Commit | Line | Data |
---|---|---|
9d9177be KW |
1 | #!perl -w |
2 | use 5.015; | |
3 | use strict; | |
4 | use warnings; | |
99f21fb9 KW |
5 | use Unicode::UCD qw(prop_aliases |
6 | prop_values | |
7 | prop_value_aliases | |
8 | prop_invlist | |
9 | prop_invmap search_invlist | |
463b4a67 | 10 | charprop |
a1c8344d | 11 | num |
2cd613ec | 12 | charblock |
99f21fb9 | 13 | ); |
3d7c117d MB |
14 | require './regen/regen_lib.pl'; |
15 | require './regen/charset_translations.pl'; | |
048bdb72 | 16 | require './lib/unicore/UCD.pl'; |
64ff2b0f | 17 | require './regen/mph.pl'; |
db95f459 | 18 | use re "/aa"; |
9d9177be KW |
19 | |
20 | # This program outputs charclass_invlists.h, which contains various inversion | |
21 | # lists in the form of C arrays that are to be used as-is for inversion lists. | |
22 | # Thus, the lists it contains are essentially pre-compiled, and need only a | |
23 | # light-weight fast wrapper to make them usable at run-time. | |
24 | ||
25 | # As such, this code knows about the internal structure of these lists, and | |
26 | # any change made to that has to be done here as well. A random number stored | |
27 | # in the headers is used to minimize the possibility of things getting | |
28 | # out-of-sync, or the wrong data structure being passed. Currently that | |
29 | # random number is: | |
99f21fb9 | 30 | |
1c7f0e60 KW |
31 | my $VERSION_DATA_STRUCTURE_TYPE = 148565664; |
32 | ||
b72e0f31 KW |
33 | # charclass_invlists.h now also contains inversion maps and enum definitions |
34 | # for those maps that have a finite number of possible values | |
99f21fb9 | 35 | |
98a1b8f7 KW |
36 | # integer or float (no exponent) |
37 | my $integer_or_float_re = qr/ ^ -? \d+ (:? \. \d+ )? $ /x; | |
38 | ||
39 | # Also includes rationals | |
40 | my $numeric_re = qr! $integer_or_float_re | ^ -? \d+ / \d+ $ !x; | |
99f21fb9 | 41 | |
cb2d98ed KW |
42 | # More than one code point may have the same code point as their fold. This |
43 | # gives the maximum number in the current Unicode release. (The folded-to | |
44 | # code point is not included in this count.) Most folds are pairs of code | |
45 | # points, like 'B' and 'b', so this number is at least one. | |
46 | my $max_fold_froms = 1; | |
47 | ||
f4b10e8e | 48 | my %keywords; |
cef72199 | 49 | my $table_name_prefix = "UNI_"; |
4eea95a6 | 50 | |
99f21fb9 KW |
51 | # Matches valid C language enum names: begins with ASCII alphabetic, then any |
52 | # ASCII \w | |
53 | my $enum_name_re = qr / ^ [[:alpha:]] \w* $ /ax; | |
54 | ||
9d9177be | 55 | my $out_fh = open_new('charclass_invlists.h', '>', |
9824c081 | 56 | {style => '*', by => 'regen/mk_invlists.pl', |
9d9177be KW |
57 | from => "Unicode::UCD"}); |
58 | ||
0f8eed22 | 59 | my $in_file_pound_if = ""; |
43b443dd | 60 | |
289ce9cc KW |
61 | my $max_hdr_len = 3; # In headings, how wide a name is allowed? |
62 | ||
9d9177be KW |
63 | print $out_fh "/* See the generating file for comments */\n\n"; |
64 | ||
a405530a KW |
65 | print $out_fh <<'EOF'; |
66 | /* This gives the number of code points that can be in the bitmap of an ANYOF | |
67 | * node. The shift number must currently be one of: 8..12. It can't be less | |
68 | * than 8 (256) because some code relies on it being at least that. Above 12 | |
69 | * (4096), and you start running into warnings that some data structure widths | |
70 | * have been exceeded, though the test suite as of this writing still passes | |
71 | * for up through 16, which is as high as anyone would ever want to go, | |
72 | * encompassing all of the Unicode BMP, and thus including all the economically | |
73 | * important world scripts. At 12 most of them are: including Arabic, | |
74 | * Cyrillic, Greek, Hebrew, Indian subcontinent, Latin, and Thai; but not Han, | |
75 | * Japanese, nor Korean. (The regarglen structure in regnodes.h is a U8, and | |
76 | * the trie types TRIEC and AHOCORASICKC are larger than U8 for shift values | |
77 | * above 12.) Be sure to benchmark before changing, as larger sizes do | |
78 | * significantly slow down the test suite */ | |
79 | ||
80 | EOF | |
81 | ||
82 | my $num_anyof_code_points = '(1 << 8)'; | |
83 | ||
84 | print $out_fh "#define NUM_ANYOF_CODE_POINTS $num_anyof_code_points\n\n"; | |
85 | ||
86 | $num_anyof_code_points = eval $num_anyof_code_points; | |
87 | ||
da477c7d KW |
88 | no warnings 'once'; |
89 | print $out_fh <<"EOF"; | |
90 | /* The precision to use in "%.*e" formats */ | |
91 | #define PL_E_FORMAT_PRECISION $Unicode::UCD::e_precision | |
92 | EOF | |
93 | ||
d74e7480 KW |
94 | # enums that should be made public |
95 | my %public_enums = ( | |
f52cc976 | 96 | _Perl_SCX => 1 |
d74e7480 KW |
97 | ); |
98 | ||
bffc0129 KW |
99 | # The symbols generated by this program are all currently defined only in a |
100 | # single dot c each. The code knows where most of them go, but this hash | |
101 | # gives overrides for the exceptions to the typical place | |
102 | my %exceptions_to_where_to_define = | |
7dddaf74 | 103 | ( |
03d17b6e | 104 | #_Perl_IVCF => 'PERL_IN_REGCOMP_C', |
bffc0129 | 105 | ); |
4761f74a | 106 | |
c0221e16 | 107 | my %where_to_define_enums = (); |
015bb97c | 108 | |
59fc10af KW |
109 | my $applies_to_all_charsets_text = "all charsets"; |
110 | ||
973a28ed KW |
111 | my %gcb_enums; |
112 | my @gcb_short_enums; | |
289ce9cc | 113 | my %gcb_abbreviations; |
6b659339 KW |
114 | my %lb_enums; |
115 | my @lb_short_enums; | |
289ce9cc | 116 | my %lb_abbreviations; |
7e54b87f KW |
117 | my %wb_enums; |
118 | my @wb_short_enums; | |
289ce9cc | 119 | my %wb_abbreviations; |
6b659339 | 120 | |
99f21fb9 KW |
121 | my @a2n; |
122 | ||
394d2d3f KW |
123 | my %prop_name_aliases; |
124 | # Invert this hash so that for each canonical name, we get a list of things | |
125 | # that map to it (excluding itself) | |
048bdb72 KW |
126 | foreach my $name (sort keys %Unicode::UCD::loose_property_name_of) { |
127 | my $canonical = $Unicode::UCD::loose_property_name_of{$name}; | |
394d2d3f KW |
128 | push @{$prop_name_aliases{$canonical}}, $name if $canonical ne $name; |
129 | } | |
130 | ||
2d74dcf2 | 131 | # Output these tables in the same vicinity as each other, so that will get |
1aefa327 KW |
132 | # paged in at about the same time. These are also assumed to be the exact |
133 | # same list as those properties used internally by perl. | |
2d74dcf2 KW |
134 | my %keep_together = ( |
135 | assigned => 1, | |
136 | ascii => 1, | |
7a6f6841 KW |
137 | upper => 1, |
138 | lower => 1, | |
139 | title => 1, | |
2d74dcf2 | 140 | cased => 1, |
7a6f6841 KW |
141 | uppercaseletter => 1, |
142 | lowercaseletter => 1, | |
143 | titlecaseletter => 1, | |
144 | casedletter => 1, | |
2d74dcf2 KW |
145 | vertspace => 1, |
146 | xposixalnum => 1, | |
147 | xposixalpha => 1, | |
148 | xposixblank => 1, | |
149 | xposixcntrl => 1, | |
150 | xposixdigit => 1, | |
151 | xposixgraph => 1, | |
152 | xposixlower => 1, | |
153 | xposixprint => 1, | |
154 | xposixpunct => 1, | |
155 | xposixspace => 1, | |
156 | xposixupper => 1, | |
157 | xposixword => 1, | |
158 | xposixxdigit => 1, | |
159 | posixalnum => 1, | |
160 | posixalpha => 1, | |
161 | posixblank => 1, | |
162 | posixcntrl => 1, | |
163 | posixdigit => 1, | |
164 | posixgraph => 1, | |
165 | posixlower => 1, | |
166 | posixprint => 1, | |
167 | posixpunct => 1, | |
168 | posixspace => 1, | |
169 | posixupper => 1, | |
170 | posixword => 1, | |
171 | posixxdigit => 1, | |
172 | _perl_any_folds => 1, | |
173 | _perl_folds_to_multi_char => 1, | |
441f4830 KW |
174 | _perl_is_in_multi_char_fold => 1, |
175 | _perl_non_final_folds => 1, | |
2d74dcf2 KW |
176 | _perl_idstart => 1, |
177 | _perl_idcont => 1, | |
178 | _perl_charname_begin => 1, | |
179 | _perl_charname_continue => 1, | |
180 | _perl_problematic_locale_foldeds_start => 1, | |
181 | _perl_problematic_locale_folds => 1, | |
182 | _perl_quotemeta => 1, | |
183 | ); | |
1aefa327 | 184 | my %perl_tags; # So can find synonyms of the above properties |
2d74dcf2 | 185 | |
2027d365 KW |
186 | my $unused_table_hdr = 'u'; # Heading for row or column for unused values |
187 | ||
99f21fb9 KW |
188 | sub uniques { |
189 | # Returns non-duplicated input values. From "Perl Best Practices: | |
190 | # Encapsulated Cleverness". p. 455 in first edition. | |
191 | ||
192 | my %seen; | |
193 | return grep { ! $seen{$_}++ } @_; | |
194 | } | |
195 | ||
18072598 KW |
196 | sub caselessly { lc $a cmp lc $b } |
197 | ||
99f21fb9 KW |
198 | sub a2n($) { |
199 | my $cp = shift; | |
200 | ||
201 | # Returns the input Unicode code point translated to native. | |
202 | ||
98a1b8f7 | 203 | return $cp if $cp !~ $integer_or_float_re || $cp > 255; |
99f21fb9 KW |
204 | return $a2n[$cp]; |
205 | } | |
206 | ||
bffc0129 KW |
207 | sub end_file_pound_if { |
208 | if ($in_file_pound_if) { | |
209 | print $out_fh "\n#endif\t/* $in_file_pound_if */\n"; | |
0f8eed22 | 210 | $in_file_pound_if = ""; |
bffc0129 KW |
211 | } |
212 | } | |
213 | ||
48737b77 KW |
214 | sub end_charset_pound_if { |
215 | print $out_fh "\n" . get_conditional_compile_line_end(); | |
216 | } | |
217 | ||
8ec55631 | 218 | sub switch_pound_if ($$;$) { |
bffc0129 KW |
219 | my $name = shift; |
220 | my $new_pound_if = shift; | |
8ec55631 KW |
221 | my $charset = shift; |
222 | ||
62a54bb7 | 223 | my @new_pound_if = ref ($new_pound_if) |
0f8eed22 | 224 | ? sort @$new_pound_if |
62a54bb7 | 225 | : $new_pound_if; |
bffc0129 KW |
226 | |
227 | # Switch to new #if given by the 2nd argument. If there is an override | |
228 | # for this, it instead switches to that. The 1st argument is the | |
0f8eed22 | 229 | # static's name, used only to check if there is an override for this |
8ec55631 KW |
230 | # |
231 | # The 'charset' parmameter, if present, is used to first end the charset | |
232 | # #if if we actually do a switch, and then restart it afterwards. This | |
233 | # code, then assumes that the charset #if's are enclosed in the file ones. | |
bffc0129 KW |
234 | |
235 | if (exists $exceptions_to_where_to_define{$name}) { | |
62a54bb7 | 236 | @new_pound_if = $exceptions_to_where_to_define{$name}; |
bffc0129 KW |
237 | } |
238 | ||
0f8eed22 | 239 | foreach my $element (@new_pound_if) { |
cef72199 KW |
240 | |
241 | # regcomp.c is arranged so that the tables are not compiled in | |
b619f93d KW |
242 | # re_comp.c, but general enums and defines (which take no space) are |
243 | # compiled */ | |
244 | my $no_xsub = 1 if $name !~ /enum|define/ | |
245 | && $element =~ / PERL_IN_ (?: REGCOMP ) _C /x; | |
0f8eed22 | 246 | $element = "defined($element)"; |
cef72199 | 247 | $element = "($element && ! defined(PERL_IN_XSUB_RE))" if $no_xsub; |
bffc0129 | 248 | } |
0f8eed22 | 249 | $new_pound_if = join " || ", @new_pound_if; |
bffc0129 | 250 | |
0f8eed22 KW |
251 | # Change to the new one if different from old |
252 | if ($in_file_pound_if ne $new_pound_if) { | |
253 | ||
8ec55631 KW |
254 | end_charset_pound_if() if defined $charset; |
255 | ||
0f8eed22 KW |
256 | # Exit any current #if |
257 | if ($in_file_pound_if) { | |
258 | end_file_pound_if; | |
62a54bb7 | 259 | } |
0f8eed22 KW |
260 | |
261 | $in_file_pound_if = $new_pound_if; | |
bffc0129 | 262 | print $out_fh "\n#if $in_file_pound_if\n"; |
8ec55631 KW |
263 | |
264 | start_charset_pound_if ($charset, 1) if defined $charset; | |
43b443dd KW |
265 | } |
266 | } | |
267 | ||
48737b77 KW |
268 | sub start_charset_pound_if ($;$) { |
269 | print $out_fh "\n" . get_conditional_compile_line_start(shift, shift); | |
270 | } | |
271 | ||
cef72199 KW |
272 | { # Closure |
273 | my $fh; | |
274 | my $in_doinit = 0; | |
275 | ||
276 | sub output_table_header($$$;$@) { | |
277 | ||
278 | # Output to $fh the heading for a table given by the other inputs | |
279 | ||
280 | $fh = shift; | |
281 | my ($type, # typedef of table, like UV, UV* | |
282 | $name, # name of table | |
283 | $comment, # Optional comment to put on header line | |
284 | @sizes # Optional sizes of each array index. If omitted, | |
285 | # there is a single index whose size is computed by | |
286 | # the C compiler. | |
287 | ) = @_; | |
288 | ||
289 | $type =~ s/ \s+ $ //x; | |
290 | ||
291 | # If a the typedef is a ptr, add in an extra const | |
292 | $type .= " const" if $type =~ / \* $ /x; | |
293 | ||
294 | $comment = "" unless defined $comment; | |
295 | $comment = " /* $comment */" if $comment; | |
296 | ||
297 | my $array_declaration; | |
298 | if (@sizes) { | |
299 | $array_declaration = ""; | |
300 | $array_declaration .= "[$_]" for @sizes; | |
301 | } | |
302 | else { | |
303 | $array_declaration = '[]'; | |
304 | } | |
305 | ||
306 | my $declaration = "$type ${name}$array_declaration"; | |
307 | ||
308 | # Things not matching this are static. Otherwise, it is an external | |
309 | # constant, initialized only under DOINIT. | |
310 | # | |
311 | # (Currently everything is static) | |
312 | if ($in_file_pound_if !~ / PERL_IN_ (?: ) _C /x) { | |
313 | $in_doinit = 0; | |
314 | print $fh "\nstatic const $declaration = {$comment\n"; | |
315 | } | |
316 | else { | |
317 | $in_doinit = 1; | |
318 | print $fh <<EOF; | |
319 | ||
320 | # ifndef DOINIT | |
321 | ||
322 | EXTCONST $declaration; | |
323 | ||
324 | # else | |
325 | ||
326 | EXTCONST $declaration = {$comment | |
327 | EOF | |
328 | } | |
329 | } | |
330 | ||
331 | sub output_table_trailer() { | |
332 | ||
333 | # Close out a table started by output_table_header() | |
334 | ||
335 | print $fh "};\n"; | |
336 | if ($in_doinit) { | |
337 | print $fh "\n# endif /* DOINIT */\n\n"; | |
338 | $in_doinit = 0; | |
339 | } | |
340 | } | |
341 | } # End closure | |
342 | ||
343 | ||
0c4ecf42 | 344 | sub output_invlist ($$;$) { |
9d9177be KW |
345 | my $name = shift; |
346 | my $invlist = shift; # Reference to inversion list array | |
0c4ecf42 | 347 | my $charset = shift // ""; # name of character set for comment |
9d9177be | 348 | |
76d3994c | 349 | die "No inversion list for $name" unless defined $invlist |
ad85f59a | 350 | && ref $invlist eq 'ARRAY'; |
76d3994c | 351 | |
9d9177be KW |
352 | # Output the inversion list $invlist using the name $name for it. |
353 | # It is output in the exact internal form for inversion lists. | |
354 | ||
a0316a6c KW |
355 | # Is the last element of the header 0, or 1 ? |
356 | my $zero_or_one = 0; | |
ad85f59a | 357 | if (@$invlist && $invlist->[0] != 0) { |
a0316a6c | 358 | unshift @$invlist, 0; |
9d9177be KW |
359 | $zero_or_one = 1; |
360 | } | |
361 | ||
cef72199 KW |
362 | $charset = "for $charset" if $charset; |
363 | output_table_header($out_fh, "UV", "${name}_invlist", $charset); | |
9d9177be | 364 | |
cef72199 KW |
365 | my $count = @$invlist; |
366 | print $out_fh <<EOF; | |
367 | \t$count,\t/* Number of elements */ | |
368 | \t$VERSION_DATA_STRUCTURE_TYPE, /* Version and data structure type */ | |
369 | \t$zero_or_one,\t/* 0 if the list starts at 0; | |
370 | \t\t 1 if it starts at the element beyond 0 */ | |
371 | EOF | |
9d9177be KW |
372 | |
373 | # The main body are the UVs passed in to this routine. Do the final | |
374 | # element separately | |
47d53124 KW |
375 | for my $i (0 .. @$invlist - 1) { |
376 | printf $out_fh "\t0x%X", $invlist->[$i]; | |
377 | print $out_fh "," if $i < @$invlist - 1; | |
378 | print $out_fh "\n"; | |
9d9177be KW |
379 | } |
380 | ||
cef72199 | 381 | output_table_trailer(); |
9d9177be KW |
382 | } |
383 | ||
99f21fb9 KW |
384 | sub output_invmap ($$$$$$$) { |
385 | my $name = shift; | |
386 | my $invmap = shift; # Reference to inversion map array | |
387 | my $prop_name = shift; | |
388 | my $input_format = shift; # The inversion map's format | |
389 | my $default = shift; # The property value for code points who | |
390 | # otherwise don't have a value specified. | |
391 | my $extra_enums = shift; # comma-separated list of our additions to the | |
392 | # property's standard possible values | |
393 | my $charset = shift // ""; # name of character set for comment | |
394 | ||
395 | # Output the inversion map $invmap for property $prop_name, but use $name | |
396 | # as the actual data structure's name. | |
397 | ||
398 | my $count = @$invmap; | |
399 | ||
400 | my $output_format; | |
18230d9d KW |
401 | my $invmap_declaration_type; |
402 | my $enum_declaration_type; | |
403 | my $aux_declaration_type; | |
99f21fb9 KW |
404 | my %enums; |
405 | my $name_prefix; | |
406 | ||
18230d9d | 407 | if ($input_format =~ / ^ [as] l? $ /x) { |
cf2cd619 KW |
408 | $prop_name = (prop_aliases($prop_name))[1] |
409 | // $prop_name =~ s/^_Perl_//r; # Get full name | |
02f811dd | 410 | my $short_name = (prop_aliases($prop_name))[0] // $prop_name; |
226b74db | 411 | my @input_enums; |
f79a09fc | 412 | |
226b74db | 413 | # Find all the possible input values. These become the enum names |
34623dbb KW |
414 | # that comprise the inversion map. For inputs that don't have sub |
415 | # lists, we can just get the unique values. Otherwise, we have to | |
416 | # expand the sublists first. | |
18230d9d | 417 | if ($input_format !~ / ^ a /x) { |
563f8b93 | 418 | if ($input_format ne 'sl') { |
18072598 | 419 | @input_enums = sort caselessly uniques(@$invmap); |
563f8b93 KW |
420 | } |
421 | else { | |
422 | foreach my $element (@$invmap) { | |
423 | if (ref $element) { | |
424 | push @input_enums, @$element; | |
425 | } | |
426 | else { | |
427 | push @input_enums, $element; | |
428 | } | |
34623dbb | 429 | } |
18072598 | 430 | @input_enums = sort caselessly uniques(@input_enums); |
34623dbb | 431 | } |
18230d9d | 432 | } |
6b659339 | 433 | |
226b74db | 434 | # The internal enums come last, and in the order specified. |
2027d365 KW |
435 | # |
436 | # The internal one named EDGE is also used a marker. Any ones that | |
437 | # come after it are used in the algorithms below, and so must be | |
438 | # defined, even if the release of Unicode this is being compiled for | |
439 | # doesn't use them. But since no code points are assigned to them in | |
440 | # such a release, those values will never be accessed. We collapse | |
441 | # all of them into a single placholder row and a column. The | |
442 | # algorithms below will fill in those cells with essentially garbage, | |
443 | # but they are never read, so it doesn't matter. This allows the | |
444 | # algorithm to remain the same from release to release. | |
445 | # | |
446 | # In one case, regexec.c also uses a placeholder which must be defined | |
447 | # here, and we put it in the unused row and column as its value is | |
448 | # never read. | |
449 | # | |
226b74db | 450 | my @enums = @input_enums; |
27a619f7 | 451 | my @extras; |
2027d365 KW |
452 | my @unused_enums; |
453 | my $unused_enum_value = @enums; | |
27a619f7 KW |
454 | if ($extra_enums ne "") { |
455 | @extras = split /,/, $extra_enums; | |
2027d365 | 456 | my $seen_EDGE = 0; |
226b74db KW |
457 | |
458 | # Don't add if already there. | |
459 | foreach my $this_extra (@extras) { | |
460 | next if grep { $_ eq $this_extra } @enums; | |
2027d365 KW |
461 | if ($this_extra eq 'EDGE') { |
462 | push @enums, $this_extra; | |
463 | $seen_EDGE = 1; | |
464 | } | |
465 | elsif ($seen_EDGE) { | |
466 | push @unused_enums, $this_extra; | |
467 | } | |
468 | else { | |
469 | push @enums, $this_extra; | |
470 | } | |
226b74db | 471 | } |
2027d365 | 472 | |
18072598 | 473 | @unused_enums = sort caselessly @unused_enums; |
2027d365 KW |
474 | $unused_enum_value = @enums; # All unused have the same value, |
475 | # one beyond the final used one | |
27a619f7 | 476 | } |
289ce9cc | 477 | |
2c490b6a KW |
478 | # These properties have extra tables written out for them that we want |
479 | # to make as compact and legible as possible. So we find short names | |
480 | # for their property values. For non-official ones we will need to | |
481 | # add a legend at the top of the table to say what the abbreviation | |
482 | # stands for. | |
483 | my $property_needs_table_re = qr/ ^ _Perl_ (?: GCB | LB | WB ) $ /x; | |
484 | ||
485 | my %short_enum_name; | |
486 | my %need_explanation; # For non-official abbreviations, we will need | |
487 | # to explain what the one we come up with | |
488 | # stands for | |
489 | my $type = lc $prop_name; | |
490 | if ($name =~ $property_needs_table_re) { | |
491 | my @short_names; # List of already used abbreviations, so we | |
492 | # don't duplicate | |
493 | for my $enum (@enums) { | |
494 | my $short_enum; | |
495 | my $is_official_name = 0; | |
496 | ||
497 | # Special case this wb property value to make the | |
498 | # name more clear | |
499 | if ($enum eq 'Perl_Tailored_HSpace') { | |
500 | $short_enum = 'hs'; | |
501 | } | |
502 | else { | |
503 | ||
504 | # Use the official short name, if found. | |
505 | ($short_enum) = prop_value_aliases($type, $enum); | |
506 | if ( defined $short_enum) { | |
507 | $is_official_name = 1; | |
508 | } | |
509 | else { | |
510 | # But if there is no official name, use the name that | |
511 | # came from the data (if any). Otherwise, the name | |
512 | # had to come from the extras list. There are two | |
513 | # types of values in that list. | |
514 | # | |
515 | # First are those enums that are not part of the | |
516 | # property, but are defined by the code in this file. | |
517 | # By convention these have all-caps names. We use the | |
518 | # lowercased name for these. | |
519 | # | |
520 | # Second are enums that are needed to get the | |
521 | # algorithms below to work and/or to get regexec.c to | |
522 | # compile, but don't exist in all Unicode releases. | |
523 | # These are handled outside this loop as | |
524 | # 'unused_enums' (as they are unused they all get | |
525 | # collapsed into a single column, and their names | |
526 | # don't matter) | |
527 | if (grep { $_ eq $enum } @input_enums) { | |
528 | $short_enum = $enum | |
529 | } | |
530 | else { | |
531 | $short_enum = lc $enum; | |
532 | } | |
533 | } | |
534 | ||
535 | # If our short name is too long, or we already know that | |
536 | # the name is an abbreviation, truncate to make sure it's | |
537 | # short enough, and remember that we did this so we can | |
538 | # later add a comment in the generated file | |
539 | if (length $short_enum > $max_hdr_len) { | |
540 | # First try using just the uppercase letters of the name; | |
541 | # if it is something like FooBar, FB is a better | |
542 | # abbreviation than Foo. That's not the case if it is | |
543 | # entirely lowercase. | |
544 | my $uc = $short_enum; | |
545 | $uc =~ s/[[:^upper:]]//g; | |
546 | $short_enum = $uc if length $uc > 1 | |
547 | && length $uc < length $short_enum; | |
548 | ||
549 | $short_enum = substr($short_enum, 0, $max_hdr_len); | |
550 | $is_official_name = 0; | |
551 | } | |
552 | } | |
553 | ||
554 | # If the name we are to display conflicts, try another. | |
555 | if (grep { $_ eq $short_enum } @short_names) { | |
556 | $is_official_name = 0; | |
557 | do { # The increment operator on strings doesn't work on | |
558 | # those containing an '_', so get rid of any final | |
559 | # portion. | |
560 | $short_enum =~ s/_//g; | |
561 | $short_enum++; | |
562 | } while grep { $_ eq $short_enum } @short_names; | |
563 | } | |
564 | ||
565 | push @short_names, $short_enum; | |
566 | $short_enum_name{$enum} = $short_enum; | |
567 | $need_explanation{$enum} = $short_enum unless $is_official_name; | |
568 | } | |
569 | } # End of calculating short enum names for certain properties | |
570 | ||
226b74db KW |
571 | # Assign a value to each element of the enum type we are creating. |
572 | # The default value always gets 0; the others are arbitrarily | |
2c490b6a KW |
573 | # assigned, but for the properties which have the extra table, it is |
574 | # in the order we have computed above so the rows and columns appear | |
575 | # alphabetically by heading abbreviation. | |
27a619f7 KW |
576 | my $enum_val = 0; |
577 | my $canonical_default = prop_value_aliases($prop_name, $default); | |
578 | $default = $canonical_default if defined $canonical_default; | |
579 | $enums{$default} = $enum_val++; | |
226b74db | 580 | |
2c490b6a KW |
581 | for my $enum (sort { ($name =~ $property_needs_table_re) |
582 | ? lc $short_enum_name{$a} | |
583 | cmp lc $short_enum_name{$b} | |
584 | : lc $a cmp lc $b | |
585 | } @enums) | |
586 | { | |
27a619f7 KW |
587 | $enums{$enum} = $enum_val++ unless exists $enums{$enum}; |
588 | } | |
589 | ||
2c490b6a KW |
590 | # Now calculate the data for the special tables output for these |
591 | # properties. | |
592 | if ($name =~ $property_needs_table_re) { | |
27a619f7 | 593 | |
226b74db KW |
594 | # The data includes the hashes %gcb_enums, %lb_enums, etc. |
595 | # Similarly we calculate column headings for the tables. | |
596 | # | |
27a619f7 | 597 | # We use string evals to allow the same code to work on |
226b74db | 598 | # all the tables |
27a619f7 | 599 | |
27a619f7 KW |
600 | # Skip if we've already done this code, which populated |
601 | # this hash | |
602 | if (eval "! \%${type}_enums") { | |
603 | ||
226b74db | 604 | # For each enum in the type ... |
2c490b6a | 605 | foreach my $enum (keys %enums) { |
27a619f7 | 606 | my $value = $enums{$enum}; |
2c490b6a | 607 | my $short_enum = $short_enum_name{$enum}; |
27a619f7 KW |
608 | |
609 | # Remember the mapping from the property value | |
610 | # (enum) name to its value. | |
611 | eval "\$${type}_enums{$enum} = $value"; | |
612 | die $@ if $@; | |
613 | ||
614 | # Remember the inverse mapping to the short name | |
615 | # so that we can properly label the generated | |
616 | # table's rows and columns | |
2c490b6a | 617 | eval "\$${type}_short_enums[$value] = '$short_enum'"; |
27a619f7 | 618 | die $@ if $@; |
2c490b6a KW |
619 | |
620 | # And note the abbreviations that need explanation | |
621 | if ($need_explanation{$enum}) { | |
622 | eval "\$${type}_abbreviations{$short_enum} = '$enum'"; | |
623 | die $@ if $@; | |
624 | } | |
7e54b87f | 625 | } |
2027d365 KW |
626 | |
627 | # Each unused enum has the same value. They all are collapsed | |
628 | # into one row and one column, named $unused_table_hdr. | |
629 | if (@unused_enums) { | |
630 | eval "\$${type}_short_enums['$unused_enum_value'] = '$unused_table_hdr'"; | |
631 | die $@ if $@; | |
632 | ||
633 | foreach my $enum (@unused_enums) { | |
634 | eval "\$${type}_enums{$enum} = $unused_enum_value"; | |
635 | die $@ if $@; | |
636 | } | |
637 | } | |
99f21fb9 | 638 | } |
19a5f1d5 | 639 | } |
99f21fb9 | 640 | |
cf2cd619 KW |
641 | # The short property names tend to be two lower case letters, but it |
642 | # looks better for those if they are upper. XXX | |
19a5f1d5 | 643 | $short_name = uc($short_name) if length($short_name) < 3 |
cf2cd619 | 644 | || substr($short_name, 0, 1) =~ /[[:lower:]]/; |
19a5f1d5 | 645 | $name_prefix = "${short_name}_"; |
cdc243dd | 646 | |
226b74db | 647 | # Start the enum definition for this map |
f99e0590 | 648 | my @enum_definition; |
19a5f1d5 KW |
649 | my @enum_list; |
650 | foreach my $enum (keys %enums) { | |
651 | $enum_list[$enums{$enum}] = $enum; | |
99f21fb9 | 652 | } |
19a5f1d5 | 653 | foreach my $i (0 .. @enum_list - 1) { |
f99e0590 | 654 | push @enum_definition, ",\n" if $i > 0; |
34623dbb | 655 | |
19a5f1d5 | 656 | my $name = $enum_list[$i]; |
f99e0590 | 657 | push @enum_definition, "\t${name_prefix}$name = $i"; |
19a5f1d5 | 658 | } |
2027d365 KW |
659 | if (@unused_enums) { |
660 | foreach my $unused (@unused_enums) { | |
661 | push @enum_definition, | |
662 | ",\n\t${name_prefix}$unused = $unused_enum_value"; | |
663 | } | |
664 | } | |
34623dbb | 665 | |
18230d9d | 666 | # For an 'l' property, we need extra enums, because some of the |
34623dbb KW |
667 | # elements are lists. Each such distinct list is placed in its own |
668 | # auxiliary map table. Here, we go through the inversion map, and for | |
669 | # each distinct list found, create an enum value for it, numbered -1, | |
670 | # -2, .... | |
671 | my %multiples; | |
672 | my $aux_table_prefix = "AUX_TABLE_"; | |
18230d9d | 673 | if ($input_format =~ /l/) { |
34623dbb KW |
674 | foreach my $element (@$invmap) { |
675 | ||
676 | # A regular scalar is not one of the lists we're looking for | |
677 | # at this stage. | |
678 | next unless ref $element; | |
679 | ||
18230d9d KW |
680 | my $joined; |
681 | if ($input_format =~ /a/) { # These are already ordered | |
682 | $joined = join ",", @$element; | |
683 | } | |
684 | else { | |
18072598 | 685 | $joined = join ",", sort caselessly @$element; |
18230d9d | 686 | } |
34623dbb KW |
687 | my $already_found = exists $multiples{$joined}; |
688 | ||
689 | my $i; | |
690 | if ($already_found) { # Use any existing one | |
691 | $i = $multiples{$joined}; | |
692 | } | |
693 | else { # Otherwise increment to get a new table number | |
694 | $i = keys(%multiples) + 1; | |
695 | $multiples{$joined} = $i; | |
696 | } | |
697 | ||
698 | # This changes the inversion map for this entry to not be the | |
699 | # list | |
700 | $element = "use_$aux_table_prefix$i"; | |
701 | ||
702 | # And add to the enum values | |
703 | if (! $already_found) { | |
f99e0590 | 704 | push @enum_definition, ",\n\t${name_prefix}$element = -$i"; |
34623dbb KW |
705 | } |
706 | } | |
707 | } | |
708 | ||
18230d9d | 709 | $enum_declaration_type = "${name_prefix}enum"; |
f99e0590 | 710 | |
c454388e KW |
711 | # Finished with the enum definition. Inversion map stuff is used only |
712 | # by regexec or utf-8 (if it is for code points) , unless it is in the | |
713 | # enum exception list | |
714 | my $where = (exists $where_to_define_enums{$name}) | |
715 | ? $where_to_define_enums{$name} | |
716 | : ($input_format =~ /a/) | |
717 | ? 'PERL_IN_UTF8_C' | |
718 | : 'PERL_IN_REGEXEC_C'; | |
719 | ||
8ec55631 KW |
720 | if (! exists $public_enums{$name}) { |
721 | switch_pound_if($name, $where, $charset); | |
722 | } | |
723 | else { | |
724 | end_charset_pound_if; | |
725 | end_file_pound_if; | |
726 | start_charset_pound_if($charset, 1); | |
727 | } | |
c454388e KW |
728 | |
729 | # If the enum only contains one element, that is a dummy, default one | |
f99e0590 KW |
730 | if (scalar @enum_definition > 1) { |
731 | ||
732 | # Currently unneeded | |
733 | #print $out_fh "\n#define ${name_prefix}ENUM_COUNT ", | |
734 | # ..scalar keys %enums, "\n"; | |
735 | ||
736 | if ($input_format =~ /l/) { | |
737 | print $out_fh | |
738 | "\n", | |
739 | "/* Negative enum values indicate the need to use an", | |
740 | " auxiliary table\n", | |
741 | " * consisting of the list of enums this one expands to.", | |
742 | " The absolute\n", | |
743 | " * values of the negative enums are indices into a table", | |
744 | " of the auxiliary\n", | |
745 | " * tables' addresses */"; | |
746 | } | |
747 | print $out_fh "\ntypedef enum {\n"; | |
748 | print $out_fh join "", @enum_definition; | |
749 | print $out_fh "\n"; | |
18230d9d | 750 | print $out_fh "} $enum_declaration_type;\n"; |
f99e0590 | 751 | } |
19a5f1d5 | 752 | |
8ec55631 | 753 | switch_pound_if($name, $where, $charset); |
d74e7480 | 754 | |
40d2776f KW |
755 | # The inversion lists here have to be UV because inversion lists are |
756 | # capable of storing any code point, and even though the the ones here | |
757 | # are only Unicode ones, which need just 21 bits, they are linked to | |
758 | # directly, rather than copied. The inversion map and aux tables also | |
759 | # only need be 21 bits, and so we can get away with declaring them | |
760 | # 32-bits to save a little space and memory (on some 64-bit | |
761 | # platforms), as they are copied. | |
18230d9d KW |
762 | $invmap_declaration_type = ($input_format =~ /s/) |
763 | ? $enum_declaration_type | |
40d2776f | 764 | : "I32"; |
18230d9d KW |
765 | $aux_declaration_type = ($input_format =~ /s/) |
766 | ? $enum_declaration_type | |
40d2776f | 767 | : "U32"; |
18230d9d | 768 | |
19a5f1d5 | 769 | $output_format = "${name_prefix}%s"; |
34623dbb KW |
770 | |
771 | # If there are auxiliary tables, output them. | |
772 | if (%multiples) { | |
773 | ||
774 | print $out_fh "\n#define HAS_${name_prefix}AUX_TABLES\n"; | |
775 | ||
776 | # Invert keys and values | |
777 | my %inverted_mults; | |
778 | while (my ($key, $value) = each %multiples) { | |
779 | $inverted_mults{$value} = $key; | |
780 | } | |
781 | ||
782 | # Output them in sorted order | |
783 | my @sorted_table_list = sort { $a <=> $b } keys %inverted_mults; | |
784 | ||
785 | # Keep track of how big each aux table is | |
786 | my @aux_counts; | |
787 | ||
788 | # Output each aux table. | |
789 | foreach my $table_number (@sorted_table_list) { | |
790 | my $table = $inverted_mults{$table_number}; | |
cef72199 | 791 | output_table_header($out_fh, |
cf2cd619 KW |
792 | $aux_declaration_type, |
793 | "$name_prefix$aux_table_prefix$table_number"); | |
34623dbb | 794 | |
cf2cd619 KW |
795 | # Earlier, we joined the elements of this table together with |
796 | # a comma | |
34623dbb KW |
797 | my @elements = split ",", $table; |
798 | ||
799 | $aux_counts[$table_number] = scalar @elements; | |
800 | for my $i (0 .. @elements - 1) { | |
801 | print $out_fh ",\n" if $i > 0; | |
18230d9d KW |
802 | if ($input_format =~ /a/) { |
803 | printf $out_fh "\t0x%X", $elements[$i]; | |
804 | } | |
805 | else { | |
806 | print $out_fh "\t${name_prefix}$elements[$i]"; | |
807 | } | |
34623dbb | 808 | } |
cef72199 KW |
809 | |
810 | print $out_fh "\n"; | |
811 | output_table_trailer(); | |
34623dbb KW |
812 | } |
813 | ||
814 | # Output the table that is indexed by the absolute value of the | |
815 | # aux table enum and contains pointers to the tables output just | |
816 | # above | |
cef72199 KW |
817 | output_table_header($out_fh, "$aux_declaration_type *", |
818 | "${name_prefix}${aux_table_prefix}ptrs"); | |
34623dbb KW |
819 | print $out_fh "\tNULL,\t/* Placeholder */\n"; |
820 | for my $i (1 .. @sorted_table_list) { | |
821 | print $out_fh ",\n" if $i > 1; | |
822 | print $out_fh "\t$name_prefix$aux_table_prefix$i"; | |
823 | } | |
cef72199 KW |
824 | print $out_fh "\n"; |
825 | output_table_trailer(); | |
34623dbb KW |
826 | |
827 | print $out_fh | |
828 | "\n/* Parallel table to the above, giving the number of elements" | |
829 | . " in each table\n * pointed to */\n"; | |
cef72199 KW |
830 | output_table_header($out_fh, "U8", |
831 | "${name_prefix}${aux_table_prefix}lengths"); | |
34623dbb KW |
832 | print $out_fh "\t0,\t/* Placeholder */\n"; |
833 | for my $i (1 .. @sorted_table_list) { | |
cf2cd619 KW |
834 | print $out_fh ",\n" if $i > 1; |
835 | print $out_fh | |
836 | "\t$aux_counts[$i]\t/* $name_prefix$aux_table_prefix$i */"; | |
34623dbb | 837 | } |
cef72199 KW |
838 | print $out_fh "\n"; |
839 | output_table_trailer(); | |
34623dbb | 840 | } # End of outputting the auxiliary and associated tables |
463b4a67 KW |
841 | |
842 | # The scx property used in regexec.c needs a specialized table which | |
843 | # is most convenient to output here, while the data structures set up | |
844 | # above are still extant. This table contains the code point that is | |
845 | # the zero digit of each script, indexed by script enum value. | |
846 | if (lc $short_name eq 'scx') { | |
847 | my @decimals_invlist = prop_invlist("Numeric_Type=Decimal"); | |
848 | my %script_zeros; | |
849 | ||
850 | # Find all the decimal digits. The 0 of each range is always the | |
851 | # 0th element, except in some early Unicode releases, so check for | |
852 | # that. | |
853 | for (my $i = 0; $i < @decimals_invlist; $i += 2) { | |
854 | my $code_point = $decimals_invlist[$i]; | |
a1c8344d | 855 | next if num(chr($code_point)) ne '0'; |
463b4a67 KW |
856 | |
857 | # Turn the scripts this zero is in into a list. | |
858 | my @scripts = split ",", | |
859 | charprop($code_point, "_Perl_SCX", '_perl_core_internal_ok'); | |
860 | $code_point = sprintf("0x%x", $code_point); | |
861 | ||
862 | foreach my $script (@scripts) { | |
863 | if (! exists $script_zeros{$script}) { | |
864 | $script_zeros{$script} = $code_point; | |
865 | } | |
866 | elsif (ref $script_zeros{$script}) { | |
867 | push $script_zeros{$script}->@*, $code_point; | |
868 | } | |
869 | else { # Turn into a list if this is the 2nd zero of the | |
870 | # script | |
871 | my $existing = $script_zeros{$script}; | |
872 | undef $script_zeros{$script}; | |
873 | push $script_zeros{$script}->@*, $existing, $code_point; | |
874 | } | |
875 | } | |
876 | } | |
877 | ||
878 | # @script_zeros contains the zero, sorted by the script's enum | |
879 | # value | |
880 | my @script_zeros; | |
881 | foreach my $script (keys %script_zeros) { | |
882 | my $enum_value = $enums{$script}; | |
883 | $script_zeros[$enum_value] = $script_zeros{$script}; | |
884 | } | |
885 | ||
886 | print $out_fh | |
887 | "\n/* This table, indexed by the script enum, gives the zero" | |
888 | . " code point for that\n * script; 0 if the script has multiple" | |
889 | . " digit sequences. Scripts without a\n * digit sequence use" | |
890 | . " ASCII [0-9], hence are marked '0' */\n"; | |
cef72199 | 891 | output_table_header($out_fh, "UV", "script_zeros"); |
463b4a67 KW |
892 | for my $i (0 .. @script_zeros - 1) { |
893 | my $code_point = $script_zeros[$i]; | |
894 | if (defined $code_point) { | |
895 | $code_point = " 0" if ref $code_point; | |
896 | print $out_fh "\t$code_point"; | |
897 | } | |
898 | elsif (lc $enum_list[$i] eq 'inherited') { | |
899 | print $out_fh "\t 0"; | |
900 | } | |
901 | else { # The only digits a script without its own set accepts | |
902 | # is [0-9] | |
903 | print $out_fh "\t'0'"; | |
904 | } | |
905 | print $out_fh "," if $i < @script_zeros - 1; | |
906 | print $out_fh "\t/* $enum_list[$i] */"; | |
907 | print $out_fh "\n"; | |
908 | } | |
cef72199 | 909 | output_table_trailer(); |
463b4a67 | 910 | } # End of special handling of scx |
99f21fb9 KW |
911 | } |
912 | else { | |
913 | die "'$input_format' invmap() format for '$prop_name' unimplemented"; | |
914 | } | |
915 | ||
916 | die "No inversion map for $prop_name" unless defined $invmap | |
917 | && ref $invmap eq 'ARRAY' | |
918 | && $count; | |
919 | ||
226b74db | 920 | # Now output the inversion map proper |
cef72199 KW |
921 | $charset = "for $charset" if $charset; |
922 | output_table_header($out_fh, $invmap_declaration_type, | |
923 | "${name}_invmap", | |
924 | $charset); | |
99f21fb9 KW |
925 | |
926 | # The main body are the scalars passed in to this routine. | |
927 | for my $i (0 .. $count - 1) { | |
928 | my $element = $invmap->[$i]; | |
02f811dd | 929 | my $full_element_name = prop_value_aliases($prop_name, $element); |
18230d9d KW |
930 | if ($input_format =~ /a/ && $element !~ /\D/) { |
931 | $element = ($element == 0) | |
932 | ? 0 | |
933 | : sprintf("0x%X", $element); | |
934 | } | |
935 | else { | |
02f811dd KW |
936 | $element = $full_element_name if defined $full_element_name; |
937 | $element = $name_prefix . $element; | |
18230d9d | 938 | } |
99f21fb9 KW |
939 | print $out_fh "\t$element"; |
940 | print $out_fh "," if $i < $count - 1; | |
941 | print $out_fh "\n"; | |
942 | } | |
cef72199 | 943 | output_table_trailer(); |
99f21fb9 KW |
944 | } |
945 | ||
5a7e5385 | 946 | sub mk_invlist_from_sorted_cp_list { |
a02047bf KW |
947 | |
948 | # Returns an inversion list constructed from the sorted input array of | |
949 | # code points | |
950 | ||
951 | my $list_ref = shift; | |
952 | ||
99f21fb9 KW |
953 | return unless @$list_ref; |
954 | ||
a02047bf KW |
955 | # Initialize to just the first element |
956 | my @invlist = ( $list_ref->[0], $list_ref->[0] + 1); | |
957 | ||
958 | # For each succeeding element, if it extends the previous range, adjust | |
959 | # up, otherwise add it. | |
960 | for my $i (1 .. @$list_ref - 1) { | |
961 | if ($invlist[-1] == $list_ref->[$i]) { | |
962 | $invlist[-1]++; | |
963 | } | |
964 | else { | |
965 | push @invlist, $list_ref->[$i], $list_ref->[$i] + 1; | |
966 | } | |
967 | } | |
968 | return @invlist; | |
969 | } | |
970 | ||
971 | # Read in the Case Folding rules, and construct arrays of code points for the | |
972 | # properties we need. | |
d2aadf62 | 973 | my ($cp_ref, $folds_ref, $format, $default) = prop_invmap("Case_Folding"); |
a02047bf KW |
974 | die "Could not find inversion map for Case_Folding" unless defined $format; |
975 | die "Incorrect format '$format' for Case_Folding inversion map" | |
347b9066 KW |
976 | unless $format eq 'al' |
977 | || $format eq 'a'; | |
d2aadf62 KW |
978 | sub _Perl_IVCF { |
979 | ||
980 | # This creates a map of the inversion of case folding. i.e., given a | |
981 | # character, it gives all the other characters that fold to it. | |
982 | # | |
983 | # Inversion maps function kind of like a hash, with the inversion list | |
984 | # specifying the buckets (keys) and the inversion maps specifying the | |
985 | # contents of the corresponding bucket. Effectively this function just | |
986 | # swaps the keys and values of the case fold hash. But there are | |
987 | # complications. Most importantly, More than one character can each have | |
988 | # the same fold. This is solved by having a list of characters that fold | |
989 | # to a given one. | |
990 | ||
991 | my %new; | |
992 | ||
993 | # Go through the inversion list. | |
994 | for (my $i = 0; $i < @$cp_ref; $i++) { | |
995 | ||
996 | # Skip if nothing folds to this | |
997 | next if $folds_ref->[$i] == 0; | |
998 | ||
999 | # This entry which is valid from here to up (but not including) the | |
1000 | # next entry is for the next $count characters, so that, for example, | |
1001 | # A-Z is represented by one entry. | |
1002 | my $cur_list = $cp_ref->[$i]; | |
1003 | my $count = $cp_ref->[$i+1] - $cur_list; | |
1004 | ||
1005 | # The fold of [$i] can be not just a single character, but a sequence | |
1006 | # of multiple ones. We deal with those here by just creating a string | |
1007 | # consisting of them. Otherwise, we use the single code point [$i] | |
1008 | # folds to. | |
1009 | my $cur_map = (ref $folds_ref->[$i]) | |
1010 | ? join "", map { chr } $folds_ref->[$i]->@* | |
1011 | : $folds_ref->[$i]; | |
1012 | ||
1013 | # Expand out this range | |
1014 | while ($count > 0) { | |
1015 | push @{$new{$cur_map}}, $cur_list; | |
1016 | ||
1017 | # A multiple-character fold is a string, and shouldn't need | |
1018 | # incrementing anyway | |
1019 | if (ref $folds_ref->[$i]) { | |
1020 | die sprintf("Case fold for %x is multiple chars; should have" | |
1021 | . " a count of 1, but instead it was $count", $count) | |
1022 | unless $count == 1; | |
1023 | } | |
1024 | else { | |
1025 | $cur_map++; | |
1026 | $cur_list++; | |
1027 | } | |
1028 | $count--; | |
1029 | } | |
1030 | } | |
1031 | ||
1032 | # Now go through and make some adjustments. We add synthetic entries for | |
59142b8b KW |
1033 | # three cases. |
1034 | # 1) If the fold of a Latin1-range character is above that range, some | |
1035 | # coding in regexec.c can be saved by creating a reverse map here. The | |
1036 | # impetus for this is that U+B5 (MICRO SIGN) folds to the Greek small | |
1037 | # mu (U+3BC). That fold isn't done at regex pattern compilation time | |
1038 | # if it means that the pattern would have to be translated into UTF-8, | |
1039 | # whose operation is slower. At run time, having this reverse | |
1040 | # translation eliminates some special cases in the code. | |
1041 | # 2) Two or more code points can fold to the same multiple character, | |
d2aadf62 KW |
1042 | # sequence, as U+FB05 and U+FB06 both fold to 'st'. This code is only |
1043 | # for single character folds, but FB05 and FB06 are single characters | |
1044 | # that are equivalent folded, so we add entries so that they are | |
1045 | # considered to fold to each other | |
59142b8b | 1046 | # 3) If two or more above-Latin1 code points fold to the same Latin1 range |
d2aadf62 KW |
1047 | # one, we also add entries so that they are considered to fold to each |
1048 | # other. This is so that under /aa or /l matching, where folding to | |
1049 | # their Latin1 range code point is illegal, they still can fold to each | |
1050 | # other. This situation happens in Unicode 3.0.1, but probably no | |
1051 | # other version. | |
1052 | foreach my $fold (keys %new) { | |
db95f459 | 1053 | my $folds_to_string = $fold =~ /\D/; |
d2aadf62 KW |
1054 | |
1055 | # If the bucket contains only one element, convert from an array to a | |
1056 | # scalar | |
1057 | if (scalar $new{$fold}->@* == 1) { | |
1058 | $new{$fold} = $new{$fold}[0]; | |
59142b8b KW |
1059 | |
1060 | # Handle case 1) above: if there were a Latin1 range code point | |
1061 | # whose fold is above that range, this creates an extra entry that | |
1062 | # maps the other direction, and would save some special case code. | |
1063 | # (The one current case of this is handled in the else clause | |
1064 | # below.) | |
1065 | $new{$new{$fold}} = $fold if $new{$fold} < 256 && $fold > 255; | |
d2aadf62 KW |
1066 | } |
1067 | else { | |
1068 | ||
59142b8b KW |
1069 | # Handle case 1) when there are multiple things that fold to an |
1070 | # above-Latin1 code point, at least one of which is in Latin1. | |
1071 | if (! $folds_to_string && $fold > 255) { | |
1072 | foreach my $cp ($new{$fold}->@*) { | |
1073 | if ($cp < 256) { | |
1074 | my @new_entry = grep { $_ != $cp } $new{$fold}->@*; | |
1075 | push @new_entry, $fold; | |
1076 | $new{$cp}->@* = @new_entry; | |
1077 | } | |
1078 | } | |
1079 | } | |
1080 | ||
d2aadf62 KW |
1081 | # Otherwise, sort numerically. This places the highest code point |
1082 | # in the list at the tail end. This is because Unicode keeps the | |
1083 | # lowercase code points as higher ordinals than the uppercase, at | |
1084 | # least for the ones that matter so far. These are synthetic | |
1085 | # entries, and we want to predictably have the lowercase (which is | |
1086 | # more likely to be what gets folded to) in the same corresponding | |
1087 | # position, so that other code can rely on that. If some new | |
1088 | # version of Unicode came along that violated this, we might have | |
1089 | # to change so that the sort is based on upper vs lower instead. | |
1090 | # (The lower-comes-after isn't true of native EBCDIC, but here we | |
1091 | # are dealing strictly with Unicode values). | |
1092 | @{$new{$fold}} = sort { $a <=> $b } $new{$fold}->@* | |
1093 | unless $folds_to_string; | |
1094 | # We will be working with a copy of this sorted entry. | |
1095 | my @source_list = $new{$fold}->@*; | |
1096 | if (! $folds_to_string) { | |
1097 | ||
1098 | # This handles situation 2) listed above, which only arises if | |
1099 | # what is being folded-to (the fold) is in the Latin1 range. | |
1100 | if ($fold > 255 ) { | |
1101 | undef @source_list; | |
1102 | } | |
1103 | else { | |
1104 | # And it only arises if there are two or more folders that | |
1105 | # fold to it above Latin1. We look at just those. | |
1106 | @source_list = grep { $_ > 255 } @source_list; | |
1107 | undef @source_list if @source_list == 1; | |
1108 | } | |
1109 | } | |
1110 | ||
1111 | # Here, we've found the items we want to set up synthetic folds | |
1112 | # for. Add entries so that each folds to each other. | |
1113 | foreach my $cp (@source_list) { | |
1114 | my @rest = grep { $cp != $_ } @source_list; | |
1115 | if (@rest == 1) { | |
1116 | $new{$cp} = $rest[0]; | |
1117 | } | |
1118 | else { | |
1119 | push @{$new{$cp}}, @rest; | |
1120 | } | |
1121 | } | |
1122 | } | |
1123 | ||
1124 | # We don't otherwise deal with multiple-character folds | |
1125 | delete $new{$fold} if $folds_to_string; | |
1126 | } | |
1127 | ||
1128 | ||
1129 | # Now we have a hash that is the inversion of the case fold property. | |
cb2d98ed KW |
1130 | # First find the maximum number of code points that fold to the same one. |
1131 | foreach my $fold_to (keys %new) { | |
1132 | if (ref $new{$fold_to}) { | |
1133 | my $folders_count = scalar @{$new{$fold_to}}; | |
1134 | $max_fold_froms = $folders_count if $folders_count > $max_fold_froms; | |
1135 | } | |
1136 | } | |
d2aadf62 | 1137 | |
cb2d98ed | 1138 | # Then convert the hash to an inversion map. |
d2aadf62 KW |
1139 | my @sorted_folds = sort { $a <=> $b } keys %new; |
1140 | my (@invlist, @invmap); | |
1141 | ||
1142 | # We know that nothing folds to the controls (whose ordinals start at 0). | |
1143 | # And the first real entries are the lowest in the hash. | |
1144 | push @invlist, 0, $sorted_folds[0]; | |
1145 | push @invmap, 0, $new{$sorted_folds[0]}; | |
1146 | ||
1147 | # Go through the remainder of the hash keys (which are the folded code | |
1148 | # points) | |
1149 | for (my $i = 1; $i < @sorted_folds; $i++) { | |
1150 | ||
1151 | # Get the current one, and the one prior to it. | |
1152 | my $fold = $sorted_folds[$i]; | |
1153 | my $prev_fold = $sorted_folds[$i-1]; | |
1154 | ||
1155 | # If the current one is not just 1 away from the prior one, we close | |
1156 | # out the range containing the previous fold, and know that the gap | |
1157 | # doesn't have anything that folds. | |
1158 | if ($fold - 1 != $prev_fold) { | |
1159 | push @invlist, $prev_fold + 1; | |
1160 | push @invmap, 0; | |
1161 | ||
1162 | # And start a new range | |
1163 | push @invlist, $fold; | |
1164 | push @invmap, $new{$fold}; | |
1165 | } | |
1166 | elsif ($new{$fold} - 1 != $new{$prev_fold}) { | |
1167 | ||
1168 | # Here the current fold is just 1 greater than the previous, but | |
1169 | # the new map isn't correspondingly 1 greater than the previous, | |
1170 | # the old range is ended, but since there is no gap, we don't have | |
1171 | # to insert anything else. | |
1172 | push @invlist, $fold; | |
1173 | push @invmap, $new{$fold}; | |
1174 | ||
1175 | } # else { Otherwise, this new entry just extends the previous } | |
1176 | ||
1177 | die "In IVCF: $invlist[-1] <= $invlist[-2]" | |
1178 | if $invlist[-1] <= $invlist[-2]; | |
1179 | } | |
1180 | ||
1181 | # And add an entry that indicates that everything above this, to infinity, | |
1182 | # does not have a case fold. | |
1183 | push @invlist, $sorted_folds[-1] + 1; | |
1184 | push @invmap, 0; | |
1185 | ||
9a9a3246 KW |
1186 | push @invlist, 0x110000; |
1187 | push @invmap, 0; | |
1188 | ||
d2aadf62 KW |
1189 | # All Unicode versions have some places where multiple code points map to |
1190 | # the same one, so the format always has an 'l' | |
1191 | return \@invlist, \@invmap, 'al', $default; | |
1192 | } | |
1193 | ||
99f21fb9 KW |
1194 | sub prop_name_for_cmp ($) { # Sort helper |
1195 | my $name = shift; | |
1196 | ||
1197 | # Returns the input lowercased, with non-alphas removed, as well as | |
1198 | # everything starting with a comma | |
1199 | ||
1200 | $name =~ s/,.*//; | |
1201 | $name =~ s/[[:^alpha:]]//g; | |
1202 | return lc $name; | |
1203 | } | |
1204 | ||
892d8259 | 1205 | sub UpperLatin1 { |
8843f0de KW |
1206 | my @return = mk_invlist_from_sorted_cp_list([ 128 .. 255 ]); |
1207 | return \@return; | |
892d8259 KW |
1208 | } |
1209 | ||
a2aeff50 KW |
1210 | sub _Perl_CCC_non0_non230 { |
1211 | ||
1212 | # Create an inversion list of code points with non-zero canonical | |
1213 | # combining class that also don't have 230 as the class number. This is | |
1214 | # part of a Unicode Standard rule | |
1215 | ||
1216 | my @nonzeros = prop_invlist("ccc=0"); | |
1217 | shift @nonzeros; # Invert so is "ccc != 0" | |
1218 | ||
1219 | my @return; | |
1220 | ||
1221 | # Expand into list of code points, while excluding those with ccc == 230 | |
1222 | for (my $i = 0; $i < @nonzeros; $i += 2) { | |
1223 | my $upper = ($i + 1) < @nonzeros | |
1224 | ? $nonzeros[$i+1] - 1 # In range | |
1225 | : $Unicode::UCD::MAX_CP; # To infinity. | |
1226 | for my $j ($nonzeros[$i] .. $upper) { | |
1227 | my @ccc_names = prop_value_aliases("ccc", charprop($j, "ccc")); | |
1228 | ||
1229 | # Final element in @ccc_names will be all numeric | |
1230 | push @return, $j if $ccc_names[-1] != 230; | |
1231 | } | |
1232 | } | |
1233 | ||
1234 | @return = sort { $a <=> $b } @return; | |
1235 | @return = mk_invlist_from_sorted_cp_list(\@return); | |
1236 | return \@return; | |
1237 | } | |
1238 | ||
289ce9cc KW |
1239 | sub output_table_common { |
1240 | ||
1241 | # Common subroutine to actually output the generated rules table. | |
1242 | ||
1243 | my ($property, | |
1244 | $table_value_defines_ref, | |
1245 | $table_ref, | |
1246 | $names_ref, | |
1247 | $abbreviations_ref) = @_; | |
1248 | my $size = @$table_ref; | |
1249 | ||
1250 | # Output the #define list, sorted by numeric value | |
1251 | if ($table_value_defines_ref) { | |
1252 | my $max_name_length = 0; | |
1253 | my @defines; | |
1254 | ||
1255 | # Put in order, and at the same time find the longest name | |
1256 | while (my ($enum, $value) = each %$table_value_defines_ref) { | |
1257 | $defines[$value] = $enum; | |
1258 | ||
1259 | my $length = length $enum; | |
1260 | $max_name_length = $length if $length > $max_name_length; | |
1261 | } | |
1262 | ||
1263 | print $out_fh "\n"; | |
1264 | ||
1265 | # Output, so that the values are vertically aligned in a column after | |
1266 | # the longest name | |
1267 | foreach my $i (0 .. @defines - 1) { | |
1268 | next unless defined $defines[$i]; | |
1269 | printf $out_fh "#define %-*s %2d\n", | |
1270 | $max_name_length, | |
1271 | $defines[$i], | |
1272 | $i; | |
1273 | } | |
1274 | } | |
1275 | ||
1276 | my $column_width = 2; # We currently allow 2 digits for the number | |
1277 | ||
a57a2641 KW |
1278 | # Being above a U8 is not currently handled |
1279 | my $table_type = 'U8'; | |
289ce9cc KW |
1280 | |
1281 | # If a name is longer than the width set aside for a column, its column | |
1282 | # needs to have increased spacing so that the name doesn't get truncated | |
1283 | # nor run into an adjacent column | |
1284 | my @spacers; | |
1285 | ||
2027d365 KW |
1286 | # Is there a row and column for unused values in this release? |
1287 | my $has_unused = $names_ref->[$size-1] eq $unused_table_hdr; | |
289ce9cc KW |
1288 | |
1289 | for my $i (0 .. $size - 1) { | |
1290 | no warnings 'numeric'; | |
289ce9cc KW |
1291 | $spacers[$i] = " " x (length($names_ref->[$i]) - $column_width); |
1292 | } | |
1293 | ||
cf2cd619 KW |
1294 | output_table_header($out_fh, $table_type, "${property}_table", undef, |
1295 | $size, $size); | |
289ce9cc KW |
1296 | |
1297 | # Calculate the column heading line | |
1298 | my $header_line = "/* " | |
1299 | . (" " x $max_hdr_len) # We let the row heading meld to | |
1300 | # the '*/' for those that are at | |
1301 | # the max | |
1302 | . " " x 3; # Space for '*/ ' | |
1303 | # Now each column | |
1304 | for my $i (0 .. $size - 1) { | |
1305 | $header_line .= sprintf "%s%*s", | |
1306 | $spacers[$i], | |
1307 | $column_width + 1, # 1 for the ',' | |
1308 | $names_ref->[$i]; | |
1309 | } | |
1310 | $header_line .= " */\n"; | |
1311 | ||
1312 | # If we have annotations, output it now. | |
2027d365 | 1313 | if ($has_unused || scalar %$abbreviations_ref) { |
289ce9cc | 1314 | my $text = ""; |
18072598 | 1315 | foreach my $abbr (sort caselessly keys %$abbreviations_ref) { |
289ce9cc KW |
1316 | $text .= "; " if $text; |
1317 | $text .= "'$abbr' stands for '$abbreviations_ref->{$abbr}'"; | |
1318 | } | |
2027d365 KW |
1319 | if ($has_unused) { |
1320 | $text .= "; $unused_table_hdr stands for 'unused in this Unicode" | |
0ebd57bb | 1321 | . " release (and the data in its row and column are garbage)" |
289ce9cc KW |
1322 | } |
1323 | ||
1324 | my $indent = " " x 3; | |
1325 | $text = $indent . "/* $text */"; | |
1326 | ||
1327 | # Wrap the text so that it is no wider than the table, which the | |
1328 | # header line gives. | |
1329 | my $output_width = length $header_line; | |
1330 | while (length $text > $output_width) { | |
1331 | my $cur_line = substr($text, 0, $output_width); | |
1332 | ||
1333 | # Find the first blank back from the right end to wrap at. | |
1334 | for (my $i = $output_width -1; $i > 0; $i--) { | |
1335 | if (substr($text, $i, 1) eq " ") { | |
1336 | print $out_fh substr($text, 0, $i), "\n"; | |
1337 | ||
1338 | # Set so will look at just the remaining tail (which will | |
1339 | # be indented and have a '*' after the indent | |
1340 | $text = $indent . " * " . substr($text, $i + 1); | |
1341 | last; | |
1342 | } | |
1343 | } | |
1344 | } | |
1345 | ||
1346 | # And any remaining | |
1347 | print $out_fh $text, "\n" if $text; | |
1348 | } | |
1349 | ||
1350 | # We calculated the header line earlier just to get its width so that we | |
1351 | # could make sure the annotations fit into that. | |
1352 | print $out_fh $header_line; | |
1353 | ||
1354 | # Now output the bulk of the table. | |
1355 | for my $i (0 .. $size - 1) { | |
1356 | ||
1357 | # First the row heading. | |
1358 | printf $out_fh "/* %-*s*/ ", $max_hdr_len, $names_ref->[$i]; | |
1359 | print $out_fh "{"; # Then the brace for this row | |
1360 | ||
1361 | # Then each column | |
1362 | for my $j (0 .. $size -1) { | |
1363 | print $out_fh $spacers[$j]; | |
1364 | printf $out_fh "%*d", $column_width, $table_ref->[$i][$j]; | |
1365 | print $out_fh "," if $j < $size - 1; | |
1366 | } | |
1367 | print $out_fh " }"; | |
1368 | print $out_fh "," if $i < $size - 1; | |
1369 | print $out_fh "\n"; | |
1370 | } | |
1371 | ||
cef72199 | 1372 | output_table_trailer(); |
289ce9cc KW |
1373 | } |
1374 | ||
973a28ed KW |
1375 | sub output_GCB_table() { |
1376 | ||
1377 | # Create and output the pair table for use in determining Grapheme Cluster | |
1378 | # Breaks, given in http://www.unicode.org/reports/tr29/. | |
b0e24409 KW |
1379 | my %gcb_actions = ( |
1380 | GCB_NOBREAK => 0, | |
1381 | GCB_BREAKABLE => 1, | |
1382 | GCB_RI_then_RI => 2, # Rules 12 and 13 | |
1383 | GCB_EX_then_EM => 3, # Rule 10 | |
c0734505 | 1384 | GCB_Maybe_Emoji_NonBreak => 4, |
b0e24409 | 1385 | ); |
973a28ed KW |
1386 | |
1387 | # The table is constructed in reverse order of the rules, to make the | |
1388 | # lower-numbered, higher priority ones override the later ones, as the | |
1389 | # algorithm stops at the earliest matching rule | |
1390 | ||
1391 | my @gcb_table; | |
1392 | my $table_size = @gcb_short_enums; | |
1393 | ||
1394 | # Otherwise, break everywhere. | |
b0e24409 | 1395 | # GB99 Any ÷ Any |
973a28ed KW |
1396 | for my $i (0 .. $table_size - 1) { |
1397 | for my $j (0 .. $table_size - 1) { | |
1398 | $gcb_table[$i][$j] = 1; | |
1399 | } | |
1400 | } | |
1401 | ||
b0e24409 KW |
1402 | # Do not break within emoji flag sequences. That is, do not break between |
1403 | # regional indicator (RI) symbols if there is an odd number of RI | |
1404 | # characters before the break point. Must be resolved in runtime code. | |
1405 | # | |
c492f156 | 1406 | # GB12 sot (RI RI)* RI × RI |
b0e24409 KW |
1407 | # GB13 [^RI] (RI RI)* RI × RI |
1408 | $gcb_table[$gcb_enums{'Regional_Indicator'}] | |
1409 | [$gcb_enums{'Regional_Indicator'}] = $gcb_actions{GCB_RI_then_RI}; | |
1410 | ||
c0734505 KW |
1411 | # Post 11.0: GB11 \p{Extended_Pictographic} Extend* ZWJ |
1412 | # × \p{Extended_Pictographic} | |
a9256a75 | 1413 | $gcb_table[$gcb_enums{'ZWJ'}][$gcb_enums{'ExtPict_XX'}] = |
c0734505 KW |
1414 | $gcb_actions{GCB_Maybe_Emoji_NonBreak}; |
1415 | ||
1416 | # This and the rule GB10 obsolete starting with Unicode 11.0, can be left | |
1417 | # in as there are no code points that match, so the code won't ever get | |
1418 | # executed. | |
b0e24409 | 1419 | # Do not break within emoji modifier sequences or emoji zwj sequences. |
c0734505 | 1420 | # Pre 11.0: GB11 ZWJ × ( Glue_After_Zwj | E_Base_GAZ ) |
b0e24409 KW |
1421 | $gcb_table[$gcb_enums{'ZWJ'}][$gcb_enums{'Glue_After_Zwj'}] = 0; |
1422 | $gcb_table[$gcb_enums{'ZWJ'}][$gcb_enums{'E_Base_GAZ'}] = 0; | |
1423 | ||
1424 | # GB10 ( E_Base | E_Base_GAZ ) Extend* × E_Modifier | |
1425 | $gcb_table[$gcb_enums{'Extend'}][$gcb_enums{'E_Modifier'}] | |
1426 | = $gcb_actions{GCB_EX_then_EM}; | |
1427 | $gcb_table[$gcb_enums{'E_Base'}][$gcb_enums{'E_Modifier'}] = 0; | |
1428 | $gcb_table[$gcb_enums{'E_Base_GAZ'}][$gcb_enums{'E_Modifier'}] = 0; | |
1429 | ||
1430 | # Do not break before extending characters or ZWJ. | |
973a28ed | 1431 | # Do not break before SpacingMarks, or after Prepend characters. |
973a28ed | 1432 | # GB9b Prepend × |
b0e24409 KW |
1433 | # GB9a × SpacingMark |
1434 | # GB9 × ( Extend | ZWJ ) | |
973a28ed | 1435 | for my $i (0 .. @gcb_table - 1) { |
289ce9cc | 1436 | $gcb_table[$gcb_enums{'Prepend'}][$i] = 0; |
b0e24409 KW |
1437 | $gcb_table[$i][$gcb_enums{'SpacingMark'}] = 0; |
1438 | $gcb_table[$i][$gcb_enums{'Extend'}] = 0; | |
1439 | $gcb_table[$i][$gcb_enums{'ZWJ'}] = 0; | |
973a28ed KW |
1440 | } |
1441 | ||
973a28ed KW |
1442 | # Do not break Hangul syllable sequences. |
1443 | # GB8 ( LVT | T) × T | |
1444 | $gcb_table[$gcb_enums{'LVT'}][$gcb_enums{'T'}] = 0; | |
1445 | $gcb_table[$gcb_enums{'T'}][$gcb_enums{'T'}] = 0; | |
1446 | ||
1447 | # GB7 ( LV | V ) × ( V | T ) | |
1448 | $gcb_table[$gcb_enums{'LV'}][$gcb_enums{'V'}] = 0; | |
1449 | $gcb_table[$gcb_enums{'LV'}][$gcb_enums{'T'}] = 0; | |
1450 | $gcb_table[$gcb_enums{'V'}][$gcb_enums{'V'}] = 0; | |
1451 | $gcb_table[$gcb_enums{'V'}][$gcb_enums{'T'}] = 0; | |
1452 | ||
1453 | # GB6 L × ( L | V | LV | LVT ) | |
1454 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'L'}] = 0; | |
1455 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'V'}] = 0; | |
1456 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'LV'}] = 0; | |
1457 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'LVT'}] = 0; | |
1458 | ||
289ce9cc KW |
1459 | # Do not break between a CR and LF. Otherwise, break before and after |
1460 | # controls. | |
973a28ed KW |
1461 | # GB5 ÷ ( Control | CR | LF ) |
1462 | # GB4 ( Control | CR | LF ) ÷ | |
1463 | for my $i (0 .. @gcb_table - 1) { | |
289ce9cc | 1464 | $gcb_table[$i][$gcb_enums{'Control'}] = 1; |
973a28ed KW |
1465 | $gcb_table[$i][$gcb_enums{'CR'}] = 1; |
1466 | $gcb_table[$i][$gcb_enums{'LF'}] = 1; | |
289ce9cc | 1467 | $gcb_table[$gcb_enums{'Control'}][$i] = 1; |
973a28ed KW |
1468 | $gcb_table[$gcb_enums{'CR'}][$i] = 1; |
1469 | $gcb_table[$gcb_enums{'LF'}][$i] = 1; | |
1470 | } | |
1471 | ||
1472 | # GB3 CR × LF | |
1473 | $gcb_table[$gcb_enums{'CR'}][$gcb_enums{'LF'}] = 0; | |
1474 | ||
b0e24409 | 1475 | # Break at the start and end of text, unless the text is empty |
973a28ed KW |
1476 | # GB1 sot ÷ |
1477 | # GB2 ÷ eot | |
1478 | for my $i (0 .. @gcb_table - 1) { | |
289ce9cc KW |
1479 | $gcb_table[$i][$gcb_enums{'EDGE'}] = 1; |
1480 | $gcb_table[$gcb_enums{'EDGE'}][$i] = 1; | |
973a28ed | 1481 | } |
289ce9cc | 1482 | $gcb_table[$gcb_enums{'EDGE'}][$gcb_enums{'EDGE'}] = 0; |
973a28ed | 1483 | |
b0e24409 | 1484 | output_table_common('GCB', \%gcb_actions, |
289ce9cc | 1485 | \@gcb_table, \@gcb_short_enums, \%gcb_abbreviations); |
973a28ed KW |
1486 | } |
1487 | ||
6b659339 KW |
1488 | sub output_LB_table() { |
1489 | ||
1490 | # Create and output the enums, #defines, and pair table for use in | |
1491 | # determining Line Breaks. This uses the default line break algorithm, | |
1492 | # given in http://www.unicode.org/reports/tr14/, but tailored by example 7 | |
1493 | # in that page, as the Unicode-furnished tests assume that tailoring. | |
1494 | ||
6b659339 KW |
1495 | # The result is really just true or false. But we follow along with tr14, |
1496 | # creating a rule which is false for something like X SP* X. That gets | |
1497 | # encoding 2. The rest of the actions are synthetic ones that indicate | |
1498 | # some context handling is required. These each are added to the | |
1499 | # underlying 0, 1, or 2, instead of replacing them, so that the underlying | |
1500 | # value can be retrieved. Actually only rules from 7 through 18 (which | |
1501 | # are the ones where space matter) are possible to have 2 added to them. | |
1502 | # The others below add just 0 or 1. It might be possible for one | |
1503 | # synthetic rule to be added to another, yielding a larger value. This | |
1504 | # doesn't happen in the Unicode 8.0 rule set, and as you can see from the | |
1505 | # names of the middle grouping below, it is impossible for that to occur | |
1506 | # for them because they all start with mutually exclusive classes. That | |
1507 | # the final rule can't be added to any of the others isn't obvious from | |
1508 | # its name, so it is assigned a power of 2 higher than the others can get | |
1509 | # to so any addition would preserve all data. (And the code will reach an | |
1510 | # assert(0) on debugging builds should this happen.) | |
1511 | my %lb_actions = ( | |
1512 | LB_NOBREAK => 0, | |
1513 | LB_BREAKABLE => 1, | |
1514 | LB_NOBREAK_EVEN_WITH_SP_BETWEEN => 2, | |
1515 | ||
b0e24409 | 1516 | LB_CM_ZWJ_foo => 3, # Rule 9 |
6b659339 KW |
1517 | LB_SP_foo => 6, # Rule 18 |
1518 | LB_PR_or_PO_then_OP_or_HY => 9, # Rule 25 | |
1519 | LB_SY_or_IS_then_various => 11, # Rule 25 | |
1520 | LB_HY_or_BA_then_foo => 13, # Rule 21 | |
b0e24409 | 1521 | LB_RI_then_RI => 15, # Rule 30a |
6b659339 | 1522 | |
b0e24409 | 1523 | LB_various_then_PO_or_PR => (1<<5), # Rule 25 |
6b659339 KW |
1524 | ); |
1525 | ||
6b659339 KW |
1526 | # Construct the LB pair table. This is based on the rules in |
1527 | # http://www.unicode.org/reports/tr14/, but modified as those rules are | |
1528 | # designed for someone taking a string of text and sequentially going | |
1529 | # through it to find the break opportunities, whereas, Perl requires | |
1530 | # determining if a given random spot is a break opportunity, without | |
1531 | # knowing all the entire string before it. | |
1532 | # | |
1533 | # The table is constructed in reverse order of the rules, to make the | |
1534 | # lower-numbered, higher priority ones override the later ones, as the | |
1535 | # algorithm stops at the earliest matching rule | |
1536 | ||
1537 | my @lb_table; | |
1538 | my $table_size = @lb_short_enums; | |
1539 | ||
1540 | # LB31. Break everywhere else | |
1541 | for my $i (0 .. $table_size - 1) { | |
1542 | for my $j (0 .. $table_size - 1) { | |
1543 | $lb_table[$i][$j] = $lb_actions{'LB_BREAKABLE'}; | |
1544 | } | |
1545 | } | |
1546 | ||
2f1eff3d UC |
1547 | # LB30b Do not break between an emoji base (or potential emoji) and an |
1548 | # emoji modifier. | |
1549 | ||
b0e24409 | 1550 | # EB × EM |
2f1eff3d | 1551 | # [\p{Extended_Pictographic}&\p{Cn}] × EM |
b0e24409 KW |
1552 | $lb_table[$lb_enums{'E_Base'}][$lb_enums{'E_Modifier'}] |
1553 | = $lb_actions{'LB_NOBREAK'}; | |
2f1eff3d UC |
1554 | $lb_table[$lb_enums{'Unassigned_Extended_Pictographic_Ideographic'}] |
1555 | [$lb_enums{'E_Modifier'}] = $lb_actions{'LB_NOBREAK'}; | |
b0e24409 KW |
1556 | |
1557 | # LB30a Break between two regional indicator symbols if and only if there | |
1558 | # are an even number of regional indicators preceding the position of the | |
1559 | # break. | |
1560 | # sot (RI RI)* RI × RI | |
1561 | # [^RI] (RI RI)* RI × RI | |
289ce9cc | 1562 | $lb_table[$lb_enums{'Regional_Indicator'}] |
b0e24409 | 1563 | [$lb_enums{'Regional_Indicator'}] = $lb_actions{'LB_RI_then_RI'}; |
6b659339 KW |
1564 | |
1565 | # LB30 Do not break between letters, numbers, or ordinary symbols and | |
b6dbf1d3 UC |
1566 | # non-East-Asian opening punctuation nor non-East-Asian closing |
1567 | # parentheses. | |
1568 | ||
1569 | # (AL | HL | NU) × [OP-[\p{ea=F}\p{ea=W}\p{ea=H}]] | |
5c7c2de1 KW |
1570 | # (what we call CP and OP here have already been modified by mktables to |
1571 | # exclude the ea items | |
289ce9cc KW |
1572 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Open_Punctuation'}] |
1573 | = $lb_actions{'LB_NOBREAK'}; | |
1574 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Open_Punctuation'}] | |
1575 | = $lb_actions{'LB_NOBREAK'}; | |
1576 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Open_Punctuation'}] | |
1577 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1578 | |
b6dbf1d3 | 1579 | # [CP-[\p{ea=F}\p{ea=W}\p{ea=H}]] × (AL | HL | NU) |
289ce9cc KW |
1580 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Alphabetic'}] |
1581 | = $lb_actions{'LB_NOBREAK'}; | |
1582 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Hebrew_Letter'}] | |
1583 | = $lb_actions{'LB_NOBREAK'}; | |
1584 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Numeric'}] | |
1585 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1586 | |
1587 | # LB29 Do not break between numeric punctuation and alphabetics (“e.g.”). | |
1588 | # IS × (AL | HL) | |
289ce9cc KW |
1589 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Alphabetic'}] |
1590 | = $lb_actions{'LB_NOBREAK'}; | |
1591 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1592 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1593 | |
1594 | # LB28 Do not break between alphabetics (“at”). | |
1595 | # (AL | HL) × (AL | HL) | |
289ce9cc KW |
1596 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Alphabetic'}] |
1597 | = $lb_actions{'LB_NOBREAK'}; | |
1598 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Alphabetic'}] | |
1599 | = $lb_actions{'LB_NOBREAK'}; | |
1600 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Hebrew_Letter'}] | |
1601 | = $lb_actions{'LB_NOBREAK'}; | |
1602 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Hebrew_Letter'}] | |
1603 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1604 | |
1605 | # LB27 Treat a Korean Syllable Block the same as ID. | |
6b659339 | 1606 | # (JL | JV | JT | H2 | H3) × PO |
289ce9cc KW |
1607 | $lb_table[$lb_enums{'JL'}][$lb_enums{'Postfix_Numeric'}] |
1608 | = $lb_actions{'LB_NOBREAK'}; | |
1609 | $lb_table[$lb_enums{'JV'}][$lb_enums{'Postfix_Numeric'}] | |
1610 | = $lb_actions{'LB_NOBREAK'}; | |
1611 | $lb_table[$lb_enums{'JT'}][$lb_enums{'Postfix_Numeric'}] | |
1612 | = $lb_actions{'LB_NOBREAK'}; | |
1613 | $lb_table[$lb_enums{'H2'}][$lb_enums{'Postfix_Numeric'}] | |
1614 | = $lb_actions{'LB_NOBREAK'}; | |
1615 | $lb_table[$lb_enums{'H3'}][$lb_enums{'Postfix_Numeric'}] | |
1616 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1617 | |
1618 | # PR × (JL | JV | JT | H2 | H3) | |
289ce9cc KW |
1619 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'JL'}] |
1620 | = $lb_actions{'LB_NOBREAK'}; | |
1621 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'JV'}] | |
1622 | = $lb_actions{'LB_NOBREAK'}; | |
1623 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'JT'}] | |
1624 | = $lb_actions{'LB_NOBREAK'}; | |
1625 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'H2'}] | |
1626 | = $lb_actions{'LB_NOBREAK'}; | |
1627 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'H3'}] | |
1628 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1629 | |
1630 | # LB26 Do not break a Korean syllable. | |
1631 | # JL × (JL | JV | H2 | H3) | |
1632 | $lb_table[$lb_enums{'JL'}][$lb_enums{'JL'}] = $lb_actions{'LB_NOBREAK'}; | |
1633 | $lb_table[$lb_enums{'JL'}][$lb_enums{'JV'}] = $lb_actions{'LB_NOBREAK'}; | |
1634 | $lb_table[$lb_enums{'JL'}][$lb_enums{'H2'}] = $lb_actions{'LB_NOBREAK'}; | |
1635 | $lb_table[$lb_enums{'JL'}][$lb_enums{'H3'}] = $lb_actions{'LB_NOBREAK'}; | |
1636 | ||
1637 | # (JV | H2) × (JV | JT) | |
1638 | $lb_table[$lb_enums{'JV'}][$lb_enums{'JV'}] = $lb_actions{'LB_NOBREAK'}; | |
1639 | $lb_table[$lb_enums{'H2'}][$lb_enums{'JV'}] = $lb_actions{'LB_NOBREAK'}; | |
1640 | $lb_table[$lb_enums{'JV'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1641 | $lb_table[$lb_enums{'H2'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1642 | ||
1643 | # (JT | H3) × JT | |
1644 | $lb_table[$lb_enums{'JT'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1645 | $lb_table[$lb_enums{'H3'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1646 | ||
1647 | # LB25 Do not break between the following pairs of classes relevant to | |
1648 | # numbers, as tailored by example 7 in | |
1649 | # http://www.unicode.org/reports/tr14/#Examples | |
1650 | # We follow that tailoring because Unicode's test cases expect it | |
1651 | # (PR | PO) × ( OP | HY )? NU | |
289ce9cc KW |
1652 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Numeric'}] |
1653 | = $lb_actions{'LB_NOBREAK'}; | |
1654 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Numeric'}] | |
1655 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1656 | |
1657 | # Given that (OP | HY )? is optional, we have to test for it in code. | |
1658 | # We add in the action (instead of overriding) for this, so that in | |
1659 | # the code we can recover the underlying break value. | |
289ce9cc | 1660 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Open_Punctuation'}] |
6b659339 | 1661 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
b6dbf1d3 UC |
1662 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'East_Asian_OP'}] |
1663 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; | |
289ce9cc | 1664 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Open_Punctuation'}] |
6b659339 | 1665 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
289ce9cc | 1666 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Hyphen'}] |
6b659339 | 1667 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
289ce9cc | 1668 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Hyphen'}] |
6b659339 KW |
1669 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
1670 | ||
1671 | # ( OP | HY ) × NU | |
289ce9cc KW |
1672 | $lb_table[$lb_enums{'Open_Punctuation'}][$lb_enums{'Numeric'}] |
1673 | = $lb_actions{'LB_NOBREAK'}; | |
b6dbf1d3 UC |
1674 | $lb_table[$lb_enums{'East_Asian_OP'}][$lb_enums{'Numeric'}] |
1675 | = $lb_actions{'LB_NOBREAK'}; | |
289ce9cc KW |
1676 | $lb_table[$lb_enums{'Hyphen'}][$lb_enums{'Numeric'}] |
1677 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1678 | |
1679 | # NU (NU | SY | IS)* × (NU | SY | IS | CL | CP ) | |
1680 | # which can be rewritten as: | |
1681 | # NU (SY | IS)* × (NU | SY | IS | CL | CP ) | |
289ce9cc KW |
1682 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Numeric'}] |
1683 | = $lb_actions{'LB_NOBREAK'}; | |
1684 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Break_Symbols'}] | |
1685 | = $lb_actions{'LB_NOBREAK'}; | |
1686 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Infix_Numeric'}] | |
1687 | = $lb_actions{'LB_NOBREAK'}; | |
1688 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Close_Punctuation'}] | |
1689 | = $lb_actions{'LB_NOBREAK'}; | |
1690 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Close_Parenthesis'}] | |
1691 | = $lb_actions{'LB_NOBREAK'}; | |
b6dbf1d3 UC |
1692 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'East_Asian_CP'}] |
1693 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1694 | |
1695 | # Like earlier where we have to test in code, we add in the action so | |
1696 | # that we can recover the underlying values. This is done in rules | |
1697 | # below, as well. The code assumes that we haven't added 2 actions. | |
1698 | # Shoul a later Unicode release break that assumption, then tests | |
1699 | # should start failing. | |
289ce9cc | 1700 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Numeric'}] |
6b659339 | 1701 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1702 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Break_Symbols'}] |
6b659339 | 1703 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1704 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Infix_Numeric'}] |
6b659339 | 1705 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1706 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Close_Punctuation'}] |
6b659339 | 1707 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1708 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Close_Parenthesis'}] |
6b659339 | 1709 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
b6dbf1d3 UC |
1710 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'East_Asian_CP'}] |
1711 | += $lb_actions{'LB_SY_or_IS_then_various'}; | |
289ce9cc | 1712 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Numeric'}] |
6b659339 | 1713 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1714 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Break_Symbols'}] |
6b659339 | 1715 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1716 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Infix_Numeric'}] |
6b659339 | 1717 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1718 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Close_Punctuation'}] |
6b659339 | 1719 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1720 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Close_Parenthesis'}] |
6b659339 | 1721 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
b6dbf1d3 UC |
1722 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'East_Asian_CP'}] |
1723 | += $lb_actions{'LB_SY_or_IS_then_various'}; | |
6b659339 KW |
1724 | |
1725 | # NU (NU | SY | IS)* (CL | CP)? × (PO | PR) | |
1726 | # which can be rewritten as: | |
1727 | # NU (SY | IS)* (CL | CP)? × (PO | PR) | |
289ce9cc KW |
1728 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Postfix_Numeric'}] |
1729 | = $lb_actions{'LB_NOBREAK'}; | |
1730 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Prefix_Numeric'}] | |
1731 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1732 | |
289ce9cc | 1733 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 | 1734 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
b6dbf1d3 UC |
1735 | $lb_table[$lb_enums{'East_Asian_CP'}][$lb_enums{'Postfix_Numeric'}] |
1736 | += $lb_actions{'LB_various_then_PO_or_PR'}; | |
289ce9cc | 1737 | $lb_table[$lb_enums{'Close_Punctuation'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 | 1738 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1739 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 | 1740 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1741 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 KW |
1742 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
1743 | ||
289ce9cc | 1744 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 | 1745 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
b6dbf1d3 UC |
1746 | $lb_table[$lb_enums{'East_Asian_CP'}][$lb_enums{'Prefix_Numeric'}] |
1747 | += $lb_actions{'LB_various_then_PO_or_PR'}; | |
289ce9cc | 1748 | $lb_table[$lb_enums{'Close_Punctuation'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 | 1749 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1750 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 | 1751 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1752 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 KW |
1753 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
1754 | ||
b0e24409 KW |
1755 | # LB24 Do not break between numeric prefix/postfix and letters, or between |
1756 | # letters and prefix/postfix. | |
1757 | # (PR | PO) × (AL | HL) | |
289ce9cc KW |
1758 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Alphabetic'}] |
1759 | = $lb_actions{'LB_NOBREAK'}; | |
1760 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1761 | = $lb_actions{'LB_NOBREAK'}; | |
289ce9cc KW |
1762 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Alphabetic'}] |
1763 | = $lb_actions{'LB_NOBREAK'}; | |
1764 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1765 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1766 | |
b0e24409 KW |
1767 | # (AL | HL) × (PR | PO) |
1768 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Prefix_Numeric'}] | |
1769 | = $lb_actions{'LB_NOBREAK'}; | |
1770 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Prefix_Numeric'}] | |
1771 | = $lb_actions{'LB_NOBREAK'}; | |
1772 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Postfix_Numeric'}] | |
1773 | = $lb_actions{'LB_NOBREAK'}; | |
1774 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Postfix_Numeric'}] | |
1775 | = $lb_actions{'LB_NOBREAK'}; | |
1776 | ||
1777 | # LB23a Do not break between numeric prefixes and ideographs, or between | |
1778 | # ideographs and numeric postfixes. | |
1779 | # PR × (ID | EB | EM) | |
1780 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Ideographic'}] | |
1781 | = $lb_actions{'LB_NOBREAK'}; | |
de524f25 KW |
1782 | $lb_table[$lb_enums{'Prefix_Numeric'}] |
1783 | [$lb_enums{'Unassigned_Extended_Pictographic_Ideographic'}] | |
1784 | = $lb_actions{'LB_NOBREAK'}; | |
b0e24409 KW |
1785 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'E_Base'}] |
1786 | = $lb_actions{'LB_NOBREAK'}; | |
1787 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'E_Modifier'}] | |
1788 | = $lb_actions{'LB_NOBREAK'}; | |
1789 | ||
1790 | # (ID | EB | EM) × PO | |
289ce9cc KW |
1791 | $lb_table[$lb_enums{'Ideographic'}][$lb_enums{'Postfix_Numeric'}] |
1792 | = $lb_actions{'LB_NOBREAK'}; | |
de524f25 KW |
1793 | $lb_table[$lb_enums{'Unassigned_Extended_Pictographic_Ideographic'}] |
1794 | [$lb_enums{'Postfix_Numeric'}] = $lb_actions{'LB_NOBREAK'}; | |
b0e24409 KW |
1795 | $lb_table[$lb_enums{'E_Base'}][$lb_enums{'Postfix_Numeric'}] |
1796 | = $lb_actions{'LB_NOBREAK'}; | |
1797 | $lb_table[$lb_enums{'E_Modifier'}][$lb_enums{'Postfix_Numeric'}] | |
1798 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1799 | |
b0e24409 | 1800 | # LB23 Do not break between digits and letters |
6b659339 | 1801 | # (AL | HL) × NU |
289ce9cc KW |
1802 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Numeric'}] |
1803 | = $lb_actions{'LB_NOBREAK'}; | |
1804 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Numeric'}] | |
1805 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1806 | |
1807 | # NU × (AL | HL) | |
289ce9cc KW |
1808 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Alphabetic'}] |
1809 | = $lb_actions{'LB_NOBREAK'}; | |
1810 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1811 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1812 | |
b6dbf1d3 UC |
1813 | # LB22 Do not break before ellipses |
1814 | for my $i (0 .. @lb_table - 1) { | |
1815 | $lb_table[$i][$lb_enums{'Inseparable'}] = $lb_actions{'LB_NOBREAK'}; | |
1816 | } | |
6b659339 KW |
1817 | |
1818 | # LB21b Don’t break between Solidus and Hebrew letters. | |
1819 | # SY × HL | |
289ce9cc KW |
1820 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Hebrew_Letter'}] |
1821 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1822 | |
1823 | # LB21a Don't break after Hebrew + Hyphen. | |
1824 | # HL (HY | BA) × | |
1825 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1826 | $lb_table[$lb_enums{'Hyphen'}][$i] |
1827 | += $lb_actions{'LB_HY_or_BA_then_foo'}; | |
1828 | $lb_table[$lb_enums{'Break_After'}][$i] | |
1829 | += $lb_actions{'LB_HY_or_BA_then_foo'}; | |
6b659339 KW |
1830 | } |
1831 | ||
1832 | # LB21 Do not break before hyphen-minus, other hyphens, fixed-width | |
1833 | # spaces, small kana, and other non-starters, or after acute accents. | |
1834 | # × BA | |
1835 | # × HY | |
1836 | # × NS | |
1837 | # BB × | |
1838 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1839 | $lb_table[$i][$lb_enums{'Break_After'}] = $lb_actions{'LB_NOBREAK'}; |
1840 | $lb_table[$i][$lb_enums{'Hyphen'}] = $lb_actions{'LB_NOBREAK'}; | |
1841 | $lb_table[$i][$lb_enums{'Nonstarter'}] = $lb_actions{'LB_NOBREAK'}; | |
1842 | $lb_table[$lb_enums{'Break_Before'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1843 | } |
1844 | ||
1845 | # LB20 Break before and after unresolved CB. | |
1846 | # ÷ CB | |
1847 | # CB ÷ | |
1848 | # Conditional breaks should be resolved external to the line breaking | |
1849 | # rules. However, the default action is to treat unresolved CB as breaking | |
1850 | # before and after. | |
1851 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1852 | $lb_table[$i][$lb_enums{'Contingent_Break'}] |
1853 | = $lb_actions{'LB_BREAKABLE'}; | |
1854 | $lb_table[$lb_enums{'Contingent_Break'}][$i] | |
1855 | = $lb_actions{'LB_BREAKABLE'}; | |
6b659339 KW |
1856 | } |
1857 | ||
1858 | # LB19 Do not break before or after quotation marks, such as ‘ ” ’. | |
1859 | # × QU | |
1860 | # QU × | |
1861 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1862 | $lb_table[$i][$lb_enums{'Quotation'}] = $lb_actions{'LB_NOBREAK'}; |
1863 | $lb_table[$lb_enums{'Quotation'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1864 | } |
1865 | ||
1866 | # LB18 Break after spaces | |
1867 | # SP ÷ | |
1868 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1869 | $lb_table[$lb_enums{'Space'}][$i] = $lb_actions{'LB_BREAKABLE'}; |
6b659339 KW |
1870 | } |
1871 | ||
1872 | # LB17 Do not break within ‘——’, even with intervening spaces. | |
1873 | # B2 SP* × B2 | |
289ce9cc | 1874 | $lb_table[$lb_enums{'Break_Both'}][$lb_enums{'Break_Both'}] |
6b659339 KW |
1875 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1876 | ||
1877 | # LB16 Do not break between closing punctuation and a nonstarter even with | |
1878 | # intervening spaces. | |
1879 | # (CL | CP) SP* × NS | |
289ce9cc | 1880 | $lb_table[$lb_enums{'Close_Punctuation'}][$lb_enums{'Nonstarter'}] |
6b659339 | 1881 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1882 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Nonstarter'}] |
6b659339 | 1883 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
b6dbf1d3 UC |
1884 | $lb_table[$lb_enums{'East_Asian_CP'}][$lb_enums{'Nonstarter'}] |
1885 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; | |
6b659339 KW |
1886 | |
1887 | ||
1888 | # LB15 Do not break within ‘”[’, even with intervening spaces. | |
1889 | # QU SP* × OP | |
289ce9cc | 1890 | $lb_table[$lb_enums{'Quotation'}][$lb_enums{'Open_Punctuation'}] |
6b659339 | 1891 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
b6dbf1d3 UC |
1892 | $lb_table[$lb_enums{'Quotation'}][$lb_enums{'East_Asian_OP'}] |
1893 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; | |
6b659339 KW |
1894 | |
1895 | # LB14 Do not break after ‘[’, even after spaces. | |
1896 | # OP SP* × | |
1897 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1898 | $lb_table[$lb_enums{'Open_Punctuation'}][$i] |
6b659339 | 1899 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
b6dbf1d3 UC |
1900 | $lb_table[$lb_enums{'East_Asian_OP'}][$i] |
1901 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; | |
6b659339 KW |
1902 | } |
1903 | ||
1904 | # LB13 Do not break before ‘]’ or ‘!’ or ‘;’ or ‘/’, even after spaces, as | |
1905 | # tailored by example 7 in http://www.unicode.org/reports/tr14/#Examples | |
1906 | # [^NU] × CL | |
1907 | # [^NU] × CP | |
1908 | # × EX | |
1909 | # [^NU] × IS | |
1910 | # [^NU] × SY | |
1911 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1912 | $lb_table[$i][$lb_enums{'Exclamation'}] |
6b659339 KW |
1913 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1914 | ||
289ce9cc | 1915 | next if $i == $lb_enums{'Numeric'}; |
6b659339 | 1916 | |
289ce9cc | 1917 | $lb_table[$i][$lb_enums{'Close_Punctuation'}] |
6b659339 | 1918 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1919 | $lb_table[$i][$lb_enums{'Close_Parenthesis'}] |
6b659339 | 1920 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
b6dbf1d3 UC |
1921 | $lb_table[$i][$lb_enums{'East_Asian_CP'}] |
1922 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; | |
289ce9cc | 1923 | $lb_table[$i][$lb_enums{'Infix_Numeric'}] |
6b659339 | 1924 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1925 | $lb_table[$i][$lb_enums{'Break_Symbols'}] |
6b659339 KW |
1926 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1927 | } | |
1928 | ||
1929 | # LB12a Do not break before NBSP and related characters, except after | |
1930 | # spaces and hyphens. | |
1931 | # [^SP BA HY] × GL | |
1932 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1933 | next if $i == $lb_enums{'Space'} |
1934 | || $i == $lb_enums{'Break_After'} | |
1935 | || $i == $lb_enums{'Hyphen'}; | |
6b659339 KW |
1936 | |
1937 | # We don't break, but if a property above has said don't break even | |
1938 | # with space between, don't override that (also in the next few rules) | |
289ce9cc | 1939 | next if $lb_table[$i][$lb_enums{'Glue'}] |
6b659339 | 1940 | == $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1941 | $lb_table[$i][$lb_enums{'Glue'}] = $lb_actions{'LB_NOBREAK'}; |
6b659339 KW |
1942 | } |
1943 | ||
1944 | # LB12 Do not break after NBSP and related characters. | |
1945 | # GL × | |
1946 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1947 | next if $lb_table[$lb_enums{'Glue'}][$i] |
6b659339 | 1948 | == $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1949 | $lb_table[$lb_enums{'Glue'}][$i] = $lb_actions{'LB_NOBREAK'}; |
6b659339 KW |
1950 | } |
1951 | ||
1952 | # LB11 Do not break before or after Word joiner and related characters. | |
1953 | # × WJ | |
1954 | # WJ × | |
1955 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1956 | if ($lb_table[$i][$lb_enums{'Word_Joiner'}] |
6b659339 KW |
1957 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) |
1958 | { | |
289ce9cc | 1959 | $lb_table[$i][$lb_enums{'Word_Joiner'}] = $lb_actions{'LB_NOBREAK'}; |
6b659339 | 1960 | } |
289ce9cc | 1961 | if ($lb_table[$lb_enums{'Word_Joiner'}][$i] |
6b659339 KW |
1962 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) |
1963 | { | |
289ce9cc | 1964 | $lb_table[$lb_enums{'Word_Joiner'}][$i] = $lb_actions{'LB_NOBREAK'}; |
6b659339 KW |
1965 | } |
1966 | } | |
1967 | ||
1968 | # Special case this here to avoid having to do a special case in the code, | |
1969 | # by making this the same as other things with a SP in front of them that | |
1970 | # don't break, we avoid an extra test | |
289ce9cc | 1971 | $lb_table[$lb_enums{'Space'}][$lb_enums{'Word_Joiner'}] |
6b659339 KW |
1972 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1973 | ||
1974 | # LB9 and LB10 are done in the same loop | |
1975 | # | |
1976 | # LB9 Do not break a combining character sequence; treat it as if it has | |
1977 | # the line breaking class of the base character in all of the | |
b0e24409 KW |
1978 | # higher-numbered rules. Treat ZWJ as if it were CM |
1979 | # Treat X (CM|ZWJ)* as if it were X. | |
6b659339 KW |
1980 | # where X is any line break class except BK, CR, LF, NL, SP, or ZW. |
1981 | ||
b0e24409 KW |
1982 | # LB10 Treat any remaining combining mark or ZWJ as AL. This catches the |
1983 | # case where a CM or ZWJ is the first character on the line or follows SP, | |
1984 | # BK, CR, LF, NL, or ZW. | |
6b659339 KW |
1985 | for my $i (0 .. @lb_table - 1) { |
1986 | ||
b0e24409 KW |
1987 | # When the CM or ZWJ is the first in the pair, we don't know without |
1988 | # looking behind whether the CM or ZWJ is going to attach to an | |
1989 | # earlier character, or not. So have to figure this out at runtime in | |
1990 | # the code | |
1991 | $lb_table[$lb_enums{'Combining_Mark'}][$i] | |
1992 | = $lb_actions{'LB_CM_ZWJ_foo'}; | |
1993 | $lb_table[$lb_enums{'ZWJ'}][$i] = $lb_actions{'LB_CM_ZWJ_foo'}; | |
289ce9cc KW |
1994 | |
1995 | if ( $i == $lb_enums{'Mandatory_Break'} | |
1996 | || $i == $lb_enums{'EDGE'} | |
1997 | || $i == $lb_enums{'Carriage_Return'} | |
1998 | || $i == $lb_enums{'Line_Feed'} | |
1999 | || $i == $lb_enums{'Next_Line'} | |
2000 | || $i == $lb_enums{'Space'} | |
2001 | || $i == $lb_enums{'ZWSpace'}) | |
6b659339 KW |
2002 | { |
2003 | # For these classes, a following CM doesn't combine, and should do | |
289ce9cc KW |
2004 | # whatever 'Alphabetic' would do. |
2005 | $lb_table[$i][$lb_enums{'Combining_Mark'}] | |
2006 | = $lb_table[$i][$lb_enums{'Alphabetic'}]; | |
b0e24409 KW |
2007 | $lb_table[$i][$lb_enums{'ZWJ'}] |
2008 | = $lb_table[$i][$lb_enums{'Alphabetic'}]; | |
6b659339 KW |
2009 | } |
2010 | else { | |
b0e24409 KW |
2011 | # For these classes, the CM or ZWJ combines, so doesn't break, |
2012 | # inheriting the type of nobreak from the master character. | |
289ce9cc | 2013 | if ($lb_table[$i][$lb_enums{'Combining_Mark'}] |
6b659339 KW |
2014 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) |
2015 | { | |
289ce9cc KW |
2016 | $lb_table[$i][$lb_enums{'Combining_Mark'}] |
2017 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 2018 | } |
b0e24409 KW |
2019 | if ($lb_table[$i][$lb_enums{'ZWJ'}] |
2020 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) | |
2021 | { | |
2022 | $lb_table[$i][$lb_enums{'ZWJ'}] | |
2023 | = $lb_actions{'LB_NOBREAK'}; | |
2024 | } | |
6b659339 KW |
2025 | } |
2026 | } | |
2027 | ||
8a6698d7 UC |
2028 | # LB8a Do not break after a zero width joiner |
2029 | # ZWJ × | |
2030 | for my $i (0 .. @lb_table - 1) { | |
2031 | $lb_table[$lb_enums{'ZWJ'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
2032 | } | |
b0e24409 | 2033 | |
6b659339 KW |
2034 | # LB8 Break before any character following a zero-width space, even if one |
2035 | # or more spaces intervene. | |
2036 | # ZW SP* ÷ | |
2037 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 2038 | $lb_table[$lb_enums{'ZWSpace'}][$i] = $lb_actions{'LB_BREAKABLE'}; |
6b659339 KW |
2039 | } |
2040 | ||
2041 | # Because of LB8-10, we need to look at context for "SP x", and this must | |
2042 | # be done in the code. So override the existing rules for that, by adding | |
2043 | # a constant to get new rules that tell the code it needs to look at | |
2044 | # context. By adding this action instead of replacing the existing one, | |
2045 | # we can get back to the original rule if necessary. | |
2046 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 2047 | $lb_table[$lb_enums{'Space'}][$i] += $lb_actions{'LB_SP_foo'}; |
6b659339 KW |
2048 | } |
2049 | ||
2050 | # LB7 Do not break before spaces or zero width space. | |
2051 | # × SP | |
2052 | # × ZW | |
2053 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
2054 | $lb_table[$i][$lb_enums{'Space'}] = $lb_actions{'LB_NOBREAK'}; |
2055 | $lb_table[$i][$lb_enums{'ZWSpace'}] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
2056 | } |
2057 | ||
2058 | # LB6 Do not break before hard line breaks. | |
2059 | # × ( BK | CR | LF | NL ) | |
2060 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
2061 | $lb_table[$i][$lb_enums{'Mandatory_Break'}] = $lb_actions{'LB_NOBREAK'}; |
2062 | $lb_table[$i][$lb_enums{'Carriage_Return'}] = $lb_actions{'LB_NOBREAK'}; | |
2063 | $lb_table[$i][$lb_enums{'Line_Feed'}] = $lb_actions{'LB_NOBREAK'}; | |
2064 | $lb_table[$i][$lb_enums{'Next_Line'}] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
2065 | } |
2066 | ||
2067 | # LB5 Treat CR followed by LF, as well as CR, LF, and NL as hard line breaks. | |
2068 | # CR × LF | |
2069 | # CR ! | |
2070 | # LF ! | |
2071 | # NL ! | |
2072 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
2073 | $lb_table[$lb_enums{'Carriage_Return'}][$i] |
2074 | = $lb_actions{'LB_BREAKABLE'}; | |
2075 | $lb_table[$lb_enums{'Line_Feed'}][$i] = $lb_actions{'LB_BREAKABLE'}; | |
2076 | $lb_table[$lb_enums{'Next_Line'}][$i] = $lb_actions{'LB_BREAKABLE'}; | |
6b659339 | 2077 | } |
289ce9cc KW |
2078 | $lb_table[$lb_enums{'Carriage_Return'}][$lb_enums{'Line_Feed'}] |
2079 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
2080 | |
2081 | # LB4 Always break after hard line breaks. | |
2082 | # BK ! | |
2083 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
2084 | $lb_table[$lb_enums{'Mandatory_Break'}][$i] |
2085 | = $lb_actions{'LB_BREAKABLE'}; | |
6b659339 KW |
2086 | } |
2087 | ||
6b659339 KW |
2088 | # LB3 Always break at the end of text. |
2089 | # ! eot | |
b0e24409 KW |
2090 | # LB2 Never break at the start of text. |
2091 | # sot × | |
6b659339 | 2092 | for my $i (0 .. @lb_table - 1) { |
289ce9cc KW |
2093 | $lb_table[$i][$lb_enums{'EDGE'}] = $lb_actions{'LB_BREAKABLE'}; |
2094 | $lb_table[$lb_enums{'EDGE'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
2095 | } |
2096 | ||
2097 | # LB1 Assign a line breaking class to each code point of the input. | |
2098 | # Resolve AI, CB, CJ, SA, SG, and XX into other line breaking classes | |
2099 | # depending on criteria outside the scope of this algorithm. | |
2100 | # | |
2101 | # In the absence of such criteria all characters with a specific | |
2102 | # combination of original class and General_Category property value are | |
2103 | # resolved as follows: | |
2104 | # Original Resolved General_Category | |
2105 | # AI, SG, XX AL Any | |
2106 | # SA CM Only Mn or Mc | |
2107 | # SA AL Any except Mn and Mc | |
2108 | # CJ NS Any | |
2109 | # | |
2110 | # This is done in mktables, so we never see any of the remapped-from | |
2111 | # classes. | |
2112 | ||
289ce9cc KW |
2113 | output_table_common('LB', \%lb_actions, |
2114 | \@lb_table, \@lb_short_enums, \%lb_abbreviations); | |
6b659339 KW |
2115 | } |
2116 | ||
7e54b87f KW |
2117 | sub output_WB_table() { |
2118 | ||
2119 | # Create and output the enums, #defines, and pair table for use in | |
2120 | # determining Word Breaks, given in http://www.unicode.org/reports/tr29/. | |
2121 | ||
2122 | # This uses the same mechanism in the other bounds tables generated by | |
2123 | # this file. The actions that could override a 0 or 1 are added to those | |
2124 | # numbers; the actions that clearly don't depend on the underlying rule | |
2125 | # simply overwrite | |
2126 | my %wb_actions = ( | |
2127 | WB_NOBREAK => 0, | |
2128 | WB_BREAKABLE => 1, | |
2129 | WB_hs_then_hs => 2, | |
b0e24409 | 2130 | WB_Ex_or_FO_or_ZWJ_then_foo => 3, |
7e54b87f KW |
2131 | WB_DQ_then_HL => 4, |
2132 | WB_HL_then_DQ => 6, | |
2133 | WB_LE_or_HL_then_MB_or_ML_or_SQ => 8, | |
2134 | WB_MB_or_ML_or_SQ_then_LE_or_HL => 10, | |
2135 | WB_MB_or_MN_or_SQ_then_NU => 12, | |
2136 | WB_NU_then_MB_or_MN_or_SQ => 14, | |
b0e24409 | 2137 | WB_RI_then_RI => 16, |
7e54b87f KW |
2138 | ); |
2139 | ||
7e54b87f KW |
2140 | # Construct the WB pair table. |
2141 | # The table is constructed in reverse order of the rules, to make the | |
2142 | # lower-numbered, higher priority ones override the later ones, as the | |
2143 | # algorithm stops at the earliest matching rule | |
2144 | ||
2145 | my @wb_table; | |
2027d365 | 2146 | my $table_size = @wb_short_enums; |
7e54b87f KW |
2147 | |
2148 | # Otherwise, break everywhere (including around ideographs). | |
b0e24409 | 2149 | # WB99 Any ÷ Any |
7e54b87f KW |
2150 | for my $i (0 .. $table_size - 1) { |
2151 | for my $j (0 .. $table_size - 1) { | |
2152 | $wb_table[$i][$j] = $wb_actions{'WB_BREAKABLE'}; | |
2153 | } | |
2154 | } | |
2155 | ||
b0e24409 KW |
2156 | # Do not break within emoji flag sequences. That is, do not break between |
2157 | # regional indicator (RI) symbols if there is an odd number of RI | |
2158 | # characters before the break point. | |
2159 | # WB16 [^RI] (RI RI)* RI × RI | |
c492f156 | 2160 | # WB15 sot (RI RI)* RI × RI |
289ce9cc | 2161 | $wb_table[$wb_enums{'Regional_Indicator'}] |
b0e24409 KW |
2162 | [$wb_enums{'Regional_Indicator'}] = $wb_actions{'WB_RI_then_RI'}; |
2163 | ||
2164 | # Do not break within emoji modifier sequences. | |
2165 | # WB14 ( E_Base | EBG ) × E_Modifier | |
2166 | $wb_table[$wb_enums{'E_Base'}][$wb_enums{'E_Modifier'}] | |
2167 | = $wb_actions{'WB_NOBREAK'}; | |
2168 | $wb_table[$wb_enums{'E_Base_GAZ'}][$wb_enums{'E_Modifier'}] | |
2169 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2170 | |
2171 | # Do not break from extenders. | |
2172 | # WB13b ExtendNumLet × (ALetter | Hebrew_Letter | Numeric | Katakana) | |
289ce9cc KW |
2173 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'ALetter'}] |
2174 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2175 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2176 | = $wb_actions{'WB_NOBREAK'}; |
289ce9cc KW |
2177 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'Hebrew_Letter'}] |
2178 | = $wb_actions{'WB_NOBREAK'}; | |
2179 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'Numeric'}] | |
2180 | = $wb_actions{'WB_NOBREAK'}; | |
2181 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'Katakana'}] | |
2182 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2183 | |
2184 | # WB13a (ALetter | Hebrew_Letter | Numeric | Katakana | ExtendNumLet) | |
d21ae9f6 | 2185 | # × ExtendNumLet |
289ce9cc KW |
2186 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'ExtendNumLet'}] |
2187 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2188 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'ExtendNumLet'}] |
c0734505 | 2189 | = $wb_actions{'WB_NOBREAK'}; |
289ce9cc KW |
2190 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'ExtendNumLet'}] |
2191 | = $wb_actions{'WB_NOBREAK'}; | |
2192 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'ExtendNumLet'}] | |
2193 | = $wb_actions{'WB_NOBREAK'}; | |
2194 | $wb_table[$wb_enums{'Katakana'}][$wb_enums{'ExtendNumLet'}] | |
2195 | = $wb_actions{'WB_NOBREAK'}; | |
2196 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'ExtendNumLet'}] | |
2197 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2198 | |
2199 | # Do not break between Katakana. | |
2200 | # WB13 Katakana × Katakana | |
289ce9cc KW |
2201 | $wb_table[$wb_enums{'Katakana'}][$wb_enums{'Katakana'}] |
2202 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2203 | |
2204 | # Do not break within sequences, such as “3.2” or “3,456.789”. | |
2205 | # WB12 Numeric × (MidNum | MidNumLet | Single_Quote) Numeric | |
289ce9cc | 2206 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'MidNumLet'}] |
7e54b87f | 2207 | += $wb_actions{'WB_NU_then_MB_or_MN_or_SQ'}; |
289ce9cc | 2208 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'MidNum'}] |
7e54b87f | 2209 | += $wb_actions{'WB_NU_then_MB_or_MN_or_SQ'}; |
289ce9cc | 2210 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'Single_Quote'}] |
7e54b87f KW |
2211 | += $wb_actions{'WB_NU_then_MB_or_MN_or_SQ'}; |
2212 | ||
2213 | # WB11 Numeric (MidNum | (MidNumLet | Single_Quote)) × Numeric | |
289ce9cc | 2214 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'Numeric'}] |
7e54b87f | 2215 | += $wb_actions{'WB_MB_or_MN_or_SQ_then_NU'}; |
289ce9cc | 2216 | $wb_table[$wb_enums{'MidNum'}][$wb_enums{'Numeric'}] |
7e54b87f | 2217 | += $wb_actions{'WB_MB_or_MN_or_SQ_then_NU'}; |
289ce9cc | 2218 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'Numeric'}] |
7e54b87f KW |
2219 | += $wb_actions{'WB_MB_or_MN_or_SQ_then_NU'}; |
2220 | ||
2221 | # Do not break within sequences of digits, or digits adjacent to letters | |
2222 | # (“3a”, or “A3”). | |
2223 | # WB10 Numeric × (ALetter | Hebrew_Letter) | |
289ce9cc KW |
2224 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'ALetter'}] |
2225 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2226 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2227 | = $wb_actions{'WB_NOBREAK'}; |
289ce9cc KW |
2228 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'Hebrew_Letter'}] |
2229 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2230 | |
2231 | # WB9 (ALetter | Hebrew_Letter) × Numeric | |
289ce9cc KW |
2232 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'Numeric'}] |
2233 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2234 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'Numeric'}] |
c0734505 | 2235 | = $wb_actions{'WB_NOBREAK'}; |
289ce9cc KW |
2236 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Numeric'}] |
2237 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2238 | |
2239 | # WB8 Numeric × Numeric | |
289ce9cc KW |
2240 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'Numeric'}] |
2241 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2242 | |
2243 | # Do not break letters across certain punctuation. | |
2244 | # WB7c Hebrew_Letter Double_Quote × Hebrew_Letter | |
289ce9cc KW |
2245 | $wb_table[$wb_enums{'Double_Quote'}][$wb_enums{'Hebrew_Letter'}] |
2246 | += $wb_actions{'WB_DQ_then_HL'}; | |
7e54b87f KW |
2247 | |
2248 | # WB7b Hebrew_Letter × Double_Quote Hebrew_Letter | |
289ce9cc KW |
2249 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Double_Quote'}] |
2250 | += $wb_actions{'WB_HL_then_DQ'}; | |
7e54b87f KW |
2251 | |
2252 | # WB7a Hebrew_Letter × Single_Quote | |
289ce9cc KW |
2253 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Single_Quote'}] |
2254 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2255 | |
2256 | # WB7 (ALetter | Hebrew_Letter) (MidLetter | MidNumLet | Single_Quote) | |
2257 | # × (ALetter | Hebrew_Letter) | |
289ce9cc | 2258 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'ALetter'}] |
7e54b87f | 2259 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
a9256a75 | 2260 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2261 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
289ce9cc | 2262 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'Hebrew_Letter'}] |
7e54b87f | 2263 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
289ce9cc | 2264 | $wb_table[$wb_enums{'MidLetter'}][$wb_enums{'ALetter'}] |
7e54b87f | 2265 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
a9256a75 | 2266 | $wb_table[$wb_enums{'MidLetter'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2267 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
289ce9cc | 2268 | $wb_table[$wb_enums{'MidLetter'}][$wb_enums{'Hebrew_Letter'}] |
7e54b87f | 2269 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
289ce9cc | 2270 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'ALetter'}] |
7e54b87f | 2271 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
a9256a75 | 2272 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2273 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
289ce9cc | 2274 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'Hebrew_Letter'}] |
7e54b87f KW |
2275 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
2276 | ||
2277 | # WB6 (ALetter | Hebrew_Letter) × (MidLetter | MidNumLet | |
2278 | # | Single_Quote) (ALetter | Hebrew_Letter) | |
289ce9cc | 2279 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'MidNumLet'}] |
7e54b87f | 2280 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
a9256a75 | 2281 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'MidNumLet'}] |
c0734505 | 2282 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
289ce9cc | 2283 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'MidNumLet'}] |
7e54b87f | 2284 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
289ce9cc | 2285 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'MidLetter'}] |
7e54b87f | 2286 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
a9256a75 | 2287 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'MidLetter'}] |
c0734505 | 2288 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
289ce9cc | 2289 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'MidLetter'}] |
7e54b87f | 2290 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
289ce9cc | 2291 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'Single_Quote'}] |
7e54b87f | 2292 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
a9256a75 | 2293 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'Single_Quote'}] |
c0734505 | 2294 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
289ce9cc | 2295 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Single_Quote'}] |
7e54b87f KW |
2296 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
2297 | ||
2298 | # Do not break between most letters. | |
2299 | # WB5 (ALetter | Hebrew_Letter) × (ALetter | Hebrew_Letter) | |
289ce9cc KW |
2300 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'ALetter'}] |
2301 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2302 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'ALetter'}] |
c0734505 | 2303 | = $wb_actions{'WB_NOBREAK'}; |
289ce9cc KW |
2304 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'Hebrew_Letter'}] |
2305 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2306 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'Hebrew_Letter'}] |
c0734505 | 2307 | = $wb_actions{'WB_NOBREAK'}; |
289ce9cc KW |
2308 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'ALetter'}] |
2309 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2310 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2311 | = $wb_actions{'WB_NOBREAK'}; |
289ce9cc KW |
2312 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Hebrew_Letter'}] |
2313 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2314 | $wb_table[$wb_enums{'ExtPict_LE'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2315 | = $wb_actions{'WB_NOBREAK'}; |
7e54b87f | 2316 | |
b0e24409 KW |
2317 | # Ignore Format and Extend characters, except after sot, CR, LF, and |
2318 | # Newline. This also has the effect of: Any × (Format | Extend | ZWJ) | |
2319 | # WB4 X (Extend | Format | ZWJ)* → X | |
7e54b87f | 2320 | for my $i (0 .. @wb_table - 1) { |
289ce9cc | 2321 | $wb_table[$wb_enums{'Extend'}][$i] |
b0e24409 | 2322 | = $wb_actions{'WB_Ex_or_FO_or_ZWJ_then_foo'}; |
289ce9cc | 2323 | $wb_table[$wb_enums{'Format'}][$i] |
b0e24409 KW |
2324 | = $wb_actions{'WB_Ex_or_FO_or_ZWJ_then_foo'}; |
2325 | $wb_table[$wb_enums{'ZWJ'}][$i] | |
2326 | = $wb_actions{'WB_Ex_or_FO_or_ZWJ_then_foo'}; | |
2327 | } | |
2328 | for my $i (0 .. @wb_table - 1) { | |
2329 | $wb_table[$i][$wb_enums{'Extend'}] = $wb_actions{'WB_NOBREAK'}; | |
2330 | $wb_table[$i][$wb_enums{'Format'}] = $wb_actions{'WB_NOBREAK'}; | |
2331 | $wb_table[$i][$wb_enums{'ZWJ'}] = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2332 | } |
2333 | ||
2334 | # Implied is that these attach to the character before them, except for | |
2335 | # the characters that mark the end of a region of text. The rules below | |
2336 | # override the ones set up here, for all the characters that need | |
2337 | # overriding. | |
2338 | for my $i (0 .. @wb_table - 1) { | |
289ce9cc KW |
2339 | $wb_table[$i][$wb_enums{'Extend'}] = $wb_actions{'WB_NOBREAK'}; |
2340 | $wb_table[$i][$wb_enums{'Format'}] = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2341 | } |
2342 | ||
c0734505 KW |
2343 | # Keep horizontal whitespace together |
2344 | # Use perl's tailoring instead | |
2345 | # WB3d WSegSpace × WSegSpace | |
2346 | #$wb_table[$wb_enums{'WSegSpace'}][$wb_enums{'WSegSpace'}] | |
2347 | # = $wb_actions{'WB_NOBREAK'}; | |
2348 | ||
b0e24409 KW |
2349 | # Do not break within emoji zwj sequences. |
2350 | # WB3c ZWJ × ( Glue_After_Zwj | EBG ) | |
2351 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'Glue_After_Zwj'}] | |
2352 | = $wb_actions{'WB_NOBREAK'}; | |
2353 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'E_Base_GAZ'}] | |
2354 | = $wb_actions{'WB_NOBREAK'}; | |
a9256a75 | 2355 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'ExtPict_XX'}] |
c0734505 | 2356 | = $wb_actions{'WB_NOBREAK'}; |
a9256a75 | 2357 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'ExtPict_LE'}] |
c0734505 | 2358 | = $wb_actions{'WB_NOBREAK'}; |
b0e24409 | 2359 | |
d21ae9f6 | 2360 | # Break before and after newlines |
7e54b87f KW |
2361 | # WB3b ÷ (Newline | CR | LF) |
2362 | # WB3a (Newline | CR | LF) ÷ | |
2363 | # et. al. | |
289ce9cc | 2364 | for my $i ('CR', 'LF', 'Newline', 'Perl_Tailored_HSpace') { |
7e54b87f KW |
2365 | for my $j (0 .. @wb_table - 1) { |
2366 | $wb_table[$j][$wb_enums{$i}] = $wb_actions{'WB_BREAKABLE'}; | |
2367 | $wb_table[$wb_enums{$i}][$j] = $wb_actions{'WB_BREAKABLE'}; | |
2368 | } | |
2369 | } | |
2370 | ||
2371 | # But do not break within white space. | |
2372 | # WB3 CR × LF | |
2373 | # et.al. | |
289ce9cc KW |
2374 | for my $i ('CR', 'LF', 'Newline', 'Perl_Tailored_HSpace') { |
2375 | for my $j ('CR', 'LF', 'Newline', 'Perl_Tailored_HSpace') { | |
7e54b87f KW |
2376 | $wb_table[$wb_enums{$i}][$wb_enums{$j}] = $wb_actions{'WB_NOBREAK'}; |
2377 | } | |
2378 | } | |
2379 | ||
b0e24409 | 2380 | # And do not break horizontal space followed by Extend or Format or ZWJ |
289ce9cc KW |
2381 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}][$wb_enums{'Extend'}] |
2382 | = $wb_actions{'WB_NOBREAK'}; | |
2383 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}][$wb_enums{'Format'}] | |
2384 | = $wb_actions{'WB_NOBREAK'}; | |
b0e24409 KW |
2385 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}][$wb_enums{'ZWJ'}] |
2386 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2387 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}] |
2388 | [$wb_enums{'Perl_Tailored_HSpace'}] | |
2389 | = $wb_actions{'WB_hs_then_hs'}; | |
7e54b87f | 2390 | |
b0e24409 KW |
2391 | # Break at the start and end of text, unless the text is empty |
2392 | # WB2 Any ÷ eot | |
2393 | # WB1 sot ÷ Any | |
7e54b87f | 2394 | for my $i (0 .. @wb_table - 1) { |
289ce9cc KW |
2395 | $wb_table[$i][$wb_enums{'EDGE'}] = $wb_actions{'WB_BREAKABLE'}; |
2396 | $wb_table[$wb_enums{'EDGE'}][$i] = $wb_actions{'WB_BREAKABLE'}; | |
7e54b87f | 2397 | } |
289ce9cc | 2398 | $wb_table[$wb_enums{'EDGE'}][$wb_enums{'EDGE'}] = 0; |
7e54b87f | 2399 | |
289ce9cc KW |
2400 | output_table_common('WB', \%wb_actions, |
2401 | \@wb_table, \@wb_short_enums, \%wb_abbreviations); | |
7e54b87f KW |
2402 | } |
2403 | ||
4eea95a6 KW |
2404 | sub sanitize_name ($) { |
2405 | # Change the non-word characters in the input string to standardized word | |
2406 | # equivalents | |
2407 | # | |
2408 | my $sanitized = shift; | |
2409 | $sanitized =~ s/=/__/; | |
2410 | $sanitized =~ s/&/_AMP_/; | |
2411 | $sanitized =~ s/\./_DOT_/; | |
2412 | $sanitized =~ s/-/_MINUS_/; | |
2413 | $sanitized =~ s!/!_SLASH_!; | |
2414 | ||
2415 | return $sanitized; | |
2416 | } | |
2417 | ||
d6945b2c YO |
2418 | sub token_name |
2419 | { | |
2420 | my $name = sanitize_name(shift); | |
2421 | warn "$name contains non-word" if $name =~ /\W/; | |
2422 | ||
2423 | return "$table_name_prefix\U$name" | |
2424 | } | |
2425 | ||
cef72199 | 2426 | switch_pound_if ('ALL', 'PERL_IN_REGCOMP_C'); |
4eea95a6 | 2427 | |
9d9177be KW |
2428 | output_invlist("Latin1", [ 0, 256 ]); |
2429 | output_invlist("AboveLatin1", [ 256 ]); | |
2430 | ||
c11f14f3 KW |
2431 | if ($num_anyof_code_points == 256) { # Same as Latin1 |
2432 | print $out_fh | |
2433 | "\nstatic const UV * const InBitmap_invlist = Latin1_invlist;\n"; | |
2434 | } | |
2435 | else { | |
2436 | output_invlist("InBitmap", [ 0, $num_anyof_code_points ]); | |
2437 | } | |
2438 | ||
bffc0129 | 2439 | end_file_pound_if; |
43b443dd | 2440 | |
3f427fd9 KW |
2441 | # We construct lists for all the POSIX and backslash sequence character |
2442 | # classes in two forms: | |
2443 | # 1) ones which match only in the ASCII range | |
2444 | # 2) ones which match either in the Latin1 range, or the entire Unicode range | |
2445 | # | |
2446 | # These get compiled in, and hence affect the memory footprint of every Perl | |
2447 | # program, even those not using Unicode. To minimize the size, currently | |
2448 | # the Latin1 version is generated for the beyond ASCII range except for those | |
2449 | # lists that are quite small for the entire range, such as for \s, which is 22 | |
2450 | # UVs long plus 4 UVs (currently) for the header. | |
2451 | # | |
2452 | # To save even more memory, the ASCII versions could be derived from the | |
2453 | # larger ones at runtime, saving some memory (minus the expense of the machine | |
2454 | # instructions to do so), but these are all small anyway, so their total is | |
2455 | # about 100 UVs. | |
2456 | # | |
2457 | # In the list of properties below that get generated, the L1 prefix is a fake | |
2458 | # property that means just the Latin1 range of the full property (whose name | |
2459 | # has an X prefix instead of L1). | |
a02047bf KW |
2460 | # |
2461 | # An initial & means to use the subroutine from this file instead of an | |
2462 | # official inversion list. | |
3f427fd9 | 2463 | |
53146480 KW |
2464 | # Below is the list of property names to generate. '&' means to use the |
2465 | # subroutine to generate the inversion list instead of the generic code | |
2466 | # below. Some properties have a comma-separated list after the name, | |
2467 | # These are extra enums to add to those found in the Unicode tables. | |
2468 | no warnings 'qw'; | |
2469 | # Ignore non-alpha in sort | |
4eea95a6 KW |
2470 | my @props; |
2471 | push @props, sort { prop_name_for_cmp($a) cmp prop_name_for_cmp($b) } qw( | |
4eea95a6 | 2472 | &UpperLatin1 |
a9256a75 | 2473 | _Perl_GCB,EDGE,E_Base,E_Base_GAZ,E_Modifier,Glue_After_Zwj,LV,Prepend,Regional_Indicator,SpacingMark,ZWJ,ExtPict_XX |
de524f25 | 2474 | _Perl_LB,EDGE,Close_Parenthesis,Hebrew_Letter,Next_Line,Regional_Indicator,ZWJ,Contingent_Break,E_Base,E_Modifier,H2,H3,JL,JT,JV,Word_Joiner,East_Asian_CP,East_Asian_OP,Unassigned_Extended_Pictographic_Ideographic |
2027d365 | 2475 | _Perl_SB,EDGE,SContinue,CR,Extend,LF |
a9256a75 | 2476 | _Perl_WB,Perl_Tailored_HSpace,EDGE,UNKNOWN,CR,Double_Quote,E_Base,E_Base_GAZ,E_Modifier,Extend,Glue_After_Zwj,Hebrew_Letter,LF,MidNumLet,Newline,Regional_Indicator,Single_Quote,ZWJ,ExtPict_XX,ExtPict_LE |
4eea95a6 KW |
2477 | _Perl_SCX,Latin,Inherited,Unknown,Kore,Jpan,Hanb,INVALID |
2478 | Lowercase_Mapping | |
2479 | Titlecase_Mapping | |
2480 | Uppercase_Mapping | |
2481 | Simple_Case_Folding | |
2482 | Case_Folding | |
2483 | &_Perl_IVCF | |
a2aeff50 | 2484 | &_Perl_CCC_non0_non230 |
4eea95a6 KW |
2485 | ); |
2486 | # NOTE that the convention is that extra enum values come | |
2487 | # after the property name, separated by commas, with the enums | |
cf2cd619 KW |
2488 | # that aren't ever defined by Unicode (with some exceptions) |
2489 | # containing at least 4 all-uppercase characters. | |
2490 | ||
2491 | # Some of the enums are current official property values that | |
2492 | # are needed for the rules in constructing certain tables in | |
2493 | # this file, and perhaps in regexec.c as well. They are here | |
2494 | # so that things don't crash when compiled on earlier Unicode | |
2495 | # releases where they don't exist. Thus the rules that use | |
2496 | # them still get compiled, but no code point actually uses | |
2497 | # them, hence they won't get exercized on such Unicode | |
2498 | # versions, but the code will still compile and run, though | |
2499 | # may not give the precise results that those versions would | |
2500 | # expect, but reasonable results nonetheless. | |
2501 | # | |
2502 | # Other enums are due to the fact that Unicode has in more | |
2503 | # recent versions added criteria to the rules in these extra | |
2504 | # tables that are based on factors outside the property | |
2505 | # values. And those have to be accounted for, essentially by | |
2506 | # here splitting certain enum equivalence classes based on | |
2507 | # those extra rules. | |
2508 | # | |
2509 | # EDGE is supposed to be a boundary between some types of | |
2510 | # enums, but khw thinks that isn't valid any more. | |
4eea95a6 KW |
2511 | |
2512 | my @bin_props; | |
1aefa327 | 2513 | my @perl_prop_synonyms; |
4eea95a6 | 2514 | my %enums; |
2d74dcf2 KW |
2515 | my @deprecated_messages = ""; # Element [0] is a placeholder |
2516 | my %deprecated_tags; | |
4eea95a6 | 2517 | |
27097618 KW |
2518 | my $float_e_format = qr/ ^ -? \d \. \d+ e [-+] \d+ $ /x; |
2519 | ||
2520 | # Create another hash that maps floating point x.yyEzz representation to what | |
2521 | # %stricter_to_file_of does for the equivalent rational. A typical entry in | |
2522 | # the latter hash is | |
2523 | # | |
2524 | # 'nv=1/2' => 'Nv/1_2', | |
2525 | # | |
2526 | # From that, this loop creates an entry | |
2527 | # | |
2528 | # 'nv=5.00e-01' => 'Nv/1_2', | |
2529 | # | |
2530 | # %stricter_to_file_of contains far more than just the rationals. Instead we | |
048bdb72 | 2531 | # use %Unicode::UCD::nv_floating_to_rational which should have an entry for each |
27097618 KW |
2532 | # nv in the former hash. |
2533 | my %floating_to_file_of; | |
048bdb72 KW |
2534 | foreach my $key (keys %Unicode::UCD::nv_floating_to_rational) { |
2535 | my $value = $Unicode::UCD::nv_floating_to_rational{$key}; | |
2536 | $floating_to_file_of{$key} = $Unicode::UCD::stricter_to_file_of{"nv=$value"}; | |
27097618 KW |
2537 | } |
2538 | ||
2cd613ec KW |
2539 | # Properties that are specified with a prop=value syntax |
2540 | my @equals_properties; | |
2541 | ||
4eea95a6 KW |
2542 | # Collect all the binary properties from data in lib/unicore |
2543 | # Sort so that complements come after the main table, and the shortest | |
8091afe3 | 2544 | # names first, finally alphabetically. Also, sort together the tables we want |
f81c4731 KW |
2545 | # to be kept together, and prefer those with 'posix' in their names, which is |
2546 | # what the C code is expecting their names to be. | |
4eea95a6 | 2547 | foreach my $property (sort |
2d74dcf2 | 2548 | { exists $keep_together{lc $b} <=> exists $keep_together{lc $a} |
f81c4731 KW |
2549 | or $b =~ /posix/i <=> $a =~ /posix/i |
2550 | or $b =~ /perl/i <=> $a =~ /perl/i | |
27097618 | 2551 | or $a =~ $float_e_format <=> $b =~ $float_e_format |
2d74dcf2 | 2552 | or $a =~ /!/ <=> $b =~ /!/ |
4eea95a6 KW |
2553 | or length $a <=> length $b |
2554 | or $a cmp $b | |
048bdb72 KW |
2555 | } keys %Unicode::UCD::loose_to_file_of, |
2556 | keys %Unicode::UCD::stricter_to_file_of, | |
27097618 | 2557 | keys %floating_to_file_of |
53146480 | 2558 | ) { |
0f5e3c71 | 2559 | |
4eea95a6 KW |
2560 | # These two hashes map properties to values that can be considered to |
2561 | # be checksums. If two properties have the same checksum, they have | |
2562 | # identical entries. Otherwise they differ in some way. | |
048bdb72 KW |
2563 | my $tag = $Unicode::UCD::loose_to_file_of{$property}; |
2564 | $tag = $Unicode::UCD::stricter_to_file_of{$property} unless defined $tag; | |
27097618 | 2565 | $tag = $floating_to_file_of{$property} unless defined $tag; |
4eea95a6 KW |
2566 | |
2567 | # The tag may contain an '!' meaning it is identical to the one formed | |
394d2d3f KW |
2568 | # by removing the !, except that it is inverted. |
2569 | my $inverted = $tag =~ s/!//; | |
4eea95a6 | 2570 | |
27097618 KW |
2571 | # This hash is lacking the property name |
2572 | $property = "nv=$property" if $property =~ $float_e_format; | |
2573 | ||
4eea95a6 KW |
2574 | # The list of 'prop=value' entries that this single entry expands to |
2575 | my @this_entries; | |
2576 | ||
2577 | # Split 'property=value' on the equals sign, with $lhs being the whole | |
2578 | # thing if there is no '=' | |
2579 | my ($lhs, $rhs) = $property =~ / ( [^=]* ) ( =? .*) /x; | |
2580 | ||
62e88327 | 2581 | # $lhs then becomes the property name. |
2cd613ec KW |
2582 | my $prop_value = $rhs =~ s/ ^ = //rx; |
2583 | ||
2584 | push @equals_properties, $lhs if $prop_value ne ""; | |
62e88327 KW |
2585 | |
2586 | # See if there are any synonyms for this property. | |
394d2d3f KW |
2587 | if (exists $prop_name_aliases{$lhs}) { |
2588 | ||
2589 | # If so, do the combinatorics so that a new entry is added for | |
2590 | # each legal property combined with the property value (which is | |
2591 | # $rhs) | |
2592 | foreach my $alias (@{$prop_name_aliases{$lhs}}) { | |
2593 | ||
2594 | # But, there are some ambiguities, like 'script' is a synonym | |
2595 | # for 'sc', and 'sc' can stand alone, meaning something | |
2596 | # entirely different than 'script'. 'script' cannot stand | |
2597 | # alone. Don't add if the potential new lhs is in the hash of | |
2598 | # stand-alone properties. | |
2599 | no warnings 'once'; | |
2600 | next if $rhs eq "" && grep { $alias eq $_ } | |
048bdb72 | 2601 | keys %Unicode::UCD::loose_property_to_file_of; |
394d2d3f KW |
2602 | |
2603 | my $new_entry = $alias . $rhs; | |
e498c235 | 2604 | push @this_entries, $new_entry; |
394d2d3f KW |
2605 | } |
2606 | } | |
2607 | ||
2608 | # Above, we added the synonyms for the base entry we're now | |
2609 | # processing. But we haven't dealt with it yet. If we already have a | |
2610 | # property with the identical characteristics, this becomes just a | |
2611 | # synonym for it. | |
62e88327 | 2612 | |
394d2d3f KW |
2613 | if (exists $enums{$tag}) { |
2614 | push @this_entries, $property; | |
2615 | } | |
2616 | else { # Otherwise, create a new entry. | |
2617 | ||
4eea95a6 KW |
2618 | # Add to the list of properties to generate inversion lists for. |
2619 | push @bin_props, uc $property; | |
2620 | ||
394d2d3f | 2621 | # Create a rule for the parser |
f4b10e8e KW |
2622 | if (! exists $keywords{$property}) { |
2623 | $keywords{$property} = token_name($property); | |
2624 | } | |
394d2d3f | 2625 | |
4eea95a6 KW |
2626 | # And create an enum for it. |
2627 | $enums{$tag} = $table_name_prefix . uc sanitize_name($property); | |
394d2d3f | 2628 | |
1aefa327 KW |
2629 | $perl_tags{$tag} = 1 if exists $keep_together{lc $property}; |
2630 | ||
394d2d3f KW |
2631 | # Some properties are deprecated. This hash tells us so, and the |
2632 | # warning message to raise if they are used. | |
048bdb72 | 2633 | if (exists $Unicode::UCD::why_deprecated{$tag}) { |
394d2d3f | 2634 | $deprecated_tags{$enums{$tag}} = scalar @deprecated_messages; |
048bdb72 | 2635 | push @deprecated_messages, $Unicode::UCD::why_deprecated{$tag}; |
394d2d3f KW |
2636 | } |
2637 | ||
2638 | # Our sort above should have made sure that we see the | |
2639 | # non-inverted version first, but this makes sure. | |
2640 | warn "$property is inverted!!!" if $inverted; | |
2641 | } | |
2642 | ||
2643 | # Everything else is #defined to be the base enum, inversion is | |
2644 | # indicated by negating the value. | |
2645 | my $defined_to = ""; | |
2646 | $defined_to .= "-" if $inverted; | |
2647 | $defined_to .= $enums{$tag}; | |
2648 | ||
2649 | # Go through the entries that evaluate to this. | |
e498c235 | 2650 | @this_entries = uniques @this_entries; |
394d2d3f KW |
2651 | foreach my $define (@this_entries) { |
2652 | ||
2653 | # There is a rule for the parser for each. | |
f4b10e8e | 2654 | $keywords{$define} = $defined_to; |
1aefa327 KW |
2655 | |
2656 | # And a #define for all simple names equivalent to a perl property, | |
2657 | # except those that begin with 'is' or 'in'; | |
2658 | if (exists $perl_tags{$tag} && $property !~ / ^ i[ns] | = /x) { | |
7e9b4fe4 | 2659 | push @perl_prop_synonyms, "#define " |
e5360b12 KW |
2660 | . $table_name_prefix |
2661 | . uc(sanitize_name($define)) | |
2662 | . " $defined_to"; | |
1aefa327 | 2663 | } |
4eea95a6 KW |
2664 | } |
2665 | } | |
2666 | ||
cf2cd619 | 2667 | @bin_props = sort { exists $keep_together{lc $b} <=> exists $keep_together{lc $a} |
2d74dcf2 | 2668 | or $a cmp $b |
4eea95a6 | 2669 | } @bin_props; |
1aefa327 | 2670 | @perl_prop_synonyms = sort(uniques(@perl_prop_synonyms)); |
4eea95a6 KW |
2671 | push @props, @bin_props; |
2672 | ||
2673 | foreach my $prop (@props) { | |
2674 | ||
2675 | # For the Latin1 properties, we change to use the eXtended version of the | |
2676 | # base property, then go through the result and get rid of everything not | |
2677 | # in Latin1 (above 255). Actually, we retain the element for the range | |
2678 | # that crosses the 255/256 boundary if it is one that matches the | |
2679 | # property. For example, in the Word property, there is a range of code | |
2680 | # points that start at U+00F8 and goes through U+02C1. Instead of | |
2681 | # artificially cutting that off at 256 because 256 is the first code point | |
2682 | # above Latin1, we let the range go to its natural ending. That gives us | |
2683 | # extra information with no added space taken. But if the range that | |
2684 | # crosses the boundary is one that doesn't match the property, we don't | |
2685 | # start a new range above 255, as that could be construed as going to | |
2686 | # infinity. For example, the Upper property doesn't include the character | |
2687 | # at 255, but does include the one at 256. We don't include the 256 one. | |
2688 | my $prop_name = $prop; | |
2689 | my $is_local_sub = $prop_name =~ s/^&//; | |
2690 | my $extra_enums = ""; | |
2691 | $extra_enums = $1 if $prop_name =~ s/, ( .* ) //x; | |
2692 | my $lookup_prop = $prop_name; | |
2693 | $prop_name = sanitize_name($prop_name); | |
cf2cd619 KW |
2694 | $prop_name = $table_name_prefix . $prop_name |
2695 | if grep { lc $lookup_prop eq lc $_ } @bin_props; | |
4eea95a6 KW |
2696 | my $l1_only = ($lookup_prop =~ s/^L1Posix/XPosix/ |
2697 | or $lookup_prop =~ s/^L1//); | |
2698 | my $nonl1_only = 0; | |
2699 | $nonl1_only = $lookup_prop =~ s/^NonL1// unless $l1_only; | |
2700 | ($lookup_prop, my $has_suffixes) = $lookup_prop =~ / (.*) ( , .* )? /x; | |
2701 | ||
4761f74a KW |
2702 | for my $charset (get_supported_code_pages()) { |
2703 | @a2n = @{get_a2n($charset)}; | |
2704 | ||
0f5e3c71 | 2705 | my @invlist; |
99f21fb9 | 2706 | my @invmap; |
5c0563e7 | 2707 | my $map_format = 0;; |
99f21fb9 | 2708 | my $map_default; |
5c0563e7 KW |
2709 | my $maps_to_code_point = 0; |
2710 | my $to_adjust = 0; | |
59fc10af | 2711 | my $same_in_all_code_pages; |
0f5e3c71 | 2712 | if ($is_local_sub) { |
8843f0de | 2713 | my @return = eval $lookup_prop; |
289ce9cc | 2714 | die $@ if $@; |
8843f0de KW |
2715 | my $invlist_ref = shift @return; |
2716 | @invlist = @$invlist_ref; | |
d2aadf62 KW |
2717 | if (@return) { # If has other values returned , must be an |
2718 | # inversion map | |
2719 | my $invmap_ref = shift @return; | |
2720 | @invmap = @$invmap_ref; | |
2721 | $map_format = shift @return; | |
2722 | $map_default = shift @return; | |
2723 | } | |
0f5e3c71 KW |
2724 | } |
2725 | else { | |
2726 | @invlist = prop_invlist($lookup_prop, '_perl_core_internal_ok'); | |
99f21fb9 | 2727 | if (! @invlist) { |
99f21fb9 | 2728 | |
ad85f59a KW |
2729 | # If couldn't find a non-empty inversion list, see if it is |
2730 | # instead an inversion map | |
2731 | my ($list_ref, $map_ref, $format, $default) | |
99f21fb9 | 2732 | = prop_invmap($lookup_prop, '_perl_core_internal_ok'); |
ad85f59a KW |
2733 | if (! $list_ref) { |
2734 | # An empty return here could mean an unknown property, or | |
2735 | # merely that the original inversion list is empty. Call | |
2736 | # in scalar context to differentiate | |
2737 | my $count = prop_invlist($lookup_prop, | |
2738 | '_perl_core_internal_ok'); | |
d99e65da KW |
2739 | if (defined $count) { |
2740 | # Short-circuit an empty inversion list. | |
2741 | output_invlist($prop_name, \@invlist, $charset); | |
59fc10af | 2742 | last; |
d99e65da | 2743 | } |
ad85f59a | 2744 | die "Could not find inversion list for '$lookup_prop'" |
ad85f59a KW |
2745 | } |
2746 | else { | |
18b852b3 KW |
2747 | @invlist = @$list_ref; |
2748 | @invmap = @$map_ref; | |
2749 | $map_format = $format; | |
2750 | $map_default = $default; | |
ad85f59a | 2751 | } |
99f21fb9 | 2752 | } |
0f5e3c71 | 2753 | } |
ad85f59a | 2754 | |
5c0563e7 KW |
2755 | if ($map_format) { |
2756 | $maps_to_code_point = $map_format =~ / a ($ | [^r] ) /x; | |
2757 | $to_adjust = $map_format =~ /a/; | |
2758 | } | |
2759 | ||
99f21fb9 KW |
2760 | # Re-order the Unicode code points to native ones for this platform. |
2761 | # This is only needed for code points below 256, because native code | |
2762 | # points are only in that range. For inversion maps of properties | |
2763 | # where the mappings are adjusted (format =~ /a/), this reordering | |
2764 | # could mess up the adjustment pattern that was in the input, so that | |
2765 | # has to be dealt with. | |
2766 | # | |
2767 | # And inversion maps that map to code points need to eventually have | |
2768 | # all those code points remapped to native, and it's better to do that | |
2769 | # here, going through the whole list not just those below 256. This | |
2770 | # is because some inversion maps have adjustments (format =~ /a/) | |
2771 | # which may be affected by the reordering. This code needs to be done | |
2772 | # both for when we are translating the inversion lists for < 256, and | |
2773 | # for the inversion maps for everything. By doing both in this loop, | |
2774 | # we can share that code. | |
2775 | # | |
2776 | # So, we go through everything for an inversion map to code points; | |
2777 | # otherwise, we can skip any remapping at all if we are going to | |
2778 | # output only the above-Latin1 values, or if the range spans the whole | |
2779 | # of 0..256, as the remap will also include all of 0..256 (256 not | |
2780 | # 255 because a re-ordering could cause 256 to need to be in the same | |
2781 | # range as 255.) | |
2b3e8a91 | 2782 | if ( (@invmap && $maps_to_code_point) |
e4e80abb KW |
2783 | || ( @invlist |
2784 | && $invlist[0] < 256 | |
2b3e8a91 | 2785 | && ( $invlist[0] != 0 |
e4e80abb | 2786 | || (scalar @invlist != 1 && $invlist[1] < 256)))) |
ceb1de32 | 2787 | { |
59fc10af | 2788 | $same_in_all_code_pages = 0; |
99f21fb9 | 2789 | if (! @invmap) { # Straight inversion list |
563f8b93 KW |
2790 | # Look at all the ranges that start before 257. |
2791 | my @latin1_list; | |
2792 | while (@invlist) { | |
2793 | last if $invlist[0] > 256; | |
2794 | my $upper = @invlist > 1 | |
2795 | ? $invlist[1] - 1 # In range | |
2796 | ||
2797 | # To infinity. You may want to stop much much | |
2798 | # earlier; going this high may expose perl | |
2799 | # deficiencies with very large numbers. | |
7d2c6c24 | 2800 | : 256; |
563f8b93 KW |
2801 | for my $j ($invlist[0] .. $upper) { |
2802 | push @latin1_list, a2n($j); | |
2803 | } | |
fb4554ea | 2804 | |
563f8b93 KW |
2805 | shift @invlist; # Shift off the range that's in the list |
2806 | shift @invlist; # Shift off the range not in the list | |
2807 | } | |
fb4554ea | 2808 | |
563f8b93 KW |
2809 | # Here @invlist contains all the ranges in the original that |
2810 | # start at code points above 256, and @latin1_list contains | |
2811 | # all the native code points for ranges that start with a | |
2812 | # Unicode code point below 257. We sort the latter and | |
2813 | # convert it to inversion list format. Then simply prepend it | |
2814 | # to the list of the higher code points. | |
2815 | @latin1_list = sort { $a <=> $b } @latin1_list; | |
2816 | @latin1_list = mk_invlist_from_sorted_cp_list(\@latin1_list); | |
2817 | unshift @invlist, @latin1_list; | |
99f21fb9 KW |
2818 | } |
2819 | else { # Is an inversion map | |
2820 | ||
2821 | # This is a similar procedure as plain inversion list, but has | |
2822 | # multiple buckets. A plain inversion list just has two | |
2823 | # buckets, 1) 'in' the list; and 2) 'not' in the list, and we | |
2824 | # pretty much can ignore the 2nd bucket, as it is completely | |
2825 | # defined by the 1st. But here, what we do is create buckets | |
2826 | # which contain the code points that map to each, translated | |
2827 | # to native and turned into an inversion list. Thus each | |
2828 | # bucket is an inversion list of native code points that map | |
2829 | # to it or don't map to it. We use these to create an | |
2830 | # inversion map for the whole property. | |
2831 | ||
2832 | # As mentioned earlier, we use this procedure to not just | |
2833 | # remap the inversion list to native values, but also the maps | |
2834 | # of code points to native ones. In the latter case we have | |
2835 | # to look at the whole of the inversion map (or at least to | |
2836 | # above Unicode; as the maps of code points above that should | |
2837 | # all be to the default). | |
c125794e KW |
2838 | my $upper_limit = (! $maps_to_code_point) |
2839 | ? 256 | |
2840 | : (Unicode::UCD::UnicodeVersion() eq '1.1.5') | |
2841 | ? 0xFFFF | |
2842 | : 0x10FFFF; | |
99f21fb9 KW |
2843 | |
2844 | my %mapped_lists; # A hash whose keys are the buckets. | |
2845 | while (@invlist) { | |
2846 | last if $invlist[0] > $upper_limit; | |
2847 | ||
2848 | # This shouldn't actually happen, as prop_invmap() returns | |
2849 | # an extra element at the end that is beyond $upper_limit | |
cf2cd619 KW |
2850 | die "inversion map (for $prop_name) that extends to" |
2851 | . " infinity is unimplemented" unless @invlist > 1; | |
99f21fb9 KW |
2852 | |
2853 | my $bucket; | |
2854 | ||
2855 | # A hash key can't be a ref (we are only expecting arrays | |
2856 | # of scalars here), so convert any such to a string that | |
2857 | # will be converted back later (using a vertical tab as | |
b148e8b1 | 2858 | # the separator). |
99f21fb9 | 2859 | if (ref $invmap[0]) { |
b148e8b1 | 2860 | $bucket = join "\cK", map { a2n($_) } @{$invmap[0]}; |
99f21fb9 | 2861 | } |
98a1b8f7 KW |
2862 | elsif ( $maps_to_code_point |
2863 | && $invmap[0] =~ $integer_or_float_re) | |
2864 | { | |
99f21fb9 KW |
2865 | |
2866 | # Do convert to native for maps to single code points. | |
2867 | # There are some properties that have a few outlier | |
2868 | # maps that aren't code points, so the above test | |
f4d6df29 KW |
2869 | # skips those. 0 is never remapped. |
2870 | $bucket = $invmap[0] == 0 ? 0 : a2n($invmap[0]); | |
99f21fb9 KW |
2871 | } else { |
2872 | $bucket = $invmap[0]; | |
2873 | } | |
2874 | ||
2875 | # We now have the bucket that all code points in the range | |
2876 | # map to, though possibly they need to be adjusted. Go | |
2877 | # through the range and put each translated code point in | |
2878 | # it into its bucket. | |
2879 | my $base_map = $invmap[0]; | |
2880 | for my $j ($invlist[0] .. $invlist[1] - 1) { | |
2881 | if ($to_adjust | |
2882 | # The 1st code point doesn't need adjusting | |
2883 | && $j > $invlist[0] | |
2884 | ||
2885 | # Skip any non-numeric maps: these are outliers | |
2886 | # that aren't code points. | |
98a1b8f7 | 2887 | && $base_map =~ $integer_or_float_re |
99f21fb9 KW |
2888 | |
2889 | # 'ne' because the default can be a string | |
2890 | && $base_map ne $map_default) | |
2891 | { | |
2892 | # We adjust, by incrementing each the bucket and | |
2893 | # the map. For code point maps, translate to | |
2894 | # native | |
2895 | $base_map++; | |
2896 | $bucket = ($maps_to_code_point) | |
2897 | ? a2n($base_map) | |
2898 | : $base_map; | |
2899 | } | |
2900 | ||
2901 | # Add the native code point to the bucket for the | |
2902 | # current map | |
2903 | push @{$mapped_lists{$bucket}}, a2n($j); | |
2904 | } # End of loop through all code points in the range | |
2905 | ||
2906 | # Get ready for the next range | |
2907 | shift @invlist; | |
2908 | shift @invmap; | |
2909 | } # End of loop through all ranges in the map. | |
2910 | ||
2911 | # Here, @invlist and @invmap retain all the ranges from the | |
2912 | # originals that start with code points above $upper_limit. | |
2913 | # Each bucket in %mapped_lists contains all the code points | |
2914 | # that map to that bucket. If the bucket is for a map to a | |
5174a821 KW |
2915 | # single code point, the bucket has been converted to native. |
2916 | # If something else (including multiple code points), no | |
2917 | # conversion is done. | |
99f21fb9 KW |
2918 | # |
2919 | # Now we recreate the inversion map into %xlated, but this | |
2920 | # time for the native character set. | |
2921 | my %xlated; | |
2922 | foreach my $bucket (keys %mapped_lists) { | |
2923 | ||
2924 | # Sort and convert this bucket to an inversion list. The | |
2925 | # result will be that ranges that start with even-numbered | |
2926 | # indexes will be for code points that map to this bucket; | |
2927 | # odd ones map to some other bucket, and are discarded | |
2928 | # below. | |
2929 | @{$mapped_lists{$bucket}} | |
2930 | = sort{ $a <=> $b} @{$mapped_lists{$bucket}}; | |
2931 | @{$mapped_lists{$bucket}} | |
cf2cd619 KW |
2932 | = mk_invlist_from_sorted_cp_list( |
2933 | \@{$mapped_lists{$bucket}}); | |
99f21fb9 KW |
2934 | |
2935 | # Add each even-numbered range in the bucket to %xlated; | |
2936 | # so that the keys of %xlated become the range start code | |
2937 | # points, and the values are their corresponding maps. | |
2938 | while (@{$mapped_lists{$bucket}}) { | |
2939 | my $range_start = $mapped_lists{$bucket}->[0]; | |
2940 | if ($bucket =~ /\cK/) { | |
2941 | @{$xlated{$range_start}} = split /\cK/, $bucket; | |
2942 | } | |
2943 | else { | |
e113b1b3 KW |
2944 | # If adjusting, and there is more than one thing |
2945 | # that maps to the same thing, they must be split | |
2946 | # so that later the adjusting doesn't think the | |
2947 | # subsequent items can go away because of the | |
2948 | # adjusting. | |
cf2cd619 KW |
2949 | my $range_end = ( $to_adjust |
2950 | && $bucket != $map_default) | |
2951 | ? $mapped_lists{$bucket}->[1] - 1 | |
2952 | : $range_start; | |
e113b1b3 KW |
2953 | for my $i ($range_start .. $range_end) { |
2954 | $xlated{$i} = $bucket; | |
2955 | } | |
99f21fb9 KW |
2956 | } |
2957 | shift @{$mapped_lists{$bucket}}; # Discard odd ranges | |
2958 | shift @{$mapped_lists{$bucket}}; # Get ready for next | |
2959 | # iteration | |
2960 | } | |
2961 | } # End of loop through all the buckets. | |
2962 | ||
2963 | # Here %xlated's keys are the range starts of all the code | |
2964 | # points in the inversion map. Construct an inversion list | |
2965 | # from them. | |
2966 | my @new_invlist = sort { $a <=> $b } keys %xlated; | |
2967 | ||
2968 | # If the list is adjusted, we want to munge this list so that | |
2969 | # we only have one entry for where consecutive code points map | |
2970 | # to consecutive values. We just skip the subsequent entries | |
2971 | # where this is the case. | |
2972 | if ($to_adjust) { | |
2973 | my @temp; | |
2974 | for my $i (0 .. @new_invlist - 1) { | |
2975 | next if $i > 0 | |
2976 | && $new_invlist[$i-1] + 1 == $new_invlist[$i] | |
98a1b8f7 KW |
2977 | && $xlated{$new_invlist[$i-1]} |
2978 | =~ $integer_or_float_re | |
2979 | && $xlated{$new_invlist[$i]} | |
2980 | =~ $integer_or_float_re | |
62e88327 KW |
2981 | && $xlated{$new_invlist[$i-1]} + 1 |
2982 | == $xlated{$new_invlist[$i]}; | |
99f21fb9 KW |
2983 | push @temp, $new_invlist[$i]; |
2984 | } | |
2985 | @new_invlist = @temp; | |
2986 | } | |
2987 | ||
2988 | # The inversion map comes from %xlated's values. We can | |
2989 | # unshift each onto the front of the untouched portion, in | |
2990 | # reverse order of the portion we did process. | |
2991 | foreach my $start (reverse @new_invlist) { | |
2992 | unshift @invmap, $xlated{$start}; | |
2993 | } | |
2994 | ||
cf2cd619 KW |
2995 | # Finally prepend the inversion list we have just constructed |
2996 | # to the one that contains anything we didn't process. | |
99f21fb9 KW |
2997 | unshift @invlist, @new_invlist; |
2998 | } | |
2999 | } | |
e4e80abb KW |
3000 | elsif (@invmap) { # inversion maps can't cope with this variable |
3001 | # being true, even if it could be true | |
3002 | $same_in_all_code_pages = 0; | |
3003 | } | |
59fc10af KW |
3004 | else { |
3005 | $same_in_all_code_pages = 1; | |
3006 | } | |
99f21fb9 KW |
3007 | |
3008 | # prop_invmap() returns an extra final entry, which we can now | |
3009 | # discard. | |
3010 | if (@invmap) { | |
3011 | pop @invlist; | |
3012 | pop @invmap; | |
ceb1de32 | 3013 | } |
0f5e3c71 KW |
3014 | |
3015 | if ($l1_only) { | |
99f21fb9 | 3016 | die "Unimplemented to do a Latin-1 only inversion map" if @invmap; |
0f5e3c71 KW |
3017 | for my $i (0 .. @invlist - 1 - 1) { |
3018 | if ($invlist[$i] > 255) { | |
3019 | ||
3020 | # In an inversion list, even-numbered elements give the code | |
3021 | # points that begin ranges that match the property; | |
3022 | # odd-numbered give ones that begin ranges that don't match. | |
3023 | # If $i is odd, we are at the first code point above 255 that | |
3024 | # doesn't match, which means the range it is ending does | |
cf2cd619 KW |
3025 | # match, and crosses the 255/256 boundary. We want to |
3026 | # include this ending point, so increment $i, so the | |
3027 | # splice below includes it. Conversely, if $i is even, it | |
3028 | # is the first code point above 255 that matches, which | |
3029 | # means there was no matching range that crossed the | |
3030 | # boundary, and we don't want to include this code point, | |
3031 | # so splice before it. | |
0f5e3c71 KW |
3032 | $i++ if $i % 2 != 0; |
3033 | ||
3034 | # Remove everything past this. | |
3035 | splice @invlist, $i; | |
99f21fb9 | 3036 | splice @invmap, $i if @invmap; |
0f5e3c71 KW |
3037 | last; |
3038 | } | |
0c4ecf42 KW |
3039 | } |
3040 | } | |
0f5e3c71 KW |
3041 | elsif ($nonl1_only) { |
3042 | my $found_nonl1 = 0; | |
3043 | for my $i (0 .. @invlist - 1 - 1) { | |
3044 | next if $invlist[$i] < 256; | |
3045 | ||
3046 | # Here, we have the first element in the array that indicates an | |
3047 | # element above Latin1. Get rid of all previous ones. | |
3048 | splice @invlist, 0, $i; | |
99f21fb9 | 3049 | splice @invmap, 0, $i if @invmap; |
0f5e3c71 KW |
3050 | |
3051 | # If this one's index is not divisible by 2, it means that this | |
3052 | # element is inverting away from being in the list, which means | |
99f21fb9 KW |
3053 | # all code points from 256 to this one are in this list (or |
3054 | # map to the default for inversion maps) | |
3055 | if ($i % 2 != 0) { | |
3056 | unshift @invlist, 256; | |
3057 | unshift @invmap, $map_default if @invmap; | |
3058 | } | |
0f5e3c71 | 3059 | $found_nonl1 = 1; |
3f427fd9 KW |
3060 | last; |
3061 | } | |
0f0b3751 KW |
3062 | if (! $found_nonl1) { |
3063 | warn "No non-Latin1 code points in $prop_name"; | |
3064 | output_invlist($prop_name, []); | |
3065 | last; | |
3066 | } | |
3f427fd9 | 3067 | } |
3f427fd9 | 3068 | |
cef72199 | 3069 | switch_pound_if ($prop_name, 'PERL_IN_REGCOMP_C'); |
59fc10af | 3070 | start_charset_pound_if($charset, 1) unless $same_in_all_code_pages; |
4761f74a | 3071 | |
59fc10af KW |
3072 | output_invlist($prop_name, \@invlist, ($same_in_all_code_pages) |
3073 | ? $applies_to_all_charsets_text | |
3074 | : $charset); | |
4761f74a KW |
3075 | |
3076 | if (@invmap) { | |
3077 | output_invmap($prop_name, \@invmap, $lookup_prop, $map_format, | |
3078 | $map_default, $extra_enums, $charset); | |
3079 | } | |
59fc10af KW |
3080 | |
3081 | last if $same_in_all_code_pages; | |
4761f74a | 3082 | end_charset_pound_if; |
0f5e3c71 | 3083 | } |
9d9177be KW |
3084 | } |
3085 | ||
4ef8bdf9 | 3086 | print $out_fh "\nconst char * const deprecated_property_msgs[] = {\n\t"; |
394d2d3f KW |
3087 | print $out_fh join ",\n\t", map { "\"$_\"" } @deprecated_messages; |
3088 | print $out_fh "\n};\n"; | |
3089 | ||
7a15fa9e KW |
3090 | switch_pound_if ('binary_invlist_enum', 'PERL_IN_REGCOMP_C'); |
3091 | ||
394d2d3f KW |
3092 | my @enums = sort values %enums; |
3093 | ||
3094 | # Save a copy of these before modification | |
3095 | my @invlist_names = map { "${_}_invlist" } @enums; | |
3096 | ||
3097 | # Post-process the enums for deprecated properties. | |
3098 | if (scalar keys %deprecated_tags) { | |
3099 | my $seen_deprecated = 0; | |
3100 | foreach my $enum (@enums) { | |
3101 | if (grep { $_ eq $enum } keys %deprecated_tags) { | |
3102 | ||
3103 | # Change the enum name for this deprecated property to a | |
3104 | # munged one to act as a placeholder in the typedef. Then | |
3105 | # make the real name be a #define whose value is such that | |
3106 | # its modulus with the number of enums yields the index into | |
3107 | # the table occupied by the placeholder. And so that dividing | |
3108 | # the #define value by the table length gives an index into | |
3109 | # the table of deprecation messages for the corresponding | |
3110 | # warning. | |
3111 | my $revised_enum = "${enum}_perl_aux"; | |
3112 | if (! $seen_deprecated) { | |
3113 | $seen_deprecated = 1; | |
3114 | print $out_fh "\n"; | |
3115 | } | |
3116 | print $out_fh "#define $enum ($revised_enum + (MAX_UNI_KEYWORD_INDEX * $deprecated_tags{$enum}))\n"; | |
3117 | $enum = $revised_enum; | |
3118 | } | |
3119 | } | |
3120 | } | |
3121 | ||
cf2cd619 KW |
3122 | print $out_fh "\ntypedef enum {\n\tPERL_BIN_PLACEHOLDER = 0,", |
3123 | " /* So no real value is zero */\n\t"; | |
394d2d3f KW |
3124 | print $out_fh join ",\n\t", @enums; |
3125 | print $out_fh "\n"; | |
3126 | print $out_fh "} binary_invlist_enum;\n"; | |
3127 | print $out_fh "\n#define MAX_UNI_KEYWORD_INDEX $enums[-1]\n"; | |
394d2d3f | 3128 | |
7a15fa9e KW |
3129 | switch_pound_if ('binary_property_tables', 'PERL_IN_REGCOMP_C'); |
3130 | ||
cef72199 | 3131 | output_table_header($out_fh, "UV *", "uni_prop_ptrs"); |
a3d5e31b | 3132 | print $out_fh "\tNULL,\t/* Placeholder */\n"; |
cef72199 | 3133 | print $out_fh "\t"; |
394d2d3f KW |
3134 | print $out_fh join ",\n\t", @invlist_names; |
3135 | print $out_fh "\n"; | |
cef72199 KW |
3136 | |
3137 | output_table_trailer(); | |
3138 | ||
7a15fa9e KW |
3139 | switch_pound_if ('synonym defines', 'PERL_IN_REGCOMP_C'); |
3140 | ||
cef72199 KW |
3141 | print $out_fh join "\n", "\n", |
3142 | #'# ifdef DOINIT', | |
3143 | #"\n", | |
e5360b12 | 3144 | "/* Synonyms for perl properties */", |
cef72199 KW |
3145 | @perl_prop_synonyms, |
3146 | #"\n", | |
3147 | #"# endif /* DOINIT */", | |
3148 | "\n"; | |
394d2d3f | 3149 | |
2cd613ec KW |
3150 | switch_pound_if ('Valid property_values', 'PERL_IN_REGCOMP_C'); |
3151 | ||
3152 | # Each entry is a pointer to a table of property values for some property. | |
3153 | # (Other properties may share this table. The next two data structures allow | |
3154 | # this sharing to be implemented.) | |
3155 | my @values_tables = "NULL /* Placeholder so zero index is an error */"; | |
3156 | ||
3157 | # Keys are all the values of a property, strung together. The value of each | |
3158 | # key is its index in @values_tables. This is because many properties have | |
3159 | # the same values, and this allows the data to appear just once. | |
3160 | my %joined_values; | |
3161 | ||
3162 | # #defines for indices into @values_tables, so can have synonyms resolved by | |
3163 | # the C compiler. | |
3164 | my @values_indices; | |
3165 | ||
3166 | # Go through each property which is specifiable by \p{prop=value}, and create | |
3167 | # a hash with the keys being the canonicalized short property names, and the | |
3168 | # values for each property being all possible values that it can take on. | |
3169 | # Both the full value and its short, canonicalized into lc, sans punctuation | |
3170 | # version are included. | |
3171 | my %all_values; | |
3172 | for my $property (sort { prop_name_for_cmp($a) cmp prop_name_for_cmp($b) } | |
3173 | uniques @equals_properties) | |
3174 | { | |
3175 | # Get and canonicalize the short name for this property. | |
3176 | my ($short_name) = prop_aliases($property); | |
3177 | $short_name = lc $short_name; | |
3178 | $short_name =~ s/[ _-]//g; | |
3179 | ||
3180 | # Now look at each value this property can take on | |
3181 | foreach my $value (prop_values($short_name)) { | |
3182 | ||
3183 | # And for each value, look at each synonym for it | |
3184 | foreach my $alias (prop_value_aliases($short_name, $value)) { | |
3185 | ||
3186 | # Add each synonym | |
3187 | push @{$all_values{$short_name}}, $alias; | |
3188 | ||
3189 | # As well as its canonicalized name. khw made the decision to not | |
3190 | # support the grandfathered L_ Gc property value | |
3191 | $alias = lc $alias; | |
3192 | $alias =~ s/[ _-]//g unless $alias =~ $numeric_re; | |
3193 | push @{$all_values{$short_name}}, $alias; | |
3194 | } | |
3195 | } | |
3196 | } | |
3197 | ||
3198 | # Also include the old style block names, using the recipe given in | |
3199 | # Unicode::UCD | |
3200 | foreach my $block (prop_values('block')) { | |
3201 | push @{$all_values{'blk'}}, charblock((prop_invlist("block=$block"))[0]); | |
3202 | } | |
3203 | ||
3204 | # Now create output tables for each property in @equals_properties (the keys | |
3205 | # in %all_values) each containing that property's possible values as computed | |
3206 | # just above. | |
3207 | PROPERTY: | |
3208 | for my $property (sort { prop_name_for_cmp($a) cmp prop_name_for_cmp($b) | |
3209 | or $a cmp $b } keys %all_values) | |
3210 | { | |
3211 | @{$all_values{$property}} = uniques(@{$all_values{$property}}); | |
3212 | ||
3213 | # String together the values for this property, sorted. This string forms | |
3214 | # a list definition, with each value as an entry in it, indented on a new | |
3215 | # line. The sorting is used to find properties that take on the exact | |
3216 | # same values to share this string. | |
3217 | my $joined = "\t\""; | |
3218 | $joined .= join "\",\n\t\"", | |
3219 | sort { ($a =~ $numeric_re && $b =~ $numeric_re) | |
3220 | ? eval $a <=> eval $b | |
3221 | : prop_name_for_cmp($a) cmp prop_name_for_cmp($b) | |
3222 | or $a cmp $b | |
3223 | } @{$all_values{$property}}; | |
3224 | # And add a trailing marker | |
3225 | $joined .= "\",\n\tNULL\n"; | |
3226 | ||
3227 | my $table_name = $table_name_prefix . $property . "_values"; | |
3228 | my $index_name = "${table_name}_index"; | |
3229 | ||
3230 | # Add a rule for the parser that is just an empty value. It will need to | |
3231 | # know to look up empty things in the prop_value_ptrs[] table. | |
3232 | ||
3233 | $keywords{"$property="} = $index_name; | |
3234 | if (exists $prop_name_aliases{$property}) { | |
3235 | foreach my $alias (@{$prop_name_aliases{$property}}) { | |
3236 | $keywords{"$alias="} = $index_name; | |
3237 | } | |
3238 | } | |
3239 | ||
3240 | # Also create rules for the synonyms of this property to point to the same | |
3241 | # thing | |
3242 | ||
3243 | # If this property's values are the same as one we've already computed, | |
3244 | # use that instead of creating a duplicate. But we add a #define to point | |
3245 | # to the proper one. | |
3246 | if (exists $joined_values{$joined}) { | |
3247 | push @values_indices, "#define $index_name $joined_values{$joined}\n"; | |
3248 | next PROPERTY; | |
3249 | } | |
3250 | ||
3251 | # And this property, now known to have unique values from any other seen | |
3252 | # so far is about to be pushed onto @values_tables. Its index is the | |
3253 | # current count. | |
3254 | push @values_indices, "#define $index_name " | |
3255 | . scalar @values_tables . "\n"; | |
3256 | $joined_values{$joined} = $index_name; | |
3257 | push @values_tables, $table_name; | |
3258 | ||
3259 | # Create the table for this set of values. | |
3260 | output_table_header($out_fh, "char *", $table_name); | |
3261 | print $out_fh $joined; | |
3262 | output_table_trailer(); | |
3263 | } # End of loop through the properties, and their values | |
3264 | ||
3265 | # We have completely determined the table of the unique property values | |
3266 | output_table_header($out_fh, "char * const *", | |
3267 | "${table_name_prefix}prop_value_ptrs"); | |
3268 | print $out_fh join ",\n", @values_tables; | |
3269 | print $out_fh "\n"; | |
3270 | output_table_trailer(); | |
3271 | ||
3272 | # And the #defines for the indices in it | |
3273 | print $out_fh "\n\n", join "", @values_indices; | |
3274 | ||
973a28ed KW |
3275 | switch_pound_if('Boundary_pair_tables', 'PERL_IN_REGEXEC_C'); |
3276 | ||
3277 | output_GCB_table(); | |
6b659339 | 3278 | output_LB_table(); |
7e54b87f | 3279 | output_WB_table(); |
6b659339 | 3280 | |
973a28ed KW |
3281 | end_file_pound_if; |
3282 | ||
cb2d98ed KW |
3283 | print $out_fh <<"EOF"; |
3284 | ||
3285 | /* More than one code point may have the same code point as their fold. This | |
3286 | * gives the maximum number in the current Unicode release. (The folded-to | |
3287 | * code point is not included in this count.) For example, both 'S' and | |
3288 | * \\x{17F} fold to 's', so the number for that fold is 2. Another way to | |
3289 | * look at it is the maximum length of all the IVCF_AUX_TABLE's */ | |
3290 | #define MAX_FOLD_FROMS $max_fold_froms | |
3291 | EOF | |
3292 | ||
2308ab83 | 3293 | my $sources_list = "lib/unicore/mktables.lst"; |
74e28a4a TC |
3294 | my @sources = qw(regen/mk_invlists.pl |
3295 | lib/unicore/mktables | |
3296 | lib/Unicode/UCD.pm | |
3297 | regen/charset_translations.pl | |
f7b69ff8 | 3298 | regen/mk_PL_charclass.pl |
74e28a4a | 3299 | ); |
9a3da3ad FC |
3300 | { |
3301 | # Depend on mktables’ own sources. It’s a shorter list of files than | |
3302 | # those that Unicode::UCD uses. | |
1ae6ead9 | 3303 | if (! open my $mktables_list, '<', $sources_list) { |
2308ab83 KW |
3304 | |
3305 | # This should force a rebuild once $sources_list exists | |
3306 | push @sources, $sources_list; | |
3307 | } | |
3308 | else { | |
3309 | while(<$mktables_list>) { | |
3310 | last if /===/; | |
3311 | chomp; | |
3312 | push @sources, "lib/unicore/$_" if /^[^#]/; | |
3313 | } | |
9a3da3ad FC |
3314 | } |
3315 | } | |
6b659339 KW |
3316 | |
3317 | read_only_bottom_close_and_rename($out_fh, \@sources); | |
394d2d3f | 3318 | |
21c34e97 KW |
3319 | my %name_to_index; |
3320 | for my $i (0 .. @enums - 1) { | |
3321 | my $loose_name = $enums[$i] =~ s/^$table_name_prefix//r; | |
3322 | $loose_name = lc $loose_name; | |
3323 | $loose_name =~ s/__/=/; | |
3324 | $loose_name =~ s/_dot_/./; | |
3325 | $loose_name =~ s/_slash_/\//g; | |
3326 | $name_to_index{$loose_name} = $i + 1; | |
3327 | } | |
3328 | # unsanitize, exclude &, maybe add these before sanitize | |
3329 | for my $i (0 .. @perl_prop_synonyms - 1) { | |
3330 | my $loose_name_pair = $perl_prop_synonyms[$i] =~ s/#\s*define\s*//r; | |
3331 | $loose_name_pair =~ s/\b$table_name_prefix//g; | |
3332 | $loose_name_pair = lc $loose_name_pair; | |
3333 | $loose_name_pair =~ s/__/=/g; | |
3334 | $loose_name_pair =~ s/_dot_/./g; | |
3335 | $loose_name_pair =~ s/_slash_/\//g; | |
3336 | my ($synonym, $primary) = split / +/, $loose_name_pair; | |
3337 | $name_to_index{$synonym} = $name_to_index{$primary}; | |
3338 | } | |
3339 | ||
3340 | my $uni_pl = open_new('lib/unicore/uni_keywords.pl', '>', | |
9824c081 | 3341 | {style => '*', by => 'regen/mk_invlists.pl', |
21c34e97 KW |
3342 | from => "Unicode::UCD"}); |
3343 | { | |
048bdb72 | 3344 | print $uni_pl "\%Unicode::UCD::uni_prop_ptrs_indices = (\n"; |
21c34e97 | 3345 | for my $name (sort keys %name_to_index) { |
21c34e97 KW |
3346 | print $uni_pl " '$name' => $name_to_index{$name},\n"; |
3347 | } | |
3348 | print $uni_pl ");\n\n1;\n"; | |
3349 | } | |
3350 | ||
3351 | read_only_bottom_close_and_rename($uni_pl, \@sources); | |
3352 | ||
22c7467e YO |
3353 | if (my $file= $ENV{DUMP_KEYWORDS_FILE}) { |
3354 | require Data::Dumper; | |
3355 | ||
3356 | open my $ofh, ">", $file | |
3357 | or die "Failed to open DUMP_KEYWORDS_FILE '$file' for write: $!"; | |
3358 | print $ofh Data::Dumper->new([\%keywords],['*keywords']) | |
3359 | ->Sortkeys(1)->Useqq(1)->Dump(); | |
3360 | close $ofh; | |
3361 | print "Wrote keywords to '$file'.\n"; | |
3362 | } | |
394d2d3f | 3363 | |
afde5508 | 3364 | my $keywords_fh = open_new('uni_keywords.h', '>', |
9824c081 | 3365 | {style => '*', by => 'regen/mk_invlists.pl', |
afde5508 | 3366 | from => "mph.pl"}); |
394d2d3f | 3367 | |
3ef83dc3 KW |
3368 | print $keywords_fh "\n#if defined(PERL_CORE) || defined(PERL_EXT_RE_BUILD)\n\n"; |
3369 | ||
44a605b0 YO |
3370 | my $mph= MinimalPerfectHash->new( |
3371 | source_hash => \%keywords, | |
3372 | match_name => "match_uniprop", | |
3373 | ); | |
3374 | $mph->make_mph_with_split_keys(); | |
3375 | print $keywords_fh $mph->make_algo(); | |
3376 | ||
3ef83dc3 | 3377 | print $keywords_fh "\n#endif /* #if defined(PERL_CORE) || defined(PERL_EXT_RE_BUILD) */\n"; |
afde5508 KW |
3378 | |
3379 | push @sources, 'regen/mph.pl'; | |
394d2d3f | 3380 | read_only_bottom_close_and_rename($keywords_fh, \@sources); |