Commit | Line | Data |
---|---|---|
a0ed51b3 | 1 | package utf8; |
cf25bb62 JH |
2 | use strict; |
3 | use warnings; | |
a0ed51b3 | 4 | |
15732964 | 5 | sub DEBUG () { 0 } |
4b17cb47 | 6 | $|=1 if DEBUG; |
a0ed51b3 LW |
7 | |
8 | sub DESTROY {} | |
9 | ||
5beb625e JH |
10 | my %Cache; |
11 | ||
86916d66 JH |
12 | sub croak { require Carp; Carp::croak(@_) } |
13 | ||
c4ab2516 | 14 | sub _loose_name ($) { |
45376db6 KW |
15 | # Given a lowercase property or property-value name, return its |
16 | # standardized version that is expected for look-up in the 'loose' hashes | |
17 | # in Heavy.pl (hence, this depends on what mktables does). This squeezes | |
18 | # out blanks, underscores and dashes. The complication stems from the | |
19 | # grandfathered-in 'L_', which retains a single trailing underscore. | |
20 | ||
21 | my $loose = $_[0] =~ s/[-\s_]//rg; | |
22 | ||
cf2cd801 | 23 | return $loose if $loose !~ / ^ (?: is | to )? l $/x; |
45376db6 KW |
24 | return 'l_' if $_[0] =~ / l .* _ /x; # If original had a trailing '_' |
25 | return $loose; | |
26 | } | |
27 | ||
5beb625e | 28 | ## |
3a2263fe RGS |
29 | ## "SWASH" == "SWATCH HASH". A "swatch" is a swatch of the Unicode landscape. |
30 | ## It's a data structure that encodes a set of Unicode characters. | |
5beb625e JH |
31 | ## |
32 | ||
99870f4d KW |
33 | { |
34 | # If a floating point number is within this distance from the value of a | |
35 | # fraction, it is considered to be that fraction, even if many more digits | |
36 | # are specified that don't exactly match. | |
37 | my $min_floating_slop; | |
38 | ||
3dd10fe8 KW |
39 | # To guard against this program calling something that in turn ends up |
40 | # calling this program with the same inputs, and hence infinitely | |
41 | # recursing, we keep a stack of the properties that are currently in | |
42 | # progress, pushed upon entry, popped upon return. | |
43 | my @recursed; | |
44 | ||
99870f4d KW |
45 | sub SWASHNEW { |
46 | my ($class, $type, $list, $minbits, $none) = @_; | |
0fada5e1 | 47 | my $user_defined = 0; |
99870f4d KW |
48 | local $^D = 0 if $^D; |
49 | ||
3dd10fe8 | 50 | $class = "" unless defined $class; |
5bcf08c4 KW |
51 | print STDERR __LINE__, ": class=$class, type=$type, list=", |
52 | (defined $list) ? $list : ':undef:', | |
53 | ", minbits=$minbits, none=$none\n" if DEBUG; | |
99870f4d KW |
54 | |
55 | ## | |
56 | ## Get the list of codepoints for the type. | |
57 | ## Called from swash_init (see utf8.c) or SWASHNEW itself. | |
58 | ## | |
59 | ## Callers of swash_init: | |
60 | ## op.c:pmtrans -- for tr/// and y/// | |
61 | ## regexec.c:regclass_swash -- for /[]/, \p, and \P | |
62 | ## utf8.c:is_utf8_common -- for common Unicode properties | |
63 | ## utf8.c:to_utf8_case -- for lc, uc, ucfirst, etc. and //i | |
681d705c | 64 | ## Unicode::UCD::prop_invlist |
62b3b855 | 65 | ## Unicode::UCD::prop_invmap |
99870f4d KW |
66 | ## |
67 | ## Given a $type, our goal is to fill $list with the set of codepoint | |
68 | ## ranges. If $type is false, $list passed is used. | |
69 | ## | |
70 | ## $minbits: | |
71 | ## For binary properties, $minbits must be 1. | |
72 | ## For character mappings (case and transliteration), $minbits must | |
73 | ## be a number except 1. | |
74 | ## | |
75 | ## $list (or that filled according to $type): | |
76 | ## Refer to perlunicode.pod, "User-Defined Character Properties." | |
77 | ## | |
78 | ## For binary properties, only characters with the property value | |
79 | ## of True should be listed. The 3rd column, if any, will be ignored | |
80 | ## | |
81 | ## $none is undocumented, so I'm (khw) trying to do some documentation | |
82 | ## of it now. It appears to be if there is a mapping in an input file | |
83 | ## that maps to 'XXXX', then that is replaced by $none+1, expressed in | |
83fe8199 | 84 | ## hexadecimal. It is used somehow in tr///. |
99870f4d KW |
85 | ## |
86 | ## To make the parsing of $type clear, this code takes the a rather | |
87 | ## unorthodox approach of last'ing out of the block once we have the | |
88 | ## info we need. Were this to be a subroutine, the 'last' would just | |
89 | ## be a 'return'. | |
90 | ## | |
fbe6f3cf KW |
91 | # If a problem is found $type is returned; |
92 | # Upon success, a new (or cached) blessed object is returned with | |
93 | # keys TYPE, BITS, EXTRAS, LIST, and NONE with values having the | |
94 | # same meanings as the input parameters. | |
a6517ebf KW |
95 | # SPECIALS contains a reference to any special-treatment hash in the |
96 | # INVERT_IT is non-zero if the result should be inverted before use | |
0fada5e1 KW |
97 | # USER_DEFINED is non-zero if the result came from a user-defined |
98 | # property. | |
99870f4d KW |
99 | my $file; ## file to load data from, and also part of the %Cache key. |
100 | my $ListSorted = 0; | |
101 | ||
102 | # Change this to get a different set of Unicode tables | |
103 | my $unicore_dir = 'unicore'; | |
a6517ebf | 104 | my $invert_it = 0; |
99870f4d KW |
105 | |
106 | if ($type) | |
cf25bb62 | 107 | { |
3dd10fe8 KW |
108 | |
109 | # Verify that this isn't a recursive call for this property. | |
110 | # Can't use croak, as it may try to recurse here itself. | |
111 | my $class_type = $class . "::$type"; | |
112 | if (grep { $_ eq $class_type } @recursed) { | |
113 | CORE::die "panic: Infinite recursion in SWASHNEW for '$type'\n"; | |
114 | } | |
115 | push @recursed, $class_type; | |
116 | ||
99870f4d KW |
117 | $type =~ s/^\s+//; |
118 | $type =~ s/\s+$//; | |
119 | ||
4ff124d1 KW |
120 | # regcomp.c surrounds the property name with '__" and '_i' if this |
121 | # is to be caseless matching. | |
122 | my $caseless = $type =~ s/^__(.*)_i$/$1/; | |
123 | ||
124 | print STDERR __LINE__, ": type=$type, caseless=$caseless\n" if DEBUG; | |
99870f4d KW |
125 | |
126 | GETFILE: | |
127 | { | |
128 | ## | |
a44f4b74 KW |
129 | ## It could be a user-defined property. Look in current |
130 | ## package if no package given | |
99870f4d KW |
131 | ## |
132 | ||
133 | my $caller1 = $type =~ s/(.+)::// ? $1 : caller(1); | |
134 | ||
d658a8a8 | 135 | if (defined $caller1 && $type =~ /^I[ns]\w+$/) { |
99870f4d KW |
136 | my $prop = "${caller1}::$type"; |
137 | if (exists &{$prop}) { | |
0e9be77f DM |
138 | # stolen from Scalar::Util::PP::tainted() |
139 | my $tainted; | |
140 | { | |
141 | local($@, $SIG{__DIE__}, $SIG{__WARN__}); | |
142 | local $^W = 0; | |
143 | no warnings; | |
144 | eval { kill 0 * $prop }; | |
145 | $tainted = 1 if $@ =~ /^Insecure/; | |
146 | } | |
147 | die "Insecure user-defined property \\p{$prop}\n" | |
148 | if $tainted; | |
99870f4d | 149 | no strict 'refs'; |
4ff124d1 | 150 | $list = &{$prop}($caseless); |
0fada5e1 | 151 | $user_defined = 1; |
99870f4d KW |
152 | last GETFILE; |
153 | } | |
154 | } | |
155 | ||
a44f4b74 KW |
156 | # During Perl's compilation, this routine may be called before |
157 | # the tables are constructed. If so, we have a chicken/egg | |
158 | # problem. If we die, the tables never get constructed, so | |
159 | # keep going, but return an empty table so only what the code | |
160 | # has compiled in internally (currently ASCII/Latin1 range | |
161 | # matching) will work. | |
1363cbd0 FC |
162 | BEGIN { |
163 | # Poor man's constant, to avoid a run-time check. | |
164 | $utf8::{miniperl} | |
165 | = \! defined &DynaLoader::boot_DynaLoader; | |
166 | } | |
167 | if (miniperl) { | |
a44f4b74 KW |
168 | eval "require '$unicore_dir/Heavy.pl'"; |
169 | last GETFILE if $@; | |
170 | } | |
171 | else { | |
172 | require "$unicore_dir/Heavy.pl"; | |
173 | } | |
1363cbd0 | 174 | BEGIN { delete $utf8::{miniperl} } |
99870f4d | 175 | |
4ff124d1 | 176 | # All property names are matched caselessly |
49801ec2 | 177 | my $property_and_table = CORE::lc $type; |
99870f4d KW |
178 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; |
179 | ||
180 | # See if is of the compound form 'property=value', where the | |
181 | # value indicates the table we should use. | |
182 | my ($property, $table, @remainder) = | |
183 | split /\s*[:=]\s*/, $property_and_table, -1; | |
3dd10fe8 KW |
184 | if (@remainder) { |
185 | pop @recursed if @recursed; | |
186 | return $type; | |
187 | } | |
99870f4d KW |
188 | |
189 | my $prefix; | |
190 | if (! defined $table) { | |
191 | ||
192 | # Here, is the single form. The property becomes empty, and | |
193 | # the whole value is the table. | |
194 | $table = $property; | |
195 | $prefix = $property = ""; | |
196 | } else { | |
197 | print STDERR __LINE__, ": $property\n" if DEBUG; | |
198 | ||
199 | # Here it is the compound property=table form. The property | |
807807b7 KW |
200 | # name is always loosely matched, and always can have an |
201 | # optional 'is' prefix (which isn't true in the single | |
202 | # form). | |
c4ab2516 | 203 | $property = _loose_name($property) =~ s/^is//r; |
99870f4d KW |
204 | |
205 | # And convert to canonical form. Quit if not valid. | |
206 | $property = $utf8::loose_property_name_of{$property}; | |
3dd10fe8 KW |
207 | if (! defined $property) { |
208 | pop @recursed if @recursed; | |
209 | return $type; | |
210 | } | |
99870f4d KW |
211 | |
212 | $prefix = "$property="; | |
213 | ||
214 | # If the rhs looks like it is a number... | |
215 | print STDERR __LINE__, ": table=$table\n" if DEBUG; | |
216 | if ($table =~ qr{ ^ [ \s 0-9 _ + / . -]+ $ }x) { | |
217 | print STDERR __LINE__, ": table=$table\n" if DEBUG; | |
218 | ||
219 | # Don't allow leading nor trailing slashes | |
3dd10fe8 KW |
220 | if ($table =~ / ^ \/ | \/ $ /x) { |
221 | pop @recursed if @recursed; | |
222 | return $type; | |
223 | } | |
99870f4d KW |
224 | |
225 | # Split on slash, in case it is a rational, like \p{1/5} | |
226 | my @parts = split qr{ \s* / \s* }x, $table, -1; | |
227 | print __LINE__, ": $type\n" if @parts > 2 && DEBUG; | |
228 | ||
229 | # Can have maximum of one slash | |
3dd10fe8 KW |
230 | if (@parts > 2) { |
231 | pop @recursed if @recursed; | |
232 | return $type; | |
233 | } | |
99870f4d KW |
234 | |
235 | foreach my $part (@parts) { | |
236 | print __LINE__, ": part=$part\n" if DEBUG; | |
237 | ||
238 | $part =~ s/^\+\s*//; # Remove leading plus | |
239 | $part =~ s/^-\s*/-/; # Remove blanks after unary | |
240 | # minus | |
241 | ||
242 | # Remove underscores between digits. | |
243 | $part =~ s/( ?<= [0-9] ) _ (?= [0-9] ) //xg; | |
244 | ||
245 | # No leading zeros (but don't make a single '0' | |
246 | # into a null string) | |
247 | $part =~ s/ ^ ( -? ) 0+ /$1/x; | |
248 | $part .= '0' if $part eq '-' || $part eq ""; | |
249 | ||
250 | # No trailing zeros after a decimal point | |
251 | $part =~ s/ ( \. .*? ) 0+ $ /$1/x; | |
252 | ||
253 | # Begin with a 0 if a leading decimal point | |
254 | $part =~ s/ ^ ( -? ) \. /${1}0./x; | |
255 | ||
256 | # Ensure not a trailing decimal point: turn into an | |
257 | # integer | |
258 | $part =~ s/ \. $ //x; | |
259 | ||
260 | print STDERR __LINE__, ": part=$part\n" if DEBUG; | |
261 | #return $type if $part eq ""; | |
262 | ||
263 | # Result better look like a number. (This test is | |
264 | # needed because, for example could have a plus in | |
265 | # the middle.) | |
3dd10fe8 KW |
266 | if ($part !~ / ^ -? [0-9]+ ( \. [0-9]+)? $ /x) { |
267 | pop @recursed if @recursed; | |
268 | return $type; | |
269 | } | |
99870f4d KW |
270 | } |
271 | ||
272 | # If a rational... | |
273 | if (@parts == 2) { | |
274 | ||
275 | # If denominator is negative, get rid of it, and ... | |
276 | if ($parts[1] =~ s/^-//) { | |
277 | ||
278 | # If numerator is also negative, convert the | |
279 | # whole thing to positive, or move the minus to | |
280 | # the numerator | |
281 | if ($parts[0] !~ s/^-//) { | |
282 | $parts[0] = '-' . $parts[0]; | |
283 | } | |
284 | } | |
285 | $table = join '/', @parts; | |
286 | } | |
287 | elsif ($property ne 'nv' || $parts[0] !~ /\./) { | |
288 | ||
289 | # Here is not numeric value, or doesn't have a | |
290 | # decimal point. No further manipulation is | |
291 | # necessary. (Note the hard-coded property name. | |
292 | # This could fail if other properties eventually | |
293 | # had fractions as well; perhaps the cjk ones | |
294 | # could evolve to do that. This hard-coding could | |
295 | # be fixed by mktables generating a list of | |
296 | # properties that could have fractions.) | |
297 | $table = $parts[0]; | |
298 | } else { | |
299 | ||
300 | # Here is a floating point numeric_value. Try to | |
301 | # convert to rational. First see if is in the list | |
302 | # of known ones. | |
303 | if (exists $utf8::nv_floating_to_rational{$parts[0]}) { | |
304 | $table = $utf8::nv_floating_to_rational{$parts[0]}; | |
305 | } else { | |
306 | ||
307 | # Here not in the list. See if is close | |
308 | # enough to something in the list. First | |
309 | # determine what 'close enough' means. It has | |
310 | # to be as tight as what mktables says is the | |
311 | # maximum slop, and as tight as how many | |
312 | # digits we were passed. That is, if the user | |
313 | # said .667, .6667, .66667, etc. we match as | |
314 | # many digits as they passed until get to | |
315 | # where it doesn't matter any more due to the | |
316 | # machine's precision. If they said .6666668, | |
317 | # we fail. | |
318 | (my $fraction = $parts[0]) =~ s/^.*\.//; | |
319 | my $epsilon = 10 ** - (length($fraction)); | |
320 | if ($epsilon > $utf8::max_floating_slop) { | |
321 | $epsilon = $utf8::max_floating_slop; | |
322 | } | |
323 | ||
324 | # But it can't be tighter than the minimum | |
325 | # precision for this machine. If haven't | |
326 | # already calculated that minimum, do so now. | |
327 | if (! defined $min_floating_slop) { | |
328 | ||
329 | # Keep going down an order of magnitude | |
330 | # until find that adding this quantity to | |
331 | # 1 remains 1; but put an upper limit on | |
332 | # this so in case this algorithm doesn't | |
333 | # work properly on some platform, that we | |
334 | # won't loop forever. | |
335 | my $count = 0; | |
336 | $min_floating_slop = 1; | |
337 | while (1+ $min_floating_slop != 1 | |
338 | && $count++ < 50) | |
339 | { | |
340 | my $next = $min_floating_slop / 10; | |
341 | last if $next == 0; # If underflows, | |
342 | # use previous one | |
343 | $min_floating_slop = $next; | |
344 | print STDERR __LINE__, ": min_float_slop=$min_floating_slop\n" if DEBUG; | |
345 | } | |
346 | ||
347 | # Back off a couple orders of magnitude, | |
348 | # just to be safe. | |
349 | $min_floating_slop *= 100; | |
350 | } | |
351 | ||
352 | if ($epsilon < $min_floating_slop) { | |
353 | $epsilon = $min_floating_slop; | |
354 | } | |
355 | print STDERR __LINE__, ": fraction=.$fraction; epsilon=$epsilon\n" if DEBUG; | |
356 | ||
357 | undef $table; | |
358 | ||
359 | # And for each possible rational in the table, | |
360 | # see if it is within epsilon of the input. | |
361 | foreach my $official | |
362 | (keys %utf8::nv_floating_to_rational) | |
363 | { | |
364 | print STDERR __LINE__, ": epsilon=$epsilon, official=$official, diff=", abs($parts[0] - $official), "\n" if DEBUG; | |
365 | if (abs($parts[0] - $official) < $epsilon) { | |
366 | $table = | |
367 | $utf8::nv_floating_to_rational{$official}; | |
368 | last; | |
369 | } | |
370 | } | |
371 | ||
372 | # Quit if didn't find one. | |
3dd10fe8 KW |
373 | if (! defined $table) { |
374 | pop @recursed if @recursed; | |
375 | return $type; | |
376 | } | |
99870f4d KW |
377 | } |
378 | } | |
379 | print STDERR __LINE__, ": $property=$table\n" if DEBUG; | |
380 | } | |
381 | } | |
382 | ||
383 | # Combine lhs (if any) and rhs to get something that matches | |
384 | # the syntax of the lookups. | |
385 | $property_and_table = "$prefix$table"; | |
386 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | |
387 | ||
388 | # First try stricter matching. | |
389 | $file = $utf8::stricter_to_file_of{$property_and_table}; | |
390 | ||
391 | # If didn't find it, try again with looser matching by editing | |
392 | # out the applicable characters on the rhs and looking up | |
393 | # again. | |
394 | if (! defined $file) { | |
c4ab2516 | 395 | $table = _loose_name($table); |
99870f4d KW |
396 | $property_and_table = "$prefix$table"; |
397 | print STDERR __LINE__, ": $property_and_table\n" if DEBUG; | |
398 | $file = $utf8::loose_to_file_of{$property_and_table}; | |
399 | } | |
400 | ||
401 | # Add the constant and go fetch it in. | |
402 | if (defined $file) { | |
a6517ebf KW |
403 | |
404 | # A beginning ! means to invert | |
405 | $invert_it = $file =~ s/^!//; | |
406 | ||
99870f4d KW |
407 | if ($utf8::why_deprecated{$file}) { |
408 | warnings::warnif('deprecated', "Use of '$type' in \\p{} or \\P{} is deprecated because: $utf8::why_deprecated{$file};"); | |
409 | } | |
4ff124d1 KW |
410 | |
411 | if ($caseless | |
412 | && exists $utf8::caseless_equivalent{$property_and_table}) | |
413 | { | |
414 | $file = $utf8::caseless_equivalent{$property_and_table}; | |
415 | } | |
99870f4d KW |
416 | $file= "$unicore_dir/lib/$file.pl"; |
417 | last GETFILE; | |
418 | } | |
419 | print STDERR __LINE__, ": didn't find $property_and_table\n" if DEBUG; | |
420 | ||
421 | ## | |
99870f4d KW |
422 | ## Last attempt -- see if it's a standard "To" name |
423 | ## (e.g. "ToLower") ToTitle is used by ucfirst(). | |
424 | ## The user-level way to access ToDigit() and ToFold() | |
425 | ## is to use Unicode::UCD. | |
426 | ## | |
cf2cd801 | 427 | # Only check if caller wants non-binary |
6aec79bd KW |
428 | my $retried = 0; |
429 | if ($minbits != 1 && $property_and_table =~ s/^to//) {{ | |
cf2cd801 KW |
430 | # Look input up in list of properties for which we have |
431 | # mapping files. | |
432 | if (defined ($file = | |
433 | $utf8::loose_property_to_file_of{$property_and_table})) | |
434 | { | |
435 | $type = $utf8::file_to_swash_name{$file}; | |
436 | print STDERR __LINE__, ": type set to $type\n" if DEBUG; | |
437 | $file = "$unicore_dir/$file.pl"; | |
438 | last GETFILE; | |
6aec79bd KW |
439 | } # If that fails see if there is a corresponding binary |
440 | # property file | |
441 | elsif (defined ($file = | |
442 | $utf8::loose_to_file_of{$property_and_table})) | |
443 | { | |
444 | ||
445 | # Here, there is no map file for the property we are | |
446 | # trying to get the map of, but this is a binary | |
447 | # property, and there is a file for it that can easily | |
448 | # be translated to a mapping. | |
449 | ||
450 | # In the case of properties that are forced to binary, | |
451 | # they are a combination. We return the actual | |
452 | # mapping instead of the binary. If the input is | |
453 | # something like 'Tocjkkiicore', it will be found in | |
454 | # %loose_property_to_file_of above as => 'To/kIICore'. | |
455 | # But the form like ToIskiicore won't be. To fix | |
456 | # this, it was easiest to do it here. These | |
457 | # properties are the complements of the default | |
458 | # property, so there is an entry in %loose_to_file_of | |
459 | # that is 'iskiicore' => '!kIICore/N', If we find such | |
460 | # an entry, strip off things and try again, which | |
461 | # should find the entry in %loose_property_to_file_of. | |
462 | # Actual binary properties that are of this form, such | |
463 | # as this entry: 'ishrkt' => '!Perl/Any' will also be | |
464 | # retried, but won't be in %loose_property_to_file_of, | |
465 | # and instead the next time through, it will find | |
466 | # 'hrkt' => '!Perl/Any' and proceed. | |
467 | redo if ! $retried | |
468 | && $file =~ /^!/ | |
469 | && $property_and_table =~ s/^is//; | |
470 | ||
471 | # This is a binary property. Setting this here causes | |
472 | # it to be stored as such in the cache, so if someone | |
473 | # comes along later looking for just a binary, they | |
474 | # get it. | |
475 | $minbits = 1; | |
476 | ||
477 | $invert_it = $file =~ s/^!//; | |
478 | $file = "$unicore_dir/lib/$file.pl"; | |
479 | last GETFILE; | |
cf2cd801 | 480 | } |
6aec79bd | 481 | } } |
99870f4d KW |
482 | |
483 | ## | |
484 | ## If we reach this line, it's because we couldn't figure | |
485 | ## out what to do with $type. Ouch. | |
486 | ## | |
487 | ||
3dd10fe8 | 488 | pop @recursed if @recursed; |
99870f4d | 489 | return $type; |
c9db9632 | 490 | } # end of GETFILE block |
cf25bb62 | 491 | |
99870f4d KW |
492 | if (defined $file) { |
493 | print STDERR __LINE__, ": found it (file='$file')\n" if DEBUG; | |
494 | ||
495 | ## | |
496 | ## If we reach here, it was due to a 'last GETFILE' above | |
497 | ## (exception: user-defined properties and mappings), so we | |
498 | ## have a filename, so now we load it if we haven't already. | |
499 | ## If we have, return the cached results. The cache key is the | |
500 | ## class and file to load. | |
501 | ## | |
502 | my $found = $Cache{$class, $file}; | |
503 | if ($found and ref($found) eq $class) { | |
a6517ebf | 504 | print STDERR __LINE__, ": Returning cached '$file' for \\p{$type}; invert_it=$invert_it\n" if DEBUG; |
3dd10fe8 | 505 | pop @recursed if @recursed; |
a6517ebf | 506 | $found->{'INVERT_IT'} = $invert_it; |
99870f4d KW |
507 | return $found; |
508 | } | |
509 | ||
510 | local $@; | |
511 | local $!; | |
512 | $list = do $file; die $@ if $@; | |
0a0ffbce | 513 | } |
cf25bb62 | 514 | |
99870f4d | 515 | $ListSorted = 1; ## we know that these lists are sorted |
c9db9632 KW |
516 | } # End of $type is non-null |
517 | ||
518 | # Here, either $type was null, or we found the requested property and | |
519 | # read it into $list | |
cf25bb62 | 520 | |
99870f4d | 521 | my $extras; |
c9db9632 | 522 | |
99870f4d KW |
523 | my $bits = $minbits; |
524 | ||
99870f4d | 525 | if ($list) { |
0e9be77f | 526 | my $taint = substr($list,0,0); # maintain taint |
99870f4d KW |
527 | my @tmp = split(/^/m, $list); |
528 | my %seen; | |
529 | no warnings; | |
0e9be77f DM |
530 | $extras = join '', $taint, grep /^[^0-9a-fA-F]/, @tmp; |
531 | $list = join '', $taint, | |
99870f4d KW |
532 | map { $_->[1] } |
533 | sort { $a->[0] <=> $b->[0] } | |
534 | map { /^([0-9a-fA-F]+)/; [ CORE::hex($1), $_ ] } | |
535 | grep { /^([0-9a-fA-F]+)/ and not $seen{$1}++ } @tmp; # XXX doesn't do ranges right | |
536 | } | |
a0ed51b3 | 537 | |
99870f4d KW |
538 | if ($none) { |
539 | my $hextra = sprintf "%04x", $none + 1; | |
540 | $list =~ s/\tXXXX$/\t$hextra/mg; | |
541 | } | |
a0ed51b3 | 542 | |
99870f4d KW |
543 | if ($minbits != 1 && $minbits < 32) { # not binary property |
544 | my $top = 0; | |
545 | while ($list =~ /^([0-9a-fA-F]+)(?:[\t]([0-9a-fA-F]+)?)(?:[ \t]([0-9a-fA-F]+))?/mg) { | |
546 | my $min = CORE::hex $1; | |
547 | my $max = defined $2 ? CORE::hex $2 : $min; | |
548 | my $val = defined $3 ? CORE::hex $3 : 0; | |
549 | $val += $max - $min if defined $3; | |
550 | $top = $val if $val > $top; | |
551 | } | |
552 | my $topbits = | |
553 | $top > 0xffff ? 32 : | |
554 | $top > 0xff ? 16 : 8; | |
555 | $bits = $topbits if $bits < $topbits; | |
556 | } | |
a0ed51b3 | 557 | |
99870f4d KW |
558 | my @extras; |
559 | if ($extras) { | |
560 | for my $x ($extras) { | |
0e9be77f | 561 | my $taint = substr($x,0,0); # maintain taint |
99870f4d KW |
562 | pos $x = 0; |
563 | while ($x =~ /^([^0-9a-fA-F\n])(.*)/mg) { | |
0e9be77f DM |
564 | my $char = "$1$taint"; |
565 | my $name = "$2$taint"; | |
566 | print STDERR __LINE__, ": char [$char] => name [$name]\n" | |
567 | if DEBUG; | |
99870f4d KW |
568 | if ($char =~ /[-+!&]/) { |
569 | my ($c,$t) = split(/::/, $name, 2); # bogus use of ::, really | |
570 | my $subobj; | |
571 | if ($c eq 'utf8') { | |
572 | $subobj = utf8->SWASHNEW($t, "", $minbits, 0); | |
573 | } | |
574 | elsif (exists &$name) { | |
575 | $subobj = utf8->SWASHNEW($name, "", $minbits, 0); | |
576 | } | |
577 | elsif ($c =~ /^([0-9a-fA-F]+)/) { | |
578 | $subobj = utf8->SWASHNEW("", $c, $minbits, 0); | |
579 | } | |
3dd10fe8 KW |
580 | if (! ref $subobj) { |
581 | pop @recursed if @recursed && $type; | |
582 | return $subobj; | |
583 | } | |
99870f4d KW |
584 | push @extras, $name => $subobj; |
585 | $bits = $subobj->{BITS} if $bits < $subobj->{BITS}; | |
04e9cbbb KW |
586 | $user_defined = $subobj->{USER_DEFINED} |
587 | if $subobj->{USER_DEFINED}; | |
99870f4d KW |
588 | } |
589 | } | |
590 | } | |
591 | } | |
a0ed51b3 | 592 | |
99870f4d | 593 | if (DEBUG) { |
0fada5e1 | 594 | print STDERR __LINE__, ": CLASS = $class, TYPE => $type, BITS => $bits, NONE => $none, INVERT_IT => $invert_it, USER_DEFINED => $user_defined"; |
99870f4d KW |
595 | print STDERR "\nLIST =>\n$list" if defined $list; |
596 | print STDERR "\nEXTRAS =>\n$extras" if defined $extras; | |
597 | print STDERR "\n"; | |
598 | } | |
a0ed51b3 | 599 | |
99870f4d KW |
600 | my $SWASH = bless { |
601 | TYPE => $type, | |
602 | BITS => $bits, | |
603 | EXTRAS => $extras, | |
604 | LIST => $list, | |
605 | NONE => $none, | |
0fada5e1 | 606 | USER_DEFINED => $user_defined, |
99870f4d KW |
607 | @extras, |
608 | } => $class; | |
609 | ||
610 | if ($file) { | |
611 | $Cache{$class, $file} = $SWASH; | |
64870e05 KW |
612 | if ($type |
613 | && exists $utf8::SwashInfo{$type} | |
614 | && exists $utf8::SwashInfo{$type}{'specials_name'}) | |
615 | { | |
616 | my $specials_name = $utf8::SwashInfo{$type}{'specials_name'}; | |
617 | no strict "refs"; | |
69b36ee0 | 618 | print STDERR "\nspecials_name => $specials_name\n" if DEBUG; |
64870e05 KW |
619 | $SWASH->{'SPECIALS'} = \%$specials_name; |
620 | } | |
a6517ebf | 621 | $SWASH->{'INVERT_IT'} = $invert_it; |
99870f4d | 622 | } |
5beb625e | 623 | |
3dd10fe8 KW |
624 | pop @recursed if @recursed && $type; |
625 | ||
99870f4d | 626 | return $SWASH; |
5beb625e | 627 | } |
a0ed51b3 LW |
628 | } |
629 | ||
979f2922 | 630 | # Now SWASHGET is recasted into a C function S_swash_get (see utf8.c). |
a0ed51b3 LW |
631 | |
632 | 1; |