candidate for TR18 compliance
[p5sagit/p5-mst-13.2.git] / lib / utf8_heavy.pl
CommitLineData
a0ed51b3 1package utf8;
cf25bb62 2use strict;
3use warnings;
12ac2576 4require "utf8_pva.pl";
a0ed51b3 5
15732964 6sub DEBUG () { 0 }
a0ed51b3 7
8sub DESTROY {}
9
5beb625e 10my %Cache;
11
12ac2576 12our (%PropertyAlias, %PA_reverse, %PropValueAlias, %PVA_reverse, %PVA_abbr_map);
13
86916d66 14sub croak { require Carp; Carp::croak(@_) }
15
5beb625e 16##
3a2263fe 17## "SWASH" == "SWATCH HASH". A "swatch" is a swatch of the Unicode landscape.
18## It's a data structure that encodes a set of Unicode characters.
5beb625e 19##
20
a0ed51b3 21sub SWASHNEW {
22 my ($class, $type, $list, $minbits, $none) = @_;
23 local $^D = 0 if $^D;
15732964 24
25 print STDERR "SWASHNEW @_\n" if DEBUG;
26
cf25bb62 27 ##
28 ## Get the list of codepoints for the type.
29 ## Called from utf8.c
30 ##
31 ## Given a $type, our goal is to fill $list with the set of codepoint
5beb625e 32 ## ranges.
cf25bb62 33 ##
34 ## To make the parsing of $type clear, this code takes the a rather
2d92f8a0 35 ## unorthodox approach of last'ing out of the block once we have the
cf25bb62 36 ## info we need. Were this to be a subroutine, the 'last' would just
37 ## be a 'return'.
38 ##
5beb625e 39 my $file; ## file to load data from, and also part of the %Cache key.
40 my $ListSorted = 0;
41
cf25bb62 42 if ($type)
43 {
44 $type =~ s/^\s+//;
45 $type =~ s/\s+$//;
46
47 print "type = $type\n" if DEBUG;
48
cf25bb62 49 GETFILE:
50 {
12ac2576 51 ##
52 ## It could be a user-defined property.
53 ##
54
55 my $caller1 = $type =~ s/(.+)::// ? $1 : caller(1);
56
57 if (defined $caller1 && $type =~ /^(?:\w+)$/) {
58 my $prop = "${caller1}::$type";
59 if (exists &{$prop}) {
60 no strict 'refs';
61
62 $list = &{$prop};
63 last GETFILE;
64 }
65 }
66
2d92f8a0 67 my $wasIs;
68
69 ($wasIs = $type =~ s/^Is(?:\s+|[-_])?//i)
5beb625e 70 or
12ac2576 71 $type =~ s/^(?:(?:General(?:\s+|_)?)?Category|gc)\s*[:=]\s*//i
5beb625e 72 or
12ac2576 73 $type =~ s/^(?:Script|sc)\s*[:=]\s*//i
5beb625e 74 or
12ac2576 75 $type =~ s/^Block\s*[:=]\s*/In/i;
76
77
78 ##
79 ## See if it's in some enumeration.
80 ##
81 if ($type =~ /^([\w\s]+)[:=]\s*(.*)/) {
82 require "utf8_pva.pl";
83 my ($enum, $val) = (lc $1, lc $2);
84 $enum =~ tr/ _-//d;
85 $val =~ tr/ _-//d;
86
87 my $pa = $PropertyAlias{$enum} ? $enum : $PA_reverse{$enum};
88 my $f = $PropValueAlias{$pa}{$val} ? $val : $PVA_reverse{$pa}{lc $val};
89
90 if ($pa and $f) {
91 $pa = "gc_sc" if $pa eq "gc" or $pa eq "sc";
92 $file = "unicore/lib/$pa/$PVA_abbr_map{$pa}{lc $f}.pl";
93 last GETFILE;
94 }
95 }
96 else {
97 my $t = lc $type;
98 $t =~ tr/ _-//d;
99
100 if ($PropValueAlias{gc}{$t} or $PropValueAlias{sc}{$t}) {
101 $file = "unicore/lib/gc_sc/$PVA_abbr_map{gc_sc}{$t}.pl";
102 last GETFILE;
103 }
104 }
cf25bb62 105
5beb625e 106 ##
107 ## See if it's in the direct mapping table.
108 ##
109 require "unicore/Exact.pl";
110 if (my $base = $utf8::Exact{$type}) {
12ac2576 111 $file = "unicore/lib/gc_sc/$base.pl";
5beb625e 112 last GETFILE;
cf25bb62 113 }
114
115 ##
5beb625e 116 ## If not there exactly, try the canonical form. The canonical
117 ## form is lowercased, with any separators (\s+|[-_]) removed.
cf25bb62 118 ##
5beb625e 119 my $canonical = lc $type;
120 $canonical =~ s/(?<=[a-z\d])(?:\s+|[-_])(?=[a-z\d])//g;
121 print "canonical = $canonical\n" if DEBUG;
cf25bb62 122
5beb625e 123 require "unicore/Canonical.pl";
12ac2576 124 if (my $base = ($utf8::Canonical{$canonical} || $utf8::Canonical{ lc $utf8::PropertyAlias{$canonical} })) {
125 $file = "unicore/lib/gc_sc/$base.pl";
5beb625e 126 last GETFILE;
cf25bb62 127 }
128
491fd90a 129 ##
3a2263fe 130 ## See if it's a user-level "To".
131 ##
132
133 my $caller0 = caller(0);
134
135 if (defined $caller0 && $type =~ /^To(?:\w+)$/) {
136 my $map = $caller0 . "::" . $type;
09e0265a 137
3a2263fe 138 if (exists &{$map}) {
139 no strict 'refs';
140
141 $list = &{$map};
142 last GETFILE;
143 }
144 }
145
cf25bb62 146 ##
3a2263fe 147 ## Last attempt -- see if it's a standard "To" name
148 ## (e.g. "ToLower") ToTitle is used by ucfirst().
149 ## The user-level way to access ToDigit() and ToFold()
150 ## is to use Unicode::UCD.
cf25bb62 151 ##
3a2263fe 152 if ($type =~ /^To(Digit|Fold|Lower|Title|Upper)$/)
cf25bb62 153 {
154 $file = "unicore/To/$1.pl";
155 ## would like to test to see if $file actually exists....
156 last GETFILE;
157 }
158
159 ##
160 ## If we reach this line, it's because we couldn't figure
161 ## out what to do with $type. Ouch.
162 ##
bc45ce41 163
164 return $type;
cf25bb62 165 }
166
491fd90a 167 if (defined $file) {
168 print "found it (file='$file')\n" if DEBUG;
169
170 ##
171 ## If we reach here, it was due to a 'last GETFILE' above
3a2263fe 172 ## (exception: user-defined properties and mappings), so we
491fd90a 173 ## have a filename, so now we load it if we haven't already.
174 ## If we have, return the cached results. The cache key is the
175 ## file to load.
176 ##
177 if ($Cache{$file} and ref($Cache{$file}) eq $class)
178 {
179 print "Returning cached '$file' for \\p{$type}\n" if DEBUG;
180 return $Cache{$class, $file};
181 }
182
183 $list = do $file;
184 }
5beb625e 185
5beb625e 186 $ListSorted = 1; ## we know that these lists are sorted
886ba366 187 }
a0ed51b3 188
15732964 189 my $extras;
485aafd9 190 my $bits = 0;
cf25bb62 191
5beb625e 192 my $ORIG = $list;
a0ed51b3 193 if ($list) {
194 my @tmp = split(/^/m, $list);
195 my %seen;
db376a24 196 no warnings;
a0ed51b3 197 $extras = join '', grep /^[^0-9a-fA-F]/, @tmp;
198 $list = join '',
df68b396 199 map { $_->[1] }
200 sort { $a->[0] <=> $b->[0] }
201 map { /^([0-9a-fA-F]+)/; [ hex($1), $_ ] }
202 grep { /^([0-9a-fA-F]+)/ and not $seen{$1}++ } @tmp; # XXX doesn't do ranges right
a0ed51b3 203 }
204
205 if ($none) {
206 my $hextra = sprintf "%04x", $none + 1;
207 $list =~ s/\tXXXX$/\t$hextra/mg;
208 }
209
210 if ($minbits < 32) {
211 my $top = 0;
3a2263fe 212 while ($list =~ /^([0-9a-fA-F]+)(?:[\t]([0-9a-fA-F]+)?)(?:[ \t]([0-9a-fA-F]+))?/mg) {
a0ed51b3 213 my $min = hex $1;
3a2263fe 214 my $max = defined $2 ? hex $2 : $min;
215 my $val = defined $3 ? hex $3 : 0;
a0ed51b3 216 $val += $max - $min if defined $3;
217 $top = $val if $val > $top;
218 }
219 $bits =
220 $top > 0xffff ? 32 :
221 $top > 0xff ? 16 :
222 $top > 1 ? 8 : 1
223 }
224 $bits = $minbits if $bits < $minbits;
225
226 my @extras;
227 for my $x ($extras) {
228 pos $x = 0;
1fef36c7 229 while ($x =~ /^([^0-9a-fA-F\n])(.*)/mg) {
a0ed51b3 230 my $char = $1;
231 my $name = $2;
6d6ae962 232 print STDERR "$1 => $2\n" if DEBUG;
09e0265a 233 if ($char =~ /[-+!&]/) {
a0ed51b3 234 my ($c,$t) = split(/::/, $name, 2); # bogus use of ::, really
11ef8fdd 235 my $subobj;
236 if ($c eq 'utf8') {
09e0265a 237 $subobj = utf8->SWASHNEW($t, "", 0, 0, 0);
238 }
239 elsif (exists &$name) {
240 $subobj = utf8->SWASHNEW($name, "", 0, 0, 0);
11ef8fdd 241 }
242 elsif ($c =~ /^([0-9a-fA-F]+)/) {
243 $subobj = utf8->SWASHNEW("", $c, 0, 0, 0);
244 }
bc45ce41 245 return $subobj unless ref $subobj;
a0ed51b3 246 push @extras, $name => $subobj;
247 $bits = $subobj->{BITS} if $bits < $subobj->{BITS};
248 }
249 }
250 }
251
15732964 252 print STDERR "CLASS = $class, TYPE => $type, BITS => $bits, NONE => $none\nEXTRAS =>\n$extras\nLIST =>\n$list\n" if DEBUG;
a0ed51b3 253
5beb625e 254 my $SWASH = bless {
a0ed51b3 255 TYPE => $type,
256 BITS => $bits,
257 EXTRAS => $extras,
258 LIST => $list,
259 NONE => $none,
260 @extras,
261 } => $class;
5beb625e 262
263 if ($file) {
264 $Cache{$class, $file} = $SWASH;
265 }
266
267 return $SWASH;
a0ed51b3 268}
269
270# NOTE: utf8.c:swash_init() assumes entries are never modified once generated.
271
272sub SWASHGET {
035d37be 273 # See utf8.c:Perl_swash_fetch for problems with this interface.
a0ed51b3 274 my ($self, $start, $len) = @_;
275 local $^D = 0 if $^D;
a0ed51b3 276 my $type = $self->{TYPE};
277 my $bits = $self->{BITS};
278 my $none = $self->{NONE};
15732964 279 print STDERR "SWASHGET @_ [$type/$bits/$none]\n" if DEBUG;
a0ed51b3 280 my $end = $start + $len;
281 my $swatch = "";
282 my $key;
283 vec($swatch, $len - 1, $bits) = 0; # Extend to correct length.
284 if ($none) {
285 for $key (0 .. $len - 1) { vec($swatch, $key, $bits) = $none }
286 }
287
288 for ($self->{LIST}) {
289 pos $_ = 0;
290 if ($bits > 1) {
291 LINE:
3a2263fe 292 while (/^([0-9a-fA-F]+)(?:[ \t]([0-9a-fA-F]+)?)?(?:[ \t]([0-9a-fA-F]+))?/mg) {
293 chomp;
294 my ($a, $b, $c) = ($1, $2, $3);
295 croak "$type: illegal mapping '$_'"
296 if $type =~ /^To/ &&
297 !(defined $a && defined $c);
298 my $min = hex $a;
299 my $max = defined $b ? hex $b : $min;
300 my $val = defined $c ? hex $c : 0;
a0ed51b3 301 next if $max < $start;
6d6ae962 302 print "$min $max $val\n" if DEBUG;
a0ed51b3 303 if ($none) {
304 if ($min < $start) {
710250cb 305 $val += $start - $min if $val < $none;
a0ed51b3 306 $min = $start;
307 }
308 for ($key = $min; $key <= $max; $key++) {
309 last LINE if $key >= $end;
6d6ae962 310 print STDERR "$key => $val\n" if DEBUG;
a0ed51b3 311 vec($swatch, $key - $start, $bits) = $val;
312 ++$val if $val < $none;
313 }
314 }
315 else {
316 if ($min < $start) {
317 $val += $start - $min;
318 $min = $start;
319 }
320 for ($key = $min; $key <= $max; $key++, $val++) {
321 last LINE if $key >= $end;
6d6ae962 322 print STDERR "$key => $val\n" if DEBUG;
a0ed51b3 323 vec($swatch, $key - $start, $bits) = $val;
324 }
325 }
326 }
327 }
328 else {
329 LINE:
99a6b1f0 330 while (/^([0-9a-fA-F]+)(?:[ \t]+([0-9a-fA-F]+))?/mg) {
3a2263fe 331 chomp;
a0ed51b3 332 my $min = hex $1;
3a2263fe 333 my $max = defined $2 ? hex $2 : $min;
a0ed51b3 334 next if $max < $start;
335 if ($min < $start) {
336 $min = $start;
337 }
338 for ($key = $min; $key <= $max; $key++) {
339 last LINE if $key >= $end;
6d6ae962 340 print STDERR "$key => 1\n" if DEBUG;
a0ed51b3 341 vec($swatch, $key - $start, 1) = 1;
342 }
343 }
344 }
345 }
346 for my $x ($self->{EXTRAS}) {
347 pos $x = 0;
09e0265a 348 while ($x =~ /^([-+!&])(.*)/mg) {
a0ed51b3 349 my $char = $1;
350 my $name = $2;
15732964 351 print STDERR "INDIRECT $1 $2\n" if DEBUG;
1fef36c7 352 my $otherbits = $self->{$name}->{BITS};
353 croak("SWASHGET size mismatch") if $bits < $otherbits;
354 my $other = $self->{$name}->SWASHGET($start, $len);
355 if ($char eq '+') {
356 if ($bits == 1 and $otherbits == 1) {
357 $swatch |= $other;
a0ed51b3 358 }
1fef36c7 359 else {
360 for ($key = 0; $key < $len; $key++) {
361 vec($swatch, $key, $bits) = vec($other, $key, $otherbits);
a0ed51b3 362 }
1fef36c7 363 }
364 }
365 elsif ($char eq '!') {
366 if ($bits == 1 and $otherbits == 1) {
367 $swatch |= ~$other;
368 }
369 else {
370 for ($key = 0; $key < $len; $key++) {
371 if (!vec($other, $key, $otherbits)) {
372 vec($swatch, $key, $bits) = 1;
a0ed51b3 373 }
374 }
375 }
1fef36c7 376 }
377 elsif ($char eq '-') {
378 if ($bits == 1 and $otherbits == 1) {
379 $swatch &= ~$other;
380 }
381 else {
382 for ($key = 0; $key < $len; $key++) {
383 if (vec($other, $key, $otherbits)) {
384 vec($swatch, $key, $bits) = 0;
a0ed51b3 385 }
386 }
387 }
388 }
09e0265a 389 elsif ($char eq '&') {
390 if ($bits == 1 and $otherbits == 1) {
391 $swatch &= $other;
392 }
393 else {
394 for ($key = 0; $key < $len; $key++) {
395 if (!vec($other, $key, $otherbits)) {
396 vec($swatch, $key, $bits) = 0;
397 }
398 }
399 }
400 }
a0ed51b3 401 }
402 }
15732964 403 if (DEBUG) {
a0ed51b3 404 print STDERR "CELLS ";
405 for ($key = 0; $key < $len; $key++) {
406 print STDERR vec($swatch, $key, $bits), " ";
407 }
408 print STDERR "\n";
409 }
410 $swatch;
411}
412
4131;