5 |
use Getopt::Std; |
use Getopt::Std; |
6 |
use Data::Dumper; |
use Data::Dumper; |
7 |
use XML::Simple; |
use XML::Simple; |
|
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
|
8 |
use Text::Iconv; |
use Text::Iconv; |
9 |
use Config::IniFiles; |
use Config::IniFiles; |
10 |
use Encode; |
use Encode; |
16 |
|
|
17 |
my $config_file = $0; |
my $config_file = $0; |
18 |
$config_file =~ s/\.pl$/.conf/; |
$config_file =~ s/\.pl$/.conf/; |
19 |
|
$config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]); |
20 |
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
21 |
|
|
22 |
my $config; |
my $config; |
151 |
} |
} |
152 |
|
|
153 |
|
|
154 |
|
# init variables for different types |
155 |
|
sub init_visible_type($) { |
156 |
|
my $type = shift; |
157 |
|
|
158 |
|
# swish, swish_exact, display, index, index_lookup |
159 |
|
# swish and display defaults |
160 |
|
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
161 |
|
if (lc($type) eq "display") { |
162 |
|
$s = 0; |
163 |
|
} elsif (lc($type) eq "swish") { |
164 |
|
$d = 0; |
165 |
|
} elsif (lc($type) eq "index") { |
166 |
|
($s,$se,$d,$i) = (0,1,0,1); |
167 |
|
} elsif (lc($type) eq "swish_exact") { |
168 |
|
($s,$se,$d,$i) = (0,1,0,0); |
169 |
|
} elsif (lc($type) =~ /^lookup/) { |
170 |
|
($s,$se,$d,$i,$il) = (0,1,0,0,1); |
171 |
|
} elsif ($type) { |
172 |
|
print STDERR "WARNING: unknown type: $type\n"; |
173 |
|
} |
174 |
|
return ($s,$se,$d,$i,$il); |
175 |
|
} |
176 |
|
|
177 |
|
|
178 |
|
# convert |
179 |
|
# |
180 |
|
# <tag> |
181 |
|
# <delimiter>, </delimiter> |
182 |
|
# <value>200a</value> |
183 |
|
# </tag> |
184 |
|
# |
185 |
|
# to |
186 |
|
# |
187 |
|
# <tag delimiter=", ">200a</tag> |
188 |
|
# |
189 |
|
# but without loosing spaces in delimiter (becasue |
190 |
|
# new XML::Simple strips spaces in attribute values |
191 |
|
# as defined in XML specification) |
192 |
|
# |
193 |
|
sub unroll_x($) { |
194 |
|
my $x = shift; |
195 |
|
|
196 |
|
if (defined $x->{value}) { |
197 |
|
my ($v,$d) = ($x->{value}->{content}, $x->{delimiter}->{content}); |
198 |
|
delete $x->{value}; |
199 |
|
delete $x->{delimiter}; |
200 |
|
$x->{content} = $v; |
201 |
|
$x->{delimiter} = $d; |
202 |
|
} |
203 |
|
return $x; |
204 |
|
} |
205 |
|
|
206 |
# begin real work: go field by field |
# begin real work: go field by field |
207 |
foreach my $field (@sorted_tags) { |
foreach my $field (@sorted_tags) { |
208 |
|
|
236 |
|
|
237 |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
238 |
|
|
239 |
|
$x = unroll_x($x); |
240 |
|
|
241 |
my $format = x($x->{content}); |
my $format = x($x->{content}); |
242 |
my $delimiter = x($x->{delimiter}) || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
243 |
|
|
244 |
my $repeat_off = 0; # init repeatable offset |
my $repeat_off = 0; # init repeatable offset |
245 |
|
|
246 |
# swish, swish_exact, display, index, index_lookup |
my ($s,$se,$d,$i,$il) = init_visible_type($x->{type}); |
|
# swish and display defaults |
|
|
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
|
|
($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index"); |
|
|
($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact"); |
|
|
($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/); |
|
247 |
|
|
248 |
# what will separate last line from this one? |
# what will separate last line from this one? |
249 |
if ($display_data && $x->{append}) { |
if ($display_data && $x->{append}) { |
250 |
$line_delimiter = ' '; |
$line_delimiter = $delimiter; |
251 |
} elsif ($display_data) { |
} elsif ($display_data) { |
252 |
$line_delimiter = '<br/>'; |
$line_delimiter = '<br/>'; |
253 |
} |
} |
257 |
|
|
258 |
# placeholder for all repeatable entries for index |
# placeholder for all repeatable entries for index |
259 |
|
|
|
sub chk_eval($) { |
|
|
my $data = shift; |
|
|
return if (! defined($data)); |
|
|
if ($data && $data =~ s/\s*eval{([^}]+)}\s*//) { |
|
|
if (eval "$1") { |
|
|
return $data; |
|
|
} else { |
|
|
return undef; |
|
|
} |
|
|
} else { |
|
|
return $data; |
|
|
} |
|
|
} |
|
|
|
|
260 |
sub mkformat($$) { |
sub mkformat($$) { |
261 |
my $x = shift || die "mkformat needs tag reference"; |
my $x = shift || die "mkformat needs tag reference"; |
262 |
my $data = shift || return; |
my $data = shift || return; |
263 |
my $format_name = x($x->{format_name}) || return chk_eval($data); |
my $format_name = x($x->{format_name}) || return $data; |
264 |
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!"; |
265 |
my $format_delimiter = x($x->{format_delimiter}); |
my $format_delimiter = x($x->{format_delimiter}); |
266 |
my @data; |
my @data; |
273 |
if ($fmt) { |
if ($fmt) { |
274 |
my $nr = scalar $fmt =~ s/%s/%s/g; |
my $nr = scalar $fmt =~ s/%s/%s/g; |
275 |
if (($#data+1) == $nr) { |
if (($#data+1) == $nr) { |
276 |
return chk_eval(sprintf($fmt,@data)); |
return sprintf($fmt,@data); |
277 |
} else { |
} else { |
278 |
#print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
#print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
279 |
return chk_eval($data); |
return $data; |
280 |
} |
} |
281 |
} else { |
} else { |
282 |
print STDERR "usage of link '$format_name' without defined format (<link> tag)\n"; |
print STDERR "usage of link '$format_name' without defined format (<link> tag)\n"; |
325 |
} |
} |
326 |
# type="swish" ; field for swish |
# type="swish" ; field for swish |
327 |
if ($swish) { |
if ($swish) { |
328 |
|
my $tmp = $swish; |
329 |
if ($filter && ($s || $se)) { |
if ($filter && ($s || $se)) { |
330 |
no strict 'refs'; |
no strict 'refs'; |
331 |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
$tmp = join(" ",&$filter($tmp)) if ($s || $se); |
|
$swish_data .= $tmp if ($s); |
|
|
$swish_exact_data .= "xxbxx $tmp xxexx " if ($se && $tmp ne ""); |
|
|
|
|
|
} else { |
|
|
$swish_data .= $swish if ($s); |
|
|
$swish_exact_data .= "xxbxx $swish xxexx " if ($se && $swish ne ""); |
|
332 |
} |
} |
333 |
|
|
|
$tmp = chk_eval($tmp); |
|
334 |
$swish_data .= $tmp if ($s && $tmp); |
$swish_data .= $tmp if ($s && $tmp); |
335 |
$swish_exact_data .= "xxbxx $tmp xxexx " if ($se && $tmp ne ""); |
$swish_exact_data .= "xxbxx $tmp xxexx " if ($tmp && $tmp ne "" && $se); |
336 |
} |
} |
337 |
|
|
338 |
# type="display" ; field for display |
# type="display" ; field for display |
365 |
no strict 'refs'; |
no strict 'refs'; |
366 |
$idisplay = &$filter($idisplay); |
$idisplay = &$filter($idisplay); |
367 |
} |
} |
|
$idisplay = chk_eval($idisplay); |
|
368 |
push @index_data, $idisplay if ($idisplay && !$iterate_by_page); |
push @index_data, $idisplay if ($idisplay && !$iterate_by_page); |
369 |
} |
} |
370 |
|
|
464 |
# now try to parse variables from configuration file |
# now try to parse variables from configuration file |
465 |
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
466 |
|
|
467 |
|
$x = unroll_x($x); |
468 |
|
|
469 |
my $delimiter = x($x->{delimiter}) || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
470 |
my $val = $cfg->val($database, x($x->{content})); |
my $val = $cfg->val($database, x($x->{content})); |
471 |
|
|
472 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
# FIXME index_lookup is not supported! |
473 |
$s = 0 if (lc($x->{type}) eq "display"); |
my ($s,$se,$d,$i,$il) = init_visible_type($x->{type}); |
|
$d = 0 if (lc($x->{type}) eq "swish"); |
|
|
# no support for swish exact in config. |
|
|
# IMHO, it's useless |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
474 |
|
|
475 |
if ($val) { |
if ($val) { |
476 |
$display_data .= $delimiter.$val if ($d); |
$display_data .= $delimiter.$val if ($d); |
477 |
$swish_data .= $val if ($s); |
$swish_data .= " ".$val if ($s); |
478 |
$index->insert($field, $val, $path) if ($i); |
$index->insert($field, $val, $path) if ($i); |
479 |
} |
} |
480 |
|
|
527 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
528 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
529 |
|
|
530 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
$xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data)); |
531 |
} |
} |
532 |
|
|
533 |
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
537 |
|
|
538 |
# add delimiters before and after word. |
# add delimiters before and after word. |
539 |
# That is required to produce exact match |
# That is required to produce exact match |
540 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
$xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data)); |
541 |
} |
} |
542 |
|
|
543 |
my $idel = $cache->{index_delimiter}->{$field}; |
my $idel = $cache->{index_delimiter}->{$field}; |
570 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
571 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
572 |
|
|
573 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
$xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data)); |
574 |
} |
} |
575 |
|
|
576 |
if ($swish_exact_data) { |
if ($swish_exact_data) { |
579 |
|
|
580 |
# add delimiters before and after word. |
# add delimiters before and after word. |
581 |
# That is required to produce exact match |
# That is required to produce exact match |
582 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
$xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data)); |
583 |
} |
} |
584 |
} |
} |
585 |
} |
} |
616 |
|
|
617 |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
618 |
|
|
619 |
my $unac_filter = $cfg_global->val('global', 'unac_filter'); |
my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter'); |
620 |
if ($unac_filter) { |
if ($my_unac_filter) { |
621 |
require $unac_filter; |
print STDERR "using $my_unac_filter to filter characters for search\n"; |
622 |
|
require $my_unac_filter; |
623 |
|
} else { |
624 |
|
print STDERR "### fallback to default my_unac_string!\n"; |
625 |
|
eval q{ |
626 |
|
sub main::my_unac_string($$) { |
627 |
|
my ($charset, $string) = (@_); |
628 |
|
return $string; |
629 |
|
} |
630 |
|
}; |
631 |
} |
} |
632 |
|
|
633 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
639 |
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
640 |
if ($lookup_file) { |
if ($lookup_file) { |
641 |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
642 |
|
if (! -e $lookup_file) { |
643 |
|
open(LOOKUP, "> $lookup_file") || die "can't create $lookup_file': $!"; |
644 |
|
close(LOOKUP); |
645 |
|
} |
646 |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
647 |
print STDERR "creating lookup file '$lookup_file'\n"; |
print STDERR "creating lookup file '$lookup_file'\n"; |
648 |
# delete memory cache for lookup file |
# delete memory cache for lookup file |
663 |
my $type_base = $type; |
my $type_base = $type; |
664 |
$type_base =~ s/_.+$//g; |
$type_base =~ s/_.+$//g; |
665 |
|
|
666 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1); |
$config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 ); |
667 |
|
|
668 |
# output current progress indicator |
# output current progress indicator |
669 |
my $last_p = 0; |
my $last_p = 0; |
933 |
|
|
934 |
all2xml.pl - read various file formats and dump XML for SWISH-E |
all2xml.pl - read various file formats and dump XML for SWISH-E |
935 |
|
|
936 |
|
=head1 SYNOPSYS |
937 |
|
|
938 |
|
$ all2xml.pl [test.conf] |
939 |
|
|
940 |
=head1 DESCRIPTION |
=head1 DESCRIPTION |
941 |
|
|
942 |
This command will read ISIS data file using OpenIsis perl module, MARC |
This command will read ISIS data file using OpenIsis perl module, MARC |
945 |
this script B<isn't general xml generator> from isis files (isis allready |
this script B<isn't general xml generator> from isis files (isis allready |
946 |
has something like that). Output of this script is tailor-made for SWISH-E. |
has something like that). Output of this script is tailor-made for SWISH-E. |
947 |
|
|
948 |
|
If no configuration file is specified, it will use default one called |
949 |
|
C<all2xml.conf>. |
950 |
|
|
951 |
=head1 BUGS |
=head1 BUGS |
952 |
|
|
953 |
Documentation is really lacking. However, in true Open Source spirit, source |
Documentation is really lacking. However, in true Open Source spirit, source |