#!perl -w
-use 5.012;
+use v5.15.8;
use strict;
use warnings;
require 'regen/regen_lib.pl';
+require 'regen/charset_translations.pl';
# This program outputs l1_charclass_tab.h, which defines the guts of the
# PL_charclass table. Each line is a bit map of properties that the Unicode
# new Unicode release, to make sure things haven't been changed by it.
my @properties = qw(
- ALNUMC_A
- ALNUMC_L1
- ALPHA_A
- ALPHA_L1
- BLANK_A
- BLANK_L1
+ NONLATIN1_SIMPLE_FOLD
+ NONLATIN1_FOLD
+ ALPHANUMERIC
+ ALPHA
+ ASCII
+ BLANK
+ CASED
CHARNAME_CONT
- CNTRL_A
- CNTRL_L1
- DIGIT_A
- GRAPH_A
- GRAPH_L1
- IDFIRST_A
- IDFIRST_L1
- LOWER_A
- LOWER_L1
- PRINT_A
- PRINT_L1
- PSXSPC_A
- PSXSPC_L1
- PUNCT_A
- PUNCT_L1
- SPACE_A
- SPACE_L1
- UPPER_A
- UPPER_L1
- WORDCHAR_A
- WORDCHAR_L1
- XDIGIT_A
+ CNTRL
+ DIGIT
+ GRAPH
+ IDFIRST
+ LOWER
+ NON_FINAL_FOLD
+ PRINT
+ PSXSPC
+ PUNCT
QUOTEMETA
+ SPACE
+ UPPER
+ WORDCHAR
+ XDIGIT
+ VERTSPACE
+ IS_IN_SOME_FOLD
+ MNEMONIC_CNTRL
);
# Read in the case fold mappings.
my %folded_closure;
-my $file="lib/unicore/CaseFolding.txt";
-open my $fh, "<", $file or die "Failed to read '$file': $!";
-while (<$fh>) {
- chomp;
-
- # Lines look like (without the initial '#'
- #0130; F; 0069 0307; # LATIN CAPITAL LETTER I WITH DOT ABOVE
- # Get rid of comments, ignore blank or comment-only lines
- my $line = $_ =~ s/ (?: \s* \# .* )? $ //rx;
- next unless length $line;
- my ($hex_from, $fold_type, @folded) = split /[\s;]+/, $line;
-
- my $from = hex $hex_from;
-
- # Perl only deals with C and F folds
- next if $fold_type ne 'C' and $fold_type ne 'F';
-
- # Get each code point in the range that participates in this line's fold.
- # The hash has keys of each code point in the range, and values of what it
- # folds to and what folds to it
- foreach my $hex_fold (@folded) {
- my $fold = hex $hex_fold;
- push @{$folded_closure{$fold}}, $from if $fold < 256;
- push @{$folded_closure{$from}}, $fold if $from < 256;
+my @hex_non_final_folds;
+my @non_latin1_simple_folds;
+my @folds;
+use Unicode::UCD;
+
+BEGIN { # Have to do this at compile time because using user-defined \p{property}
+
+ # Use the Unicode data file if we are on an ASCII platform (which its data
+ # is for), and it is in the modern format (starting in Unicode 3.1.0) and
+ # it is available. This avoids being affected by potential bugs
+ # introduced by other layers of Perl
+ my $file="lib/unicore/CaseFolding.txt";
+
+ if (ord('A') == 65
+ && pack("C*", split /\./, Unicode::UCD::UnicodeVersion()) ge v3.1.0
+ && open my $fh, "<", $file)
+ {
+ @folds = <$fh>;
+ }
+ else {
+ my ($invlist_ref, $invmap_ref, undef, $default)
+ = Unicode::UCD::prop_invmap('Case_Folding');
+ for my $i (0 .. @$invlist_ref - 1 - 1) {
+ next if $invmap_ref->[$i] == $default;
+ my $adjust = -1;
+ for my $j ($invlist_ref->[$i] .. $invlist_ref->[$i+1] -1) {
+ $adjust++;
+
+ # Single-code point maps go to a 'C' type
+ if (! ref $invmap_ref->[$i]) {
+ push @folds, sprintf("%04X; C; %04X\n",
+ $j,
+ $invmap_ref->[$i] + $adjust);
+ }
+ else { # Multi-code point maps go to 'F'. prop_invmap()
+ # guarantees that no adjustment is needed for these,
+ # as the range will contain just one element
+ push @folds, sprintf("%04X; F; %s\n",
+ $j,
+ join " ", map { sprintf "%04X", $_ }
+ @{$invmap_ref->[$i]});
+ }
+ }
+ }
+ }
+
+ for (@folds) {
+ chomp;
+
+ # Lines look like (without the initial '#'
+ #0130; F; 0069 0307; # LATIN CAPITAL LETTER I WITH DOT ABOVE
+ # Get rid of comments, ignore blank or comment-only lines
+ my $line = $_ =~ s/ (?: \s* \# .* )? $ //rx;
+ next unless length $line;
+ my ($hex_from, $fold_type, @folded) = split /[\s;]+/, $line;
+
+ my $from = hex $hex_from;
+
+ # Perl only deals with S, C, and F folds
+ next if $fold_type ne 'C' and $fold_type ne 'F' and $fold_type ne 'S';
+
+ # Get each code point in the range that participates in this line's fold.
+ # The hash has keys of each code point in the range, and values of what it
+ # folds to and what folds to it
+ for my $i (0 .. @folded - 1) {
+ my $hex_fold = $folded[$i];
+ my $fold = hex $hex_fold;
+ push @{$folded_closure{$fold}}, $from if $fold < 256;
+ push @{$folded_closure{$from}}, $fold if $from < 256;
+
+ if (($fold_type eq 'C' || $fold_type eq 'S')
+ && ($fold < 256 != $from < 256))
+ {
+ # Fold is simple (hence can't be a non-final fold, so the 'if'
+ # above is mutualy exclusive from the 'if below) and crosses
+ # 255/256 boundary. We keep track of the Latin1 code points
+ # in such folds.
+ push @non_latin1_simple_folds, ($fold < 256)
+ ? $fold
+ : $from;
+ }
+ elsif ($i < @folded-1
+ && $fold < 256
+ && ! grep { $_ eq $hex_fold } @hex_non_final_folds)
+ {
+ push @hex_non_final_folds, $hex_fold;
+
+ # Also add the upper case, which in the latin1 range folds to
+ # $fold
+ push @hex_non_final_folds, sprintf "%04X", ord uc chr $fold;
+ }
+ }
+ }
+
+ # Now having read all the lines, combine them into the full closure of each
+ # code point in the range by adding lists together that share a common
+ # element
+ foreach my $folded (keys %folded_closure) {
+ foreach my $from (grep { $_ < 256 } @{$folded_closure{$folded}}) {
+ push @{$folded_closure{$from}}, @{$folded_closure{$folded}};
+ }
+ }
+
+ # We have the single-character folds that cross the 255/256, like KELVIN
+ # SIGN => 'k', but we need the closure, so add like 'K' to it
+ foreach my $folded (@non_latin1_simple_folds) {
+ foreach my $fold (@{$folded_closure{$folded}}) {
+ if ($fold < 256 && ! grep { $fold == $_ } @non_latin1_simple_folds) {
+ push @non_latin1_simple_folds, $fold;
+ }
+ }
}
}
-# Now having read all the lines, combine them into the full closure of each
-# code point in the range by adding lists together that share a common element
-foreach my $folded (keys %folded_closure) {
- foreach my $from (grep { $_ < 256 } @{$folded_closure{$folded}}) {
- push @{$folded_closure{$from}}, @{$folded_closure{$folded}};
+sub Is_Non_Latin1_Fold {
+ my @return;
+
+ foreach my $folded (keys %folded_closure) {
+ push @return, sprintf("%X", $folded), if grep { $_ > 255 }
+ @{$folded_closure{$folded}};
}
+ return join("\n", @return) . "\n";
}
-my @bits; # Bit map for each code point
+sub Is_Non_Latin1_Simple_Fold { # Latin1 code points that are folded to by
+ # non-Latin1 code points as single character
+ # folds
+ return join("\n", map { sprintf "%X", $_ } @non_latin1_simple_folds) . "\n";
+}
-foreach my $folded (keys %folded_closure) {
- $bits[$folded] = "_CC_NONLATIN1_FOLD" if grep { $_ > 255 }
- @{$folded_closure{$folded}};
+sub Is_Non_Final_Fold {
+ return join("\n", @hex_non_final_folds) . "\n";
}
+my @bits; # Bit map for each code point
+
# For each character, calculate which properties it matches.
for my $ord (0..255) {
my $char = chr($ord);
- utf8::upgrade($char); # Important to use Unicode semantics!
+ utf8::upgrade($char); # Important to use Unicode rules!
# Look at all the properties we care about here.
for my $property (@properties) {
if (! ($name =~ s/_L1$//)) {
# Here, isn't an _L1. If its _A, it's automatically false for
- # non-ascii. The only one current one without a suffix is valid
- # over the whole range.
- next if $name =~ s/_A$// && $ord >= 128;
-
+ # non-ascii. The only current ones (besides ASCII) without a
+ # suffix are valid over the whole range.
+ next if $name =~ s/_A$// && $char !~ /\p{ASCII}/;
}
my $re;
if ($name eq 'PUNCT') {;
# just \pP outside it.
$re = qr/\p{Punct}|[^\P{Symbol}\P{ASCII}]/;
} elsif ($name eq 'CHARNAME_CONT') {;
- $re = qr/[-\p{XPosixWord} ():\xa0]/;
+ $re = qr/\p{_Perl_Charname_Continue}/,
} elsif ($name eq 'SPACE') {;
$re = qr/\p{XPerlSpace}/;
} elsif ($name eq 'IDFIRST') {
$re = qr/[\v\p{Space}]/;
} elsif ($name eq 'WORDCHAR') {
$re = qr/\p{XPosixWord}/;
- } elsif ($name eq 'ALNUMC') {
+ } elsif ($name eq 'ALPHANUMERIC') {
# Like \w, but no underscore
$re = qr/\p{Alnum}/;
- } elsif ($name eq 'OCTAL') {
- $re = qr/[0-7]/;
} elsif ($name eq 'QUOTEMETA') {
$re = qr/\p{_Perl_Quotemeta}/;
+ } elsif ($name eq 'NONLATIN1_FOLD') {
+ $re = qr/\p{Is_Non_Latin1_Fold}/;
+ } elsif ($name eq 'NONLATIN1_SIMPLE_FOLD') {
+ $re = qr/\p{Is_Non_Latin1_Simple_Fold}/;
+ } elsif ($name eq 'NON_FINAL_FOLD') {
+ $re = qr/\p{Is_Non_Final_Fold}/;
+ } elsif ($name eq 'IS_IN_SOME_FOLD') {
+ $re = qr/\p{_Perl_Any_Folds}/;
+ } elsif ($name eq 'MNEMONIC_CNTRL') {
+ # These are the control characters that there are mnemonics for
+ $re = qr/[\a\b\e\f\n\r\t]/;
} else { # The remainder have the same name and values as Unicode
$re = eval "qr/\\p{$name}/";
use Carp;
#print "$ord, $name $property, $re\n";
if ($char =~ $re) { # Add this property if matches
$bits[$ord] .= '|' if $bits[$ord];
- $bits[$ord] .= "_CC_$property";
+ $bits[$ord] .= "(1U<<_CC_$property)";
}
}
#print __LINE__, " $ord $char $bits[$ord]\n";
}
-# Names of C0 controls
-my @C0 = qw (
- NUL
- SOH
- STX
- ETX
- EOT
- ENQ
- ACK
- BEL
- BS
- HT
- LF
- VT
- FF
- CR
- SO
- SI
- DLE
- DC1
- DC2
- DC3
- DC4
- NAK
- SYN
- ETB
- CAN
- EOM
- SUB
- ESC
- FS
- GS
- RS
- US
- );
-
-# Names of C1 controls, plus the adjacent DEL
-my @C1 = qw(
- DEL
- PAD
- HOP
- BPH
- NBH
- IND
- NEL
- SSA
- ESA
- HTS
- HTJ
- VTS
- PLD
- PLU
- RI
- SS2
- SS3
- DCS
- PU1
- PU2
- STS
- CCH
- MW
- SPA
- EPA
- SOS
- SGC
- SCI
- CSI
- ST
- OSC
- PM
- APC
- );
-
my $out_fh = open_new('l1_char_class_tab.h', '>',
{style => '*', by => $0,
- from => "property definitions and $file"});
+ from => "property definitions"});
+
+print $out_fh <<END;
+/* For code points whose position is not the same as Unicode, both are shown
+ * in the comment*/
+END
# Output the table using fairly short names for each char.
-for my $ord (0..255) {
- my $name;
- if ($ord < 32) { # A C0 control
- $name = $C0[$ord];
- } elsif ($ord > 32 && $ord < 127) { # Graphic
- $name = "'" . chr($ord) . "'";
- } elsif ($ord >= 127 && $ord <= 0x9f) {
- $name = $C1[$ord - 127]; # A C1 control + DEL
- } else { # SPACE, or, if Latin1, shorten the name */
- use charnames();
- $name = charnames::viacode($ord);
- $name =~ s/LATIN CAPITAL LETTER //
- || $name =~ s/LATIN SMALL LETTER (.*)/\L$1/;
+foreach my $charset (get_supported_code_pages()) {
+ my @a2n = @{get_a2n($charset)};
+ my @out;
+
+ print $out_fh "\n" . get_conditional_compile_line_start($charset);
+ for my $ord (0..255) {
+ my $name;
+ my $char = chr $ord;
+ if ($char =~ /\p{PosixGraph}/) {
+ my $quote = $char eq "'" ? '"' : "'";
+ $name = $quote . chr($ord) . $quote;
+ }
+ elsif ($char =~ /\p{XPosixGraph}/) {
+ use charnames();
+ $name = charnames::viacode($ord);
+ $name =~ s/LATIN CAPITAL LETTER //
+ or $name =~ s/LATIN SMALL LETTER (.*)/\L$1/
+ or $name =~ s/ SIGN\b//
+ or $name =~ s/EXCLAMATION MARK/'!'/
+ or $name =~ s/QUESTION MARK/'?'/
+ or $name =~ s/QUOTATION MARK/QUOTE/
+ or $name =~ s/ INDICATOR//;
+ $name =~ s/\bWITH\b/\L$&/;
+ $name =~ s/\bONE\b/1/;
+ $name =~ s/\b(TWO|HALF)\b/2/;
+ $name =~ s/\bTHREE\b/3/;
+ $name =~ s/\b QUARTER S? \b/4/x;
+ $name =~ s/VULGAR FRACTION (.) (.)/$1\/$2/;
+ $name =~ s/\bTILDE\b/'~'/i
+ or $name =~ s/\bCIRCUMFLEX\b/'^'/i
+ or $name =~ s/\bSTROKE\b/'\/'/i
+ or $name =~ s/ ABOVE\b//i;
+ }
+ else {
+ use Unicode::UCD qw(prop_invmap);
+ my ($list_ref, $map_ref, $format) = prop_invmap("Name_Alias");
+ if ($format !~ /^s/) {
+ use Carp;
+ carp "Unexpected format '$format' for 'Name_Alias";
+ last;
+ }
+ my $which = Unicode::UCD::search_invlist($list_ref, $ord);
+ if (! defined $which) {
+ use Carp;
+ carp "No name found for code pont $ord";
+ }
+ else {
+ my $map = $map_ref->[$which];
+ if (! ref $map) {
+ $name = $map;
+ }
+ else {
+ # Just pick the first abbreviation if more than one
+ my @names = grep { $_ =~ /abbreviation/ } @$map;
+ $name = $names[0];
+ }
+ $name =~ s/:.*//;
+ }
+ }
+ my $index = $a2n[$ord];
+ $out[$index] = ($ord == $index)
+ ? sprintf "/* U+%02X %s */ %s,\n", $ord, $name, $bits[$ord]
+ : sprintf "/* 0x%02X U+%02X %s */ %s,\n", $index, $ord, $name, $bits[$ord];
}
- printf $out_fh "/* U+%02X %s */ %s,\n", $ord, $name, $bits[$ord];
+ print $out_fh join "", @out;
+ print $out_fh "\n" . get_conditional_compile_line_end();
}
read_only_bottom_close_and_rename($out_fh)