Commit | Line | Data |
---|---|---|
61507e3c RK |
1 | #! /usr/bin/perl -w |
2 | # | |
e5a5a138 RK |
3 | # This file is part of DisOrder. |
4 | # Copyright (C) 2007 Richard Kettlewell | |
5 | # | |
6 | # This program is free software; you can redistribute it and/or modify | |
7 | # it under the terms of the GNU General Public License as published by | |
8 | # the Free Software Foundation; either version 2 of the License, or | |
9 | # (at your option) any later version. | |
10 | # | |
11 | # This program is distributed in the hope that it will be useful, but | |
12 | # WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | # General Public License for more details. | |
15 | # | |
16 | # You should have received a copy of the GNU General Public License | |
17 | # along with this program; if not, write to the Free Software | |
18 | # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 | |
19 | # USA | |
20 | # | |
21 | # | |
22 | # Generate Unicode support tables | |
23 | # | |
24 | # This script will download data from unicode.org if the required files | |
25 | # aren't in the current directory. | |
26 | # | |
27 | # After modifying this script you should run: | |
28 | # make -C lib rebuild-unicode check | |
29 | # | |
30 | # Things not supported yet: | |
31 | # - SpecialCasing.txt data for case mapping | |
32 | # - Title case offsets | |
33 | # - Some kind of hinting for composition | |
e5a5a138 RK |
34 | # - ... |
35 | # | |
36 | # NB the generated files DO NOT offer a stable ABI and so are not immediately | |
37 | # suitable for use in a general-purpose library. Things that would need to | |
38 | # be done: | |
39 | # - Hide unidata.h from applications; it will never be ABI- or even API-stable. | |
40 | # - Stablized General_Category values | |
41 | # - Extend the unicode.h API to general utility rather than just what | |
42 | # DisOrder needs. | |
43 | # - ... | |
44 | # | |
61507e3c | 45 | use strict; |
35b651f0 | 46 | use File::Basename; |
61507e3c RK |
47 | |
48 | sub out { | |
49 | print @_ or die "$!\n"; | |
50 | } | |
51 | ||
52 | sub key { | |
53 | my $d = shift; | |
54 | local $_; | |
55 | ||
56 | return join("-", map($d->{$_}, sort keys %$d)); | |
57 | } | |
58 | ||
e5a5a138 RK |
59 | # Size of a subtable |
60 | # | |
61 | # This can be varied to trade off the number of subtables against their size. | |
c2e01e0a RK |
62 | # 16 gave the smallest results last time I checked (on a Mac with a 32-bit |
63 | # build). | |
64 | our $modulus = 16; | |
65 | ||
66 | if(@ARGV) { | |
67 | $modulus = shift; | |
68 | } | |
e5a5a138 | 69 | |
1a05e381 RK |
70 | # Where to break the table. There is a huge empty section of the Unicode |
71 | # code space and we deal with this by simply leaving it out of the table. | |
72 | # This complicates the lookup function a little but should not affect | |
73 | # performance in the cases we care about. | |
74 | our $break_start = 0x30000; | |
75 | our $break_end = 0xE0000; | |
76 | ||
77 | # Similarly we simply omit the very top of the table and sort it out in the | |
78 | # lookup function. | |
79 | our $break_top = 0xE0200; | |
80 | ||
61507e3c RK |
81 | my %cats = (); # known general categories |
82 | my %data = (); # mapping of codepoints to information | |
61507e3c | 83 | my $max = 0; # maximum codepoint |
e5a5a138 RK |
84 | my $maxccc = 0; # maximum combining class |
85 | my $maxud = 0; | |
86 | my $minud = 0; # max/min upper case offset | |
87 | my $maxld = 0; | |
88 | my $minld = 0; # max/min lower case offset | |
89 | ||
90 | # Make sure we have our desired input files. We explicitly specify a | |
91 | # Unicode standard version to make sure that a given version of DisOrder | |
92 | # supports a given version of Unicode. | |
0b7052da | 93 | sub input { |
35b651f0 RK |
94 | my $path = shift; |
95 | my $lpath = basename($path); | |
96 | if(!-e $lpath) { | |
97 | system("wget http://www.unicode.org/Public/5.0.0/ucd/$path"); | |
98 | chmod(0444, $lpath) or die "$lpath: $!\n"; | |
e5a5a138 | 99 | } |
0b7052da | 100 | open(STDIN, "<$lpath") or die "$lpath: $!\n"; |
bcf9ed7f | 101 | print STDERR "Reading $lpath...\n"; |
e5a5a138 | 102 | } |
61507e3c | 103 | |
e5a5a138 RK |
104 | |
105 | # Read the main data file | |
0b7052da | 106 | input("UnicodeData.txt"); |
1a05e381 | 107 | my ($start, $end); |
16506c9d RK |
108 | my $maxcompat = 0; |
109 | my $maxcanon = 0; | |
110 | my $hangul_syllable_decomps = 0; | |
111 | my $hangul_choseong_decomps = 0; | |
61507e3c RK |
112 | while(<>) { |
113 | my @f = split(/;/, $_); | |
114 | my $c = hex($f[0]); # codepoint | |
61507e3c | 115 | my $name = $f[1]; |
1a05e381 RK |
116 | die "$f[0] $name is in the break\n" |
117 | if $c >= $break_start && $c < $break_end; | |
e5a5a138 | 118 | my $gc = $f[2]; # General_Category |
bcf9ed7f | 119 | # Variuos GCs we don't expect to see in UnicodeData.txt |
0a4af692 | 120 | $cats{$gc} = 1; # always record all GCs |
1a05e381 RK |
121 | if($name =~ /first>/i) { |
122 | $start = $c; | |
123 | next; | |
124 | } elsif($name =~ /last>/i) { | |
125 | $end = $c; | |
126 | } else { | |
127 | $start = $end = $c; | |
128 | } | |
bcf9ed7f | 129 | die "unexpected Cn" if $gc eq 'Cn'; |
e5a5a138 RK |
130 | my $ccc = $f[3]; # Canonical_Combining_Class |
131 | my $dm = $f[5]; # Decomposition_Type + Decomposition_Mapping | |
132 | my $sum = hex($f[12]) || $c; # Simple_Uppercase_Mapping | |
133 | my $slm = hex($f[13]) || $c; # Simple_Lowercase_Mapping | |
61507e3c RK |
134 | # recalculate the upper/lower case mappings as offsets |
135 | my $ud = $sum - $c; | |
136 | my $ld = $slm - $c; | |
e5a5a138 RK |
137 | # update bounds on various values |
138 | $maxccc = $ccc if $ccc > $maxccc; # assumed never to be -ve | |
139 | $minud = $ud if $ud < $minud; | |
140 | $maxud = $ud if $ud > $maxud; | |
141 | $minld = $ld if $ld < $minld; | |
142 | $maxld = $ld if $ld > $maxld; | |
1a05e381 | 143 | if($start != $end) { |
0e843521 | 144 | printf STDERR "> range %04X-%04X is %s\n", $start, $end, $gc; |
1a05e381 RK |
145 | } |
146 | for($c = $start; $c <= $end; ++$c) { | |
0e843521 RK |
147 | my $d = { |
148 | "gc" => $gc, | |
149 | "ccc" => $ccc, | |
150 | "ud" => $ud, | |
151 | "ld" => $ld, | |
152 | }; | |
153 | if($dm ne '') { | |
16506c9d | 154 | my $maxref; |
f98fcddb RK |
155 | if($dm =~ /</) { |
156 | # This is a compatibility decomposition | |
0e843521 | 157 | $dm =~ s/^<.*>\s*//; |
f98fcddb | 158 | $d->{compat} = 1; |
16506c9d RK |
159 | $maxref = \$maxcompat; |
160 | } else { | |
161 | $maxref = \$maxcanon; | |
0e843521 | 162 | } |
99695df9 | 163 | $d->{decomp} = [map(hex($_), split(/\s+/, $dm))]; |
16506c9d RK |
164 | my $len = scalar @{$d->{decomp}}; |
165 | $$maxref = $len if $len > $$maxref; | |
166 | if(!$d->{compat}) { | |
167 | if(${$d->{decomp}}[0] >= 0xAC00 && ${$d->{decomp}}[0] <= 0xD7A3) { | |
168 | ++$hangul_syllable_decomps; | |
169 | } | |
170 | if(${$d->{decomp}}[0] >= 0x1100 && ${$d->{decomp}}[0] <= 0x115F) { | |
171 | ++$hangul_choseong_decomps; | |
172 | } | |
173 | } | |
0e843521 | 174 | } |
1a05e381 RK |
175 | $data{$c} = $d; |
176 | } | |
61507e3c | 177 | $cats{$gc} = 1; |
1a05e381 | 178 | $max = $end if $end > $max; |
61507e3c RK |
179 | } |
180 | ||
0b7052da RK |
181 | sub read_prop_with_ranges { |
182 | my $path = shift; | |
183 | my $propkey = shift; | |
184 | input($path); | |
185 | while(<>) { | |
186 | chomp; | |
187 | s/\s*\#.*//; | |
188 | next if $_ eq ''; | |
189 | my ($range, $propval) = split(/\s*;\s*/, $_); | |
190 | if($range =~ /(.*)\.\.(.*)/) { | |
191 | for my $c (hex($1) .. hex($2)) { | |
0e843521 | 192 | $data{$c}->{$propkey} = $propval; |
0b7052da RK |
193 | } |
194 | } else { | |
195 | my $c = hex($range); | |
0e843521 | 196 | $data{$c}->{$propkey} = $propval; |
35b651f0 | 197 | } |
0b7052da RK |
198 | } |
199 | } | |
200 | ||
349b7b74 | 201 | # Grapheme_Break etc |
0b7052da | 202 | read_prop_with_ranges("auxiliary/GraphemeBreakProperty.txt", "gbreak"); |
0b7052da | 203 | read_prop_with_ranges("auxiliary/WordBreakProperty.txt", "wbreak"); |
349b7b74 | 204 | read_prop_with_ranges("auxiliary/SentenceBreakProperty.txt", "sbreak"); |
0b7052da | 205 | |
349b7b74 RK |
206 | # Compute the full list and fill in the Extend category properly |
207 | my %gbreak = (); | |
208 | my %wbreak = (); | |
209 | my %sbreak = (); | |
0b7052da | 210 | for my $c (keys %data) { |
349b7b74 RK |
211 | if(!exists $data{$c}->{gbreak}) { |
212 | $data{$c}->{gbreak} = 'Other'; | |
213 | } | |
214 | $gbreak{$data{$c}->{gbreak}} = 1; | |
215 | ||
0b7052da | 216 | if(!exists $data{$c}->{wbreak}) { |
349b7b74 | 217 | if($data{$c}->{gbreak} eq 'Extend') { |
0b7052da RK |
218 | $data{$c}->{wbreak} = 'Extend'; |
219 | } else { | |
220 | $data{$c}->{wbreak} = 'Other'; | |
35b651f0 RK |
221 | } |
222 | } | |
349b7b74 RK |
223 | $wbreak{$data{$c}->{wbreak}} = 1; |
224 | ||
225 | if(!exists $data{$c}->{sbreak}) { | |
226 | if($data{$c}->{gbreak} eq 'Extend') { | |
227 | $data{$c}->{sbreak} = 'Extend'; | |
228 | } else { | |
229 | $data{$c}->{sbreak} = 'Other'; | |
230 | } | |
231 | } | |
232 | $sbreak{$data{$c}->{sbreak}} = 1; | |
35b651f0 RK |
233 | } |
234 | ||
99695df9 RK |
235 | # Various derived properties |
236 | input("DerivedNormalizationProps.txt"); | |
237 | while(<>) { | |
238 | chomp; | |
239 | s/\s*\#.*//; | |
240 | next if $_ eq ''; | |
241 | my @f = split(/\s*;\s*/, $_); | |
242 | if(@f == 2) { | |
243 | push(@f, 1); | |
244 | } | |
245 | my ($range, $propkey, $propval) = @f; | |
246 | if($range =~ /(.*)\.\.(.*)/) { | |
247 | for my $c (hex($1) .. hex($2)) { | |
248 | $data{$c}->{$propkey} = $propval | |
249 | } | |
250 | } else { | |
251 | my $c = hex($range); | |
252 | $data{$c}->{$propkey} = $propval | |
253 | } | |
254 | } | |
255 | ||
e5a5a138 RK |
256 | # Round up the maximum value to a whole number of subtables |
257 | $max += ($modulus - 1) - ($max % $modulus); | |
61507e3c | 258 | |
bcf9ed7f RK |
259 | # Private use characters |
260 | # We only fill in values below $max, utf32__unidata() | |
261 | my $Co = { | |
262 | "gc" => "Co", | |
263 | "ccc" => 0, | |
264 | "ud" => 0, | |
265 | "ld" => 0 | |
266 | }; | |
267 | for(my $c = 0xE000; $c <= 0xF8FF && $c <= $max; ++$c) { | |
268 | $data{$c} = $Co; | |
269 | } | |
270 | for(my $c = 0xF0000; $c <= 0xFFFFD && $c <= $max; ++$c) { | |
271 | $data{$c} = $Co; | |
272 | } | |
273 | for(my $c = 0x100000; $c <= 0x10FFFD && $c <= $max; ++$c) { | |
274 | $data{$c} = $Co; | |
275 | } | |
276 | ||
277 | # Anything left is not assigned | |
278 | my $Cn = { | |
279 | "gc" => "Cn", # not assigned | |
280 | "ccc" => 0, | |
281 | "ud" => 0, | |
282 | "ld" => 0 | |
283 | }; | |
61507e3c RK |
284 | for(my $c = 0; $c <= $max; ++$c) { |
285 | if(!exists $data{$c}) { | |
bcf9ed7f RK |
286 | $data{$c} = $Cn; |
287 | } | |
288 | if(!exists $data{$c}->{wbreak}) { | |
289 | $data{$c}->{wbreak} = 'Other'; | |
290 | } | |
291 | if(!exists $data{$c}->{gbreak}) { | |
292 | $data{$c}->{gbreak} = 'Other'; | |
293 | } | |
294 | if(!exists $data{$c}->{sbreak}) { | |
295 | $data{$c}->{sbreak} = 'Other'; | |
61507e3c RK |
296 | } |
297 | } | |
298 | $cats{'Cn'} = 1; | |
299 | ||
e5a5a138 | 300 | # Read the casefolding data too |
0b7052da | 301 | input("CaseFolding.txt"); |
e5a5a138 RK |
302 | while(<>) { |
303 | chomp; | |
304 | next if /^\#/ or $_ eq ''; | |
305 | my @f = split(/\s*;\s*/, $_); | |
306 | # Full case folding means use status C and F. | |
307 | # We discard status T, Turkish users may wish to change this. | |
308 | if($f[1] eq 'C' or $f[1] eq 'F') { | |
309 | my $c = hex($f[0]); | |
310 | $data{$c}->{casefold} = $f[2]; | |
311 | # We are particularly interest in combining characters that | |
312 | # case-fold to non-combining characters, or characters that | |
313 | # case-fold to sequences with combining characters in non-initial | |
314 | # positions, as these required decomposiiton before case-folding | |
315 | my @d = map(hex($_), split(/\s+/, $data{$c}->{casefold})); | |
316 | if($data{$c}->{ccc} != 0) { | |
317 | # This is a combining character | |
318 | if($data{$d[0]}->{ccc} == 0) { | |
319 | # The first character of its case-folded form is NOT | |
320 | # a combining character. The field name is the example | |
321 | # explicitly mentioned in the spec. | |
322 | $data{$c}->{ypogegrammeni} = 1; | |
323 | } | |
324 | } else { | |
325 | # This is a non-combining character; inspect the non-initial | |
326 | # code points of the case-folded sequence | |
327 | shift(@d); | |
328 | if(grep($data{$_}->{ccc} != 0, @d)) { | |
329 | # Some non-initial code point in the case-folded for is NOT a | |
330 | # a combining character. | |
331 | $data{$c}->{ypogegrammeni} = 1; | |
332 | } | |
333 | } | |
334 | } | |
335 | } | |
336 | ||
337 | # Generate the header file | |
bcf9ed7f | 338 | print STDERR "Generating unidata.h...\n"; |
61507e3c RK |
339 | open(STDOUT, ">unidata.h") or die "unidata.h: $!\n"; |
340 | ||
e5a5a138 RK |
341 | out("/* Automatically generated file, see scripts/make-unidata */\n", |
342 | "#ifndef UNIDATA_H\n", | |
61507e3c RK |
343 | "#define UNIDATA_H\n"); |
344 | ||
e5a5a138 | 345 | # TODO choose stable values for General_Category |
14523635 | 346 | out("enum unicode_General_Category {\n", |
61507e3c | 347 | join(",\n", |
14523635 | 348 | map(" unicode_General_Category_$_", sort keys %cats)), "\n};\n"); |
e5a5a138 | 349 | |
349b7b74 RK |
350 | out("enum unicode_Grapheme_Break {\n", |
351 | join(",\n", | |
352 | map(" unicode_Grapheme_Break_$_", sort keys %gbreak)), | |
353 | "\n};\n"); | |
354 | out("extern const char *const unicode_Grapheme_Break_names[];\n"); | |
355 | ||
0b7052da RK |
356 | out("enum unicode_Word_Break {\n", |
357 | join(",\n", | |
349b7b74 | 358 | map(" unicode_Word_Break_$_", sort keys %wbreak)), |
0b7052da | 359 | "\n};\n"); |
bb48024f | 360 | out("extern const char *const unicode_Word_Break_names[];\n"); |
0b7052da | 361 | |
349b7b74 RK |
362 | out("enum unicode_Sentence_Break {\n", |
363 | join(",\n", | |
364 | map(" unicode_Sentence_Break_$_", sort keys %sbreak)), | |
365 | "\n};\n"); | |
366 | out("extern const char *const unicode_Sentence_Break_names[];\n"); | |
367 | ||
e5a5a138 | 368 | out("enum unicode_flags {\n", |
f98fcddb RK |
369 | " unicode_normalize_before_casefold = 1,\n", |
370 | " unicode_compatibility_decomposition = 2\n", | |
e5a5a138 RK |
371 | "};\n", |
372 | "\n"); | |
373 | ||
374 | # Choose the narrowest type that will fit the required values | |
375 | sub choosetype { | |
376 | my ($min, $max) = @_; | |
377 | if($min >= 0) { | |
378 | return "char" if $max <= 127; | |
379 | return "unsigned char" if $max <= 255; | |
380 | return "int16_t" if $max < 32767; | |
381 | return "uint16_t" if $max < 65535; | |
382 | return "int32_t"; | |
383 | } else { | |
384 | return "char" if $min >= -127 && $max <= 127; | |
385 | return "int16_t" if $min >= -32767 && $max <= 32767; | |
386 | return "int32_t"; | |
387 | } | |
388 | } | |
389 | ||
61507e3c | 390 | out("struct unidata {\n", |
99695df9 RK |
391 | # decomposition (canonical or compatibility; |
392 | # unicode_compatibility_decomposition distinguishes) or NULL | |
f98fcddb | 393 | " const uint32_t *decomp;\n", |
99695df9 RK |
394 | |
395 | # case-folded string or NULL | |
e5a5a138 | 396 | " const uint32_t *casefold;\n", |
99695df9 RK |
397 | |
398 | # composed characters that start with this code point. This only | |
399 | # includes primary composites, i.e. the decomposition mapping is | |
400 | # canonical and this code point is not in the exclusion table. | |
401 | " const uint32_t *composed;\n", | |
402 | ||
1a05e381 RK |
403 | # " ".choosetype($minud, $maxud)." upper_offset;\n", |
404 | # " ".choosetype($minld, $maxld)." lower_offset;\n", | |
99695df9 RK |
405 | |
406 | # canonical combining class | |
e5a5a138 | 407 | " ".choosetype(0, $maxccc)." ccc;\n", |
14523635 | 408 | " char general_category;\n", |
99695df9 RK |
409 | |
410 | # see unicode_flags enum | |
e5a5a138 | 411 | " uint8_t flags;\n", |
349b7b74 | 412 | " char grapheme_break;\n", |
0b7052da | 413 | " char word_break;\n", |
349b7b74 | 414 | " char sentence_break;\n", |
61507e3c | 415 | "};\n"); |
f98fcddb | 416 | # decomp and casefold do have have non-BMP characters, so we |
e5a5a138 RK |
417 | # can't use a simple 16-bit table. We could use UTF-8 or UTF-16 |
418 | # though, saving a bit of space (probably not that much...) at the | |
419 | # cost of marginally reduced performance and additional complexity | |
61507e3c RK |
420 | |
421 | out("extern const struct unidata *const unidata[];\n"); | |
422 | ||
18cda350 RK |
423 | out("extern const struct unicode_utf8_row {\n", |
424 | " uint8_t count;\n", | |
425 | " uint8_t min2, max2;\n", | |
426 | "} unicode_utf8_valid[];\n"); | |
427 | ||
61507e3c | 428 | out("#define UNICODE_NCHARS ", ($max + 1), "\n"); |
e5a5a138 | 429 | out("#define UNICODE_MODULUS $modulus\n"); |
1a05e381 RK |
430 | out("#define UNICODE_BREAK_START $break_start\n"); |
431 | out("#define UNICODE_BREAK_END $break_end\n"); | |
432 | out("#define UNICODE_BREAK_TOP $break_top\n"); | |
61507e3c RK |
433 | |
434 | out("#endif\n"); | |
435 | ||
436 | close STDOUT or die "unidata.h: $!\n"; | |
437 | ||
bcf9ed7f | 438 | print STDERR "Generating unidata.c...\n"; |
61507e3c RK |
439 | open(STDOUT, ">unidata.c") or die "unidata.c: $!\n"; |
440 | ||
e5a5a138 | 441 | out("/* Automatically generated file, see scripts/make-unidata */\n", |
05b75f8d | 442 | "#include \"common.h\"\n", |
e5a5a138 RK |
443 | "#include \"unidata.h\"\n"); |
444 | ||
349b7b74 | 445 | # Short aliases to keep .c file small |
e5a5a138 | 446 | |
14523635 RK |
447 | out(map(sprintf("#define %s unicode_General_Category_%s\n", $_, $_), |
448 | sort keys %cats)); | |
449 | out(map(sprintf("#define GB%s unicode_Grapheme_Break_%s\n", $_, $_), | |
450 | sort keys %gbreak)); | |
451 | out(map(sprintf("#define WB%s unicode_Word_Break_%s\n", $_, $_), | |
452 | sort keys %wbreak)); | |
453 | out(map(sprintf("#define SB%s unicode_Sentence_Break_%s\n", $_, $_), | |
454 | sort keys %sbreak)); | |
99695df9 RK |
455 | out("#define NBC unicode_normalize_before_casefold\n"); |
456 | out("#define CD unicode_compatibility_decomposition\n"); | |
e5a5a138 | 457 | |
349b7b74 RK |
458 | # Names for *_Break properties |
459 | out("const char *const unicode_Grapheme_Break_names[] = {\n", | |
460 | join(",\n", | |
461 | map(" \"$_\"", sort keys %gbreak)), | |
462 | "\n};\n"); | |
bb48024f RK |
463 | out("const char *const unicode_Word_Break_names[] = {\n", |
464 | join(",\n", | |
349b7b74 RK |
465 | map(" \"$_\"", sort keys %wbreak)), |
466 | "\n};\n"); | |
467 | out("const char *const unicode_Sentence_Break_names[] = {\n", | |
468 | join(",\n", | |
469 | map(" \"$_\"", sort keys %sbreak)), | |
bb48024f RK |
470 | "\n};\n"); |
471 | ||
99695df9 RK |
472 | our $ddnum = 0; |
473 | our $ddsaved = 0; | |
474 | our %ddnums = (); | |
475 | my $ddfirst = 1; | |
476 | out("static const uint32_t "); | |
477 | sub dedupe { | |
478 | my $s = join(",", @_); | |
479 | if(!exists $ddnums{$s}) { | |
480 | if($ddfirst) { | |
481 | $ddfirst = 0; | |
482 | } else { | |
483 | out(",\n"); | |
484 | } | |
485 | out("dd$ddnum\[]={$s}"); | |
486 | $ddnums{$s} = $ddnum++; | |
487 | } else { | |
488 | ++$ddsaved; | |
489 | } | |
490 | return "dd$ddnums{$s}"; | |
491 | } | |
492 | ||
16506c9d | 493 | # Generate the decomposition mapping tables. |
99695df9 | 494 | print STDERR "> decomposition mappings\n"; |
e5a5a138 | 495 | for(my $c = 0; $c <= $max; ++$c) { |
f98fcddb | 496 | if(exists $data{$c} && exists $data{$c}->{decomp}) { |
99695df9 RK |
497 | $data{$c}->{decompsym} = dedupe(@{$data{$c}->{decomp}}, 0); |
498 | } | |
499 | } | |
500 | ||
501 | print STDERR "> composition mappings\n"; | |
502 | # First we must generate the mapping of each code point to possible | |
16506c9d | 503 | # compositions. |
99695df9 RK |
504 | for(my $c = 0; $c <= $max; ++$c) { |
505 | if(exists $data{$c} | |
506 | && exists $data{$c}->{decomp} | |
507 | && !exists $data{$c}->{compat} | |
508 | && !$data{$c}->{Full_Composition_Exclusion}) { | |
509 | # $c has a non-excluded canonical decomposition, i.e. it is | |
510 | # a primary composite. Find the first code point of the decomposition | |
511 | my $first = ${$data{$c}->{decomp}}[0]; | |
512 | if(!exists $data{$first}->{compose}) { | |
16506c9d | 513 | $data{$first}->{compose} = [$c]; |
e5a5a138 | 514 | } else { |
16506c9d | 515 | push(@{$data{$first}->{compose}}, $c); |
e5a5a138 | 516 | } |
e5a5a138 RK |
517 | } |
518 | } | |
16506c9d | 519 | # Then we can generate the tables. |
99695df9 RK |
520 | for(my $c = 0; $c <= $max; ++$c) { |
521 | if(exists $data{$c} && exists $data{$c}->{compose}) { | |
522 | $data{$c}->{compsym} = dedupe(@{$data{$c}->{compose}}, 0); | |
523 | } | |
524 | } | |
e5a5a138 | 525 | |
16506c9d | 526 | # The case folding table. |
99695df9 | 527 | print STDERR "> case-fold mappings\n"; |
e5a5a138 | 528 | for(my $c = 0; $c <= $max; ++$c) { |
bcf9ed7f | 529 | if(exists $data{$c} && exists $data{$c}->{casefold}) { |
99695df9 RK |
530 | $data{$c}->{cfsym} = dedupe(map(hex($_), split(/\s+/, |
531 | $data{$c}->{casefold})), | |
532 | 0); | |
e5a5a138 RK |
533 | } |
534 | } | |
99695df9 RK |
535 | |
536 | # End of de-dupable arrays | |
e5a5a138 RK |
537 | out(";\n"); |
538 | ||
539 | # Visit all the $modulus-character blocks in turn and generate the | |
540 | # required subtables. As above we spot duplicates to save space. In | |
541 | # Unicode 5.0.0 with $modulus=128 and current table data this saves | |
542 | # 1372 subtables or at least three and a half megabytes on 32-bit | |
543 | # platforms. | |
99695df9 | 544 | print STDERR "> subtables\n"; |
61507e3c RK |
545 | my %subtable = (); # base->subtable number |
546 | my %subtableno = (); # subtable number -> content | |
547 | my $subtablecounter = 0; # counter for subtable numbers | |
e5a5a138 RK |
548 | my $subtablessaved = 0; # number of tables saved |
549 | for(my $base = 0; $base <= $max; $base += $modulus) { | |
1a05e381 RK |
550 | next if $base >= $break_start && $base < $break_end; |
551 | next if $base >= $break_top; | |
61507e3c | 552 | my @t; |
e5a5a138 | 553 | for(my $c = $base; $c < $base + $modulus; ++$c) { |
61507e3c | 554 | my $d = $data{$c}; |
f98fcddb | 555 | my $decompsym = ($data{$c}->{decompsym} or "0"); |
e5a5a138 | 556 | my $cfsym = ($data{$c}->{cfsym} or "0"); |
99695df9 | 557 | my $compsym = ($data{$c}->{compsym} or "0"); |
35b651f0 RK |
558 | my @flags = (); |
559 | if($data{$c}->{ypogegrammeni}) { | |
99695df9 | 560 | push(@flags, "NBC"); |
35b651f0 | 561 | } |
f98fcddb | 562 | if($data{$c}->{compat}) { |
99695df9 | 563 | push(@flags, "CD"); |
f98fcddb | 564 | } |
35b651f0 | 565 | my $flags = @flags ? join("|", @flags) : 0; |
e5a5a138 RK |
566 | push(@t, "{". |
567 | join(",", | |
f98fcddb | 568 | $decompsym, |
e5a5a138 | 569 | $cfsym, |
99695df9 | 570 | $compsym, |
1a05e381 RK |
571 | # $d->{ud}, |
572 | # $d->{ld}, | |
0b7052da RK |
573 | $d->{ccc}, |
574 | $d->{gc}, | |
e5a5a138 | 575 | $flags, |
349b7b74 RK |
576 | "GB$d->{gbreak}", |
577 | "WB$d->{wbreak}", | |
578 | "SB$d->{sbreak}", | |
e5a5a138 | 579 | )."}"); |
61507e3c RK |
580 | } |
581 | my $t = join(",\n", @t); | |
582 | if(!exists $subtable{$t}) { | |
0e843521 | 583 | out(sprintf("/* %04X-%04X */\n", $base, $base + $modulus - 1)); |
e5a5a138 | 584 | out("static const struct unidata st$subtablecounter\[] = {\n", |
61507e3c RK |
585 | "$t\n", |
586 | "};\n"); | |
587 | $subtable{$t} = $subtablecounter++; | |
e5a5a138 RK |
588 | } else { |
589 | ++$subtablessaved; | |
61507e3c RK |
590 | } |
591 | $subtableno{$base} = $subtable{$t}; | |
592 | } | |
593 | ||
99695df9 | 594 | print STDERR "> main table\n"; |
bcf9ed7f | 595 | out("const struct unidata *const unidata[]={\n"); |
e5a5a138 | 596 | for(my $base = 0; $base <= $max; $base += $modulus) { |
1a05e381 RK |
597 | next if $base >= $break_start && $base < $break_end; |
598 | next if $base >= $break_top; | |
0a4af692 | 599 | #out("st$subtableno{$base} /* ".sprintf("%04x", $base)." */,\n"); |
e5a5a138 | 600 | out("st$subtableno{$base},\n"); |
61507e3c RK |
601 | } |
602 | out("};\n"); | |
603 | ||
99695df9 | 604 | print STDERR "> UTF-8 table\n"; |
18cda350 RK |
605 | out("const struct unicode_utf8_row unicode_utf8_valid[] = {\n"); |
606 | for(my $c = 0; $c <= 0x7F; ++$c) { | |
607 | out(" { 1, 0, 0 }, /* $c */\n"); | |
608 | } | |
609 | for(my $c = 0x80; $c < 0xC2; ++$c) { | |
610 | out(" { 0, 0, 0 }, /* $c */\n"); | |
611 | } | |
612 | for(my $c = 0xC2; $c <= 0xDF; ++$c) { | |
613 | out(" { 2, 0x80, 0xBF }, /* $c */\n"); | |
614 | } | |
615 | for(my $c = 0xE0; $c <= 0xE0; ++$c) { | |
616 | out(" { 3, 0xA0, 0xBF }, /* $c */\n"); | |
617 | } | |
618 | for(my $c = 0xE1; $c <= 0xEC; ++$c) { | |
619 | out(" { 3, 0x80, 0xBF }, /* $c */\n"); | |
620 | } | |
621 | for(my $c = 0xED; $c <= 0xED; ++$c) { | |
622 | out(" { 3, 0x80, 0x9F }, /* $c */\n"); | |
623 | } | |
624 | for(my $c = 0xEE; $c <= 0xEF; ++$c) { | |
625 | out(" { 3, 0x80, 0xBF }, /* $c */\n"); | |
626 | } | |
627 | for(my $c = 0xF0; $c <= 0xF0; ++$c) { | |
628 | out(" { 4, 0x90, 0xBF }, /* $c */\n"); | |
629 | } | |
630 | for(my $c = 0xF1; $c <= 0xF3; ++$c) { | |
631 | out(" { 4, 0x80, 0xBF }, /* $c */\n"); | |
632 | } | |
633 | for(my $c = 0xF4; $c <= 0xF4; ++$c) { | |
634 | out(" { 4, 0x80, 0x8F }, /* $c */\n"); | |
635 | } | |
636 | for(my $c = 0xF5; $c <= 0xFF; ++$c) { | |
637 | out(" { 0, 0, 0 }, /* $c */\n"); | |
638 | } | |
639 | out("};\n"); | |
640 | ||
61507e3c RK |
641 | close STDOUT or die "unidata.c: $!\n"; |
642 | ||
99695df9 | 643 | print STDERR "Done.\n\n"; |
c2e01e0a | 644 | printf STDERR "modulus=%d\n", $modulus; |
bcf9ed7f RK |
645 | printf STDERR "max=%04X\n", $max; |
646 | print STDERR "subtables=$subtablecounter, subtablessaved=$subtablessaved\n"; | |
99695df9 | 647 | print STDERR "ddsaved=$ddsaved\n"; |
16506c9d RK |
648 | print STDERR "maxcompat=$maxcompat maxcanon=$maxcanon\n"; |
649 | print STDERR "$hangul_syllable_decomps canonical decompositions to Hangul syllables\n"; | |
650 | print STDERR "$hangul_choseong_decomps canonical decompositions to Hangul Choseong\n"; | |
651 | ||
652 | die "We assumed that canonical decompositions were never more than 2 long!\n" | |
653 | if $maxcanon > 2; | |
654 | ||
655 | die "We assumed no canonical decompositions to Hangul syllables/Choseong!\n" | |
656 | if $hangul_syllable_decomps || $hangul_choseong_decomps; |