X-Git-Url: http://git.shiar.nl/sheet.git/blobdiff_plain/efc30e581f90743618f1bbf44fe0bfae075549ef..b690e02814ec7b65b10789fac83cd7c8b2b72b30:/charset-encoding.inc.pl

diff --git a/charset-encoding.inc.pl b/charset-encoding.inc.pl
index 9ae8523..6e46477 100644
--- a/charset-encoding.inc.pl
+++ b/charset-encoding.inc.pl
@@ -1,83 +1,219 @@
 use 5.014;
 use warnings;
+use utf8;
 
 +{
-	'ascii'       => ['' => '00-7F'],
-	'cp437'       => ['cp850' => 0, 'ascii' => '00-1F+80'], # ascii range overridden later
-	'gsm0338'     => ['ascii' => '00-7F'],
-	'dingbats'    => ['' => '20-7F+A0'],
-	'MacDingbats' => ['dingbats' => '80-9F'],
-	'AdobeZdingbat'=> ['MacDingbats' => '80-9F'], # should be identical but maps to private use
-	'symbol'      => ['' => '20-7F+A0'],
-	'AdobeSymbol' => ['symbol' => '20-7F+A0', '' => '20-7F+A0'], # minor differences, irrelevant except for different '€'
-	'wingdings'   => ['' => '20'],
-	'wingdings2'  => ['' => '20'],
-	'wingdings3'  => ['' => '20'],
-	'webdings'    => ['' => '20'],
-
-	'iso-8859-2'  => ['iso-8859-1' => 'A0'],
-	'iso-8859-3'  => ['iso-8859-1' => 'A0'], #TODO: also apply to iso-8859-9
-	'iso-8859-4'  => ['iso-8859-2' => 'A0'],
-	'iso-8859-5'  => ['iso-8859-1' => 'A0'],
-	'iso-8859-6'  => ['cp1256' => '80', 'iso-8859-1' => 'A0'],
-	'iso-8859-7'  => ['iso-8859-1' => 'A0'],
-	'iso-8859-8'  => ['iso-8859-1' => 'A0'],
-	'iso-8859-9'  => ['iso-8859-1' => 'D0-DF+F0'],
-	'iso-8859-10' => ['iso-8859-4' => 'A0'],
-	'iso-8859-11' => ['iso-8859-1' => 'A0'],
-	'iso-8859-13' => ['iso-8859-4' => 'A0'],
-	'iso-8859-14' => ['iso-8859-1' => 'A0'],
-	'iso-8859-15' => ['iso-8859-1' => 'A0-BF'],
-	'iso-8859-16' => ['iso-8859-2' => 'A0'],
-	'hp-roman8'   => ['iso-8859-1' => 'A0'],
-
-	'cp1252'      => ['iso-8859-1' => '80-9F'],
-	'cp1250'      => ['iso-8859-2' => '80-BF', 'cp1252' => '80'],
-	'cp1254'      => ['iso-8859-9' => '80-9F', 'cp1252' => '80-9F+D0'],
-	'cp874'       => ['iso-8859-11' => '80-9F', 'cp1252' => '80'], # windows-874 actually cp1162
-	'cp1257'      => ['iso-8859-13' => '80-9F+FF', 'cp1252' => '80'],
-	'cp1251'      => ['cp1252' => '80'],
-	'cp1253'      => ['cp1252' => '80'],
-	'cp1255'      => ['iso-8859-8' => '80-DF', 'cp1252' => '80'],
-	'cp1256'      => ['cp1252' => '80'],
-	'cp1258'      => ['cp1252' => '80-9F+C0'],
-
-	'cp850'       => ['cp437' => '90'],
-	'cp860'       => ['cp437' => '80-AF'],
-	'cp861'       => ['cp865' => '80-AF'],
-	'cp863'       => ['cp437' => '80-AF'],
-	'cp865'       => ['cp437' => '90-AF'],
-	'cp852'       => ['cp850' => '80', 'cp437' => '80'],
-	'cp857'       => ['cp850' => '80-AF+D0-EF', 'cp437' => '80'],
-	'cp775'       => ['cp850' => '80'],  # partial cp437
-	'cp866'       => ['cp437' => '80-AF+E0'],
-	'cp855'       => ['cp437' => '80'],
-	'cp1006'      => ['iso-8859-6' => 'A0', 'cp437' => '80'],
-	'cp737'       => ['cp437' => '80-AF+E0'],
-	'cp869'       => ['cp437' => '80'],
-	'cp862'       => ['cp437' => '80-9F'],
-	'cp864'       => ['MacArabic' => '80', 'iso-8859-6' => '80', 'cp437' => '80'], #TODO: compare form variants
-
-	'koi8-u'      => ['koi8-r' => '90-BF'],
-	'koi8-f'      => ['koi8-u' => '90-BF'],
-
-	'MacRomanian' => ['MacRoman' => 'A0-BF+D0-DF'],
-	'MacRumanian' => ['MacRomanian' => 'A0-BF+D0-DF', 'MacRoman' => 'A0-BF+D0-DF'],
-	'MacCroatian' => ['MacRoman' => 'A0'],
-	'MacCentralEurRoman' => ['MacRoman' => '80'],
-	'MacIcelandic'=> ['MacRoman' => 'A0-AF+D0-EF'],
-	'MacTurkish'  => ['MacRoman' => 'D0-DF'], # F5 is unassigned
-	'MacSami'     => ['MacIcelandic' => '90-BF+D0-DF+F0', 'MacRoman' => '90'],
-	'MacGreek'    => ['MacRoman' => '80'],
-	'MacCyrillic' => ['MacRoman' => '80'],
-	'MacHebrew'   => ['iso-8859-8' => '80', 'MacRoman' => '80-8F+A0'], # partial ascii
-	'MacArabic'   => ['iso-8859-6' => '80', 'cp864' => '80', 'MacRoman' => '80'], #TODO: multiple parents
-	'MacFarsi'    => ['MacArabic' => 'B0-BF', 'MacRoman' => '80'],
-
-	'cp37'        => ['posix-bc' => '00'],
-	'posix-bc'    => ['cp1047' => '40'],
-	'cp500'       => ['cp37' => '40-5F+B0-BF'],
-	'cp1047'      => ['cp37' => '10-2F+50-5F+A0-BF'],
-	'cp1026'      => ['cp37' => '40'],
-	'cp875'       => ['cp37' => '30'],
+	default    => [qw( u+0-27F utf8+realsize iso-8859-1 iso-8859-15 cp1252 cp437 cp850 )],
+	unicode    => [qw( uu+cols=32+realsize u+0-FFF u+1000-18AF u+18B0-1FFF )],
+	us         => [qw( cp437 cp863 gsm0338 AdobeStandardEncoding )],
+	ebcdic     => [qw( cp37 cp500 cp1047 posix-bc cp1026 cp875 )],
+	iso        => [map {"iso-8859-$_"} 1 .. 11, 13 .. 16],
+	dos        => [qw( cp437 cp865 cp861 cp860 cp863 cp850 cp857 cp852 cp775
+	                   cp737 cp869 cp866 cp855 cp862 cp864 )],
+	aix        => [qw( cp1006 )],
+	win        => [qw( cp1252 cp1250 cp1254 cp1257 cp1258 cp1253 cp1251 cp1255 cp1256 cp874 )],
+	mac        => [qw( MacRoman MacRomanian MacRumanian MacCroatian MacCentralEurRoman MacTurkish MacIcelandic MacSami
+	                   MacGreek MacCyrillic MacHebrew MacArabic MacFarsi MacThai )],
+
+	# languages
+	westeur    => [qw( iso-8859-1 iso-8859-15 cp1252 iso-8859-14 cp850 hp-roman8 nextstep MacRoman )],
+	centeur    => [qw( iso-8859-2 iso-8859-16 cp1250 cp852 MacRomanian MacCroatian MacCentralEurRoman )], # MacRumanian only for DB
+	turkish    => [qw( iso-8859-9 iso-8859-3 cp1254 cp857 MacTurkish )],
+	norteur    => [qw( baltic nordic )],
+	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
+	                   +400 +2DE0 +A640-A69F +500-52F )], # MacUkrainian is broken
+	arabic     => [qw( iso-8859-6 cp1256 MacArabic cp864 cp1006 MacFarsi
+	                   +600 +8A0-8BF+8E0 +750-77F )],
+	greek      => [qw( iso-8859-7 cp1253 MacGreek cp737 cp869  +370 +1F00 )],
+	hebrew     => [qw( iso-8859-8 cp1255 MacHebrew cp862  +590 )],
+	thai       => [qw( iso-8859-11 cp874 MacThai )], # U+0E00 identical to iso-8859-11
+	vietnamese => [qw( viscii cp1258 MacVietnamese )],
+	symbols    => [qw( symbol dingbats MacDingbats wingdings wingdings2 wingdings3 webdings )],
+
+	# iso-code shorthand
+	1 => 'westeur',
+	2 => 'centeur',
+	3 => 'turkish',
+	4 => 'baltic',
+	5 => 'cyrillic',
+	6 => 'arabic',
+	7 => 'greek',
+	8 => 'hebrew',
+	9 => 'turkish',
+	10 => 'nordic',
+	11 => 'thai',
+
+	'ascii'        => {inherit => ['' => '00-7F']},
+	'cp437'        => {
+		inherit => ['cp850' => 0, 'ascii' => '00-1F+80'], # ascii range overridden later
+		replace => {
+			0    => ' ☺☻♥♦♣♠•◘○◙♂♀♪♫☼►◄↕‼¶§▬↨↑↓→←∟↔▲▼', # visible variants of control characters
+			0xED => 'ϕ', # non-greek usage and appearance
+		},
+	},
+	'gsm0338'      => {inherit => ['ascii' => '00-7F']},
+	'dingbats'     => {inherit => ['' => '20-7F+A0']},
+	'macdingbats'  => {inherit => ['dingbats' => '80-9F']},
+	'adobezdingbat'=> {inherit => ['MacDingbats' => '80-9F']}, # should be identical but maps to private use
+	'symbol'       => {
+		inherit => ['' => '20-7F+A0'],
+		replace => {
+			0x60 => '│', # replace radical extender by closest unicode equivalent
+			0xBD => '⏐⎯', # arrow extenders
+			0xD2 => '®©™', # serif variants
+			0xE0 => '◊', # replace lookalike, should match AdobeSymbol
+			0xE2 => '®©™', # sans-serif variants
+			0xE6 => '⎛⎜⎝⎡⎢⎣⎧⎨⎩⎪',
+			0xF0 => '€',
+			0xF4 => '⎮⌡⎞⎟⎠⎤⎥⎦⎫⎬⎭',
+		},
+	},
+	'adobesymbol'  => {inherit => ['symbol' => '20-7F+A0', '' => '20-7F+A0']}, # minor differences, irrelevant except for different '€'
+	'wingdings'    => {inherit => ['' => '20'], setup => sub {require Encode::Wingdings}},
+	'wingdings2'   => {inherit => ['' => '20'], setup => sub {require Encode::Wingdings2}},
+	'wingdings3'   => {inherit => ['' => '20'], setup => sub {require Encode::Wingdings3}},
+	'webdings'     => {inherit => ['' => '20'], setup => sub {require Encode::Webdings}},
+
+	'iso-8859-2'   => {inherit => ['iso-8859-1' => 'A0']},
+	'iso-8859-3'   => {inherit => ['iso-8859-1' => 'A0']}, #TODO: also apply to iso-8859-9
+	'iso-8859-4'   => {inherit => ['iso-8859-2' => 'A0']},
+	'iso-8859-5'   => {inherit => ['iso-8859-1' => 'A0']},
+	'iso-8859-6'   => {inherit => ['cp1256' => '80', 'iso-8859-1' => 'A0']},
+	'iso-8859-7'   => {inherit => ['iso-8859-1' => 'A0']},
+	'iso-8859-8'   => {inherit => ['iso-8859-1' => 'A0']},
+	'iso-8859-9'   => {inherit => ['iso-8859-1' => 'D0-DF+F0']},
+	'iso-8859-10'  => {inherit => ['iso-8859-4' => 'A0']},
+	'iso-8859-11'  => {inherit => ['iso-8859-1' => 'A0']},
+	'iso-8859-13'  => {inherit => ['iso-8859-4' => 'A0']},
+	'iso-8859-14'  => {inherit => ['iso-8859-1' => 'A0']},
+	'iso-8859-15'  => {inherit => ['iso-8859-1' => 'A0-BF']},
+	'iso-8859-16'  => {inherit => ['iso-8859-2' => 'A0']},
+	'hp-roman8'    => {inherit => ['iso-8859-1' => 'A0']},
+
+	'cp1252'       => {inherit => ['iso-8859-1' => '80-9F']},
+	'cp1250'       => {inherit => ['iso-8859-2' => '80-BF', 'cp1252' => '80']},
+	'cp1254'       => {inherit => ['iso-8859-9' => '80-9F', 'cp1252' => '80-9F+D0']},
+	'cp874'        => {inherit => ['iso-8859-11' => '80-9F', 'cp1252' => '80']}, # windows-874 actually cp1162
+	'cp1257'       => {inherit => ['iso-8859-13' => '80-9F+FF', 'cp1252' => '80']},
+	'cp1251'       => {inherit => ['cp1252' => '80']},
+	'cp1253'       => {inherit => ['cp1252' => '80']},
+	'cp1255'       => {inherit => ['iso-8859-8' => '80-DF', 'cp1252' => '80']},
+	'cp1256'       => {inherit => ['cp1252' => '80']},
+	'cp1258'       => {inherit => ['cp1252' => '80-9F+C0']},
+
+	'cp850'        => {inherit => ['cp437' => '90']},
+	'cp860'        => {inherit => ['cp437' => '80-AF']},
+	'cp861'        => {inherit => ['cp865' => '80-AF']},
+	'cp863'        => {inherit => ['cp437' => '80-AF']},
+	'cp865'        => {inherit => ['cp437' => '90-AF']},
+	'cp852'        => {inherit => ['cp850' => '80', 'cp437' => '80']},
+	'cp857'        => {inherit => ['cp850' => '80-AF+D0-EF', 'cp437' => '80']},
+	'cp775'        => {inherit => ['cp850' => '80']},  # partial cp437
+	'cp866'        => {inherit => ['cp437' => '80-AF+E0']},
+	'cp855'        => {inherit => ['cp437' => '80']},
+	'cp1006'       => {inherit => ['iso-8859-6' => 'A0', 'cp437' => '80']},
+	'cp737'        => {inherit => ['cp437' => '80-AF+E0']},
+	'cp869'        => {inherit => ['cp437' => '80']},
+	'cp862'        => {inherit => ['cp437' => '80-9F']},
+	'cp864'        => {inherit => ['MacArabic' => '80', 'iso-8859-6' => '80', 'cp437' => '80']}, #TODO: compare form variants
+
+	'koi8-u'       => {inherit => ['koi8-r' => '90-BF']},
+	'koi8-f'       => {inherit => ['koi8-u' => '90-BF']},
+
+	'macromanian'  => {inherit => ['MacRoman' => 'A0-BF+D0-DF']},
+	'macrumanian'  => {inherit => ['MacRomanian' => 'A0-BF+D0-DF', 'MacRoman' => 'A0-BF+D0-DF']},
+	'maccroatian'  => {inherit => ['MacRoman' => 'A0']},
+	'maccentraleurroman' => {inherit => ['MacRoman' => '80']},
+	'macicelandic' => {inherit => ['MacRoman' => 'A0-AF+D0-EF']},
+	'macturkish'   => {inherit => ['MacRoman' => 'D0-DF']}, # F5 is unassigned
+	'macsami'      => {inherit => ['MacIcelandic' => '90-BF+D0-DF+F0', 'MacRoman' => '90']},
+	'macgreek'     => {inherit => ['MacRoman' => '80']},
+	'maccyrillic'  => {inherit => ['MacRoman' => '80']},
+	'machebrew'    => {
+		inherit => ['iso-8859-8' => '80', 'MacRoman' => '80-8F+A0'], # ignore partial ascii
+		varchar => 1,
+		replace => {
+			# strip private use characters for unneeded roundtrip
+			0xDE => chr(0x5C7), # qamats qatan
+			0xC0 => 'לֹ', # lamed holam
+			#TODO: private use for canoral codes (obsolete nikud positioning)
+		},
+	},
+	'macarabic'    => {inherit => ['iso-8859-6' => '80', 'cp864' => '80', 'MacRoman' => '80']}, #TODO: multiple parents
+	'macfarsi'     => {inherit => ['MacArabic' => 'B0-BF', 'MacRoman' => '80']},
+	'macthai'      => {
+		inherit => ['iso-8859-11' => '80-9F+D0'],
+		varchar => 1,
+		replace => {
+			# strip appended private use characters for unneeded roundtrip
+			0x83 => "\x{E48}", 0x88 => "\x{E48}", 0x98 => "\x{E48}",
+			0x84 => "\x{E49}", 0x89 => "\x{E49}", 0x99 => "\x{E49}",
+			0x85 => "\x{E4A}", 0x8A => "\x{E4A}", 0x9A => "\x{E4A}",
+			0x86 => "\x{E4B}", 0x8B => "\x{E4B}", 0x9B => "\x{E4B}",
+			0x87 => "\x{E4C}", 0x8C => "\x{E4C}", 0x9C => "\x{E4C}",
+			0x8F => "\x{E4D}",
+			0x92 => "\x{E31}", 0x93 => "\x{E47}", 0x94 => "\x{E34}",
+			0x95 => "\x{E35}", 0x96 => "\x{E36}", 0x97 => "\x{E37}",
+		},
+	},
+
+	'cp37'         => {inherit => ['posix-bc' => '00']},
+	'posix-bc'     => {inherit => ['cp1047' => '40']},
+	'cp500'        => {inherit => ['cp37' => '40-5F+B0-BF']},
+	'cp1047'       => {inherit => ['cp37' => '10-2F+50-5F+A0-BF']},
+	'cp1026'       => {inherit => ['cp37' => '40']},
+	'cp875'        => {inherit => ['cp37' => '30']},
+
+	''             => {setup => sub {
+		my $row = shift;
+		$row->{offset} = delete $row->{startpoint};
+		$row->{set} = 'Unicode characters';
+		my $block = $row->{offset} >> 8;
+		$row->{endpoint} ||= ($block + 1 << 8) - 1;
+		$block == ($row->{endpoint} >> 8) or undef $block;
+
+		$row->{table} = join '', map { chr } $row->{offset} .. $row->{endpoint};
+		utf8::upgrade($row->{table});  # prevent latin1 output
+
+		$row->{endpoint} -= $row->{offset};
+
+		if (defined $block) {
+			$row->{set} = sprintf 'Unicode block U+%02Xxx', $block;
+			$row->{offset} %= 0x100;
+		}
+
+		return $row;
+	}},
+	u              => {setup => sub {
+		my $row = shift;
+		state $celldata = do 'charset-unicode.inc.pl'
+			or Alert('Table data could not be read', $@ || $!);
+		$row->{cell} = $celldata;
+
+		$row->{endpoint} ||= 0x1FFF;
+		$row->{set} = 'Unicode ' . (
+			$row->{startpoint} <  0x1000 && $row->{endpoint} < 0x1000 ? 'BMP' :
+			$row->{startpoint} >= 0x1000 && $row->{endpoint} < 0x2000 ? 'SMP' :
+			'allocations'
+		);
+		return $row;
+	}},
+	uu             => {setup => sub {
+		my $row = shift;
+		$row->{cell} = do 'charset-ucplanes.inc.pl'
+			or Alert('Table data could not be read', $@ || $!);
+		$row->{endpoint} ||= 0x3FF;
+		$row->{set} = 'Unicode planes';
+		return $row;
+	}},
+	utf8           => {setup => sub {
+		my $row = shift;
+		$row->{set} = 'UTF-8';
+		$row->{cell} = do 'charset-utf8.inc.pl'
+			or Alert('Table data could not be read', $@ || $!);
+		return $row;
+	}},
+	'utf-8'        => 'utf8',
 };