72 |
# if you are tight on memory, turn this off |
# if you are tight on memory, turn this off |
73 |
my $use_lhash_cache = 1; |
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 |
# lookup key |
115 |
my $lookup_key; |
my $lookup_key; |
116 |
|
|
117 |
# cache for field in pages |
# cache for field in pages |
118 |
$cache->{display_data} = (); |
delete $cache->{display_data}; |
119 |
$cache->{swish_data} = (); |
delete $cache->{swish_data}; |
120 |
$cache->{swish_exact_data} = (); |
delete $cache->{swish_exact_data}; |
121 |
|
delete $cache->{index_data}; |
122 |
|
delete $cache->{index_delimiter}; |
123 |
my @page_fields; # names of fields |
my @page_fields; # names of fields |
124 |
|
|
125 |
|
|
140 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
141 |
} |
} |
142 |
if ($field_name) { |
if ($field_name) { |
143 |
return x($field_name); |
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 |
|
|
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); |
173 |
my %page_max = (); |
my %page_max = (); |
174 |
# default line_delimiter if using |
# default line_delimiter if using |
175 |
my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>'; |
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 |
|
|
194 |
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
my ($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 |
$il = 1 if (lc($x->{type}) =~ /^lookup/); |
($s,$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,$display) = (1,1); |
210 |
|
|
211 |
# placeholder for all repeatable entries for index |
# placeholder for all repeatable entries for index |
|
my @index_data; |
|
212 |
|
|
213 |
sub mkformat { |
sub chk_eval($) { |
214 |
|
my $data = shift; |
215 |
|
if ($data =~ s/eval{([^}]+)}//) { |
216 |
|
if (eval "$1") { |
217 |
|
return $data; |
218 |
|
} else { |
219 |
|
return undef; |
220 |
|
} |
221 |
|
} else { |
222 |
|
return $data; |
223 |
|
} |
224 |
|
} |
225 |
|
|
226 |
|
sub mkformat($$) { |
227 |
my $x = shift || die "mkformat needs tag reference"; |
my $x = shift || die "mkformat needs tag reference"; |
228 |
my $data = shift || return; |
my $data = shift || return; |
229 |
my $format_name = x($x->{format_name}) || return $data; |
my $format_name = x($x->{format_name}) || return chk_eval($data); |
230 |
my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!"; |
my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!"; |
231 |
my $format_delimiter = x($x->{format_delimiter}); |
my $format_delimiter = x($x->{format_delimiter}); |
232 |
my @data; |
my @data; |
239 |
if ($fmt) { |
if ($fmt) { |
240 |
my $nr = scalar $fmt =~ s/%s/%s/g; |
my $nr = scalar $fmt =~ s/%s/%s/g; |
241 |
if (($#data+1) == $nr) { |
if (($#data+1) == $nr) { |
242 |
return sprintf($fmt,@data); |
return chk_eval(sprintf($fmt,@data)); |
243 |
} else { |
} else { |
244 |
print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
#print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
245 |
return $data; |
return chk_eval($data); |
246 |
} |
} |
247 |
} else { |
} else { |
248 |
print STDERR "usage of link '$format_name' without defined format (<link> tag)\n"; |
print STDERR "usage of link '$format_name' without defined format (<link> tag)\n"; |
295 |
no strict 'refs'; |
no strict 'refs'; |
296 |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
297 |
$swish_data .= $tmp if ($s); |
$swish_data .= $tmp if ($s); |
298 |
$swish_exact_data .= $tmp if ($se); |
$swish_exact_data .= "xxbxx $tmp xxexx " if ($se && $tmp ne ""); |
299 |
|
|
300 |
} else { |
} else { |
301 |
$swish_data .= $swish if ($s); |
$swish_data .= $swish if ($s); |
302 |
$swish_exact_data .= $swish if ($se); |
$swish_exact_data .= "xxbxx $swish xxexx " if ($se && $swish ne ""); |
303 |
} |
} |
304 |
} |
} |
305 |
|
|
306 |
# type="display" ; field for display |
# type="display" ; field for display |
307 |
if ($d && $display) { |
if ($d && $display) { |
308 |
|
my $ldel = $delimiter; |
309 |
if ($line_delimiter && $display_data) { |
if ($line_delimiter && $display_data) { |
310 |
$display_data .= $line_delimiter; |
$ldel = $line_delimiter; |
311 |
} |
} |
312 |
if ($filter) { |
if ($filter) { |
313 |
no strict 'refs'; |
no strict 'refs'; |
314 |
if ($display_data) { |
my @arr; |
315 |
$display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display))); |
foreach my $tmp (&$filter($display)) { |
316 |
} else { |
my $tmp2 = mkformat($x,$tmp); |
317 |
$display_data = join($delimiter,mkformat($x,&$filter($display))); |
push @arr,$tmp2 if ($tmp2); |
318 |
} |
} |
319 |
|
$display_data .= $ldel if ($display_data && @arr); |
320 |
|
$display_data .= join($delimiter,@arr); |
321 |
} else { |
} else { |
322 |
if ($display_data) { |
$display_data .= $ldel if ($display_data); |
323 |
$display_data .= $delimiter.mkformat($x,$display); |
my $tmp = mkformat($x,$display); |
324 |
} else { |
$display_data .= $tmp if ($tmp); |
|
$display_data = mkformat($x,$display); |
|
|
} |
|
325 |
} |
} |
326 |
} |
} |
327 |
|
|
328 |
# type="index" ; insert into index |
# type="index" ; insert into index |
329 |
|
my $idisplay; |
330 |
if ($i && $display) { |
if ($i && $display) { |
331 |
|
$idisplay = $display; |
332 |
if ($filter) { |
if ($filter) { |
333 |
no strict 'refs'; |
no strict 'refs'; |
334 |
$display = &$filter($display); |
$idisplay = &$filter($idisplay); |
|
} |
|
|
if ($x->{append} && @index_data) { |
|
|
$index_data[$#index_data].=$display; |
|
|
} else { |
|
|
push @index_data, $display; |
|
335 |
} |
} |
336 |
|
push @index_data, $idisplay if (! $iterate_by_page); |
337 |
} |
} |
338 |
|
|
339 |
# store fields in lookup |
# store fields in lookup |
342 |
if ($lookup_key) { |
if ($lookup_key) { |
343 |
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
344 |
} else { |
} else { |
345 |
$lookup_key = $display; |
if ($filter) { |
346 |
|
no strict 'refs'; |
347 |
|
$lookup_key = &$filter($display); |
348 |
|
} else { |
349 |
|
$lookup_key = $display; |
350 |
|
} |
351 |
} |
} |
352 |
} elsif (lc($x->{type}) eq "lookup_val") { |
} elsif (lc($x->{type}) eq "lookup_val") { |
353 |
if ($lookup_key) { |
if ($lookup_key) { |
354 |
$lhash{$lookup_key} = $display; |
if ($filter) { |
355 |
|
no strict 'refs'; |
356 |
|
$lhash{$lookup_key} = &$filter($display); |
357 |
|
} else { |
358 |
|
$lhash{$lookup_key} = $display; |
359 |
|
} |
360 |
} else { |
} else { |
361 |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
362 |
} |
} |
369 |
sub iterate_fld($$$$$$) { |
sub iterate_fld($$$$$$) { |
370 |
my ($cache,$what,$field,$page,$data,$append) = @_; |
my ($cache,$what,$field,$page,$data,$append) = @_; |
371 |
return if (!$data); |
return if (!$data); |
372 |
my $line_delimiter = $page_line_delimiter; |
|
373 |
$line_delimiter = " " if ($append); |
my $ldel = $page_line_delimiter; |
374 |
|
$ldel = " " if ($append); |
375 |
|
#print STDERR "line delimiter: ",Dumper($ldel) if ($ldel); |
376 |
if (! $cache->{$what}->{$field}->[$page]) { |
if (! $cache->{$what}->{$field}->[$page]) { |
377 |
$cache->{$what}->{$field}->[$page] = $data; |
$cache->{$what}->{$field}->[$page] = $data; |
378 |
} else { |
} else { |
379 |
$cache->{$what}->{$field}->[$page] .= $line_delimiter.$data; |
$cache->{$what}->{$field}->[$page] .= $ldel.$data; |
380 |
} |
} |
381 |
} |
} |
382 |
|
|
383 |
if ($display_data) { |
if ($display_data) { |
|
print STDERR "line delimiter: ",Dumper($line_delimiter) if ($line_delimiter); |
|
384 |
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
385 |
} |
} |
386 |
$display_data = ""; |
$display_data = ""; |
392 |
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
393 |
$swish_exact_data = ""; |
$swish_exact_data = ""; |
394 |
} |
} |
395 |
|
|
396 |
|
if ($idisplay) { |
397 |
|
my $ldel=$page_line_delimiter; |
398 |
|
my @index_data; |
399 |
|
if ($cache->{index_data}->{$field}->[$page]) { |
400 |
|
|
401 |
|
@index_data = @{$cache->{index_data}->{$field}->[$page]}; |
402 |
|
} |
403 |
|
if ($x->{append}) { |
404 |
|
if (@index_data) { |
405 |
|
$index_data[$#index_data] .= $idisplay; |
406 |
|
} else { |
407 |
|
push @index_data, $idisplay; |
408 |
|
} |
409 |
|
} else { |
410 |
|
push @index_data, $idisplay; |
411 |
|
} |
412 |
|
$idisplay = ""; |
413 |
|
@{$cache->{index_data}->{$field}->[$page]} = @index_data; |
414 |
|
} |
415 |
} |
} |
416 |
} |
} |
417 |
|
|
418 |
# fill data in index |
if (! $iterate_by_page) { |
419 |
foreach my $d (@index_data) { |
my $idel = $x->{index_delimiter}; |
420 |
$index->insert($field, $d, $path); |
# fill data in index |
421 |
|
foreach my $tmp (@index_data) { |
422 |
|
my $i = $d = $tmp; |
423 |
|
if ($idel && $tmp =~ m/$idel/) { |
424 |
|
($i,$d) = split(/$idel/,$tmp); |
425 |
|
} |
426 |
|
$index->insert($field, $i, $d, $path); |
427 |
|
} |
428 |
|
@index_data = (); |
429 |
} |
} |
430 |
} |
} |
431 |
|
|
473 |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
474 |
#print STDERR Dumper($cache->{display_data}); |
#print STDERR Dumper($cache->{display_data}); |
475 |
for (my $page=0; $page <= $nr_pages; $page++) { |
for (my $page=0; $page <= $nr_pages; $page++) { |
476 |
|
my $display_data; |
477 |
my $display_data = $cache->{display_data}->{$field}->[$page]; |
if ($cache->{format}->{$field}) { |
478 |
|
my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]); |
479 |
|
$display_data=$tmp if ($tmp); |
480 |
|
} else { |
481 |
|
$display_data = $cache->{display_data}->{$field}->[$page]; |
482 |
|
} |
483 |
if ($display_data) { # default |
if ($display_data) { # default |
484 |
if ($field eq "headline") { |
if ($field eq "headline") { |
485 |
$xml .= xmlify("headline", $display_data); |
$xml .= xmlify("headline", $display_data); |
507 |
|
|
508 |
# add delimiters before and after word. |
# add delimiters before and after word. |
509 |
# That is required to produce exact match |
# That is required to produce exact match |
510 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
511 |
|
} |
512 |
|
|
513 |
|
my $idel = $cache->{index_delimiter}->{$field}; |
514 |
|
foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) { |
515 |
|
my $i = $tmp; |
516 |
|
my $d = $tmp; |
517 |
|
if ($idel && $tmp =~ m/$idel/) { |
518 |
|
($i,$d) = split(/$idel/,$tmp); |
519 |
|
} |
520 |
|
$index->insert($field, $i, $d, $path); |
521 |
|
#print STDERR "index [$idel] $field: $i --> $d [$path]\n"; |
522 |
} |
} |
523 |
} |
} |
524 |
|
|
549 |
|
|
550 |
# add delimiters before and after word. |
# add delimiters before and after word. |
551 |
# That is required to produce exact match |
# That is required to produce exact match |
552 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
553 |
} |
} |
554 |
} |
} |
555 |
} |
} |
602 |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
603 |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
604 |
print STDERR "creating lookup file '$lookup_file'\n"; |
print STDERR "creating lookup file '$lookup_file'\n"; |
605 |
|
# delete memory cache for lookup file |
606 |
|
delete $cache->{lhash}; |
607 |
} |
} |
608 |
|
|
609 |
# open existing lookup file |
# open existing lookup file |
652 |
# now read database |
# now read database |
653 |
print STDERR "using: $type...\n"; |
print STDERR "using: $type...\n"; |
654 |
|
|
655 |
|
# erase cache for tags by order in this database |
656 |
|
delete $cache->{tags_by_order}; |
657 |
|
|
658 |
if ($type_base eq "isis") { |
if ($type_base eq "isis") { |
659 |
|
|
660 |
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 "$database doesn't have 'isis_db' defined!"; |
740 |
print STDERR "\n"; |
print STDERR "\n"; |
741 |
|
|
742 |
} elsif ($type_base eq "excel") { |
} elsif ($type_base eq "excel") { |
743 |
use Spreadsheet::ParseExcel; |
require Spreadsheet::ParseExcel; |
744 |
use Spreadsheet::ParseExcel::Utility qw(int2col); |
require Spreadsheet::ParseExcel::Utility; |
745 |
|
import Spreadsheet::ParseExcel::Utility qw(int2col); |
746 |
|
|
747 |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
748 |
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
796 |
} |
} |
797 |
} elsif ($type_base eq "marc") { |
} elsif ($type_base eq "marc") { |
798 |
|
|
799 |
use MARC; |
require MARC; |
800 |
|
|
801 |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
802 |
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
803 |
|
|
804 |
# optional argument is format |
# optional argument is format |
805 |
my $format = x($config->{format}) || 'usmarc'; |
my $format = x($config->{marc_format}) || 'usmarc'; |
806 |
|
|
807 |
print STDERR "Reading MARC file '$marc_file'\n"; |
print STDERR "Reading MARC file '$marc_file'\n"; |
808 |
|
|
809 |
my $marc = new MARC; |
my $marc = new MARC; |
810 |
my $nr = $marc->openmarc({ |
my $nr = $marc->openmarc({ |
811 |
file=>$marc_file, format=>$format |
file=>$marc_file, format=>$format |
812 |
}) || die "Can't open MARC file '$marc_file'"; |
}) || die "Can't open MARC file '$marc_file' with format '$format'"; |
813 |
|
|
814 |
|
# read MARC file in memory |
815 |
|
$marc->nextmarc(-1); |
816 |
|
|
817 |
my $i=0; # record nr. |
my $max_rec = $marc->marc_count(); |
818 |
|
|
819 |
my $rec; |
for(my $i=1; $i<=$max_rec; $i++) { |
820 |
|
|
821 |
while ($marc->nextmarc(1)) { |
progress($i,$max_rec); |
822 |
|
|
823 |
# XXX |
# store value for marc_sf.pm |
824 |
fakeprogress($i++); |
$main::cache->{marc_record} = $i; |
825 |
|
|
826 |
my $swishpath = $database."#".$i; |
my $swishpath = $database."#".$i; |
827 |
|
|
833 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
834 |
} |
} |
835 |
} |
} |
836 |
|
|
837 |
|
print STDERR "\n"; |
838 |
|
|
839 |
} elsif ($type_base eq "feed") { |
} elsif ($type_base eq "feed") { |
840 |
|
|
841 |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |