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; |
118 |
delete $cache->{display_data}; |
delete $cache->{display_data}; |
119 |
delete $cache->{swish_data}; |
delete $cache->{swish_data}; |
120 |
delete $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 |
($s,$se,$d,$i) = (0,0,0,1) if (lc($x->{type}) eq "index"); |
($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index"); |
198 |
($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact"); |
($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 $data = shift; |
215 |
|
return if (! defined($data)); |
216 |
|
if ($data && $data =~ s/\s*eval{([^}]+)}\s*//) { |
217 |
|
if (eval "$1") { |
218 |
|
return $data; |
219 |
|
} else { |
220 |
|
return undef; |
221 |
|
} |
222 |
|
} else { |
223 |
|
return $data; |
224 |
|
} |
225 |
|
} |
226 |
|
|
227 |
|
sub mkformat($$) { |
228 |
my $x = shift || die "mkformat needs tag reference"; |
my $x = shift || die "mkformat needs tag reference"; |
229 |
my $data = shift || return; |
my $data = shift || return; |
230 |
my $format_name = x($x->{format_name}) || return $data; |
my $format_name = x($x->{format_name}) || return chk_eval($data); |
231 |
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!"; |
232 |
my $format_delimiter = x($x->{format_delimiter}); |
my $format_delimiter = x($x->{format_delimiter}); |
233 |
my @data; |
my @data; |
240 |
if ($fmt) { |
if ($fmt) { |
241 |
my $nr = scalar $fmt =~ s/%s/%s/g; |
my $nr = scalar $fmt =~ s/%s/%s/g; |
242 |
if (($#data+1) == $nr) { |
if (($#data+1) == $nr) { |
243 |
return sprintf($fmt,@data); |
return chk_eval(sprintf($fmt,@data)); |
244 |
} else { |
} else { |
245 |
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"; |
246 |
return $data; |
return chk_eval($data); |
247 |
} |
} |
248 |
} else { |
} else { |
249 |
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"; |
271 |
$display = $new_display; |
$display = $new_display; |
272 |
$cache->{lhash}->{$display} = $new_display; |
$cache->{lhash}->{$display} = $new_display; |
273 |
} else { |
} else { |
274 |
print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
# print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
275 |
$display = ""; |
$display = ""; |
276 |
$cache->{lhash}->{$display} = $null; |
$cache->{lhash}->{$display} = $null; |
277 |
} |
} |
296 |
no strict 'refs'; |
no strict 'refs'; |
297 |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
298 |
$swish_data .= $tmp if ($s); |
$swish_data .= $tmp if ($s); |
299 |
$swish_exact_data .= $tmp if ($se); |
$swish_exact_data .= "xxbxx $tmp xxexx " if ($se && $tmp ne ""); |
300 |
|
|
301 |
} else { |
} else { |
302 |
$swish_data .= $swish if ($s); |
$swish_data .= $swish if ($s); |
303 |
$swish_exact_data .= $swish if ($se); |
$swish_exact_data .= "xxbxx $swish xxexx " if ($se && $swish ne ""); |
304 |
} |
} |
305 |
|
|
306 |
|
$tmp = chk_eval($tmp); |
307 |
|
$swish_data .= $tmp if ($s && $tmp); |
308 |
|
$swish_exact_data .= "xxbxx $tmp xxexx " if ($se && $tmp ne ""); |
309 |
} |
} |
310 |
|
|
311 |
# type="display" ; field for display |
# type="display" ; field for display |
312 |
if ($d && $display) { |
if ($d && $display) { |
313 |
|
my $ldel = $delimiter; |
314 |
if ($line_delimiter && $display_data) { |
if ($line_delimiter && $display_data) { |
315 |
$display_data .= $line_delimiter; |
$ldel = $line_delimiter; |
316 |
} |
} |
317 |
if ($filter) { |
if ($filter) { |
318 |
no strict 'refs'; |
no strict 'refs'; |
319 |
if ($display_data) { |
my @arr; |
320 |
$display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display))); |
foreach my $tmp (&$filter($display)) { |
321 |
} else { |
my $tmp2 = mkformat($x,$tmp); |
322 |
$display_data = join($delimiter,mkformat($x,&$filter($display))); |
push @arr,$tmp2 if ($tmp2); |
323 |
} |
} |
324 |
|
$display_data .= $ldel if ($display_data && @arr); |
325 |
|
$display_data .= join($delimiter,@arr); |
326 |
} else { |
} else { |
327 |
if ($display_data) { |
$display_data .= $ldel if ($display_data); |
328 |
$display_data .= $delimiter.mkformat($x,$display); |
my $tmp = mkformat($x,$display); |
329 |
} else { |
$display_data .= $tmp if ($tmp); |
|
$display_data = mkformat($x,$display); |
|
|
} |
|
330 |
} |
} |
331 |
} |
} |
332 |
|
|
333 |
# type="index" ; insert into index |
# type="index" ; insert into index |
334 |
|
my $idisplay; |
335 |
if ($i && $display) { |
if ($i && $display) { |
336 |
|
$idisplay = $display; |
337 |
if ($filter) { |
if ($filter) { |
338 |
no strict 'refs'; |
no strict 'refs'; |
339 |
$display = &$filter($display); |
$idisplay = &$filter($idisplay); |
|
} |
|
|
if ($x->{append} && @index_data) { |
|
|
$index_data[$#index_data].=$display; |
|
|
} else { |
|
|
push @index_data, $display; |
|
340 |
} |
} |
341 |
|
$idisplay = chk_eval($idisplay); |
342 |
|
push @index_data, $idisplay if ($idisplay && !$iterate_by_page); |
343 |
} |
} |
344 |
|
|
345 |
# store fields in lookup |
# store fields in lookup |
348 |
if ($lookup_key) { |
if ($lookup_key) { |
349 |
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!)"; |
350 |
} else { |
} else { |
351 |
$lookup_key = $display; |
if ($filter) { |
352 |
|
no strict 'refs'; |
353 |
|
$lookup_key = &$filter($display); |
354 |
|
} else { |
355 |
|
$lookup_key = $display; |
356 |
|
} |
357 |
} |
} |
358 |
} elsif (lc($x->{type}) eq "lookup_val") { |
} elsif (lc($x->{type}) eq "lookup_val") { |
359 |
if ($lookup_key) { |
if ($lookup_key) { |
360 |
$lhash{$lookup_key} = $display; |
if ($filter) { |
361 |
|
no strict 'refs'; |
362 |
|
$lhash{$lookup_key} = &$filter($display); |
363 |
|
} else { |
364 |
|
$lhash{$lookup_key} = $display; |
365 |
|
} |
366 |
} else { |
} else { |
367 |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
368 |
} |
} |
375 |
sub iterate_fld($$$$$$) { |
sub iterate_fld($$$$$$) { |
376 |
my ($cache,$what,$field,$page,$data,$append) = @_; |
my ($cache,$what,$field,$page,$data,$append) = @_; |
377 |
return if (!$data); |
return if (!$data); |
378 |
my $line_delimiter = $page_line_delimiter; |
|
379 |
$line_delimiter = " " if ($append); |
my $ldel = $page_line_delimiter; |
380 |
|
$ldel = " " if ($append); |
381 |
|
#print STDERR "line delimiter: ",Dumper($ldel) if ($ldel); |
382 |
if (! $cache->{$what}->{$field}->[$page]) { |
if (! $cache->{$what}->{$field}->[$page]) { |
383 |
$cache->{$what}->{$field}->[$page] = $data; |
$cache->{$what}->{$field}->[$page] = $data; |
384 |
} else { |
} else { |
385 |
$cache->{$what}->{$field}->[$page] .= $line_delimiter.$data; |
$cache->{$what}->{$field}->[$page] .= $ldel.$data; |
386 |
} |
} |
387 |
} |
} |
388 |
|
|
389 |
if ($display_data) { |
if ($display_data) { |
|
print STDERR "line delimiter: ",Dumper($line_delimiter) if ($line_delimiter); |
|
390 |
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
391 |
} |
} |
392 |
$display_data = ""; |
$display_data = ""; |
398 |
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
399 |
$swish_exact_data = ""; |
$swish_exact_data = ""; |
400 |
} |
} |
401 |
|
|
402 |
|
if ($idisplay) { |
403 |
|
my $ldel=$page_line_delimiter; |
404 |
|
my @index_data; |
405 |
|
if ($cache->{index_data}->{$field}->[$page]) { |
406 |
|
|
407 |
|
@index_data = @{$cache->{index_data}->{$field}->[$page]}; |
408 |
|
} |
409 |
|
if ($x->{append}) { |
410 |
|
if (@index_data) { |
411 |
|
$index_data[$#index_data] .= $idisplay; |
412 |
|
} else { |
413 |
|
push @index_data, $idisplay; |
414 |
|
} |
415 |
|
} else { |
416 |
|
push @index_data, $idisplay; |
417 |
|
} |
418 |
|
$idisplay = ""; |
419 |
|
@{$cache->{index_data}->{$field}->[$page]} = @index_data; |
420 |
|
} |
421 |
} |
} |
422 |
} |
} |
423 |
|
|
424 |
# fill data in index |
if (! $iterate_by_page) { |
425 |
foreach my $d (@index_data) { |
my $idel = $x->{index_delimiter}; |
426 |
$index->insert($field, $d, $path); |
# fill data in index |
427 |
|
foreach my $tmp (@index_data) { |
428 |
|
my $i = $d = $tmp; |
429 |
|
if ($idel && $tmp =~ m/$idel/) { |
430 |
|
($i,$d) = split(/$idel/,$tmp); |
431 |
|
} |
432 |
|
$index->insert($field, $i, $d, $path); |
433 |
|
} |
434 |
|
@index_data = (); |
435 |
} |
} |
436 |
} |
} |
437 |
|
|
479 |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
480 |
#print STDERR Dumper($cache->{display_data}); |
#print STDERR Dumper($cache->{display_data}); |
481 |
for (my $page=0; $page <= $nr_pages; $page++) { |
for (my $page=0; $page <= $nr_pages; $page++) { |
482 |
|
my $display_data; |
483 |
my $display_data = $cache->{display_data}->{$field}->[$page]; |
if ($cache->{format}->{$field}) { |
484 |
|
my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]); |
485 |
|
$display_data=$tmp if ($tmp); |
486 |
|
} else { |
487 |
|
$display_data = $cache->{display_data}->{$field}->[$page]; |
488 |
|
} |
489 |
if ($display_data) { # default |
if ($display_data) { # default |
490 |
if ($field eq "headline") { |
if ($field eq "headline") { |
491 |
$xml .= xmlify("headline", $display_data); |
$xml .= xmlify("headline", $display_data); |
513 |
|
|
514 |
# add delimiters before and after word. |
# add delimiters before and after word. |
515 |
# That is required to produce exact match |
# That is required to produce exact match |
516 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
517 |
|
} |
518 |
|
|
519 |
|
my $idel = $cache->{index_delimiter}->{$field}; |
520 |
|
foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) { |
521 |
|
my $i = $tmp; |
522 |
|
my $d = $tmp; |
523 |
|
if ($idel && $tmp =~ m/$idel/) { |
524 |
|
($i,$d) = split(/$idel/,$tmp); |
525 |
|
} |
526 |
|
$index->insert($field, $i, $d, $path); |
527 |
|
#print STDERR "index [$idel] $field: $i --> $d [$path]\n"; |
528 |
} |
} |
529 |
} |
} |
530 |
|
|
555 |
|
|
556 |
# add delimiters before and after word. |
# add delimiters before and after word. |
557 |
# That is required to produce exact match |
# That is required to produce exact match |
558 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
559 |
} |
} |
560 |
} |
} |
561 |
} |
} |
608 |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
609 |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
610 |
print STDERR "creating lookup file '$lookup_file'\n"; |
print STDERR "creating lookup file '$lookup_file'\n"; |
611 |
|
# delete memory cache for lookup file |
612 |
|
delete $cache->{lhash}; |
613 |
} |
} |
614 |
|
|
615 |
# open existing lookup file |
# open existing lookup file |
746 |
print STDERR "\n"; |
print STDERR "\n"; |
747 |
|
|
748 |
} elsif ($type_base eq "excel") { |
} elsif ($type_base eq "excel") { |
749 |
use Spreadsheet::ParseExcel; |
require Spreadsheet::ParseExcel; |
750 |
use Spreadsheet::ParseExcel::Utility qw(int2col); |
require Spreadsheet::ParseExcel::Utility; |
751 |
|
import Spreadsheet::ParseExcel::Utility qw(int2col); |
752 |
|
|
753 |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
754 |
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 "$database doesn't have 'excel_file' defined!"; |
802 |
} |
} |
803 |
} elsif ($type_base eq "marc") { |
} elsif ($type_base eq "marc") { |
804 |
|
|
805 |
use MARC; |
require MARC; |
806 |
|
|
807 |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
808 |
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!"; |
809 |
|
|
810 |
# optional argument is format |
# optional argument is format |
811 |
my $format = x($config->{format}) || 'usmarc'; |
my $format = x($config->{marc_format}) || 'usmarc'; |
812 |
|
|
813 |
print STDERR "Reading MARC file '$marc_file'\n"; |
print STDERR "Reading MARC file '$marc_file'\n"; |
814 |
|
|
815 |
my $marc = new MARC; |
my $marc = new MARC; |
816 |
my $nr = $marc->openmarc({ |
my $nr = $marc->openmarc({ |
817 |
file=>$marc_file, format=>$format |
file=>$marc_file, format=>$format |
818 |
}) || die "Can't open MARC file '$marc_file'"; |
}) || die "Can't open MARC file '$marc_file' with format '$format'"; |
819 |
|
|
820 |
|
# read MARC file in memory |
821 |
|
$marc->nextmarc(-1); |
822 |
|
|
823 |
my $i=0; # record nr. |
my $max_rec = $marc->marc_count(); |
824 |
|
|
825 |
my $rec; |
for(my $i=1; $i<=$max_rec; $i++) { |
826 |
|
|
827 |
while ($marc->nextmarc(1)) { |
progress($i,$max_rec); |
828 |
|
|
829 |
# XXX |
# store value for marc_sf.pm |
830 |
fakeprogress($i++); |
$main::cache->{marc_record} = $i; |
831 |
|
|
832 |
my $swishpath = $database."#".$i; |
my $swishpath = $database."#".$i; |
833 |
|
|
839 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
840 |
} |
} |
841 |
} |
} |
842 |
|
|
843 |
|
print STDERR "\n"; |
844 |
|
|
845 |
} elsif ($type_base eq "feed") { |
} elsif ($type_base eq "feed") { |
846 |
|
|
847 |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |