/[webpac]/branches/ffzg/all2xml.pl
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Annotation of /branches/ffzg/all2xml.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 188 - (hide annotations)
Sat Nov 29 19:07:00 2003 UTC (20 years, 3 months ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 24159 byte(s)
implemented index_delimiter which enables to to format index entries in format
(values to be inserted in index);;(values to be displayed) if there is
definition of index_delimiter=";;". This will allow you to index (and
search) through values from original database and still have ability to
display lookup 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 54 sub data2xml {
76    
77 dpavlin 10 use xmlify;
78    
79 dpavlin 54 my $type = shift @_;
80 dpavlin 3 my $row = shift @_;
81 dpavlin 13 my $add_xml = shift @_;
82 dpavlin 59 # needed to read values from configuration file
83     my $cfg = shift @_;
84     my $database = shift @_;
85 dpavlin 3
86     my $xml;
87    
88 dpavlin 10 use parse_format;
89 dpavlin 3
90 dpavlin 13 my $html = ""; # html formatted display output
91 dpavlin 10
92 dpavlin 13 my %field_usage; # counter for usage of each field
93    
94 dpavlin 32 # sort subrouting using order="" attribute
95     sub by_order {
96 dpavlin 98 my $va = $config->{indexer}->{$a}->{order} ||
97     $config->{indexer}->{$a};
98     my $vb = $config->{indexer}->{$b}->{order} ||
99     $config->{indexer}->{$b};
100 dpavlin 29
101 dpavlin 98 return $va <=> $vb;
102 dpavlin 32 }
103 dpavlin 3
104 dpavlin 170 my @sorted_tags;
105 dpavlin 180 if ($cache->{tags_by_order}) {
106     @sorted_tags = @{$cache->{tags_by_order}};
107 dpavlin 170 } else {
108     @sorted_tags = sort by_order keys %{$config->{indexer}};
109 dpavlin 180 $cache->{tags_by_order} = \@sorted_tags;
110 dpavlin 170 }
111 dpavlin 32
112 dpavlin 177 # lookup key
113     my $lookup_key;
114    
115 dpavlin 178 # cache for field in pages
116 dpavlin 180 delete $cache->{display_data};
117     delete $cache->{swish_data};
118     delete $cache->{swish_exact_data};
119 dpavlin 182 delete $cache->{index_data};
120 dpavlin 188 delete $cache->{index_delimiter};
121 dpavlin 178 my @page_fields; # names of fields
122    
123    
124     # subs used to produce output
125    
126     sub get_field_name($$$) {
127     my ($config,$field,$field_usage) = @_;
128    
129     # find field name (signular, plural)
130     my $field_name = "";
131     if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) {
132     $field_name = $config->{indexer}->{$field}->{name_singular};
133     } elsif ($config->{indexer}->{$field}->{name_plural}) {
134     $field_name = $config->{indexer}->{$field}->{name_plural};
135     } elsif ($config->{indexer}->{$field}->{name}) {
136     $field_name = $config->{indexer}->{$field}->{name};
137     } else {
138     print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
139     }
140     if ($field_name) {
141     return x($field_name);
142     }
143     }
144    
145    
146     # begin real work: go field by field
147 dpavlin 170 foreach my $field (@sorted_tags) {
148    
149 dpavlin 40 $field=x($field);
150 dpavlin 13 $field_usage{$field}++;
151    
152 dpavlin 10 my $swish_data = "";
153 dpavlin 163 my $swish_exact_data = "";
154 dpavlin 3 my $display_data = "";
155 dpavlin 182 my @index_data;
156 dpavlin 35 my $line_delimiter;
157 dpavlin 3
158 dpavlin 34 my ($swish,$display);
159    
160 dpavlin 54 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
161 dpavlin 178
162     # is this field page-by-page?
163     my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
164     push @page_fields,$field if ($iterate_by_page);
165     my %page_max = ();
166     # default line_delimiter if using
167     my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
168 dpavlin 188 $cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter};
169 dpavlin 178
170 dpavlin 54 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
171 dpavlin 3
172 dpavlin 40 my $format = x($x->{content});
173     my $delimiter = x($x->{delimiter}) || ' ';
174 dpavlin 3
175 dpavlin 178 my $repeat_off = 0; # init repeatable offset
176 dpavlin 29
177 dpavlin 177 # swish, swish_exact, display, index, index_lookup
178     # swish and display defaults
179     my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
180 dpavlin 29 $s = 0 if (lc($x->{type}) eq "display");
181     $d = 0 if (lc($x->{type}) eq "swish");
182 dpavlin 188 ($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index");
183 dpavlin 181 ($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact");
184 dpavlin 182 ($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/);
185 dpavlin 29
186 dpavlin 35 # what will separate last line from this one?
187 dpavlin 182 if ($display_data && $x->{append}) {
188 dpavlin 34 $line_delimiter = ' ';
189 dpavlin 35 } elsif ($display_data) {
190 dpavlin 34 $line_delimiter = '<br/>';
191     }
192 dpavlin 29
193 dpavlin 34 # init vars so that we go into while...
194     ($swish,$display) = (1,1);
195 dpavlin 29
196 dpavlin 97 # placeholder for all repeatable entries for index
197    
198 dpavlin 153 sub mkformat {
199     my $x = shift || die "mkformat needs tag reference";
200     my $data = shift || return;
201     my $format_name = x($x->{format_name}) || return $data;
202     my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
203     my $format_delimiter = x($x->{format_delimiter});
204     my @data;
205     if ($format_delimiter) {
206     @data = split(/$format_delimiter/,$data);
207     } else {
208     push @data,$data;
209     }
210    
211     if ($fmt) {
212     my $nr = scalar $fmt =~ s/%s/%s/g;
213     if (($#data+1) == $nr) {
214     return sprintf($fmt,@data);
215     } else {
216     print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
217     return $data;
218     }
219     } else {
220     print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
221     }
222     }
223    
224 dpavlin 90 # while because of repeatable fields
225     while ($swish || $display) {
226 dpavlin 178 my $page = $repeat_off;
227     $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
228 dpavlin 54 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
229 dpavlin 90 if ($repeat_off > 1000) {
230     print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
231     last;
232     }
233 dpavlin 177
234     # is this field is lookup?
235     if ($display && $x->{lookup}) {
236 dpavlin 178 my $null = "<!-- null -->";
237 dpavlin 177 if ($use_lhash_cache) {
238     if (!defined($cache->{lhash}->{$display})) {
239     my $new_display = $lhash{$display};
240 dpavlin 178 if (defined($new_display)) {
241 dpavlin 177 #print STDERR "lookup cache store '$display' = '$new_display'\n";
242     $display = $new_display;
243     $cache->{lhash}->{$display} = $new_display;
244     } else {
245     print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
246     $display = "";
247 dpavlin 178 $cache->{lhash}->{$display} = $null;
248 dpavlin 177 }
249     } else {
250     $display = $cache->{lhash}->{$display};
251     }
252     } else {
253 dpavlin 178 $display = $lhash{$display} || $null;
254 dpavlin 177 }
255     }
256    
257 dpavlin 29 # filter="name" ; filter this field through
258     # filter/[name].pm
259     my $filter = $x->{filter};
260 dpavlin 170 if ($filter && !$cache->{filter_loaded}->{$filter}) {
261 dpavlin 29 require "filter/".$filter.".pm";
262 dpavlin 170 $cache->{filter_loaded}->{$filter}++;
263 dpavlin 20 }
264 dpavlin 29 # type="swish" ; field for swish
265 dpavlin 163 if ($swish) {
266     if ($filter && ($s || $se)) {
267 dpavlin 29 no strict 'refs';
268 dpavlin 163 my $tmp = join(" ",&$filter($swish)) if ($s || $se);
269     $swish_data .= $tmp if ($s);
270     $swish_exact_data .= $tmp if ($se);
271    
272 dpavlin 29 } else {
273 dpavlin 163 $swish_data .= $swish if ($s);
274     $swish_exact_data .= $swish if ($se);
275 dpavlin 29 }
276     }
277 dpavlin 17
278 dpavlin 29 # type="display" ; field for display
279     if ($d && $display) {
280 dpavlin 35 if ($line_delimiter && $display_data) {
281     $display_data .= $line_delimiter;
282     }
283 dpavlin 29 if ($filter) {
284     no strict 'refs';
285 dpavlin 138 if ($display_data) {
286 dpavlin 153 $display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display)));
287 dpavlin 138 } else {
288 dpavlin 153 $display_data = join($delimiter,mkformat($x,&$filter($display)));
289 dpavlin 138 }
290 dpavlin 29 } else {
291 dpavlin 138 if ($display_data) {
292 dpavlin 153 $display_data .= $delimiter.mkformat($x,$display);
293 dpavlin 138 } else {
294 dpavlin 153 $display_data = mkformat($x,$display);
295 dpavlin 138 }
296 dpavlin 29 }
297 dpavlin 20 }
298 dpavlin 29
299     # type="index" ; insert into index
300 dpavlin 182 my $idisplay;
301 dpavlin 29 if ($i && $display) {
302 dpavlin 182 $idisplay = $display;
303 dpavlin 177 if ($filter) {
304     no strict 'refs';
305 dpavlin 182 $idisplay = &$filter($idisplay);
306 dpavlin 177 }
307 dpavlin 182 push @index_data, $idisplay if (! $iterate_by_page);
308 dpavlin 97 }
309    
310 dpavlin 177 # store fields in lookup
311     if ($il && $display) {
312     if (lc($x->{type}) eq "lookup_key") {
313     if ($lookup_key) {
314     print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
315     } else {
316     $lookup_key = $display;
317     }
318     } elsif (lc($x->{type}) eq "lookup_val") {
319     if ($lookup_key) {
320     $lhash{$lookup_key} = $display;
321     } else {
322     print STDERR "WARNING: no lookup_key defined for '$display'?";
323     }
324 dpavlin 20 }
325 dpavlin 178
326 dpavlin 20 }
327 dpavlin 178
328     # store data for page-by-page repeatable fields
329     if ($iterate_by_page) {
330     sub iterate_fld($$$$$$) {
331     my ($cache,$what,$field,$page,$data,$append) = @_;
332     return if (!$data);
333 dpavlin 182
334     my $ldel = $page_line_delimiter;
335     $ldel = " " if ($append);
336     #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel);
337 dpavlin 178 if (! $cache->{$what}->{$field}->[$page]) {
338     $cache->{$what}->{$field}->[$page] = $data;
339     } else {
340 dpavlin 182 $cache->{$what}->{$field}->[$page] .= $ldel.$data;
341 dpavlin 178 }
342     }
343    
344     if ($display_data) {
345     iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append});
346     }
347     $display_data = "";
348     if ($swish_data) {
349     iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append});
350     $swish_data = "";
351     }
352     if ($swish_exact_data) {
353     iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append});
354     $swish_exact_data = "";
355     }
356 dpavlin 182
357     if ($idisplay) {
358     my $ldel=$page_line_delimiter;
359     my @index_data;
360     if ($cache->{index_data}->{$field}->[$page]) {
361    
362     @index_data = @{$cache->{index_data}->{$field}->[$page]};
363     }
364     if ($x->{append}) {
365     $index_data[$#index_data] .= $idisplay;
366     } else {
367     push @index_data, $idisplay;
368     }
369     $idisplay = "";
370     @{$cache->{index_data}->{$field}->[$page]} = @index_data;
371     }
372 dpavlin 178 }
373 dpavlin 17 }
374 dpavlin 177
375 dpavlin 182 if (! $iterate_by_page) {
376 dpavlin 188 my $idel = $x->{index_delimiter};
377 dpavlin 182 # fill data in index
378 dpavlin 188 foreach my $tmp (@index_data) {
379     my $i = $d = $tmp;
380     if ($idel && $tmp =~ m/$idel/) {
381     ($i,$d) = split(/$idel/,$tmp);
382     }
383     $index->insert($field, $i, $d, $path);
384 dpavlin 182 }
385     @index_data = ();
386 dpavlin 177 }
387 dpavlin 3 }
388 dpavlin 9
389 dpavlin 59 # now try to parse variables from configuration file
390     foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
391 dpavlin 13
392 dpavlin 62 my $delimiter = x($x->{delimiter}) || ' ';
393 dpavlin 59 my $val = $cfg->val($database, x($x->{content}));
394    
395     my ($s,$d,$i) = (1,1,0); # swish, display default
396     $s = 0 if (lc($x->{type}) eq "display");
397     $d = 0 if (lc($x->{type}) eq "swish");
398 dpavlin 163 # no support for swish exact in config.
399     # IMHO, it's useless
400 dpavlin 59 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
401    
402     if ($val) {
403 dpavlin 62 $display_data .= $delimiter.$val if ($d);
404 dpavlin 59 $swish_data .= $val if ($s);
405     $index->insert($field, $val, $path) if ($i);
406     }
407    
408 dpavlin 178 if ($iterate_by_page) {
409     # FIXME data from config tag will appear just
410     # on first page!!!
411     my $page = 0;
412     if ($display_data) {
413     $cache->{display_data}->{$field}->[$page] = $display_data;
414     $display_data = "";
415     }
416     if ($swish_data) {
417     $cache->{swish_data}->{$field}->[$page] = $swish_data;
418     $swish_data = "";
419     }
420     if ($swish_exact_data) {
421     $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
422     $swish_exact_data = "";
423     }
424     }
425 dpavlin 59 }
426    
427 dpavlin 178 # save data page-by-page
428     foreach my $field (@page_fields) {
429     my $nr_pages = $page_max{$field} || next;
430     #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
431     #print STDERR Dumper($cache->{display_data});
432     for (my $page=0; $page <= $nr_pages; $page++) {
433 dpavlin 59
434 dpavlin 178 my $display_data = $cache->{display_data}->{$field}->[$page];
435     if ($display_data) { # default
436     if ($field eq "headline") {
437     $xml .= xmlify("headline", $display_data);
438     } else {
439 dpavlin 29
440 dpavlin 178 # fallback to empty field name if needed
441     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
442     $html .= "#-#".$display_data."###\n";
443     }
444     }
445    
446     my $swish_data = $cache->{swish_data}->{$field}->[$page];
447     if ($swish_data) {
448     # remove extra spaces
449     $swish_data =~ s/ +/ /g;
450     $swish_data =~ s/ +$//g;
451 dpavlin 13
452 dpavlin 178 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
453 dpavlin 13 }
454 dpavlin 178
455     my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
456     if ($swish_exact_data) {
457     $swish_exact_data =~ s/ +/ /g;
458     $swish_exact_data =~ s/ +$//g;
459    
460     # add delimiters before and after word.
461     # That is required to produce exact match
462     $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
463 dpavlin 13 }
464 dpavlin 182
465 dpavlin 188 my $idel = $cache->{index_delimiter}->{$field};
466     foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) {
467     my $i = $tmp;
468     my $d = $tmp;
469     if ($idel && $tmp =~ m/$idel/) {
470     ($i,$d) = split(/$idel/,$tmp);
471     }
472     $index->insert($field, $i, $d, $path);
473     #print STDERR "index [$idel] $field: $i --> $d [$path]\n";
474 dpavlin 182 }
475 dpavlin 13 }
476 dpavlin 178
477 dpavlin 10 }
478 dpavlin 178
479     if (! $iterate_by_page) {
480     if ($display_data) {
481     if ($field eq "headline") {
482     $xml .= xmlify("headline", $display_data);
483     } else {
484 dpavlin 20
485 dpavlin 178 # fallback to empty field name if needed
486     $html .= get_field_name($config,$field,$field_usage{$field}) || '';
487     $html .= "#-#".$display_data."###\n";
488     }
489     }
490     if ($swish_data) {
491     # remove extra spaces
492     $swish_data =~ s/ +/ /g;
493     $swish_data =~ s/ +$//g;
494 dpavlin 10
495 dpavlin 178 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
496     }
497 dpavlin 9
498 dpavlin 178 if ($swish_exact_data) {
499     $swish_exact_data =~ s/ +/ /g;
500     $swish_exact_data =~ s/ +$//g;
501    
502     # add delimiters before and after word.
503     # That is required to produce exact match
504     $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
505     }
506 dpavlin 163 }
507 dpavlin 3 }
508 dpavlin 13
509     # dump formatted output in <html>
510     if ($html) {
511 dpavlin 81 #$xml .= xmlify("html",$html);
512     $xml .= "<html><![CDATA[ $html ]]></html>";
513 dpavlin 13 }
514    
515 dpavlin 3 if ($xml) {
516 dpavlin 13 $xml .= $add_xml if ($add_xml);
517 dpavlin 10 return "<xml>\n$xml</xml>\n";
518 dpavlin 3 } else {
519     return;
520     }
521     }
522    
523     ##########################################################################
524    
525 dpavlin 54 # read configuration for this script
526 dpavlin 13 my $cfg = new Config::IniFiles( -file => $config_file );
527 dpavlin 1
528 dpavlin 54 # read global.conf configuration
529     my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
530    
531 dpavlin 50 # open index
532     $index = new index_DBI(
533 dpavlin 54 $cfg_global->val('global', 'dbi_dbd'),
534     $cfg_global->val('global', 'dbi_dsn'),
535     $cfg_global->val('global', 'dbi_user'),
536     $cfg_global->val('global', 'dbi_passwd') || '',
537 dpavlin 50 );
538    
539 dpavlin 97 my $show_progress = $cfg_global->val('global', 'show_progress');
540    
541 dpavlin 164 my $unac_filter = $cfg_global->val('global', 'unac_filter');
542     if ($unac_filter) {
543     require $unac_filter;
544     }
545    
546 dpavlin 13 foreach my $database ($cfg->Sections) {
547 dpavlin 1
548 dpavlin 54 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
549 dpavlin 40 my $add_xml = $cfg -> val($database, 'xml'); # optional
550 dpavlin 1
551 dpavlin 177 # create new lookup file
552     my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
553     if ($lookup_file) {
554     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
555     tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
556     print STDERR "creating lookup file '$lookup_file'\n";
557     }
558    
559     # open existing lookup file
560     $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
561     if ($lookup_file) {
562     #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
563     tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
564     print STDERR "opening lookup file '$lookup_file'\n";
565     }
566    
567 dpavlin 54 print STDERR "reading ./import_xml/$type.xml\n";
568 dpavlin 1
569 dpavlin 58 # extract just type basic
570     my $type_base = $type;
571     $type_base =~ s/_.+$//g;
572 dpavlin 40
573 dpavlin 153 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
574 dpavlin 58
575 dpavlin 54 # output current progress indicator
576     my $last_p = 0;
577     sub progress {
578 dpavlin 101 return if (! $show_progress);
579 dpavlin 54 my $current = shift;
580 dpavlin 57 my $total = shift || 1;
581 dpavlin 54 my $p = int($current * 100 / $total);
582     if ($p != $last_p) {
583     printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
584     $last_p = $p;
585     }
586 dpavlin 3 }
587    
588 dpavlin 67 my $fake_dir = 1;
589     sub fakeprogress {
590 dpavlin 104 return if (! $show_progress);
591 dpavlin 67 my $current = shift @_;
592    
593     my @ind = ('-','\\','|','/','-','\\','|','/', '-');
594    
595     $last_p += $fake_dir;
596     $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
597     if ($last_p % 10 == 0) {
598     printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
599     }
600     }
601    
602 dpavlin 54 # now read database
603     print STDERR "using: $type...\n";
604 dpavlin 1
605 dpavlin 180 # erase cache for tags by order in this database
606     delete $cache->{tags_by_order};
607    
608 dpavlin 58 if ($type_base eq "isis") {
609    
610 dpavlin 54 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
611 dpavlin 1
612 dpavlin 54 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
613     my $db = OpenIsis::open( $isis_db );
614 dpavlin 10
615 dpavlin 106 # check if .txt database for OpenIsis is zero length,
616     # if so, erase it and re-open database
617     sub check_txt_db {
618     my $isis_db = shift || die "need isis database name";
619 dpavlin 109 my $reopen = 0;
620    
621 dpavlin 106 if (-e $isis_db.".TXT") {
622 dpavlin 109 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
623 dpavlin 108 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
624 dpavlin 109 $reopen++;
625 dpavlin 106 }
626 dpavlin 109 if (-e $isis_db.".PTR") {
627     print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
628     unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
629     $reopen++;
630     }
631     return OpenIsis::open( $isis_db ) if ($reopen);
632 dpavlin 106 }
633    
634     # EOF error
635     if ($db == -1) {
636     $db = check_txt_db($isis_db);
637 dpavlin 108 if ($db == -1) {
638 dpavlin 106 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
639     next;
640     }
641     }
642    
643 dpavlin 101 # OpenIsis::ERR_BADF
644     if ($db == -4) {
645     print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
646     next;
647     # OpenIsis::ERR_IO
648     } elsif ($db == -5) {
649     print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
650     next;
651     } elsif ($db < 0) {
652     print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
653     next;
654     }
655    
656 dpavlin 54 my $max_rowid = OpenIsis::maxRowid( $db );
657 dpavlin 3
658 dpavlin 106 # if 0 records, try to rease isis .txt database
659     if ($max_rowid == 0) {
660     # force removal of database
661     $db = check_txt_db($isis_db);
662     $max_rowid = OpenIsis::maxRowid( $db );
663     }
664    
665 dpavlin 54 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
666    
667     my $path = $database;
668    
669     for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
670     my $row = OpenIsis::read( $db, $row_id );
671     if ($row && $row->{mfn}) {
672    
673     progress($row->{mfn}, $max_rowid);
674    
675     my $swishpath = $path."#".int($row->{mfn});
676    
677 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
678 dpavlin 54 $xml = $cp2utf->convert($xml);
679     use bytes; # as opposed to chars
680     print "Path-Name: $swishpath\n";
681     print "Content-Length: ".(length($xml)+1)."\n";
682     print "Document-Type: XML\n\n$xml\n";
683     }
684 dpavlin 3 }
685 dpavlin 54 }
686 dpavlin 108 # for this to work with current version of OpenIsis (0.9.0)
687     # you might need my patch from
688     # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
689     OpenIsis::close($db);
690 dpavlin 54 print STDERR "\n";
691 dpavlin 3
692 dpavlin 58 } elsif ($type_base eq "excel") {
693 dpavlin 54 use Spreadsheet::ParseExcel;
694     use Spreadsheet::ParseExcel::Utility qw(int2col);
695    
696     $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
697     my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
698 dpavlin 43
699 dpavlin 54 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
700 dpavlin 56 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
701 dpavlin 54
702     my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
703    
704     my $sheet_nr = 0;
705     foreach my $oWks (@{$oBook->{Worksheet}}) {
706     #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
707     last if ($oWks->{Name} eq $sheet);
708     $sheet_nr++;
709     }
710    
711     my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
712    
713     print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
714     defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
715     my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
716    
717 dpavlin 56 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
718 dpavlin 54 my $row;
719     for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
720     my $cell = $oWorksheet->{Cells}[$iR][$iC];
721     if ($cell) {
722     $row->{int2col($iC)} = $cell->Value;
723     }
724     }
725    
726     progress($iR, $end_row);
727    
728     # print "row[$iR/$end_row] ";
729     # foreach (keys %{$row}) {
730     # print "$_: ",$row->{$_},"\t";
731     # }
732     # print "\n";
733    
734     my $swishpath = $database."#".$iR;
735    
736     next if (! $row);
737    
738 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
739 dpavlin 44 $xml = $cp2utf->convert($xml);
740 dpavlin 35 use bytes; # as opposed to chars
741 dpavlin 43 print "Path-Name: $swishpath\n";
742 dpavlin 10 print "Content-Length: ".(length($xml)+1)."\n";
743     print "Document-Type: XML\n\n$xml\n";
744 dpavlin 3 }
745 dpavlin 1 }
746 dpavlin 62 } elsif ($type_base eq "marc") {
747 dpavlin 67
748 dpavlin 62 use MARC;
749    
750     $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
751     my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
752    
753     # optional argument is format
754     my $format = x($config->{format}) || 'usmarc';
755    
756     print STDERR "Reading MARC file '$marc_file'\n";
757    
758     my $marc = new MARC;
759     my $nr = $marc->openmarc({
760     file=>$marc_file, format=>$format
761     }) || die "Can't open MARC file '$marc_file'";
762    
763     my $i=0; # record nr.
764    
765     my $rec;
766    
767     while ($marc->nextmarc(1)) {
768    
769     # XXX
770 dpavlin 67 fakeprogress($i++);
771 dpavlin 62
772     my $swishpath = $database."#".$i;
773    
774     if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
775     $xml = $cp2utf->convert($xml);
776     use bytes; # as opposed to chars
777     print "Path-Name: $swishpath\n";
778     print "Content-Length: ".(length($xml)+1)."\n";
779     print "Document-Type: XML\n\n$xml\n";
780     }
781     }
782 dpavlin 67 } elsif ($type_base eq "feed") {
783    
784     $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
785     my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
786    
787     print STDERR "Reading feed from program '$prog'\n";
788    
789     open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
790    
791 dpavlin 74 my $i=1; # record nr.
792 dpavlin 67
793     my $data;
794     my $line=1;
795    
796     while (<FEED>) {
797     chomp;
798    
799     if (/^$/) {
800 dpavlin 74 my $swishpath = $database."#".$i++;
801 dpavlin 67
802     if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
803     $xml = $cp2utf->convert($xml);
804     use bytes; # as opposed to chars
805     print "Path-Name: $swishpath\n";
806     print "Content-Length: ".(length($xml)+1)."\n";
807     print "Document-Type: XML\n\n$xml\n";
808     }
809     $line = 1;
810     $data = {};
811     next;
812     }
813    
814 dpavlin 74 $line = $1 if (s/^(\d+):\s*//);
815 dpavlin 67 $data->{$line++} = $_;
816    
817 dpavlin 74 fakeprogress($i);
818 dpavlin 67
819     }
820 dpavlin 177 # close lookup
821     untie %lhash if (%lhash);
822 dpavlin 1 }
823     }
824 dpavlin 3
825 dpavlin 10 # call this to commit index
826     $index->close;
827 dpavlin 3
828     1;
829     __END__
830     ##########################################################################
831    
832     =head1 NAME
833    
834 dpavlin 81 all2xml.pl - read various file formats and dump XML for SWISH-E
835 dpavlin 3
836     =head1 DESCRIPTION
837    
838 dpavlin 81 This command will read ISIS data file using OpenIsis perl module, MARC
839     records using MARC module and optionally Micro$oft Excel files to
840     create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
841     this script B<isn't general xml generator> from isis files (isis allready
842     has something like that). Output of this script is tailor-made for SWISH-E.
843 dpavlin 3
844 dpavlin 81 =head1 BUGS
845    
846     Documentation is really lacking. However, in true Open Source spirit, source
847     is best documentation. I even made considerable effort to comment parts
848     which are not intuitively clear, so...
849    
850 dpavlin 3 =head1 AUTHOR
851    
852     Dobrica Pavlinusic <dpavlin@rot13.org>
853    
854     =head1 COPYRIGHT
855    
856     GNU Public License (GPL) v2 or later
857    
858     =head1 SEE ALSO
859    
860     SWISH-E web site at http://www.swish-e.org
861    
862     =cut

Properties

Name Value
cvs2svn:cvs-rev 1.53
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26