5 require 'regen/regen_lib.pl';
7 # This program outputs l1_charclass_tab.h, which defines the guts of the
8 # PL_charclass table. Each line is a bit map of properties that the Unicode
9 # code point at the corresponding position in the table array has. The first
10 # line corresponds to code point U+0000, NULL, the last line to U+00FF. For
11 # an application to see if the code point "i" has a particular property, it
13 # 'PL_charclass[i] & BIT'
14 # The bit names are of the form '_CC_property_suffix', where 'CC' stands for
15 # character class, and 'property' is the corresponding property, and 'suffix'
16 # is one of '_A' to mean the property is true only if the corresponding code
17 # point is ASCII, and '_L1' means that the range includes any Latin1
18 # character (ISO-8859-1 including the C0 and C1 controls). A property without
19 # these suffixes does not have different forms for both ranges.
21 # This program need be run only when adding new properties to it, or upon a
22 # new Unicode release, to make sure things haven't been changed by it.
48 BACKSLASH_FOO_LBRACE_IS_META
51 # Read in the case fold mappings.
53 my @hex_non_final_folds;
57 BEGIN { # Have to do this at compile time because using user-defined \p{property}
59 # Use the Unicode data file if we are on an ASCII platform (which its data
60 # is for), and it is in the modern format (starting in Unicode 3.1.0) and
61 # it is available. This avoids being affected by potential bugs
62 # introduced by other layers of Perl
63 my $file="lib/unicore/CaseFolding.txt";
66 && pack("C*", split /\./, Unicode::UCD::UnicodeVersion()) ge v3.1.0
67 && open my $fh, "<", $file)
72 my ($invlist_ref, $invmap_ref, undef, $default)
73 = Unicode::UCD::prop_invmap('Case_Folding');
74 for my $i (0 .. @$invlist_ref - 1 - 1) {
75 next if $invmap_ref->[$i] == $default;
77 for my $j ($invlist_ref->[$i] .. $invlist_ref->[$i+1] -1) {
80 # Single-code point maps go to a 'C' type
81 if (! ref $invmap_ref->[$i]) {
82 push @folds, sprintf("%04X; C; %04X\n",
84 $invmap_ref->[$i] + $adjust);
86 else { # Multi-code point maps go to 'F'. prop_invmap()
87 # guarantees that no adjustment is needed for these,
88 # as the range will contain just one element
89 push @folds, sprintf("%04X; F; %s\n",
91 join " ", map { sprintf "%04X", $_ }
92 @{$invmap_ref->[$i]});
101 # Lines look like (without the initial '#'
102 #0130; F; 0069 0307; # LATIN CAPITAL LETTER I WITH DOT ABOVE
103 # Get rid of comments, ignore blank or comment-only lines
104 my $line = $_ =~ s/ (?: \s* \# .* )? $ //rx;
105 next unless length $line;
106 my ($hex_from, $fold_type, @folded) = split /[\s;]+/, $line;
108 my $from = hex $hex_from;
110 # Perl only deals with C and F folds
111 next if $fold_type ne 'C' and $fold_type ne 'F';
113 # Get each code point in the range that participates in this line's fold.
114 # The hash has keys of each code point in the range, and values of what it
115 # folds to and what folds to it
116 for my $i (0 .. @folded - 1) {
117 my $hex_fold = $folded[$i];
118 my $fold = hex $hex_fold;
119 push @{$folded_closure{$fold}}, $from if $fold < 256;
120 push @{$folded_closure{$from}}, $fold if $from < 256;
124 && ! grep { $_ eq $hex_fold } @hex_non_final_folds)
126 push @hex_non_final_folds, $hex_fold;
128 # Also add the upper case, which in the latin1 range folds to
130 push @hex_non_final_folds, sprintf "%04X", ord uc chr $fold;
135 # Now having read all the lines, combine them into the full closure of each
136 # code point in the range by adding lists together that share a common
138 foreach my $folded (keys %folded_closure) {
139 foreach my $from (grep { $_ < 256 } @{$folded_closure{$folded}}) {
140 push @{$folded_closure{$from}}, @{$folded_closure{$folded}};
145 sub Is_Non_Latin1_Fold {
148 foreach my $folded (keys %folded_closure) {
149 push @return, sprintf("%X", $folded), if grep { $_ > 255 }
150 @{$folded_closure{$folded}};
152 return join("\n", @return) . "\n";
155 sub Is_Non_Final_Fold {
156 return join("\n", @hex_non_final_folds) . "\n";
159 my @bits; # Bit map for each code point
161 # For each character, calculate which properties it matches.
162 for my $ord (0..255) {
163 my $char = chr($ord);
164 utf8::upgrade($char); # Important to use Unicode rules!
166 # Look at all the properties we care about here.
167 for my $property (@properties) {
168 my $name = $property;
170 # Remove the suffix to get the actual property name.
171 # Currently the suffixes are '_L1', '_A', and none.
172 # If is a latin1 version, no further checking is needed.
173 if (! ($name =~ s/_L1$//)) {
175 # Here, isn't an _L1. If its _A, it's automatically false for
176 # non-ascii. The only current ones (besides ASCII) without a
177 # suffix are valid over the whole range.
178 next if $name =~ s/_A$// && $char !~ /\p{ASCII}/;
181 if ($name eq 'PUNCT') {;
183 # Sadly, this is inconsistent: \pP and \pS for the ascii range,
184 # just \pP outside it.
185 $re = qr/\p{Punct}|[^\P{Symbol}\P{ASCII}]/;
186 } elsif ($name eq 'CHARNAME_CONT') {;
187 $re = qr/\p{_Perl_Charname_Continue}/,
188 } elsif ($name eq 'SPACE') {;
189 $re = qr/\p{XPerlSpace}/;
190 } elsif ($name eq 'IDFIRST') {
191 $re = qr/[_\p{Alpha}]/;
192 } elsif ($name eq 'PSXSPC') {
193 $re = qr/[\v\p{Space}]/;
194 } elsif ($name eq 'WORDCHAR') {
195 $re = qr/\p{XPosixWord}/;
196 } elsif ($name eq 'ALPHANUMERIC') {
197 # Like \w, but no underscore
199 } elsif ($name eq 'QUOTEMETA') {
200 $re = qr/\p{_Perl_Quotemeta}/;
201 } elsif ($name eq 'NONLATIN1_FOLD') {
202 $re = qr/\p{Is_Non_Latin1_Fold}/;
203 } elsif ($name eq 'NON_FINAL_FOLD') {
204 $re = qr/\p{Is_Non_Final_Fold}/;
205 } elsif ($name eq 'IS_IN_SOME_FOLD') {
206 $re = qr/\p{_Perl_Any_Folds}/;
207 } elsif ($name eq 'BACKSLASH_FOO_LBRACE_IS_META') {
209 # This is true for FOO where FOO is the varying character in:
211 # and the sequence has non-literal meaning to Perl; so it is true
212 # for 'x' because \x{ is special, but not 'a' because \a{ isn't.
214 } else { # The remainder have the same name and values as Unicode
215 $re = eval "qr/\\p{$name}/";
217 carp $@ if ! defined $re;
219 #print "$ord, $name $property, $re\n";
220 if ($char =~ $re) { # Add this property if matches
221 $bits[$ord] .= '|' if $bits[$ord];
222 $bits[$ord] .= "(1U<<_CC_$property)";
225 #print __LINE__, " $ord $char $bits[$ord]\n";
228 my $out_fh = open_new('l1_char_class_tab.h', '>',
229 {style => '*', by => $0,
230 from => "property definitions"});
232 # Output the table using fairly short names for each char.
233 for my $ord (0..255) {
236 if ($char =~ /\p{PosixGraph}/) {
237 my $quote = $char eq "'" ? '"' : "'";
238 $name = $quote . chr($ord) . $quote;
240 elsif ($char =~ /\p{XPosixGraph}/) {
242 $name = charnames::viacode($ord);
243 $name =~ s/LATIN CAPITAL LETTER //
244 or $name =~ s/LATIN SMALL LETTER (.*)/\L$1/
245 or $name =~ s/ SIGN\b//
246 or $name =~ s/EXCLAMATION MARK/'!'/
247 or $name =~ s/QUESTION MARK/'?'/
248 or $name =~ s/QUOTATION MARK/QUOTE/
249 or $name =~ s/ INDICATOR//;
250 $name =~ s/\bWITH\b/\L$&/;
251 $name =~ s/\bONE\b/1/;
252 $name =~ s/\b(TWO|HALF)\b/2/;
253 $name =~ s/\bTHREE\b/3/;
254 $name =~ s/\b QUARTER S? \b/4/x;
255 $name =~ s/VULGAR FRACTION (.) (.)/$1\/$2/;
256 $name =~ s/\bTILDE\b/'~'/i
257 or $name =~ s/\bCIRCUMFLEX\b/'^'/i
258 or $name =~ s/\bSTROKE\b/'\/'/i
259 or $name =~ s/ ABOVE\b//i;
262 use Unicode::UCD qw(prop_invmap);
263 my ($list_ref, $map_ref, $format) = prop_invmap("Name_Alias");
264 if ($format !~ /^s/) {
266 carp "Unexpected format '$format' for 'Name_Alias";
269 my $which = Unicode::UCD::search_invlist($list_ref, $ord);
270 if (! defined $which) {
272 carp "No name found for code pont $ord";
275 my $map = $map_ref->[$which];
280 # Just pick the first abbreviation if more than one
281 my @names = grep { $_ =~ /abbreviation/ } @$map;
287 printf $out_fh "/* U+%02X %s */ %s,\n", $ord, $name, $bits[$ord];
290 read_only_bottom_close_and_rename($out_fh)