/[webpac]/trunk/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 /trunk/all2xml.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 259 - (hide annotations)
Thu Mar 11 18:23:59 2004 UTC (20 years ago) by dpavlin
File MIME type: text/plain
File size: 25885 byte(s)
ported 257:258 from hidra branch
all2xml.pl - fix for swish without filter
openisis/perl/OpenIsis.pm - removed warning

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

Properties

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

  ViewVC Help
Powered by ViewVC 1.1.26