/[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 599 - (hide annotations)
Wed Dec 8 18:23:25 2004 UTC (19 years, 3 months ago) by dpavlin
File MIME type: text/plain
File size: 26943 byte(s)
fix repeatable field names which have non 7-bit ascii characters

1 dpavlin 1 #!/usr/bin/perl -w
2    
3     use strict;
4     use OpenIsis;
5     use Getopt::Std;
6     use Data::Dumper;
7     use XML::Simple;
8 dpavlin 10 use Text::Iconv;
9 dpavlin 13 use Config::IniFiles;
10 dpavlin 40 use Encode;
11 dpavlin 177 #use GDBM_File;
12     use Fcntl; # for O_RDWR
13     use TDB_File;
14 dpavlin 1
15 dpavlin 10 $|=1;
16 dpavlin 9
17 dpavlin 13 my $config_file = $0;
18     $config_file =~ s/\.pl$/.conf/;
19 dpavlin 320 $config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]);
20 dpavlin 13 die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
21    
22 dpavlin 10 my $config;
23    
24 dpavlin 58 #use index_DBI; # default DBI module for index
25     use index_DBI_cache; # faster DBI module using memory cache
26 dpavlin 50 my $index;
27 dpavlin 10
28 dpavlin 1 my %opts;
29    
30 dpavlin 7 # usage:
31     # -d directory name
32     # -m multiple directories
33     # -q quiet
34     # -s run swish
35 dpavlin 1
36 dpavlin 7 getopts('d:m:qs', \%opts);
37    
38 dpavlin 43 my $path; # this is name of database
39 dpavlin 1
40 dpavlin 57 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
41 dpavlin 1
42 dpavlin 40 # this is encoding of all files on disk, including import_xml/*.xml file and
43     # filter/*.pm files! It will be used to store strings in perl internally!
44     my $codepage = 'ISO-8859-2';
45 dpavlin 1
46 dpavlin 40 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
47     # this function will convert data from XML files to local encoding
48     sub x {
49     return $utf2cp->convert($_[0]);
50     }
51 dpavlin 3
52 dpavlin 54 # decode isis/excel or other import codepage
53     my $import2cp;
54 dpavlin 10
55 dpavlin 40 # outgoing xml must be in UTF-8
56     my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
57 dpavlin 29
58 dpavlin 54 # mapping between data type and tag which specify
59     # format in XML file
60     my %type2tag = (
61     'isis' => 'isis',
62 dpavlin 62 'excel' => 'column',
63     'marc' => 'marc',
64 dpavlin 67 'feed' => 'feed'
65 dpavlin 54 );
66 dpavlin 3
67 dpavlin 170 my $cache; # for cacheing
68    
69 dpavlin 177 # lookup hash (tied to file)
70     my %lhash;
71     # this option will cache all lookup entries in memory.
72     # if you are tight on memory, turn this off
73     my $use_lhash_cache = 1;
74    
75 dpavlin 195 my $last_field_name; # cache to prevent repeated fields
76    
77 dpavlin 54 sub data2xml {
78    
79 dpavlin 10 use xmlify;
80    
81 dpavlin 54 my $type = shift @_;
82 dpavlin 3 my $row = shift @_;
83 dpavlin 13 my $add_xml = shift @_;
84 dpavlin 59 # needed to read values from configuration file
85     my $cfg = shift @_;
86     my $database = shift @_;
87 dpavlin 3
88     my $xml;
89    
90 dpavlin 10 use parse_format;
91 dpavlin 3
92 dpavlin 13 my $html = ""; # html formatted display output
93 dpavlin 10
94 dpavlin 13 my %field_usage; # counter for usage of each field
95    
96 dpavlin 32 # sort subrouting using order="" attribute
97     sub by_order {
98 dpavlin 98 my $va = $config->{indexer}->{$a}->{order} ||
99     $config->{indexer}->{$a};
100     my $vb = $config->{indexer}->{$b}->{order} ||
101     $config->{indexer}->{$b};
102 dpavlin 29
103 dpavlin 98 return $va <=> $vb;
104 dpavlin 32 }
105 dpavlin 3
106 dpavlin 170 my @sorted_tags;
107 dpavlin 180 if ($cache->{tags_by_order}) {
108     @sorted_tags = @{$cache->{tags_by_order}};
109 dpavlin 170 } else {
110     @sorted_tags = sort by_order keys %{$config->{indexer}};
111 dpavlin 180 $cache->{tags_by_order} = \@sorted_tags;
112 dpavlin 170 }
113 dpavlin 32
114 dpavlin 177 # lookup key
115     my $lookup_key;
116    
117 dpavlin 178 # cache for field in pages
118 dpavlin 180 delete $cache->{display_data};
119     delete $cache->{swish_data};
120     delete $cache->{swish_exact_data};
121 dpavlin 182 delete $cache->{index_data};
122 dpavlin 188 delete $cache->{index_delimiter};
123 dpavlin 178 my @page_fields; # names of fields
124    
125    
126     # subs used to produce output
127    
128     sub get_field_name($$$) {
129     my ($config,$field,$field_usage) = @_;
130    
131     # find field name (signular, plural)
132     my $field_name = "";
133     if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) {
134     $field_name = $config->{indexer}->{$field}->{name_singular};
135     } elsif ($config->{indexer}->{$field}->{name_plural}) {
136     $field_name = $config->{indexer}->{$field}->{name_plural};
137     } elsif ($config->{indexer}->{$field}->{name}) {
138     $field_name = $config->{indexer}->{$field}->{name};
139     } else {
140     print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
141     }
142 dpavlin 599
143 dpavlin 178 if ($field_name) {
144 dpavlin 599 $field_name = x($field_name);
145 dpavlin 195 if (! $last_field_name) {
146 dpavlin 599 $last_field_name = $field_name;
147 dpavlin 195 return $last_field_name;
148     } elsif ($field_name ne $last_field_name) {
149 dpavlin 599 $last_field_name = $field_name;
150 dpavlin 195 return $last_field_name;
151     }
152 dpavlin 178 }
153     }
154    
155 dpavlin 279
156     # init variables for different types
157 dpavlin 274 sub init_visible_type($) {
158     my $type = shift;
159 dpavlin 178
160 dpavlin 274 # swish, swish_exact, display, index, index_lookup
161     # swish and display defaults
162     my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
163     if (lc($type) eq "display") {
164     $s = 0;
165     } elsif (lc($type) eq "swish") {
166     $d = 0;
167     } elsif (lc($type) eq "index") {
168     ($s,$se,$d,$i) = (0,1,0,1);
169     } elsif (lc($type) eq "swish_exact") {
170     ($s,$se,$d,$i) = (0,1,0,0);
171     } elsif (lc($type) =~ /^lookup/) {
172     ($s,$se,$d,$i,$il) = (0,1,0,0,1);
173 dpavlin 333 } elsif ($type) {
174     print STDERR "WARNING: unknown type: $type\n";
175 dpavlin 274 }
176     return ($s,$se,$d,$i,$il);
177     }
178    
179 dpavlin 279
180     # convert
181     #
182     # <tag>
183     # <delimiter>, </delimiter>
184     # <value>200a</value>
185     # </tag>
186     #
187     # to
188     #
189     # <tag delimiter=", ">200a</tag>
190     #
191     # but without loosing spaces in delimiter (becasue
192     # new XML::Simple strips spaces in attribute values
193     # as defined in XML specification)
194     #
195     sub unroll_x($) {
196     my $x = shift;
197    
198     if (defined $x->{value}) {
199     my ($v,$d) = ($x->{value}->{content}, $x->{delimiter}->{content});
200     delete $x->{value};
201     delete $x->{delimiter};
202     $x->{content} = $v;
203     $x->{delimiter} = $d;
204     }
205     return $x;
206     }
207    
208 dpavlin 178 # begin real work: go field by field
209 dpavlin 170 foreach my $field (@sorted_tags) {
210    
211 dpavlin 40 $field=x($field);
212 dpavlin 13 $field_usage{$field}++;
213    
214 dpavlin 10 my $swish_data = "";
215 dpavlin 163 my $swish_exact_data = "";
216 dpavlin 3 my $display_data = "";
217 dpavlin 182 my @index_data;
218 dpavlin 35 my $line_delimiter;
219 dpavlin 3
220 dpavlin 34 my ($swish,$display);
221    
222 dpavlin 54 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
223 dpavlin 178
224     # is this field page-by-page?
225     my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
226     push @page_fields,$field if ($iterate_by_page);
227     my %page_max = ();
228     # default line_delimiter if using
229     my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
230 dpavlin 188 $cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter};
231 dpavlin 178
232 dpavlin 195 my $format_name = $config->{indexer}->{$field}->{format_name};
233     my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter};
234     if ($format_name && $format_delimiter) {
235     $cache->{format}->{$field}->{format_name} = $format_name;
236     $cache->{format}->{$field}->{format_delimiter} = $format_delimiter;
237     }
238    
239 dpavlin 54 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
240 dpavlin 3
241 dpavlin 279 $x = unroll_x($x);
242    
243 dpavlin 40 my $format = x($x->{content});
244     my $delimiter = x($x->{delimiter}) || ' ';
245 dpavlin 3
246 dpavlin 178 my $repeat_off = 0; # init repeatable offset
247 dpavlin 29
248 dpavlin 274 my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
249 dpavlin 29
250 dpavlin 35 # what will separate last line from this one?
251 dpavlin 182 if ($display_data && $x->{append}) {
252 dpavlin 290 $line_delimiter = $delimiter;
253 dpavlin 35 } elsif ($display_data) {
254 dpavlin 34 $line_delimiter = '<br/>';
255     }
256 dpavlin 29
257 dpavlin 34 # init vars so that we go into while...
258     ($swish,$display) = (1,1);
259 dpavlin 29
260 dpavlin 97 # placeholder for all repeatable entries for index
261    
262 dpavlin 195 sub mkformat($$) {
263 dpavlin 153 my $x = shift || die "mkformat needs tag reference";
264     my $data = shift || return;
265 dpavlin 263 my $format_name = x($x->{format_name}) || return $data;
266 dpavlin 153 my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
267     my $format_delimiter = x($x->{format_delimiter});
268     my @data;
269     if ($format_delimiter) {
270     @data = split(/$format_delimiter/,$data);
271     } else {
272     push @data,$data;
273     }
274    
275     if ($fmt) {
276     my $nr = scalar $fmt =~ s/%s/%s/g;
277     if (($#data+1) == $nr) {
278 dpavlin 263 return sprintf($fmt,@data);
279 dpavlin 153 } else {
280 dpavlin 207 #print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
281 dpavlin 263 return $data;
282 dpavlin 153 }
283     } else {
284     print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
285     }
286     }
287    
288 dpavlin 90 # while because of repeatable fields
289     while ($swish || $display) {
290 dpavlin 178 my $page = $repeat_off;
291     $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
292 dpavlin 54 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
293 dpavlin 90 if ($repeat_off > 1000) {
294     print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
295     last;
296     }
297 dpavlin 177
298     # is this field is lookup?
299     if ($display && $x->{lookup}) {
300 dpavlin 178 my $null = "<!-- null -->";
301 dpavlin 177 if ($use_lhash_cache) {
302     if (!defined($cache->{lhash}->{$display})) {
303     my $new_display = $lhash{$display};
304 dpavlin 178 if (defined($new_display)) {
305 dpavlin 177 #print STDERR "lookup cache store '$display' = '$new_display'\n";
306     $display = $new_display;
307     $cache->{lhash}->{$display} = $new_display;
308     } else {
309 dpavlin 255 # print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
310 dpavlin 177 $display = "";
311 dpavlin 178 $cache->{lhash}->{$display} = $null;
312 dpavlin 177 }
313     } else {
314     $display = $cache->{lhash}->{$display};
315     }
316     } else {
317 dpavlin 178 $display = $lhash{$display} || $null;
318 dpavlin 177 }
319     }
320    
321 dpavlin 29 # filter="name" ; filter this field through
322     # filter/[name].pm
323     my $filter = $x->{filter};
324 dpavlin 170 if ($filter && !$cache->{filter_loaded}->{$filter}) {
325 dpavlin 29 require "filter/".$filter.".pm";
326 dpavlin 170 $cache->{filter_loaded}->{$filter}++;
327 dpavlin 20 }
328 dpavlin 29 # type="swish" ; field for swish
329 dpavlin 163 if ($swish) {
330 dpavlin 259 my $tmp = $swish;
331 dpavlin 163 if ($filter && ($s || $se)) {
332 dpavlin 29 no strict 'refs';
333 dpavlin 259 $tmp = join(" ",&$filter($tmp)) if ($s || $se);
334 dpavlin 29 }
335 dpavlin 234
336 dpavlin 255 $swish_data .= $tmp if ($s && $tmp);
337 dpavlin 256 $swish_exact_data .= "xxbxx $tmp xxexx " if ($tmp && $tmp ne "" && $se);
338 dpavlin 29 }
339 dpavlin 17
340 dpavlin 29 # type="display" ; field for display
341     if ($d && $display) {
342 dpavlin 207 my $ldel = $delimiter;
343 dpavlin 35 if ($line_delimiter && $display_data) {
344 dpavlin 207 $ldel = $line_delimiter;
345 dpavlin 35 }
346 dpavlin 29 if ($filter) {
347     no strict 'refs';
348 dpavlin 207 my @arr;
349     foreach my $tmp (&$filter($display)) {
350     my $tmp2 = mkformat($x,$tmp);
351     push @arr,$tmp2 if ($tmp2);
352 dpavlin 138 }
353 dpavlin 207 $display_data .= $ldel if ($display_data && @arr);
354     $display_data .= join($delimiter,@arr);
355 dpavlin 29 } else {
356 dpavlin 207 $display_data .= $ldel if ($display_data);
357     my $tmp = mkformat($x,$display);
358 dpavlin 224 $display_data .= $tmp if ($tmp);
359 dpavlin 29 }
360 dpavlin 20 }
361 dpavlin 29
362     # type="index" ; insert into index
363 dpavlin 182 my $idisplay;
364 dpavlin 29 if ($i && $display) {
365 dpavlin 182 $idisplay = $display;
366 dpavlin 177 if ($filter) {
367     no strict 'refs';
368 dpavlin 182 $idisplay = &$filter($idisplay);
369 dpavlin 177 }
370 dpavlin 255 push @index_data, $idisplay if ($idisplay && !$iterate_by_page);
371 dpavlin 97 }
372    
373 dpavlin 177 # store fields in lookup
374     if ($il && $display) {
375     if (lc($x->{type}) eq "lookup_key") {
376     if ($lookup_key) {
377     print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
378     } else {
379 dpavlin 233 if ($filter) {
380     no strict 'refs';
381     $lookup_key = &$filter($display);
382     } else {
383     $lookup_key = $display;
384     }
385 dpavlin 177 }
386     } elsif (lc($x->{type}) eq "lookup_val") {
387     if ($lookup_key) {
388 dpavlin 233 if ($filter) {
389     no strict 'refs';
390     $lhash{$lookup_key} = &$filter($display);
391     } else {
392     $lhash{$lookup_key} = $display;
393     }
394 dpavlin 177 } else {
395     print STDERR "WARNING: no lookup_key defined for '$display'?";
396     }
397 dpavlin 20 }
398 dpavlin 178
399 dpavlin 20 }
400 dpavlin 178
401     # store data for page-by-page repeatable fields
402     if ($iterate_by_page) {
403     sub iterate_fld($$$$$$) {
404     my ($cache,$what,$field,$page,$data,$append) = @_;
405     return if (!$data);
406 dpavlin 182
407     my $ldel = $page_line_delimiter;
408     $ldel = " " if ($append);
409     #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel);
410 dpavlin 178 if (! $cache->{$what}->{$field}->[$page]) {
411     $cache->{$what}->{$field}->[$page] = $data;
412     } else {
413 dpavlin 182 $cache->{$what}->{$field}->[$page] .= $ldel.$data;
414 dpavlin 178 }
415     }
416    
417     if ($display_data) {
418     iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append});
419     }
420     $display_data = "";
421     if ($swish_data) {
422     iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append});
423     $swish_data = "";
424     }
425     if ($swish_exact_data) {
426     iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append});
427     $swish_exact_data = "";
428     }
429 dpavlin 182
430     if ($idisplay) {
431     my $ldel=$page_line_delimiter;
432     my @index_data;
433     if ($cache->{index_data}->{$field}->[$page]) {
434    
435     @index_data = @{$cache->{index_data}->{$field}->[$page]};
436     }
437     if ($x->{append}) {
438 dpavlin 195 if (@index_data) {
439     $index_data[$#index_data] .= $idisplay;
440     } else {
441     push @index_data, $idisplay;
442     }
443 dpavlin 182 } else {
444     push @index_data, $idisplay;
445     }
446     $idisplay = "";
447     @{$cache->{index_data}->{$field}->[$page]} = @index_data;
448     }
449 dpavlin 178 }
450 dpavlin 17 }
451 dpavlin 177
452 dpavlin 182 if (! $iterate_by_page) {
453 dpavlin 188 my $idel = $x->{index_delimiter};
454 dpavlin 182 # fill data in index
455 dpavlin 188 foreach my $tmp (@index_data) {
456     my $i = $d = $tmp;
457     if ($idel && $tmp =~ m/$idel/) {
458     ($i,$d) = split(/$idel/,$tmp);
459     }
460     $index->insert($field, $i, $d, $path);
461 dpavlin 182 }
462     @index_data = ();
463 dpavlin 177 }
464 dpavlin 3 }
465 dpavlin 9
466 dpavlin 59 # now try to parse variables from configuration file
467     foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
468 dpavlin 13
469 dpavlin 279 $x = unroll_x($x);
470    
471 dpavlin 62 my $delimiter = x($x->{delimiter}) || ' ';
472 dpavlin 59 my $val = $cfg->val($database, x($x->{content}));
473    
474 dpavlin 274 # FIXME index_lookup is not supported!
475     my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
476 dpavlin 59
477     if ($val) {
478 dpavlin 62 $display_data .= $delimiter.$val if ($d);
479 dpavlin 274 $swish_data .= " ".$val if ($s);
480 dpavlin 59 $index->insert($field, $val, $path) if ($i);
481     }
482    
483 dpavlin 178 if ($iterate_by_page) {
484     # FIXME data from config tag will appear just
485     # on first page!!!
486     my $page = 0;
487     if ($display_data) {
488     $cache->{display_data}->{$field}->[$page] = $display_data;
489     $display_data = "";
490     }
491     if ($swish_data) {
492     $cache->{swish_data}->{$field}->[$page] = $swish_data;
493     $swish_data = "";
494     }
495     if ($swish_exact_data) {
496     $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
497     $swish_exact_data = "";
498     }
499     }
500 dpavlin 59 }
501    
502 dpavlin 178 # save data page-by-page
503     foreach my $field (@page_fields) {
504     my $nr_pages = $page_max{$field} || next;
505     #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
506     #print STDERR Dumper($cache->{display_data});
507     for (my $page=0; $page <= $nr_pages; $page++) {
508 dpavlin 195 my $display_data;
509     if ($cache->{format}->{$field}) {
510 dpavlin 207 my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
511     $display_data=$tmp if ($tmp);
512 dpavlin 195 } else {
513     $display_data = $cache->{display_data}->{$field}->[$page];
514     }
515 dpavlin 178 if ($display_data) { # default
516     if ($field eq "headline") {
517     $xml .= xmlify("headline", $display_data);
518     } else {
519 dpavlin 29
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    
526     my $swish_data = $cache->{swish_data}->{$field}->[$page];
527     if ($swish_data) {
528     # remove extra spaces
529     $swish_data =~ s/ +/ /g;
530     $swish_data =~ s/ +$//g;
531 dpavlin 13
532 dpavlin 320 $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
533 dpavlin 13 }
534 dpavlin 178
535     my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
536     if ($swish_exact_data) {
537     $swish_exact_data =~ s/ +/ /g;
538     $swish_exact_data =~ s/ +$//g;
539    
540     # add delimiters before and after word.
541     # That is required to produce exact match
542 dpavlin 320 $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
543 dpavlin 13 }
544 dpavlin 182
545 dpavlin 188 my $idel = $cache->{index_delimiter}->{$field};
546     foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) {
547     my $i = $tmp;
548     my $d = $tmp;
549     if ($idel && $tmp =~ m/$idel/) {
550     ($i,$d) = split(/$idel/,$tmp);
551     }
552     $index->insert($field, $i, $d, $path);
553     #print STDERR "index [$idel] $field: $i --> $d [$path]\n";
554 dpavlin 182 }
555 dpavlin 13 }
556 dpavlin 178
557 dpavlin 10 }
558 dpavlin 178
559     if (! $iterate_by_page) {
560     if ($display_data) {
561     if ($field eq "headline") {
562     $xml .= xmlify("headline", $display_data);
563     } else {
564 dpavlin 20
565 dpavlin 178 # fallback to empty field name if needed
566     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
567     $html .= "#-#".$display_data."###\n";
568     }
569     }
570     if ($swish_data) {
571     # remove extra spaces
572     $swish_data =~ s/ +/ /g;
573     $swish_data =~ s/ +$//g;
574 dpavlin 10
575 dpavlin 320 $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
576 dpavlin 178 }
577 dpavlin 9
578 dpavlin 178 if ($swish_exact_data) {
579     $swish_exact_data =~ s/ +/ /g;
580     $swish_exact_data =~ s/ +$//g;
581    
582     # add delimiters before and after word.
583     # That is required to produce exact match
584 dpavlin 320 $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
585 dpavlin 178 }
586 dpavlin 163 }
587 dpavlin 3 }
588 dpavlin 13
589     # dump formatted output in <html>
590     if ($html) {
591 dpavlin 81 #$xml .= xmlify("html",$html);
592     $xml .= "<html><![CDATA[ $html ]]></html>";
593 dpavlin 13 }
594    
595 dpavlin 3 if ($xml) {
596 dpavlin 13 $xml .= $add_xml if ($add_xml);
597 dpavlin 10 return "<xml>\n$xml</xml>\n";
598 dpavlin 3 } else {
599     return;
600     }
601     }
602    
603     ##########################################################################
604    
605 dpavlin 54 # read configuration for this script
606 dpavlin 13 my $cfg = new Config::IniFiles( -file => $config_file );
607 dpavlin 1
608 dpavlin 54 # read global.conf configuration
609     my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
610    
611 dpavlin 50 # open index
612     $index = new index_DBI(
613 dpavlin 54 $cfg_global->val('global', 'dbi_dbd'),
614     $cfg_global->val('global', 'dbi_dsn'),
615     $cfg_global->val('global', 'dbi_user'),
616     $cfg_global->val('global', 'dbi_passwd') || '',
617 dpavlin 50 );
618    
619 dpavlin 97 my $show_progress = $cfg_global->val('global', 'show_progress');
620    
621 dpavlin 320 my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter');
622     if ($my_unac_filter) {
623     print STDERR "using $my_unac_filter to filter characters for search\n";
624     require $my_unac_filter;
625     } else {
626     print STDERR "### fallback to default my_unac_string!\n";
627     eval q{
628     sub main::my_unac_string($$) {
629     my ($charset, $string) = (@_);
630     return $string;
631     }
632     };
633 dpavlin 164 }
634    
635 dpavlin 13 foreach my $database ($cfg->Sections) {
636 dpavlin 1
637 dpavlin 54 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
638 dpavlin 40 my $add_xml = $cfg -> val($database, 'xml'); # optional
639 dpavlin 1
640 dpavlin 177 # create new lookup file
641     my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
642     if ($lookup_file) {
643     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
644 dpavlin 379 if (! -e $lookup_file) {
645     open(LOOKUP, "> $lookup_file") || die "can't create $lookup_file': $!";
646     close(LOOKUP);
647     }
648 dpavlin 177 tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
649     print STDERR "creating lookup file '$lookup_file'\n";
650 dpavlin 231 # delete memory cache for lookup file
651     delete $cache->{lhash};
652 dpavlin 177 }
653    
654     # open existing lookup file
655     $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
656     if ($lookup_file) {
657     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
658     tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
659     print STDERR "opening lookup file '$lookup_file'\n";
660     }
661    
662 dpavlin 54 print STDERR "reading ./import_xml/$type.xml\n";
663 dpavlin 1
664 dpavlin 58 # extract just type basic
665     my $type_base = $type;
666     $type_base =~ s/_.+$//g;
667 dpavlin 40
668 dpavlin 279 $config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 );
669 dpavlin 58
670 dpavlin 54 # output current progress indicator
671     my $last_p = 0;
672     sub progress {
673 dpavlin 101 return if (! $show_progress);
674 dpavlin 54 my $current = shift;
675 dpavlin 57 my $total = shift || 1;
676 dpavlin 54 my $p = int($current * 100 / $total);
677     if ($p != $last_p) {
678     printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
679     $last_p = $p;
680     }
681 dpavlin 3 }
682    
683 dpavlin 67 my $fake_dir = 1;
684     sub fakeprogress {
685 dpavlin 104 return if (! $show_progress);
686 dpavlin 67 my $current = shift @_;
687    
688     my @ind = ('-','\\','|','/','-','\\','|','/', '-');
689    
690     $last_p += $fake_dir;
691     $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
692     if ($last_p % 10 == 0) {
693     printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
694     }
695     }
696    
697 dpavlin 54 # now read database
698     print STDERR "using: $type...\n";
699 dpavlin 1
700 dpavlin 180 # erase cache for tags by order in this database
701     delete $cache->{tags_by_order};
702    
703 dpavlin 58 if ($type_base eq "isis") {
704    
705 dpavlin 54 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
706 dpavlin 1
707 dpavlin 54 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
708     my $db = OpenIsis::open( $isis_db );
709 dpavlin 10
710 dpavlin 106 # check if .txt database for OpenIsis is zero length,
711     # if so, erase it and re-open database
712     sub check_txt_db {
713     my $isis_db = shift || die "need isis database name";
714 dpavlin 109 my $reopen = 0;
715    
716 dpavlin 106 if (-e $isis_db.".TXT") {
717 dpavlin 109 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
718 dpavlin 108 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
719 dpavlin 109 $reopen++;
720 dpavlin 106 }
721 dpavlin 109 if (-e $isis_db.".PTR") {
722     print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
723     unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
724     $reopen++;
725     }
726     return OpenIsis::open( $isis_db ) if ($reopen);
727 dpavlin 106 }
728    
729     # EOF error
730     if ($db == -1) {
731     $db = check_txt_db($isis_db);
732 dpavlin 108 if ($db == -1) {
733 dpavlin 106 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
734     next;
735     }
736     }
737    
738 dpavlin 101 # OpenIsis::ERR_BADF
739     if ($db == -4) {
740     print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
741     next;
742     # OpenIsis::ERR_IO
743     } elsif ($db == -5) {
744     print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
745     next;
746     } elsif ($db < 0) {
747     print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
748     next;
749     }
750    
751 dpavlin 54 my $max_rowid = OpenIsis::maxRowid( $db );
752 dpavlin 3
753 dpavlin 106 # if 0 records, try to rease isis .txt database
754     if ($max_rowid == 0) {
755     # force removal of database
756     $db = check_txt_db($isis_db);
757     $max_rowid = OpenIsis::maxRowid( $db );
758     }
759    
760 dpavlin 54 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
761    
762     my $path = $database;
763    
764     for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
765     my $row = OpenIsis::read( $db, $row_id );
766     if ($row && $row->{mfn}) {
767    
768     progress($row->{mfn}, $max_rowid);
769    
770     my $swishpath = $path."#".int($row->{mfn});
771    
772 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
773 dpavlin 54 $xml = $cp2utf->convert($xml);
774     use bytes; # as opposed to chars
775     print "Path-Name: $swishpath\n";
776     print "Content-Length: ".(length($xml)+1)."\n";
777     print "Document-Type: XML\n\n$xml\n";
778     }
779 dpavlin 3 }
780 dpavlin 54 }
781 dpavlin 108 # for this to work with current version of OpenIsis (0.9.0)
782     # you might need my patch from
783     # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
784     OpenIsis::close($db);
785 dpavlin 54 print STDERR "\n";
786 dpavlin 3
787 dpavlin 58 } elsif ($type_base eq "excel") {
788 dpavlin 255 require Spreadsheet::ParseExcel;
789     require Spreadsheet::ParseExcel::Utility;
790     import Spreadsheet::ParseExcel::Utility qw(int2col);
791 dpavlin 54
792     $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
793     my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
794 dpavlin 43
795 dpavlin 54 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
796 dpavlin 56 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
797 dpavlin 54
798     my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
799    
800     my $sheet_nr = 0;
801     foreach my $oWks (@{$oBook->{Worksheet}}) {
802     #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
803     last if ($oWks->{Name} eq $sheet);
804     $sheet_nr++;
805     }
806    
807     my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
808    
809     print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
810     defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
811     my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
812    
813 dpavlin 56 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
814 dpavlin 54 my $row;
815     for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
816     my $cell = $oWorksheet->{Cells}[$iR][$iC];
817     if ($cell) {
818 dpavlin 488 # this conversion is a cludge.
819     # Files from Excell could have
820     # characters which don't fit into
821     # destination encoding.
822     $row->{int2col($iC)} = $utf2cp->convert($cell->Value) || $cell->Value;
823 dpavlin 54 }
824     }
825    
826     progress($iR, $end_row);
827    
828     # print "row[$iR/$end_row] ";
829     # foreach (keys %{$row}) {
830     # print "$_: ",$row->{$_},"\t";
831     # }
832     # print "\n";
833    
834     my $swishpath = $database."#".$iR;
835    
836     next if (! $row);
837    
838 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
839 dpavlin 44 $xml = $cp2utf->convert($xml);
840 dpavlin 35 use bytes; # as opposed to chars
841 dpavlin 43 print "Path-Name: $swishpath\n";
842 dpavlin 10 print "Content-Length: ".(length($xml)+1)."\n";
843     print "Document-Type: XML\n\n$xml\n";
844 dpavlin 3 }
845 dpavlin 1 }
846 dpavlin 62 } elsif ($type_base eq "marc") {
847 dpavlin 67
848 dpavlin 255 require MARC;
849 dpavlin 62
850     $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
851     my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
852    
853     # optional argument is format
854 dpavlin 218 my $format = x($config->{marc_format}) || 'usmarc';
855 dpavlin 62
856     print STDERR "Reading MARC file '$marc_file'\n";
857    
858     my $marc = new MARC;
859     my $nr = $marc->openmarc({
860     file=>$marc_file, format=>$format
861 dpavlin 218 }) || die "Can't open MARC file '$marc_file' with format '$format'";
862 dpavlin 62
863 dpavlin 215 # read MARC file in memory
864     $marc->nextmarc(-1);
865 dpavlin 62
866 dpavlin 215 my $max_rec = $marc->marc_count();
867 dpavlin 62
868 dpavlin 215 for(my $i=1; $i<=$max_rec; $i++) {
869 dpavlin 62
870 dpavlin 215 progress($i,$max_rec);
871 dpavlin 62
872 dpavlin 215 # store value for marc_sf.pm
873     $main::cache->{marc_record} = $i;
874    
875 dpavlin 62 my $swishpath = $database."#".$i;
876    
877     if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
878     $xml = $cp2utf->convert($xml);
879     use bytes; # as opposed to chars
880     print "Path-Name: $swishpath\n";
881     print "Content-Length: ".(length($xml)+1)."\n";
882     print "Document-Type: XML\n\n$xml\n";
883     }
884     }
885 dpavlin 215
886     print STDERR "\n";
887    
888 dpavlin 67 } elsif ($type_base eq "feed") {
889    
890     $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
891     my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
892    
893     print STDERR "Reading feed from program '$prog'\n";
894    
895     open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
896    
897 dpavlin 74 my $i=1; # record nr.
898 dpavlin 67
899     my $data;
900     my $line=1;
901    
902     while (<FEED>) {
903     chomp;
904    
905     if (/^$/) {
906 dpavlin 74 my $swishpath = $database."#".$i++;
907 dpavlin 67
908     if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
909     $xml = $cp2utf->convert($xml);
910     use bytes; # as opposed to chars
911     print "Path-Name: $swishpath\n";
912     print "Content-Length: ".(length($xml)+1)."\n";
913     print "Document-Type: XML\n\n$xml\n";
914     }
915     $line = 1;
916     $data = {};
917     next;
918     }
919    
920 dpavlin 74 $line = $1 if (s/^(\d+):\s*//);
921 dpavlin 67 $data->{$line++} = $_;
922    
923 dpavlin 74 fakeprogress($i);
924 dpavlin 67
925     }
926 dpavlin 177 # close lookup
927     untie %lhash if (%lhash);
928 dpavlin 1 }
929     }
930 dpavlin 3
931 dpavlin 10 # call this to commit index
932     $index->close;
933 dpavlin 3
934     1;
935     __END__
936     ##########################################################################
937    
938     =head1 NAME
939    
940 dpavlin 81 all2xml.pl - read various file formats and dump XML for SWISH-E
941 dpavlin 3
942 dpavlin 298 =head1 SYNOPSYS
943    
944     $ all2xml.pl [test.conf]
945    
946 dpavlin 3 =head1 DESCRIPTION
947    
948 dpavlin 81 This command will read ISIS data file using OpenIsis perl module, MARC
949     records using MARC module and optionally Micro$oft Excel files to
950     create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
951     this script B<isn't general xml generator> from isis files (isis allready
952     has something like that). Output of this script is tailor-made for SWISH-E.
953 dpavlin 3
954 dpavlin 298 If no configuration file is specified, it will use default one called
955     C<all2xml.conf>.
956    
957 dpavlin 81 =head1 BUGS
958    
959     Documentation is really lacking. However, in true Open Source spirit, source
960     is best documentation. I even made considerable effort to comment parts
961     which are not intuitively clear, so...
962    
963 dpavlin 3 =head1 AUTHOR
964    
965     Dobrica Pavlinusic <dpavlin@rot13.org>
966    
967     =head1 COPYRIGHT
968    
969     GNU Public License (GPL) v2 or later
970    
971     =head1 SEE ALSO
972    
973     SWISH-E web site at http://www.swish-e.org
974    
975     =cut

Properties

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

  ViewVC Help
Powered by ViewVC 1.1.26