--- trunk/all2xml.pl	2005/02/27 23:07:35	678
+++ trunk/all2xml.pl	2006/10/29 16:34:25	750
@@ -1,7 +1,7 @@
 #!/usr/bin/perl -w
 
 use strict;
-use Biblio::Isis;
+use Biblio::Isis 0.23;
 use Getopt::Std;
 use Data::Dumper;
 use XML::Simple;
@@ -64,7 +64,6 @@
 	'excel' => 'column',
 	'marc' => 'marc',
 	'feed' => 'feed',
-	'dbf' => 'isis',	# special case, re-use isis import_xml
 );
 
 my $cache;	# for cacheing
@@ -77,6 +76,9 @@
 
 my $last_field_name;	# cache to prevent repeated fields
 
+my $broken_cdata = XMLin('<foo><![CDATA[<bar>]]></foo>') eq '<bar>>';
+warn "XML::Simple on this system seems broken with <![CDATA[..]]>.\n" if ($broken_cdata);
+
 sub data2xml {
 
 	use xmlify;
@@ -127,6 +129,7 @@
 	delete $cache->{swish_exact_data};
 	delete $cache->{index_data};
 	delete $cache->{index_delimiter};
+	delete $cache->{distinct};
 	my @page_fields;	# names of fields
 
 
@@ -207,6 +210,7 @@
 			delete $x->{value};
 			delete $x->{delimiter};
 			$x->{content} = $v;
+			$d =~ s#>$## if ($d && $broken_cdata);
 			$x->{delimiter} = $d;
 		}
 		return $x;
@@ -226,7 +230,7 @@
 
 		my ($swish,$display);
 
-		my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
+		my $tag = $cfg->val($database, 'import_xml_tag') || $type2tag{$type} || die "can't find which tag to use for type $type";
 
 		# is this field page-by-page?
 		my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
@@ -235,6 +239,11 @@
 		# default line_delimiter if using
 		my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
 		$cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter};
+		my $distinct = $config->{indexer}->{$field}->{distinct};
+		if ($distinct && !$iterate_by_page) {
+			warn "WARNING: distinct is currently not supported without iterate_by_page!\n";
+			$distinct = 0;
+		}
 
 		my $format_name = $config->{indexer}->{$field}->{format_name};
 		my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter};
@@ -413,9 +422,10 @@
 						$ldel = " " if ($append);
 #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel);
 						if (! $cache->{$what}->{$field}->[$page]) {
-							$cache->{$what}->{$field}->[$page] = $data;
-						} else {
-							$cache->{$what}->{$field}->[$page] .= $ldel.$data;
+							push @{$cache->{$what}->{$field}->[$page]}, {
+								data => $data,
+								delimiter => $ldel,
+							};
 						}
 					}
 
@@ -482,7 +492,7 @@
 			if ($val) {
 				$display_data .= $delimiter.$val if ($d);
 				$swish_data .= " ".$val if ($s);
-				$index->insert($field, $val, $path) if ($i);
+				$index->insert($field, $val, $val, $path) if ($i);
 			}
 
 			if ($iterate_by_page) {
@@ -490,15 +500,15 @@
 				# on first page!!!
 				my $page = 0;
 				if ($display_data) {
-					$cache->{display_data}->{$field}->[$page] = $display_data;
+					push @{$cache->{display_data}->{$field}->[$page]}, { data => $display_data };
 					$display_data = "";
 				}
 				if ($swish_data) {
-					$cache->{swish_data}->{$field}->[$page] = $swish_data;
+					push @{$cache->{swish_data}->{$field}->[$page]}, { data => $swish_data };
 					$swish_data = "";
 				}
 				if ($swish_exact_data) {
-					$cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
+					push @{$cache->{swish_exact_data}->{$field}->[$page]}, { data => $swish_exact_data };
 					$swish_exact_data = "";
 				}
 			}
@@ -509,14 +519,28 @@
 			my $nr_pages = $page_max{$field} || next;
 #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
 #print STDERR Dumper($cache->{display_data});
+			my $seen;	# used for distinct
 			for (my $page=0; $page <= $nr_pages; $page++) {
 				my $display_data;
-				if ($cache->{format}->{$field}) {
-					my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
-					$display_data=$tmp if ($tmp);
-				} else {
-					$display_data = $cache->{display_data}->{$field}->[$page];
+				my $delimiter = '';
+				foreach my $element (@{ $cache->{display_data}->{$field}->[$page] }) {
+					my $data = $element->{data};
+					die "BUG! no data in element?" unless ($data);
+
+					if ($distinct) {
+						next if ($cache->{distinct}->{$field}->{ $data });
+						$cache->{distinct}->{$field}->{ $data } = 1;
+					}
+
+					if ($cache->{format}->{$field}) {
+						my $tmp = mkformat($cache->{format}->{$field},$data);
+						$display_data .= $delimiter . $tmp if ($tmp);
+					} else {
+						$display_data .= $delimiter . $data;
+					}
+					$delimiter = $element->{delimiter} if ($element->{delimiter});
 				}
+
 				if ($display_data) { # default
 					if ($field eq "headline") {
 						$xml .= xmlify("headline", $display_data);
@@ -528,7 +552,7 @@
 					}
 				}
 				
-				my $swish_data = $cache->{swish_data}->{$field}->[$page];
+				my $swish_data = join(" ",map { $_->{data} } @{ $cache->{swish_data}->{$field}->[$page] });
 				if ($swish_data) {
 					# remove extra spaces
 					$swish_data =~ s/ +/ /g;
@@ -537,7 +561,7 @@
 					$xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
 				}
 
-				my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
+				my $swish_exact_data = join(" ", map { $_->{data} } @{ $cache->{swish_exact_data}->{$field}->[$page] });
 				if ($swish_exact_data) {
 					$swish_exact_data =~ s/ +/ /g;
 					$swish_exact_data =~ s/ +$//g;
@@ -668,7 +692,8 @@
 		print STDERR "opening lookup file '$lookup_file'\n";
 	}
 
-	my $import_xml_file = "./import_xml/$type.xml";
+	my $import_xml_type = $cfg->val($database, 'import_xml_file') || $type;
+	my $import_xml_file = "./import_xml/$import_xml_type.xml";
 
 	if (! -r $import_xml_file) {
 		print STDERR "ERROR: file $import_xml_file not readable skipping!\n";
@@ -681,7 +706,15 @@
 	my $type_base = $type;
 	$type_base =~ s/_.+$//g;
 
-	$config=XMLin($import_xml_file, ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 );
+	my $tag = $cfg->val($database, 'import_xml_tag') || $type2tag{$type_base} || die "can't find which tag to use for type $type";
+	$config=XMLin($import_xml_file, ForceArray => [ $tag, 'config', 'format' ], ForceContent => 1 );
+
+	# check for broken XML::Simple
+	if ( $broken_cdata ) {
+		map {
+			$config->{format}->{$_}->{content} =~ s#>$##;
+		} keys %{ $config->{format} };
+	}
 
 	# helper for progress bar
 	sub fmt_time {
@@ -743,7 +776,6 @@
 	}
 
 	# now read database
-print STDERR "using: $type...\n";
 
 	# erase cache for tags by order in this database
 	delete $cache->{tags_by_order};
@@ -753,7 +785,10 @@
 		my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
 
 		$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
-		my $db = new Biblio::Isis( isisdb => $isis_db );
+		my $db = new Biblio::Isis(
+			isisdb => $isis_db,
+			join_subfields_with => ' ; ',
+		);
 
 		if (! $db) {
 			print STDERR "FATAL: can't read ISIS database: $isis_db, skipping...\n";
@@ -795,7 +830,8 @@
 		my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
 
 		my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
-		my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
+		my $start_row = x($config->{start_row}) || die "no start_row in $type.xml";
+		$start_row--;
 
 		my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";