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 |
|
|
66 |
|
|
67 |
my $cache; # for cacheing |
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 |
|
my $last_field_name; # cache to prevent repeated fields |
76 |
|
|
77 |
sub data2xml { |
sub data2xml { |
78 |
|
|
79 |
use xmlify; |
use xmlify; |
104 |
} |
} |
105 |
|
|
106 |
my @sorted_tags; |
my @sorted_tags; |
107 |
if ($cache->{tags_by_order}->{$type}) { |
if ($cache->{tags_by_order}) { |
108 |
@sorted_tags = @{$cache->{tags_by_order}->{$type}}; |
@sorted_tags = @{$cache->{tags_by_order}}; |
109 |
} else { |
} else { |
110 |
@sorted_tags = sort by_order keys %{$config->{indexer}}; |
@sorted_tags = sort by_order keys %{$config->{indexer}}; |
111 |
$cache->{tags_by_order}->{$type} = \@sorted_tags; |
$cache->{tags_by_order} = \@sorted_tags; |
112 |
|
} |
113 |
|
|
114 |
|
# lookup key |
115 |
|
my $lookup_key; |
116 |
|
|
117 |
|
# cache for field in pages |
118 |
|
delete $cache->{display_data}; |
119 |
|
delete $cache->{swish_data}; |
120 |
|
delete $cache->{swish_exact_data}; |
121 |
|
delete $cache->{index_data}; |
122 |
|
delete $cache->{index_delimiter}; |
123 |
|
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 |
|
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 |
|
} |
151 |
} |
} |
152 |
|
|
153 |
|
|
154 |
|
# begin real work: go field by field |
155 |
foreach my $field (@sorted_tags) { |
foreach my $field (@sorted_tags) { |
156 |
|
|
157 |
$field=x($field); |
$field=x($field); |
160 |
my $swish_data = ""; |
my $swish_data = ""; |
161 |
my $swish_exact_data = ""; |
my $swish_exact_data = ""; |
162 |
my $display_data = ""; |
my $display_data = ""; |
163 |
|
my @index_data; |
164 |
my $line_delimiter; |
my $line_delimiter; |
165 |
|
|
166 |
my ($swish,$display); |
my ($swish,$display); |
167 |
|
|
168 |
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"; |
169 |
|
|
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 |
|
$cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter}; |
177 |
|
|
178 |
|
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 |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
186 |
|
|
187 |
my $format = x($x->{content}); |
my $format = x($x->{content}); |
188 |
my $delimiter = x($x->{delimiter}) || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
189 |
|
|
190 |
my $repeat_off = 0; # repeatable offset |
my $repeat_off = 0; # init repeatable offset |
191 |
|
|
192 |
my ($s,$se,$d,$i) = (1,0,1,0); # swish, display default |
# 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 |
$s = 0 if (lc($x->{type}) eq "display"); |
$s = 0 if (lc($x->{type}) eq "display"); |
196 |
$d = 0 if (lc($x->{type}) eq "swish"); |
$d = 0 if (lc($x->{type}) eq "swish"); |
197 |
$se = 1 if (lc($x->{type}) eq "swish_exact"); |
($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index"); |
198 |
($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"); |
199 |
|
($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/); |
200 |
|
|
201 |
# what will separate last line from this one? |
# what will separate last line from this one? |
202 |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
if ($display_data && $x->{append}) { |
203 |
$line_delimiter = ' '; |
$line_delimiter = ' '; |
204 |
} elsif ($display_data) { |
} elsif ($display_data) { |
205 |
$line_delimiter = '<br/>'; |
$line_delimiter = '<br/>'; |
209 |
($swish,$display) = (1,1); |
($swish,$display) = (1,1); |
210 |
|
|
211 |
# placeholder for all repeatable entries for index |
# placeholder for all repeatable entries for index |
|
my @index_data; |
|
|
my $index_filter; |
|
212 |
|
|
213 |
sub mkformat { |
sub mkformat($$) { |
214 |
my $x = shift || die "mkformat needs tag reference"; |
my $x = shift || die "mkformat needs tag reference"; |
215 |
my $data = shift || return; |
my $data = shift || return; |
216 |
my $format_name = x($x->{format_name}) || return $data; |
my $format_name = x($x->{format_name}) || return $data; |
238 |
|
|
239 |
# while because of repeatable fields |
# while because of repeatable fields |
240 |
while ($swish || $display) { |
while ($swish || $display) { |
241 |
|
my $page = $repeat_off; |
242 |
|
$page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0)); |
243 |
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
244 |
if ($repeat_off > 1000) { |
if ($repeat_off > 1000) { |
245 |
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n"; |
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n"; |
246 |
last; |
last; |
247 |
} |
} |
248 |
|
|
249 |
|
# is this field is lookup? |
250 |
|
if ($display && $x->{lookup}) { |
251 |
|
my $null = "<!-- null -->"; |
252 |
|
if ($use_lhash_cache) { |
253 |
|
if (!defined($cache->{lhash}->{$display})) { |
254 |
|
my $new_display = $lhash{$display}; |
255 |
|
if (defined($new_display)) { |
256 |
|
#print STDERR "lookup cache store '$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 |
|
$cache->{lhash}->{$display} = $null; |
263 |
|
} |
264 |
|
} else { |
265 |
|
$display = $cache->{lhash}->{$display}; |
266 |
|
} |
267 |
|
} else { |
268 |
|
$display = $lhash{$display} || $null; |
269 |
|
} |
270 |
|
} |
271 |
|
|
272 |
# filter="name" ; filter this field through |
# filter="name" ; filter this field through |
273 |
# filter/[name].pm |
# filter/[name].pm |
274 |
my $filter = $x->{filter}; |
my $filter = $x->{filter}; |
294 |
if ($d && $display) { |
if ($d && $display) { |
295 |
if ($line_delimiter && $display_data) { |
if ($line_delimiter && $display_data) { |
296 |
$display_data .= $line_delimiter; |
$display_data .= $line_delimiter; |
|
undef $line_delimiter; |
|
297 |
} |
} |
298 |
if ($filter) { |
if ($filter) { |
299 |
no strict 'refs'; |
no strict 'refs'; |
312 |
} |
} |
313 |
|
|
314 |
# type="index" ; insert into index |
# type="index" ; insert into index |
315 |
|
my $idisplay; |
316 |
if ($i && $display) { |
if ($i && $display) { |
317 |
push @index_data, $display; |
$idisplay = $display; |
318 |
$index_filter = $filter if ($filter); |
if ($filter) { |
319 |
|
no strict 'refs'; |
320 |
|
$idisplay = &$filter($idisplay); |
321 |
|
} |
322 |
|
push @index_data, $idisplay if (! $iterate_by_page); |
323 |
} |
} |
|
} |
|
324 |
|
|
325 |
# fill data in index |
# store fields in lookup |
326 |
if (@index_data) { |
if ($il && $display) { |
327 |
if ($index_filter) { |
if (lc($x->{type}) eq "lookup_key") { |
328 |
no strict 'refs'; |
if ($lookup_key) { |
329 |
foreach my $d (@index_data) { |
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
330 |
$index->insert($field, &$index_filter($d), $path); |
} else { |
331 |
|
$lookup_key = $display; |
332 |
|
} |
333 |
|
} elsif (lc($x->{type}) eq "lookup_val") { |
334 |
|
if ($lookup_key) { |
335 |
|
$lhash{$lookup_key} = $display; |
336 |
|
} else { |
337 |
|
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
338 |
|
} |
339 |
} |
} |
340 |
} else { |
|
341 |
foreach my $d (@index_data) { |
} |
342 |
$index->insert($field, $d, $path); |
|
343 |
|
# store data for page-by-page repeatable fields |
344 |
|
if ($iterate_by_page) { |
345 |
|
sub iterate_fld($$$$$$) { |
346 |
|
my ($cache,$what,$field,$page,$data,$append) = @_; |
347 |
|
return if (!$data); |
348 |
|
|
349 |
|
my $ldel = $page_line_delimiter; |
350 |
|
$ldel = " " if ($append); |
351 |
|
#print STDERR "line delimiter: ",Dumper($ldel) if ($ldel); |
352 |
|
if (! $cache->{$what}->{$field}->[$page]) { |
353 |
|
$cache->{$what}->{$field}->[$page] = $data; |
354 |
|
} else { |
355 |
|
$cache->{$what}->{$field}->[$page] .= $ldel.$data; |
356 |
|
} |
357 |
|
} |
358 |
|
|
359 |
|
if ($display_data) { |
360 |
|
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
361 |
|
} |
362 |
|
$display_data = ""; |
363 |
|
if ($swish_data) { |
364 |
|
iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append}); |
365 |
|
$swish_data = ""; |
366 |
|
} |
367 |
|
if ($swish_exact_data) { |
368 |
|
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
369 |
|
$swish_exact_data = ""; |
370 |
} |
} |
371 |
|
|
372 |
|
if ($idisplay) { |
373 |
|
my $ldel=$page_line_delimiter; |
374 |
|
my @index_data; |
375 |
|
if ($cache->{index_data}->{$field}->[$page]) { |
376 |
|
|
377 |
|
@index_data = @{$cache->{index_data}->{$field}->[$page]}; |
378 |
|
} |
379 |
|
if ($x->{append}) { |
380 |
|
if (@index_data) { |
381 |
|
$index_data[$#index_data] .= $idisplay; |
382 |
|
} else { |
383 |
|
push @index_data, $idisplay; |
384 |
|
} |
385 |
|
} else { |
386 |
|
push @index_data, $idisplay; |
387 |
|
} |
388 |
|
$idisplay = ""; |
389 |
|
@{$cache->{index_data}->{$field}->[$page]} = @index_data; |
390 |
|
} |
391 |
|
} |
392 |
|
} |
393 |
|
|
394 |
|
if (! $iterate_by_page) { |
395 |
|
my $idel = $x->{index_delimiter}; |
396 |
|
# fill data in index |
397 |
|
foreach my $tmp (@index_data) { |
398 |
|
my $i = $d = $tmp; |
399 |
|
if ($idel && $tmp =~ m/$idel/) { |
400 |
|
($i,$d) = split(/$idel/,$tmp); |
401 |
|
} |
402 |
|
$index->insert($field, $i, $d, $path); |
403 |
} |
} |
404 |
|
@index_data = (); |
405 |
} |
} |
406 |
} |
} |
407 |
|
|
424 |
$index->insert($field, $val, $path) if ($i); |
$index->insert($field, $val, $path) if ($i); |
425 |
} |
} |
426 |
|
|
427 |
|
if ($iterate_by_page) { |
428 |
|
# FIXME data from config tag will appear just |
429 |
|
# on first page!!! |
430 |
|
my $page = 0; |
431 |
|
if ($display_data) { |
432 |
|
$cache->{display_data}->{$field}->[$page] = $display_data; |
433 |
|
$display_data = ""; |
434 |
|
} |
435 |
|
if ($swish_data) { |
436 |
|
$cache->{swish_data}->{$field}->[$page] = $swish_data; |
437 |
|
$swish_data = ""; |
438 |
|
} |
439 |
|
if ($swish_exact_data) { |
440 |
|
$cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data; |
441 |
|
$swish_exact_data = ""; |
442 |
|
} |
443 |
|
} |
444 |
} |
} |
445 |
|
|
446 |
|
# save data page-by-page |
447 |
if ($display_data) { |
foreach my $field (@page_fields) { |
448 |
|
my $nr_pages = $page_max{$field} || next; |
449 |
if ($field eq "headline") { |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
450 |
$xml .= xmlify("headline", $display_data); |
#print STDERR Dumper($cache->{display_data}); |
451 |
} else { |
for (my $page=0; $page <= $nr_pages; $page++) { |
452 |
|
my $display_data; |
453 |
# find field name (signular, plural) |
if ($cache->{format}->{$field}) { |
454 |
my $field_name = ""; |
$display_data=mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]); |
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
|
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
|
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
|
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
|
|
} elsif ($config->{indexer}->{$field}->{name}) { |
|
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
|
455 |
} else { |
} else { |
456 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
$display_data = $cache->{display_data}->{$field}->[$page]; |
457 |
|
} |
458 |
|
if ($display_data) { # default |
459 |
|
if ($field eq "headline") { |
460 |
|
$xml .= xmlify("headline", $display_data); |
461 |
|
} else { |
462 |
|
|
463 |
|
# fallback to empty field name if needed |
464 |
|
$html .= get_field_name($config,$field,$field_usage{$field}) || ''; |
465 |
|
$html .= "#-#".$display_data."###\n"; |
466 |
|
} |
467 |
|
} |
468 |
|
|
469 |
|
my $swish_data = $cache->{swish_data}->{$field}->[$page]; |
470 |
|
if ($swish_data) { |
471 |
|
# remove extra spaces |
472 |
|
$swish_data =~ s/ +/ /g; |
473 |
|
$swish_data =~ s/ +$//g; |
474 |
|
|
475 |
|
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
476 |
|
} |
477 |
|
|
478 |
|
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
479 |
|
if ($swish_exact_data) { |
480 |
|
$swish_exact_data =~ s/ +/ /g; |
481 |
|
$swish_exact_data =~ s/ +$//g; |
482 |
|
|
483 |
|
# add delimiters before and after word. |
484 |
|
# That is required to produce exact match |
485 |
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
486 |
} |
} |
487 |
if ($field_name) { |
|
488 |
$html .= x($field_name); |
my $idel = $cache->{index_delimiter}->{$field}; |
489 |
|
foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) { |
490 |
|
my $i = $tmp; |
491 |
|
my $d = $tmp; |
492 |
|
if ($idel && $tmp =~ m/$idel/) { |
493 |
|
($i,$d) = split(/$idel/,$tmp); |
494 |
|
} |
495 |
|
$index->insert($field, $i, $d, $path); |
496 |
|
#print STDERR "index [$idel] $field: $i --> $d [$path]\n"; |
497 |
} |
} |
|
$html .= $display_data."###\n"; |
|
498 |
} |
} |
499 |
|
|
500 |
} |
} |
501 |
if ($swish_data) { |
|
502 |
# remove extra spaces |
if (! $iterate_by_page) { |
503 |
$swish_data =~ s/ +/ /g; |
if ($display_data) { |
504 |
$swish_data =~ s/ +$//g; |
if ($field eq "headline") { |
505 |
|
$xml .= xmlify("headline", $display_data); |
506 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
} else { |
|
} |
|
507 |
|
|
508 |
if ($swish_exact_data) { |
# fallback to empty field name if needed |
509 |
$swish_exact_data =~ s/ +/ /g; |
$html .= get_field_name($config,$field,$field_usage{$field}) || ''; |
510 |
$swish_exact_data =~ s/ +$//g; |
$html .= "#-#".$display_data."###\n"; |
511 |
|
} |
512 |
|
} |
513 |
|
if ($swish_data) { |
514 |
|
# remove extra spaces |
515 |
|
$swish_data =~ s/ +/ /g; |
516 |
|
$swish_data =~ s/ +$//g; |
517 |
|
|
518 |
# add delimiters before and after word. |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
519 |
# That is required to produce exact match |
} |
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
|
|
} |
|
520 |
|
|
521 |
|
if ($swish_exact_data) { |
522 |
|
$swish_exact_data =~ s/ +/ /g; |
523 |
|
$swish_exact_data =~ s/ +$//g; |
524 |
|
|
525 |
|
# add delimiters before and after word. |
526 |
|
# That is required to produce exact match |
527 |
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
528 |
|
} |
529 |
|
} |
530 |
} |
} |
531 |
|
|
532 |
# dump formatted output in <html> |
# dump formatted output in <html> |
571 |
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"; |
572 |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
573 |
|
|
574 |
|
# create new lookup file |
575 |
|
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
576 |
|
if ($lookup_file) { |
577 |
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
578 |
|
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
579 |
|
print STDERR "creating lookup file '$lookup_file'\n"; |
580 |
|
} |
581 |
|
|
582 |
|
# open existing lookup file |
583 |
|
$lookup_file = $cfg -> val($database, 'lookup_open'); # optional |
584 |
|
if ($lookup_file) { |
585 |
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644; |
586 |
|
tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644; |
587 |
|
print STDERR "opening lookup file '$lookup_file'\n"; |
588 |
|
} |
589 |
|
|
590 |
print STDERR "reading ./import_xml/$type.xml\n"; |
print STDERR "reading ./import_xml/$type.xml\n"; |
591 |
|
|
592 |
# extract just type basic |
# extract just type basic |
625 |
# now read database |
# now read database |
626 |
print STDERR "using: $type...\n"; |
print STDERR "using: $type...\n"; |
627 |
|
|
628 |
|
# erase cache for tags by order in this database |
629 |
|
delete $cache->{tags_by_order}; |
630 |
|
|
631 |
if ($type_base eq "isis") { |
if ($type_base eq "isis") { |
632 |
|
|
633 |
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 "$database doesn't have 'isis_db' defined!"; |
840 |
fakeprogress($i); |
fakeprogress($i); |
841 |
|
|
842 |
} |
} |
843 |
|
# close lookup |
844 |
|
untie %lhash if (%lhash); |
845 |
} |
} |
846 |
} |
} |
847 |
|
|