4 require './regen/regen_lib.pl';
5 require './regen/charset_translations.pl';
7 use charnames qw(:loose);
9 my $out_fh = open_new('unicode_constants.h', '>',
10 {style => '*', by => $0,
11 from => "Unicode data"});
15 #ifndef PERL_UNICODE_CONSTANTS_H_ /* Guard against nested #includes */
16 #define PERL_UNICODE_CONSTANTS_H_ 1
18 /* This file contains #defines for the version of Unicode being used and
19 * various Unicode code points. The values the code point macros expand to
20 * are the native Unicode code point, or all or portions of the UTF-8 encoding
21 * for the code point. In the former case, the macro name has the suffix
22 * "_NATIVE"; otherwise, the suffix "_UTF8".
24 * The macros that have the suffix "_UTF8" may have further suffixes, as
26 * "_FIRST_BYTE" if the value is just the first byte of the UTF-8
27 * representation; the value will be a numeric constant.
28 * "_TAIL" if instead it represents all but the first byte. This, and
29 * with no additional suffix are both string constants */
32 =for apidoc_section \$unicode
34 =for apidoc AmnU|const char *|BOM_UTF8
36 This is a macro that evaluates to a string constant of the UTF-8 bytes that
37 define the Unicode BYTE ORDER MARK (U+FEFF) for the platform that perl
38 is compiled on. This allows code to use a mnemonic for this character that
39 works on both ASCII and EBCDIC platforms.
40 S<C<sizeof(BOM_UTF8) - 1>> can be used to get its length in
43 =for apidoc AmnU|const char *|REPLACEMENT_CHARACTER_UTF8
45 This is a macro that evaluates to a string constant of the UTF-8 bytes that
46 define the Unicode REPLACEMENT CHARACTER (U+FFFD) for the platform that perl
47 is compiled on. This allows code to use a mnemonic for this character that
48 works on both ASCII and EBCDIC platforms.
49 S<C<sizeof(REPLACEMENT_CHARACTER_UTF8) - 1>> can be used to get its length in
57 sub backslash_x_form($$;$) {
58 # Output the code point represented by the byte string $bytes as a
59 # sequence of \x{} constants. $bytes should be the UTF-8 for the code
60 # point if the final parameter is absent or empty. Otherwise it should be
61 # the Latin1 code point itself.
63 # The output is translated into the character set '$charset'.
65 my ($bytes, $charset, $non_utf8) = @_;
67 die "Must be utf8 if above 255" if $bytes > 255;
68 my $a2n = get_a2n($charset);
69 return sprintf "\\x%02X", $a2n->[$bytes];
72 return join "", map { sprintf "\\x%02X", ord $_ }
73 split //, cp_2_utfbytes($bytes, $charset);
77 my $version = Unicode::UCD::UnicodeVersion();
78 my ($major, $dot, $dotdot) = $version =~ / (.*?) \. (.*?) (?: \. (.*) )? $ /x;
79 $dotdot = 0 unless defined $dotdot;
82 #define UNICODE_MAJOR_VERSION $major
83 #define UNICODE_DOT_VERSION $dot
84 #define UNICODE_DOT_DOT_VERSION $dotdot
88 # The data are at __DATA__ in this file.
92 foreach my $charset (get_supported_code_pages()) {
93 print $out_fh "\n" . get_conditional_compile_line_start($charset);
95 my @a2n = @{get_a2n($charset)};
100 # Convert any '#' comments to /* ... */; empty lines and comments are
101 # output as blank lines
102 if ($_ =~ m/ ^ \s* (?: \# ( .* ) )? $ /x) {
103 my $comment_body = $1 // "";
104 if ($comment_body ne "") {
105 print $out_fh "/* $comment_body */\n";
113 unless ($_ =~ m/ ^ ( [^\ ]* ) # Name or code point token
114 (?: [\ ]+ ( [^ ]* ) )? # optional flag
115 (?: [\ ]+ ( .* ) )? # name if unnamed; flag is required
118 die "Unexpected syntax at line $.: $_\n";
123 my $desired_name = $3;
127 my $U_cp; # code point in Unicode (not-native) terms
129 if ($name_or_cp =~ /^U\+(.*)/) {
131 $name = charnames::viacode($name_or_cp);
132 if (! defined $name) {
133 next if $flag =~ /skip_if_undef/;
134 die "Unknown code point '$name_or_cp' at line $.: $_\n" unless $desired_name;
140 die "Unknown name '$name' at line $.: $_\n" unless defined $name;
141 $U_cp = charnames::vianame($name =~ s/_/ /gr);
148 $name = $desired_name if $name eq "" && $desired_name;
149 $name =~ s/[- ]/_/g; # The macro name can have no blanks nor dashes
153 if (defined $flag && $flag eq 'native') {
154 die "Are you sure you want to run this on an above-Latin1 code point?" if $cp > 0xff;
156 $str = sprintf "0x%02X", $cp; # Is a numeric constant
159 $str = backslash_x_form($U_cp, $charset);
162 if (! defined $flag || $flag =~ /^ string (_skip_if_undef)? $/x) {
163 $str = "\"$str\""; # Will be a string constant
164 } elsif ($flag eq 'tail') {
165 $str =~ s/\\x..//; # Remove the first byte
167 $str = "\"$str\""; # Will be a string constant
169 elsif ($flag eq 'first') {
170 $str =~ s/ \\x ( .. ) .* /$1/x; # Get the two nibbles of the 1st byte
171 $suffix .= '_FIRST_BYTE';
172 $str = "0x$str"; # Is a numeric constant
175 die "Unknown flag at line $.: $_\n";
178 printf $out_fh "# define %s%s %s /* U+%04X */\n", $name, $suffix, $str, $U_cp;
182 for my $i (0x20 .. 0x7E) {
183 $max_PRINT_A = $a2n[$i] if $a2n[$i] > $max_PRINT_A;
185 $max_PRINT_A = sprintf "0x%02X", $max_PRINT_A;
186 print $out_fh <<"EOT";
188 # ifdef PERL_IN_REGCOMP_C
189 # define MAX_PRINT_A $max_PRINT_A /* The max code point that isPRINT_A */
193 print $out_fh get_conditional_compile_line_end();
197 use Unicode::UCD 'prop_invlist';
200 my @other_invlist = prop_invlist("Other");
201 for (my $i = 0; $i < @other_invlist; $i += 2) {
202 $count += ((defined $other_invlist[$i+1])
203 ? $other_invlist[$i+1]
205 - $other_invlist[$i];
207 $count = 0x110000 - $count;
208 print $out_fh <<~"EOT";
210 /* The number of code points not matching \\pC */
211 #ifdef PERL_IN_REGCOMP_C
212 # define NON_OTHER_COUNT $count
216 # If this release has both the CWCM and CWCF properties, find the highest code
217 # point which changes under any case change. We can use this to short-circuit
219 my @cwcm = prop_invlist('CWCM');
221 my @cwcf = prop_invlist('CWCF');
223 my $max = ($cwcm[-1] < $cwcf[-1])
226 $max = sprintf "0x%X", $max - 1;
227 print $out_fh <<~"EOS";
229 /* The highest code point that has any type of case change */
230 #ifdef PERL_IN_UTF8_C
231 # define HIGHEST_CASE_CHANGING_CP $max
237 print $out_fh "\n#endif /* PERL_UNICODE_CONSTANTS_H_ */\n";
239 read_only_bottom_close_and_rename($out_fh);
243 # Note that any apidoc comments you want in the file need to be added to one
244 # of the prints above
246 # A blank line is output as-is.
247 # Comments (lines whose first non-blank is a '#') are converted to C-style,
248 # though empty comments are converted to blank lines. Otherwise, each line
249 # represents one #define, and begins with either a Unicode character name with
250 # the blanks and dashes in it squeezed out or replaced by underscores; or it
251 # may be a hexadecimal Unicode code point of the form U+xxxx. In the latter
252 # case, the name will be looked-up to use as the name of the macro. In either
253 # case, the macro name will have suffixes as listed above, and all blanks and
254 # dashes will be replaced by underscores.
256 # Each line may optionally have one of the following flags on it, separated by
257 # white space from the initial token.
258 # string indicates that the output is to be of the string form
259 # described in the comments above that are placed in the file.
260 # string_skip_ifundef is the same as 'string', but instead of dying if the
261 # code point doesn't exist, the line is just skipped: no output is
263 # first indicates that the output is to be of the FIRST_BYTE form.
264 # tail indicates that the output is of the _TAIL form.
265 # native indicates that the output is the code point, converted to the
266 # platform's native character set if applicable
268 # If the code point has no official name, the desired name may be appended
269 # after the flag, which will be ignored if there is an official name.
271 # This program is used to make it convenient to create compile time constants
272 # of UTF-8, and to generate proper EBCDIC as well as ASCII without manually
273 # having to figure things out.
281 U+1E9E string_skip_if_undef
296 U+10FFFF string MAX_UNICODE