/[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 196 - (hide annotations)
Mon Dec 15 00:12:16 2003 UTC (20 years, 3 months ago) by dpavlin
File MIME type: text/plain
File size: 25247 byte(s)
correct support for swish_exact when there are repeatable fields

1 dpavlin 1 #!/usr/bin/perl -w
2    
3     use strict;
4     use OpenIsis;
5     use Getopt::Std;
6     use Data::Dumper;
7     use XML::Simple;
8 dpavlin 5 use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
9 dpavlin 10 use Text::Iconv;
10 dpavlin 13 use Config::IniFiles;
11 dpavlin 40 use Encode;
12 dpavlin 177 #use GDBM_File;
13     use Fcntl; # for O_RDWR
14     use TDB_File;
15 dpavlin 1
16 dpavlin 10 $|=1;
17 dpavlin 9
18 dpavlin 13 my $config_file = $0;
19     $config_file =~ s/\.pl$/.conf/;
20     die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
21    
22 dpavlin 10 my $config;
23    
24 dpavlin 58 #use index_DBI; # default DBI module for index
25     use index_DBI_cache; # faster DBI module using memory cache
26 dpavlin 50 my $index;
27 dpavlin 10
28 dpavlin 1 my %opts;
29    
30 dpavlin 7 # usage:
31     # -d directory name
32     # -m multiple directories
33     # -q quiet
34     # -s run swish
35 dpavlin 1
36 dpavlin 7 getopts('d:m:qs', \%opts);
37    
38 dpavlin 43 my $path; # this is name of database
39 dpavlin 1
40 dpavlin 57 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
41 dpavlin 1
42 dpavlin 40 # this is encoding of all files on disk, including import_xml/*.xml file and
43     # filter/*.pm files! It will be used to store strings in perl internally!
44     my $codepage = 'ISO-8859-2';
45 dpavlin 1
46 dpavlin 40 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
47     # this function will convert data from XML files to local encoding
48     sub x {
49     return $utf2cp->convert($_[0]);
50     }
51 dpavlin 3
52 dpavlin 54 # decode isis/excel or other import codepage
53     my $import2cp;
54 dpavlin 10
55 dpavlin 40 # outgoing xml must be in UTF-8
56     my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
57 dpavlin 29
58 dpavlin 54 # mapping between data type and tag which specify
59     # format in XML file
60     my %type2tag = (
61     'isis' => 'isis',
62 dpavlin 62 'excel' => 'column',
63     'marc' => 'marc',
64 dpavlin 67 'feed' => 'feed'
65 dpavlin 54 );
66 dpavlin 3
67 dpavlin 170 my $cache; # for cacheing
68    
69 dpavlin 177 # lookup hash (tied to file)
70     my %lhash;
71     # this option will cache all lookup entries in memory.
72     # if you are tight on memory, turn this off
73     my $use_lhash_cache = 1;
74    
75 dpavlin 195 my $last_field_name; # cache to prevent repeated fields
76    
77 dpavlin 54 sub data2xml {
78    
79 dpavlin 10 use xmlify;
80    
81 dpavlin 54 my $type = shift @_;
82 dpavlin 3 my $row = shift @_;
83 dpavlin 13 my $add_xml = shift @_;
84 dpavlin 59 # needed to read values from configuration file
85     my $cfg = shift @_;
86     my $database = shift @_;
87 dpavlin 3
88     my $xml;
89    
90 dpavlin 10 use parse_format;
91 dpavlin 3
92 dpavlin 13 my $html = ""; # html formatted display output
93 dpavlin 10
94 dpavlin 13 my %field_usage; # counter for usage of each field
95    
96 dpavlin 32 # sort subrouting using order="" attribute
97     sub by_order {
98 dpavlin 98 my $va = $config->{indexer}->{$a}->{order} ||
99     $config->{indexer}->{$a};
100     my $vb = $config->{indexer}->{$b}->{order} ||
101     $config->{indexer}->{$b};
102 dpavlin 29
103 dpavlin 98 return $va <=> $vb;
104 dpavlin 32 }
105 dpavlin 3
106 dpavlin 170 my @sorted_tags;
107 dpavlin 180 if ($cache->{tags_by_order}) {
108     @sorted_tags = @{$cache->{tags_by_order}};
109 dpavlin 170 } else {
110     @sorted_tags = sort by_order keys %{$config->{indexer}};
111 dpavlin 180 $cache->{tags_by_order} = \@sorted_tags;
112 dpavlin 170 }
113 dpavlin 32
114 dpavlin 177 # lookup key
115     my $lookup_key;
116    
117 dpavlin 178 # cache for field in pages
118 dpavlin 180 delete $cache->{display_data};
119     delete $cache->{swish_data};
120     delete $cache->{swish_exact_data};
121 dpavlin 182 delete $cache->{index_data};
122 dpavlin 188 delete $cache->{index_delimiter};
123 dpavlin 178 my @page_fields; # names of fields
124    
125    
126     # subs used to produce output
127    
128     sub get_field_name($$$) {
129     my ($config,$field,$field_usage) = @_;
130    
131     # find field name (signular, plural)
132     my $field_name = "";
133     if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) {
134     $field_name = $config->{indexer}->{$field}->{name_singular};
135     } elsif ($config->{indexer}->{$field}->{name_plural}) {
136     $field_name = $config->{indexer}->{$field}->{name_plural};
137     } elsif ($config->{indexer}->{$field}->{name}) {
138     $field_name = $config->{indexer}->{$field}->{name};
139     } else {
140     print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
141     }
142     if ($field_name) {
143 dpavlin 195 if (! $last_field_name) {
144     $last_field_name = x($field_name);
145     return $last_field_name;
146     } elsif ($field_name ne $last_field_name) {
147     $last_field_name = x($field_name);
148     return $last_field_name;
149     }
150 dpavlin 178 }
151     }
152    
153    
154     # begin real work: go field by field
155 dpavlin 170 foreach my $field (@sorted_tags) {
156    
157 dpavlin 40 $field=x($field);
158 dpavlin 13 $field_usage{$field}++;
159    
160 dpavlin 10 my $swish_data = "";
161 dpavlin 163 my $swish_exact_data = "";
162 dpavlin 3 my $display_data = "";
163 dpavlin 182 my @index_data;
164 dpavlin 35 my $line_delimiter;
165 dpavlin 3
166 dpavlin 34 my ($swish,$display);
167    
168 dpavlin 54 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
169 dpavlin 178
170     # is this field page-by-page?
171     my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
172     push @page_fields,$field if ($iterate_by_page);
173     my %page_max = ();
174     # default line_delimiter if using
175     my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
176 dpavlin 188 $cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter};
177 dpavlin 178
178 dpavlin 195 my $format_name = $config->{indexer}->{$field}->{format_name};
179     my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter};
180     if ($format_name && $format_delimiter) {
181     $cache->{format}->{$field}->{format_name} = $format_name;
182     $cache->{format}->{$field}->{format_delimiter} = $format_delimiter;
183     }
184    
185 dpavlin 54 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
186 dpavlin 3
187 dpavlin 40 my $format = x($x->{content});
188     my $delimiter = x($x->{delimiter}) || ' ';
189 dpavlin 3
190 dpavlin 178 my $repeat_off = 0; # init repeatable offset
191 dpavlin 29
192 dpavlin 177 # swish, swish_exact, display, index, index_lookup
193     # swish and display defaults
194     my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
195 dpavlin 29 $s = 0 if (lc($x->{type}) eq "display");
196     $d = 0 if (lc($x->{type}) eq "swish");
197 dpavlin 188 ($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index");
198 dpavlin 181 ($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact");
199 dpavlin 182 ($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/);
200 dpavlin 29
201 dpavlin 35 # what will separate last line from this one?
202 dpavlin 182 if ($display_data && $x->{append}) {
203 dpavlin 34 $line_delimiter = ' ';
204 dpavlin 35 } elsif ($display_data) {
205 dpavlin 34 $line_delimiter = '<br/>';
206     }
207 dpavlin 29
208 dpavlin 34 # init vars so that we go into while...
209     ($swish,$display) = (1,1);
210 dpavlin 29
211 dpavlin 97 # placeholder for all repeatable entries for index
212    
213 dpavlin 195 sub mkformat($$) {
214 dpavlin 153 my $x = shift || die "mkformat needs tag reference";
215     my $data = shift || return;
216     my $format_name = x($x->{format_name}) || return $data;
217     my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
218     my $format_delimiter = x($x->{format_delimiter});
219     my @data;
220     if ($format_delimiter) {
221     @data = split(/$format_delimiter/,$data);
222     } else {
223     push @data,$data;
224     }
225    
226     if ($fmt) {
227     my $nr = scalar $fmt =~ s/%s/%s/g;
228     if (($#data+1) == $nr) {
229     return sprintf($fmt,@data);
230     } else {
231     print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
232     return $data;
233     }
234     } else {
235     print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
236     }
237     }
238    
239 dpavlin 90 # while because of repeatable fields
240     while ($swish || $display) {
241 dpavlin 178 my $page = $repeat_off;
242     $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
243 dpavlin 54 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
244 dpavlin 90 if ($repeat_off > 1000) {
245     print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
246     last;
247     }
248 dpavlin 177
249     # is this field is lookup?
250     if ($display && $x->{lookup}) {
251 dpavlin 178 my $null = "<!-- null -->";
252 dpavlin 177 if ($use_lhash_cache) {
253     if (!defined($cache->{lhash}->{$display})) {
254     my $new_display = $lhash{$display};
255 dpavlin 178 if (defined($new_display)) {
256 dpavlin 177 #print STDERR "lookup cache store '$display' = '$new_display'\n";
257     $display = $new_display;
258     $cache->{lhash}->{$display} = $new_display;
259     } else {
260     print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
261     $display = "";
262 dpavlin 178 $cache->{lhash}->{$display} = $null;
263 dpavlin 177 }
264     } else {
265     $display = $cache->{lhash}->{$display};
266     }
267     } else {
268 dpavlin 178 $display = $lhash{$display} || $null;
269 dpavlin 177 }
270     }
271    
272 dpavlin 29 # filter="name" ; filter this field through
273     # filter/[name].pm
274     my $filter = $x->{filter};
275 dpavlin 170 if ($filter && !$cache->{filter_loaded}->{$filter}) {
276 dpavlin 29 require "filter/".$filter.".pm";
277 dpavlin 170 $cache->{filter_loaded}->{$filter}++;
278 dpavlin 20 }
279 dpavlin 29 # type="swish" ; field for swish
280 dpavlin 163 if ($swish) {
281     if ($filter && ($s || $se)) {
282 dpavlin 29 no strict 'refs';
283 dpavlin 163 my $tmp = join(" ",&$filter($swish)) if ($s || $se);
284     $swish_data .= $tmp if ($s);
285 dpavlin 196 if ($se) {
286     if ($swish_exact_data) {
287     $swish_exact_data .= "xxexx xxbxx ".$tmp;
288     } else {
289     $swish_exact_data .= $tmp;
290     }
291     }
292 dpavlin 163
293 dpavlin 29 } else {
294 dpavlin 163 $swish_data .= $swish if ($s);
295 dpavlin 196 if ($se) {
296     if ($swish_exact_data) {
297     $swish_exact_data .= "xxexx xxbxx ".$swish;
298     } else {
299     $swish_exact_data .= $swish;
300     }
301     }
302 dpavlin 29 }
303     }
304 dpavlin 17
305 dpavlin 29 # type="display" ; field for display
306     if ($d && $display) {
307 dpavlin 35 if ($line_delimiter && $display_data) {
308     $display_data .= $line_delimiter;
309     }
310 dpavlin 29 if ($filter) {
311     no strict 'refs';
312 dpavlin 138 if ($display_data) {
313 dpavlin 153 $display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display)));
314 dpavlin 138 } else {
315 dpavlin 153 $display_data = join($delimiter,mkformat($x,&$filter($display)));
316 dpavlin 138 }
317 dpavlin 29 } else {
318 dpavlin 138 if ($display_data) {
319 dpavlin 153 $display_data .= $delimiter.mkformat($x,$display);
320 dpavlin 138 } else {
321 dpavlin 153 $display_data = mkformat($x,$display);
322 dpavlin 138 }
323 dpavlin 29 }
324 dpavlin 20 }
325 dpavlin 29
326     # type="index" ; insert into index
327 dpavlin 182 my $idisplay;
328 dpavlin 29 if ($i && $display) {
329 dpavlin 182 $idisplay = $display;
330 dpavlin 177 if ($filter) {
331     no strict 'refs';
332 dpavlin 182 $idisplay = &$filter($idisplay);
333 dpavlin 177 }
334 dpavlin 182 push @index_data, $idisplay if (! $iterate_by_page);
335 dpavlin 97 }
336    
337 dpavlin 177 # store fields in lookup
338     if ($il && $display) {
339     if (lc($x->{type}) eq "lookup_key") {
340     if ($lookup_key) {
341     print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
342     } else {
343     $lookup_key = $display;
344     }
345     } elsif (lc($x->{type}) eq "lookup_val") {
346     if ($lookup_key) {
347     $lhash{$lookup_key} = $display;
348     } else {
349     print STDERR "WARNING: no lookup_key defined for '$display'?";
350     }
351 dpavlin 20 }
352 dpavlin 178
353 dpavlin 20 }
354 dpavlin 178
355     # store data for page-by-page repeatable fields
356     if ($iterate_by_page) {
357     sub iterate_fld($$$$$$) {
358     my ($cache,$what,$field,$page,$data,$append) = @_;
359     return if (!$data);
360 dpavlin 182
361     my $ldel = $page_line_delimiter;
362     $ldel = " " if ($append);
363     #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel);
364 dpavlin 178 if (! $cache->{$what}->{$field}->[$page]) {
365     $cache->{$what}->{$field}->[$page] = $data;
366     } else {
367 dpavlin 182 $cache->{$what}->{$field}->[$page] .= $ldel.$data;
368 dpavlin 178 }
369     }
370    
371     if ($display_data) {
372     iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append});
373     }
374     $display_data = "";
375     if ($swish_data) {
376     iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append});
377     $swish_data = "";
378     }
379     if ($swish_exact_data) {
380     iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append});
381     $swish_exact_data = "";
382     }
383 dpavlin 182
384     if ($idisplay) {
385     my $ldel=$page_line_delimiter;
386     my @index_data;
387     if ($cache->{index_data}->{$field}->[$page]) {
388    
389     @index_data = @{$cache->{index_data}->{$field}->[$page]};
390     }
391     if ($x->{append}) {
392 dpavlin 195 if (@index_data) {
393     $index_data[$#index_data] .= $idisplay;
394     } else {
395     push @index_data, $idisplay;
396     }
397 dpavlin 182 } else {
398     push @index_data, $idisplay;
399     }
400     $idisplay = "";
401     @{$cache->{index_data}->{$field}->[$page]} = @index_data;
402     }
403 dpavlin 178 }
404 dpavlin 17 }
405 dpavlin 177
406 dpavlin 182 if (! $iterate_by_page) {
407 dpavlin 188 my $idel = $x->{index_delimiter};
408 dpavlin 182 # fill data in index
409 dpavlin 188 foreach my $tmp (@index_data) {
410     my $i = $d = $tmp;
411     if ($idel && $tmp =~ m/$idel/) {
412     ($i,$d) = split(/$idel/,$tmp);
413     }
414     $index->insert($field, $i, $d, $path);
415 dpavlin 182 }
416     @index_data = ();
417 dpavlin 177 }
418 dpavlin 3 }
419 dpavlin 9
420 dpavlin 59 # now try to parse variables from configuration file
421     foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
422 dpavlin 13
423 dpavlin 62 my $delimiter = x($x->{delimiter}) || ' ';
424 dpavlin 59 my $val = $cfg->val($database, x($x->{content}));
425    
426     my ($s,$d,$i) = (1,1,0); # swish, display default
427     $s = 0 if (lc($x->{type}) eq "display");
428     $d = 0 if (lc($x->{type}) eq "swish");
429 dpavlin 163 # no support for swish exact in config.
430     # IMHO, it's useless
431 dpavlin 59 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
432    
433     if ($val) {
434 dpavlin 62 $display_data .= $delimiter.$val if ($d);
435 dpavlin 59 $swish_data .= $val if ($s);
436     $index->insert($field, $val, $path) if ($i);
437     }
438    
439 dpavlin 178 if ($iterate_by_page) {
440     # FIXME data from config tag will appear just
441     # on first page!!!
442     my $page = 0;
443     if ($display_data) {
444     $cache->{display_data}->{$field}->[$page] = $display_data;
445     $display_data = "";
446     }
447     if ($swish_data) {
448     $cache->{swish_data}->{$field}->[$page] = $swish_data;
449     $swish_data = "";
450     }
451     if ($swish_exact_data) {
452     $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
453     $swish_exact_data = "";
454     }
455     }
456 dpavlin 59 }
457    
458 dpavlin 178 # save data page-by-page
459     foreach my $field (@page_fields) {
460     my $nr_pages = $page_max{$field} || next;
461     #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
462     #print STDERR Dumper($cache->{display_data});
463     for (my $page=0; $page <= $nr_pages; $page++) {
464 dpavlin 195 my $display_data;
465     if ($cache->{format}->{$field}) {
466     $display_data=mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
467     } else {
468     $display_data = $cache->{display_data}->{$field}->[$page];
469     }
470 dpavlin 178 if ($display_data) { # default
471     if ($field eq "headline") {
472     $xml .= xmlify("headline", $display_data);
473     } else {
474 dpavlin 29
475 dpavlin 178 # fallback to empty field name if needed
476     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
477     $html .= "#-#".$display_data."###\n";
478     }
479     }
480    
481     my $swish_data = $cache->{swish_data}->{$field}->[$page];
482     if ($swish_data) {
483     # remove extra spaces
484     $swish_data =~ s/ +/ /g;
485     $swish_data =~ s/ +$//g;
486 dpavlin 13
487 dpavlin 178 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
488 dpavlin 13 }
489 dpavlin 178
490     my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
491     if ($swish_exact_data) {
492     $swish_exact_data =~ s/ +/ /g;
493     $swish_exact_data =~ s/ +$//g;
494    
495     # add delimiters before and after word.
496     # That is required to produce exact match
497     $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
498 dpavlin 13 }
499 dpavlin 182
500 dpavlin 188 my $idel = $cache->{index_delimiter}->{$field};
501     foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) {
502     my $i = $tmp;
503     my $d = $tmp;
504     if ($idel && $tmp =~ m/$idel/) {
505     ($i,$d) = split(/$idel/,$tmp);
506     }
507     $index->insert($field, $i, $d, $path);
508     #print STDERR "index [$idel] $field: $i --> $d [$path]\n";
509 dpavlin 182 }
510 dpavlin 13 }
511 dpavlin 178
512 dpavlin 10 }
513 dpavlin 178
514     if (! $iterate_by_page) {
515     if ($display_data) {
516     if ($field eq "headline") {
517     $xml .= xmlify("headline", $display_data);
518     } else {
519 dpavlin 20
520 dpavlin 178 # fallback to empty field name if needed
521     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
522     $html .= "#-#".$display_data."###\n";
523     }
524     }
525     if ($swish_data) {
526     # remove extra spaces
527     $swish_data =~ s/ +/ /g;
528     $swish_data =~ s/ +$//g;
529 dpavlin 10
530 dpavlin 178 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
531     }
532 dpavlin 9
533 dpavlin 178 if ($swish_exact_data) {
534     $swish_exact_data =~ s/ +/ /g;
535     $swish_exact_data =~ s/ +$//g;
536    
537     # add delimiters before and after word.
538     # That is required to produce exact match
539     $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
540     }
541 dpavlin 163 }
542 dpavlin 3 }
543 dpavlin 13
544     # dump formatted output in <html>
545     if ($html) {
546 dpavlin 81 #$xml .= xmlify("html",$html);
547     $xml .= "<html><![CDATA[ $html ]]></html>";
548 dpavlin 13 }
549    
550 dpavlin 3 if ($xml) {
551 dpavlin 13 $xml .= $add_xml if ($add_xml);
552 dpavlin 10 return "<xml>\n$xml</xml>\n";
553 dpavlin 3 } else {
554     return;
555     }
556     }
557    
558     ##########################################################################
559    
560 dpavlin 54 # read configuration for this script
561 dpavlin 13 my $cfg = new Config::IniFiles( -file => $config_file );
562 dpavlin 1
563 dpavlin 54 # read global.conf configuration
564     my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
565    
566 dpavlin 50 # open index
567     $index = new index_DBI(
568 dpavlin 54 $cfg_global->val('global', 'dbi_dbd'),
569     $cfg_global->val('global', 'dbi_dsn'),
570     $cfg_global->val('global', 'dbi_user'),
571     $cfg_global->val('global', 'dbi_passwd') || '',
572 dpavlin 50 );
573    
574 dpavlin 97 my $show_progress = $cfg_global->val('global', 'show_progress');
575    
576 dpavlin 164 my $unac_filter = $cfg_global->val('global', 'unac_filter');
577     if ($unac_filter) {
578     require $unac_filter;
579     }
580    
581 dpavlin 13 foreach my $database ($cfg->Sections) {
582 dpavlin 1
583 dpavlin 54 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
584 dpavlin 40 my $add_xml = $cfg -> val($database, 'xml'); # optional
585 dpavlin 1
586 dpavlin 177 # create new lookup file
587     my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
588     if ($lookup_file) {
589     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
590     tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
591     print STDERR "creating lookup file '$lookup_file'\n";
592     }
593    
594     # open existing lookup file
595     $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
596     if ($lookup_file) {
597     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
598     tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
599     print STDERR "opening lookup file '$lookup_file'\n";
600     }
601    
602 dpavlin 54 print STDERR "reading ./import_xml/$type.xml\n";
603 dpavlin 1
604 dpavlin 58 # extract just type basic
605     my $type_base = $type;
606     $type_base =~ s/_.+$//g;
607 dpavlin 40
608 dpavlin 153 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
609 dpavlin 58
610 dpavlin 54 # output current progress indicator
611     my $last_p = 0;
612     sub progress {
613 dpavlin 101 return if (! $show_progress);
614 dpavlin 54 my $current = shift;
615 dpavlin 57 my $total = shift || 1;
616 dpavlin 54 my $p = int($current * 100 / $total);
617     if ($p != $last_p) {
618     printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
619     $last_p = $p;
620     }
621 dpavlin 3 }
622    
623 dpavlin 67 my $fake_dir = 1;
624     sub fakeprogress {
625 dpavlin 104 return if (! $show_progress);
626 dpavlin 67 my $current = shift @_;
627    
628     my @ind = ('-','\\','|','/','-','\\','|','/', '-');
629    
630     $last_p += $fake_dir;
631     $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
632     if ($last_p % 10 == 0) {
633     printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
634     }
635     }
636    
637 dpavlin 54 # now read database
638     print STDERR "using: $type...\n";
639 dpavlin 1
640 dpavlin 180 # erase cache for tags by order in this database
641     delete $cache->{tags_by_order};
642    
643 dpavlin 58 if ($type_base eq "isis") {
644    
645 dpavlin 54 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
646 dpavlin 1
647 dpavlin 54 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
648     my $db = OpenIsis::open( $isis_db );
649 dpavlin 10
650 dpavlin 106 # check if .txt database for OpenIsis is zero length,
651     # if so, erase it and re-open database
652     sub check_txt_db {
653     my $isis_db = shift || die "need isis database name";
654 dpavlin 109 my $reopen = 0;
655    
656 dpavlin 106 if (-e $isis_db.".TXT") {
657 dpavlin 109 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
658 dpavlin 108 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
659 dpavlin 109 $reopen++;
660 dpavlin 106 }
661 dpavlin 109 if (-e $isis_db.".PTR") {
662     print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
663     unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
664     $reopen++;
665     }
666     return OpenIsis::open( $isis_db ) if ($reopen);
667 dpavlin 106 }
668    
669     # EOF error
670     if ($db == -1) {
671     $db = check_txt_db($isis_db);
672 dpavlin 108 if ($db == -1) {
673 dpavlin 106 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
674     next;
675     }
676     }
677    
678 dpavlin 101 # OpenIsis::ERR_BADF
679     if ($db == -4) {
680     print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
681     next;
682     # OpenIsis::ERR_IO
683     } elsif ($db == -5) {
684     print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
685     next;
686     } elsif ($db < 0) {
687     print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
688     next;
689     }
690    
691 dpavlin 54 my $max_rowid = OpenIsis::maxRowid( $db );
692 dpavlin 3
693 dpavlin 106 # if 0 records, try to rease isis .txt database
694     if ($max_rowid == 0) {
695     # force removal of database
696     $db = check_txt_db($isis_db);
697     $max_rowid = OpenIsis::maxRowid( $db );
698     }
699    
700 dpavlin 54 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
701    
702     my $path = $database;
703    
704     for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
705     my $row = OpenIsis::read( $db, $row_id );
706     if ($row && $row->{mfn}) {
707    
708     progress($row->{mfn}, $max_rowid);
709    
710     my $swishpath = $path."#".int($row->{mfn});
711    
712 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
713 dpavlin 54 $xml = $cp2utf->convert($xml);
714     use bytes; # as opposed to chars
715     print "Path-Name: $swishpath\n";
716     print "Content-Length: ".(length($xml)+1)."\n";
717     print "Document-Type: XML\n\n$xml\n";
718     }
719 dpavlin 3 }
720 dpavlin 54 }
721 dpavlin 108 # for this to work with current version of OpenIsis (0.9.0)
722     # you might need my patch from
723     # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
724     OpenIsis::close($db);
725 dpavlin 54 print STDERR "\n";
726 dpavlin 3
727 dpavlin 58 } elsif ($type_base eq "excel") {
728 dpavlin 54 use Spreadsheet::ParseExcel;
729     use Spreadsheet::ParseExcel::Utility qw(int2col);
730    
731     $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
732     my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
733 dpavlin 43
734 dpavlin 54 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
735 dpavlin 56 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
736 dpavlin 54
737     my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
738    
739     my $sheet_nr = 0;
740     foreach my $oWks (@{$oBook->{Worksheet}}) {
741     #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
742     last if ($oWks->{Name} eq $sheet);
743     $sheet_nr++;
744     }
745    
746     my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
747    
748     print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
749     defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
750     my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
751    
752 dpavlin 56 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
753 dpavlin 54 my $row;
754     for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
755     my $cell = $oWorksheet->{Cells}[$iR][$iC];
756     if ($cell) {
757     $row->{int2col($iC)} = $cell->Value;
758     }
759     }
760    
761     progress($iR, $end_row);
762    
763     # print "row[$iR/$end_row] ";
764     # foreach (keys %{$row}) {
765     # print "$_: ",$row->{$_},"\t";
766     # }
767     # print "\n";
768    
769     my $swishpath = $database."#".$iR;
770    
771     next if (! $row);
772    
773 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
774 dpavlin 44 $xml = $cp2utf->convert($xml);
775 dpavlin 35 use bytes; # as opposed to chars
776 dpavlin 43 print "Path-Name: $swishpath\n";
777 dpavlin 10 print "Content-Length: ".(length($xml)+1)."\n";
778     print "Document-Type: XML\n\n$xml\n";
779 dpavlin 3 }
780 dpavlin 1 }
781 dpavlin 62 } elsif ($type_base eq "marc") {
782 dpavlin 67
783 dpavlin 62 use MARC;
784    
785     $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
786     my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
787    
788     # optional argument is format
789     my $format = x($config->{format}) || 'usmarc';
790    
791     print STDERR "Reading MARC file '$marc_file'\n";
792    
793     my $marc = new MARC;
794     my $nr = $marc->openmarc({
795     file=>$marc_file, format=>$format
796     }) || die "Can't open MARC file '$marc_file'";
797    
798     my $i=0; # record nr.
799    
800     my $rec;
801    
802     while ($marc->nextmarc(1)) {
803    
804     # XXX
805 dpavlin 67 fakeprogress($i++);
806 dpavlin 62
807     my $swishpath = $database."#".$i;
808    
809     if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
810     $xml = $cp2utf->convert($xml);
811     use bytes; # as opposed to chars
812     print "Path-Name: $swishpath\n";
813     print "Content-Length: ".(length($xml)+1)."\n";
814     print "Document-Type: XML\n\n$xml\n";
815     }
816     }
817 dpavlin 67 } elsif ($type_base eq "feed") {
818    
819     $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
820     my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
821    
822     print STDERR "Reading feed from program '$prog'\n";
823    
824     open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
825    
826 dpavlin 74 my $i=1; # record nr.
827 dpavlin 67
828     my $data;
829     my $line=1;
830    
831     while (<FEED>) {
832     chomp;
833    
834     if (/^$/) {
835 dpavlin 74 my $swishpath = $database."#".$i++;
836 dpavlin 67
837     if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
838     $xml = $cp2utf->convert($xml);
839     use bytes; # as opposed to chars
840     print "Path-Name: $swishpath\n";
841     print "Content-Length: ".(length($xml)+1)."\n";
842     print "Document-Type: XML\n\n$xml\n";
843     }
844     $line = 1;
845     $data = {};
846     next;
847     }
848    
849 dpavlin 74 $line = $1 if (s/^(\d+):\s*//);
850 dpavlin 67 $data->{$line++} = $_;
851    
852 dpavlin 74 fakeprogress($i);
853 dpavlin 67
854     }
855 dpavlin 177 # close lookup
856     untie %lhash if (%lhash);
857 dpavlin 1 }
858     }
859 dpavlin 3
860 dpavlin 10 # call this to commit index
861     $index->close;
862 dpavlin 3
863     1;
864     __END__
865     ##########################################################################
866    
867     =head1 NAME
868    
869 dpavlin 81 all2xml.pl - read various file formats and dump XML for SWISH-E
870 dpavlin 3
871     =head1 DESCRIPTION
872    
873 dpavlin 81 This command will read ISIS data file using OpenIsis perl module, MARC
874     records using MARC module and optionally Micro$oft Excel files to
875     create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
876     this script B<isn't general xml generator> from isis files (isis allready
877     has something like that). Output of this script is tailor-made for SWISH-E.
878 dpavlin 3
879 dpavlin 81 =head1 BUGS
880    
881     Documentation is really lacking. However, in true Open Source spirit, source
882     is best documentation. I even made considerable effort to comment parts
883     which are not intuitively clear, so...
884    
885 dpavlin 3 =head1 AUTHOR
886    
887     Dobrica Pavlinusic <dpavlin@rot13.org>
888    
889     =head1 COPYRIGHT
890    
891     GNU Public License (GPL) v2 or later
892    
893     =head1 SEE ALSO
894    
895     SWISH-E web site at http://www.swish-e.org
896    
897     =cut

Properties

Name Value
cvs2svn:cvs-rev 1.55
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26