 108 # for this to work with current version of OpenIsis (0.9.0)
710     # you might need my patch from
711     # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
712     OpenIsis::close($db);
713 dpavlin 54 print STDERR "\n";
714 dpavlin 3
715 dpavlin 58 } elsif ($type_base eq "excel") {
716 dpavlin 54 use Spreadsheet::ParseExcel;
717     use Spreadsheet::ParseExcel::Utility qw(int2col);
718    
719     $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
720     my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
721 dpavlin 43
722 dpavlin 54 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
723 dpavlin 56 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
724 dpavlin 54
725     my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
726    
727     my $sheet_nr = 0;
728     foreach my $oWks (@{$oBook->{Worksheet}}) {
729     #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
730     last if ($oWks->{Name} eq $sheet);
731     $sheet_nr++;
732     }
733    
734     my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
735    
736     print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
737     defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
738     my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
739    
740 dpavlin 56 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
741 dpavlin 54 my $row;
742     for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
743     my $cell = $oWorksheet->{Cells}[$iR][$iC];
744     if ($cell) {
745     $row->{int2col($iC)} = $cell->Value;
746     }
747     }
748    
749     progress($iR, $end_row);
750    
751     # print "row[$iR/$end_row] ";
752     # foreach (keys %{$row}) {
753     # print "$_: ",$row->{$_},"\t";
754     # }
755     # print "\n";
756    
757     my $swishpath = $database."#".$iR;
758    
759     next if (! $row);
760    
761 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
762 dpavlin 44 $xml = $cp2utf->convert($xml);
763 dpavlin 35 use bytes; # as opposed to chars
764 dpavlin 43 print "Path-Name: $swishpath\n";
765 dpavlin 10 print "Content-Length: ".(length($xml)+1)."\n";
766     print "Document-Type: XML\n\n$xml\n";
767 dpavlin 3 }
768 dpavlin 1 }
769 dpavlin 62 } elsif ($type_base eq "marc") {
770 dpavlin 67
771 dpavlin 62 use MARC;
772    
773     $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
774     my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
775    
776     # optional argument is format
777     my $format = x($config->{format}) || 'usmarc';
778    
779     print STDERR "Reading MARC file '$marc_file'\n";
780    
781     my $marc = new MARC;
782     my $nr = $marc->openmarc({
783     file=>$marc_file, format=>$format
784     }) || die "Can't open MARC file '$marc_file'";
785    
786     my $i=0; # record nr.
787    
788     my $rec;
789    
790     while ($marc->nextmarc(1)) {
791    
792     # XXX
793 dpavlin 67 fakeprogress($i++);
794 dpavlin 62
795     my $swishpath = $database."#".$i;
796    
797     if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
798     $xml = $cp2utf->convert($xml);
799     use bytes; # as opposed to chars
800     print "Path-Name: $swishpath\n";
801     print "Content-Length: ".(length($xml)+1)."\n";
802     print "Document-Type: XML\n\n$xml\n";
803     }
804     }
805 dpavlin 67 } elsif ($type_base eq "feed") {
806    
807     $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
808     my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
809    
810     print STDERR "Reading feed from program '$prog'\n";
811    
812     open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
813    
814 dpavlin 74 my $i=1; # record nr.
815 dpavlin 67
816     my $data;
817     my $line=1;
818    
819     while (<FEED>) {
820     chomp;
821    
822     if (/^$/) {
823 dpavlin 74 my $swishpath = $database."#".$i++;
824 dpavlin 67
825     if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
826     $xml = $cp2utf->convert($xml);
827     use bytes; # as opposed to chars
828     print "Path-Name: $swishpath\n";
829     print "Content-Length: ".(length($xml)+1)."\n";
830     print "Document-Type: XML\n\n$xml\n";
831     }
832     $line = 1;
833     $data = {};
834     next;
835     }
836    
837 dpavlin 74 $line = $1 if (s/^(\d+):\s*//);
838 dpavlin 67 $data->{$line++} = $_;
839    
840 dpavlin 74 fakeprogress($i);
841 dpavlin 67
842     }
843 dpavlin 177 # close lookup
844     untie %lhash if (%lhash);
845 dpavlin 1 }
846     }
847 dpavlin 3
848 dpavlin 10 # call this to commit index
849     $index->close;
850 dpavlin 3
851     1;
852     __END__
853     ##########################################################################
854    
855     =head1 NAME
856    
857 dpavlin 81 all2xml.pl - read various file formats and dump XML for SWISH-E
858 dpavlin 3
859     =head1 DESCRIPTION
860    
861 dpavlin 81 This command will read ISIS data file using OpenIsis perl module, MARC
862     records using MARC module and optionally Micro$oft Excel files to
863     create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
864     this script B<isn't general xml generator> from isis files (isis allready
865     has something like that). Output of this script is tailor-made for SWISH-E.
866 dpavlin 3
867 dpavlin 81 =head1 BUGS
868    
869     Documentation is really lacking. However, in true Open Source spirit, source
870     is best documentation. I even made considerable effort to comment parts
871     which are not intuitively clear, so...
872    
873 dpavlin 3 =head1 AUTHOR
874    
875     Dobrica Pavlinusic <dpavlin@rot13.org>
876    
877     =head1 COPYRIGHT
878    
879     GNU Public License (GPL) v2 or later
880    
881     =head1 SEE ALSO
882    
883     SWISH-E web site at http://www.swish-e.org
884    
885     =cut

Properties

Name Value
cvs2svn:cvs-rev 1.57
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26