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; |
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}; |
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 |
|
|
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 |
($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/); |
($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/); |
200 |
|
|
210 |
|
|
211 |
# placeholder for all repeatable entries for index |
# placeholder for all repeatable entries for index |
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 $data; |
228 |
if (($#data+1) == $nr) { |
if (($#data+1) == $nr) { |
229 |
return sprintf($fmt,@data); |
return sprintf($fmt,@data); |
230 |
} else { |
} else { |
231 |
print STDERR "mkformat: [$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"; |
232 |
return $data; |
return $data; |
233 |
} |
} |
234 |
} else { |
} else { |
257 |
$display = $new_display; |
$display = $new_display; |
258 |
$cache->{lhash}->{$display} = $new_display; |
$cache->{lhash}->{$display} = $new_display; |
259 |
} else { |
} else { |
260 |
print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
# print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
261 |
$display = ""; |
$display = ""; |
262 |
$cache->{lhash}->{$display} = $null; |
$cache->{lhash}->{$display} = $null; |
263 |
} |
} |
278 |
} |
} |
279 |
# type="swish" ; field for swish |
# type="swish" ; field for swish |
280 |
if ($swish) { |
if ($swish) { |
281 |
|
my $tmp = $swish; |
282 |
if ($filter && ($s || $se)) { |
if ($filter && ($s || $se)) { |
283 |
no strict 'refs'; |
no strict 'refs'; |
284 |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
$tmp = join(" ",&$filter($tmp)) if ($s || $se); |
|
$swish_data .= $tmp if ($s); |
|
|
$swish_exact_data .= $tmp if ($se); |
|
|
|
|
|
} else { |
|
|
$swish_data .= $swish if ($s); |
|
|
$swish_exact_data .= $swish if ($se); |
|
285 |
} |
} |
286 |
|
|
287 |
|
$swish_data .= $tmp if ($s && $tmp); |
288 |
|
$swish_exact_data .= "xxbxx $tmp xxexx " if ($tmp && $tmp ne "" && $se); |
289 |
} |
} |
290 |
|
|
291 |
# type="display" ; field for display |
# type="display" ; field for display |
292 |
if ($d && $display) { |
if ($d && $display) { |
293 |
|
my $ldel = $delimiter; |
294 |
if ($line_delimiter && $display_data) { |
if ($line_delimiter && $display_data) { |
295 |
$display_data .= $line_delimiter; |
$ldel = $line_delimiter; |
296 |
} |
} |
297 |
if ($filter) { |
if ($filter) { |
298 |
no strict 'refs'; |
no strict 'refs'; |
299 |
if ($display_data) { |
my @arr; |
300 |
$display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display))); |
foreach my $tmp (&$filter($display)) { |
301 |
} else { |
my $tmp2 = mkformat($x,$tmp); |
302 |
$display_data = join($delimiter,mkformat($x,&$filter($display))); |
push @arr,$tmp2 if ($tmp2); |
303 |
} |
} |
304 |
|
$display_data .= $ldel if ($display_data && @arr); |
305 |
|
$display_data .= join($delimiter,@arr); |
306 |
} else { |
} else { |
307 |
if ($display_data) { |
$display_data .= $ldel if ($display_data); |
308 |
$display_data .= $delimiter.mkformat($x,$display); |
my $tmp = mkformat($x,$display); |
309 |
} else { |
$display_data .= $tmp if ($tmp); |
|
$display_data = mkformat($x,$display); |
|
|
} |
|
310 |
} |
} |
311 |
} |
} |
312 |
|
|
318 |
no strict 'refs'; |
no strict 'refs'; |
319 |
$idisplay = &$filter($idisplay); |
$idisplay = &$filter($idisplay); |
320 |
} |
} |
321 |
push @index_data, $idisplay if (! $iterate_by_page); |
push @index_data, $idisplay if ($idisplay && !$iterate_by_page); |
322 |
} |
} |
323 |
|
|
324 |
# store fields in lookup |
# store fields in lookup |
327 |
if ($lookup_key) { |
if ($lookup_key) { |
328 |
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!)"; |
329 |
} else { |
} else { |
330 |
$lookup_key = $display; |
if ($filter) { |
331 |
|
no strict 'refs'; |
332 |
|
$lookup_key = &$filter($display); |
333 |
|
} else { |
334 |
|
$lookup_key = $display; |
335 |
|
} |
336 |
} |
} |
337 |
} elsif (lc($x->{type}) eq "lookup_val") { |
} elsif (lc($x->{type}) eq "lookup_val") { |
338 |
if ($lookup_key) { |
if ($lookup_key) { |
339 |
$lhash{$lookup_key} = $display; |
if ($filter) { |
340 |
|
no strict 'refs'; |
341 |
|
$lhash{$lookup_key} = &$filter($display); |
342 |
|
} else { |
343 |
|
$lhash{$lookup_key} = $display; |
344 |
|
} |
345 |
} else { |
} else { |
346 |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
347 |
} |
} |
386 |
@index_data = @{$cache->{index_data}->{$field}->[$page]}; |
@index_data = @{$cache->{index_data}->{$field}->[$page]}; |
387 |
} |
} |
388 |
if ($x->{append}) { |
if ($x->{append}) { |
389 |
$index_data[$#index_data] .= $idisplay; |
if (@index_data) { |
390 |
|
$index_data[$#index_data] .= $idisplay; |
391 |
|
} else { |
392 |
|
push @index_data, $idisplay; |
393 |
|
} |
394 |
} else { |
} else { |
395 |
push @index_data, $idisplay; |
push @index_data, $idisplay; |
396 |
} |
} |
401 |
} |
} |
402 |
|
|
403 |
if (! $iterate_by_page) { |
if (! $iterate_by_page) { |
404 |
|
my $idel = $x->{index_delimiter}; |
405 |
# fill data in index |
# fill data in index |
406 |
foreach my $d (@index_data) { |
foreach my $tmp (@index_data) { |
407 |
$index->insert($field, $d, $path); |
my $i = $d = $tmp; |
408 |
|
if ($idel && $tmp =~ m/$idel/) { |
409 |
|
($i,$d) = split(/$idel/,$tmp); |
410 |
|
} |
411 |
|
$index->insert($field, $i, $d, $path); |
412 |
} |
} |
413 |
@index_data = (); |
@index_data = (); |
414 |
} |
} |
458 |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
459 |
#print STDERR Dumper($cache->{display_data}); |
#print STDERR Dumper($cache->{display_data}); |
460 |
for (my $page=0; $page <= $nr_pages; $page++) { |
for (my $page=0; $page <= $nr_pages; $page++) { |
461 |
|
my $display_data; |
462 |
my $display_data = $cache->{display_data}->{$field}->[$page]; |
if ($cache->{format}->{$field}) { |
463 |
|
my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]); |
464 |
|
$display_data=$tmp if ($tmp); |
465 |
|
} else { |
466 |
|
$display_data = $cache->{display_data}->{$field}->[$page]; |
467 |
|
} |
468 |
if ($display_data) { # default |
if ($display_data) { # default |
469 |
if ($field eq "headline") { |
if ($field eq "headline") { |
470 |
$xml .= xmlify("headline", $display_data); |
$xml .= xmlify("headline", $display_data); |
492 |
|
|
493 |
# add delimiters before and after word. |
# add delimiters before and after word. |
494 |
# That is required to produce exact match |
# That is required to produce exact match |
495 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
496 |
} |
} |
497 |
|
|
498 |
foreach my $d (@{$cache->{index_data}->{$field}->[$page]}) { |
my $idel = $cache->{index_delimiter}->{$field}; |
499 |
$index->insert($field, $d, $path); |
foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) { |
500 |
#print STDERR "index $field: $d [$path]\n"; |
my $i = $tmp; |
501 |
|
my $d = $tmp; |
502 |
|
if ($idel && $tmp =~ m/$idel/) { |
503 |
|
($i,$d) = split(/$idel/,$tmp); |
504 |
|
} |
505 |
|
$index->insert($field, $i, $d, $path); |
506 |
|
#print STDERR "index [$idel] $field: $i --> $d [$path]\n"; |
507 |
} |
} |
508 |
} |
} |
509 |
|
|
534 |
|
|
535 |
# add delimiters before and after word. |
# add delimiters before and after word. |
536 |
# That is required to produce exact match |
# That is required to produce exact match |
537 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
538 |
} |
} |
539 |
} |
} |
540 |
} |
} |
587 |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
588 |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
589 |
print STDERR "creating lookup file '$lookup_file'\n"; |
print STDERR "creating lookup file '$lookup_file'\n"; |
590 |
|
# delete memory cache for lookup file |
591 |
|
delete $cache->{lhash}; |
592 |
} |
} |
593 |
|
|
594 |
# open existing lookup file |
# open existing lookup file |
725 |
print STDERR "\n"; |
print STDERR "\n"; |
726 |
|
|
727 |
} elsif ($type_base eq "excel") { |
} elsif ($type_base eq "excel") { |
728 |
use Spreadsheet::ParseExcel; |
require Spreadsheet::ParseExcel; |
729 |
use Spreadsheet::ParseExcel::Utility qw(int2col); |
require Spreadsheet::ParseExcel::Utility; |
730 |
|
import Spreadsheet::ParseExcel::Utility qw(int2col); |
731 |
|
|
732 |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
733 |
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 "$database doesn't have 'excel_file' defined!"; |
781 |
} |
} |
782 |
} elsif ($type_base eq "marc") { |
} elsif ($type_base eq "marc") { |
783 |
|
|
784 |
use MARC; |
require MARC; |
785 |
|
|
786 |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
787 |
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!"; |
788 |
|
|
789 |
# optional argument is format |
# optional argument is format |
790 |
my $format = x($config->{format}) || 'usmarc'; |
my $format = x($config->{marc_format}) || 'usmarc'; |
791 |
|
|
792 |
print STDERR "Reading MARC file '$marc_file'\n"; |
print STDERR "Reading MARC file '$marc_file'\n"; |
793 |
|
|
794 |
my $marc = new MARC; |
my $marc = new MARC; |
795 |
my $nr = $marc->openmarc({ |
my $nr = $marc->openmarc({ |
796 |
file=>$marc_file, format=>$format |
file=>$marc_file, format=>$format |
797 |
}) || die "Can't open MARC file '$marc_file'"; |
}) || die "Can't open MARC file '$marc_file' with format '$format'"; |
798 |
|
|
799 |
|
# read MARC file in memory |
800 |
|
$marc->nextmarc(-1); |
801 |
|
|
802 |
my $i=0; # record nr. |
my $max_rec = $marc->marc_count(); |
803 |
|
|
804 |
my $rec; |
for(my $i=1; $i<=$max_rec; $i++) { |
805 |
|
|
806 |
while ($marc->nextmarc(1)) { |
progress($i,$max_rec); |
807 |
|
|
808 |
# XXX |
# store value for marc_sf.pm |
809 |
fakeprogress($i++); |
$main::cache->{marc_record} = $i; |
810 |
|
|
811 |
my $swishpath = $database."#".$i; |
my $swishpath = $database."#".$i; |
812 |
|
|
818 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
819 |
} |
} |
820 |
} |
} |
821 |
|
|
822 |
|
print STDERR "\n"; |
823 |
|
|
824 |
} elsif ($type_base eq "feed") { |
} elsif ($type_base eq "feed") { |
825 |
|
|
826 |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |