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 |
99f21fb9 | 12 | ); |
3d7c117d MB |
13 | require './regen/regen_lib.pl'; |
14 | require './regen/charset_translations.pl'; | |
4eea95a6 | 15 | require './lib/unicore/Heavy.pl'; |
db95f459 | 16 | use re "/aa"; |
9d9177be KW |
17 | |
18 | # This program outputs charclass_invlists.h, which contains various inversion | |
19 | # lists in the form of C arrays that are to be used as-is for inversion lists. | |
20 | # Thus, the lists it contains are essentially pre-compiled, and need only a | |
21 | # light-weight fast wrapper to make them usable at run-time. | |
22 | ||
23 | # As such, this code knows about the internal structure of these lists, and | |
24 | # any change made to that has to be done here as well. A random number stored | |
25 | # in the headers is used to minimize the possibility of things getting | |
26 | # out-of-sync, or the wrong data structure being passed. Currently that | |
27 | # random number is: | |
99f21fb9 | 28 | |
1c7f0e60 KW |
29 | my $VERSION_DATA_STRUCTURE_TYPE = 148565664; |
30 | ||
b72e0f31 KW |
31 | # charclass_invlists.h now also contains inversion maps and enum definitions |
32 | # for those maps that have a finite number of possible values | |
99f21fb9 | 33 | |
99f21fb9 | 34 | # integer or float |
db95f459 | 35 | my $numeric_re = qr/ ^ -? \d+ (:? \. \d+ )? $ /x; |
99f21fb9 | 36 | |
f4b10e8e | 37 | my %keywords; |
cef72199 | 38 | my $table_name_prefix = "UNI_"; |
4eea95a6 | 39 | |
99f21fb9 KW |
40 | # Matches valid C language enum names: begins with ASCII alphabetic, then any |
41 | # ASCII \w | |
42 | my $enum_name_re = qr / ^ [[:alpha:]] \w* $ /ax; | |
43 | ||
9d9177be | 44 | my $out_fh = open_new('charclass_invlists.h', '>', |
74e28a4a | 45 | {style => '*', by => 'regen/mk_invlists.pl', |
9d9177be KW |
46 | from => "Unicode::UCD"}); |
47 | ||
0f8eed22 | 48 | my $in_file_pound_if = ""; |
43b443dd | 49 | |
289ce9cc KW |
50 | my $max_hdr_len = 3; # In headings, how wide a name is allowed? |
51 | ||
9d9177be KW |
52 | print $out_fh "/* See the generating file for comments */\n\n"; |
53 | ||
d74e7480 KW |
54 | # enums that should be made public |
55 | my %public_enums = ( | |
f52cc976 | 56 | _Perl_SCX => 1 |
d74e7480 KW |
57 | ); |
58 | ||
bffc0129 KW |
59 | # The symbols generated by this program are all currently defined only in a |
60 | # single dot c each. The code knows where most of them go, but this hash | |
61 | # gives overrides for the exceptions to the typical place | |
62 | my %exceptions_to_where_to_define = | |
7dddaf74 | 63 | ( |
03d17b6e | 64 | #_Perl_IVCF => 'PERL_IN_REGCOMP_C', |
bffc0129 | 65 | ); |
4761f74a | 66 | |
c0221e16 | 67 | my %where_to_define_enums = (); |
015bb97c | 68 | |
59fc10af KW |
69 | my $applies_to_all_charsets_text = "all charsets"; |
70 | ||
973a28ed KW |
71 | my %gcb_enums; |
72 | my @gcb_short_enums; | |
289ce9cc | 73 | my %gcb_abbreviations; |
6b659339 KW |
74 | my %lb_enums; |
75 | my @lb_short_enums; | |
289ce9cc | 76 | my %lb_abbreviations; |
7e54b87f KW |
77 | my %wb_enums; |
78 | my @wb_short_enums; | |
289ce9cc | 79 | my %wb_abbreviations; |
6b659339 | 80 | |
99f21fb9 KW |
81 | my @a2n; |
82 | ||
394d2d3f KW |
83 | my %prop_name_aliases; |
84 | # Invert this hash so that for each canonical name, we get a list of things | |
85 | # that map to it (excluding itself) | |
86 | foreach my $name (sort keys %utf8::loose_property_name_of) { | |
87 | my $canonical = $utf8::loose_property_name_of{$name}; | |
88 | push @{$prop_name_aliases{$canonical}}, $name if $canonical ne $name; | |
89 | } | |
90 | ||
2d74dcf2 | 91 | # Output these tables in the same vicinity as each other, so that will get |
1aefa327 KW |
92 | # paged in at about the same time. These are also assumed to be the exact |
93 | # same list as those properties used internally by perl. | |
2d74dcf2 KW |
94 | my %keep_together = ( |
95 | assigned => 1, | |
96 | ascii => 1, | |
7a6f6841 KW |
97 | upper => 1, |
98 | lower => 1, | |
99 | title => 1, | |
2d74dcf2 | 100 | cased => 1, |
7a6f6841 KW |
101 | uppercaseletter => 1, |
102 | lowercaseletter => 1, | |
103 | titlecaseletter => 1, | |
104 | casedletter => 1, | |
2d74dcf2 KW |
105 | vertspace => 1, |
106 | xposixalnum => 1, | |
107 | xposixalpha => 1, | |
108 | xposixblank => 1, | |
109 | xposixcntrl => 1, | |
110 | xposixdigit => 1, | |
111 | xposixgraph => 1, | |
112 | xposixlower => 1, | |
113 | xposixprint => 1, | |
114 | xposixpunct => 1, | |
115 | xposixspace => 1, | |
116 | xposixupper => 1, | |
117 | xposixword => 1, | |
118 | xposixxdigit => 1, | |
119 | posixalnum => 1, | |
120 | posixalpha => 1, | |
121 | posixblank => 1, | |
122 | posixcntrl => 1, | |
123 | posixdigit => 1, | |
124 | posixgraph => 1, | |
125 | posixlower => 1, | |
126 | posixprint => 1, | |
127 | posixpunct => 1, | |
128 | posixspace => 1, | |
129 | posixupper => 1, | |
130 | posixword => 1, | |
131 | posixxdigit => 1, | |
132 | _perl_any_folds => 1, | |
133 | _perl_folds_to_multi_char => 1, | |
134 | _perl_idstart => 1, | |
135 | _perl_idcont => 1, | |
136 | _perl_charname_begin => 1, | |
137 | _perl_charname_continue => 1, | |
138 | _perl_problematic_locale_foldeds_start => 1, | |
139 | _perl_problematic_locale_folds => 1, | |
140 | _perl_quotemeta => 1, | |
141 | ); | |
1aefa327 | 142 | my %perl_tags; # So can find synonyms of the above properties |
2d74dcf2 | 143 | |
2027d365 KW |
144 | my $unused_table_hdr = 'u'; # Heading for row or column for unused values |
145 | ||
99f21fb9 KW |
146 | sub uniques { |
147 | # Returns non-duplicated input values. From "Perl Best Practices: | |
148 | # Encapsulated Cleverness". p. 455 in first edition. | |
149 | ||
150 | my %seen; | |
151 | return grep { ! $seen{$_}++ } @_; | |
152 | } | |
153 | ||
154 | sub a2n($) { | |
155 | my $cp = shift; | |
156 | ||
157 | # Returns the input Unicode code point translated to native. | |
158 | ||
159 | return $cp if $cp !~ $numeric_re || $cp > 255; | |
160 | return $a2n[$cp]; | |
161 | } | |
162 | ||
bffc0129 KW |
163 | sub end_file_pound_if { |
164 | if ($in_file_pound_if) { | |
165 | print $out_fh "\n#endif\t/* $in_file_pound_if */\n"; | |
0f8eed22 | 166 | $in_file_pound_if = ""; |
bffc0129 KW |
167 | } |
168 | } | |
169 | ||
48737b77 KW |
170 | sub end_charset_pound_if { |
171 | print $out_fh "\n" . get_conditional_compile_line_end(); | |
172 | } | |
173 | ||
8ec55631 | 174 | sub switch_pound_if ($$;$) { |
bffc0129 KW |
175 | my $name = shift; |
176 | my $new_pound_if = shift; | |
8ec55631 KW |
177 | my $charset = shift; |
178 | ||
62a54bb7 | 179 | my @new_pound_if = ref ($new_pound_if) |
0f8eed22 | 180 | ? sort @$new_pound_if |
62a54bb7 | 181 | : $new_pound_if; |
bffc0129 KW |
182 | |
183 | # Switch to new #if given by the 2nd argument. If there is an override | |
184 | # for this, it instead switches to that. The 1st argument is the | |
0f8eed22 | 185 | # static's name, used only to check if there is an override for this |
8ec55631 KW |
186 | # |
187 | # The 'charset' parmameter, if present, is used to first end the charset | |
188 | # #if if we actually do a switch, and then restart it afterwards. This | |
189 | # code, then assumes that the charset #if's are enclosed in the file ones. | |
bffc0129 KW |
190 | |
191 | if (exists $exceptions_to_where_to_define{$name}) { | |
62a54bb7 | 192 | @new_pound_if = $exceptions_to_where_to_define{$name}; |
bffc0129 KW |
193 | } |
194 | ||
0f8eed22 | 195 | foreach my $element (@new_pound_if) { |
cef72199 KW |
196 | |
197 | # regcomp.c is arranged so that the tables are not compiled in | |
198 | # re_comp.c */ | |
199 | my $no_xsub = 1 if $element =~ / PERL_IN_ (?: REGCOMP ) _C /x; | |
0f8eed22 | 200 | $element = "defined($element)"; |
cef72199 | 201 | $element = "($element && ! defined(PERL_IN_XSUB_RE))" if $no_xsub; |
bffc0129 | 202 | } |
0f8eed22 | 203 | $new_pound_if = join " || ", @new_pound_if; |
bffc0129 | 204 | |
0f8eed22 KW |
205 | # Change to the new one if different from old |
206 | if ($in_file_pound_if ne $new_pound_if) { | |
207 | ||
8ec55631 KW |
208 | end_charset_pound_if() if defined $charset; |
209 | ||
0f8eed22 KW |
210 | # Exit any current #if |
211 | if ($in_file_pound_if) { | |
212 | end_file_pound_if; | |
62a54bb7 | 213 | } |
0f8eed22 KW |
214 | |
215 | $in_file_pound_if = $new_pound_if; | |
bffc0129 | 216 | print $out_fh "\n#if $in_file_pound_if\n"; |
8ec55631 KW |
217 | |
218 | start_charset_pound_if ($charset, 1) if defined $charset; | |
43b443dd KW |
219 | } |
220 | } | |
221 | ||
48737b77 KW |
222 | sub start_charset_pound_if ($;$) { |
223 | print $out_fh "\n" . get_conditional_compile_line_start(shift, shift); | |
224 | } | |
225 | ||
cef72199 KW |
226 | { # Closure |
227 | my $fh; | |
228 | my $in_doinit = 0; | |
229 | ||
230 | sub output_table_header($$$;$@) { | |
231 | ||
232 | # Output to $fh the heading for a table given by the other inputs | |
233 | ||
234 | $fh = shift; | |
235 | my ($type, # typedef of table, like UV, UV* | |
236 | $name, # name of table | |
237 | $comment, # Optional comment to put on header line | |
238 | @sizes # Optional sizes of each array index. If omitted, | |
239 | # there is a single index whose size is computed by | |
240 | # the C compiler. | |
241 | ) = @_; | |
242 | ||
243 | $type =~ s/ \s+ $ //x; | |
244 | ||
245 | # If a the typedef is a ptr, add in an extra const | |
246 | $type .= " const" if $type =~ / \* $ /x; | |
247 | ||
248 | $comment = "" unless defined $comment; | |
249 | $comment = " /* $comment */" if $comment; | |
250 | ||
251 | my $array_declaration; | |
252 | if (@sizes) { | |
253 | $array_declaration = ""; | |
254 | $array_declaration .= "[$_]" for @sizes; | |
255 | } | |
256 | else { | |
257 | $array_declaration = '[]'; | |
258 | } | |
259 | ||
260 | my $declaration = "$type ${name}$array_declaration"; | |
261 | ||
262 | # Things not matching this are static. Otherwise, it is an external | |
263 | # constant, initialized only under DOINIT. | |
264 | # | |
265 | # (Currently everything is static) | |
266 | if ($in_file_pound_if !~ / PERL_IN_ (?: ) _C /x) { | |
267 | $in_doinit = 0; | |
268 | print $fh "\nstatic const $declaration = {$comment\n"; | |
269 | } | |
270 | else { | |
271 | $in_doinit = 1; | |
272 | print $fh <<EOF; | |
273 | ||
274 | # ifndef DOINIT | |
275 | ||
276 | EXTCONST $declaration; | |
277 | ||
278 | # else | |
279 | ||
280 | EXTCONST $declaration = {$comment | |
281 | EOF | |
282 | } | |
283 | } | |
284 | ||
285 | sub output_table_trailer() { | |
286 | ||
287 | # Close out a table started by output_table_header() | |
288 | ||
289 | print $fh "};\n"; | |
290 | if ($in_doinit) { | |
291 | print $fh "\n# endif /* DOINIT */\n\n"; | |
292 | $in_doinit = 0; | |
293 | } | |
294 | } | |
295 | } # End closure | |
296 | ||
297 | ||
0c4ecf42 | 298 | sub output_invlist ($$;$) { |
9d9177be KW |
299 | my $name = shift; |
300 | my $invlist = shift; # Reference to inversion list array | |
0c4ecf42 | 301 | my $charset = shift // ""; # name of character set for comment |
9d9177be | 302 | |
76d3994c | 303 | die "No inversion list for $name" unless defined $invlist |
ad85f59a | 304 | && ref $invlist eq 'ARRAY'; |
76d3994c | 305 | |
9d9177be KW |
306 | # Output the inversion list $invlist using the name $name for it. |
307 | # It is output in the exact internal form for inversion lists. | |
308 | ||
a0316a6c KW |
309 | # Is the last element of the header 0, or 1 ? |
310 | my $zero_or_one = 0; | |
ad85f59a | 311 | if (@$invlist && $invlist->[0] != 0) { |
a0316a6c | 312 | unshift @$invlist, 0; |
9d9177be KW |
313 | $zero_or_one = 1; |
314 | } | |
315 | ||
cef72199 KW |
316 | $charset = "for $charset" if $charset; |
317 | output_table_header($out_fh, "UV", "${name}_invlist", $charset); | |
9d9177be | 318 | |
cef72199 KW |
319 | my $count = @$invlist; |
320 | print $out_fh <<EOF; | |
321 | \t$count,\t/* Number of elements */ | |
322 | \t$VERSION_DATA_STRUCTURE_TYPE, /* Version and data structure type */ | |
323 | \t$zero_or_one,\t/* 0 if the list starts at 0; | |
324 | \t\t 1 if it starts at the element beyond 0 */ | |
325 | EOF | |
9d9177be KW |
326 | |
327 | # The main body are the UVs passed in to this routine. Do the final | |
328 | # element separately | |
47d53124 KW |
329 | for my $i (0 .. @$invlist - 1) { |
330 | printf $out_fh "\t0x%X", $invlist->[$i]; | |
331 | print $out_fh "," if $i < @$invlist - 1; | |
332 | print $out_fh "\n"; | |
9d9177be KW |
333 | } |
334 | ||
cef72199 | 335 | output_table_trailer(); |
9d9177be KW |
336 | } |
337 | ||
99f21fb9 KW |
338 | sub output_invmap ($$$$$$$) { |
339 | my $name = shift; | |
340 | my $invmap = shift; # Reference to inversion map array | |
341 | my $prop_name = shift; | |
342 | my $input_format = shift; # The inversion map's format | |
343 | my $default = shift; # The property value for code points who | |
344 | # otherwise don't have a value specified. | |
345 | my $extra_enums = shift; # comma-separated list of our additions to the | |
346 | # property's standard possible values | |
347 | my $charset = shift // ""; # name of character set for comment | |
348 | ||
349 | # Output the inversion map $invmap for property $prop_name, but use $name | |
350 | # as the actual data structure's name. | |
351 | ||
352 | my $count = @$invmap; | |
353 | ||
354 | my $output_format; | |
18230d9d KW |
355 | my $invmap_declaration_type; |
356 | my $enum_declaration_type; | |
357 | my $aux_declaration_type; | |
99f21fb9 KW |
358 | my %enums; |
359 | my $name_prefix; | |
360 | ||
18230d9d | 361 | if ($input_format =~ / ^ [as] l? $ /x) { |
02f811dd KW |
362 | $prop_name = (prop_aliases($prop_name))[1] // $prop_name =~ s/^_Perl_//r; # Get full name |
363 | my $short_name = (prop_aliases($prop_name))[0] // $prop_name; | |
226b74db | 364 | my @input_enums; |
f79a09fc | 365 | |
226b74db | 366 | # Find all the possible input values. These become the enum names |
34623dbb KW |
367 | # that comprise the inversion map. For inputs that don't have sub |
368 | # lists, we can just get the unique values. Otherwise, we have to | |
369 | # expand the sublists first. | |
18230d9d | 370 | if ($input_format !~ / ^ a /x) { |
563f8b93 KW |
371 | if ($input_format ne 'sl') { |
372 | @input_enums = sort(uniques(@$invmap)); | |
373 | } | |
374 | else { | |
375 | foreach my $element (@$invmap) { | |
376 | if (ref $element) { | |
377 | push @input_enums, @$element; | |
378 | } | |
379 | else { | |
380 | push @input_enums, $element; | |
381 | } | |
34623dbb | 382 | } |
563f8b93 | 383 | @input_enums = sort(uniques(@input_enums)); |
34623dbb | 384 | } |
18230d9d | 385 | } |
6b659339 | 386 | |
226b74db | 387 | # The internal enums come last, and in the order specified. |
2027d365 KW |
388 | # |
389 | # The internal one named EDGE is also used a marker. Any ones that | |
390 | # come after it are used in the algorithms below, and so must be | |
391 | # defined, even if the release of Unicode this is being compiled for | |
392 | # doesn't use them. But since no code points are assigned to them in | |
393 | # such a release, those values will never be accessed. We collapse | |
394 | # all of them into a single placholder row and a column. The | |
395 | # algorithms below will fill in those cells with essentially garbage, | |
396 | # but they are never read, so it doesn't matter. This allows the | |
397 | # algorithm to remain the same from release to release. | |
398 | # | |
399 | # In one case, regexec.c also uses a placeholder which must be defined | |
400 | # here, and we put it in the unused row and column as its value is | |
401 | # never read. | |
402 | # | |
226b74db | 403 | my @enums = @input_enums; |
27a619f7 | 404 | my @extras; |
2027d365 KW |
405 | my @unused_enums; |
406 | my $unused_enum_value = @enums; | |
27a619f7 KW |
407 | if ($extra_enums ne "") { |
408 | @extras = split /,/, $extra_enums; | |
2027d365 | 409 | my $seen_EDGE = 0; |
226b74db KW |
410 | |
411 | # Don't add if already there. | |
412 | foreach my $this_extra (@extras) { | |
413 | next if grep { $_ eq $this_extra } @enums; | |
2027d365 KW |
414 | if ($this_extra eq 'EDGE') { |
415 | push @enums, $this_extra; | |
416 | $seen_EDGE = 1; | |
417 | } | |
418 | elsif ($seen_EDGE) { | |
419 | push @unused_enums, $this_extra; | |
420 | } | |
421 | else { | |
422 | push @enums, $this_extra; | |
423 | } | |
226b74db | 424 | } |
2027d365 KW |
425 | |
426 | @unused_enums = sort @unused_enums; | |
427 | $unused_enum_value = @enums; # All unused have the same value, | |
428 | # one beyond the final used one | |
27a619f7 | 429 | } |
289ce9cc | 430 | |
226b74db KW |
431 | # Assign a value to each element of the enum type we are creating. |
432 | # The default value always gets 0; the others are arbitrarily | |
433 | # assigned. | |
27a619f7 KW |
434 | my $enum_val = 0; |
435 | my $canonical_default = prop_value_aliases($prop_name, $default); | |
436 | $default = $canonical_default if defined $canonical_default; | |
437 | $enums{$default} = $enum_val++; | |
226b74db | 438 | |
27a619f7 KW |
439 | for my $enum (@enums) { |
440 | $enums{$enum} = $enum_val++ unless exists $enums{$enum}; | |
441 | } | |
442 | ||
226b74db | 443 | # Calculate the data for the special tables output for these properties. |
27a619f7 KW |
444 | if ($name =~ / ^ _Perl_ (?: GCB | LB | WB ) $ /x) { |
445 | ||
226b74db KW |
446 | # The data includes the hashes %gcb_enums, %lb_enums, etc. |
447 | # Similarly we calculate column headings for the tables. | |
448 | # | |
27a619f7 | 449 | # We use string evals to allow the same code to work on |
226b74db | 450 | # all the tables |
27a619f7 KW |
451 | my $type = lc $prop_name; |
452 | ||
27a619f7 KW |
453 | # Skip if we've already done this code, which populated |
454 | # this hash | |
455 | if (eval "! \%${type}_enums") { | |
456 | ||
226b74db | 457 | # For each enum in the type ... |
27a619f7 KW |
458 | foreach my $enum (sort keys %enums) { |
459 | my $value = $enums{$enum}; | |
460 | my $short; | |
461 | my $abbreviated_from; | |
462 | ||
463 | # Special case this wb property value to make the | |
464 | # name more clear | |
465 | if ($enum eq 'Perl_Tailored_HSpace') { | |
466 | $short = 'hs'; | |
467 | $abbreviated_from = $enum; | |
468 | } | |
27a619f7 | 469 | else { |
226b74db KW |
470 | |
471 | # Use the official short name, if found. | |
27a619f7 KW |
472 | ($short) = prop_value_aliases($type, $enum); |
473 | ||
226b74db KW |
474 | if (! defined $short) { |
475 | ||
476 | # But if there is no official name, use the name | |
477 | # that came from the data (if any). Otherwise, | |
478 | # the name had to come from the extras list. | |
479 | # There are two types of values in that list. | |
480 | # | |
481 | # First are those enums that are not part of the | |
482 | # property, but are defined by this code. By | |
2027d365 KW |
483 | # convention these have all-caps names. We use |
484 | # the lowercased name for these. | |
226b74db | 485 | # |
2027d365 KW |
486 | # Second are enums that are needed to get the |
487 | # algorithms below to work and/or to get regexec.c | |
488 | # to compile, but don't exist in all Unicode | |
489 | # releases. These are handled outside this loop | |
490 | # as 'unused_enums' | |
226b74db KW |
491 | if (grep { $_ eq $enum } @input_enums) { |
492 | $short = $enum | |
493 | } | |
226b74db KW |
494 | else { |
495 | $short = lc $enum; | |
496 | } | |
497 | } | |
27a619f7 KW |
498 | } |
499 | ||
500 | # If our short name is too long, or we already | |
501 | # know that the name is an abbreviation, truncate | |
502 | # to make sure it's short enough, and remember | |
226b74db KW |
503 | # that we did this so we can later add a comment in the |
504 | # generated file | |
27a619f7 KW |
505 | if ( $abbreviated_from |
506 | || length $short > $max_hdr_len) | |
507 | { | |
508 | $short = substr($short, 0, $max_hdr_len); | |
509 | $abbreviated_from = $enum | |
510 | unless $abbreviated_from; | |
511 | # If the name we are to display conflicts, try | |
512 | # another. | |
513 | while (eval "exists | |
514 | \$${type}_abbreviations{$short}") | |
515 | { | |
289ce9cc | 516 | die $@ if $@; |
256fceb3 KW |
517 | |
518 | # The increment operator on strings doesn't work | |
519 | # on those containing an '_', so just use the | |
520 | # final portion. | |
521 | my @short = split '_', $short; | |
522 | $short[-1]++; | |
523 | $short = join "_", @short; | |
289ce9cc | 524 | } |
19a5f1d5 | 525 | |
27a619f7 | 526 | eval "\$${type}_abbreviations{$short} = '$enum'"; |
19a5f1d5 | 527 | die $@ if $@; |
7e54b87f | 528 | } |
27a619f7 KW |
529 | |
530 | # Remember the mapping from the property value | |
531 | # (enum) name to its value. | |
532 | eval "\$${type}_enums{$enum} = $value"; | |
533 | die $@ if $@; | |
534 | ||
535 | # Remember the inverse mapping to the short name | |
536 | # so that we can properly label the generated | |
537 | # table's rows and columns | |
538 | eval "\$${type}_short_enums[$value] = '$short'"; | |
539 | die $@ if $@; | |
7e54b87f | 540 | } |
2027d365 KW |
541 | |
542 | # Each unused enum has the same value. They all are collapsed | |
543 | # into one row and one column, named $unused_table_hdr. | |
544 | if (@unused_enums) { | |
545 | eval "\$${type}_short_enums['$unused_enum_value'] = '$unused_table_hdr'"; | |
546 | die $@ if $@; | |
547 | ||
548 | foreach my $enum (@unused_enums) { | |
549 | eval "\$${type}_enums{$enum} = $unused_enum_value"; | |
550 | die $@ if $@; | |
551 | } | |
552 | } | |
99f21fb9 | 553 | } |
19a5f1d5 | 554 | } |
99f21fb9 | 555 | |
19a5f1d5 KW |
556 | # The short names tend to be two lower case letters, but it looks |
557 | # better for those if they are upper. XXX | |
558 | $short_name = uc($short_name) if length($short_name) < 3 | |
226b74db | 559 | || substr($short_name, 0, 1) =~ /[[:lower:]]/; |
19a5f1d5 | 560 | $name_prefix = "${short_name}_"; |
cdc243dd | 561 | |
226b74db | 562 | # Start the enum definition for this map |
f99e0590 | 563 | my @enum_definition; |
19a5f1d5 KW |
564 | my @enum_list; |
565 | foreach my $enum (keys %enums) { | |
566 | $enum_list[$enums{$enum}] = $enum; | |
99f21fb9 | 567 | } |
19a5f1d5 | 568 | foreach my $i (0 .. @enum_list - 1) { |
f99e0590 | 569 | push @enum_definition, ",\n" if $i > 0; |
34623dbb | 570 | |
19a5f1d5 | 571 | my $name = $enum_list[$i]; |
f99e0590 | 572 | push @enum_definition, "\t${name_prefix}$name = $i"; |
19a5f1d5 | 573 | } |
2027d365 KW |
574 | if (@unused_enums) { |
575 | foreach my $unused (@unused_enums) { | |
576 | push @enum_definition, | |
577 | ",\n\t${name_prefix}$unused = $unused_enum_value"; | |
578 | } | |
579 | } | |
34623dbb | 580 | |
18230d9d | 581 | # For an 'l' property, we need extra enums, because some of the |
34623dbb KW |
582 | # elements are lists. Each such distinct list is placed in its own |
583 | # auxiliary map table. Here, we go through the inversion map, and for | |
584 | # each distinct list found, create an enum value for it, numbered -1, | |
585 | # -2, .... | |
586 | my %multiples; | |
587 | my $aux_table_prefix = "AUX_TABLE_"; | |
18230d9d | 588 | if ($input_format =~ /l/) { |
34623dbb KW |
589 | foreach my $element (@$invmap) { |
590 | ||
591 | # A regular scalar is not one of the lists we're looking for | |
592 | # at this stage. | |
593 | next unless ref $element; | |
594 | ||
18230d9d KW |
595 | my $joined; |
596 | if ($input_format =~ /a/) { # These are already ordered | |
597 | $joined = join ",", @$element; | |
598 | } | |
599 | else { | |
600 | $joined = join ",", sort @$element; | |
601 | } | |
34623dbb KW |
602 | my $already_found = exists $multiples{$joined}; |
603 | ||
604 | my $i; | |
605 | if ($already_found) { # Use any existing one | |
606 | $i = $multiples{$joined}; | |
607 | } | |
608 | else { # Otherwise increment to get a new table number | |
609 | $i = keys(%multiples) + 1; | |
610 | $multiples{$joined} = $i; | |
611 | } | |
612 | ||
613 | # This changes the inversion map for this entry to not be the | |
614 | # list | |
615 | $element = "use_$aux_table_prefix$i"; | |
616 | ||
617 | # And add to the enum values | |
618 | if (! $already_found) { | |
f99e0590 | 619 | push @enum_definition, ",\n\t${name_prefix}$element = -$i"; |
34623dbb KW |
620 | } |
621 | } | |
622 | } | |
623 | ||
18230d9d | 624 | $enum_declaration_type = "${name_prefix}enum"; |
f99e0590 | 625 | |
c454388e KW |
626 | # Finished with the enum definition. Inversion map stuff is used only |
627 | # by regexec or utf-8 (if it is for code points) , unless it is in the | |
628 | # enum exception list | |
629 | my $where = (exists $where_to_define_enums{$name}) | |
630 | ? $where_to_define_enums{$name} | |
631 | : ($input_format =~ /a/) | |
632 | ? 'PERL_IN_UTF8_C' | |
633 | : 'PERL_IN_REGEXEC_C'; | |
634 | ||
8ec55631 KW |
635 | if (! exists $public_enums{$name}) { |
636 | switch_pound_if($name, $where, $charset); | |
637 | } | |
638 | else { | |
639 | end_charset_pound_if; | |
640 | end_file_pound_if; | |
641 | start_charset_pound_if($charset, 1); | |
642 | } | |
c454388e KW |
643 | |
644 | # If the enum only contains one element, that is a dummy, default one | |
f99e0590 KW |
645 | if (scalar @enum_definition > 1) { |
646 | ||
647 | # Currently unneeded | |
648 | #print $out_fh "\n#define ${name_prefix}ENUM_COUNT ", | |
649 | # ..scalar keys %enums, "\n"; | |
650 | ||
651 | if ($input_format =~ /l/) { | |
652 | print $out_fh | |
653 | "\n", | |
654 | "/* Negative enum values indicate the need to use an", | |
655 | " auxiliary table\n", | |
656 | " * consisting of the list of enums this one expands to.", | |
657 | " The absolute\n", | |
658 | " * values of the negative enums are indices into a table", | |
659 | " of the auxiliary\n", | |
660 | " * tables' addresses */"; | |
661 | } | |
662 | print $out_fh "\ntypedef enum {\n"; | |
663 | print $out_fh join "", @enum_definition; | |
664 | print $out_fh "\n"; | |
18230d9d | 665 | print $out_fh "} $enum_declaration_type;\n"; |
f99e0590 | 666 | } |
19a5f1d5 | 667 | |
8ec55631 | 668 | switch_pound_if($name, $where, $charset); |
d74e7480 | 669 | |
18230d9d KW |
670 | $invmap_declaration_type = ($input_format =~ /s/) |
671 | ? $enum_declaration_type | |
341bb5b7 | 672 | : "int"; |
18230d9d KW |
673 | $aux_declaration_type = ($input_format =~ /s/) |
674 | ? $enum_declaration_type | |
e39a4130 | 675 | : "unsigned int"; |
18230d9d | 676 | |
19a5f1d5 | 677 | $output_format = "${name_prefix}%s"; |
34623dbb KW |
678 | |
679 | # If there are auxiliary tables, output them. | |
680 | if (%multiples) { | |
681 | ||
682 | print $out_fh "\n#define HAS_${name_prefix}AUX_TABLES\n"; | |
683 | ||
684 | # Invert keys and values | |
685 | my %inverted_mults; | |
686 | while (my ($key, $value) = each %multiples) { | |
687 | $inverted_mults{$value} = $key; | |
688 | } | |
689 | ||
690 | # Output them in sorted order | |
691 | my @sorted_table_list = sort { $a <=> $b } keys %inverted_mults; | |
692 | ||
693 | # Keep track of how big each aux table is | |
694 | my @aux_counts; | |
695 | ||
696 | # Output each aux table. | |
697 | foreach my $table_number (@sorted_table_list) { | |
698 | my $table = $inverted_mults{$table_number}; | |
cef72199 KW |
699 | output_table_header($out_fh, |
700 | $aux_declaration_type, | |
701 | "$name_prefix$aux_table_prefix$table_number"); | |
34623dbb KW |
702 | |
703 | # Earlier, we joined the elements of this table together with a comma | |
704 | my @elements = split ",", $table; | |
705 | ||
706 | $aux_counts[$table_number] = scalar @elements; | |
707 | for my $i (0 .. @elements - 1) { | |
708 | print $out_fh ",\n" if $i > 0; | |
18230d9d KW |
709 | if ($input_format =~ /a/) { |
710 | printf $out_fh "\t0x%X", $elements[$i]; | |
711 | } | |
712 | else { | |
713 | print $out_fh "\t${name_prefix}$elements[$i]"; | |
714 | } | |
34623dbb | 715 | } |
cef72199 KW |
716 | |
717 | print $out_fh "\n"; | |
718 | output_table_trailer(); | |
34623dbb KW |
719 | } |
720 | ||
721 | # Output the table that is indexed by the absolute value of the | |
722 | # aux table enum and contains pointers to the tables output just | |
723 | # above | |
cef72199 KW |
724 | output_table_header($out_fh, "$aux_declaration_type *", |
725 | "${name_prefix}${aux_table_prefix}ptrs"); | |
34623dbb KW |
726 | print $out_fh "\tNULL,\t/* Placeholder */\n"; |
727 | for my $i (1 .. @sorted_table_list) { | |
728 | print $out_fh ",\n" if $i > 1; | |
729 | print $out_fh "\t$name_prefix$aux_table_prefix$i"; | |
730 | } | |
cef72199 KW |
731 | print $out_fh "\n"; |
732 | output_table_trailer(); | |
34623dbb KW |
733 | |
734 | print $out_fh | |
735 | "\n/* Parallel table to the above, giving the number of elements" | |
736 | . " in each table\n * pointed to */\n"; | |
cef72199 KW |
737 | output_table_header($out_fh, "U8", |
738 | "${name_prefix}${aux_table_prefix}lengths"); | |
34623dbb KW |
739 | print $out_fh "\t0,\t/* Placeholder */\n"; |
740 | for my $i (1 .. @sorted_table_list) { | |
741 | print $out_fh ",\n" if $i > 1; | |
742 | print $out_fh "\t$aux_counts[$i]\t/* $name_prefix$aux_table_prefix$i */"; | |
743 | } | |
cef72199 KW |
744 | print $out_fh "\n"; |
745 | output_table_trailer(); | |
34623dbb | 746 | } # End of outputting the auxiliary and associated tables |
463b4a67 KW |
747 | |
748 | # The scx property used in regexec.c needs a specialized table which | |
749 | # is most convenient to output here, while the data structures set up | |
750 | # above are still extant. This table contains the code point that is | |
751 | # the zero digit of each script, indexed by script enum value. | |
752 | if (lc $short_name eq 'scx') { | |
753 | my @decimals_invlist = prop_invlist("Numeric_Type=Decimal"); | |
754 | my %script_zeros; | |
755 | ||
756 | # Find all the decimal digits. The 0 of each range is always the | |
757 | # 0th element, except in some early Unicode releases, so check for | |
758 | # that. | |
759 | for (my $i = 0; $i < @decimals_invlist; $i += 2) { | |
760 | my $code_point = $decimals_invlist[$i]; | |
a1c8344d | 761 | next if num(chr($code_point)) ne '0'; |
463b4a67 KW |
762 | |
763 | # Turn the scripts this zero is in into a list. | |
764 | my @scripts = split ",", | |
765 | charprop($code_point, "_Perl_SCX", '_perl_core_internal_ok'); | |
766 | $code_point = sprintf("0x%x", $code_point); | |
767 | ||
768 | foreach my $script (@scripts) { | |
769 | if (! exists $script_zeros{$script}) { | |
770 | $script_zeros{$script} = $code_point; | |
771 | } | |
772 | elsif (ref $script_zeros{$script}) { | |
773 | push $script_zeros{$script}->@*, $code_point; | |
774 | } | |
775 | else { # Turn into a list if this is the 2nd zero of the | |
776 | # script | |
777 | my $existing = $script_zeros{$script}; | |
778 | undef $script_zeros{$script}; | |
779 | push $script_zeros{$script}->@*, $existing, $code_point; | |
780 | } | |
781 | } | |
782 | } | |
783 | ||
784 | # @script_zeros contains the zero, sorted by the script's enum | |
785 | # value | |
786 | my @script_zeros; | |
787 | foreach my $script (keys %script_zeros) { | |
788 | my $enum_value = $enums{$script}; | |
789 | $script_zeros[$enum_value] = $script_zeros{$script}; | |
790 | } | |
791 | ||
792 | print $out_fh | |
793 | "\n/* This table, indexed by the script enum, gives the zero" | |
794 | . " code point for that\n * script; 0 if the script has multiple" | |
795 | . " digit sequences. Scripts without a\n * digit sequence use" | |
796 | . " ASCII [0-9], hence are marked '0' */\n"; | |
cef72199 | 797 | output_table_header($out_fh, "UV", "script_zeros"); |
463b4a67 KW |
798 | for my $i (0 .. @script_zeros - 1) { |
799 | my $code_point = $script_zeros[$i]; | |
800 | if (defined $code_point) { | |
801 | $code_point = " 0" if ref $code_point; | |
802 | print $out_fh "\t$code_point"; | |
803 | } | |
804 | elsif (lc $enum_list[$i] eq 'inherited') { | |
805 | print $out_fh "\t 0"; | |
806 | } | |
807 | else { # The only digits a script without its own set accepts | |
808 | # is [0-9] | |
809 | print $out_fh "\t'0'"; | |
810 | } | |
811 | print $out_fh "," if $i < @script_zeros - 1; | |
812 | print $out_fh "\t/* $enum_list[$i] */"; | |
813 | print $out_fh "\n"; | |
814 | } | |
cef72199 | 815 | output_table_trailer(); |
463b4a67 | 816 | } # End of special handling of scx |
99f21fb9 KW |
817 | } |
818 | else { | |
819 | die "'$input_format' invmap() format for '$prop_name' unimplemented"; | |
820 | } | |
821 | ||
822 | die "No inversion map for $prop_name" unless defined $invmap | |
823 | && ref $invmap eq 'ARRAY' | |
824 | && $count; | |
825 | ||
226b74db | 826 | # Now output the inversion map proper |
cef72199 KW |
827 | $charset = "for $charset" if $charset; |
828 | output_table_header($out_fh, $invmap_declaration_type, | |
829 | "${name}_invmap", | |
830 | $charset); | |
99f21fb9 KW |
831 | |
832 | # The main body are the scalars passed in to this routine. | |
833 | for my $i (0 .. $count - 1) { | |
834 | my $element = $invmap->[$i]; | |
02f811dd | 835 | my $full_element_name = prop_value_aliases($prop_name, $element); |
18230d9d KW |
836 | if ($input_format =~ /a/ && $element !~ /\D/) { |
837 | $element = ($element == 0) | |
838 | ? 0 | |
839 | : sprintf("0x%X", $element); | |
840 | } | |
841 | else { | |
02f811dd KW |
842 | $element = $full_element_name if defined $full_element_name; |
843 | $element = $name_prefix . $element; | |
18230d9d | 844 | } |
99f21fb9 KW |
845 | print $out_fh "\t$element"; |
846 | print $out_fh "," if $i < $count - 1; | |
847 | print $out_fh "\n"; | |
848 | } | |
cef72199 | 849 | output_table_trailer(); |
99f21fb9 KW |
850 | } |
851 | ||
5a7e5385 | 852 | sub mk_invlist_from_sorted_cp_list { |
a02047bf KW |
853 | |
854 | # Returns an inversion list constructed from the sorted input array of | |
855 | # code points | |
856 | ||
857 | my $list_ref = shift; | |
858 | ||
99f21fb9 KW |
859 | return unless @$list_ref; |
860 | ||
a02047bf KW |
861 | # Initialize to just the first element |
862 | my @invlist = ( $list_ref->[0], $list_ref->[0] + 1); | |
863 | ||
864 | # For each succeeding element, if it extends the previous range, adjust | |
865 | # up, otherwise add it. | |
866 | for my $i (1 .. @$list_ref - 1) { | |
867 | if ($invlist[-1] == $list_ref->[$i]) { | |
868 | $invlist[-1]++; | |
869 | } | |
870 | else { | |
871 | push @invlist, $list_ref->[$i], $list_ref->[$i] + 1; | |
872 | } | |
873 | } | |
874 | return @invlist; | |
875 | } | |
876 | ||
877 | # Read in the Case Folding rules, and construct arrays of code points for the | |
878 | # properties we need. | |
d2aadf62 | 879 | my ($cp_ref, $folds_ref, $format, $default) = prop_invmap("Case_Folding"); |
a02047bf KW |
880 | die "Could not find inversion map for Case_Folding" unless defined $format; |
881 | die "Incorrect format '$format' for Case_Folding inversion map" | |
347b9066 KW |
882 | unless $format eq 'al' |
883 | || $format eq 'a'; | |
a02047bf KW |
884 | my @has_multi_char_fold; |
885 | my @is_non_final_fold; | |
886 | ||
887 | for my $i (0 .. @$folds_ref - 1) { | |
888 | next unless ref $folds_ref->[$i]; # Skip single-char folds | |
889 | push @has_multi_char_fold, $cp_ref->[$i]; | |
890 | ||
b6a6e956 | 891 | # Add to the non-finals list each code point that is in a non-final |
a02047bf KW |
892 | # position |
893 | for my $j (0 .. @{$folds_ref->[$i]} - 2) { | |
e498c235 | 894 | push @is_non_final_fold, $folds_ref->[$i][$j]; |
a02047bf | 895 | } |
e498c235 | 896 | @is_non_final_fold = uniques @is_non_final_fold; |
a02047bf KW |
897 | } |
898 | ||
a02047bf KW |
899 | sub _Perl_Non_Final_Folds { |
900 | @is_non_final_fold = sort { $a <=> $b } @is_non_final_fold; | |
8843f0de KW |
901 | my @return = mk_invlist_from_sorted_cp_list(\@is_non_final_fold); |
902 | return \@return; | |
a02047bf KW |
903 | } |
904 | ||
d2aadf62 KW |
905 | sub _Perl_IVCF { |
906 | ||
907 | # This creates a map of the inversion of case folding. i.e., given a | |
908 | # character, it gives all the other characters that fold to it. | |
909 | # | |
910 | # Inversion maps function kind of like a hash, with the inversion list | |
911 | # specifying the buckets (keys) and the inversion maps specifying the | |
912 | # contents of the corresponding bucket. Effectively this function just | |
913 | # swaps the keys and values of the case fold hash. But there are | |
914 | # complications. Most importantly, More than one character can each have | |
915 | # the same fold. This is solved by having a list of characters that fold | |
916 | # to a given one. | |
917 | ||
918 | my %new; | |
919 | ||
920 | # Go through the inversion list. | |
921 | for (my $i = 0; $i < @$cp_ref; $i++) { | |
922 | ||
923 | # Skip if nothing folds to this | |
924 | next if $folds_ref->[$i] == 0; | |
925 | ||
926 | # This entry which is valid from here to up (but not including) the | |
927 | # next entry is for the next $count characters, so that, for example, | |
928 | # A-Z is represented by one entry. | |
929 | my $cur_list = $cp_ref->[$i]; | |
930 | my $count = $cp_ref->[$i+1] - $cur_list; | |
931 | ||
932 | # The fold of [$i] can be not just a single character, but a sequence | |
933 | # of multiple ones. We deal with those here by just creating a string | |
934 | # consisting of them. Otherwise, we use the single code point [$i] | |
935 | # folds to. | |
936 | my $cur_map = (ref $folds_ref->[$i]) | |
937 | ? join "", map { chr } $folds_ref->[$i]->@* | |
938 | : $folds_ref->[$i]; | |
939 | ||
940 | # Expand out this range | |
941 | while ($count > 0) { | |
942 | push @{$new{$cur_map}}, $cur_list; | |
943 | ||
944 | # A multiple-character fold is a string, and shouldn't need | |
945 | # incrementing anyway | |
946 | if (ref $folds_ref->[$i]) { | |
947 | die sprintf("Case fold for %x is multiple chars; should have" | |
948 | . " a count of 1, but instead it was $count", $count) | |
949 | unless $count == 1; | |
950 | } | |
951 | else { | |
952 | $cur_map++; | |
953 | $cur_list++; | |
954 | } | |
955 | $count--; | |
956 | } | |
957 | } | |
958 | ||
959 | # Now go through and make some adjustments. We add synthetic entries for | |
960 | # two cases. | |
961 | # 1) Two or more code points can fold to the same multiple character, | |
962 | # sequence, as U+FB05 and U+FB06 both fold to 'st'. This code is only | |
963 | # for single character folds, but FB05 and FB06 are single characters | |
964 | # that are equivalent folded, so we add entries so that they are | |
965 | # considered to fold to each other | |
966 | # 2) If two or more above-Latin1 code points fold to the same Latin1 range | |
967 | # one, we also add entries so that they are considered to fold to each | |
968 | # other. This is so that under /aa or /l matching, where folding to | |
969 | # their Latin1 range code point is illegal, they still can fold to each | |
970 | # other. This situation happens in Unicode 3.0.1, but probably no | |
971 | # other version. | |
972 | foreach my $fold (keys %new) { | |
db95f459 | 973 | my $folds_to_string = $fold =~ /\D/; |
d2aadf62 KW |
974 | |
975 | # If the bucket contains only one element, convert from an array to a | |
976 | # scalar | |
977 | if (scalar $new{$fold}->@* == 1) { | |
978 | $new{$fold} = $new{$fold}[0]; | |
979 | } | |
980 | else { | |
981 | ||
982 | # Otherwise, sort numerically. This places the highest code point | |
983 | # in the list at the tail end. This is because Unicode keeps the | |
984 | # lowercase code points as higher ordinals than the uppercase, at | |
985 | # least for the ones that matter so far. These are synthetic | |
986 | # entries, and we want to predictably have the lowercase (which is | |
987 | # more likely to be what gets folded to) in the same corresponding | |
988 | # position, so that other code can rely on that. If some new | |
989 | # version of Unicode came along that violated this, we might have | |
990 | # to change so that the sort is based on upper vs lower instead. | |
991 | # (The lower-comes-after isn't true of native EBCDIC, but here we | |
992 | # are dealing strictly with Unicode values). | |
993 | @{$new{$fold}} = sort { $a <=> $b } $new{$fold}->@* | |
994 | unless $folds_to_string; | |
995 | # We will be working with a copy of this sorted entry. | |
996 | my @source_list = $new{$fold}->@*; | |
997 | if (! $folds_to_string) { | |
998 | ||
999 | # This handles situation 2) listed above, which only arises if | |
1000 | # what is being folded-to (the fold) is in the Latin1 range. | |
1001 | if ($fold > 255 ) { | |
1002 | undef @source_list; | |
1003 | } | |
1004 | else { | |
1005 | # And it only arises if there are two or more folders that | |
1006 | # fold to it above Latin1. We look at just those. | |
1007 | @source_list = grep { $_ > 255 } @source_list; | |
1008 | undef @source_list if @source_list == 1; | |
1009 | } | |
1010 | } | |
1011 | ||
1012 | # Here, we've found the items we want to set up synthetic folds | |
1013 | # for. Add entries so that each folds to each other. | |
1014 | foreach my $cp (@source_list) { | |
1015 | my @rest = grep { $cp != $_ } @source_list; | |
1016 | if (@rest == 1) { | |
1017 | $new{$cp} = $rest[0]; | |
1018 | } | |
1019 | else { | |
1020 | push @{$new{$cp}}, @rest; | |
1021 | } | |
1022 | } | |
1023 | } | |
1024 | ||
1025 | # We don't otherwise deal with multiple-character folds | |
1026 | delete $new{$fold} if $folds_to_string; | |
1027 | } | |
1028 | ||
1029 | ||
1030 | # Now we have a hash that is the inversion of the case fold property. | |
1031 | # Convert it to an inversion map. | |
1032 | ||
1033 | my @sorted_folds = sort { $a <=> $b } keys %new; | |
1034 | my (@invlist, @invmap); | |
1035 | ||
1036 | # We know that nothing folds to the controls (whose ordinals start at 0). | |
1037 | # And the first real entries are the lowest in the hash. | |
1038 | push @invlist, 0, $sorted_folds[0]; | |
1039 | push @invmap, 0, $new{$sorted_folds[0]}; | |
1040 | ||
1041 | # Go through the remainder of the hash keys (which are the folded code | |
1042 | # points) | |
1043 | for (my $i = 1; $i < @sorted_folds; $i++) { | |
1044 | ||
1045 | # Get the current one, and the one prior to it. | |
1046 | my $fold = $sorted_folds[$i]; | |
1047 | my $prev_fold = $sorted_folds[$i-1]; | |
1048 | ||
1049 | # If the current one is not just 1 away from the prior one, we close | |
1050 | # out the range containing the previous fold, and know that the gap | |
1051 | # doesn't have anything that folds. | |
1052 | if ($fold - 1 != $prev_fold) { | |
1053 | push @invlist, $prev_fold + 1; | |
1054 | push @invmap, 0; | |
1055 | ||
1056 | # And start a new range | |
1057 | push @invlist, $fold; | |
1058 | push @invmap, $new{$fold}; | |
1059 | } | |
1060 | elsif ($new{$fold} - 1 != $new{$prev_fold}) { | |
1061 | ||
1062 | # Here the current fold is just 1 greater than the previous, but | |
1063 | # the new map isn't correspondingly 1 greater than the previous, | |
1064 | # the old range is ended, but since there is no gap, we don't have | |
1065 | # to insert anything else. | |
1066 | push @invlist, $fold; | |
1067 | push @invmap, $new{$fold}; | |
1068 | ||
1069 | } # else { Otherwise, this new entry just extends the previous } | |
1070 | ||
1071 | die "In IVCF: $invlist[-1] <= $invlist[-2]" | |
1072 | if $invlist[-1] <= $invlist[-2]; | |
1073 | } | |
1074 | ||
1075 | # And add an entry that indicates that everything above this, to infinity, | |
1076 | # does not have a case fold. | |
1077 | push @invlist, $sorted_folds[-1] + 1; | |
1078 | push @invmap, 0; | |
1079 | ||
1080 | # All Unicode versions have some places where multiple code points map to | |
1081 | # the same one, so the format always has an 'l' | |
1082 | return \@invlist, \@invmap, 'al', $default; | |
1083 | } | |
1084 | ||
99f21fb9 KW |
1085 | sub prop_name_for_cmp ($) { # Sort helper |
1086 | my $name = shift; | |
1087 | ||
1088 | # Returns the input lowercased, with non-alphas removed, as well as | |
1089 | # everything starting with a comma | |
1090 | ||
1091 | $name =~ s/,.*//; | |
1092 | $name =~ s/[[:^alpha:]]//g; | |
1093 | return lc $name; | |
1094 | } | |
1095 | ||
892d8259 | 1096 | sub UpperLatin1 { |
8843f0de KW |
1097 | my @return = mk_invlist_from_sorted_cp_list([ 128 .. 255 ]); |
1098 | return \@return; | |
892d8259 KW |
1099 | } |
1100 | ||
289ce9cc KW |
1101 | sub output_table_common { |
1102 | ||
1103 | # Common subroutine to actually output the generated rules table. | |
1104 | ||
1105 | my ($property, | |
1106 | $table_value_defines_ref, | |
1107 | $table_ref, | |
1108 | $names_ref, | |
1109 | $abbreviations_ref) = @_; | |
1110 | my $size = @$table_ref; | |
1111 | ||
1112 | # Output the #define list, sorted by numeric value | |
1113 | if ($table_value_defines_ref) { | |
1114 | my $max_name_length = 0; | |
1115 | my @defines; | |
1116 | ||
1117 | # Put in order, and at the same time find the longest name | |
1118 | while (my ($enum, $value) = each %$table_value_defines_ref) { | |
1119 | $defines[$value] = $enum; | |
1120 | ||
1121 | my $length = length $enum; | |
1122 | $max_name_length = $length if $length > $max_name_length; | |
1123 | } | |
1124 | ||
1125 | print $out_fh "\n"; | |
1126 | ||
1127 | # Output, so that the values are vertically aligned in a column after | |
1128 | # the longest name | |
1129 | foreach my $i (0 .. @defines - 1) { | |
1130 | next unless defined $defines[$i]; | |
1131 | printf $out_fh "#define %-*s %2d\n", | |
1132 | $max_name_length, | |
1133 | $defines[$i], | |
1134 | $i; | |
1135 | } | |
1136 | } | |
1137 | ||
1138 | my $column_width = 2; # We currently allow 2 digits for the number | |
1139 | ||
1140 | # If the maximum value in the table is 1, it can be a bool. (Being above | |
1141 | # a U8 is not currently handled | |
1142 | my $max_element = 0; | |
1143 | for my $i (0 .. $size - 1) { | |
1144 | for my $j (0 .. $size - 1) { | |
1145 | next if $max_element >= $table_ref->[$i][$j]; | |
1146 | $max_element = $table_ref->[$i][$j]; | |
1147 | } | |
1148 | } | |
1149 | die "Need wider table column width given '$max_element" | |
1150 | if length $max_element > $column_width; | |
1151 | ||
1152 | my $table_type = ($max_element == 1) | |
1153 | ? 'bool' | |
1154 | : 'U8'; | |
1155 | ||
1156 | # If a name is longer than the width set aside for a column, its column | |
1157 | # needs to have increased spacing so that the name doesn't get truncated | |
1158 | # nor run into an adjacent column | |
1159 | my @spacers; | |
1160 | ||
2027d365 KW |
1161 | # Is there a row and column for unused values in this release? |
1162 | my $has_unused = $names_ref->[$size-1] eq $unused_table_hdr; | |
289ce9cc KW |
1163 | |
1164 | for my $i (0 .. $size - 1) { | |
1165 | no warnings 'numeric'; | |
289ce9cc KW |
1166 | $spacers[$i] = " " x (length($names_ref->[$i]) - $column_width); |
1167 | } | |
1168 | ||
cef72199 | 1169 | output_table_header($out_fh, $table_type, "${property}_table", undef, $size, $size); |
289ce9cc KW |
1170 | |
1171 | # Calculate the column heading line | |
1172 | my $header_line = "/* " | |
1173 | . (" " x $max_hdr_len) # We let the row heading meld to | |
1174 | # the '*/' for those that are at | |
1175 | # the max | |
1176 | . " " x 3; # Space for '*/ ' | |
1177 | # Now each column | |
1178 | for my $i (0 .. $size - 1) { | |
1179 | $header_line .= sprintf "%s%*s", | |
1180 | $spacers[$i], | |
1181 | $column_width + 1, # 1 for the ',' | |
1182 | $names_ref->[$i]; | |
1183 | } | |
1184 | $header_line .= " */\n"; | |
1185 | ||
1186 | # If we have annotations, output it now. | |
2027d365 | 1187 | if ($has_unused || scalar %$abbreviations_ref) { |
289ce9cc KW |
1188 | my $text = ""; |
1189 | foreach my $abbr (sort keys %$abbreviations_ref) { | |
1190 | $text .= "; " if $text; | |
1191 | $text .= "'$abbr' stands for '$abbreviations_ref->{$abbr}'"; | |
1192 | } | |
2027d365 KW |
1193 | if ($has_unused) { |
1194 | $text .= "; $unused_table_hdr stands for 'unused in this Unicode" | |
1195 | . " release (and the data in the row or column are garbage)" | |
289ce9cc KW |
1196 | } |
1197 | ||
1198 | my $indent = " " x 3; | |
1199 | $text = $indent . "/* $text */"; | |
1200 | ||
1201 | # Wrap the text so that it is no wider than the table, which the | |
1202 | # header line gives. | |
1203 | my $output_width = length $header_line; | |
1204 | while (length $text > $output_width) { | |
1205 | my $cur_line = substr($text, 0, $output_width); | |
1206 | ||
1207 | # Find the first blank back from the right end to wrap at. | |
1208 | for (my $i = $output_width -1; $i > 0; $i--) { | |
1209 | if (substr($text, $i, 1) eq " ") { | |
1210 | print $out_fh substr($text, 0, $i), "\n"; | |
1211 | ||
1212 | # Set so will look at just the remaining tail (which will | |
1213 | # be indented and have a '*' after the indent | |
1214 | $text = $indent . " * " . substr($text, $i + 1); | |
1215 | last; | |
1216 | } | |
1217 | } | |
1218 | } | |
1219 | ||
1220 | # And any remaining | |
1221 | print $out_fh $text, "\n" if $text; | |
1222 | } | |
1223 | ||
1224 | # We calculated the header line earlier just to get its width so that we | |
1225 | # could make sure the annotations fit into that. | |
1226 | print $out_fh $header_line; | |
1227 | ||
1228 | # Now output the bulk of the table. | |
1229 | for my $i (0 .. $size - 1) { | |
1230 | ||
1231 | # First the row heading. | |
1232 | printf $out_fh "/* %-*s*/ ", $max_hdr_len, $names_ref->[$i]; | |
1233 | print $out_fh "{"; # Then the brace for this row | |
1234 | ||
1235 | # Then each column | |
1236 | for my $j (0 .. $size -1) { | |
1237 | print $out_fh $spacers[$j]; | |
1238 | printf $out_fh "%*d", $column_width, $table_ref->[$i][$j]; | |
1239 | print $out_fh "," if $j < $size - 1; | |
1240 | } | |
1241 | print $out_fh " }"; | |
1242 | print $out_fh "," if $i < $size - 1; | |
1243 | print $out_fh "\n"; | |
1244 | } | |
1245 | ||
cef72199 | 1246 | output_table_trailer(); |
289ce9cc KW |
1247 | } |
1248 | ||
973a28ed KW |
1249 | sub output_GCB_table() { |
1250 | ||
1251 | # Create and output the pair table for use in determining Grapheme Cluster | |
1252 | # Breaks, given in http://www.unicode.org/reports/tr29/. | |
b0e24409 KW |
1253 | my %gcb_actions = ( |
1254 | GCB_NOBREAK => 0, | |
1255 | GCB_BREAKABLE => 1, | |
1256 | GCB_RI_then_RI => 2, # Rules 12 and 13 | |
1257 | GCB_EX_then_EM => 3, # Rule 10 | |
c0734505 | 1258 | GCB_Maybe_Emoji_NonBreak => 4, |
b0e24409 | 1259 | ); |
973a28ed KW |
1260 | |
1261 | # The table is constructed in reverse order of the rules, to make the | |
1262 | # lower-numbered, higher priority ones override the later ones, as the | |
1263 | # algorithm stops at the earliest matching rule | |
1264 | ||
1265 | my @gcb_table; | |
1266 | my $table_size = @gcb_short_enums; | |
1267 | ||
1268 | # Otherwise, break everywhere. | |
b0e24409 | 1269 | # GB99 Any ÷ Any |
973a28ed KW |
1270 | for my $i (0 .. $table_size - 1) { |
1271 | for my $j (0 .. $table_size - 1) { | |
1272 | $gcb_table[$i][$j] = 1; | |
1273 | } | |
1274 | } | |
1275 | ||
b0e24409 KW |
1276 | # Do not break within emoji flag sequences. That is, do not break between |
1277 | # regional indicator (RI) symbols if there is an odd number of RI | |
1278 | # characters before the break point. Must be resolved in runtime code. | |
1279 | # | |
c492f156 | 1280 | # GB12 sot (RI RI)* RI × RI |
b0e24409 KW |
1281 | # GB13 [^RI] (RI RI)* RI × RI |
1282 | $gcb_table[$gcb_enums{'Regional_Indicator'}] | |
1283 | [$gcb_enums{'Regional_Indicator'}] = $gcb_actions{GCB_RI_then_RI}; | |
1284 | ||
c0734505 KW |
1285 | # Post 11.0: GB11 \p{Extended_Pictographic} Extend* ZWJ |
1286 | # × \p{Extended_Pictographic} | |
1287 | $gcb_table[$gcb_enums{'ZWJ'}][$gcb_enums{'XPG_XX'}] = | |
1288 | $gcb_actions{GCB_Maybe_Emoji_NonBreak}; | |
1289 | ||
1290 | # This and the rule GB10 obsolete starting with Unicode 11.0, can be left | |
1291 | # in as there are no code points that match, so the code won't ever get | |
1292 | # executed. | |
b0e24409 | 1293 | # Do not break within emoji modifier sequences or emoji zwj sequences. |
c0734505 | 1294 | # Pre 11.0: GB11 ZWJ × ( Glue_After_Zwj | E_Base_GAZ ) |
b0e24409 KW |
1295 | $gcb_table[$gcb_enums{'ZWJ'}][$gcb_enums{'Glue_After_Zwj'}] = 0; |
1296 | $gcb_table[$gcb_enums{'ZWJ'}][$gcb_enums{'E_Base_GAZ'}] = 0; | |
1297 | ||
1298 | # GB10 ( E_Base | E_Base_GAZ ) Extend* × E_Modifier | |
1299 | $gcb_table[$gcb_enums{'Extend'}][$gcb_enums{'E_Modifier'}] | |
1300 | = $gcb_actions{GCB_EX_then_EM}; | |
1301 | $gcb_table[$gcb_enums{'E_Base'}][$gcb_enums{'E_Modifier'}] = 0; | |
1302 | $gcb_table[$gcb_enums{'E_Base_GAZ'}][$gcb_enums{'E_Modifier'}] = 0; | |
1303 | ||
1304 | # Do not break before extending characters or ZWJ. | |
973a28ed | 1305 | # Do not break before SpacingMarks, or after Prepend characters. |
973a28ed | 1306 | # GB9b Prepend × |
b0e24409 KW |
1307 | # GB9a × SpacingMark |
1308 | # GB9 × ( Extend | ZWJ ) | |
973a28ed | 1309 | for my $i (0 .. @gcb_table - 1) { |
289ce9cc | 1310 | $gcb_table[$gcb_enums{'Prepend'}][$i] = 0; |
b0e24409 KW |
1311 | $gcb_table[$i][$gcb_enums{'SpacingMark'}] = 0; |
1312 | $gcb_table[$i][$gcb_enums{'Extend'}] = 0; | |
1313 | $gcb_table[$i][$gcb_enums{'ZWJ'}] = 0; | |
973a28ed KW |
1314 | } |
1315 | ||
973a28ed KW |
1316 | # Do not break Hangul syllable sequences. |
1317 | # GB8 ( LVT | T) × T | |
1318 | $gcb_table[$gcb_enums{'LVT'}][$gcb_enums{'T'}] = 0; | |
1319 | $gcb_table[$gcb_enums{'T'}][$gcb_enums{'T'}] = 0; | |
1320 | ||
1321 | # GB7 ( LV | V ) × ( V | T ) | |
1322 | $gcb_table[$gcb_enums{'LV'}][$gcb_enums{'V'}] = 0; | |
1323 | $gcb_table[$gcb_enums{'LV'}][$gcb_enums{'T'}] = 0; | |
1324 | $gcb_table[$gcb_enums{'V'}][$gcb_enums{'V'}] = 0; | |
1325 | $gcb_table[$gcb_enums{'V'}][$gcb_enums{'T'}] = 0; | |
1326 | ||
1327 | # GB6 L × ( L | V | LV | LVT ) | |
1328 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'L'}] = 0; | |
1329 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'V'}] = 0; | |
1330 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'LV'}] = 0; | |
1331 | $gcb_table[$gcb_enums{'L'}][$gcb_enums{'LVT'}] = 0; | |
1332 | ||
289ce9cc KW |
1333 | # Do not break between a CR and LF. Otherwise, break before and after |
1334 | # controls. | |
973a28ed KW |
1335 | # GB5 ÷ ( Control | CR | LF ) |
1336 | # GB4 ( Control | CR | LF ) ÷ | |
1337 | for my $i (0 .. @gcb_table - 1) { | |
289ce9cc | 1338 | $gcb_table[$i][$gcb_enums{'Control'}] = 1; |
973a28ed KW |
1339 | $gcb_table[$i][$gcb_enums{'CR'}] = 1; |
1340 | $gcb_table[$i][$gcb_enums{'LF'}] = 1; | |
289ce9cc | 1341 | $gcb_table[$gcb_enums{'Control'}][$i] = 1; |
973a28ed KW |
1342 | $gcb_table[$gcb_enums{'CR'}][$i] = 1; |
1343 | $gcb_table[$gcb_enums{'LF'}][$i] = 1; | |
1344 | } | |
1345 | ||
1346 | # GB3 CR × LF | |
1347 | $gcb_table[$gcb_enums{'CR'}][$gcb_enums{'LF'}] = 0; | |
1348 | ||
b0e24409 | 1349 | # Break at the start and end of text, unless the text is empty |
973a28ed KW |
1350 | # GB1 sot ÷ |
1351 | # GB2 ÷ eot | |
1352 | for my $i (0 .. @gcb_table - 1) { | |
289ce9cc KW |
1353 | $gcb_table[$i][$gcb_enums{'EDGE'}] = 1; |
1354 | $gcb_table[$gcb_enums{'EDGE'}][$i] = 1; | |
973a28ed | 1355 | } |
289ce9cc | 1356 | $gcb_table[$gcb_enums{'EDGE'}][$gcb_enums{'EDGE'}] = 0; |
973a28ed | 1357 | |
b0e24409 | 1358 | output_table_common('GCB', \%gcb_actions, |
289ce9cc | 1359 | \@gcb_table, \@gcb_short_enums, \%gcb_abbreviations); |
973a28ed KW |
1360 | } |
1361 | ||
6b659339 KW |
1362 | sub output_LB_table() { |
1363 | ||
1364 | # Create and output the enums, #defines, and pair table for use in | |
1365 | # determining Line Breaks. This uses the default line break algorithm, | |
1366 | # given in http://www.unicode.org/reports/tr14/, but tailored by example 7 | |
1367 | # in that page, as the Unicode-furnished tests assume that tailoring. | |
1368 | ||
6b659339 KW |
1369 | # The result is really just true or false. But we follow along with tr14, |
1370 | # creating a rule which is false for something like X SP* X. That gets | |
1371 | # encoding 2. The rest of the actions are synthetic ones that indicate | |
1372 | # some context handling is required. These each are added to the | |
1373 | # underlying 0, 1, or 2, instead of replacing them, so that the underlying | |
1374 | # value can be retrieved. Actually only rules from 7 through 18 (which | |
1375 | # are the ones where space matter) are possible to have 2 added to them. | |
1376 | # The others below add just 0 or 1. It might be possible for one | |
1377 | # synthetic rule to be added to another, yielding a larger value. This | |
1378 | # doesn't happen in the Unicode 8.0 rule set, and as you can see from the | |
1379 | # names of the middle grouping below, it is impossible for that to occur | |
1380 | # for them because they all start with mutually exclusive classes. That | |
1381 | # the final rule can't be added to any of the others isn't obvious from | |
1382 | # its name, so it is assigned a power of 2 higher than the others can get | |
1383 | # to so any addition would preserve all data. (And the code will reach an | |
1384 | # assert(0) on debugging builds should this happen.) | |
1385 | my %lb_actions = ( | |
1386 | LB_NOBREAK => 0, | |
1387 | LB_BREAKABLE => 1, | |
1388 | LB_NOBREAK_EVEN_WITH_SP_BETWEEN => 2, | |
1389 | ||
b0e24409 | 1390 | LB_CM_ZWJ_foo => 3, # Rule 9 |
6b659339 KW |
1391 | LB_SP_foo => 6, # Rule 18 |
1392 | LB_PR_or_PO_then_OP_or_HY => 9, # Rule 25 | |
1393 | LB_SY_or_IS_then_various => 11, # Rule 25 | |
1394 | LB_HY_or_BA_then_foo => 13, # Rule 21 | |
b0e24409 | 1395 | LB_RI_then_RI => 15, # Rule 30a |
6b659339 | 1396 | |
b0e24409 | 1397 | LB_various_then_PO_or_PR => (1<<5), # Rule 25 |
6b659339 KW |
1398 | ); |
1399 | ||
6b659339 KW |
1400 | # Construct the LB pair table. This is based on the rules in |
1401 | # http://www.unicode.org/reports/tr14/, but modified as those rules are | |
1402 | # designed for someone taking a string of text and sequentially going | |
1403 | # through it to find the break opportunities, whereas, Perl requires | |
1404 | # determining if a given random spot is a break opportunity, without | |
1405 | # knowing all the entire string before it. | |
1406 | # | |
1407 | # The table is constructed in reverse order of the rules, to make the | |
1408 | # lower-numbered, higher priority ones override the later ones, as the | |
1409 | # algorithm stops at the earliest matching rule | |
1410 | ||
1411 | my @lb_table; | |
1412 | my $table_size = @lb_short_enums; | |
1413 | ||
1414 | # LB31. Break everywhere else | |
1415 | for my $i (0 .. $table_size - 1) { | |
1416 | for my $j (0 .. $table_size - 1) { | |
1417 | $lb_table[$i][$j] = $lb_actions{'LB_BREAKABLE'}; | |
1418 | } | |
1419 | } | |
1420 | ||
b0e24409 KW |
1421 | # LB30b Do not break between an emoji base and an emoji modifier. |
1422 | # EB × EM | |
1423 | $lb_table[$lb_enums{'E_Base'}][$lb_enums{'E_Modifier'}] | |
1424 | = $lb_actions{'LB_NOBREAK'}; | |
1425 | ||
1426 | # LB30a Break between two regional indicator symbols if and only if there | |
1427 | # are an even number of regional indicators preceding the position of the | |
1428 | # break. | |
1429 | # sot (RI RI)* RI × RI | |
1430 | # [^RI] (RI RI)* RI × RI | |
289ce9cc | 1431 | $lb_table[$lb_enums{'Regional_Indicator'}] |
b0e24409 | 1432 | [$lb_enums{'Regional_Indicator'}] = $lb_actions{'LB_RI_then_RI'}; |
6b659339 KW |
1433 | |
1434 | # LB30 Do not break between letters, numbers, or ordinary symbols and | |
1435 | # opening or closing parentheses. | |
1436 | # (AL | HL | NU) × OP | |
289ce9cc KW |
1437 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Open_Punctuation'}] |
1438 | = $lb_actions{'LB_NOBREAK'}; | |
1439 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Open_Punctuation'}] | |
1440 | = $lb_actions{'LB_NOBREAK'}; | |
1441 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Open_Punctuation'}] | |
1442 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1443 | |
1444 | # CP × (AL | HL | NU) | |
289ce9cc KW |
1445 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Alphabetic'}] |
1446 | = $lb_actions{'LB_NOBREAK'}; | |
1447 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Hebrew_Letter'}] | |
1448 | = $lb_actions{'LB_NOBREAK'}; | |
1449 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Numeric'}] | |
1450 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1451 | |
1452 | # LB29 Do not break between numeric punctuation and alphabetics (“e.g.”). | |
1453 | # IS × (AL | HL) | |
289ce9cc KW |
1454 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Alphabetic'}] |
1455 | = $lb_actions{'LB_NOBREAK'}; | |
1456 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1457 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1458 | |
1459 | # LB28 Do not break between alphabetics (“at”). | |
1460 | # (AL | HL) × (AL | HL) | |
289ce9cc KW |
1461 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Alphabetic'}] |
1462 | = $lb_actions{'LB_NOBREAK'}; | |
1463 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Alphabetic'}] | |
1464 | = $lb_actions{'LB_NOBREAK'}; | |
1465 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Hebrew_Letter'}] | |
1466 | = $lb_actions{'LB_NOBREAK'}; | |
1467 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Hebrew_Letter'}] | |
1468 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1469 | |
1470 | # LB27 Treat a Korean Syllable Block the same as ID. | |
1471 | # (JL | JV | JT | H2 | H3) × IN | |
289ce9cc KW |
1472 | $lb_table[$lb_enums{'JL'}][$lb_enums{'Inseparable'}] |
1473 | = $lb_actions{'LB_NOBREAK'}; | |
1474 | $lb_table[$lb_enums{'JV'}][$lb_enums{'Inseparable'}] | |
1475 | = $lb_actions{'LB_NOBREAK'}; | |
1476 | $lb_table[$lb_enums{'JT'}][$lb_enums{'Inseparable'}] | |
1477 | = $lb_actions{'LB_NOBREAK'}; | |
1478 | $lb_table[$lb_enums{'H2'}][$lb_enums{'Inseparable'}] | |
1479 | = $lb_actions{'LB_NOBREAK'}; | |
1480 | $lb_table[$lb_enums{'H3'}][$lb_enums{'Inseparable'}] | |
1481 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1482 | |
1483 | # (JL | JV | JT | H2 | H3) × PO | |
289ce9cc KW |
1484 | $lb_table[$lb_enums{'JL'}][$lb_enums{'Postfix_Numeric'}] |
1485 | = $lb_actions{'LB_NOBREAK'}; | |
1486 | $lb_table[$lb_enums{'JV'}][$lb_enums{'Postfix_Numeric'}] | |
1487 | = $lb_actions{'LB_NOBREAK'}; | |
1488 | $lb_table[$lb_enums{'JT'}][$lb_enums{'Postfix_Numeric'}] | |
1489 | = $lb_actions{'LB_NOBREAK'}; | |
1490 | $lb_table[$lb_enums{'H2'}][$lb_enums{'Postfix_Numeric'}] | |
1491 | = $lb_actions{'LB_NOBREAK'}; | |
1492 | $lb_table[$lb_enums{'H3'}][$lb_enums{'Postfix_Numeric'}] | |
1493 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1494 | |
1495 | # PR × (JL | JV | JT | H2 | H3) | |
289ce9cc KW |
1496 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'JL'}] |
1497 | = $lb_actions{'LB_NOBREAK'}; | |
1498 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'JV'}] | |
1499 | = $lb_actions{'LB_NOBREAK'}; | |
1500 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'JT'}] | |
1501 | = $lb_actions{'LB_NOBREAK'}; | |
1502 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'H2'}] | |
1503 | = $lb_actions{'LB_NOBREAK'}; | |
1504 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'H3'}] | |
1505 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1506 | |
1507 | # LB26 Do not break a Korean syllable. | |
1508 | # JL × (JL | JV | H2 | H3) | |
1509 | $lb_table[$lb_enums{'JL'}][$lb_enums{'JL'}] = $lb_actions{'LB_NOBREAK'}; | |
1510 | $lb_table[$lb_enums{'JL'}][$lb_enums{'JV'}] = $lb_actions{'LB_NOBREAK'}; | |
1511 | $lb_table[$lb_enums{'JL'}][$lb_enums{'H2'}] = $lb_actions{'LB_NOBREAK'}; | |
1512 | $lb_table[$lb_enums{'JL'}][$lb_enums{'H3'}] = $lb_actions{'LB_NOBREAK'}; | |
1513 | ||
1514 | # (JV | H2) × (JV | JT) | |
1515 | $lb_table[$lb_enums{'JV'}][$lb_enums{'JV'}] = $lb_actions{'LB_NOBREAK'}; | |
1516 | $lb_table[$lb_enums{'H2'}][$lb_enums{'JV'}] = $lb_actions{'LB_NOBREAK'}; | |
1517 | $lb_table[$lb_enums{'JV'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1518 | $lb_table[$lb_enums{'H2'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1519 | ||
1520 | # (JT | H3) × JT | |
1521 | $lb_table[$lb_enums{'JT'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1522 | $lb_table[$lb_enums{'H3'}][$lb_enums{'JT'}] = $lb_actions{'LB_NOBREAK'}; | |
1523 | ||
1524 | # LB25 Do not break between the following pairs of classes relevant to | |
1525 | # numbers, as tailored by example 7 in | |
1526 | # http://www.unicode.org/reports/tr14/#Examples | |
1527 | # We follow that tailoring because Unicode's test cases expect it | |
1528 | # (PR | PO) × ( OP | HY )? NU | |
289ce9cc KW |
1529 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Numeric'}] |
1530 | = $lb_actions{'LB_NOBREAK'}; | |
1531 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Numeric'}] | |
1532 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1533 | |
1534 | # Given that (OP | HY )? is optional, we have to test for it in code. | |
1535 | # We add in the action (instead of overriding) for this, so that in | |
1536 | # the code we can recover the underlying break value. | |
289ce9cc | 1537 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Open_Punctuation'}] |
6b659339 | 1538 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
289ce9cc | 1539 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Open_Punctuation'}] |
6b659339 | 1540 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
289ce9cc | 1541 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Hyphen'}] |
6b659339 | 1542 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
289ce9cc | 1543 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Hyphen'}] |
6b659339 KW |
1544 | += $lb_actions{'LB_PR_or_PO_then_OP_or_HY'}; |
1545 | ||
1546 | # ( OP | HY ) × NU | |
289ce9cc KW |
1547 | $lb_table[$lb_enums{'Open_Punctuation'}][$lb_enums{'Numeric'}] |
1548 | = $lb_actions{'LB_NOBREAK'}; | |
1549 | $lb_table[$lb_enums{'Hyphen'}][$lb_enums{'Numeric'}] | |
1550 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1551 | |
1552 | # NU (NU | SY | IS)* × (NU | SY | IS | CL | CP ) | |
1553 | # which can be rewritten as: | |
1554 | # NU (SY | IS)* × (NU | SY | IS | CL | CP ) | |
289ce9cc KW |
1555 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Numeric'}] |
1556 | = $lb_actions{'LB_NOBREAK'}; | |
1557 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Break_Symbols'}] | |
1558 | = $lb_actions{'LB_NOBREAK'}; | |
1559 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Infix_Numeric'}] | |
1560 | = $lb_actions{'LB_NOBREAK'}; | |
1561 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Close_Punctuation'}] | |
1562 | = $lb_actions{'LB_NOBREAK'}; | |
1563 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Close_Parenthesis'}] | |
1564 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1565 | |
1566 | # Like earlier where we have to test in code, we add in the action so | |
1567 | # that we can recover the underlying values. This is done in rules | |
1568 | # below, as well. The code assumes that we haven't added 2 actions. | |
1569 | # Shoul a later Unicode release break that assumption, then tests | |
1570 | # should start failing. | |
289ce9cc | 1571 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Numeric'}] |
6b659339 | 1572 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1573 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Break_Symbols'}] |
6b659339 | 1574 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1575 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Infix_Numeric'}] |
6b659339 | 1576 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1577 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Close_Punctuation'}] |
6b659339 | 1578 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1579 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Close_Parenthesis'}] |
6b659339 | 1580 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1581 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Numeric'}] |
6b659339 | 1582 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1583 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Break_Symbols'}] |
6b659339 | 1584 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1585 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Infix_Numeric'}] |
6b659339 | 1586 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1587 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Close_Punctuation'}] |
6b659339 | 1588 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
289ce9cc | 1589 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Close_Parenthesis'}] |
6b659339 KW |
1590 | += $lb_actions{'LB_SY_or_IS_then_various'}; |
1591 | ||
1592 | # NU (NU | SY | IS)* (CL | CP)? × (PO | PR) | |
1593 | # which can be rewritten as: | |
1594 | # NU (SY | IS)* (CL | CP)? × (PO | PR) | |
289ce9cc KW |
1595 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Postfix_Numeric'}] |
1596 | = $lb_actions{'LB_NOBREAK'}; | |
1597 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Prefix_Numeric'}] | |
1598 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1599 | |
289ce9cc | 1600 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 | 1601 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1602 | $lb_table[$lb_enums{'Close_Punctuation'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 | 1603 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1604 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 | 1605 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1606 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Postfix_Numeric'}] |
6b659339 KW |
1607 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
1608 | ||
289ce9cc | 1609 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 | 1610 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1611 | $lb_table[$lb_enums{'Close_Punctuation'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 | 1612 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1613 | $lb_table[$lb_enums{'Infix_Numeric'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 | 1614 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
289ce9cc | 1615 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Prefix_Numeric'}] |
6b659339 KW |
1616 | += $lb_actions{'LB_various_then_PO_or_PR'}; |
1617 | ||
b0e24409 KW |
1618 | # LB24 Do not break between numeric prefix/postfix and letters, or between |
1619 | # letters and prefix/postfix. | |
1620 | # (PR | PO) × (AL | HL) | |
289ce9cc KW |
1621 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Alphabetic'}] |
1622 | = $lb_actions{'LB_NOBREAK'}; | |
1623 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1624 | = $lb_actions{'LB_NOBREAK'}; | |
289ce9cc KW |
1625 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Alphabetic'}] |
1626 | = $lb_actions{'LB_NOBREAK'}; | |
1627 | $lb_table[$lb_enums{'Postfix_Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1628 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1629 | |
b0e24409 KW |
1630 | # (AL | HL) × (PR | PO) |
1631 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Prefix_Numeric'}] | |
1632 | = $lb_actions{'LB_NOBREAK'}; | |
1633 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Prefix_Numeric'}] | |
1634 | = $lb_actions{'LB_NOBREAK'}; | |
1635 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Postfix_Numeric'}] | |
1636 | = $lb_actions{'LB_NOBREAK'}; | |
1637 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Postfix_Numeric'}] | |
1638 | = $lb_actions{'LB_NOBREAK'}; | |
1639 | ||
1640 | # LB23a Do not break between numeric prefixes and ideographs, or between | |
1641 | # ideographs and numeric postfixes. | |
1642 | # PR × (ID | EB | EM) | |
1643 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'Ideographic'}] | |
1644 | = $lb_actions{'LB_NOBREAK'}; | |
1645 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'E_Base'}] | |
1646 | = $lb_actions{'LB_NOBREAK'}; | |
1647 | $lb_table[$lb_enums{'Prefix_Numeric'}][$lb_enums{'E_Modifier'}] | |
1648 | = $lb_actions{'LB_NOBREAK'}; | |
1649 | ||
1650 | # (ID | EB | EM) × PO | |
289ce9cc KW |
1651 | $lb_table[$lb_enums{'Ideographic'}][$lb_enums{'Postfix_Numeric'}] |
1652 | = $lb_actions{'LB_NOBREAK'}; | |
b0e24409 KW |
1653 | $lb_table[$lb_enums{'E_Base'}][$lb_enums{'Postfix_Numeric'}] |
1654 | = $lb_actions{'LB_NOBREAK'}; | |
1655 | $lb_table[$lb_enums{'E_Modifier'}][$lb_enums{'Postfix_Numeric'}] | |
1656 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1657 | |
b0e24409 | 1658 | # LB23 Do not break between digits and letters |
6b659339 | 1659 | # (AL | HL) × NU |
289ce9cc KW |
1660 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Numeric'}] |
1661 | = $lb_actions{'LB_NOBREAK'}; | |
1662 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Numeric'}] | |
1663 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1664 | |
1665 | # NU × (AL | HL) | |
289ce9cc KW |
1666 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Alphabetic'}] |
1667 | = $lb_actions{'LB_NOBREAK'}; | |
1668 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Hebrew_Letter'}] | |
1669 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1670 | |
1671 | # LB22 Do not break between two ellipses, or between letters, numbers or | |
1672 | # exclamations and ellipsis. | |
1673 | # (AL | HL) × IN | |
289ce9cc KW |
1674 | $lb_table[$lb_enums{'Alphabetic'}][$lb_enums{'Inseparable'}] |
1675 | = $lb_actions{'LB_NOBREAK'}; | |
1676 | $lb_table[$lb_enums{'Hebrew_Letter'}][$lb_enums{'Inseparable'}] | |
1677 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1678 | |
289ce9cc KW |
1679 | # Exclamation × IN |
1680 | $lb_table[$lb_enums{'Exclamation'}][$lb_enums{'Inseparable'}] | |
1681 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1682 | |
b0e24409 | 1683 | # (ID | EB | EM) × IN |
289ce9cc KW |
1684 | $lb_table[$lb_enums{'Ideographic'}][$lb_enums{'Inseparable'}] |
1685 | = $lb_actions{'LB_NOBREAK'}; | |
b0e24409 KW |
1686 | $lb_table[$lb_enums{'E_Base'}][$lb_enums{'Inseparable'}] |
1687 | = $lb_actions{'LB_NOBREAK'}; | |
1688 | $lb_table[$lb_enums{'E_Modifier'}][$lb_enums{'Inseparable'}] | |
1689 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1690 | |
1691 | # IN × IN | |
289ce9cc KW |
1692 | $lb_table[$lb_enums{'Inseparable'}][$lb_enums{'Inseparable'}] |
1693 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1694 | |
1695 | # NU × IN | |
289ce9cc KW |
1696 | $lb_table[$lb_enums{'Numeric'}][$lb_enums{'Inseparable'}] |
1697 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1698 | |
1699 | # LB21b Don’t break between Solidus and Hebrew letters. | |
1700 | # SY × HL | |
289ce9cc KW |
1701 | $lb_table[$lb_enums{'Break_Symbols'}][$lb_enums{'Hebrew_Letter'}] |
1702 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1703 | |
1704 | # LB21a Don't break after Hebrew + Hyphen. | |
1705 | # HL (HY | BA) × | |
1706 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1707 | $lb_table[$lb_enums{'Hyphen'}][$i] |
1708 | += $lb_actions{'LB_HY_or_BA_then_foo'}; | |
1709 | $lb_table[$lb_enums{'Break_After'}][$i] | |
1710 | += $lb_actions{'LB_HY_or_BA_then_foo'}; | |
6b659339 KW |
1711 | } |
1712 | ||
1713 | # LB21 Do not break before hyphen-minus, other hyphens, fixed-width | |
1714 | # spaces, small kana, and other non-starters, or after acute accents. | |
1715 | # × BA | |
1716 | # × HY | |
1717 | # × NS | |
1718 | # BB × | |
1719 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1720 | $lb_table[$i][$lb_enums{'Break_After'}] = $lb_actions{'LB_NOBREAK'}; |
1721 | $lb_table[$i][$lb_enums{'Hyphen'}] = $lb_actions{'LB_NOBREAK'}; | |
1722 | $lb_table[$i][$lb_enums{'Nonstarter'}] = $lb_actions{'LB_NOBREAK'}; | |
1723 | $lb_table[$lb_enums{'Break_Before'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1724 | } |
1725 | ||
1726 | # LB20 Break before and after unresolved CB. | |
1727 | # ÷ CB | |
1728 | # CB ÷ | |
1729 | # Conditional breaks should be resolved external to the line breaking | |
1730 | # rules. However, the default action is to treat unresolved CB as breaking | |
1731 | # before and after. | |
1732 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1733 | $lb_table[$i][$lb_enums{'Contingent_Break'}] |
1734 | = $lb_actions{'LB_BREAKABLE'}; | |
1735 | $lb_table[$lb_enums{'Contingent_Break'}][$i] | |
1736 | = $lb_actions{'LB_BREAKABLE'}; | |
6b659339 KW |
1737 | } |
1738 | ||
1739 | # LB19 Do not break before or after quotation marks, such as ‘ ” ’. | |
1740 | # × QU | |
1741 | # QU × | |
1742 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1743 | $lb_table[$i][$lb_enums{'Quotation'}] = $lb_actions{'LB_NOBREAK'}; |
1744 | $lb_table[$lb_enums{'Quotation'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1745 | } |
1746 | ||
1747 | # LB18 Break after spaces | |
1748 | # SP ÷ | |
1749 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1750 | $lb_table[$lb_enums{'Space'}][$i] = $lb_actions{'LB_BREAKABLE'}; |
6b659339 KW |
1751 | } |
1752 | ||
1753 | # LB17 Do not break within ‘——’, even with intervening spaces. | |
1754 | # B2 SP* × B2 | |
289ce9cc | 1755 | $lb_table[$lb_enums{'Break_Both'}][$lb_enums{'Break_Both'}] |
6b659339 KW |
1756 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1757 | ||
1758 | # LB16 Do not break between closing punctuation and a nonstarter even with | |
1759 | # intervening spaces. | |
1760 | # (CL | CP) SP* × NS | |
289ce9cc | 1761 | $lb_table[$lb_enums{'Close_Punctuation'}][$lb_enums{'Nonstarter'}] |
6b659339 | 1762 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1763 | $lb_table[$lb_enums{'Close_Parenthesis'}][$lb_enums{'Nonstarter'}] |
6b659339 KW |
1764 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1765 | ||
1766 | ||
1767 | # LB15 Do not break within ‘”[’, even with intervening spaces. | |
1768 | # QU SP* × OP | |
289ce9cc | 1769 | $lb_table[$lb_enums{'Quotation'}][$lb_enums{'Open_Punctuation'}] |
6b659339 KW |
1770 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1771 | ||
1772 | # LB14 Do not break after ‘[’, even after spaces. | |
1773 | # OP SP* × | |
1774 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1775 | $lb_table[$lb_enums{'Open_Punctuation'}][$i] |
6b659339 KW |
1776 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1777 | } | |
1778 | ||
1779 | # LB13 Do not break before ‘]’ or ‘!’ or ‘;’ or ‘/’, even after spaces, as | |
1780 | # tailored by example 7 in http://www.unicode.org/reports/tr14/#Examples | |
1781 | # [^NU] × CL | |
1782 | # [^NU] × CP | |
1783 | # × EX | |
1784 | # [^NU] × IS | |
1785 | # [^NU] × SY | |
1786 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1787 | $lb_table[$i][$lb_enums{'Exclamation'}] |
6b659339 KW |
1788 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1789 | ||
289ce9cc | 1790 | next if $i == $lb_enums{'Numeric'}; |
6b659339 | 1791 | |
289ce9cc | 1792 | $lb_table[$i][$lb_enums{'Close_Punctuation'}] |
6b659339 | 1793 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1794 | $lb_table[$i][$lb_enums{'Close_Parenthesis'}] |
6b659339 | 1795 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1796 | $lb_table[$i][$lb_enums{'Infix_Numeric'}] |
6b659339 | 1797 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1798 | $lb_table[$i][$lb_enums{'Break_Symbols'}] |
6b659339 KW |
1799 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1800 | } | |
1801 | ||
1802 | # LB12a Do not break before NBSP and related characters, except after | |
1803 | # spaces and hyphens. | |
1804 | # [^SP BA HY] × GL | |
1805 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1806 | next if $i == $lb_enums{'Space'} |
1807 | || $i == $lb_enums{'Break_After'} | |
1808 | || $i == $lb_enums{'Hyphen'}; | |
6b659339 KW |
1809 | |
1810 | # We don't break, but if a property above has said don't break even | |
1811 | # with space between, don't override that (also in the next few rules) | |
289ce9cc | 1812 | next if $lb_table[$i][$lb_enums{'Glue'}] |
6b659339 | 1813 | == $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1814 | $lb_table[$i][$lb_enums{'Glue'}] = $lb_actions{'LB_NOBREAK'}; |
6b659339 KW |
1815 | } |
1816 | ||
1817 | # LB12 Do not break after NBSP and related characters. | |
1818 | # GL × | |
1819 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1820 | next if $lb_table[$lb_enums{'Glue'}][$i] |
6b659339 | 1821 | == $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
289ce9cc | 1822 | $lb_table[$lb_enums{'Glue'}][$i] = $lb_actions{'LB_NOBREAK'}; |
6b659339 KW |
1823 | } |
1824 | ||
1825 | # LB11 Do not break before or after Word joiner and related characters. | |
1826 | # × WJ | |
1827 | # WJ × | |
1828 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1829 | if ($lb_table[$i][$lb_enums{'Word_Joiner'}] |
6b659339 KW |
1830 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) |
1831 | { | |
289ce9cc | 1832 | $lb_table[$i][$lb_enums{'Word_Joiner'}] = $lb_actions{'LB_NOBREAK'}; |
6b659339 | 1833 | } |
289ce9cc | 1834 | if ($lb_table[$lb_enums{'Word_Joiner'}][$i] |
6b659339 KW |
1835 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) |
1836 | { | |
289ce9cc | 1837 | $lb_table[$lb_enums{'Word_Joiner'}][$i] = $lb_actions{'LB_NOBREAK'}; |
6b659339 KW |
1838 | } |
1839 | } | |
1840 | ||
1841 | # Special case this here to avoid having to do a special case in the code, | |
1842 | # by making this the same as other things with a SP in front of them that | |
1843 | # don't break, we avoid an extra test | |
289ce9cc | 1844 | $lb_table[$lb_enums{'Space'}][$lb_enums{'Word_Joiner'}] |
6b659339 KW |
1845 | = $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}; |
1846 | ||
1847 | # LB9 and LB10 are done in the same loop | |
1848 | # | |
1849 | # LB9 Do not break a combining character sequence; treat it as if it has | |
1850 | # the line breaking class of the base character in all of the | |
b0e24409 KW |
1851 | # higher-numbered rules. Treat ZWJ as if it were CM |
1852 | # Treat X (CM|ZWJ)* as if it were X. | |
6b659339 KW |
1853 | # where X is any line break class except BK, CR, LF, NL, SP, or ZW. |
1854 | ||
b0e24409 KW |
1855 | # LB10 Treat any remaining combining mark or ZWJ as AL. This catches the |
1856 | # case where a CM or ZWJ is the first character on the line or follows SP, | |
1857 | # BK, CR, LF, NL, or ZW. | |
6b659339 KW |
1858 | for my $i (0 .. @lb_table - 1) { |
1859 | ||
b0e24409 KW |
1860 | # When the CM or ZWJ is the first in the pair, we don't know without |
1861 | # looking behind whether the CM or ZWJ is going to attach to an | |
1862 | # earlier character, or not. So have to figure this out at runtime in | |
1863 | # the code | |
1864 | $lb_table[$lb_enums{'Combining_Mark'}][$i] | |
1865 | = $lb_actions{'LB_CM_ZWJ_foo'}; | |
1866 | $lb_table[$lb_enums{'ZWJ'}][$i] = $lb_actions{'LB_CM_ZWJ_foo'}; | |
289ce9cc KW |
1867 | |
1868 | if ( $i == $lb_enums{'Mandatory_Break'} | |
1869 | || $i == $lb_enums{'EDGE'} | |
1870 | || $i == $lb_enums{'Carriage_Return'} | |
1871 | || $i == $lb_enums{'Line_Feed'} | |
1872 | || $i == $lb_enums{'Next_Line'} | |
1873 | || $i == $lb_enums{'Space'} | |
1874 | || $i == $lb_enums{'ZWSpace'}) | |
6b659339 KW |
1875 | { |
1876 | # For these classes, a following CM doesn't combine, and should do | |
289ce9cc KW |
1877 | # whatever 'Alphabetic' would do. |
1878 | $lb_table[$i][$lb_enums{'Combining_Mark'}] | |
1879 | = $lb_table[$i][$lb_enums{'Alphabetic'}]; | |
b0e24409 KW |
1880 | $lb_table[$i][$lb_enums{'ZWJ'}] |
1881 | = $lb_table[$i][$lb_enums{'Alphabetic'}]; | |
6b659339 KW |
1882 | } |
1883 | else { | |
b0e24409 KW |
1884 | # For these classes, the CM or ZWJ combines, so doesn't break, |
1885 | # inheriting the type of nobreak from the master character. | |
289ce9cc | 1886 | if ($lb_table[$i][$lb_enums{'Combining_Mark'}] |
6b659339 KW |
1887 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) |
1888 | { | |
289ce9cc KW |
1889 | $lb_table[$i][$lb_enums{'Combining_Mark'}] |
1890 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 | 1891 | } |
b0e24409 KW |
1892 | if ($lb_table[$i][$lb_enums{'ZWJ'}] |
1893 | != $lb_actions{'LB_NOBREAK_EVEN_WITH_SP_BETWEEN'}) | |
1894 | { | |
1895 | $lb_table[$i][$lb_enums{'ZWJ'}] | |
1896 | = $lb_actions{'LB_NOBREAK'}; | |
1897 | } | |
6b659339 KW |
1898 | } |
1899 | } | |
1900 | ||
8a6698d7 UC |
1901 | # LB8a Do not break after a zero width joiner |
1902 | # ZWJ × | |
1903 | for my $i (0 .. @lb_table - 1) { | |
1904 | $lb_table[$lb_enums{'ZWJ'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
1905 | } | |
b0e24409 | 1906 | |
6b659339 KW |
1907 | # LB8 Break before any character following a zero-width space, even if one |
1908 | # or more spaces intervene. | |
1909 | # ZW SP* ÷ | |
1910 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1911 | $lb_table[$lb_enums{'ZWSpace'}][$i] = $lb_actions{'LB_BREAKABLE'}; |
6b659339 KW |
1912 | } |
1913 | ||
1914 | # Because of LB8-10, we need to look at context for "SP x", and this must | |
1915 | # be done in the code. So override the existing rules for that, by adding | |
1916 | # a constant to get new rules that tell the code it needs to look at | |
1917 | # context. By adding this action instead of replacing the existing one, | |
1918 | # we can get back to the original rule if necessary. | |
1919 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc | 1920 | $lb_table[$lb_enums{'Space'}][$i] += $lb_actions{'LB_SP_foo'}; |
6b659339 KW |
1921 | } |
1922 | ||
1923 | # LB7 Do not break before spaces or zero width space. | |
1924 | # × SP | |
1925 | # × ZW | |
1926 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1927 | $lb_table[$i][$lb_enums{'Space'}] = $lb_actions{'LB_NOBREAK'}; |
1928 | $lb_table[$i][$lb_enums{'ZWSpace'}] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1929 | } |
1930 | ||
1931 | # LB6 Do not break before hard line breaks. | |
1932 | # × ( BK | CR | LF | NL ) | |
1933 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1934 | $lb_table[$i][$lb_enums{'Mandatory_Break'}] = $lb_actions{'LB_NOBREAK'}; |
1935 | $lb_table[$i][$lb_enums{'Carriage_Return'}] = $lb_actions{'LB_NOBREAK'}; | |
1936 | $lb_table[$i][$lb_enums{'Line_Feed'}] = $lb_actions{'LB_NOBREAK'}; | |
1937 | $lb_table[$i][$lb_enums{'Next_Line'}] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1938 | } |
1939 | ||
1940 | # LB5 Treat CR followed by LF, as well as CR, LF, and NL as hard line breaks. | |
1941 | # CR × LF | |
1942 | # CR ! | |
1943 | # LF ! | |
1944 | # NL ! | |
1945 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1946 | $lb_table[$lb_enums{'Carriage_Return'}][$i] |
1947 | = $lb_actions{'LB_BREAKABLE'}; | |
1948 | $lb_table[$lb_enums{'Line_Feed'}][$i] = $lb_actions{'LB_BREAKABLE'}; | |
1949 | $lb_table[$lb_enums{'Next_Line'}][$i] = $lb_actions{'LB_BREAKABLE'}; | |
6b659339 | 1950 | } |
289ce9cc KW |
1951 | $lb_table[$lb_enums{'Carriage_Return'}][$lb_enums{'Line_Feed'}] |
1952 | = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1953 | |
1954 | # LB4 Always break after hard line breaks. | |
1955 | # BK ! | |
1956 | for my $i (0 .. @lb_table - 1) { | |
289ce9cc KW |
1957 | $lb_table[$lb_enums{'Mandatory_Break'}][$i] |
1958 | = $lb_actions{'LB_BREAKABLE'}; | |
6b659339 KW |
1959 | } |
1960 | ||
6b659339 KW |
1961 | # LB3 Always break at the end of text. |
1962 | # ! eot | |
b0e24409 KW |
1963 | # LB2 Never break at the start of text. |
1964 | # sot × | |
6b659339 | 1965 | for my $i (0 .. @lb_table - 1) { |
289ce9cc KW |
1966 | $lb_table[$i][$lb_enums{'EDGE'}] = $lb_actions{'LB_BREAKABLE'}; |
1967 | $lb_table[$lb_enums{'EDGE'}][$i] = $lb_actions{'LB_NOBREAK'}; | |
6b659339 KW |
1968 | } |
1969 | ||
1970 | # LB1 Assign a line breaking class to each code point of the input. | |
1971 | # Resolve AI, CB, CJ, SA, SG, and XX into other line breaking classes | |
1972 | # depending on criteria outside the scope of this algorithm. | |
1973 | # | |
1974 | # In the absence of such criteria all characters with a specific | |
1975 | # combination of original class and General_Category property value are | |
1976 | # resolved as follows: | |
1977 | # Original Resolved General_Category | |
1978 | # AI, SG, XX AL Any | |
1979 | # SA CM Only Mn or Mc | |
1980 | # SA AL Any except Mn and Mc | |
1981 | # CJ NS Any | |
1982 | # | |
1983 | # This is done in mktables, so we never see any of the remapped-from | |
1984 | # classes. | |
1985 | ||
289ce9cc KW |
1986 | output_table_common('LB', \%lb_actions, |
1987 | \@lb_table, \@lb_short_enums, \%lb_abbreviations); | |
6b659339 KW |
1988 | } |
1989 | ||
7e54b87f KW |
1990 | sub output_WB_table() { |
1991 | ||
1992 | # Create and output the enums, #defines, and pair table for use in | |
1993 | # determining Word Breaks, given in http://www.unicode.org/reports/tr29/. | |
1994 | ||
1995 | # This uses the same mechanism in the other bounds tables generated by | |
1996 | # this file. The actions that could override a 0 or 1 are added to those | |
1997 | # numbers; the actions that clearly don't depend on the underlying rule | |
1998 | # simply overwrite | |
1999 | my %wb_actions = ( | |
2000 | WB_NOBREAK => 0, | |
2001 | WB_BREAKABLE => 1, | |
2002 | WB_hs_then_hs => 2, | |
b0e24409 | 2003 | WB_Ex_or_FO_or_ZWJ_then_foo => 3, |
7e54b87f KW |
2004 | WB_DQ_then_HL => 4, |
2005 | WB_HL_then_DQ => 6, | |
2006 | WB_LE_or_HL_then_MB_or_ML_or_SQ => 8, | |
2007 | WB_MB_or_ML_or_SQ_then_LE_or_HL => 10, | |
2008 | WB_MB_or_MN_or_SQ_then_NU => 12, | |
2009 | WB_NU_then_MB_or_MN_or_SQ => 14, | |
b0e24409 | 2010 | WB_RI_then_RI => 16, |
7e54b87f KW |
2011 | ); |
2012 | ||
7e54b87f KW |
2013 | # Construct the WB pair table. |
2014 | # The table is constructed in reverse order of the rules, to make the | |
2015 | # lower-numbered, higher priority ones override the later ones, as the | |
2016 | # algorithm stops at the earliest matching rule | |
2017 | ||
2018 | my @wb_table; | |
2027d365 | 2019 | my $table_size = @wb_short_enums; |
7e54b87f KW |
2020 | |
2021 | # Otherwise, break everywhere (including around ideographs). | |
b0e24409 | 2022 | # WB99 Any ÷ Any |
7e54b87f KW |
2023 | for my $i (0 .. $table_size - 1) { |
2024 | for my $j (0 .. $table_size - 1) { | |
2025 | $wb_table[$i][$j] = $wb_actions{'WB_BREAKABLE'}; | |
2026 | } | |
2027 | } | |
2028 | ||
b0e24409 KW |
2029 | # Do not break within emoji flag sequences. That is, do not break between |
2030 | # regional indicator (RI) symbols if there is an odd number of RI | |
2031 | # characters before the break point. | |
2032 | # WB16 [^RI] (RI RI)* RI × RI | |
c492f156 | 2033 | # WB15 sot (RI RI)* RI × RI |
289ce9cc | 2034 | $wb_table[$wb_enums{'Regional_Indicator'}] |
b0e24409 KW |
2035 | [$wb_enums{'Regional_Indicator'}] = $wb_actions{'WB_RI_then_RI'}; |
2036 | ||
2037 | # Do not break within emoji modifier sequences. | |
2038 | # WB14 ( E_Base | EBG ) × E_Modifier | |
2039 | $wb_table[$wb_enums{'E_Base'}][$wb_enums{'E_Modifier'}] | |
2040 | = $wb_actions{'WB_NOBREAK'}; | |
2041 | $wb_table[$wb_enums{'E_Base_GAZ'}][$wb_enums{'E_Modifier'}] | |
2042 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2043 | |
2044 | # Do not break from extenders. | |
2045 | # WB13b ExtendNumLet × (ALetter | Hebrew_Letter | Numeric | Katakana) | |
289ce9cc KW |
2046 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'ALetter'}] |
2047 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2048 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'XPG_LE'}] |
2049 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2050 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'Hebrew_Letter'}] |
2051 | = $wb_actions{'WB_NOBREAK'}; | |
2052 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'Numeric'}] | |
2053 | = $wb_actions{'WB_NOBREAK'}; | |
2054 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'Katakana'}] | |
2055 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2056 | |
2057 | # WB13a (ALetter | Hebrew_Letter | Numeric | Katakana | ExtendNumLet) | |
d21ae9f6 | 2058 | # × ExtendNumLet |
289ce9cc KW |
2059 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'ExtendNumLet'}] |
2060 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2061 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'ExtendNumLet'}] |
2062 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2063 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'ExtendNumLet'}] |
2064 | = $wb_actions{'WB_NOBREAK'}; | |
2065 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'ExtendNumLet'}] | |
2066 | = $wb_actions{'WB_NOBREAK'}; | |
2067 | $wb_table[$wb_enums{'Katakana'}][$wb_enums{'ExtendNumLet'}] | |
2068 | = $wb_actions{'WB_NOBREAK'}; | |
2069 | $wb_table[$wb_enums{'ExtendNumLet'}][$wb_enums{'ExtendNumLet'}] | |
2070 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2071 | |
2072 | # Do not break between Katakana. | |
2073 | # WB13 Katakana × Katakana | |
289ce9cc KW |
2074 | $wb_table[$wb_enums{'Katakana'}][$wb_enums{'Katakana'}] |
2075 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2076 | |
2077 | # Do not break within sequences, such as “3.2” or “3,456.789”. | |
2078 | # WB12 Numeric × (MidNum | MidNumLet | Single_Quote) Numeric | |
289ce9cc | 2079 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'MidNumLet'}] |
7e54b87f | 2080 | += $wb_actions{'WB_NU_then_MB_or_MN_or_SQ'}; |
289ce9cc | 2081 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'MidNum'}] |
7e54b87f | 2082 | += $wb_actions{'WB_NU_then_MB_or_MN_or_SQ'}; |
289ce9cc | 2083 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'Single_Quote'}] |
7e54b87f KW |
2084 | += $wb_actions{'WB_NU_then_MB_or_MN_or_SQ'}; |
2085 | ||
2086 | # WB11 Numeric (MidNum | (MidNumLet | Single_Quote)) × Numeric | |
289ce9cc | 2087 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'Numeric'}] |
7e54b87f | 2088 | += $wb_actions{'WB_MB_or_MN_or_SQ_then_NU'}; |
289ce9cc | 2089 | $wb_table[$wb_enums{'MidNum'}][$wb_enums{'Numeric'}] |
7e54b87f | 2090 | += $wb_actions{'WB_MB_or_MN_or_SQ_then_NU'}; |
289ce9cc | 2091 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'Numeric'}] |
7e54b87f KW |
2092 | += $wb_actions{'WB_MB_or_MN_or_SQ_then_NU'}; |
2093 | ||
2094 | # Do not break within sequences of digits, or digits adjacent to letters | |
2095 | # (“3a”, or “A3”). | |
2096 | # WB10 Numeric × (ALetter | Hebrew_Letter) | |
289ce9cc KW |
2097 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'ALetter'}] |
2098 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2099 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'XPG_LE'}] |
2100 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2101 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'Hebrew_Letter'}] |
2102 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2103 | |
2104 | # WB9 (ALetter | Hebrew_Letter) × Numeric | |
289ce9cc KW |
2105 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'Numeric'}] |
2106 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2107 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'Numeric'}] |
2108 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2109 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Numeric'}] |
2110 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2111 | |
2112 | # WB8 Numeric × Numeric | |
289ce9cc KW |
2113 | $wb_table[$wb_enums{'Numeric'}][$wb_enums{'Numeric'}] |
2114 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2115 | |
2116 | # Do not break letters across certain punctuation. | |
2117 | # WB7c Hebrew_Letter Double_Quote × Hebrew_Letter | |
289ce9cc KW |
2118 | $wb_table[$wb_enums{'Double_Quote'}][$wb_enums{'Hebrew_Letter'}] |
2119 | += $wb_actions{'WB_DQ_then_HL'}; | |
7e54b87f KW |
2120 | |
2121 | # WB7b Hebrew_Letter × Double_Quote Hebrew_Letter | |
289ce9cc KW |
2122 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Double_Quote'}] |
2123 | += $wb_actions{'WB_HL_then_DQ'}; | |
7e54b87f KW |
2124 | |
2125 | # WB7a Hebrew_Letter × Single_Quote | |
289ce9cc KW |
2126 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Single_Quote'}] |
2127 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2128 | |
2129 | # WB7 (ALetter | Hebrew_Letter) (MidLetter | MidNumLet | Single_Quote) | |
2130 | # × (ALetter | Hebrew_Letter) | |
289ce9cc | 2131 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'ALetter'}] |
7e54b87f | 2132 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
c0734505 KW |
2133 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'XPG_LE'}] |
2134 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; | |
289ce9cc | 2135 | $wb_table[$wb_enums{'MidNumLet'}][$wb_enums{'Hebrew_Letter'}] |
7e54b87f | 2136 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
289ce9cc | 2137 | $wb_table[$wb_enums{'MidLetter'}][$wb_enums{'ALetter'}] |
7e54b87f | 2138 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
c0734505 KW |
2139 | $wb_table[$wb_enums{'MidLetter'}][$wb_enums{'XPG_LE'}] |
2140 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; | |
289ce9cc | 2141 | $wb_table[$wb_enums{'MidLetter'}][$wb_enums{'Hebrew_Letter'}] |
7e54b87f | 2142 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
289ce9cc | 2143 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'ALetter'}] |
7e54b87f | 2144 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
c0734505 KW |
2145 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'XPG_LE'}] |
2146 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; | |
289ce9cc | 2147 | $wb_table[$wb_enums{'Single_Quote'}][$wb_enums{'Hebrew_Letter'}] |
7e54b87f KW |
2148 | += $wb_actions{'WB_MB_or_ML_or_SQ_then_LE_or_HL'}; |
2149 | ||
2150 | # WB6 (ALetter | Hebrew_Letter) × (MidLetter | MidNumLet | |
2151 | # | Single_Quote) (ALetter | Hebrew_Letter) | |
289ce9cc | 2152 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'MidNumLet'}] |
7e54b87f | 2153 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
c0734505 KW |
2154 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'MidNumLet'}] |
2155 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; | |
289ce9cc | 2156 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'MidNumLet'}] |
7e54b87f | 2157 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
289ce9cc | 2158 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'MidLetter'}] |
7e54b87f | 2159 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
c0734505 KW |
2160 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'MidLetter'}] |
2161 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; | |
289ce9cc | 2162 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'MidLetter'}] |
7e54b87f | 2163 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
289ce9cc | 2164 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'Single_Quote'}] |
7e54b87f | 2165 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
c0734505 KW |
2166 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'Single_Quote'}] |
2167 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; | |
289ce9cc | 2168 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Single_Quote'}] |
7e54b87f KW |
2169 | += $wb_actions{'WB_LE_or_HL_then_MB_or_ML_or_SQ'}; |
2170 | ||
2171 | # Do not break between most letters. | |
2172 | # WB5 (ALetter | Hebrew_Letter) × (ALetter | Hebrew_Letter) | |
289ce9cc KW |
2173 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'ALetter'}] |
2174 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2175 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'ALetter'}] |
2176 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2177 | $wb_table[$wb_enums{'ALetter'}][$wb_enums{'Hebrew_Letter'}] |
2178 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2179 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'Hebrew_Letter'}] |
2180 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2181 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'ALetter'}] |
2182 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2183 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'XPG_LE'}] |
2184 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2185 | $wb_table[$wb_enums{'Hebrew_Letter'}][$wb_enums{'Hebrew_Letter'}] |
2186 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2187 | $wb_table[$wb_enums{'XPG_LE'}][$wb_enums{'XPG_LE'}] |
2188 | = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f | 2189 | |
b0e24409 KW |
2190 | # Ignore Format and Extend characters, except after sot, CR, LF, and |
2191 | # Newline. This also has the effect of: Any × (Format | Extend | ZWJ) | |
2192 | # WB4 X (Extend | Format | ZWJ)* → X | |
7e54b87f | 2193 | for my $i (0 .. @wb_table - 1) { |
289ce9cc | 2194 | $wb_table[$wb_enums{'Extend'}][$i] |
b0e24409 | 2195 | = $wb_actions{'WB_Ex_or_FO_or_ZWJ_then_foo'}; |
289ce9cc | 2196 | $wb_table[$wb_enums{'Format'}][$i] |
b0e24409 KW |
2197 | = $wb_actions{'WB_Ex_or_FO_or_ZWJ_then_foo'}; |
2198 | $wb_table[$wb_enums{'ZWJ'}][$i] | |
2199 | = $wb_actions{'WB_Ex_or_FO_or_ZWJ_then_foo'}; | |
2200 | } | |
2201 | for my $i (0 .. @wb_table - 1) { | |
2202 | $wb_table[$i][$wb_enums{'Extend'}] = $wb_actions{'WB_NOBREAK'}; | |
2203 | $wb_table[$i][$wb_enums{'Format'}] = $wb_actions{'WB_NOBREAK'}; | |
2204 | $wb_table[$i][$wb_enums{'ZWJ'}] = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2205 | } |
2206 | ||
2207 | # Implied is that these attach to the character before them, except for | |
2208 | # the characters that mark the end of a region of text. The rules below | |
2209 | # override the ones set up here, for all the characters that need | |
2210 | # overriding. | |
2211 | for my $i (0 .. @wb_table - 1) { | |
289ce9cc KW |
2212 | $wb_table[$i][$wb_enums{'Extend'}] = $wb_actions{'WB_NOBREAK'}; |
2213 | $wb_table[$i][$wb_enums{'Format'}] = $wb_actions{'WB_NOBREAK'}; | |
7e54b87f KW |
2214 | } |
2215 | ||
c0734505 KW |
2216 | # Keep horizontal whitespace together |
2217 | # Use perl's tailoring instead | |
2218 | # WB3d WSegSpace × WSegSpace | |
2219 | #$wb_table[$wb_enums{'WSegSpace'}][$wb_enums{'WSegSpace'}] | |
2220 | # = $wb_actions{'WB_NOBREAK'}; | |
2221 | ||
b0e24409 KW |
2222 | # Do not break within emoji zwj sequences. |
2223 | # WB3c ZWJ × ( Glue_After_Zwj | EBG ) | |
2224 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'Glue_After_Zwj'}] | |
2225 | = $wb_actions{'WB_NOBREAK'}; | |
2226 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'E_Base_GAZ'}] | |
2227 | = $wb_actions{'WB_NOBREAK'}; | |
c0734505 KW |
2228 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'XPG_XX'}] |
2229 | = $wb_actions{'WB_NOBREAK'}; | |
2230 | $wb_table[$wb_enums{'ZWJ'}][$wb_enums{'XPG_LE'}] | |
2231 | = $wb_actions{'WB_NOBREAK'}; | |
b0e24409 | 2232 | |
d21ae9f6 | 2233 | # Break before and after newlines |
7e54b87f KW |
2234 | # WB3b ÷ (Newline | CR | LF) |
2235 | # WB3a (Newline | CR | LF) ÷ | |
2236 | # et. al. | |
289ce9cc | 2237 | for my $i ('CR', 'LF', 'Newline', 'Perl_Tailored_HSpace') { |
7e54b87f KW |
2238 | for my $j (0 .. @wb_table - 1) { |
2239 | $wb_table[$j][$wb_enums{$i}] = $wb_actions{'WB_BREAKABLE'}; | |
2240 | $wb_table[$wb_enums{$i}][$j] = $wb_actions{'WB_BREAKABLE'}; | |
2241 | } | |
2242 | } | |
2243 | ||
2244 | # But do not break within white space. | |
2245 | # WB3 CR × LF | |
2246 | # et.al. | |
289ce9cc KW |
2247 | for my $i ('CR', 'LF', 'Newline', 'Perl_Tailored_HSpace') { |
2248 | for my $j ('CR', 'LF', 'Newline', 'Perl_Tailored_HSpace') { | |
7e54b87f KW |
2249 | $wb_table[$wb_enums{$i}][$wb_enums{$j}] = $wb_actions{'WB_NOBREAK'}; |
2250 | } | |
2251 | } | |
2252 | ||
b0e24409 | 2253 | # And do not break horizontal space followed by Extend or Format or ZWJ |
289ce9cc KW |
2254 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}][$wb_enums{'Extend'}] |
2255 | = $wb_actions{'WB_NOBREAK'}; | |
2256 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}][$wb_enums{'Format'}] | |
2257 | = $wb_actions{'WB_NOBREAK'}; | |
b0e24409 KW |
2258 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}][$wb_enums{'ZWJ'}] |
2259 | = $wb_actions{'WB_NOBREAK'}; | |
289ce9cc KW |
2260 | $wb_table[$wb_enums{'Perl_Tailored_HSpace'}] |
2261 | [$wb_enums{'Perl_Tailored_HSpace'}] | |
2262 | = $wb_actions{'WB_hs_then_hs'}; | |
7e54b87f | 2263 | |
b0e24409 KW |
2264 | # Break at the start and end of text, unless the text is empty |
2265 | # WB2 Any ÷ eot | |
2266 | # WB1 sot ÷ Any | |
7e54b87f | 2267 | for my $i (0 .. @wb_table - 1) { |
289ce9cc KW |
2268 | $wb_table[$i][$wb_enums{'EDGE'}] = $wb_actions{'WB_BREAKABLE'}; |
2269 | $wb_table[$wb_enums{'EDGE'}][$i] = $wb_actions{'WB_BREAKABLE'}; | |
7e54b87f | 2270 | } |
289ce9cc | 2271 | $wb_table[$wb_enums{'EDGE'}][$wb_enums{'EDGE'}] = 0; |
7e54b87f | 2272 | |
289ce9cc KW |
2273 | output_table_common('WB', \%wb_actions, |
2274 | \@wb_table, \@wb_short_enums, \%wb_abbreviations); | |
7e54b87f KW |
2275 | } |
2276 | ||
4eea95a6 KW |
2277 | sub sanitize_name ($) { |
2278 | # Change the non-word characters in the input string to standardized word | |
2279 | # equivalents | |
2280 | # | |
2281 | my $sanitized = shift; | |
2282 | $sanitized =~ s/=/__/; | |
2283 | $sanitized =~ s/&/_AMP_/; | |
2284 | $sanitized =~ s/\./_DOT_/; | |
2285 | $sanitized =~ s/-/_MINUS_/; | |
2286 | $sanitized =~ s!/!_SLASH_!; | |
2287 | ||
2288 | return $sanitized; | |
2289 | } | |
2290 | ||
cef72199 | 2291 | switch_pound_if ('ALL', 'PERL_IN_REGCOMP_C'); |
4eea95a6 | 2292 | |
9d9177be KW |
2293 | output_invlist("Latin1", [ 0, 256 ]); |
2294 | output_invlist("AboveLatin1", [ 256 ]); | |
2295 | ||
bffc0129 | 2296 | end_file_pound_if; |
43b443dd | 2297 | |
3f427fd9 KW |
2298 | # We construct lists for all the POSIX and backslash sequence character |
2299 | # classes in two forms: | |
2300 | # 1) ones which match only in the ASCII range | |
2301 | # 2) ones which match either in the Latin1 range, or the entire Unicode range | |
2302 | # | |
2303 | # These get compiled in, and hence affect the memory footprint of every Perl | |
2304 | # program, even those not using Unicode. To minimize the size, currently | |
2305 | # the Latin1 version is generated for the beyond ASCII range except for those | |
2306 | # lists that are quite small for the entire range, such as for \s, which is 22 | |
2307 | # UVs long plus 4 UVs (currently) for the header. | |
2308 | # | |
2309 | # To save even more memory, the ASCII versions could be derived from the | |
2310 | # larger ones at runtime, saving some memory (minus the expense of the machine | |
2311 | # instructions to do so), but these are all small anyway, so their total is | |
2312 | # about 100 UVs. | |
2313 | # | |
2314 | # In the list of properties below that get generated, the L1 prefix is a fake | |
2315 | # property that means just the Latin1 range of the full property (whose name | |
2316 | # has an X prefix instead of L1). | |
a02047bf KW |
2317 | # |
2318 | # An initial & means to use the subroutine from this file instead of an | |
2319 | # official inversion list. | |
3f427fd9 | 2320 | |
53146480 KW |
2321 | # Below is the list of property names to generate. '&' means to use the |
2322 | # subroutine to generate the inversion list instead of the generic code | |
2323 | # below. Some properties have a comma-separated list after the name, | |
2324 | # These are extra enums to add to those found in the Unicode tables. | |
2325 | no warnings 'qw'; | |
2326 | # Ignore non-alpha in sort | |
4eea95a6 KW |
2327 | my @props; |
2328 | push @props, sort { prop_name_for_cmp($a) cmp prop_name_for_cmp($b) } qw( | |
2329 | &NonL1_Perl_Non_Final_Folds | |
2330 | &UpperLatin1 | |
2027d365 KW |
2331 | _Perl_GCB,EDGE,E_Base,E_Base_GAZ,E_Modifier,Glue_After_Zwj,LV,Prepend,Regional_Indicator,SpacingMark,ZWJ,XPG_XX |
2332 | _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 | |
2333 | _Perl_SB,EDGE,SContinue,CR,Extend,LF | |
2334 | _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,XPG_XX,XPG_LE | |
4eea95a6 KW |
2335 | _Perl_SCX,Latin,Inherited,Unknown,Kore,Jpan,Hanb,INVALID |
2336 | Lowercase_Mapping | |
2337 | Titlecase_Mapping | |
2338 | Uppercase_Mapping | |
2339 | Simple_Case_Folding | |
2340 | Case_Folding | |
2341 | &_Perl_IVCF | |
2342 | ); | |
2343 | # NOTE that the convention is that extra enum values come | |
2344 | # after the property name, separated by commas, with the enums | |
2345 | # that aren't ever defined by Unicode coming last, at least 4 | |
2346 | # all-uppercase characters. The others are enum names that | |
2347 | # are needed by perl, but aren't in all Unicode releases. | |
2348 | ||
2349 | my @bin_props; | |
1aefa327 | 2350 | my @perl_prop_synonyms; |
4eea95a6 | 2351 | my %enums; |
2d74dcf2 KW |
2352 | my @deprecated_messages = ""; # Element [0] is a placeholder |
2353 | my %deprecated_tags; | |
4eea95a6 | 2354 | |
27097618 KW |
2355 | my $float_e_format = qr/ ^ -? \d \. \d+ e [-+] \d+ $ /x; |
2356 | ||
2357 | # Create another hash that maps floating point x.yyEzz representation to what | |
2358 | # %stricter_to_file_of does for the equivalent rational. A typical entry in | |
2359 | # the latter hash is | |
2360 | # | |
2361 | # 'nv=1/2' => 'Nv/1_2', | |
2362 | # | |
2363 | # From that, this loop creates an entry | |
2364 | # | |
2365 | # 'nv=5.00e-01' => 'Nv/1_2', | |
2366 | # | |
2367 | # %stricter_to_file_of contains far more than just the rationals. Instead we | |
2368 | # use %utf8::nv_floating_to_rational which should have an entry for each | |
2369 | # nv in the former hash. | |
2370 | my %floating_to_file_of; | |
2371 | foreach my $key (keys %utf8::nv_floating_to_rational) { | |
2372 | my $value = $utf8::nv_floating_to_rational{$key}; | |
2373 | $floating_to_file_of{$key} = $utf8::stricter_to_file_of{"nv=$value"}; | |
2374 | } | |
2375 | ||
4eea95a6 KW |
2376 | # Collect all the binary properties from data in lib/unicore |
2377 | # Sort so that complements come after the main table, and the shortest | |
8091afe3 | 2378 | # names first, finally alphabetically. Also, sort together the tables we want |
f81c4731 KW |
2379 | # to be kept together, and prefer those with 'posix' in their names, which is |
2380 | # what the C code is expecting their names to be. | |
4eea95a6 | 2381 | foreach my $property (sort |
2d74dcf2 | 2382 | { exists $keep_together{lc $b} <=> exists $keep_together{lc $a} |
f81c4731 KW |
2383 | or $b =~ /posix/i <=> $a =~ /posix/i |
2384 | or $b =~ /perl/i <=> $a =~ /perl/i | |
27097618 | 2385 | or $a =~ $float_e_format <=> $b =~ $float_e_format |
2d74dcf2 | 2386 | or $a =~ /!/ <=> $b =~ /!/ |
4eea95a6 KW |
2387 | or length $a <=> length $b |
2388 | or $a cmp $b | |
2389 | } keys %utf8::loose_to_file_of, | |
27097618 KW |
2390 | keys %utf8::stricter_to_file_of, |
2391 | keys %floating_to_file_of | |
53146480 | 2392 | ) { |
0f5e3c71 | 2393 | |
4eea95a6 KW |
2394 | # These two hashes map properties to values that can be considered to |
2395 | # be checksums. If two properties have the same checksum, they have | |
2396 | # identical entries. Otherwise they differ in some way. | |
2397 | my $tag = $utf8::loose_to_file_of{$property}; | |
2398 | $tag = $utf8::stricter_to_file_of{$property} unless defined $tag; | |
27097618 | 2399 | $tag = $floating_to_file_of{$property} unless defined $tag; |
4eea95a6 KW |
2400 | |
2401 | # The tag may contain an '!' meaning it is identical to the one formed | |
394d2d3f KW |
2402 | # by removing the !, except that it is inverted. |
2403 | my $inverted = $tag =~ s/!//; | |
4eea95a6 | 2404 | |
27097618 KW |
2405 | # This hash is lacking the property name |
2406 | $property = "nv=$property" if $property =~ $float_e_format; | |
2407 | ||
4eea95a6 KW |
2408 | # The list of 'prop=value' entries that this single entry expands to |
2409 | my @this_entries; | |
2410 | ||
2411 | # Split 'property=value' on the equals sign, with $lhs being the whole | |
2412 | # thing if there is no '=' | |
2413 | my ($lhs, $rhs) = $property =~ / ( [^=]* ) ( =? .*) /x; | |
2414 | ||
394d2d3f KW |
2415 | # $lhs then becomes the property name. See if there are any synonyms |
2416 | # for this property. | |
2417 | if (exists $prop_name_aliases{$lhs}) { | |
2418 | ||
2419 | # If so, do the combinatorics so that a new entry is added for | |
2420 | # each legal property combined with the property value (which is | |
2421 | # $rhs) | |
2422 | foreach my $alias (@{$prop_name_aliases{$lhs}}) { | |
2423 | ||
2424 | # But, there are some ambiguities, like 'script' is a synonym | |
2425 | # for 'sc', and 'sc' can stand alone, meaning something | |
2426 | # entirely different than 'script'. 'script' cannot stand | |
2427 | # alone. Don't add if the potential new lhs is in the hash of | |
2428 | # stand-alone properties. | |
2429 | no warnings 'once'; | |
2430 | next if $rhs eq "" && grep { $alias eq $_ } | |
2431 | keys %utf8::loose_property_to_file_of; | |
2432 | ||
2433 | my $new_entry = $alias . $rhs; | |
e498c235 | 2434 | push @this_entries, $new_entry; |
394d2d3f KW |
2435 | } |
2436 | } | |
2437 | ||
2438 | # Above, we added the synonyms for the base entry we're now | |
2439 | # processing. But we haven't dealt with it yet. If we already have a | |
2440 | # property with the identical characteristics, this becomes just a | |
2441 | # synonym for it. | |
2442 | if (exists $enums{$tag}) { | |
2443 | push @this_entries, $property; | |
2444 | } | |
2445 | else { # Otherwise, create a new entry. | |
2446 | ||
4eea95a6 KW |
2447 | # Add to the list of properties to generate inversion lists for. |
2448 | push @bin_props, uc $property; | |
2449 | ||
394d2d3f | 2450 | # Create a rule for the parser |
f4b10e8e KW |
2451 | if (! exists $keywords{$property}) { |
2452 | $keywords{$property} = token_name($property); | |
2453 | } | |
394d2d3f | 2454 | |
4eea95a6 KW |
2455 | # And create an enum for it. |
2456 | $enums{$tag} = $table_name_prefix . uc sanitize_name($property); | |
394d2d3f | 2457 | |
1aefa327 KW |
2458 | $perl_tags{$tag} = 1 if exists $keep_together{lc $property}; |
2459 | ||
394d2d3f KW |
2460 | # Some properties are deprecated. This hash tells us so, and the |
2461 | # warning message to raise if they are used. | |
2462 | if (exists $utf8::why_deprecated{$tag}) { | |
2463 | $deprecated_tags{$enums{$tag}} = scalar @deprecated_messages; | |
2464 | push @deprecated_messages, $utf8::why_deprecated{$tag}; | |
2465 | } | |
2466 | ||
2467 | # Our sort above should have made sure that we see the | |
2468 | # non-inverted version first, but this makes sure. | |
2469 | warn "$property is inverted!!!" if $inverted; | |
2470 | } | |
2471 | ||
2472 | # Everything else is #defined to be the base enum, inversion is | |
2473 | # indicated by negating the value. | |
2474 | my $defined_to = ""; | |
2475 | $defined_to .= "-" if $inverted; | |
2476 | $defined_to .= $enums{$tag}; | |
2477 | ||
2478 | # Go through the entries that evaluate to this. | |
e498c235 | 2479 | @this_entries = uniques @this_entries; |
394d2d3f KW |
2480 | foreach my $define (@this_entries) { |
2481 | ||
2482 | # There is a rule for the parser for each. | |
f4b10e8e | 2483 | $keywords{$define} = $defined_to; |
1aefa327 KW |
2484 | |
2485 | # And a #define for all simple names equivalent to a perl property, | |
2486 | # except those that begin with 'is' or 'in'; | |
2487 | if (exists $perl_tags{$tag} && $property !~ / ^ i[ns] | = /x) { | |
2488 | push @perl_prop_synonyms, "#define " | |
2489 | . $table_name_prefix | |
2490 | . uc(sanitize_name($define)) | |
2491 | . " $defined_to"; | |
2492 | } | |
4eea95a6 KW |
2493 | } |
2494 | } | |
2495 | ||
2d74dcf2 KW |
2496 | @bin_props = sort { exists $keep_together{lc $b} <=> exists $keep_together{lc $a} |
2497 | or $a cmp $b | |
4eea95a6 | 2498 | } @bin_props; |
1aefa327 | 2499 | @perl_prop_synonyms = sort(uniques(@perl_prop_synonyms)); |
4eea95a6 KW |
2500 | push @props, @bin_props; |
2501 | ||
2502 | foreach my $prop (@props) { | |
2503 | ||
2504 | # For the Latin1 properties, we change to use the eXtended version of the | |
2505 | # base property, then go through the result and get rid of everything not | |
2506 | # in Latin1 (above 255). Actually, we retain the element for the range | |
2507 | # that crosses the 255/256 boundary if it is one that matches the | |
2508 | # property. For example, in the Word property, there is a range of code | |
2509 | # points that start at U+00F8 and goes through U+02C1. Instead of | |
2510 | # artificially cutting that off at 256 because 256 is the first code point | |
2511 | # above Latin1, we let the range go to its natural ending. That gives us | |
2512 | # extra information with no added space taken. But if the range that | |
2513 | # crosses the boundary is one that doesn't match the property, we don't | |
2514 | # start a new range above 255, as that could be construed as going to | |
2515 | # infinity. For example, the Upper property doesn't include the character | |
2516 | # at 255, but does include the one at 256. We don't include the 256 one. | |
2517 | my $prop_name = $prop; | |
2518 | my $is_local_sub = $prop_name =~ s/^&//; | |
2519 | my $extra_enums = ""; | |
2520 | $extra_enums = $1 if $prop_name =~ s/, ( .* ) //x; | |
2521 | my $lookup_prop = $prop_name; | |
2522 | $prop_name = sanitize_name($prop_name); | |
2523 | $prop_name = $table_name_prefix . $prop_name if grep { lc $lookup_prop eq lc $_ } @bin_props; | |
2524 | my $l1_only = ($lookup_prop =~ s/^L1Posix/XPosix/ | |
2525 | or $lookup_prop =~ s/^L1//); | |
2526 | my $nonl1_only = 0; | |
2527 | $nonl1_only = $lookup_prop =~ s/^NonL1// unless $l1_only; | |
2528 | ($lookup_prop, my $has_suffixes) = $lookup_prop =~ / (.*) ( , .* )? /x; | |
2529 | ||
4761f74a KW |
2530 | for my $charset (get_supported_code_pages()) { |
2531 | @a2n = @{get_a2n($charset)}; | |
2532 | ||
0f5e3c71 | 2533 | my @invlist; |
99f21fb9 KW |
2534 | my @invmap; |
2535 | my $map_format; | |
2536 | my $map_default; | |
2537 | my $maps_to_code_point; | |
2538 | my $to_adjust; | |
59fc10af | 2539 | my $same_in_all_code_pages; |
0f5e3c71 | 2540 | if ($is_local_sub) { |
8843f0de | 2541 | my @return = eval $lookup_prop; |
289ce9cc | 2542 | die $@ if $@; |
8843f0de KW |
2543 | my $invlist_ref = shift @return; |
2544 | @invlist = @$invlist_ref; | |
d2aadf62 KW |
2545 | if (@return) { # If has other values returned , must be an |
2546 | # inversion map | |
2547 | my $invmap_ref = shift @return; | |
2548 | @invmap = @$invmap_ref; | |
2549 | $map_format = shift @return; | |
2550 | $map_default = shift @return; | |
2551 | } | |
0f5e3c71 KW |
2552 | } |
2553 | else { | |
2554 | @invlist = prop_invlist($lookup_prop, '_perl_core_internal_ok'); | |
99f21fb9 | 2555 | if (! @invlist) { |
99f21fb9 | 2556 | |
ad85f59a KW |
2557 | # If couldn't find a non-empty inversion list, see if it is |
2558 | # instead an inversion map | |
2559 | my ($list_ref, $map_ref, $format, $default) | |
99f21fb9 | 2560 | = prop_invmap($lookup_prop, '_perl_core_internal_ok'); |
ad85f59a KW |
2561 | if (! $list_ref) { |
2562 | # An empty return here could mean an unknown property, or | |
2563 | # merely that the original inversion list is empty. Call | |
2564 | # in scalar context to differentiate | |
2565 | my $count = prop_invlist($lookup_prop, | |
2566 | '_perl_core_internal_ok'); | |
d99e65da KW |
2567 | if (defined $count) { |
2568 | # Short-circuit an empty inversion list. | |
2569 | output_invlist($prop_name, \@invlist, $charset); | |
59fc10af | 2570 | last; |
d99e65da | 2571 | } |
ad85f59a | 2572 | die "Could not find inversion list for '$lookup_prop'" |
ad85f59a KW |
2573 | } |
2574 | else { | |
18b852b3 KW |
2575 | @invlist = @$list_ref; |
2576 | @invmap = @$map_ref; | |
2577 | $map_format = $format; | |
2578 | $map_default = $default; | |
b148e8b1 | 2579 | $maps_to_code_point = $map_format =~ / a ($ | [^r] ) /x; |
18b852b3 | 2580 | $to_adjust = $map_format =~ /a/; |
ad85f59a | 2581 | } |
99f21fb9 | 2582 | } |
0f5e3c71 | 2583 | } |
ad85f59a | 2584 | |
99f21fb9 KW |
2585 | # Re-order the Unicode code points to native ones for this platform. |
2586 | # This is only needed for code points below 256, because native code | |
2587 | # points are only in that range. For inversion maps of properties | |
2588 | # where the mappings are adjusted (format =~ /a/), this reordering | |
2589 | # could mess up the adjustment pattern that was in the input, so that | |
2590 | # has to be dealt with. | |
2591 | # | |
2592 | # And inversion maps that map to code points need to eventually have | |
2593 | # all those code points remapped to native, and it's better to do that | |
2594 | # here, going through the whole list not just those below 256. This | |
2595 | # is because some inversion maps have adjustments (format =~ /a/) | |
2596 | # which may be affected by the reordering. This code needs to be done | |
2597 | # both for when we are translating the inversion lists for < 256, and | |
2598 | # for the inversion maps for everything. By doing both in this loop, | |
2599 | # we can share that code. | |
2600 | # | |
2601 | # So, we go through everything for an inversion map to code points; | |
2602 | # otherwise, we can skip any remapping at all if we are going to | |
2603 | # output only the above-Latin1 values, or if the range spans the whole | |
2604 | # of 0..256, as the remap will also include all of 0..256 (256 not | |
2605 | # 255 because a re-ordering could cause 256 to need to be in the same | |
2606 | # range as 255.) | |
2b3e8a91 | 2607 | if ( (@invmap && $maps_to_code_point) |
e4e80abb KW |
2608 | || ( @invlist |
2609 | && $invlist[0] < 256 | |
2b3e8a91 | 2610 | && ( $invlist[0] != 0 |
e4e80abb | 2611 | || (scalar @invlist != 1 && $invlist[1] < 256)))) |
ceb1de32 | 2612 | { |
59fc10af | 2613 | $same_in_all_code_pages = 0; |
99f21fb9 | 2614 | if (! @invmap) { # Straight inversion list |
563f8b93 KW |
2615 | # Look at all the ranges that start before 257. |
2616 | my @latin1_list; | |
2617 | while (@invlist) { | |
2618 | last if $invlist[0] > 256; | |
2619 | my $upper = @invlist > 1 | |
2620 | ? $invlist[1] - 1 # In range | |
2621 | ||
2622 | # To infinity. You may want to stop much much | |
2623 | # earlier; going this high may expose perl | |
2624 | # deficiencies with very large numbers. | |
7d2c6c24 | 2625 | : 256; |
563f8b93 KW |
2626 | for my $j ($invlist[0] .. $upper) { |
2627 | push @latin1_list, a2n($j); | |
2628 | } | |
fb4554ea | 2629 | |
563f8b93 KW |
2630 | shift @invlist; # Shift off the range that's in the list |
2631 | shift @invlist; # Shift off the range not in the list | |
2632 | } | |
fb4554ea | 2633 | |
563f8b93 KW |
2634 | # Here @invlist contains all the ranges in the original that |
2635 | # start at code points above 256, and @latin1_list contains | |
2636 | # all the native code points for ranges that start with a | |
2637 | # Unicode code point below 257. We sort the latter and | |
2638 | # convert it to inversion list format. Then simply prepend it | |
2639 | # to the list of the higher code points. | |
2640 | @latin1_list = sort { $a <=> $b } @latin1_list; | |
2641 | @latin1_list = mk_invlist_from_sorted_cp_list(\@latin1_list); | |
2642 | unshift @invlist, @latin1_list; | |
99f21fb9 KW |
2643 | } |
2644 | else { # Is an inversion map | |
2645 | ||
2646 | # This is a similar procedure as plain inversion list, but has | |
2647 | # multiple buckets. A plain inversion list just has two | |
2648 | # buckets, 1) 'in' the list; and 2) 'not' in the list, and we | |
2649 | # pretty much can ignore the 2nd bucket, as it is completely | |
2650 | # defined by the 1st. But here, what we do is create buckets | |
2651 | # which contain the code points that map to each, translated | |
2652 | # to native and turned into an inversion list. Thus each | |
2653 | # bucket is an inversion list of native code points that map | |
2654 | # to it or don't map to it. We use these to create an | |
2655 | # inversion map for the whole property. | |
2656 | ||
2657 | # As mentioned earlier, we use this procedure to not just | |
2658 | # remap the inversion list to native values, but also the maps | |
2659 | # of code points to native ones. In the latter case we have | |
2660 | # to look at the whole of the inversion map (or at least to | |
2661 | # above Unicode; as the maps of code points above that should | |
2662 | # all be to the default). | |
c125794e KW |
2663 | my $upper_limit = (! $maps_to_code_point) |
2664 | ? 256 | |
2665 | : (Unicode::UCD::UnicodeVersion() eq '1.1.5') | |
2666 | ? 0xFFFF | |
2667 | : 0x10FFFF; | |
99f21fb9 KW |
2668 | |
2669 | my %mapped_lists; # A hash whose keys are the buckets. | |
2670 | while (@invlist) { | |
2671 | last if $invlist[0] > $upper_limit; | |
2672 | ||
2673 | # This shouldn't actually happen, as prop_invmap() returns | |
2674 | # an extra element at the end that is beyond $upper_limit | |
7e2c536f | 2675 | die "inversion map (for $prop_name) that extends to infinity is unimplemented" unless @invlist > 1; |
99f21fb9 KW |
2676 | |
2677 | my $bucket; | |
2678 | ||
2679 | # A hash key can't be a ref (we are only expecting arrays | |
2680 | # of scalars here), so convert any such to a string that | |
2681 | # will be converted back later (using a vertical tab as | |
b148e8b1 | 2682 | # the separator). |
99f21fb9 | 2683 | if (ref $invmap[0]) { |
b148e8b1 | 2684 | $bucket = join "\cK", map { a2n($_) } @{$invmap[0]}; |
99f21fb9 KW |
2685 | } |
2686 | elsif ($maps_to_code_point && $invmap[0] =~ $numeric_re) { | |
2687 | ||
2688 | # Do convert to native for maps to single code points. | |
2689 | # There are some properties that have a few outlier | |
2690 | # maps that aren't code points, so the above test | |
2691 | # skips those. | |
2692 | $bucket = a2n($invmap[0]); | |
2693 | } else { | |
2694 | $bucket = $invmap[0]; | |
2695 | } | |
2696 | ||
2697 | # We now have the bucket that all code points in the range | |
2698 | # map to, though possibly they need to be adjusted. Go | |
2699 | # through the range and put each translated code point in | |
2700 | # it into its bucket. | |
2701 | my $base_map = $invmap[0]; | |
2702 | for my $j ($invlist[0] .. $invlist[1] - 1) { | |
2703 | if ($to_adjust | |
2704 | # The 1st code point doesn't need adjusting | |
2705 | && $j > $invlist[0] | |
2706 | ||
2707 | # Skip any non-numeric maps: these are outliers | |
2708 | # that aren't code points. | |
2709 | && $base_map =~ $numeric_re | |
2710 | ||
2711 | # 'ne' because the default can be a string | |
2712 | && $base_map ne $map_default) | |
2713 | { | |
2714 | # We adjust, by incrementing each the bucket and | |
2715 | # the map. For code point maps, translate to | |
2716 | # native | |
2717 | $base_map++; | |
2718 | $bucket = ($maps_to_code_point) | |
2719 | ? a2n($base_map) | |
2720 | : $base_map; | |
2721 | } | |
2722 | ||
2723 | # Add the native code point to the bucket for the | |
2724 | # current map | |
2725 | push @{$mapped_lists{$bucket}}, a2n($j); | |
2726 | } # End of loop through all code points in the range | |
2727 | ||
2728 | # Get ready for the next range | |
2729 | shift @invlist; | |
2730 | shift @invmap; | |
2731 | } # End of loop through all ranges in the map. | |
2732 | ||
2733 | # Here, @invlist and @invmap retain all the ranges from the | |
2734 | # originals that start with code points above $upper_limit. | |
2735 | # Each bucket in %mapped_lists contains all the code points | |
2736 | # that map to that bucket. If the bucket is for a map to a | |
5174a821 KW |
2737 | # single code point, the bucket has been converted to native. |
2738 | # If something else (including multiple code points), no | |
2739 | # conversion is done. | |
99f21fb9 KW |
2740 | # |
2741 | # Now we recreate the inversion map into %xlated, but this | |
2742 | # time for the native character set. | |
2743 | my %xlated; | |
2744 | foreach my $bucket (keys %mapped_lists) { | |
2745 | ||
2746 | # Sort and convert this bucket to an inversion list. The | |
2747 | # result will be that ranges that start with even-numbered | |
2748 | # indexes will be for code points that map to this bucket; | |
2749 | # odd ones map to some other bucket, and are discarded | |
2750 | # below. | |
2751 | @{$mapped_lists{$bucket}} | |
2752 | = sort{ $a <=> $b} @{$mapped_lists{$bucket}}; | |
2753 | @{$mapped_lists{$bucket}} | |
2754 | = mk_invlist_from_sorted_cp_list(\@{$mapped_lists{$bucket}}); | |
2755 | ||
2756 | # Add each even-numbered range in the bucket to %xlated; | |
2757 | # so that the keys of %xlated become the range start code | |
2758 | # points, and the values are their corresponding maps. | |
2759 | while (@{$mapped_lists{$bucket}}) { | |
2760 | my $range_start = $mapped_lists{$bucket}->[0]; | |
2761 | if ($bucket =~ /\cK/) { | |
2762 | @{$xlated{$range_start}} = split /\cK/, $bucket; | |
2763 | } | |
2764 | else { | |
e113b1b3 KW |
2765 | # If adjusting, and there is more than one thing |
2766 | # that maps to the same thing, they must be split | |
2767 | # so that later the adjusting doesn't think the | |
2768 | # subsequent items can go away because of the | |
2769 | # adjusting. | |
2770 | my $range_end = ($to_adjust && $bucket != $map_default) | |
2771 | ? $mapped_lists{$bucket}->[1] - 1 | |
2772 | : $range_start; | |
2773 | for my $i ($range_start .. $range_end) { | |
2774 | $xlated{$i} = $bucket; | |
2775 | } | |
99f21fb9 KW |
2776 | } |
2777 | shift @{$mapped_lists{$bucket}}; # Discard odd ranges | |
2778 | shift @{$mapped_lists{$bucket}}; # Get ready for next | |
2779 | # iteration | |
2780 | } | |
2781 | } # End of loop through all the buckets. | |
2782 | ||
2783 | # Here %xlated's keys are the range starts of all the code | |
2784 | # points in the inversion map. Construct an inversion list | |
2785 | # from them. | |
2786 | my @new_invlist = sort { $a <=> $b } keys %xlated; | |
2787 | ||
2788 | # If the list is adjusted, we want to munge this list so that | |
2789 | # we only have one entry for where consecutive code points map | |
2790 | # to consecutive values. We just skip the subsequent entries | |
2791 | # where this is the case. | |
2792 | if ($to_adjust) { | |
2793 | my @temp; | |
2794 | for my $i (0 .. @new_invlist - 1) { | |
2795 | next if $i > 0 | |
2796 | && $new_invlist[$i-1] + 1 == $new_invlist[$i] | |
2797 | && $xlated{$new_invlist[$i-1]} =~ $numeric_re | |
2798 | && $xlated{$new_invlist[$i]} =~ $numeric_re | |
2799 | && $xlated{$new_invlist[$i-1]} + 1 == $xlated{$new_invlist[$i]}; | |
2800 | push @temp, $new_invlist[$i]; | |
2801 | } | |
2802 | @new_invlist = @temp; | |
2803 | } | |
2804 | ||
2805 | # The inversion map comes from %xlated's values. We can | |
2806 | # unshift each onto the front of the untouched portion, in | |
2807 | # reverse order of the portion we did process. | |
2808 | foreach my $start (reverse @new_invlist) { | |
2809 | unshift @invmap, $xlated{$start}; | |
2810 | } | |
2811 | ||
2812 | # Finally prepend the inversion list we have just constructed to the | |
2813 | # one that contains anything we didn't process. | |
2814 | unshift @invlist, @new_invlist; | |
2815 | } | |
2816 | } | |
e4e80abb KW |
2817 | elsif (@invmap) { # inversion maps can't cope with this variable |
2818 | # being true, even if it could be true | |
2819 | $same_in_all_code_pages = 0; | |
2820 | } | |
59fc10af KW |
2821 | else { |
2822 | $same_in_all_code_pages = 1; | |
2823 | } | |
99f21fb9 KW |
2824 | |
2825 | # prop_invmap() returns an extra final entry, which we can now | |
2826 | # discard. | |
2827 | if (@invmap) { | |
2828 | pop @invlist; | |
2829 | pop @invmap; | |
ceb1de32 | 2830 | } |
0f5e3c71 KW |
2831 | |
2832 | if ($l1_only) { | |
99f21fb9 | 2833 | die "Unimplemented to do a Latin-1 only inversion map" if @invmap; |
0f5e3c71 KW |
2834 | for my $i (0 .. @invlist - 1 - 1) { |
2835 | if ($invlist[$i] > 255) { | |
2836 | ||
2837 | # In an inversion list, even-numbered elements give the code | |
2838 | # points that begin ranges that match the property; | |
2839 | # odd-numbered give ones that begin ranges that don't match. | |
2840 | # If $i is odd, we are at the first code point above 255 that | |
2841 | # doesn't match, which means the range it is ending does | |
2842 | # match, and crosses the 255/256 boundary. We want to include | |
2843 | # this ending point, so increment $i, so the splice below | |
2844 | # includes it. Conversely, if $i is even, it is the first | |
2845 | # code point above 255 that matches, which means there was no | |
2846 | # matching range that crossed the boundary, and we don't want | |
2847 | # to include this code point, so splice before it. | |
2848 | $i++ if $i % 2 != 0; | |
2849 | ||
2850 | # Remove everything past this. | |
2851 | splice @invlist, $i; | |
99f21fb9 | 2852 | splice @invmap, $i if @invmap; |
0f5e3c71 KW |
2853 | last; |
2854 | } | |
0c4ecf42 KW |
2855 | } |
2856 | } | |
0f5e3c71 KW |
2857 | elsif ($nonl1_only) { |
2858 | my $found_nonl1 = 0; | |
2859 | for my $i (0 .. @invlist - 1 - 1) { | |
2860 | next if $invlist[$i] < 256; | |
2861 | ||
2862 | # Here, we have the first element in the array that indicates an | |
2863 | # element above Latin1. Get rid of all previous ones. | |
2864 | splice @invlist, 0, $i; | |
99f21fb9 | 2865 | splice @invmap, 0, $i if @invmap; |
0f5e3c71 KW |
2866 | |
2867 | # If this one's index is not divisible by 2, it means that this | |
2868 | # element is inverting away from being in the list, which means | |
99f21fb9 KW |
2869 | # all code points from 256 to this one are in this list (or |
2870 | # map to the default for inversion maps) | |
2871 | if ($i % 2 != 0) { | |
2872 | unshift @invlist, 256; | |
2873 | unshift @invmap, $map_default if @invmap; | |
2874 | } | |
0f5e3c71 | 2875 | $found_nonl1 = 1; |
3f427fd9 KW |
2876 | last; |
2877 | } | |
0f0b3751 KW |
2878 | if (! $found_nonl1) { |
2879 | warn "No non-Latin1 code points in $prop_name"; | |
2880 | output_invlist($prop_name, []); | |
2881 | last; | |
2882 | } | |
3f427fd9 | 2883 | } |
3f427fd9 | 2884 | |
cef72199 | 2885 | switch_pound_if ($prop_name, 'PERL_IN_REGCOMP_C'); |
59fc10af | 2886 | start_charset_pound_if($charset, 1) unless $same_in_all_code_pages; |
4761f74a | 2887 | |
59fc10af KW |
2888 | output_invlist($prop_name, \@invlist, ($same_in_all_code_pages) |
2889 | ? $applies_to_all_charsets_text | |
2890 | : $charset); | |
4761f74a KW |
2891 | |
2892 | if (@invmap) { | |
2893 | output_invmap($prop_name, \@invmap, $lookup_prop, $map_format, | |
2894 | $map_default, $extra_enums, $charset); | |
2895 | } | |
59fc10af KW |
2896 | |
2897 | last if $same_in_all_code_pages; | |
4761f74a | 2898 | end_charset_pound_if; |
0f5e3c71 | 2899 | } |
9d9177be KW |
2900 | } |
2901 | ||
cef72199 | 2902 | switch_pound_if ('binary_property_tables', 'PERL_IN_REGCOMP_C'); |
394d2d3f KW |
2903 | |
2904 | print $out_fh "\nconst char * deprecated_property_msgs[] = {\n\t"; | |
2905 | print $out_fh join ",\n\t", map { "\"$_\"" } @deprecated_messages; | |
2906 | print $out_fh "\n};\n"; | |
2907 | ||
394d2d3f KW |
2908 | my @enums = sort values %enums; |
2909 | ||
2910 | # Save a copy of these before modification | |
2911 | my @invlist_names = map { "${_}_invlist" } @enums; | |
2912 | ||
2913 | # Post-process the enums for deprecated properties. | |
2914 | if (scalar keys %deprecated_tags) { | |
2915 | my $seen_deprecated = 0; | |
2916 | foreach my $enum (@enums) { | |
2917 | if (grep { $_ eq $enum } keys %deprecated_tags) { | |
2918 | ||
2919 | # Change the enum name for this deprecated property to a | |
2920 | # munged one to act as a placeholder in the typedef. Then | |
2921 | # make the real name be a #define whose value is such that | |
2922 | # its modulus with the number of enums yields the index into | |
2923 | # the table occupied by the placeholder. And so that dividing | |
2924 | # the #define value by the table length gives an index into | |
2925 | # the table of deprecation messages for the corresponding | |
2926 | # warning. | |
2927 | my $revised_enum = "${enum}_perl_aux"; | |
2928 | if (! $seen_deprecated) { | |
2929 | $seen_deprecated = 1; | |
2930 | print $out_fh "\n"; | |
2931 | } | |
2932 | print $out_fh "#define $enum ($revised_enum + (MAX_UNI_KEYWORD_INDEX * $deprecated_tags{$enum}))\n"; | |
2933 | $enum = $revised_enum; | |
2934 | } | |
2935 | } | |
2936 | } | |
2937 | ||
2938 | print $out_fh "\ntypedef enum {\n\tPERL_BIN_PLACEHOLDER = 0,\n\t"; | |
2939 | print $out_fh join ",\n\t", @enums; | |
2940 | print $out_fh "\n"; | |
2941 | print $out_fh "} binary_invlist_enum;\n"; | |
2942 | print $out_fh "\n#define MAX_UNI_KEYWORD_INDEX $enums[-1]\n"; | |
394d2d3f | 2943 | |
cef72199 | 2944 | output_table_header($out_fh, "UV *", "uni_prop_ptrs"); |
394d2d3f | 2945 | print $out_fh "\tNULL,\t/* Placeholder */\n\t"; |
cef72199 | 2946 | print $out_fh "\t"; |
394d2d3f KW |
2947 | print $out_fh join ",\n\t", @invlist_names; |
2948 | print $out_fh "\n"; | |
cef72199 KW |
2949 | |
2950 | output_table_trailer(); | |
2951 | ||
2952 | print $out_fh join "\n", "\n", | |
2953 | #'# ifdef DOINIT', | |
2954 | #"\n", | |
2955 | "/* Synonyms for perl properties */", | |
2956 | @perl_prop_synonyms, | |
2957 | #"\n", | |
2958 | #"# endif /* DOINIT */", | |
2959 | "\n"; | |
394d2d3f | 2960 | |
973a28ed KW |
2961 | switch_pound_if('Boundary_pair_tables', 'PERL_IN_REGEXEC_C'); |
2962 | ||
2963 | output_GCB_table(); | |
6b659339 | 2964 | output_LB_table(); |
7e54b87f | 2965 | output_WB_table(); |
6b659339 | 2966 | |
973a28ed KW |
2967 | end_file_pound_if; |
2968 | ||
2308ab83 | 2969 | my $sources_list = "lib/unicore/mktables.lst"; |
74e28a4a TC |
2970 | my @sources = qw(regen/mk_invlists.pl |
2971 | lib/unicore/mktables | |
2972 | lib/Unicode/UCD.pm | |
2973 | regen/charset_translations.pl | |
f7b69ff8 | 2974 | regen/mk_PL_charclass.pl |
74e28a4a | 2975 | ); |
9a3da3ad FC |
2976 | { |
2977 | # Depend on mktables’ own sources. It’s a shorter list of files than | |
2978 | # those that Unicode::UCD uses. | |
1ae6ead9 | 2979 | if (! open my $mktables_list, '<', $sources_list) { |
2308ab83 KW |
2980 | |
2981 | # This should force a rebuild once $sources_list exists | |
2982 | push @sources, $sources_list; | |
2983 | } | |
2984 | else { | |
2985 | while(<$mktables_list>) { | |
2986 | last if /===/; | |
2987 | chomp; | |
2988 | push @sources, "lib/unicore/$_" if /^[^#]/; | |
2989 | } | |
9a3da3ad FC |
2990 | } |
2991 | } | |
6b659339 KW |
2992 | |
2993 | read_only_bottom_close_and_rename($out_fh, \@sources); | |
394d2d3f | 2994 | |
afde5508 | 2995 | require './regen/mph.pl'; |
394d2d3f KW |
2996 | |
2997 | sub token_name | |
2998 | { | |
2999 | my $name = sanitize_name(shift); | |
db95f459 | 3000 | warn "$name contains non-word" if $name =~ /\W/; |
394d2d3f | 3001 | |
afde5508 | 3002 | return "$table_name_prefix\U$name" |
394d2d3f KW |
3003 | } |
3004 | ||
afde5508 | 3005 | my $keywords_fh = open_new('uni_keywords.h', '>', |
394d2d3f | 3006 | {style => '*', by => 'regen/mk_invlists.pl', |
afde5508 | 3007 | from => "mph.pl"}); |
394d2d3f | 3008 | |
27097618 KW |
3009 | no warnings 'once'; |
3010 | print $keywords_fh <<"EOF"; | |
5ae55d32 | 3011 | /* The precision to use in "%.*e" formats */ |
27097618 KW |
3012 | #define PL_E_FORMAT_PRECISION $utf8::e_precision |
3013 | ||
3014 | EOF | |
3015 | ||
f4b10e8e | 3016 | my ($second_level, $seed1, $length_all_keys, $smart_blob, $rows) = MinimalPerfectHash::make_mph_from_hash(\%keywords); |
afde5508 KW |
3017 | print $keywords_fh MinimalPerfectHash::make_algo($second_level, $seed1, $length_all_keys, $smart_blob, $rows, undef, undef, undef, 'match_uniprop' ); |
3018 | ||
3019 | push @sources, 'regen/mph.pl'; | |
394d2d3f | 3020 | read_only_bottom_close_and_rename($keywords_fh, \@sources); |