9 |
use Text::Iconv; |
use Text::Iconv; |
10 |
use Config::IniFiles; |
use Config::IniFiles; |
11 |
use Encode; |
use Encode; |
12 |
|
#use GDBM_File; |
13 |
|
use Fcntl; # for O_RDWR |
14 |
|
use TDB_File; |
15 |
|
|
16 |
$|=1; |
$|=1; |
17 |
|
|
64 |
'feed' => 'feed' |
'feed' => 'feed' |
65 |
); |
); |
66 |
|
|
67 |
|
my $cache; # for cacheing |
68 |
|
|
69 |
|
# 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 |
sub data2xml { |
sub data2xml { |
76 |
|
|
77 |
use xmlify; |
use xmlify; |
101 |
return $va <=> $vb; |
return $va <=> $vb; |
102 |
} |
} |
103 |
|
|
104 |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
my @sorted_tags; |
105 |
|
if ($cache->{tags_by_order}) { |
106 |
|
@sorted_tags = @{$cache->{tags_by_order}}; |
107 |
|
} else { |
108 |
|
@sorted_tags = sort by_order keys %{$config->{indexer}}; |
109 |
|
$cache->{tags_by_order} = \@sorted_tags; |
110 |
|
} |
111 |
|
|
112 |
|
# lookup key |
113 |
|
my $lookup_key; |
114 |
|
|
115 |
|
# cache for field in pages |
116 |
|
delete $cache->{display_data}; |
117 |
|
delete $cache->{swish_data}; |
118 |
|
delete $cache->{swish_exact_data}; |
119 |
|
delete $cache->{index_data}; |
120 |
|
my @page_fields; # names of fields |
121 |
|
|
122 |
|
|
123 |
|
# subs used to produce output |
124 |
|
|
125 |
|
sub get_field_name($$$) { |
126 |
|
my ($config,$field,$field_usage) = @_; |
127 |
|
|
128 |
|
# find field name (signular, plural) |
129 |
|
my $field_name = ""; |
130 |
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) { |
131 |
|
$field_name = $config->{indexer}->{$field}->{name_singular}; |
132 |
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
133 |
|
$field_name = $config->{indexer}->{$field}->{name_plural}; |
134 |
|
} elsif ($config->{indexer}->{$field}->{name}) { |
135 |
|
$field_name = $config->{indexer}->{$field}->{name}; |
136 |
|
} else { |
137 |
|
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
138 |
|
} |
139 |
|
if ($field_name) { |
140 |
|
return x($field_name); |
141 |
|
} |
142 |
|
} |
143 |
|
|
144 |
|
|
145 |
|
# begin real work: go field by field |
146 |
|
foreach my $field (@sorted_tags) { |
147 |
|
|
148 |
$field=x($field); |
$field=x($field); |
149 |
$field_usage{$field}++; |
$field_usage{$field}++; |
151 |
my $swish_data = ""; |
my $swish_data = ""; |
152 |
my $swish_exact_data = ""; |
my $swish_exact_data = ""; |
153 |
my $display_data = ""; |
my $display_data = ""; |
154 |
|
my @index_data; |
155 |
my $line_delimiter; |
my $line_delimiter; |
156 |
|
|
157 |
my ($swish,$display); |
my ($swish,$display); |
158 |
|
|
159 |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
160 |
|
|
161 |
|
# is this field page-by-page? |
162 |
|
my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page}; |
163 |
|
push @page_fields,$field if ($iterate_by_page); |
164 |
|
my %page_max = (); |
165 |
|
# default line_delimiter if using |
166 |
|
my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>'; |
167 |
|
|
168 |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
169 |
|
|
170 |
my $format = x($x->{content}); |
my $format = x($x->{content}); |
171 |
my $delimiter = x($x->{delimiter}) || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
172 |
|
|
173 |
my $repeat_off = 0; # repeatable offset |
my $repeat_off = 0; # init repeatable offset |
174 |
|
|
175 |
my ($s,$se,$d,$i) = (1,0,1,0); # swish, display default |
# swish, swish_exact, display, index, index_lookup |
176 |
|
# swish and display defaults |
177 |
|
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
178 |
$s = 0 if (lc($x->{type}) eq "display"); |
$s = 0 if (lc($x->{type}) eq "display"); |
179 |
$d = 0 if (lc($x->{type}) eq "swish"); |
$d = 0 if (lc($x->{type}) eq "swish"); |
180 |
$se = 1 if (lc($x->{type}) eq "swish_exact"); |
($s,$se,$d,$i) = (0,0,0,1) if (lc($x->{type}) eq "index"); |
181 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact"); |
182 |
|
($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/); |
183 |
|
|
184 |
# what will separate last line from this one? |
# what will separate last line from this one? |
185 |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
if ($display_data && $x->{append}) { |
186 |
$line_delimiter = ' '; |
$line_delimiter = ' '; |
187 |
} elsif ($display_data) { |
} elsif ($display_data) { |
188 |
$line_delimiter = '<br/>'; |
$line_delimiter = '<br/>'; |
192 |
($swish,$display) = (1,1); |
($swish,$display) = (1,1); |
193 |
|
|
194 |
# placeholder for all repeatable entries for index |
# placeholder for all repeatable entries for index |
|
my @index_data; |
|
|
my $index_filter; |
|
195 |
|
|
196 |
sub mkformat { |
sub mkformat { |
197 |
my $x = shift || die "mkformat needs tag reference"; |
my $x = shift || die "mkformat needs tag reference"; |
221 |
|
|
222 |
# while because of repeatable fields |
# while because of repeatable fields |
223 |
while ($swish || $display) { |
while ($swish || $display) { |
224 |
|
my $page = $repeat_off; |
225 |
|
$page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0)); |
226 |
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
227 |
if ($repeat_off > 1000) { |
if ($repeat_off > 1000) { |
228 |
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n"; |
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n"; |
229 |
last; |
last; |
230 |
} |
} |
231 |
|
|
232 |
|
# is this field is lookup? |
233 |
|
if ($display && $x->{lookup}) { |
234 |
|
my $null = "<!-- null -->"; |
235 |
|
if ($use_lhash_cache) { |
236 |
|
if (!defined($cache->{lhash}->{$display})) { |
237 |
|
my $new_display = $lhash{$display}; |
238 |
|
if (defined($new_display)) { |
239 |
|
#print STDERR "lookup cache store '$display' = '$new_display'\n"; |
240 |
|
$display = $new_display; |
241 |
|
$cache->{lhash}->{$display} = $new_display; |
242 |
|
} else { |
243 |
|
print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
244 |
|
$display = ""; |
245 |
|
$cache->{lhash}->{$display} = $null; |
246 |
|
} |
247 |
|
} else { |
248 |
|
$display = $cache->{lhash}->{$display}; |
249 |
|
} |
250 |
|
} else { |
251 |
|
$display = $lhash{$display} || $null; |
252 |
|
} |
253 |
|
} |
254 |
|
|
255 |
# filter="name" ; filter this field through |
# filter="name" ; filter this field through |
256 |
# filter/[name].pm |
# filter/[name].pm |
257 |
my $filter = $x->{filter}; |
my $filter = $x->{filter}; |
258 |
if ($filter) { |
if ($filter && !$cache->{filter_loaded}->{$filter}) { |
259 |
require "filter/".$filter.".pm"; |
require "filter/".$filter.".pm"; |
260 |
|
$cache->{filter_loaded}->{$filter}++; |
261 |
} |
} |
262 |
# type="swish" ; field for swish |
# type="swish" ; field for swish |
263 |
if ($swish) { |
if ($swish) { |
277 |
if ($d && $display) { |
if ($d && $display) { |
278 |
if ($line_delimiter && $display_data) { |
if ($line_delimiter && $display_data) { |
279 |
$display_data .= $line_delimiter; |
$display_data .= $line_delimiter; |
|
undef $line_delimiter; |
|
280 |
} |
} |
281 |
if ($filter) { |
if ($filter) { |
282 |
no strict 'refs'; |
no strict 'refs'; |
295 |
} |
} |
296 |
|
|
297 |
# type="index" ; insert into index |
# type="index" ; insert into index |
298 |
|
my $idisplay; |
299 |
if ($i && $display) { |
if ($i && $display) { |
300 |
push @index_data, $display; |
$idisplay = $display; |
301 |
$index_filter = $filter if ($filter); |
if ($filter) { |
302 |
|
no strict 'refs'; |
303 |
|
$idisplay = &$filter($idisplay); |
304 |
|
} |
305 |
|
push @index_data, $idisplay if (! $iterate_by_page); |
306 |
} |
} |
|
} |
|
307 |
|
|
308 |
# fill data in index |
# store fields in lookup |
309 |
if (@index_data) { |
if ($il && $display) { |
310 |
if ($index_filter) { |
if (lc($x->{type}) eq "lookup_key") { |
311 |
no strict 'refs'; |
if ($lookup_key) { |
312 |
foreach my $d (@index_data) { |
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
313 |
$index->insert($field, &$index_filter($d), $path); |
} else { |
314 |
|
$lookup_key = $display; |
315 |
|
} |
316 |
|
} elsif (lc($x->{type}) eq "lookup_val") { |
317 |
|
if ($lookup_key) { |
318 |
|
$lhash{$lookup_key} = $display; |
319 |
|
} else { |
320 |
|
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
321 |
|
} |
322 |
} |
} |
323 |
} else { |
|
324 |
foreach my $d (@index_data) { |
} |
325 |
$index->insert($field, $d, $path); |
|
326 |
|
# store data for page-by-page repeatable fields |
327 |
|
if ($iterate_by_page) { |
328 |
|
sub iterate_fld($$$$$$) { |
329 |
|
my ($cache,$what,$field,$page,$data,$append) = @_; |
330 |
|
return if (!$data); |
331 |
|
|
332 |
|
my $ldel = $page_line_delimiter; |
333 |
|
$ldel = " " if ($append); |
334 |
|
#print STDERR "line delimiter: ",Dumper($ldel) if ($ldel); |
335 |
|
if (! $cache->{$what}->{$field}->[$page]) { |
336 |
|
$cache->{$what}->{$field}->[$page] = $data; |
337 |
|
} else { |
338 |
|
$cache->{$what}->{$field}->[$page] .= $ldel.$data; |
339 |
|
} |
340 |
|
} |
341 |
|
|
342 |
|
if ($display_data) { |
343 |
|
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
344 |
|
} |
345 |
|
$display_data = ""; |
346 |
|
if ($swish_data) { |
347 |
|
iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append}); |
348 |
|
$swish_data = ""; |
349 |
|
} |
350 |
|
if ($swish_exact_data) { |
351 |
|
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
352 |
|
$swish_exact_data = ""; |
353 |
|
} |
354 |
|
|
355 |
|
if ($idisplay) { |
356 |
|
my $ldel=$page_line_delimiter; |
357 |
|
my @index_data; |
358 |
|
if ($cache->{index_data}->{$field}->[$page]) { |
359 |
|
|
360 |
|
@index_data = @{$cache->{index_data}->{$field}->[$page]}; |
361 |
|
} |
362 |
|
if ($x->{append}) { |
363 |
|
$index_data[$#index_data] .= $idisplay; |
364 |
|
} else { |
365 |
|
push @index_data, $idisplay; |
366 |
|
} |
367 |
|
$idisplay = ""; |
368 |
|
@{$cache->{index_data}->{$field}->[$page]} = @index_data; |
369 |
} |
} |
370 |
} |
} |
371 |
} |
} |
372 |
|
|
373 |
|
if (! $iterate_by_page) { |
374 |
|
# fill data in index |
375 |
|
foreach my $d (@index_data) { |
376 |
|
$index->insert($field, $d, $path); |
377 |
|
} |
378 |
|
@index_data = (); |
379 |
|
} |
380 |
} |
} |
381 |
|
|
382 |
# now try to parse variables from configuration file |
# now try to parse variables from configuration file |
398 |
$index->insert($field, $val, $path) if ($i); |
$index->insert($field, $val, $path) if ($i); |
399 |
} |
} |
400 |
|
|
401 |
|
if ($iterate_by_page) { |
402 |
|
# FIXME data from config tag will appear just |
403 |
|
# on first page!!! |
404 |
|
my $page = 0; |
405 |
|
if ($display_data) { |
406 |
|
$cache->{display_data}->{$field}->[$page] = $display_data; |
407 |
|
$display_data = ""; |
408 |
|
} |
409 |
|
if ($swish_data) { |
410 |
|
$cache->{swish_data}->{$field}->[$page] = $swish_data; |
411 |
|
$swish_data = ""; |
412 |
|
} |
413 |
|
if ($swish_exact_data) { |
414 |
|
$cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data; |
415 |
|
$swish_exact_data = ""; |
416 |
|
} |
417 |
|
} |
418 |
} |
} |
419 |
|
|
420 |
|
# save data page-by-page |
421 |
|
foreach my $field (@page_fields) { |
422 |
|
my $nr_pages = $page_max{$field} || next; |
423 |
|
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
424 |
|
#print STDERR Dumper($cache->{display_data}); |
425 |
|
for (my $page=0; $page <= $nr_pages; $page++) { |
426 |
|
|
427 |
|
my $display_data = $cache->{display_data}->{$field}->[$page]; |
428 |
|
if ($display_data) { # default |
429 |
|
if ($field eq "headline") { |
430 |
|
$xml .= xmlify("headline", $display_data); |
431 |
|
} else { |
432 |
|
|
433 |
if ($display_data) { |
# fallback to empty field name if needed |
434 |
|
$html .= get_field_name($config,$field,$field_usage{$field}) || ''; |
435 |
if ($field eq "headline") { |
$html .= "#-#".$display_data."###\n"; |
436 |
$xml .= xmlify("headline", $display_data); |
} |
437 |
} else { |
} |
438 |
|
|
439 |
# find field name (signular, plural) |
my $swish_data = $cache->{swish_data}->{$field}->[$page]; |
440 |
my $field_name = ""; |
if ($swish_data) { |
441 |
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
# remove extra spaces |
442 |
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
$swish_data =~ s/ +/ /g; |
443 |
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
$swish_data =~ s/ +$//g; |
444 |
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
|
445 |
} elsif ($config->{indexer}->{$field}->{name}) { |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
446 |
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
} |
447 |
} else { |
|
448 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
449 |
|
if ($swish_exact_data) { |
450 |
|
$swish_exact_data =~ s/ +/ /g; |
451 |
|
$swish_exact_data =~ s/ +$//g; |
452 |
|
|
453 |
|
# add delimiters before and after word. |
454 |
|
# That is required to produce exact match |
455 |
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
456 |
} |
} |
457 |
if ($field_name) { |
|
458 |
$html .= x($field_name); |
foreach my $d (@{$cache->{index_data}->{$field}->[$page]}) { |
459 |
|
$index->insert($field, $d, $path); |
460 |
|
#print STDERR "index $field: $d [$path]\n"; |
461 |
} |
} |
|
$html .= $display_data."###\n"; |
|
462 |
} |
} |
463 |
|
|
464 |
} |
} |
465 |
if ($swish_data) { |
|
466 |
# remove extra spaces |
if (! $iterate_by_page) { |
467 |
$swish_data =~ s/ +/ /g; |
if ($display_data) { |
468 |
$swish_data =~ s/ +$//g; |
if ($field eq "headline") { |
469 |
|
$xml .= xmlify("headline", $display_data); |
470 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
} else { |
|
} |
|
471 |
|
|
472 |
if ($swish_exact_data) { |
# fallback to empty field name if needed |
473 |
$swish_exact_data =~ s/ +/ /g; |
$html .= get_field_name($config,$field,$field_usage{$field}) || ''; |
474 |
$swish_exact_data =~ s/ +$//g; |
$html .= "#-#".$display_data."###\n"; |
475 |
|
} |
476 |
|
} |
477 |
|
if ($swish_data) { |
478 |
|
# remove extra spaces |
479 |
|
$swish_data =~ s/ +/ /g; |
480 |
|
$swish_data =~ s/ +$//g; |
481 |
|
|
482 |
# add delimiters before and after word. |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
483 |
# That is required to produce exact match |
} |
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
|
|
} |
|
484 |
|
|
485 |
|
if ($swish_exact_data) { |
486 |
|
$swish_exact_data =~ s/ +/ /g; |
487 |
|
$swish_exact_data =~ s/ +$//g; |
488 |
|
|
489 |
|
# add delimiters before and after word. |
490 |
|
# That is required to produce exact match |
491 |
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
492 |
|
} |
493 |
|
} |
494 |
} |
} |
495 |
|
|
496 |
# dump formatted output in <html> |
# dump formatted output in <html> |
525 |
|
|
526 |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
527 |
|
|
528 |
|
my $unac_filter = $cfg_global->val('global', 'unac_filter'); |
529 |
|
if ($unac_filter) { |
530 |
|
require $unac_filter; |
531 |
|
} |
532 |
|
|
533 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
534 |
|
|
535 |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
536 |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
537 |
|
|
538 |
|
# create new lookup file |
539 |
|
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
540 |
|
if ($lookup_file) { |
541 |
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
542 |
|
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
543 |
|
print STDERR "creating lookup file '$lookup_file'\n"; |
544 |
|
} |
545 |
|
|
546 |
|
# open existing lookup file |
547 |
|
$lookup_file = $cfg -> val($database, 'lookup_open'); # optional |
548 |
|
if ($lookup_file) { |
549 |
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644; |
550 |
|
tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644; |
551 |
|
print STDERR "opening lookup file '$lookup_file'\n"; |
552 |
|
} |
553 |
|
|
554 |
print STDERR "reading ./import_xml/$type.xml\n"; |
print STDERR "reading ./import_xml/$type.xml\n"; |
555 |
|
|
556 |
# extract just type basic |
# extract just type basic |
589 |
# now read database |
# now read database |
590 |
print STDERR "using: $type...\n"; |
print STDERR "using: $type...\n"; |
591 |
|
|
592 |
|
# erase cache for tags by order in this database |
593 |
|
delete $cache->{tags_by_order}; |
594 |
|
|
595 |
if ($type_base eq "isis") { |
if ($type_base eq "isis") { |
596 |
|
|
597 |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
804 |
fakeprogress($i); |
fakeprogress($i); |
805 |
|
|
806 |
} |
} |
807 |
|
# close lookup |
808 |
|
untie %lhash if (%lhash); |
809 |
} |
} |
810 |
} |
} |
811 |
|
|