11 our (%PropertyAlias, %PA_reverse, %PropValueAlias, %PVA_reverse, %PVA_abbr_map);
13 sub croak { require Carp; Carp::croak(@_) }
16 ## "SWASH" == "SWATCH HASH". A "swatch" is a swatch of the Unicode landscape.
17 ## It's a data structure that encodes a set of Unicode characters.
21 my ($class, $type, $list, $minbits, $none) = @_;
24 print STDERR "SWASHNEW @_\n" if DEBUG;
27 ## Get the list of codepoints for the type.
28 ## Called from swash_init (see utf8.c) or SWASHNEW itself.
30 ## Callers of swash_init:
31 ## op.c:pmtrans -- for tr/// and y///
32 ## regexec.c:regclass_swash -- for /[]/, \p, and \P
33 ## utf8.c:is_utf8_common -- for common Unicode properties
34 ## utf8.c:to_utf8_case -- for lc, uc, ucfirst, etc. and //i
36 ## Given a $type, our goal is to fill $list with the set of codepoint
37 ## ranges. If $type is false, $list passed is used.
40 ## For binary properties, $minbits must be 1.
41 ## For character mappings (case and transliteration), $minbits must
42 ## be a number except 1.
44 ## $list (or that filled according to $type):
45 ## Refer to perlunicode.pod, "User-Defined Character Properties."
47 ## For binary properties, only characters with the property value
48 ## of True should be listed. The 3rd column, if any, will be ignored.
50 ## To make the parsing of $type clear, this code takes the a rather
51 ## unorthodox approach of last'ing out of the block once we have the
52 ## info we need. Were this to be a subroutine, the 'last' would just
55 my $file; ## file to load data from, and also part of the %Cache key.
63 print STDERR "type = $type\n" if DEBUG;
68 ## It could be a user-defined property.
71 my $caller1 = $type =~ s/(.+)::// ? $1 : caller(1);
73 if (defined $caller1 && $type =~ /^(?:\w+)$/) {
74 my $prop = "${caller1}::$type";
75 if (exists &{$prop}) {
85 ($wasIs = $type =~ s/^Is(?:\s+|[-_])?//i)
87 $type =~ s/^(?:(?:General(?:\s+|_)?)?Category|gc)\s*[:=]\s*//i
89 $type =~ s/^(?:Script|sc)\s*[:=]\s*//i
91 $type =~ s/^Block\s*[:=]\s*/In/i;
95 ## See if it's in some enumeration.
97 require "unicore/PVA.pl";
98 if ($type =~ /^([\w\s]+)[:=]\s*(.*)/) {
99 my ($enum, $val) = (lc $1, lc $2);
103 my $pa = $PropertyAlias{$enum} ? $enum : $PA_reverse{$enum};
104 my $f = $PropValueAlias{$pa}{$val} ? $val : $PVA_reverse{$pa}{lc $val};
107 $pa = "gc_sc" if $pa eq "gc" or $pa eq "sc";
108 $file = "unicore/lib/$pa/$PVA_abbr_map{$pa}{lc $f}.pl";
116 if ($PropValueAlias{gc}{$t} or $PropValueAlias{sc}{$t}) {
117 $file = "unicore/lib/gc_sc/$PVA_abbr_map{gc_sc}{$t}.pl";
123 ## See if it's in the direct mapping table.
125 require "unicore/Exact.pl";
126 if (my $base = $utf8::Exact{$type}) {
127 $file = "unicore/lib/gc_sc/$base.pl";
132 ## If not there exactly, try the canonical form. The canonical
133 ## form is lowercased, with any separators (\s+|[-_]) removed.
135 my $canonical = lc $type;
136 $canonical =~ s/(?<=[a-z\d])(?:\s+|[-_])(?=[a-z\d])//g;
137 print STDERR "canonical = $canonical\n" if DEBUG;
139 require "unicore/Canonical.pl";
140 if (my $base = ($utf8::Canonical{$canonical} || $utf8::Canonical{ lc $utf8::PropertyAlias{$canonical} })) {
141 $file = "unicore/lib/gc_sc/$base.pl";
146 ## See if it's a user-level "To".
149 my $caller0 = caller(0);
151 if (defined $caller0 && $type =~ /^To(?:\w+)$/) {
152 my $map = $caller0 . "::" . $type;
154 if (exists &{$map}) {
163 ## Last attempt -- see if it's a standard "To" name
164 ## (e.g. "ToLower") ToTitle is used by ucfirst().
165 ## The user-level way to access ToDigit() and ToFold()
166 ## is to use Unicode::UCD.
168 if ($type =~ /^To(Digit|Fold|Lower|Title|Upper)$/) {
169 $file = "unicore/To/$1.pl";
170 ## would like to test to see if $file actually exists....
175 ## If we reach this line, it's because we couldn't figure
176 ## out what to do with $type. Ouch.
183 print STDERR "found it (file='$file')\n" if DEBUG;
186 ## If we reach here, it was due to a 'last GETFILE' above
187 ## (exception: user-defined properties and mappings), so we
188 ## have a filename, so now we load it if we haven't already.
189 ## If we have, return the cached results. The cache key is the
190 ## class and file to load.
192 my $found = $Cache{$class, $file};
193 if ($found and ref($found) eq $class) {
194 print STDERR "Returning cached '$file' for \\p{$type}\n" if DEBUG;
198 $list = do $file; die $@ if $@;
201 $ListSorted = 1; ## we know that these lists are sorted
209 my @tmp = split(/^/m, $list);
212 $extras = join '', grep /^[^0-9a-fA-F]/, @tmp;
215 sort { $a->[0] <=> $b->[0] }
216 map { /^([0-9a-fA-F]+)/; [ hex($1), $_ ] }
217 grep { /^([0-9a-fA-F]+)/ and not $seen{$1}++ } @tmp; # XXX doesn't do ranges right
221 my $hextra = sprintf "%04x", $none + 1;
222 $list =~ s/\tXXXX$/\t$hextra/mg;
225 if ($minbits != 1 && $minbits < 32) { # not binary property
227 while ($list =~ /^([0-9a-fA-F]+)(?:[\t]([0-9a-fA-F]+)?)(?:[ \t]([0-9a-fA-F]+))?/mg) {
229 my $max = defined $2 ? hex $2 : $min;
230 my $val = defined $3 ? hex $3 : 0;
231 $val += $max - $min if defined $3;
232 $top = $val if $val > $top;
236 $top > 0xff ? 16 : 8;
237 $bits = $topbits if $bits < $topbits;
241 for my $x ($extras) {
243 while ($x =~ /^([^0-9a-fA-F\n])(.*)/mg) {
246 print STDERR "$1 => $2\n" if DEBUG;
247 if ($char =~ /[-+!&]/) {
248 my ($c,$t) = split(/::/, $name, 2); # bogus use of ::, really
251 $subobj = utf8->SWASHNEW($t, "", $minbits, 0);
253 elsif (exists &$name) {
254 $subobj = utf8->SWASHNEW($name, "", $minbits, 0);
256 elsif ($c =~ /^([0-9a-fA-F]+)/) {
257 $subobj = utf8->SWASHNEW("", $c, $minbits, 0);
259 return $subobj unless ref $subobj;
260 push @extras, $name => $subobj;
261 $bits = $subobj->{BITS} if $bits < $subobj->{BITS};
266 print STDERR "CLASS = $class, TYPE => $type, BITS => $bits, NONE => $none\nEXTRAS =>\n$extras\nLIST =>\n$list\n" if DEBUG;
278 $Cache{$class, $file} = $SWASH;
284 # Now SWASHGET is recasted into a C function S_swash_get (see utf8.c).