/[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 255 - (hide annotations)
Tue Mar 9 12:17:05 2004 UTC (20 years ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 26092 byte(s)
ported r248:252 from hidra branch:

r248: much improved installation instructions, especially for Debian
      GNU/Linux distributions
r249: changed use of Spreadsheet::ParseExcel and MARC to require/import so
      that dependency on those modules can be resolved in runtime.
r250: finished installation documentation
r251: removing dependency on HTML::Parser would ease installation
r252: smaller eval{} fiexes. eval{} logic should really move to
      parse_format.pm

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

Properties

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

  ViewVC Help
Powered by ViewVC 1.1.26