[p5sagit/p5-mst-13.2.git] / lib / unicore / mktables.PL

#!../../miniperl

use bytes;

$UnicodeData = "Unicode.txt";
$SyllableData = "syllables.txt";
$PropData = "PropList.txt";


# Note: we try to keep filenames unique within first 8 chars.  Using
# subdirectories for the following helps.
mkdir "In", 0755;
mkdir "Is", 0755;
mkdir "To", 0755;

@todo = (
# typical

    # 005F: SPACING UNDERSCROE
    ['IsWord',   '$cat =~ /^[LMN]/ or $code eq "005F"',	''],
    ['IsAlnum',  '$cat =~ /^[LMN]/',	''],
    ['IsAlpha',  '$cat =~ /^[LM]/',	''],
    # 0009: HORIZONTAL TABULATION
    # 000A: LINE FEED
    # 000B: VERTICAL TABULATION
    # 000C: FORM FEED
    # 000D: CARRIAGE RETURN
    # 0020: SPACE
    ['IsSpace',  '$cat  =~ /^Z/ ||
                  $code =~ /^(0009|000A|000B|000C|000D)$/',	''],
    ['IsSpacePerl',
                 '$cat  =~ /^Z/ ||
                  $code =~ /^(0009|000A|000C|000D)$/',		''],
    ['IsBlank',  '$code =~ /^(0020|0009)$/ ||
		  $cat  =~ /^Z[^lp]$/',	''],
    ['IsDigit',  '$cat =~ /^Nd$/',	''],
    ['IsUpper',  '$cat =~ /^L[ut]$/',	''],
    ['IsLower',  '$cat =~ /^Ll$/',	''],
    ['IsASCII',  '$code le "007f"',	''],
    ['IsCntrl',  '$cat =~ /^C/',	''],
    ['IsGraph',  '$cat =~ /^([LMNPS]|Co)/',	''],
    ['IsPrint',  '$cat =~ /^([LMNPS]|Co|Zs)/',	''],
    ['IsPunct',  '$cat =~ /^P/',	''],
    # 003[0-9]: DIGIT ZERO..NINE, 00[46][1-6]: A..F, a..f
    ['IsXDigit', '$code =~ /^00(3[0-9]|[46][1-6])$/',	''],
    ['ToUpper',  '$up',			'$up'],
    ['ToLower',  '$down',		'$down'],
    ['ToTitle',  '$title',		'$title'],
    ['ToDigit',  '$dec ne ""',		'$dec'],

# Name

    ['Name',	'$name',		'$name'],

# Category

    ['Category', '$cat',		'$cat'],

# Normative

    ['IsM',	'$cat =~ /^M/',		''],	# Mark
    ['IsMn',	'$cat eq "Mn"',		''],	# Mark, Non-Spacing 
    ['IsMc',	'$cat eq "Mc"',		''],	# Mark, Combining
    ['IsMe',	'$cat eq "Me"',		''],    # Mark, Enclosing

    ['IsN',	'$cat =~ /^N/',		''],	# Number
    ['IsNd',	'$cat eq "Nd"',		''],	# Number, Decimal Digit
    ['IsNo',	'$cat eq "No"',		''],	# Number, Other
    ['IsNl',	'$cat eq "Nl"',		''],    # Number, Letter

    ['IsZ',	'$cat =~ /^Z/',		''],	# Separator
    ['IsZs',	'$cat eq "Zs"',		''],	# Separator, Space
    ['IsZl',	'$cat eq "Zl"',		''],	# Separator, Line
    ['IsZp',	'$cat eq "Zp"',		''],	# Separator, Paragraph

    ['IsC',	'$cat =~ /^C/',		''],	# Crazy
    ['IsCc',	'$cat eq "Cc"',		''],	# Other, Control or Format
    ['IsCo',	'$cat eq "Co"',		''],	# Other, Private Use
    ['IsCn',	'$cat eq "Cn"',		''],	# Other, Not Assigned
    ['IsCf',	'$cat eq "Cf"',		''],    # Other, Format
    ['IsCs',	'$cat eq "Cs"',		''],    # Other, Surrogate
    ['IsCn',	'Unassigned Code Value',$PropData],	# Other, Not Assigned
 
# Informative

    ['IsL',	'$cat =~ /^L/',		''],	# Letter
    ['IsLu',	'$cat eq "Lu"',		''],	# Letter, Uppercase
    ['IsLl',	'$cat eq "Ll"',		''],	# Letter, Lowercase
    ['IsLt',	'$cat eq "Lt"',		''],	# Letter, Titlecase 
    ['IsLm',	'$cat eq "Lm"',		''],	# Letter, Modifier
    ['IsLo',	'$cat eq "Lo"',		''],	# Letter, Other 

    ['IsP',	'$cat =~ /^P/',		''],	# Punctuation
    ['IsPd',	'$cat eq "Pd"',		''],	# Punctuation, Dash
    ['IsPs',	'$cat eq "Ps"',		''],	# Punctuation, Open
    ['IsPe',	'$cat eq "Pe"',		''],	# Punctuation, Close
    ['IsPo',	'$cat eq "Po"',		''],	# Punctuation, Other
    ['IsPc',	'$cat eq "Pc"',		''],	# Punctuation, Connector
    ['IsPi',	'$cat eq "Pi"',		''],	# Punctuation, Initial quote
    ['IsPf',	'$cat eq "Pf"',		''],	# Punctuation, Final quote

    ['IsS',	'$cat =~ /^S/',		''],	# Symbol
    ['IsSm',	'$cat eq "Sm"',		''],	# Symbol, Math
    ['IsSk',	'$cat eq "Sk"',		''],	# Symbol, Modifier
    ['IsSc',	'$cat eq "Sc"',		''],	# Symbol, Currency
    ['IsSo',	'$cat eq "So"',		''],	# Symbol, Other

# Combining class
    ['CombiningClass', '$comb',		'$comb'],

# BIDIRECTIONAL PROPERTIES
 
    ['Bidirectional', '$bid',		'$bid'],

# Strong types:

    ['IsBidiL',	'$bid eq "L"',		''],	# Left-Right; Most alphabetic,
						# syllabic, and logographic
						# characters (e.g., CJK
						# ideographs)
    ['IsBidiR',	'$bid eq "R"',		''],	# Right-Left; Arabic, Hebrew,
						# and punctuation specific to
						# those scripts

    ['IsBidiLRE', '$bid eq "LRE"',       ''],    # Left-to-Right Embedding
    ['IsBidiLRO', '$bid eq "LRO"',       ''],    # Left-to-Right Override
    ['IsBidiAL', '$bid eq "AL"',         ''],    # Right-to-Left Arabic
    ['IsBidiRLE', '$bid eq "RLE"',       ''],    # Right-to-Left Embedding
    ['IsBidiRLO', '$bid eq "RLO"',       ''],    # Right-to-Left Override
    ['IsBidiPDF', '$bid eq "PDF"',       ''],    # Pop Directional Format
    ['IsBidiNSM', '$bid eq "NSM"',       ''],    # Non-Spacing Mark
    ['IsBidiBN', '$bid eq "BN"',         ''],    # Boundary Neutral

# Weak types:

    ['IsBidiEN','$bid eq "EN"',		''],	# European Number
    ['IsBidiES','$bid eq "ES"',		''],	# European Number Separator
    ['IsBidiET','$bid eq "ET"',		''],	# European Number Terminator
    ['IsBidiAN','$bid eq "AN"',		''],	# Arabic Number
    ['IsBidiCS','$bid eq "CS"',		''],	# Common Number Separator

# Separators:

    ['IsBidiB',	'$bid eq "B"',		''],	# Block Separator
    ['IsBidiS',	'$bid eq "S"',		''],	# Segment Separator

# Neutrals:

    ['IsBidiWS','$bid eq "WS"',		''],	# Whitespace
    ['IsBidiON','$bid eq "ON"',		''],	# Other Neutrals ; All other
						# characters: punctuation,
						# symbols

# Decomposition

    ['Decomposition',	'$decomp',	'$decomp'],
    ['IsDecoCanon',	'$decomp && $decomp !~ /^</',	''],
    ['IsDecoCompat',	'$decomp =~ /^</',		''],
    ['IsDCfont',	'$decomp =~ /^<font>/',		''],
    ['IsDCnoBreak',	'$decomp =~ /^<noBreak>/',	''],
    ['IsDCinitial',	'$decomp =~ /^<initial>/',	''],
    ['IsDCmedial',	'$decomp =~ /^<medial>/',	''],
    ['IsDCfinal',	'$decomp =~ /^<final>/',	''],
    ['IsDCisolated',	'$decomp =~ /^<isolated>/',	''],
    ['IsDCcircle',	'$decomp =~ /^<circle>/',	''],
    ['IsDCsuper',	'$decomp =~ /^<super>/',	''],
    ['IsDCsub',		'$decomp =~ /^<sub>/',		''],
    ['IsDCvertical',	'$decomp =~ /^<vertical>/',	''],
    ['IsDCwide',	'$decomp =~ /^<wide>/',		''],
    ['IsDCnarrow',	'$decomp =~ /^<narrow>/',	''],
    ['IsDCsmall',	'$decomp =~ /^<small>/',	''],
    ['IsDCsquare',	'$decomp =~ /^<square>/',	''],
    ['IsDCfraction',	'$decomp =~ /^<fraction>/',	''],
    ['IsDCcompat',	'$decomp =~ /^<compat>/',	''],

# Number

    ['Number', 	'$num ne ""',		'$num'],

# Mirrored

    ['IsMirrored', '$mir eq "Y"',	''],

# Arabic

    ['ArabLink', 	'1',		'$link'],
    ['ArabLnkGrp', 	'1',		'$linkgroup'],

# Jamo

    ['JamoShort',	'1',		'$short'],

# Syllables

    syllable_defs(),

# Line break properties - Normative

    ['IsLbrkBK','$brk eq "BK"',		''],	# Mandatory Break
    ['IsLbrkCR','$brk eq "CR"',		''],	# Carriage Return
    ['IsLbrkLF','$brk eq "LF"',		''],	# Line Feed
    ['IsLbrkCM','$brk eq "CM"',		''],	# Attached Characters and Combining Marks
    ['IsLbrkSG','$brk eq "SG"',		''],	# Surrogates
    ['IsLbrkGL','$brk eq "GL"',		''],	# Non-breaking (Glue)
    ['IsLbrkCB','$brk eq "CB"',		''],	# Contingent Break Opportunity
    ['IsLbrkSP','$brk eq "SP"',		''],	# Space
    ['IsLbrkZW','$brk eq "ZW"',		''],	# Zero Width Space

# Line break properties - Informative
    ['IsLbrkXX','$brk eq "XX"',		''],	# Unknown
    ['IsLbrkOP','$brk eq "OP"',		''],	# Opening Punctuation
    ['IsLbrkCL','$brk eq "CL"',		''],	# Closing Punctuation
    ['IsLbrkQU','$brk eq "QU"',		''],	# Ambiguous Quotation
    ['IsLbrkNS','$brk eq "NS"',		''],	# Non Starter
    ['IsLbrkEX','$brk eq "EX"',		''],	# Exclamation/Interrogation
    ['IsLbrkSY','$brk eq "SY"',		''],	# Symbols Allowing Breaks
    ['IsLbrkIS','$brk eq "IS"',		''],	# Infix Separator (Numeric)
    ['IsLbrkPR','$brk eq "PR"',		''],	# Prefix (Numeric)
    ['IsLbrkPO','$brk eq "PO"',		''],	# Postfix (Numeric)
    ['IsLbrkNU','$brk eq "NU"',		''],	# Numeric
    ['IsLbrkAL','$brk eq "AL"',		''],	# Ordinary Alphabetic and Symbol Characters
    ['IsLbrkID','$brk eq "ID"',		''],	# Ideographic
    ['IsLbrkIN','$brk eq "IN"',		''],	# Inseparable
    ['IsLbrkHY','$brk eq "HY"',		''],	# Hyphen
    ['IsLbrkBB','$brk eq "BB"',		''],	# Break Opportunity Before
    ['IsLbrkBA','$brk eq "BA"',		''],	# Break Opportunity After
    ['IsLbrkSA','$brk eq "SA"',		''],	# Complex Context (South East Asian)
    ['IsLbrkAI','$brk eq "AI"',		''],	# Ambiguous (Alphabetic or Ideographic)
    ['IsLbrkB2','$brk eq "B2"',		''],	# Break Opportunity Before and After
);

# This is not written for speed...

my %InId;
my $InId = 0;

foreach $file (@todo) {
    my ($table, $wanted, $val) = @$file;
    next if @ARGV and not grep { $_ eq $table } @ARGV;
    print $table, "\n";
    $table =~ s/\W+//g;
    if ($table =~ /^(Is|To)(.+)/) {
	open(OUT, ">$1/$2.pl") or die "Can't create $1/$2.pl: $!\n";
    }
    else {
	open(OUT, ">$table.pl") or die "Can't create $table.pl: $!\n";
    }
    print OUT <<EOH;
# !!!!!!!   DO NOT EDIT THIS FILE   !!!!!!! 
# This file is built by $0 from e.g. $UnicodeData.
# Any changes made here will be lost!
EOH
    print OUT <<"END";
return <<'END';
END
    print OUT proplist($table, $wanted, $val);
    print OUT "END\n";
    close OUT;
}

# Do Scripts before Blocks so that in case of naming conflicts
# the more natural one (Script) wins over the artificial one (Block).

print "Scripts\n";
open(UD, 'Scripts.txt') or die "Can't open Scripts.txt: $!\n";
open(OUT, ">Scripts.pl") or die "Can't create Scripts.pl: $!\n";
print OUT <<EOH;
# !!!!!!!   DO NOT EDIT THIS FILE   !!!!!!! 
# This file is built by $0 from e.g. $UnicodeData.
# Any changes made here will be lost!
EOH
print OUT <<"END";
return <<'END';
END

my %Scripts;

while (<UD>) {
    next if /^#/;
    next if /^$/;
    chomp;
    ($code, $last, $name) = /^([0-9a-f]+)(?:\.\.([0-9a-f]+))?\s+;\s+(.+)\s+\#/i;
    if ($name) {
	my $InName = lc($name);
	$InName =~ s/\b(\w)/uc($1)/ge;
	$InName =~ s/\W+//g;
	my $id;
        unless (exists $InId{$InName}) {
	    print "\t$InName\n";
	    $id = $Scripts{$InName} = $InId{$InName} = $InId++;
	    open(SCRIPT, ">In/$id.pl") or die "create In/$id.pl: $!\n";
	    print SCRIPT <<EOH;
# !!!!!!!   DO NOT EDIT THIS FILE   !!!!!!! 
# This file is built by $0 from e.g. $UnicodeData.
# Any changes made here will be lost!
return <<'END';
EOH
            close(SCRIPT);
	} else {
	    $id = $InId{$InName};
	}
	$last = "" unless defined $last;
	print OUT "$code\t$last\t$name\t# $InName In/$id.pl\n";
        open(SCRIPT, ">>In/$id.pl");
	print SCRIPT <<END;
$code	$last
END
	close SCRIPT;
    }
}

for my $id (values %InId) {
    open(SCRIPT, ">>In/$id.pl");
    print SCRIPT <<END2;
END
END2
    close(SCRIPT);
}

print OUT "END\n";
close OUT;

# Must treat blocks specially.

exit if @ARGV and not grep { $_ eq Block } @ARGV;
print "Blocks\n";
open(UD, 'Blocks.txt') or die "Can't open Blocks.txt: $!\n";
open(OUT, ">Blocks.pl") or die "Can't create Blocks.pl: $!\n";
print OUT <<EOH;
# !!!!!!!   DO NOT EDIT THIS FILE   !!!!!!! 
# This file is built by $0 from e.g. $UnicodeData.
# Any changes made here will be lost!
EOH
print OUT <<"END";
return <<'END';
END

while (<UD>) {
    next if /^#/;
    next if /^$/;
    chomp;
    ($code, $last, $name) = /^([0-9a-f]+)\.\.([0-9a-f]+); (.+)/i;
    if ($name) {
	my $InName = $name;
	$InName =~ s/\W+//g;
	print "\t$InName\n";
	my $id;
	# TODO: only the first one of Private Use blocks qualifies
        unless (exists $InId{$InName}) {
	    $InId{$InName} = $InId++;
	} elsif (exists $Scripts{$InName}) {
	    $InName .= 'Block';
	    $InId{$InName} = $InId++;
	}
	$id = $InId{$InName};
	open(BLOCK, ">In/$id.pl") or die "create In/$id.pl: $!\n";
	print OUT "$code\t$last\t$name\t# $InName In/$id.pl\n";
	print BLOCK <<EOH;
# !!!!!!!   DO NOT EDIT THIS FILE   !!!!!!! 
# This file is built by $0 from e.g. $UnicodeData.
# Any changes made here will be lost!
EOH
	print BLOCK <<"END2";
return <<'END';
$code	$last
END
END2
	close BLOCK;
    }
}

print OUT "END\n";
close OUT;

open(INID, ">In.pl");

print INID <<EOH;
# !!!!!!!   DO NOT EDIT THIS FILE   !!!!!!! 
# This file is built by $0 from e.g. $UnicodeData.
# Any changes made here will be lost!
%utf8::In = (
EOH

# Order doesn't matter but let's prettyprint anyway.
foreach my $in (sort { $InId{$a} <=> $InId{$b} } keys %InId) {
    printf INID "%-40s => %3d,\n", "'$in'", $InId{$in};
}

print INID ");\n";

close(INID);

##################################################

sub proplist {
    my ($table, $wanted, $val) = @_;
    my @wanted;
    my $out;
    my $split;

    return listFromPropFile($wanted) if $val eq $PropData;

    if ($table =~ /^Arab/) {
	open(UD, "ArabShap.txt") or warn "Can't open $table: $!";

	$split = '($code, $name, $link, $linkgroup) = split(/; */);';
    }
    elsif ($table =~ /^Jamo/) {
	open(UD, "Jamo.txt") or warn "Can't open $table: $!";

	$split = '($code, $short, $name) = split(/; */); $code =~ s/^U\+//;';
    }
    elsif ($table =~ /^IsSyl/) {
	open(UD, $SyllableData) or warn "Can't open $table: $!";

	$split = '($code, $short, $syl) = split(/; */); $code =~ s/^U\+//;';
    }
    elsif ($table =~ /^IsLbrk/) {
	open(UD, "LineBrk.txt") or warn "Can't open $table: $!";

	$split = '($code, $brk, $name) = /^([0-9a-f]+);(\w+) # (.+)/i;';
    }
    else {
	open(UD, $UnicodeData) or warn "Can't open $UnicodeData: $!";

	$split = '($code, $name, $cat, $comb, $bid, $decomp, $dec, $dig, $num, $mir, $uni1,
		$comment, $up, $down, $title) = split(/;/);';
    }

    if ($table =~ /^(?:To|Is)[A-Z]/) {
	eval <<"END";
	    while (<UD>) {
		next if /^#/;
		next if /^\\s/;
		s/\\s+\$//;
		$split
		if ($wanted) {
		    push(\@wanted, [hex \$code, hex $val, \$name =~ /, First>\$/]);
		}
	    }
END
	die $@ if $@;

	while (@wanted) {
	    $beg = shift @wanted;
	    $last = $beg;
	    while (@wanted and $wanted[0]->[0] == $last->[0] + 1 and
		(not $val or $wanted[0]->[1] == $last->[1] + 1)) {
		    $last = shift @wanted;
	    }
	    $out .= sprintf "%04x", $beg->[0];
	    if ($beg->[2]) {
		$last = shift @wanted;
	    }
	    if ($beg == $last) {
		$out .= "\t";
	    }
	    else {
		$out .= sprintf "\t%04x", $last->[0];
	    }
	    $out .= sprintf "\t%04x", $beg->[1] if $val;
	    $out .= "\n";
	}
    }
    else {
	eval <<"END";
	    while (<UD>) {
		next if /^#/;
		next if /^\\s*\$/;
		chop;
		$split
		if ($wanted) {
		    push(\@wanted, [hex \$code, $val, \$name =~ /, First>\$/]);
		}
	    }
END
	die $@ if $@;

	while (@wanted) {
	    $beg = shift @wanted;
	    $last = $beg;
	    while (@wanted and $wanted[0]->[0] == $last->[0] + 1 and
		($wanted[0]->[1] eq $last->[1])) {
		    $last = shift @wanted;
	    }
	    $out .= sprintf "%04x", $beg->[0];
	    if ($beg->[2]) {
		$last = shift @wanted;
	    }
	    if ($beg == $last) {
		$out .= "\t";
	    }
	    else {
		$out .= sprintf "\t%04x", $last->[0];
	    }
	    $out .= sprintf "\t%s\n", $beg->[1];
	}
    }
    $out;
}

sub listFromPropFile {
    my ($wanted) = @_;
    my $out;

    open (UD, $PropData) or die "Can't open $PropData: $!\n";
    local($/) = "\n" . '*' x 43 . "\n\nProperty dump for:";   # not 42?

    <UD>;
    while (<UD>) {
        chomp;
        if (s/0x[\d\w]+\s+\((.*?)\)// and $wanted eq $1) {
            s/\(\d+ chars\)//g;
            s/^\s+//mg;
            s/\s+$//mg;
            s/\.\./\t/g;
	    $out = lc $_;
	    last;
        }
    }
    close (UD);
    "$out\n";
}

sub syllable_defs {
    my @defs;
    my %seen;

    open (SD, $SyllableData) or die "Can't open $SyllableData: $!\n";
    while (<SD>) {
        next if /^\s*(#|$)/;
        s/\s+$//;
        ($code, $name, $syl) = split /; */;
        next unless $syl;
        push (@defs, ["IsSyl$syl", qq{\$syl eq "$syl"}, ''])
                                                     unless $seen{$syl}++;
    }
    close (SD);
    return (@defs);
}

# eof
Commit	Line	Data
a0ed51b3	1	#!../../miniperl
a0ed51b3	2
7c6f5cd2	3	use bytes;
7c6f5cd2	4
190eec7c	5	$UnicodeData = "Unicode.txt";
7c6f5cd2	6	$SyllableData = "syllables.txt";
8836d2a5	7	$PropData = "PropList.txt";
7c6f5cd2	8
11695a73	9
a0ed51b3	10	# Note: we try to keep filenames unique within first 8 chars. Using
a0ed51b3	11	# subdirectories for the following helps.
289d4f09	12	mkdir "In", 0755;
	13	mkdir "Is", 0755;
	14	mkdir "To", 0755;
a0ed51b3	15
	16	@todo = (
	17	# typical
	18
f59877d4	19	# 005F: SPACING UNDERSCROE
	20	['IsWord', '$cat =~ /^[LMN]/ or $code eq "005F"', ''],
	21	['IsAlnum', '$cat =~ /^[LMN]/', ''],
	22	['IsAlpha', '$cat =~ /^[LM]/', ''],
	23	# 0009: HORIZONTAL TABULATION
	24	# 000A: LINE FEED
	25	# 000B: VERTICAL TABULATION
	26	# 000C: FORM FEED
	27	# 000D: CARRIAGE RETURN
3bec3564	28	# 0020: SPACE
f59877d4	29	['IsSpace', '$cat =~ /^Z/ \|\|
f59877d4	30	$code =~ /^(0009\|000A\|000B\|000C\|000D)$/', ''],
3bec3564	31	['IsSpacePerl',
	32	'$cat =~ /^Z/ \|\|
	33	$code =~ /^(0009\|000A\|000C\|000D)$/', ''],
f5868911	34	['IsBlank', '$code =~ /^(0020\|0009)$/ \|\|
f5868911	35	$cat =~ /^Z[^lp]$/', ''],
b8c5462f	36	['IsDigit', '$cat =~ /^Nd$/', ''],
7c6f5cd2	37	['IsUpper', '$cat =~ /^L[ut]$/', ''],
b8c5462f	38	['IsLower', '$cat =~ /^Ll$/', ''],
f59877d4	39	['IsASCII', '$code le "007f"', ''],
b8c5462f	40	['IsCntrl', '$cat =~ /^C/', ''],
f59877d4	41	['IsGraph', '$cat =~ /^([LMNPS]\|Co)/', ''],
	42	['IsPrint', '$cat =~ /^([LMNPS]\|Co\|Zs)/', ''],
	43	['IsPunct', '$cat =~ /^P/', ''],
	44	# 003[0-9]: DIGIT ZERO..NINE, 00[46][1-6]: A..F, a..f
b8c5462f	45	['IsXDigit', '$code =~ /^00(3[0-9]\|[46][1-6])$/', ''],
	46	['ToUpper', '$up', '$up'],
	47	['ToLower', '$down', '$down'],
	48	['ToTitle', '$title', '$title'],
	49	['ToDigit', '$dec ne ""', '$dec'],
a0ed51b3	50
	51	# Name
	52
	53	['Name', '$name', '$name'],
	54
	55	# Category
	56
	57	['Category', '$cat', '$cat'],
	58
	59	# Normative
	60
	61	['IsM', '$cat =~ /^M/', ''], # Mark
	62	['IsMn', '$cat eq "Mn"', ''], # Mark, Non-Spacing
	63	['IsMc', '$cat eq "Mc"', ''], # Mark, Combining
7c6f5cd2	64	['IsMe', '$cat eq "Me"', ''], # Mark, Enclosing
a0ed51b3	65
	66	['IsN', '$cat =~ /^N/', ''], # Number
	67	['IsNd', '$cat eq "Nd"', ''], # Number, Decimal Digit
	68	['IsNo', '$cat eq "No"', ''], # Number, Other
7c6f5cd2	69	['IsNl', '$cat eq "Nl"', ''], # Number, Letter
a0ed51b3	70
7c6f5cd2	71	['IsZ', '$cat =~ /^Z/', ''], # Separator
a0ed51b3	72	['IsZs', '$cat eq "Zs"', ''], # Separator, Space
	73	['IsZl', '$cat eq "Zl"', ''], # Separator, Line
	74	['IsZp', '$cat eq "Zp"', ''], # Separator, Paragraph
	75
	76	['IsC', '$cat =~ /^C/', ''], # Crazy
	77	['IsCc', '$cat eq "Cc"', ''], # Other, Control or Format
	78	['IsCo', '$cat eq "Co"', ''], # Other, Private Use
	79	['IsCn', '$cat eq "Cn"', ''], # Other, Not Assigned
7c6f5cd2	80	['IsCf', '$cat eq "Cf"', ''], # Other, Format
	81	['IsCs', '$cat eq "Cs"', ''], # Other, Surrogate
	82	['IsCn', 'Unassigned Code Value',$PropData], # Other, Not Assigned
a0ed51b3	83
	84	# Informative
	85
	86	['IsL', '$cat =~ /^L/', ''], # Letter
	87	['IsLu', '$cat eq "Lu"', ''], # Letter, Uppercase
	88	['IsLl', '$cat eq "Ll"', ''], # Letter, Lowercase
	89	['IsLt', '$cat eq "Lt"', ''], # Letter, Titlecase
	90	['IsLm', '$cat eq "Lm"', ''], # Letter, Modifier
	91	['IsLo', '$cat eq "Lo"', ''], # Letter, Other
	92
	93	['IsP', '$cat =~ /^P/', ''], # Punctuation
	94	['IsPd', '$cat eq "Pd"', ''], # Punctuation, Dash
	95	['IsPs', '$cat eq "Ps"', ''], # Punctuation, Open
	96	['IsPe', '$cat eq "Pe"', ''], # Punctuation, Close
	97	['IsPo', '$cat eq "Po"', ''], # Punctuation, Other
7c6f5cd2	98	['IsPc', '$cat eq "Pc"', ''], # Punctuation, Connector
	99	['IsPi', '$cat eq "Pi"', ''], # Punctuation, Initial quote
	100	['IsPf', '$cat eq "Pf"', ''], # Punctuation, Final quote
a0ed51b3	101
	102	['IsS', '$cat =~ /^S/', ''], # Symbol
	103	['IsSm', '$cat eq "Sm"', ''], # Symbol, Math
7c6f5cd2	104	['IsSk', '$cat eq "Sk"', ''], # Symbol, Modifier
a0ed51b3	105	['IsSc', '$cat eq "Sc"', ''], # Symbol, Currency
	106	['IsSo', '$cat eq "So"', ''], # Symbol, Other
	107
	108	# Combining class
	109	['CombiningClass', '$comb', '$comb'],
	110
	111	# BIDIRECTIONAL PROPERTIES
	112
	113	['Bidirectional', '$bid', '$bid'],
	114
	115	# Strong types:
	116
	117	['IsBidiL', '$bid eq "L"', ''], # Left-Right; Most alphabetic,
	118	# syllabic, and logographic
	119	# characters (e.g., CJK
	120	# ideographs)
	121	['IsBidiR', '$bid eq "R"', ''], # Right-Left; Arabic, Hebrew,
	122	# and punctuation specific to
	123	# those scripts
	124
7c6f5cd2	125	['IsBidiLRE', '$bid eq "LRE"', ''], # Left-to-Right Embedding
	126	['IsBidiLRO', '$bid eq "LRO"', ''], # Left-to-Right Override
	127	['IsBidiAL', '$bid eq "AL"', ''], # Right-to-Left Arabic
	128	['IsBidiRLE', '$bid eq "RLE"', ''], # Right-to-Left Embedding
	129	['IsBidiRLO', '$bid eq "RLO"', ''], # Right-to-Left Override
	130	['IsBidiPDF', '$bid eq "PDF"', ''], # Pop Directional Format
	131	['IsBidiNSM', '$bid eq "NSM"', ''], # Non-Spacing Mark
	132	['IsBidiBN', '$bid eq "BN"', ''], # Boundary Neutral
	133
a0ed51b3	134	# Weak types:
	135
	136	['IsBidiEN','$bid eq "EN"', ''], # European Number
	137	['IsBidiES','$bid eq "ES"', ''], # European Number Separator
	138	['IsBidiET','$bid eq "ET"', ''], # European Number Terminator
	139	['IsBidiAN','$bid eq "AN"', ''], # Arabic Number
	140	['IsBidiCS','$bid eq "CS"', ''], # Common Number Separator
	141
	142	# Separators:
	143
	144	['IsBidiB', '$bid eq "B"', ''], # Block Separator
	145	['IsBidiS', '$bid eq "S"', ''], # Segment Separator
	146
	147	# Neutrals:
	148
	149	['IsBidiWS','$bid eq "WS"', ''], # Whitespace
	150	['IsBidiON','$bid eq "ON"', ''], # Other Neutrals ; All other
	151	# characters: punctuation,
	152	# symbols
	153
	154	# Decomposition
	155
	156	['Decomposition', '$decomp', '$decomp'],
	157	['IsDecoCanon', '$decomp && $decomp !~ /^</', ''],
	158	['IsDecoCompat', '$decomp =~ /^</', ''],
	159	['IsDCfont', '$decomp =~ /^<font>/', ''],
	160	['IsDCnoBreak', '$decomp =~ /^<noBreak>/', ''],
	161	['IsDCinitial', '$decomp =~ /^<initial>/', ''],
f59877d4	162	['IsDCmedial', '$decomp =~ /^<medial>/', ''],
a0ed51b3	163	['IsDCfinal', '$decomp =~ /^<final>/', ''],
	164	['IsDCisolated', '$decomp =~ /^<isolated>/', ''],
	165	['IsDCcircle', '$decomp =~ /^<circle>/', ''],
	166	['IsDCsuper', '$decomp =~ /^<super>/', ''],
	167	['IsDCsub', '$decomp =~ /^<sub>/', ''],
	168	['IsDCvertical', '$decomp =~ /^<vertical>/', ''],
	169	['IsDCwide', '$decomp =~ /^<wide>/', ''],
	170	['IsDCnarrow', '$decomp =~ /^<narrow>/', ''],
	171	['IsDCsmall', '$decomp =~ /^<small>/', ''],
	172	['IsDCsquare', '$decomp =~ /^<square>/', ''],
7c6f5cd2	173	['IsDCfraction', '$decomp =~ /^<fraction>/', ''],
a0ed51b3	174	['IsDCcompat', '$decomp =~ /^<compat>/', ''],
	175
	176	# Number
	177
e1b504a6	178	['Number', '$num ne ""', '$num'],
a0ed51b3	179
	180	# Mirrored
	181
	182	['IsMirrored', '$mir eq "Y"', ''],
	183
	184	# Arabic
	185
	186	['ArabLink', '1', '$link'],
	187	['ArabLnkGrp', '1', '$linkgroup'],
	188
	189	# Jamo
	190
	191	['JamoShort', '1', '$short'],
499bfa7a	192
	193	# Syllables
	194
7c6f5cd2	195	syllable_defs(),
7c6f5cd2	196
a77b4ae5	197	# Line break properties - Normative
	198
	199	['IsLbrkBK','$brk eq "BK"', ''], # Mandatory Break
	200	['IsLbrkCR','$brk eq "CR"', ''], # Carriage Return
	201	['IsLbrkLF','$brk eq "LF"', ''], # Line Feed
	202	['IsLbrkCM','$brk eq "CM"', ''], # Attached Characters and Combining Marks
	203	['IsLbrkSG','$brk eq "SG"', ''], # Surrogates
	204	['IsLbrkGL','$brk eq "GL"', ''], # Non-breaking (Glue)
	205	['IsLbrkCB','$brk eq "CB"', ''], # Contingent Break Opportunity
	206	['IsLbrkSP','$brk eq "SP"', ''], # Space
	207	['IsLbrkZW','$brk eq "ZW"', ''], # Zero Width Space
	208
	209	# Line break properties - Informative
	210	['IsLbrkXX','$brk eq "XX"', ''], # Unknown
	211	['IsLbrkOP','$brk eq "OP"', ''], # Opening Punctuation
	212	['IsLbrkCL','$brk eq "CL"', ''], # Closing Punctuation
	213	['IsLbrkQU','$brk eq "QU"', ''], # Ambiguous Quotation
	214	['IsLbrkNS','$brk eq "NS"', ''], # Non Starter
	215	['IsLbrkEX','$brk eq "EX"', ''], # Exclamation/Interrogation
	216	['IsLbrkSY','$brk eq "SY"', ''], # Symbols Allowing Breaks
	217	['IsLbrkIS','$brk eq "IS"', ''], # Infix Separator (Numeric)
	218	['IsLbrkPR','$brk eq "PR"', ''], # Prefix (Numeric)
	219	['IsLbrkPO','$brk eq "PO"', ''], # Postfix (Numeric)
	220	['IsLbrkNU','$brk eq "NU"', ''], # Numeric
	221	['IsLbrkAL','$brk eq "AL"', ''], # Ordinary Alphabetic and Symbol Characters
	222	['IsLbrkID','$brk eq "ID"', ''], # Ideographic
	223	['IsLbrkIN','$brk eq "IN"', ''], # Inseparable
	224	['IsLbrkHY','$brk eq "HY"', ''], # Hyphen
	225	['IsLbrkBB','$brk eq "BB"', ''], # Break Opportunity Before
	226	['IsLbrkBA','$brk eq "BA"', ''], # Break Opportunity After
	227	['IsLbrkSA','$brk eq "SA"', ''], # Complex Context (South East Asian)
	228	['IsLbrkAI','$brk eq "AI"', ''], # Ambiguous (Alphabetic or Ideographic)
	229	['IsLbrkB2','$brk eq "B2"', ''], # Break Opportunity Before and After
a0ed51b3	230	);
	231
	232	# This is not written for speed...
	233
9fdf68be	234	my %InId;
	235	my $InId = 0;
	236
a0ed51b3	237	foreach $file (@todo) {
	238	my ($table, $wanted, $val) = @$file;
	239	next if @ARGV and not grep { $_ eq $table } @ARGV;
9fdf68be	240	print $table, "\n";
9fdf68be	241	$table =~ s/\W+//g;
2796c109	242	if ($table =~ /^(Is\|To)(.+)/) {
a0ed51b3	243	open(OUT, ">$1/$2.pl") or die "Can't create $1/$2.pl: $!\n";
	244	}
	245	else {
	246	open(OUT, ">$table.pl") or die "Can't create $table.pl: $!\n";
	247	}
14055466	248	print OUT <<EOH;
	249	# !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
	250	# This file is built by $0 from e.g. $UnicodeData.
	251	# Any changes made here will be lost!
	252	EOH
a0ed51b3	253	print OUT <<"END";
	254	return <<'END';
	255	END
	256	print OUT proplist($table, $wanted, $val);
	257	print OUT "END\n";
	258	close OUT;
	259	}
	260
2796c109	261	# Do Scripts before Blocks so that in case of naming conflicts
	262	# the more natural one (Script) wins over the artificial one (Block).
	263
	264	print "Scripts\n";
	265	open(UD, 'Scripts.txt') or die "Can't open Scripts.txt: $!\n";
	266	open(OUT, ">Scripts.pl") or die "Can't create Scripts.pl: $!\n";
	267	print OUT <<EOH;
	268	# !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
	269	# This file is built by $0 from e.g. $UnicodeData.
	270	# Any changes made here will be lost!
	271	EOH
	272	print OUT <<"END";
	273	return <<'END';
	274	END
	275
	276	my %Scripts;
	277
	278	while (<UD>) {
	279	next if /^#/;
	280	next if /^$/;
	281	chomp;
5f9563ea	282	($code, $last, $name) = /^([0-9a-f]+)(?:\.\.([0-9a-f]+))?\s+;\s+(.+)\s+\#/i;
2796c109	283	if ($name) {
	284	my $InName = lc($name);
	285	$InName =~ s/\b(\w)/uc($1)/ge;
	286	$InName =~ s/\W+//g;
	287	my $id;
	288	unless (exists $InId{$InName}) {
	289	print "\t$InName\n";
	290	$id = $Scripts{$InName} = $InId{$InName} = $InId++;
	291	open(SCRIPT, ">In/$id.pl") or die "create In/$id.pl: $!\n";
	292	print SCRIPT <<EOH;
	293	# !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
	294	# This file is built by $0 from e.g. $UnicodeData.
	295	# Any changes made here will be lost!
	296	return <<'END';
	297	EOH
	298	close(SCRIPT);
	299	} else {
	300	$id = $InId{$InName};
	301	}
5f9563ea	302	$last = "" unless defined $last;
2796c109	303	print OUT "$code\t$last\t$name\t# $InName In/$id.pl\n";
	304	open(SCRIPT, ">>In/$id.pl");
	305	print SCRIPT <<END;
	306	$code $last
	307	END
	308	close SCRIPT;
	309	}
	310	}
	311
	312	for my $id (values %InId) {
	313	open(SCRIPT, ">>In/$id.pl");
	314	print SCRIPT <<END2;
	315	END
	316	END2
	317	close(SCRIPT);
	318	}
	319
	320	print OUT "END\n";
	321	close OUT;
	322
a0ed51b3	323	# Must treat blocks specially.
	324
	325	exit if @ARGV and not grep { $_ eq Block } @ARGV;
9fdf68be	326	print "Blocks\n";
7c6f5cd2	327	open(UD, 'Blocks.txt') or die "Can't open Blocks.txt: $!\n";
9fdf68be	328	open(OUT, ">Blocks.pl") or die "Can't create Blocks.pl: $!\n";
14055466	329	print OUT <<EOH;
	330	# !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
	331	# This file is built by $0 from e.g. $UnicodeData.
	332	# Any changes made here will be lost!
	333	EOH
a0ed51b3	334	print OUT <<"END";
	335	return <<'END';
	336	END
	337
	338	while (<UD>) {
	339	next if /^#/;
	340	next if /^$/;
	341	chomp;
9fdf68be	342	($code, $last, $name) = /^([0-9a-f]+)\.\.([0-9a-f]+); (.+)/i;
a0ed51b3	343	if ($name) {
2796c109	344	my $InName = $name;
	345	$InName =~ s/\W+//g;
	346	print "\t$InName\n";
9fdf68be	347	my $id;
2796c109	348	# TODO: only the first one of Private Use blocks qualifies
	349	unless (exists $InId{$InName}) {
	350	$InId{$InName} = $InId++;
	351	} elsif (exists $Scripts{$InName}) {
	352	$InName .= 'Block';
	353	$InId{$InName} = $InId++;
9fdf68be	354	}
2796c109	355	$id = $InId{$InName};
	356	open(BLOCK, ">In/$id.pl") or die "create In/$id.pl: $!\n";
	357	print OUT "$code\t$last\t$name\t# $InName In/$id.pl\n";
14055466	358	print BLOCK <<EOH;
	359	# !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
	360	# This file is built by $0 from e.g. $UnicodeData.
	361	# Any changes made here will be lost!
	362	EOH
a0ed51b3	363	print BLOCK <<"END2";
	364	return <<'END';
	365	$code $last
	366	END
	367	END2
	368	close BLOCK;
	369	}
	370	}
	371
	372	print OUT "END\n";
	373	close OUT;
	374
9fdf68be	375	open(INID, ">In.pl");
	376
	377	print INID <<EOH;
	378	# !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
	379	# This file is built by $0 from e.g. $UnicodeData.
	380	# Any changes made here will be lost!
	381	%utf8::In = (
	382	EOH
	383
	384	# Order doesn't matter but let's prettyprint anyway.
56ca2fc0	385	foreach my $in (sort { $InId{$a} <=> $InId{$b} } keys %InId) {
f178ed66	386	printf INID "%-40s => %3d,\n", "'$in'", $InId{$in};
9fdf68be	387	}
	388
	389	print INID ");\n";
	390
	391	close(INID);
	392
a0ed51b3	393	##################################################
	394
	395	sub proplist {
	396	my ($table, $wanted, $val) = @_;
	397	my @wanted;
	398	my $out;
	399	my $split;
	400
7c6f5cd2	401	return listFromPropFile($wanted) if $val eq $PropData;
7c6f5cd2	402
a0ed51b3	403	if ($table =~ /^Arab/) {
d357d9fe	404	open(UD, "ArabShap.txt") or warn "Can't open $table: $!";
a0ed51b3	405
	406	$split = '($code, $name, $link, $linkgroup) = split(/; */);';
	407	}
	408	elsif ($table =~ /^Jamo/) {
505afebf	409	open(UD, "Jamo.txt") or warn "Can't open $table: $!";
a0ed51b3	410
	411	$split = '($code, $short, $name) = split(/; */); $code =~ s/^U\+//;';
	412	}
499bfa7a	413	elsif ($table =~ /^IsSyl/) {
7c6f5cd2	414	open(UD, $SyllableData) or warn "Can't open $table: $!";
499bfa7a	415
	416	$split = '($code, $short, $syl) = split(/; */); $code =~ s/^U\+//;';
	417	}
a77b4ae5	418	elsif ($table =~ /^IsLbrk/) {
	419	open(UD, "LineBrk.txt") or warn "Can't open $table: $!";
	420
50fc4248	421	$split = '($code, $brk, $name) = /^([0-9a-f]+);(\w+) # (.+)/i;';
a77b4ae5	422	}
a0ed51b3	423	else {
11695a73	424	open(UD, $UnicodeData) or warn "Can't open $UnicodeData: $!";
a0ed51b3	425
	426	$split = '($code, $name, $cat, $comb, $bid, $decomp, $dec, $dig, $num, $mir, $uni1,
	427	$comment, $up, $down, $title) = split(/;/);';
	428	}
	429
	430	if ($table =~ /^(?:To\|Is)[A-Z]/) {
	431	eval <<"END";
	432	while (<UD>) {
	433	next if /^#/;
7c6f5cd2	434	next if /^\\s/;
7c6f5cd2	435	s/\\s+\$//;
a0ed51b3	436	$split
	437	if ($wanted) {
	438	push(\@wanted, [hex \$code, hex $val, \$name =~ /, First>\$/]);
	439	}
	440	}
	441	END
	442	die $@ if $@;
	443
	444	while (@wanted) {
	445	$beg = shift @wanted;
	446	$last = $beg;
	447	while (@wanted and $wanted[0]->[0] == $last->[0] + 1 and
	448	(not $val or $wanted[0]->[1] == $last->[1] + 1)) {
	449	$last = shift @wanted;
	450	}
	451	$out .= sprintf "%04x", $beg->[0];
	452	if ($beg->[2]) {
	453	$last = shift @wanted;
	454	}
	455	if ($beg == $last) {
	456	$out .= "\t";
	457	}
	458	else {
	459	$out .= sprintf "\t%04x", $last->[0];
	460	}
	461	$out .= sprintf "\t%04x", $beg->[1] if $val;
	462	$out .= "\n";
	463	}
	464	}
	465	else {
	466	eval <<"END";
	467	while (<UD>) {
	468	next if /^#/;
7c6f5cd2	469	next if /^\\s*\$/;
a0ed51b3	470	chop;
	471	$split
	472	if ($wanted) {
	473	push(\@wanted, [hex \$code, $val, \$name =~ /, First>\$/]);
	474	}
	475	}
	476	END
	477	die $@ if $@;
	478
	479	while (@wanted) {
	480	$beg = shift @wanted;
	481	$last = $beg;
	482	while (@wanted and $wanted[0]->[0] == $last->[0] + 1 and
	483	($wanted[0]->[1] eq $last->[1])) {
	484	$last = shift @wanted;
	485	}
	486	$out .= sprintf "%04x", $beg->[0];
	487	if ($beg->[2]) {
	488	$last = shift @wanted;
	489	}
	490	if ($beg == $last) {
	491	$out .= "\t";
	492	}
	493	else {
	494	$out .= sprintf "\t%04x", $last->[0];
	495	}
	496	$out .= sprintf "\t%s\n", $beg->[1];
	497	}
	498	}
	499	$out;
	500	}
11695a73	501
7c6f5cd2	502	sub listFromPropFile {
	503	my ($wanted) = @_;
	504	my $out;
	505
	506	open (UD, $PropData) or die "Can't open $PropData: $!\n";
	507	local($/) = "\n" . '*' x 43 . "\n\nProperty dump for:"; # not 42?
	508
	509	<UD>;
	510	while (<UD>) {
	511	chomp;
	512	if (s/0x[\d\w]+\s+\((.*?)\)// and $wanted eq $1) {
	513	s/\(\d+ chars\)//g;
	514	s/^\s+//mg;
	515	s/\s+$//mg;
	516	s/\.\./\t/g;
	517	$out = lc $_;
	518	last;
	519	}
	520	}
	521	close (UD);
	522	"$out\n";
	523	}
	524
	525	sub syllable_defs {
	526	my @defs;
	527	my %seen;
	528
	529	open (SD, $SyllableData) or die "Can't open $SyllableData: $!\n";
	530	while (<SD>) {
	531	next if /^\s*(#\|$)/;
	532	s/\s+$//;
	533	($code, $name, $syl) = split /; */;
	534	next unless $syl;
	535	push (@defs, ["IsSyl$syl", qq{\$syl eq "$syl"}, ''])
	536	unless $seen{$syl}++;
	537	}
	538	close (SD);
	539	return (@defs);
	540	}
	541
6dd159d1	542	# eof