This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
utf8_heavy.pl: Correct debugging statement
[perl5.git] / lib / utf8_heavy.pl
CommitLineData
a0ed51b3 1package utf8;
cf25bb62
JH
2use strict;
3use warnings;
a0ed51b3 4
15732964 5sub DEBUG () { 0 }
a0ed51b3
LW
6
7sub DESTROY {}
8
5beb625e
JH
9my %Cache;
10
86916d66
JH
11sub croak { require Carp; Carp::croak(@_) }
12
c4ab2516 13sub _loose_name ($) {
45376db6
KW
14 # Given a lowercase property or property-value name, return its
15 # standardized version that is expected for look-up in the 'loose' hashes
16 # in Heavy.pl (hence, this depends on what mktables does). This squeezes
17 # out blanks, underscores and dashes. The complication stems from the
18 # grandfathered-in 'L_', which retains a single trailing underscore.
19
20 my $loose = $_[0] =~ s/[-\s_]//rg;
21
22 return $loose if $loose !~ / ^ (?: is )? l $/x;
23 return 'l_' if $_[0] =~ / l .* _ /x; # If original had a trailing '_'
24 return $loose;
25}
26
5beb625e 27##
3a2263fe
RGS
28## "SWASH" == "SWATCH HASH". A "swatch" is a swatch of the Unicode landscape.
29## It's a data structure that encodes a set of Unicode characters.
5beb625e
JH
30##
31
99870f4d
KW
32{
33 # If a floating point number is within this distance from the value of a
34 # fraction, it is considered to be that fraction, even if many more digits
35 # are specified that don't exactly match.
36 my $min_floating_slop;
37
3dd10fe8
KW
38 # To guard against this program calling something that in turn ends up
39 # calling this program with the same inputs, and hence infinitely
40 # recursing, we keep a stack of the properties that are currently in
41 # progress, pushed upon entry, popped upon return.
42 my @recursed;
43
99870f4d
KW
44 sub SWASHNEW {
45 my ($class, $type, $list, $minbits, $none) = @_;
0fada5e1 46 my $user_defined = 0;
99870f4d
KW
47 local $^D = 0 if $^D;
48
3dd10fe8 49 $class = "" unless defined $class;
5bcf08c4
KW
50 print STDERR __LINE__, ": class=$class, type=$type, list=",
51 (defined $list) ? $list : ':undef:',
52 ", minbits=$minbits, none=$none\n" if DEBUG;
99870f4d
KW
53
54 ##
55 ## Get the list of codepoints for the type.
56 ## Called from swash_init (see utf8.c) or SWASHNEW itself.
57 ##
58 ## Callers of swash_init:
59 ## op.c:pmtrans -- for tr/// and y///
60 ## regexec.c:regclass_swash -- for /[]/, \p, and \P
61 ## utf8.c:is_utf8_common -- for common Unicode properties
62 ## utf8.c:to_utf8_case -- for lc, uc, ucfirst, etc. and //i
63 ##
64 ## Given a $type, our goal is to fill $list with the set of codepoint
65 ## ranges. If $type is false, $list passed is used.
66 ##
67 ## $minbits:
68 ## For binary properties, $minbits must be 1.
69 ## For character mappings (case and transliteration), $minbits must
70 ## be a number except 1.
71 ##
72 ## $list (or that filled according to $type):
73 ## Refer to perlunicode.pod, "User-Defined Character Properties."
74 ##
75 ## For binary properties, only characters with the property value
76 ## of True should be listed. The 3rd column, if any, will be ignored
77 ##
78 ## $none is undocumented, so I'm (khw) trying to do some documentation
79 ## of it now. It appears to be if there is a mapping in an input file
80 ## that maps to 'XXXX', then that is replaced by $none+1, expressed in
83fe8199 81 ## hexadecimal. It is used somehow in tr///.
99870f4d
KW
82 ##
83 ## To make the parsing of $type clear, this code takes the a rather
84 ## unorthodox approach of last'ing out of the block once we have the
85 ## info we need. Were this to be a subroutine, the 'last' would just
86 ## be a 'return'.
87 ##
fbe6f3cf
KW
88 # If a problem is found $type is returned;
89 # Upon success, a new (or cached) blessed object is returned with
90 # keys TYPE, BITS, EXTRAS, LIST, and NONE with values having the
91 # same meanings as the input parameters.
a6517ebf
KW
92 # SPECIALS contains a reference to any special-treatment hash in the
93 # INVERT_IT is non-zero if the result should be inverted before use
0fada5e1
KW
94 # USER_DEFINED is non-zero if the result came from a user-defined
95 # property.
99870f4d
KW
96 my $file; ## file to load data from, and also part of the %Cache key.
97 my $ListSorted = 0;
98
99 # Change this to get a different set of Unicode tables
100 my $unicore_dir = 'unicore';
a6517ebf 101 my $invert_it = 0;
99870f4d
KW
102
103 if ($type)
cf25bb62 104 {
3dd10fe8
KW
105
106 # Verify that this isn't a recursive call for this property.
107 # Can't use croak, as it may try to recurse here itself.
108 my $class_type = $class . "::$type";
109 if (grep { $_ eq $class_type } @recursed) {
110 CORE::die "panic: Infinite recursion in SWASHNEW for '$type'\n";
111 }
112 push @recursed, $class_type;
113
99870f4d
KW
114 $type =~ s/^\s+//;
115 $type =~ s/\s+$//;
116
4ff124d1
KW
117 # regcomp.c surrounds the property name with '__" and '_i' if this
118 # is to be caseless matching.
119 my $caseless = $type =~ s/^__(.*)_i$/$1/;
120
121 print STDERR __LINE__, ": type=$type, caseless=$caseless\n" if DEBUG;
99870f4d
KW
122
123 GETFILE:
124 {
125 ##
a44f4b74
KW
126 ## It could be a user-defined property. Look in current
127 ## package if no package given
99870f4d
KW
128 ##
129
130 my $caller1 = $type =~ s/(.+)::// ? $1 : caller(1);
131
d658a8a8 132 if (defined $caller1 && $type =~ /^I[ns]\w+$/) {
99870f4d
KW
133 my $prop = "${caller1}::$type";
134 if (exists &{$prop}) {
0e9be77f
DM
135 # stolen from Scalar::Util::PP::tainted()
136 my $tainted;
137 {
138 local($@, $SIG{__DIE__}, $SIG{__WARN__});
139 local $^W = 0;
140 no warnings;
141 eval { kill 0 * $prop };
142 $tainted = 1 if $@ =~ /^Insecure/;
143 }
144 die "Insecure user-defined property \\p{$prop}\n"
145 if $tainted;
99870f4d 146 no strict 'refs';
4ff124d1 147 $list = &{$prop}($caseless);
0fada5e1 148 $user_defined = 1;
99870f4d
KW
149 last GETFILE;
150 }
151 }
152
a44f4b74
KW
153 # During Perl's compilation, this routine may be called before
154 # the tables are constructed. If so, we have a chicken/egg
155 # problem. If we die, the tables never get constructed, so
156 # keep going, but return an empty table so only what the code
157 # has compiled in internally (currently ASCII/Latin1 range
158 # matching) will work.
1363cbd0
FC
159 BEGIN {
160 # Poor man's constant, to avoid a run-time check.
161 $utf8::{miniperl}
162 = \! defined &DynaLoader::boot_DynaLoader;
163 }
164 if (miniperl) {
a44f4b74
KW
165 eval "require '$unicore_dir/Heavy.pl'";
166 last GETFILE if $@;
167 }
168 else {
169 require "$unicore_dir/Heavy.pl";
170 }
1363cbd0 171 BEGIN { delete $utf8::{miniperl} }
99870f4d 172
4ff124d1 173 # All property names are matched caselessly
49801ec2 174 my $property_and_table = CORE::lc $type;
99870f4d
KW
175 print STDERR __LINE__, ": $property_and_table\n" if DEBUG;
176
177 # See if is of the compound form 'property=value', where the
178 # value indicates the table we should use.
179 my ($property, $table, @remainder) =
180 split /\s*[:=]\s*/, $property_and_table, -1;
3dd10fe8
KW
181 if (@remainder) {
182 pop @recursed if @recursed;
183 return $type;
184 }
99870f4d
KW
185
186 my $prefix;
187 if (! defined $table) {
188
189 # Here, is the single form. The property becomes empty, and
190 # the whole value is the table.
191 $table = $property;
192 $prefix = $property = "";
193 } else {
194 print STDERR __LINE__, ": $property\n" if DEBUG;
195
196 # Here it is the compound property=table form. The property
807807b7
KW
197 # name is always loosely matched, and always can have an
198 # optional 'is' prefix (which isn't true in the single
199 # form).
c4ab2516 200 $property = _loose_name($property) =~ s/^is//r;
99870f4d
KW
201
202 # And convert to canonical form. Quit if not valid.
203 $property = $utf8::loose_property_name_of{$property};
3dd10fe8
KW
204 if (! defined $property) {
205 pop @recursed if @recursed;
206 return $type;
207 }
99870f4d
KW
208
209 $prefix = "$property=";
210
211 # If the rhs looks like it is a number...
212 print STDERR __LINE__, ": table=$table\n" if DEBUG;
213 if ($table =~ qr{ ^ [ \s 0-9 _ + / . -]+ $ }x) {
214 print STDERR __LINE__, ": table=$table\n" if DEBUG;
215
216 # Don't allow leading nor trailing slashes
3dd10fe8
KW
217 if ($table =~ / ^ \/ | \/ $ /x) {
218 pop @recursed if @recursed;
219 return $type;
220 }
99870f4d
KW
221
222 # Split on slash, in case it is a rational, like \p{1/5}
223 my @parts = split qr{ \s* / \s* }x, $table, -1;
224 print __LINE__, ": $type\n" if @parts > 2 && DEBUG;
225
226 # Can have maximum of one slash
3dd10fe8
KW
227 if (@parts > 2) {
228 pop @recursed if @recursed;
229 return $type;
230 }
99870f4d
KW
231
232 foreach my $part (@parts) {
233 print __LINE__, ": part=$part\n" if DEBUG;
234
235 $part =~ s/^\+\s*//; # Remove leading plus
236 $part =~ s/^-\s*/-/; # Remove blanks after unary
237 # minus
238
239 # Remove underscores between digits.
240 $part =~ s/( ?<= [0-9] ) _ (?= [0-9] ) //xg;
241
242 # No leading zeros (but don't make a single '0'
243 # into a null string)
244 $part =~ s/ ^ ( -? ) 0+ /$1/x;
245 $part .= '0' if $part eq '-' || $part eq "";
246
247 # No trailing zeros after a decimal point
248 $part =~ s/ ( \. .*? ) 0+ $ /$1/x;
249
250 # Begin with a 0 if a leading decimal point
251 $part =~ s/ ^ ( -? ) \. /${1}0./x;
252
253 # Ensure not a trailing decimal point: turn into an
254 # integer
255 $part =~ s/ \. $ //x;
256
257 print STDERR __LINE__, ": part=$part\n" if DEBUG;
258 #return $type if $part eq "";
259
260 # Result better look like a number. (This test is
261 # needed because, for example could have a plus in
262 # the middle.)
3dd10fe8
KW
263 if ($part !~ / ^ -? [0-9]+ ( \. [0-9]+)? $ /x) {
264 pop @recursed if @recursed;
265 return $type;
266 }
99870f4d
KW
267 }
268
269 # If a rational...
270 if (@parts == 2) {
271
272 # If denominator is negative, get rid of it, and ...
273 if ($parts[1] =~ s/^-//) {
274
275 # If numerator is also negative, convert the
276 # whole thing to positive, or move the minus to
277 # the numerator
278 if ($parts[0] !~ s/^-//) {
279 $parts[0] = '-' . $parts[0];
280 }
281 }
282 $table = join '/', @parts;
283 }
284 elsif ($property ne 'nv' || $parts[0] !~ /\./) {
285
286 # Here is not numeric value, or doesn't have a
287 # decimal point. No further manipulation is
288 # necessary. (Note the hard-coded property name.
289 # This could fail if other properties eventually
290 # had fractions as well; perhaps the cjk ones
291 # could evolve to do that. This hard-coding could
292 # be fixed by mktables generating a list of
293 # properties that could have fractions.)
294 $table = $parts[0];
295 } else {
296
297 # Here is a floating point numeric_value. Try to
298 # convert to rational. First see if is in the list
299 # of known ones.
300 if (exists $utf8::nv_floating_to_rational{$parts[0]}) {
301 $table = $utf8::nv_floating_to_rational{$parts[0]};
302 } else {
303
304 # Here not in the list. See if is close
305 # enough to something in the list. First
306 # determine what 'close enough' means. It has
307 # to be as tight as what mktables says is the
308 # maximum slop, and as tight as how many
309 # digits we were passed. That is, if the user
310 # said .667, .6667, .66667, etc. we match as
311 # many digits as they passed until get to
312 # where it doesn't matter any more due to the
313 # machine's precision. If they said .6666668,
314 # we fail.
315 (my $fraction = $parts[0]) =~ s/^.*\.//;
316 my $epsilon = 10 ** - (length($fraction));
317 if ($epsilon > $utf8::max_floating_slop) {
318 $epsilon = $utf8::max_floating_slop;
319 }
320
321 # But it can't be tighter than the minimum
322 # precision for this machine. If haven't
323 # already calculated that minimum, do so now.
324 if (! defined $min_floating_slop) {
325
326 # Keep going down an order of magnitude
327 # until find that adding this quantity to
328 # 1 remains 1; but put an upper limit on
329 # this so in case this algorithm doesn't
330 # work properly on some platform, that we
331 # won't loop forever.
332 my $count = 0;
333 $min_floating_slop = 1;
334 while (1+ $min_floating_slop != 1
335 && $count++ < 50)
336 {
337 my $next = $min_floating_slop / 10;
338 last if $next == 0; # If underflows,
339 # use previous one
340 $min_floating_slop = $next;
341 print STDERR __LINE__, ": min_float_slop=$min_floating_slop\n" if DEBUG;
342 }
343
344 # Back off a couple orders of magnitude,
345 # just to be safe.
346 $min_floating_slop *= 100;
347 }
348
349 if ($epsilon < $min_floating_slop) {
350 $epsilon = $min_floating_slop;
351 }
352 print STDERR __LINE__, ": fraction=.$fraction; epsilon=$epsilon\n" if DEBUG;
353
354 undef $table;
355
356 # And for each possible rational in the table,
357 # see if it is within epsilon of the input.
358 foreach my $official
359 (keys %utf8::nv_floating_to_rational)
360 {
361 print STDERR __LINE__, ": epsilon=$epsilon, official=$official, diff=", abs($parts[0] - $official), "\n" if DEBUG;
362 if (abs($parts[0] - $official) < $epsilon) {
363 $table =
364 $utf8::nv_floating_to_rational{$official};
365 last;
366 }
367 }
368
369 # Quit if didn't find one.
3dd10fe8
KW
370 if (! defined $table) {
371 pop @recursed if @recursed;
372 return $type;
373 }
99870f4d
KW
374 }
375 }
376 print STDERR __LINE__, ": $property=$table\n" if DEBUG;
377 }
378 }
379
380 # Combine lhs (if any) and rhs to get something that matches
381 # the syntax of the lookups.
382 $property_and_table = "$prefix$table";
383 print STDERR __LINE__, ": $property_and_table\n" if DEBUG;
384
385 # First try stricter matching.
386 $file = $utf8::stricter_to_file_of{$property_and_table};
387
388 # If didn't find it, try again with looser matching by editing
389 # out the applicable characters on the rhs and looking up
390 # again.
391 if (! defined $file) {
c4ab2516 392 $table = _loose_name($table);
99870f4d
KW
393 $property_and_table = "$prefix$table";
394 print STDERR __LINE__, ": $property_and_table\n" if DEBUG;
395 $file = $utf8::loose_to_file_of{$property_and_table};
396 }
397
398 # Add the constant and go fetch it in.
399 if (defined $file) {
a6517ebf
KW
400
401 # A beginning ! means to invert
402 $invert_it = $file =~ s/^!//;
403
99870f4d
KW
404 if ($utf8::why_deprecated{$file}) {
405 warnings::warnif('deprecated', "Use of '$type' in \\p{} or \\P{} is deprecated because: $utf8::why_deprecated{$file};");
406 }
4ff124d1
KW
407
408 if ($caseless
409 && exists $utf8::caseless_equivalent{$property_and_table})
410 {
411 $file = $utf8::caseless_equivalent{$property_and_table};
412 }
99870f4d
KW
413 $file= "$unicore_dir/lib/$file.pl";
414 last GETFILE;
415 }
416 print STDERR __LINE__, ": didn't find $property_and_table\n" if DEBUG;
417
418 ##
99870f4d
KW
419 ## Last attempt -- see if it's a standard "To" name
420 ## (e.g. "ToLower") ToTitle is used by ucfirst().
421 ## The user-level way to access ToDigit() and ToFold()
422 ## is to use Unicode::UCD.
423 ##
424 if ($type =~ /^To(Digit|Fold|Lower|Title|Upper)$/) {
08432a98
KW
425
426 # Fail if wanting a binary property, as these aren't.
427 if ($minbits == 1) {
428 pop @recursed if @recursed;
429 return $type;
430 }
99870f4d
KW
431 $file = "$unicore_dir/To/$1.pl";
432 ## would like to test to see if $file actually exists....
433 last GETFILE;
434 }
435
436 ##
437 ## If we reach this line, it's because we couldn't figure
438 ## out what to do with $type. Ouch.
439 ##
440
3dd10fe8 441 pop @recursed if @recursed;
99870f4d 442 return $type;
cf25bb62
JH
443 }
444
99870f4d
KW
445 if (defined $file) {
446 print STDERR __LINE__, ": found it (file='$file')\n" if DEBUG;
447
448 ##
449 ## If we reach here, it was due to a 'last GETFILE' above
450 ## (exception: user-defined properties and mappings), so we
451 ## have a filename, so now we load it if we haven't already.
452 ## If we have, return the cached results. The cache key is the
453 ## class and file to load.
454 ##
455 my $found = $Cache{$class, $file};
456 if ($found and ref($found) eq $class) {
a6517ebf 457 print STDERR __LINE__, ": Returning cached '$file' for \\p{$type}; invert_it=$invert_it\n" if DEBUG;
3dd10fe8 458 pop @recursed if @recursed;
a6517ebf 459 $found->{'INVERT_IT'} = $invert_it;
99870f4d
KW
460 return $found;
461 }
462
463 local $@;
464 local $!;
465 $list = do $file; die $@ if $@;
0a0ffbce 466 }
cf25bb62 467
99870f4d 468 $ListSorted = 1; ## we know that these lists are sorted
cf25bb62
JH
469 }
470
99870f4d
KW
471 my $extras;
472 my $bits = $minbits;
473
99870f4d 474 if ($list) {
0e9be77f 475 my $taint = substr($list,0,0); # maintain taint
99870f4d
KW
476 my @tmp = split(/^/m, $list);
477 my %seen;
478 no warnings;
0e9be77f
DM
479 $extras = join '', $taint, grep /^[^0-9a-fA-F]/, @tmp;
480 $list = join '', $taint,
99870f4d
KW
481 map { $_->[1] }
482 sort { $a->[0] <=> $b->[0] }
483 map { /^([0-9a-fA-F]+)/; [ CORE::hex($1), $_ ] }
484 grep { /^([0-9a-fA-F]+)/ and not $seen{$1}++ } @tmp; # XXX doesn't do ranges right
485 }
a0ed51b3 486
99870f4d
KW
487 if ($none) {
488 my $hextra = sprintf "%04x", $none + 1;
489 $list =~ s/\tXXXX$/\t$hextra/mg;
490 }
a0ed51b3 491
99870f4d
KW
492 if ($minbits != 1 && $minbits < 32) { # not binary property
493 my $top = 0;
494 while ($list =~ /^([0-9a-fA-F]+)(?:[\t]([0-9a-fA-F]+)?)(?:[ \t]([0-9a-fA-F]+))?/mg) {
495 my $min = CORE::hex $1;
496 my $max = defined $2 ? CORE::hex $2 : $min;
497 my $val = defined $3 ? CORE::hex $3 : 0;
498 $val += $max - $min if defined $3;
499 $top = $val if $val > $top;
500 }
501 my $topbits =
502 $top > 0xffff ? 32 :
503 $top > 0xff ? 16 : 8;
504 $bits = $topbits if $bits < $topbits;
505 }
a0ed51b3 506
99870f4d
KW
507 my @extras;
508 if ($extras) {
509 for my $x ($extras) {
0e9be77f 510 my $taint = substr($x,0,0); # maintain taint
99870f4d
KW
511 pos $x = 0;
512 while ($x =~ /^([^0-9a-fA-F\n])(.*)/mg) {
0e9be77f
DM
513 my $char = "$1$taint";
514 my $name = "$2$taint";
515 print STDERR __LINE__, ": char [$char] => name [$name]\n"
516 if DEBUG;
99870f4d
KW
517 if ($char =~ /[-+!&]/) {
518 my ($c,$t) = split(/::/, $name, 2); # bogus use of ::, really
519 my $subobj;
520 if ($c eq 'utf8') {
521 $subobj = utf8->SWASHNEW($t, "", $minbits, 0);
522 }
523 elsif (exists &$name) {
524 $subobj = utf8->SWASHNEW($name, "", $minbits, 0);
525 }
526 elsif ($c =~ /^([0-9a-fA-F]+)/) {
527 $subobj = utf8->SWASHNEW("", $c, $minbits, 0);
528 }
3dd10fe8
KW
529 if (! ref $subobj) {
530 pop @recursed if @recursed && $type;
531 return $subobj;
532 }
99870f4d
KW
533 push @extras, $name => $subobj;
534 $bits = $subobj->{BITS} if $bits < $subobj->{BITS};
535 }
536 }
537 }
538 }
a0ed51b3 539
99870f4d 540 if (DEBUG) {
0fada5e1 541 print STDERR __LINE__, ": CLASS = $class, TYPE => $type, BITS => $bits, NONE => $none, INVERT_IT => $invert_it, USER_DEFINED => $user_defined";
99870f4d
KW
542 print STDERR "\nLIST =>\n$list" if defined $list;
543 print STDERR "\nEXTRAS =>\n$extras" if defined $extras;
544 print STDERR "\n";
545 }
a0ed51b3 546
99870f4d
KW
547 my $SWASH = bless {
548 TYPE => $type,
549 BITS => $bits,
550 EXTRAS => $extras,
551 LIST => $list,
552 NONE => $none,
0fada5e1 553 USER_DEFINED => $user_defined,
99870f4d
KW
554 @extras,
555 } => $class;
556
557 if ($file) {
558 $Cache{$class, $file} = $SWASH;
64870e05
KW
559 if ($type
560 && exists $utf8::SwashInfo{$type}
561 && exists $utf8::SwashInfo{$type}{'specials_name'})
562 {
563 my $specials_name = $utf8::SwashInfo{$type}{'specials_name'};
564 no strict "refs";
69b36ee0 565 print STDERR "\nspecials_name => $specials_name\n" if DEBUG;
64870e05
KW
566 $SWASH->{'SPECIALS'} = \%$specials_name;
567 }
a6517ebf 568 $SWASH->{'INVERT_IT'} = $invert_it;
99870f4d 569 }
5beb625e 570
3dd10fe8
KW
571 pop @recursed if @recursed && $type;
572
99870f4d 573 return $SWASH;
5beb625e 574 }
a0ed51b3
LW
575}
576
979f2922 577# Now SWASHGET is recasted into a C function S_swash_get (see utf8.c).
a0ed51b3
LW
578
5791;