/[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 601 - (hide annotations)
Wed Dec 8 18:45:32 2004 UTC (19 years, 3 months ago) by dpavlin
File MIME type: text/plain
File size: 26873 byte(s)
updated branches to trunk

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 10 use Text::Iconv;
9 dpavlin 13 use Config::IniFiles;
10 dpavlin 40 use Encode;
11 dpavlin 177 #use GDBM_File;
12     use Fcntl; # for O_RDWR
13     use TDB_File;
14 dpavlin 1
15 dpavlin 10 $|=1;
16 dpavlin 9
17 dpavlin 13 my $config_file = $0;
18     $config_file =~ s/\.pl$/.conf/;
19 dpavlin 321 $config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]);
20 dpavlin 13 die "FATAL: can't find configuration file '$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 dpavlin 601
143 dpavlin 178 if ($field_name) {
144 dpavlin 601 $field_name = x($field_name);
145 dpavlin 195 if (! $last_field_name) {
146 dpavlin 601 $last_field_name = $field_name;
147 dpavlin 195 return $last_field_name;
148     } elsif ($field_name ne $last_field_name) {
149 dpavlin 601 $last_field_name = $field_name;
150 dpavlin 195 return $last_field_name;
151     }
152 dpavlin 178 }
153     }
154    
155 dpavlin 279
156     # init variables for different types
157 dpavlin 274 sub init_visible_type($) {
158     my $type = shift;
159 dpavlin 178
160 dpavlin 274 # swish, swish_exact, display, index, index_lookup
161     # swish and display defaults
162     my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
163     if (lc($type) eq "display") {
164     $s = 0;
165     } elsif (lc($type) eq "swish") {
166     $d = 0;
167     } elsif (lc($type) eq "index") {
168     ($s,$se,$d,$i) = (0,1,0,1);
169     } elsif (lc($type) eq "swish_exact") {
170     ($s,$se,$d,$i) = (0,1,0,0);
171     } elsif (lc($type) =~ /^lookup/) {
172     ($s,$se,$d,$i,$il) = (0,1,0,0,1);
173     }
174     return ($s,$se,$d,$i,$il);
175     }
176    
177 dpavlin 279
178     # convert
179     #
180     # <tag>
181     # <delimiter>, </delimiter>
182     # <value>200a</value>
183     # </tag>
184     #
185     # to
186     #
187     # <tag delimiter=", ">200a</tag>
188     #
189     # but without loosing spaces in delimiter (becasue
190     # new XML::Simple strips spaces in attribute values
191     # as defined in XML specification)
192     #
193     sub unroll_x($) {
194     my $x = shift;
195    
196     if (defined $x->{value}) {
197     my ($v,$d) = ($x->{value}->{content}, $x->{delimiter}->{content});
198     delete $x->{value};
199     delete $x->{delimiter};
200     $x->{content} = $v;
201     $x->{delimiter} = $d;
202     }
203     return $x;
204     }
205    
206 dpavlin 178 # begin real work: go field by field
207 dpavlin 170 foreach my $field (@sorted_tags) {
208    
209 dpavlin 40 $field=x($field);
210 dpavlin 13 $field_usage{$field}++;
211    
212 dpavlin 10 my $swish_data = "";
213 dpavlin 163 my $swish_exact_data = "";
214 dpavlin 3 my $display_data = "";
215 dpavlin 182 my @index_data;
216 dpavlin 35 my $line_delimiter;
217 dpavlin 3
218 dpavlin 34 my ($swish,$display);
219    
220 dpavlin 54 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
221 dpavlin 178
222     # is this field page-by-page?
223     my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
224     push @page_fields,$field if ($iterate_by_page);
225     my %page_max = ();
226     # default line_delimiter if using
227     my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
228 dpavlin 188 $cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter};
229 dpavlin 178
230 dpavlin 195 my $format_name = $config->{indexer}->{$field}->{format_name};
231     my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter};
232     if ($format_name && $format_delimiter) {
233     $cache->{format}->{$field}->{format_name} = $format_name;
234     $cache->{format}->{$field}->{format_delimiter} = $format_delimiter;
235     }
236    
237 dpavlin 54 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
238 dpavlin 3
239 dpavlin 279 $x = unroll_x($x);
240    
241 dpavlin 40 my $format = x($x->{content});
242     my $delimiter = x($x->{delimiter}) || ' ';
243 dpavlin 3
244 dpavlin 178 my $repeat_off = 0; # init repeatable offset
245 dpavlin 29
246 dpavlin 274 my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
247 dpavlin 29
248 dpavlin 35 # what will separate last line from this one?
249 dpavlin 182 if ($display_data && $x->{append}) {
250 dpavlin 291 $line_delimiter = $delimiter;
251 dpavlin 35 } elsif ($display_data) {
252 dpavlin 34 $line_delimiter = '<br/>';
253     }
254 dpavlin 29
255 dpavlin 34 # init vars so that we go into while...
256     ($swish,$display) = (1,1);
257 dpavlin 29
258 dpavlin 97 # placeholder for all repeatable entries for index
259    
260 dpavlin 195 sub mkformat($$) {
261 dpavlin 153 my $x = shift || die "mkformat needs tag reference";
262     my $data = shift || return;
263 dpavlin 263 my $format_name = x($x->{format_name}) || return $data;
264 dpavlin 153 my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
265     my $format_delimiter = x($x->{format_delimiter});
266     my @data;
267     if ($format_delimiter) {
268     @data = split(/$format_delimiter/,$data);
269     } else {
270     push @data,$data;
271     }
272    
273     if ($fmt) {
274     my $nr = scalar $fmt =~ s/%s/%s/g;
275     if (($#data+1) == $nr) {
276 dpavlin 263 return sprintf($fmt,@data);
277 dpavlin 153 } else {
278 dpavlin 207 #print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
279 dpavlin 263 return $data;
280 dpavlin 153 }
281     } else {
282     print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
283     }
284     }
285    
286 dpavlin 90 # while because of repeatable fields
287     while ($swish || $display) {
288 dpavlin 178 my $page = $repeat_off;
289     $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
290 dpavlin 54 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
291 dpavlin 90 if ($repeat_off > 1000) {
292     print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
293     last;
294     }
295 dpavlin 177
296     # is this field is lookup?
297     if ($display && $x->{lookup}) {
298 dpavlin 178 my $null = "<!-- null -->";
299 dpavlin 177 if ($use_lhash_cache) {
300     if (!defined($cache->{lhash}->{$display})) {
301     my $new_display = $lhash{$display};
302 dpavlin 178 if (defined($new_display)) {
303 dpavlin 177 #print STDERR "lookup cache store '$display' = '$new_display'\n";
304     $display = $new_display;
305     $cache->{lhash}->{$display} = $new_display;
306     } else {
307 dpavlin 255 # print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
308 dpavlin 177 $display = "";
309 dpavlin 178 $cache->{lhash}->{$display} = $null;
310 dpavlin 177 }
311     } else {
312     $display = $cache->{lhash}->{$display};
313     }
314     } else {
315 dpavlin 178 $display = $lhash{$display} || $null;
316 dpavlin 177 }
317     }
318    
319 dpavlin 29 # filter="name" ; filter this field through
320     # filter/[name].pm
321     my $filter = $x->{filter};
322 dpavlin 170 if ($filter && !$cache->{filter_loaded}->{$filter}) {
323 dpavlin 29 require "filter/".$filter.".pm";
324 dpavlin 170 $cache->{filter_loaded}->{$filter}++;
325 dpavlin 20 }
326 dpavlin 29 # type="swish" ; field for swish
327 dpavlin 163 if ($swish) {
328 dpavlin 259 my $tmp = $swish;
329 dpavlin 163 if ($filter && ($s || $se)) {
330 dpavlin 29 no strict 'refs';
331 dpavlin 259 $tmp = join(" ",&$filter($tmp)) if ($s || $se);
332 dpavlin 29 }
333 dpavlin 234
334 dpavlin 255 $swish_data .= $tmp if ($s && $tmp);
335 dpavlin 256 $swish_exact_data .= "xxbxx $tmp xxexx " if ($tmp && $tmp ne "" && $se);
336 dpavlin 29 }
337 dpavlin 17
338 dpavlin 29 # type="display" ; field for display
339     if ($d && $display) {
340 dpavlin 207 my $ldel = $delimiter;
341 dpavlin 35 if ($line_delimiter && $display_data) {
342 dpavlin 207 $ldel = $line_delimiter;
343 dpavlin 35 }
344 dpavlin 29 if ($filter) {
345     no strict 'refs';
346 dpavlin 207 my @arr;
347     foreach my $tmp (&$filter($display)) {
348     my $tmp2 = mkformat($x,$tmp);
349     push @arr,$tmp2 if ($tmp2);
350 dpavlin 138 }
351 dpavlin 207 $display_data .= $ldel if ($display_data && @arr);
352     $display_data .= join($delimiter,@arr);
353 dpavlin 29 } else {
354 dpavlin 207 $display_data .= $ldel if ($display_data);
355     my $tmp = mkformat($x,$display);
356 dpavlin 224 $display_data .= $tmp if ($tmp);
357 dpavlin 29 }
358 dpavlin 20 }
359 dpavlin 29
360     # type="index" ; insert into index
361 dpavlin 182 my $idisplay;
362 dpavlin 29 if ($i && $display) {
363 dpavlin 182 $idisplay = $display;
364 dpavlin 177 if ($filter) {
365     no strict 'refs';
366 dpavlin 182 $idisplay = &$filter($idisplay);
367 dpavlin 177 }
368 dpavlin 255 push @index_data, $idisplay if ($idisplay && !$iterate_by_page);
369 dpavlin 97 }
370    
371 dpavlin 177 # store fields in lookup
372     if ($il && $display) {
373     if (lc($x->{type}) eq "lookup_key") {
374     if ($lookup_key) {
375     print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
376     } else {
377 dpavlin 233 if ($filter) {
378     no strict 'refs';
379     $lookup_key = &$filter($display);
380     } else {
381     $lookup_key = $display;
382     }
383 dpavlin 177 }
384     } elsif (lc($x->{type}) eq "lookup_val") {
385     if ($lookup_key) {
386 dpavlin 233 if ($filter) {
387     no strict 'refs';
388     $lhash{$lookup_key} = &$filter($display);
389     } else {
390     $lhash{$lookup_key} = $display;
391     }
392 dpavlin 177 } else {
393     print STDERR "WARNING: no lookup_key defined for '$display'?";
394     }
395 dpavlin 20 }
396 dpavlin 178
397 dpavlin 20 }
398 dpavlin 178
399     # store data for page-by-page repeatable fields
400     if ($iterate_by_page) {
401     sub iterate_fld($$$$$$) {
402     my ($cache,$what,$field,$page,$data,$append) = @_;
403     return if (!$data);
404 dpavlin 182
405     my $ldel = $page_line_delimiter;
406     $ldel = " " if ($append);
407     #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel);
408 dpavlin 178 if (! $cache->{$what}->{$field}->[$page]) {
409     $cache->{$what}->{$field}->[$page] = $data;
410     } else {
411 dpavlin 182 $cache->{$what}->{$field}->[$page] .= $ldel.$data;
412 dpavlin 178 }
413     }
414    
415     if ($display_data) {
416     iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append});
417     }
418     $display_data = "";
419     if ($swish_data) {
420     iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append});
421     $swish_data = "";
422     }
423     if ($swish_exact_data) {
424     iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append});
425     $swish_exact_data = "";
426     }
427 dpavlin 182
428     if ($idisplay) {
429     my $ldel=$page_line_delimiter;
430     my @index_data;
431     if ($cache->{index_data}->{$field}->[$page]) {
432    
433     @index_data = @{$cache->{index_data}->{$field}->[$page]};
434     }
435     if ($x->{append}) {
436 dpavlin 195 if (@index_data) {
437     $index_data[$#index_data] .= $idisplay;
438     } else {
439     push @index_data, $idisplay;
440     }
441 dpavlin 182 } else {
442     push @index_data, $idisplay;
443     }
444     $idisplay = "";
445     @{$cache->{index_data}->{$field}->[$page]} = @index_data;
446     }
447 dpavlin 178 }
448 dpavlin 17 }
449 dpavlin 177
450 dpavlin 182 if (! $iterate_by_page) {
451 dpavlin 188 my $idel = $x->{index_delimiter};
452 dpavlin 182 # fill data in index
453 dpavlin 188 foreach my $tmp (@index_data) {
454     my $i = $d = $tmp;
455     if ($idel && $tmp =~ m/$idel/) {
456     ($i,$d) = split(/$idel/,$tmp);
457     }
458     $index->insert($field, $i, $d, $path);
459 dpavlin 182 }
460     @index_data = ();
461 dpavlin 177 }
462 dpavlin 3 }
463 dpavlin 9
464 dpavlin 59 # now try to parse variables from configuration file
465     foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
466 dpavlin 13
467 dpavlin 279 $x = unroll_x($x);
468    
469 dpavlin 62 my $delimiter = x($x->{delimiter}) || ' ';
470 dpavlin 59 my $val = $cfg->val($database, x($x->{content}));
471    
472 dpavlin 274 # FIXME index_lookup is not supported!
473     my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
474 dpavlin 59
475     if ($val) {
476 dpavlin 62 $display_data .= $delimiter.$val if ($d);
477 dpavlin 274 $swish_data .= " ".$val if ($s);
478 dpavlin 59 $index->insert($field, $val, $path) if ($i);
479     }
480    
481 dpavlin 178 if ($iterate_by_page) {
482     # FIXME data from config tag will appear just
483     # on first page!!!
484     my $page = 0;
485     if ($display_data) {
486     $cache->{display_data}->{$field}->[$page] = $display_data;
487     $display_data = "";
488     }
489     if ($swish_data) {
490     $cache->{swish_data}->{$field}->[$page] = $swish_data;
491     $swish_data = "";
492     }
493     if ($swish_exact_data) {
494     $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
495     $swish_exact_data = "";
496     }
497     }
498 dpavlin 59 }
499    
500 dpavlin 178 # save data page-by-page
501     foreach my $field (@page_fields) {
502     my $nr_pages = $page_max{$field} || next;
503     #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
504     #print STDERR Dumper($cache->{display_data});
505     for (my $page=0; $page <= $nr_pages; $page++) {
506 dpavlin 195 my $display_data;
507     if ($cache->{format}->{$field}) {
508 dpavlin 207 my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
509     $display_data=$tmp if ($tmp);
510 dpavlin 195 } else {
511     $display_data = $cache->{display_data}->{$field}->[$page];
512     }
513 dpavlin 178 if ($display_data) { # default
514     if ($field eq "headline") {
515     $xml .= xmlify("headline", $display_data);
516     } else {
517 dpavlin 29
518 dpavlin 178 # fallback to empty field name if needed
519     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
520     $html .= "#-#".$display_data."###\n";
521     }
522     }
523    
524     my $swish_data = $cache->{swish_data}->{$field}->[$page];
525     if ($swish_data) {
526     # remove extra spaces
527     $swish_data =~ s/ +/ /g;
528     $swish_data =~ s/ +$//g;
529 dpavlin 13
530 dpavlin 321 $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
531 dpavlin 13 }
532 dpavlin 178
533     my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
534     if ($swish_exact_data) {
535     $swish_exact_data =~ s/ +/ /g;
536     $swish_exact_data =~ s/ +$//g;
537    
538     # add delimiters before and after word.
539     # That is required to produce exact match
540 dpavlin 321 $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
541 dpavlin 13 }
542 dpavlin 182
543 dpavlin 188 my $idel = $cache->{index_delimiter}->{$field};
544     foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) {
545     my $i = $tmp;
546     my $d = $tmp;
547     if ($idel && $tmp =~ m/$idel/) {
548     ($i,$d) = split(/$idel/,$tmp);
549     }
550     $index->insert($field, $i, $d, $path);
551     #print STDERR "index [$idel] $field: $i --> $d [$path]\n";
552 dpavlin 182 }
553 dpavlin 13 }
554 dpavlin 178
555 dpavlin 10 }
556 dpavlin 178
557     if (! $iterate_by_page) {
558     if ($display_data) {
559     if ($field eq "headline") {
560     $xml .= xmlify("headline", $display_data);
561     } else {
562 dpavlin 20
563 dpavlin 178 # fallback to empty field name if needed
564     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
565     $html .= "#-#".$display_data."###\n";
566     }
567     }
568     if ($swish_data) {
569     # remove extra spaces
570     $swish_data =~ s/ +/ /g;
571     $swish_data =~ s/ +$//g;
572 dpavlin 10
573 dpavlin 321 $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
574 dpavlin 178 }
575 dpavlin 9
576 dpavlin 178 if ($swish_exact_data) {
577     $swish_exact_data =~ s/ +/ /g;
578     $swish_exact_data =~ s/ +$//g;
579    
580     # add delimiters before and after word.
581     # That is required to produce exact match
582 dpavlin 321 $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
583 dpavlin 178 }
584 dpavlin 163 }
585 dpavlin 3 }
586 dpavlin 13
587     # dump formatted output in <html>
588     if ($html) {
589 dpavlin 81 #$xml .= xmlify("html",$html);
590     $xml .= "<html><![CDATA[ $html ]]></html>";
591 dpavlin 13 }
592    
593 dpavlin 3 if ($xml) {
594 dpavlin 13 $xml .= $add_xml if ($add_xml);
595 dpavlin 10 return "<xml>\n$xml</xml>\n";
596 dpavlin 3 } else {
597     return;
598     }
599     }
600    
601     ##########################################################################
602    
603 dpavlin 54 # read configuration for this script
604 dpavlin 13 my $cfg = new Config::IniFiles( -file => $config_file );
605 dpavlin 1
606 dpavlin 54 # read global.conf configuration
607     my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
608    
609 dpavlin 50 # open index
610     $index = new index_DBI(
611 dpavlin 54 $cfg_global->val('global', 'dbi_dbd'),
612     $cfg_global->val('global', 'dbi_dsn'),
613     $cfg_global->val('global', 'dbi_user'),
614     $cfg_global->val('global', 'dbi_passwd') || '',
615 dpavlin 50 );
616    
617 dpavlin 97 my $show_progress = $cfg_global->val('global', 'show_progress');
618    
619 dpavlin 321 my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter');
620     if ($my_unac_filter) {
621     print STDERR "using $my_unac_filter to filter characters for search\n";
622     require $my_unac_filter;
623     } else {
624     print STDERR "### fallback to default my_unac_string!\n";
625     eval q{
626     sub main::my_unac_string($$) {
627     my ($charset, $string) = (@_);
628     return $string;
629     }
630     };
631 dpavlin 164 }
632    
633 dpavlin 13 foreach my $database ($cfg->Sections) {
634 dpavlin 1
635 dpavlin 54 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
636 dpavlin 40 my $add_xml = $cfg -> val($database, 'xml'); # optional
637 dpavlin 1
638 dpavlin 177 # create new lookup file
639     my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
640     if ($lookup_file) {
641     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
642 dpavlin 458 if (! -e $lookup_file) {
643     open(LOOKUP, "> $lookup_file") || die "can't create $lookup_file': $!";
644     close(LOOKUP);
645     }
646 dpavlin 177 tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
647     print STDERR "creating lookup file '$lookup_file'\n";
648 dpavlin 231 # delete memory cache for lookup file
649     delete $cache->{lhash};
650 dpavlin 177 }
651    
652     # open existing lookup file
653     $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
654     if ($lookup_file) {
655     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
656     tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
657     print STDERR "opening lookup file '$lookup_file'\n";
658     }
659    
660 dpavlin 54 print STDERR "reading ./import_xml/$type.xml\n";
661 dpavlin 1
662 dpavlin 58 # extract just type basic
663     my $type_base = $type;
664     $type_base =~ s/_.+$//g;
665 dpavlin 40
666 dpavlin 279 $config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 );
667 dpavlin 58
668 dpavlin 54 # output current progress indicator
669     my $last_p = 0;
670     sub progress {
671 dpavlin 101 return if (! $show_progress);
672 dpavlin 54 my $current = shift;
673 dpavlin 57 my $total = shift || 1;
674 dpavlin 54 my $p = int($current * 100 / $total);
675     if ($p != $last_p) {
676     printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
677     $last_p = $p;
678     }
679 dpavlin 3 }
680    
681 dpavlin 67 my $fake_dir = 1;
682     sub fakeprogress {
683 dpavlin 104 return if (! $show_progress);
684 dpavlin 67 my $current = shift @_;
685    
686     my @ind = ('-','\\','|','/','-','\\','|','/', '-');
687    
688     $last_p += $fake_dir;
689     $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
690     if ($last_p % 10 == 0) {
691     printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
692     }
693     }
694    
695 dpavlin 54 # now read database
696     print STDERR "using: $type...\n";
697 dpavlin 1
698 dpavlin 180 # erase cache for tags by order in this database
699     delete $cache->{tags_by_order};
700    
701 dpavlin 58 if ($type_base eq "isis") {
702    
703 dpavlin 54 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
704 dpavlin 1
705 dpavlin 54 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
706     my $db = OpenIsis::open( $isis_db );
707 dpavlin 10
708 dpavlin 106 # check if .txt database for OpenIsis is zero length,
709     # if so, erase it and re-open database
710     sub check_txt_db {
711     my $isis_db = shift || die "need isis database name";
712 dpavlin 109 my $reopen = 0;
713    
714 dpavlin 106 if (-e $isis_db.".TXT") {
715 dpavlin 109 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
716 dpavlin 108 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
717 dpavlin 109 $reopen++;
718 dpavlin 106 }
719 dpavlin 109 if (-e $isis_db.".PTR") {
720     print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
721     unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
722     $reopen++;
723     }
724     return OpenIsis::open( $isis_db ) if ($reopen);
725 dpavlin 106 }
726    
727     # EOF error
728     if ($db == -1) {
729     $db = check_txt_db($isis_db);
730 dpavlin 108 if ($db == -1) {
731 dpavlin 106 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
732     next;
733     }
734     }
735    
736 dpavlin 101 # OpenIsis::ERR_BADF
737     if ($db == -4) {
738     print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
739     next;
740     # OpenIsis::ERR_IO
741     } elsif ($db == -5) {
742     print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
743     next;
744     } elsif ($db < 0) {
745     print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
746     next;
747     }
748    
749 dpavlin 54 my $max_rowid = OpenIsis::maxRowid( $db );
750 dpavlin 3
751 dpavlin 106 # if 0 records, try to rease isis .txt database
752     if ($max_rowid == 0) {
753     # force removal of database
754     $db = check_txt_db($isis_db);
755     $max_rowid = OpenIsis::maxRowid( $db );
756     }
757    
758 dpavlin 54 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
759    
760     my $path = $database;
761    
762     for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
763     my $row = OpenIsis::read( $db, $row_id );
764     if ($row && $row->{mfn}) {
765    
766     progress($row->{mfn}, $max_rowid);
767    
768     my $swishpath = $path."#".int($row->{mfn});
769    
770 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
771 dpavlin 54 $xml = $cp2utf->convert($xml);
772     use bytes; # as opposed to chars
773     print "Path-Name: $swishpath\n";
774     print "Content-Length: ".(length($xml)+1)."\n";
775     print "Document-Type: XML\n\n$xml\n";
776     }
777 dpavlin 3 }
778 dpavlin 54 }
779 dpavlin 108 # for this to work with current version of OpenIsis (0.9.0)
780     # you might need my patch from
781     # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
782     OpenIsis::close($db);
783 dpavlin 54 print STDERR "\n";
784 dpavlin 3
785 dpavlin 58 } elsif ($type_base eq "excel") {
786 dpavlin 255 require Spreadsheet::ParseExcel;
787     require Spreadsheet::ParseExcel::Utility;
788     import Spreadsheet::ParseExcel::Utility qw(int2col);
789 dpavlin 54
790     $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
791     my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
792 dpavlin 43
793 dpavlin 54 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
794 dpavlin 56 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
795 dpavlin 54
796     my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
797    
798     my $sheet_nr = 0;
799     foreach my $oWks (@{$oBook->{Worksheet}}) {
800     #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
801     last if ($oWks->{Name} eq $sheet);
802     $sheet_nr++;
803     }
804    
805     my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
806    
807     print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
808     defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
809     my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
810    
811 dpavlin 56 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
812 dpavlin 54 my $row;
813     for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
814     my $cell = $oWorksheet->{Cells}[$iR][$iC];
815     if ($cell) {
816 dpavlin 489 # this conversion is a cludge.
817     # Files from Excell could have
818     # characters which don't fit into
819     # destination encoding.
820     $row->{int2col($iC)} = $utf2cp->convert($cell->Value) || $cell->Value;
821 dpavlin 54 }
822     }
823    
824     progress($iR, $end_row);
825    
826     # print "row[$iR/$end_row] ";
827     # foreach (keys %{$row}) {
828     # print "$_: ",$row->{$_},"\t";
829     # }
830     # print "\n";
831    
832     my $swishpath = $database."#".$iR;
833    
834     next if (! $row);
835    
836 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
837 dpavlin 44 $xml = $cp2utf->convert($xml);
838 dpavlin 35 use bytes; # as opposed to chars
839 dpavlin 43 print "Path-Name: $swishpath\n";
840 dpavlin 10 print "Content-Length: ".(length($xml)+1)."\n";
841     print "Document-Type: XML\n\n$xml\n";
842 dpavlin 3 }
843 dpavlin 1 }
844 dpavlin 62 } elsif ($type_base eq "marc") {
845 dpavlin 67
846 dpavlin 255 require MARC;
847 dpavlin 62
848     $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
849     my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
850    
851     # optional argument is format
852 dpavlin 218 my $format = x($config->{marc_format}) || 'usmarc';
853 dpavlin 62
854     print STDERR "Reading MARC file '$marc_file'\n";
855    
856     my $marc = new MARC;
857     my $nr = $marc->openmarc({
858     file=>$marc_file, format=>$format
859 dpavlin 218 }) || die "Can't open MARC file '$marc_file' with format '$format'";
860 dpavlin 62
861 dpavlin 215 # read MARC file in memory
862     $marc->nextmarc(-1);
863 dpavlin 62
864 dpavlin 215 my $max_rec = $marc->marc_count();
865 dpavlin 62
866 dpavlin 215 for(my $i=1; $i<=$max_rec; $i++) {
867 dpavlin 62
868 dpavlin 215 progress($i,$max_rec);
869 dpavlin 62
870 dpavlin 215 # store value for marc_sf.pm
871     $main::cache->{marc_record} = $i;
872    
873 dpavlin 62 my $swishpath = $database."#".$i;
874    
875     if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
876     $xml = $cp2utf->convert($xml);
877     use bytes; # as opposed to chars
878     print "Path-Name: $swishpath\n";
879     print "Content-Length: ".(length($xml)+1)."\n";
880     print "Document-Type: XML\n\n$xml\n";
881     }
882     }
883 dpavlin 215
884     print STDERR "\n";
885    
886 dpavlin 67 } elsif ($type_base eq "feed") {
887    
888     $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
889     my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
890    
891     print STDERR "Reading feed from program '$prog'\n";
892    
893     open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
894    
895 dpavlin 74 my $i=1; # record nr.
896 dpavlin 67
897     my $data;
898     my $line=1;
899    
900     while (<FEED>) {
901     chomp;
902    
903     if (/^$/) {
904 dpavlin 74 my $swishpath = $database."#".$i++;
905 dpavlin 67
906     if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
907     $xml = $cp2utf->convert($xml);
908     use bytes; # as opposed to chars
909     print "Path-Name: $swishpath\n";
910     print "Content-Length: ".(length($xml)+1)."\n";
911     print "Document-Type: XML\n\n$xml\n";
912     }
913     $line = 1;
914     $data = {};
915     next;
916     }
917    
918 dpavlin 74 $line = $1 if (s/^(\d+):\s*//);
919 dpavlin 67 $data->{$line++} = $_;
920    
921 dpavlin 74 fakeprogress($i);
922 dpavlin 67
923     }
924 dpavlin 177 # close lookup
925     untie %lhash if (%lhash);
926 dpavlin 1 }
927     }
928 dpavlin 3
929 dpavlin 10 # call this to commit index
930     $index->close;
931 dpavlin 3
932     1;
933     __END__
934     ##########################################################################
935    
936     =head1 NAME
937    
938 dpavlin 81 all2xml.pl - read various file formats and dump XML for SWISH-E
939 dpavlin 3
940 dpavlin 303 =head1 SYNOPSYS
941    
942     $ all2xml.pl [test.conf]
943    
944 dpavlin 3 =head1 DESCRIPTION
945    
946 dpavlin 81 This command will read ISIS data file using OpenIsis perl module, MARC
947     records using MARC module and optionally Micro$oft Excel files to
948     create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
949     this script B<isn't general xml generator> from isis files (isis allready
950     has something like that). Output of this script is tailor-made for SWISH-E.
951 dpavlin 3
952 dpavlin 303 If no configuration file is specified, it will use default one called
953     C<all2xml.conf>.
954    
955 dpavlin 81 =head1 BUGS
956    
957     Documentation is really lacking. However, in true Open Source spirit, source
958     is best documentation. I even made considerable effort to comment parts
959     which are not intuitively clear, so...
960    
961 dpavlin 3 =head1 AUTHOR
962    
963     Dobrica Pavlinusic <dpavlin@rot13.org>
964    
965     =head1 COPYRIGHT
966    
967     GNU Public License (GPL) v2 or later
968    
969     =head1 SEE ALSO
970    
971     SWISH-E web site at http://www.swish-e.org
972    
973     =cut

Properties

Name Value
cvs2svn:cvs-rev 1.64
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26