5 require 'regen/regen_lib.pl';
6 require 'regen/charset_translations.pl';
8 # This program outputs l1_charclass_tab.h, which defines the guts of the
9 # PL_charclass table. Each line is a bit map of properties that the Unicode
10 # code point at the corresponding position in the table array has. The first
11 # line corresponds to code point U+0000, NULL, the last line to U+00FF. For
12 # an application to see if the code point "i" has a particular property, it
14 # 'PL_charclass[i] & BIT'
15 # The bit names are of the form '_CC_property_suffix', where 'CC' stands for
16 # character class, and 'property' is the corresponding property, and 'suffix'
17 # is one of '_A' to mean the property is true only if the corresponding code
18 # point is ASCII, and '_L1' means that the range includes any Latin1
19 # character (ISO-8859-1 including the C0 and C1 controls). A property without
20 # these suffixes does not have different forms for both ranges.
22 # This program need be run only when adding new properties to it, or upon a
23 # new Unicode release, to make sure things haven't been changed by it.
50 BACKSLASH_FOO_LBRACE_IS_META
53 # Read in the case fold mappings.
55 my @hex_non_final_folds;
56 my @non_latin1_simple_folds;
60 BEGIN { # Have to do this at compile time because using user-defined \p{property}
62 # Use the Unicode data file if we are on an ASCII platform (which its data
63 # is for), and it is in the modern format (starting in Unicode 3.1.0) and
64 # it is available. This avoids being affected by potential bugs
65 # introduced by other layers of Perl
66 my $file="lib/unicore/CaseFolding.txt";
69 && pack("C*", split /\./, Unicode::UCD::UnicodeVersion()) ge v3.1.0
70 && open my $fh, "<", $file)
75 my ($invlist_ref, $invmap_ref, undef, $default)
76 = Unicode::UCD::prop_invmap('Case_Folding');
77 for my $i (0 .. @$invlist_ref - 1 - 1) {
78 next if $invmap_ref->[$i] == $default;
80 for my $j ($invlist_ref->[$i] .. $invlist_ref->[$i+1] -1) {
83 # Single-code point maps go to a 'C' type
84 if (! ref $invmap_ref->[$i]) {
85 push @folds, sprintf("%04X; C; %04X\n",
87 $invmap_ref->[$i] + $adjust);
89 else { # Multi-code point maps go to 'F'. prop_invmap()
90 # guarantees that no adjustment is needed for these,
91 # as the range will contain just one element
92 push @folds, sprintf("%04X; F; %s\n",
94 join " ", map { sprintf "%04X", $_ }
95 @{$invmap_ref->[$i]});
104 # Lines look like (without the initial '#'
105 #0130; F; 0069 0307; # LATIN CAPITAL LETTER I WITH DOT ABOVE
106 # Get rid of comments, ignore blank or comment-only lines
107 my $line = $_ =~ s/ (?: \s* \# .* )? $ //rx;
108 next unless length $line;
109 my ($hex_from, $fold_type, @folded) = split /[\s;]+/, $line;
111 my $from = hex $hex_from;
113 # Perl only deals with S, C, and F folds
114 next if $fold_type ne 'C' and $fold_type ne 'F' and $fold_type ne 'S';
116 # Get each code point in the range that participates in this line's fold.
117 # The hash has keys of each code point in the range, and values of what it
118 # folds to and what folds to it
119 for my $i (0 .. @folded - 1) {
120 my $hex_fold = $folded[$i];
121 my $fold = hex $hex_fold;
122 push @{$folded_closure{$fold}}, $from if $fold < 256;
123 push @{$folded_closure{$from}}, $fold if $from < 256;
125 if (($fold_type eq 'C' || $fold_type eq 'S')
126 && ($fold < 256 != $from < 256))
128 # Fold is simple (hence can't be a non-final fold, so the 'if'
129 # above is mutualy exclusive from the 'if below) and crosses
130 # 255/256 boundary. We keep track of the Latin1 code points
132 push @non_latin1_simple_folds, ($fold < 256)
136 elsif ($i < @folded-1
138 && ! grep { $_ eq $hex_fold } @hex_non_final_folds)
140 push @hex_non_final_folds, $hex_fold;
142 # Also add the upper case, which in the latin1 range folds to
144 push @hex_non_final_folds, sprintf "%04X", ord uc chr $fold;
149 # Now having read all the lines, combine them into the full closure of each
150 # code point in the range by adding lists together that share a common
152 foreach my $folded (keys %folded_closure) {
153 foreach my $from (grep { $_ < 256 } @{$folded_closure{$folded}}) {
154 push @{$folded_closure{$from}}, @{$folded_closure{$folded}};
158 # We have the single-character folds that cross the 255/256, like KELVIN
159 # SIGN => 'k', but we need the closure, so add like 'K' to it
160 foreach my $folded (@non_latin1_simple_folds) {
161 foreach my $fold (@{$folded_closure{$folded}}) {
162 if ($fold < 256 && ! grep { $fold == $_ } @non_latin1_simple_folds) {
163 push @non_latin1_simple_folds, $fold;
169 sub Is_Non_Latin1_Fold {
172 foreach my $folded (keys %folded_closure) {
173 push @return, sprintf("%X", $folded), if grep { $_ > 255 }
174 @{$folded_closure{$folded}};
176 return join("\n", @return) . "\n";
179 sub Is_Non_Latin1_Simple_Fold { # Latin1 code points that are folded to by
180 # non-Latin1 code points as single character
182 return join("\n", map { sprintf "%X", $_ } @non_latin1_simple_folds) . "\n";
185 sub Is_Non_Final_Fold {
186 return join("\n", @hex_non_final_folds) . "\n";
189 my @bits; # Bit map for each code point
191 # For each character, calculate which properties it matches.
192 for my $ord (0..255) {
193 my $char = chr($ord);
194 utf8::upgrade($char); # Important to use Unicode rules!
196 # Look at all the properties we care about here.
197 for my $property (@properties) {
198 my $name = $property;
200 # Remove the suffix to get the actual property name.
201 # Currently the suffixes are '_L1', '_A', and none.
202 # If is a latin1 version, no further checking is needed.
203 if (! ($name =~ s/_L1$//)) {
205 # Here, isn't an _L1. If its _A, it's automatically false for
206 # non-ascii. The only current ones (besides ASCII) without a
207 # suffix are valid over the whole range.
208 next if $name =~ s/_A$// && $char !~ /\p{ASCII}/;
211 if ($name eq 'PUNCT') {;
213 # Sadly, this is inconsistent: \pP and \pS for the ascii range,
214 # just \pP outside it.
215 $re = qr/\p{Punct}|[^\P{Symbol}\P{ASCII}]/;
216 } elsif ($name eq 'CHARNAME_CONT') {;
217 $re = qr/\p{_Perl_Charname_Continue}/,
218 } elsif ($name eq 'SPACE') {;
219 $re = qr/\p{XPerlSpace}/;
220 } elsif ($name eq 'IDFIRST') {
221 $re = qr/[_\p{Alpha}]/;
222 } elsif ($name eq 'PSXSPC') {
223 $re = qr/[\v\p{Space}]/;
224 } elsif ($name eq 'WORDCHAR') {
225 $re = qr/\p{XPosixWord}/;
226 } elsif ($name eq 'ALPHANUMERIC') {
227 # Like \w, but no underscore
229 } elsif ($name eq 'QUOTEMETA') {
230 $re = qr/\p{_Perl_Quotemeta}/;
231 } elsif ($name eq 'NONLATIN1_FOLD') {
232 $re = qr/\p{Is_Non_Latin1_Fold}/;
233 } elsif ($name eq 'NONLATIN1_SIMPLE_FOLD') {
234 $re = qr/\p{Is_Non_Latin1_Simple_Fold}/;
235 } elsif ($name eq 'NON_FINAL_FOLD') {
236 $re = qr/\p{Is_Non_Final_Fold}/;
237 } elsif ($name eq 'IS_IN_SOME_FOLD') {
238 $re = qr/\p{_Perl_Any_Folds}/;
239 } elsif ($name eq 'BACKSLASH_FOO_LBRACE_IS_META') {
241 # This is true for FOO where FOO is the varying character in:
243 # and the sequence has non-literal meaning to Perl; so it is true
244 # for 'x' because \x{ is special, but not 'a' because \a{ isn't.
246 } else { # The remainder have the same name and values as Unicode
247 $re = eval "qr/\\p{$name}/";
249 carp $@ if ! defined $re;
251 #print "$ord, $name $property, $re\n";
252 if ($char =~ $re) { # Add this property if matches
253 $bits[$ord] .= '|' if $bits[$ord];
254 $bits[$ord] .= "(1U<<_CC_$property)";
257 #print __LINE__, " $ord $char $bits[$ord]\n";
260 my $out_fh = open_new('l1_char_class_tab.h', '>',
261 {style => '*', by => $0,
262 from => "property definitions"});
265 /* For code points whose position is not the same as Unicode, both are shown
269 # Output the table using fairly short names for each char.
270 foreach my $charset (get_supported_code_pages()) {
271 my @a2n = get_a2n($charset);
274 print $out_fh "\n" . get_conditional_compile_line_start($charset);
275 for my $ord (0..255) {
278 if ($char =~ /\p{PosixGraph}/) {
279 my $quote = $char eq "'" ? '"' : "'";
280 $name = $quote . chr($ord) . $quote;
282 elsif ($char =~ /\p{XPosixGraph}/) {
284 $name = charnames::viacode($ord);
285 $name =~ s/LATIN CAPITAL LETTER //
286 or $name =~ s/LATIN SMALL LETTER (.*)/\L$1/
287 or $name =~ s/ SIGN\b//
288 or $name =~ s/EXCLAMATION MARK/'!'/
289 or $name =~ s/QUESTION MARK/'?'/
290 or $name =~ s/QUOTATION MARK/QUOTE/
291 or $name =~ s/ INDICATOR//;
292 $name =~ s/\bWITH\b/\L$&/;
293 $name =~ s/\bONE\b/1/;
294 $name =~ s/\b(TWO|HALF)\b/2/;
295 $name =~ s/\bTHREE\b/3/;
296 $name =~ s/\b QUARTER S? \b/4/x;
297 $name =~ s/VULGAR FRACTION (.) (.)/$1\/$2/;
298 $name =~ s/\bTILDE\b/'~'/i
299 or $name =~ s/\bCIRCUMFLEX\b/'^'/i
300 or $name =~ s/\bSTROKE\b/'\/'/i
301 or $name =~ s/ ABOVE\b//i;
304 use Unicode::UCD qw(prop_invmap);
305 my ($list_ref, $map_ref, $format) = prop_invmap("Name_Alias");
306 if ($format !~ /^s/) {
308 carp "Unexpected format '$format' for 'Name_Alias";
311 my $which = Unicode::UCD::search_invlist($list_ref, $ord);
312 if (! defined $which) {
314 carp "No name found for code pont $ord";
317 my $map = $map_ref->[$which];
322 # Just pick the first abbreviation if more than one
323 my @names = grep { $_ =~ /abbreviation/ } @$map;
329 my $index = $a2n[$ord];
330 $out[$index] = ($ord == $index)
331 ? sprintf "/* U+%02X %s */ %s,\n", $ord, $name, $bits[$ord]
332 : sprintf "/* 0x%02X U+%02X %s */ %s,\n", $index, $ord, $name, $bits[$ord];
334 print $out_fh join "", @out;
335 print $out_fh "\n" . get_conditional_compile_line_end();
338 read_only_bottom_close_and_rename($out_fh)