/[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 207 - (hide annotations)
Sat Jan 31 21:03:06 2004 UTC (15 years, 10 months ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 25315 byte(s)
thesaurus is finally working... It contains recursive entries to parnet
term, and we actually needed to display narrower terms, so mem_lookup was
created. Important changes:
- you can write eval{"901a" eq "Mikrotezaurus"} within <isis>
  tag and if expression evaluates to false, no content will be outputed
  (It's used to hide microtesarus terms from lover level descriptors)
- mem_lookup.pm now supports formats: you can write something like
  [a:5614];;[d:[a:5614]] and it will correctly embed values

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

Properties

Name Value
cvs2svn:cvs-rev 1.58
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26