--- trunk/all2xml.pl 2004/03/14 11:50:29 274
+++ branches/biomed/all2xml.pl 2004/09/29 18:56:26 489
@@ -5,7 +5,6 @@
use Getopt::Std;
use Data::Dumper;
use XML::Simple;
-use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
use Text::Iconv;
use Config::IniFiles;
use Encode;
@@ -17,6 +16,7 @@
my $config_file = $0;
$config_file =~ s/\.pl$/.conf/;
+$config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]);
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
my $config;
@@ -150,6 +150,8 @@
}
}
+
+ # init variables for different types
sub init_visible_type($) {
my $type = shift;
@@ -166,10 +168,41 @@
($s,$se,$d,$i) = (0,1,0,0);
} elsif (lc($type) =~ /^lookup/) {
($s,$se,$d,$i,$il) = (0,1,0,0,1);
+ } elsif ($type) {
+ print STDERR "WARNING: unknown type: $type\n";
}
return ($s,$se,$d,$i,$il);
}
+
+ # convert
+ #
+ #
+ # ,
+ # 200a
+ #
+ #
+ # to
+ #
+ # 200a
+ #
+ # but without loosing spaces in delimiter (becasue
+ # new XML::Simple strips spaces in attribute values
+ # as defined in XML specification)
+ #
+ sub unroll_x($) {
+ my $x = shift;
+
+ if (defined $x->{value}) {
+ my ($v,$d) = ($x->{value}->{content}, $x->{delimiter}->{content});
+ delete $x->{value};
+ delete $x->{delimiter};
+ $x->{content} = $v;
+ $x->{delimiter} = $d;
+ }
+ return $x;
+ }
+
# begin real work: go field by field
foreach my $field (@sorted_tags) {
@@ -203,6 +236,8 @@
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
+ $x = unroll_x($x);
+
my $format = x($x->{content});
my $delimiter = x($x->{delimiter}) || ' ';
@@ -212,7 +247,7 @@
# what will separate last line from this one?
if ($display_data && $x->{append}) {
- $line_delimiter = ' ';
+ $line_delimiter = $delimiter;
} elsif ($display_data) {
$line_delimiter = '
';
}
@@ -429,6 +464,8 @@
# now try to parse variables from configuration file
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
+ $x = unroll_x($x);
+
my $delimiter = x($x->{delimiter}) || ' ';
my $val = $cfg->val($database, x($x->{content}));
@@ -490,7 +527,7 @@
$swish_data =~ s/ +/ /g;
$swish_data =~ s/ +$//g;
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
}
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
@@ -500,7 +537,7 @@
# add delimiters before and after word.
# That is required to produce exact match
- $xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data));
+ $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
}
my $idel = $cache->{index_delimiter}->{$field};
@@ -533,7 +570,7 @@
$swish_data =~ s/ +/ /g;
$swish_data =~ s/ +$//g;
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
}
if ($swish_exact_data) {
@@ -542,7 +579,7 @@
# add delimiters before and after word.
# That is required to produce exact match
- $xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data));
+ $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
}
}
}
@@ -579,9 +616,18 @@
my $show_progress = $cfg_global->val('global', 'show_progress');
-my $unac_filter = $cfg_global->val('global', 'unac_filter');
-if ($unac_filter) {
- require $unac_filter;
+my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter');
+if ($my_unac_filter) {
+ print STDERR "using $my_unac_filter to filter characters for search\n";
+ require $my_unac_filter;
+} else {
+ print STDERR "### fallback to default my_unac_string!\n";
+ eval q{
+ sub main::my_unac_string($$) {
+ my ($charset, $string) = (@_);
+ return $string;
+ }
+ };
}
foreach my $database ($cfg->Sections) {
@@ -593,6 +639,10 @@
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
if ($lookup_file) {
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
+ if (! -e $lookup_file) {
+ open(LOOKUP, "> $lookup_file") || die "can't create $lookup_file': $!";
+ close(LOOKUP);
+ }
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
print STDERR "creating lookup file '$lookup_file'\n";
# delete memory cache for lookup file
@@ -613,7 +663,7 @@
my $type_base = $type;
$type_base =~ s/_.+$//g;
- $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
+ $config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 );
# output current progress indicator
my $last_p = 0;
@@ -763,7 +813,11 @@
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
my $cell = $oWorksheet->{Cells}[$iR][$iC];
if ($cell) {
- $row->{int2col($iC)} = $cell->Value;
+ # this conversion is a cludge.
+ # Files from Excell could have
+ # characters which don't fit into
+ # destination encoding.
+ $row->{int2col($iC)} = $utf2cp->convert($cell->Value) || $cell->Value;
}
}
@@ -883,6 +937,10 @@
all2xml.pl - read various file formats and dump XML for SWISH-E
+=head1 SYNOPSYS
+
+ $ all2xml.pl [test.conf]
+
=head1 DESCRIPTION
This command will read ISIS data file using OpenIsis perl module, MARC
@@ -891,6 +949,9 @@
this script B from isis files (isis allready
has something like that). Output of this script is tailor-made for SWISH-E.
+If no configuration file is specified, it will use default one called
+C.
+
=head1 BUGS
Documentation is really lacking. However, in true Open Source spirit, source