Update search_xapian.pl

This patch remedies two issues:
1. Long-text fields that have more than 200 characters (e.g. the abstract of an eprint) are now indexed. Before, they were just left out, resulting in wrong search results when one did a simple search (e.g. using a search term such as abstract:dyscalculia)
2. Sorting by date after simple search now works correctly. Before, sort order was chaotic because dates were weighted by term within-document-frequency (wdf).  Dates are now stripped down to years and stored using the add_boolean_term method (name is misleading, has nothing to do with booleans), that assigns wdf = 0 to the term.

lib/cfg.d/search_xapian.pl

@@ -99,15 +99,31 @@
 			next if !EPrints::Utils::is_set( $value );
 			$tg->index_text( $value );
 			$tg->increase_termpos();
-			next if length($value) > 200; # Xapian term length limit-ish
+
+			# Allow indexing of long text fields (e.g. abstracts that are longer than 200 chars)
+			# The Xapian length limit applies to only a single term with 200 consecutive chars without white-space
+			# next if length($value) > 200; # Xapian term length limit-ish
 			if( $field->isa( "EPrints::MetaField::Text" ) || $field->isa( "EPrints::MetaField::Name" ) )
 			{
 				$tg->index_text( $value, 2, $prefix );
 				$tg->increase_termpos();
 			}
 			else
 			{
-				$doc->add_term( $prefix . $value );
+				# Improve indexing: Long term values must be filtered
+				# Dates are stripped to year only and are added with wdf = 0
+				# This enables that records are correctly sorted by date
+				next if length($value) > 200; # Xapian term length limit-ish
+
+				if ($field->name eq 'date')
+				{
+					$value =~ /^(\d{4})/;
+					$doc->add_boolean_term( $prefix . $value );
+				}
+				else 
+				{
+					$doc->add_term( $prefix . $value );
+				}
 			}
 		}
 		foreach my $langid (@{$repo->config( "languages" )})