From: Mischa POSLAWSKY <perl@shiar.org>
Date: Sat, 8 Apr 2017 15:02:37 +0000 (+0200)
Subject: charset: relevant unicode blocks in language comparisons
X-Git-Tag: v1.10~207
X-Git-Url: http://git.shiar.nl/sheet.git/commitdiff_plain/276a672c5fd261773a452d3bca765c7c7e4e562f

charset: relevant unicode blocks in language comparisons
---

diff --git a/charset.plp b/charset.plp
index 7dc1aee..6119859 100644
--- a/charset.plp
+++ b/charset.plp
@@ -48,10 +48,12 @@ sub tabinput {
 		turkish    => [qw( iso-8859-9 iso-8859-3 cp1254 cp857 MacTurkish )],
 		baltic     => [qw( iso-8859-4 iso-8859-13 cp1257 cp775 )],
 		nordic     => [qw( iso-8859-10 cp865 cp861 MacIcelandic MacSami )],
-		cyrillic   => [qw( koi8-r koi8-u koi8-f iso-8859-5 cp1251 MacCyrillic cp866 cp855 )], # MacUkrainian is broken
-		arabic     => [qw( iso-8859-6 cp1256 MacArabic cp864 cp1006 MacFarsi )],
-		greek      => [qw( iso-8859-7 cp1253 MacGreek cp737 cp869 )],
-		hebrew     => [qw( iso-8859-8 cp1255 MacHebrew cp862 )],
+		cyrillic   => [qw( koi8-r koi8-u koi8-f iso-8859-5 cp1251 MacCyrillic cp866 cp855
+		                   U4 U2DE-2DF UA64-A69 U50-52 )], # MacUkrainian is broken
+		arabic     => [qw( iso-8859-6 cp1256 MacArabic cp864 cp1006 MacFarsi
+		                   U6 U8A-8F+0-31+64 U75-77 )],
+		greek      => [qw( iso-8859-7 cp1253 MacGreek cp737 cp869  U37-3F U1F )],
+		hebrew     => [qw( iso-8859-8 cp1255 MacHebrew cp862  U59-5F )],
 		thai       => [qw( iso-8859-11 cp874 MacThai )],
 		vietnamese => [qw( viscii cp1258 MacVietnamese )],
 		symbol     => [qw( symbol dingbats MacDingbats AdobeZdingbat AdobeSymbol )],
@@ -193,13 +195,14 @@ sub tabinput {
 		}
 	}
 
-	if ($input =~ /^U([0-9a-f]+)(?:-([0-9a-f]+))?/) {
+	if ($input =~ /^U([0-9a-fA-F]+)(?:-([0-9a-fA-F]+))?/) {
 		my $start = hex($1) << ($2 ? 4 : 8);
-		my $end = $2 ? hex($2) << 4 : $start + 240;
-		$row{table} = join '', map { chr } $start .. $end+15;
+		my $end = $2 ? (hex($2) << 4) + $row{cols} - 1 : $start + 255;
+		$row{table} = join '', map { chr } $start .. $end;
 		utf8::upgrade($row{table});  # prevent latin1 output
-		$row{endpoint} = $end + 14 - $start;
+		$row{endpoint} = $end - $start;
 		$row{set} = sprintf 'Unicode block U+%02Xxx', $start >> 8;
+		$row{offset} = $start % 256;
 	}
 	elsif ($input eq 'U') {
 		$row{set} = 'Unicode planes';