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 H_UNICODE_CONSTANTS /* Guard against nested #includes */
16 #define H_UNICODE_CONSTANTS 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 =head1 Unicode Support
34 =for apidoc AmU|placeholder|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 AmU|placeholder|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 my $version = Unicode::UCD::UnicodeVersion();
58 my ($major, $dot, $dotdot) = $version =~ / (.*?) \. (.*?) (?: \. (.*) )? $ /x;
59 $dotdot = 0 unless defined $dotdot;
62 #define UNICODE_MAJOR_VERSION $major
63 #define UNICODE_DOT_VERSION $dot
64 #define UNICODE_DOT_DOT_VERSION $dotdot
68 # The data are at __DATA__ in this file.
72 foreach my $charset (get_supported_code_pages()) {
73 print $out_fh "\n" . get_conditional_compile_line_start($charset);
75 my @a2n = @{get_a2n($charset)};
80 # Convert any '#' comments to /* ... */; empty lines and comments are
81 # output as blank lines
82 if ($_ =~ m/ ^ \s* (?: \# ( .* ) )? $ /x) {
83 my $comment_body = $1 // "";
84 if ($comment_body ne "") {
85 print $out_fh "/* $comment_body */\n";
93 unless ($_ =~ m/ ^ ( [^\ ]* ) # Name or code point token
94 (?: [\ ]+ ( [^ ]* ) )? # optional flag
95 (?: [\ ]+ ( .* ) )? # name if unnamed; flag is required
98 die "Unexpected syntax at line $.: $_\n";
103 my $desired_name = $3;
107 my $U_cp; # code point in Unicode (not-native) terms
109 if ($name_or_cp =~ /^U\+(.*)/) {
111 $name = charnames::viacode($name_or_cp);
112 if (! defined $name) {
113 next if $flag =~ /skip_if_undef/;
114 die "Unknown code point '$name_or_cp' at line $.: $_\n" unless $desired_name;
120 die "Unknown name '$name' at line $.: $_\n" unless defined $name;
121 $U_cp = charnames::vianame($name =~ s/_/ /gr);
128 $name = $desired_name if $name eq "" && $desired_name;
129 $name =~ s/[- ]/_/g; # The macro name can have no blanks nor dashes
133 if (defined $flag && $flag eq 'native') {
134 die "Are you sure you want to run this on an above-Latin1 code point?" if $cp > 0xff;
136 $str = sprintf "0x%02X", $cp; # Is a numeric constant
139 $str = join "", map { sprintf "\\x%02X", ord $_ } split //, cp_2_utfbytes($U_cp, $charset);
142 if (! defined $flag || $flag =~ /^ string (_skip_if_undef)? $/x) {
143 $str = "\"$str\""; # Will be a string constant
144 } elsif ($flag eq 'tail') {
145 $str =~ s/\\x..//; # Remove the first byte
147 $str = "\"$str\""; # Will be a string constant
149 elsif ($flag eq 'first') {
150 $str =~ s/ \\x ( .. ) .* /$1/x; # Get the two nibbles of the 1st byte
151 $suffix .= '_FIRST_BYTE';
152 $str = "0x$str"; # Is a numeric constant
155 die "Unknown flag at line $.: $_\n";
158 printf $out_fh "# define %s%s %s /* U+%04X */\n", $name, $suffix, $str, $U_cp;
162 for my $i (0x20 .. 0x7E) {
163 $max_PRINT_A = $a2n[$i] if $a2n[$i] > $max_PRINT_A;
165 printf $out_fh "# define MAX_PRINT_A_FOR_USE_ONLY_BY_REGCOMP_DOT_C 0x%02X /* The max code point that isPRINT_A */\n", $max_PRINT_A;
167 print $out_fh "\n" . get_conditional_compile_line_end();
171 use Unicode::UCD 'prop_invlist';
174 my @other_invlist = prop_invlist("Other");
175 for (my $i = 0; $i < @other_invlist; $i += 2) {
176 $count += ((defined $other_invlist[$i+1])
177 ? $other_invlist[$i+1]
179 - $other_invlist[$i];
181 printf $out_fh "\n/* The number of code points not matching \\pC */\n"
182 . "#define NON_OTHER_COUNT_FOR_USE_ONLY_BY_REGCOMP_DOT_C %d\n",
185 # If this release has both the CWCM and CWCF properties, find the highest code
186 # point which changes under any case change. We can use this to short-circuit
188 my @cwcm = prop_invlist('CWCM');
190 my @cwcf = prop_invlist('CWCF');
192 my $max = ($cwcm[-1] < $cwcf[-1])
195 printf $out_fh "\n/* The highest code point that has any type of case change */\n"
196 . "#define HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C 0x%X\n",
201 print $out_fh "\n#endif /* H_UNICODE_CONSTANTS */\n";
203 read_only_bottom_close_and_rename($out_fh);
207 # Note that any apidoc comments you want in the file need to be added to one
208 # of the prints above
210 # A blank line is output as-is.
211 # Comments (lines whose first non-blank is a '#') are converted to C-style,
212 # though empty comments are converted to blank lines. Otherwise, each line
213 # represents one #define, and begins with either a Unicode character name with
214 # the blanks and dashes in it squeezed out or replaced by underscores; or it
215 # may be a hexadecimal Unicode code point of the form U+xxxx. In the latter
216 # case, the name will be looked-up to use as the name of the macro. In either
217 # case, the macro name will have suffixes as listed above, and all blanks and
218 # dashes will be replaced by underscores.
220 # Each line may optionally have one of the following flags on it, separated by
221 # white space from the initial token.
222 # string indicates that the output is to be of the string form
223 # described in the comments above that are placed in the file.
224 # string_skip_ifundef is the same as 'string', but instead of dying if the
225 # code point doesn't exist, the line is just skipped: no output is
227 # first indicates that the output is to be of the FIRST_BYTE form.
228 # tail indicates that the output is of the _TAIL form.
229 # native indicates that the output is the code point, converted to the
230 # platform's native character set if applicable
232 # If the code point has no official name, the desired name may be appended
233 # after the flag, which will be ignored if there is an official name.
235 # This program is used to make it convenient to create compile time constants
236 # of UTF-8, and to generate proper EBCDIC as well as ASCII without manually
237 # having to figure things out.
247 U+1E9E string_skip_if_undef