--- branches/tehnika/all2xml.pl	2004/03/12 15:33:13	264
+++ branches/tehnika/all2xml.pl	2004/12/08 18:45:32	601
@@ -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 $config;
@@ -139,18 +139,70 @@
 		} else {
 			print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
 		}
+
 		if ($field_name) {
+			$field_name = x($field_name);
 			if (! $last_field_name) {
-				$last_field_name = x($field_name);
+				$last_field_name = $field_name;
 				return $last_field_name;
 			} elsif ($field_name ne $last_field_name) {
-				$last_field_name = x($field_name);
+				$last_field_name = $field_name;
 				return $last_field_name;
 			}
 		}
 	}
 
 
+	# init variables for different types
+	sub init_visible_type($) {
+		my $type = shift;
+
+		# swish, swish_exact, display, index, index_lookup
+		# swish and display defaults
+		my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
+		if (lc($type) eq "display") {
+			$s = 0;
+		} elsif (lc($type) eq "swish") {
+			$d = 0;
+		} elsif (lc($type) eq "index") {
+			($s,$se,$d,$i) = (0,1,0,1);
+		} elsif (lc($type) eq "swish_exact") {
+			($s,$se,$d,$i) = (0,1,0,0);
+		} elsif (lc($type) =~ /^lookup/) {
+			($s,$se,$d,$i,$il) = (0,1,0,0,1);
+		}
+		return ($s,$se,$d,$i,$il);
+	}
+
+
+	# convert
+	#
+	# <tag>
+	#  <delimiter>, </delimiter>
+	#  <value>200a</value>
+	# </tag>
+	# 
+	# to
+	#
+	# <tag delimiter=", ">200a</tag>
+	#
+	# 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) {
 
@@ -184,23 +236,18 @@
 
 		foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
 
+			$x = unroll_x($x);
+
 			my $format = x($x->{content});
 			my $delimiter = x($x->{delimiter}) || ' ';
 
 			my $repeat_off = 0;	# init repeatable offset
 
-			# swish, swish_exact, display, index, index_lookup
-			# 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/);
+			my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
 
 			# what will separate last line from this one?
 			if ($display_data && $x->{append}) {
-				$line_delimiter = ' ';
+				$line_delimiter = $delimiter;
 			} elsif ($display_data) {
 				$line_delimiter = '<br/>';
 			}
@@ -417,19 +464,17 @@
 		# 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}));
 
-			my ($s,$d,$i) = (1,1,0);	# swish, display default
-			$s = 0 if (lc($x->{type}) eq "display");
-			$d = 0 if (lc($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");
+			# FIXME index_lookup is not supported!
+			my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
 
 			if ($val) {
 				$display_data .= $delimiter.$val if ($d);
-				$swish_data .= $val if ($s);
+				$swish_data .= " ".$val if ($s);
 				$index->insert($field, $val, $path) if ($i);
 			}
 
@@ -482,7 +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];
@@ -492,7 +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};
@@ -525,7 +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) {
@@ -534,7 +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));
 			}
 		}
 	}
@@ -571,9 +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) {
@@ -585,6 +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
@@ -605,7 +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;
@@ -755,7 +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;
 				}
 			}
 
@@ -875,6 +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
@@ -883,6 +949,9 @@
 this script B<isn't general xml generator> 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<all2xml.conf>.
+
 =head1 BUGS
 
 Documentation is really lacking. However, in true Open Source spirit, source