icuSources/data/brkitr/rules/line_normal_cj.txt

   1 # Copyright (c) 2002-2016  International Business Machines Corporation and
   2 # others. All Rights Reserved.
   3 #
   4 #  file:  line_normal_cj.txt
   5 #
   6 #         Line Breaking Rules
   7 #         Implement default line breaking as defined by
   8 #         Unicode Standard Annex #14 Revision 35 for Unicode 8.0
   9 #         http://www.unicode.org/reports/tr14/
  10 #
  11 #         Includes the Emoji breaking proposals from Unicode L2/16-011R3.
  12 #         http://www.unicode.org/L2/L2016/16011r3-break-prop-emoji.pdf
  13 #
  14 #         tailored as noted in 2nd paragraph below.
  15 #
  16 #         TODO:  Rule LB 8 remains as it was in Unicode 5.2
  17 #         This is only because of a limitation of ICU break engine implementation,
  18 #         not because the older behavior is desirable.
  19 #
  20 #         This tailors the line break behavior to correspond to CSS
  21 #         line-break=normal (BCP47 -u-lb-normal) as defined for Chinese & Japanese.
  22 #         It sets characters of class CJ to behave like ID.
  23 #         In addition, it allows breaks:
  24 #         * before hyphens 2010 & 2013 (both BA) and 301C, 30A0 (both NS)
  25
  26 #
  27 #  Character Classes defined by TR 14.
  28 #
  29
  30 !!chain;
  31
  32 !!lookAheadHardBreak;
  33 #
  34 #  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
  35 #                          and only used for the line break rules.
  36 #
  37 #           It is used in the implementation of rule LB 10
  38 #           which says to treat any combining mark that is not attached to a base
  39 #           character as if it were of class AL  (alphabetic).
  40 #
  41 #           The problem occurs in the reverse rules.
  42 #
  43 #           Consider a sequence like, with correct breaks as shown
  44 #               LF  ID  CM  AL  AL
  45 #                  ^       ^       ^
  46 #           Then consider the sequence without the initial ID (ideographic)
  47 #                 LF  CM  AL  AL
  48 #                    ^           ^
  49 #           Our CM, which in the first example was attached to the ideograph,
  50 #           is now unattached, becomes an alpha, and joins in with the other
  51 #           alphas.
  52 #
  53 #           When iterating forwards, these sequences do not present any problems
  54 #           When iterating backwards, we need to look ahead when encountering
  55 #           a CM to see whether it attaches to something further on or not.
  56 #           (Look-ahead in a reverse rule is looking towards the start)
  57 #
  58 #           If the CM is unattached, we need to force a break.
  59 #
  60 #           !!lookAheadHardBreak forces the run time state machine to
  61 #           stop immediately when a look ahead rule ( '/' operator) matches,
  62 #           and set the match position to that of the look-ahead operator,
  63 #           no matter what other rules may be in play at the time.
  64 #
  65 #           See rule LB 19 for an example.
  66 #
  67
  68 # Temporary definitions of Emoji Base and Emoji Modifiers, until properties are available.
  69
  70 $EB = [\u261D\u26F9\u270A-\u270D\U0001F385\U0001F3C3-\U0001F3C4\U0001F3CA-\U0001F3CB\U0001F442-\U0001F443\U0001F446-\U0001F450\U0001F466-\U0001F469\U0001F46E\U0001F470-\U0001F478\U0001F47C\U0001F481-\U0001F483\U0001F485-\U0001F487\U0001F4AA\U0001F575\U0001F590\U0001F595-\U0001F596\U0001F645-\U0001F647\U0001F64B-\U0001F64F\U0001F6A3\U0001F6B4-\U0001F6B6\U0001F6C0\U0001F918];
  71 $EM = [\U0001F3FB-\U0001F3FF];
  72
  73 $AI = [[:LineBreak =  Ambiguous:] - [\u2640\u2642]];
  74 $AL = [[:LineBreak =  Alphabetic:] - [$EM\u2764]];
  75 $BAX = [\u2010 \u2013];
  76 $BA = [[:LineBreak =  Break_After:] - $BAX];
  77 $BB = [:LineBreak =  Break_Before:];
  78 $BK = [:LineBreak =  Mandatory_Break:];
  79 $B2 = [:LineBreak =  Break_Both:];
  80 $CB = [:LineBreak =  Contingent_Break:];
  81 $CJ = [:LineBreak =  Conditional_Japanese_Starter:];
  82 $CL = [:LineBreak =  Close_Punctuation:];
  83 $CM = [[:LineBreak =  Combining_Mark:] \u200d];
  84 $CP = [:LineBreak =  Close_Parenthesis:];
  85 $CR = [:LineBreak =  Carriage_Return:];
  86 $EX = [:LineBreak =  Exclamation:];
  87 $GL = [:LineBreak =  Glue:];
  88 $HL = [:LineBreak =  Hebrew_Letter:];
  89 $HY = [:LineBreak =  Hyphen:];
  90 $H2 = [:LineBreak =  H2:];
  91 $H3 = [:LineBreak =  H3:];
  92 $ID = [[:LineBreak =  Ideographic:] $CJ [\u2640\u2642\u2764] - $EB];
  93 $IN = [:LineBreak =  Inseperable:];
  94 $IS = [:LineBreak =  Infix_Numeric:];
  95 $JL = [:LineBreak =  JL:];
  96 $JV = [:LineBreak =  JV:];
  97 $JT = [:LineBreak =  JT:];
  98 $LF = [:LineBreak =  Line_Feed:];
  99 $NL = [:LineBreak =  Next_Line:];
 100 $NSX = [\u301C \u30A0];
 101 $NS = [[:LineBreak =  Nonstarter:] - $NSX];
 102 $NU = [:LineBreak =  Numeric:];
 103 $OP = [:LineBreak =  Open_Punctuation:];
 104 $PO = [:LineBreak =  Postfix_Numeric:];
 105 $PR = [:LineBreak =  Prefix_Numeric:];
 106 $QU = [:LineBreak =  Quotation:];
 107 $RI = [:LineBreak =  Regional_Indicator:];
 108 $SA = [:LineBreak =  Complex_Context:];
 109 $SG = [:LineBreak =  Surrogate:];
 110 $SP = [:LineBreak =  Space:];
 111 $SY = [:LineBreak =  Break_Symbols:];
 112 $WJ = [:LineBreak =  Word_Joiner:];
 113 $XX = [:LineBreak =  Unknown:];
 114 $ZW = [:LineBreak =  ZWSpace:];
 115 $ZWJ = [\u200d];
 116
 117 #   Dictionary character set, for triggering language-based break engines. Currently
 118 #   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
 119 #   5.0 or later as the definition of Complex_Context was corrected to include all
 120 #   characters requiring dictionary break.
 121
 122 $dictionary = [:LineBreak = Complex_Context:];
 123
 124 #
 125 #  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
 126 #                               SA  (South East Asian: Thai, Lao, Khmer)
 127 #                               SG  (Unpaired Surrogates)
 128 #                               XX  (Unknown, unassigned)
 129 #                         as $AL  (Alphabetic)
 130 #
 131 $ALPlus = [$AL $AI $SA $SG $XX];
 132
 133 #
 134 #  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
 135 #
 136 $ALcm = $ALPlus $CM*;
 137 $BAcm = $BA $CM*;
 138 $BAXcm = $BAX $CM*;
 139 $BBcm = $BB $CM*;
 140 $B2cm = $B2 $CM*;
 141 $CLcm = $CL $CM*;
 142 $CPcm = $CP $CM*;
 143 $EXcm = $EX $CM*;
 144 $GLcm = $GL $CM*;
 145 $HLcm = $HL $CM*;
 146 $HYcm = $HY $CM*;
 147 $H2cm = $H2 $CM*;
 148 $H3cm = $H3 $CM*;
 149 $INcm = $IN $CM*;
 150 $IScm = $IS $CM*;
 151 $JLcm = $JL $CM*;
 152 $JVcm = $JV $CM*;
 153 $JTcm = $JT $CM*;
 154 $NScm = $NS $CM*;
 155 $NSXcm = $NSX $CM*;
 156 $NUcm = $NU $CM*;
 157 $OPcm = $OP $CM*;
 158 $POcm = $PO $CM*;
 159 $PRcm = $PR $CM*;
 160 $QUcm = $QU $CM*;
 161 $RIcm = $RI $CM*;
 162 $SYcm = $SY $CM*;
 163 $WJcm = $WJ $CM*;
 164
 165 ## -------------------------------------------------
 166
 167 !!forward;
 168
 169 #
 170 #  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
 171 #
 172 $ALPlus $CM+;
 173 $BA $CM+;
 174 $BAX $CM+;
 175 $BB $CM+;
 176 $B2 $CM+;
 177 $CL $CM+;
 178 $CP $CM+;
 179 $EB $CM+;
 180 $EM $CM+;
 181 $EX $CM+;
 182 $GL $CM+;
 183 $HL $CM+;
 184 $HY $CM+;
 185 $H2 $CM+;
 186 $H3 $CM+;
 187 $ID $CM+;
 188 $IN $CM+;
 189 $IS $CM+;
 190 $JL $CM+;
 191 $JV $CM+;
 192 $JT $CM+;
 193 $NS $CM+;
 194 $NSX $CM+;
 195 $NU $CM+;
 196 $OP $CM+;
 197 $PO $CM+;
 198 $PR $CM+;
 199 $QU $CM+;
 200 $RI $CM+;
 201 $SY $CM+;
 202 $WJ $CM+;
 203
 204 #
 205 # CAN_CM  is the set of characters that may combine with CM combining chars.
 206 #         Note that Linebreak UAX 14's concept of a combining char and the rules
 207 #         for what they can combine with are _very_ different from the rest of Unicode.
 208 #
 209 #         Note that $CM itself is left out of this set.  If CM is needed as a base
 210 #         it must be listed separately in the rule.
 211 #
 212 $CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
 213 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
 214
 215 #
 216 # AL_FOLLOW  set of chars that can unconditionally follow an AL
 217 #            Needed in rules where stand-alone $CM s are treated as AL.
 218 #            Chaining is disabled with CM because it causes other failures,
 219 #            so for this one case we need to manually list out longer sequences.
 220 #
 221 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
 222 $AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
 223 $AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
 224
 225
 226 #
 227 #  Rule LB 4, 5    Mandatory (Hard) breaks.
 228 #
 229 $LB4Breaks    = [$BK $CR $LF $NL];
 230 $LB4NonBreaks = [^$BK $CR $LF $NL $CM];
 231 $CR $LF {100};
 232
 233 #
 234 #  LB 6    Do not break before hard line breaks.
 235 #
 236 $LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
 237 $CAN_CM $CM*    $LB4Breaks {100};
 238 ^$CM+           $LB4Breaks {100};
 239
 240 # LB 7         x SP
 241 #              x ZW
 242 $LB4NonBreaks [$SP $ZW];
 243 $CAN_CM $CM*  [$SP $ZW];
 244 ^$CM+         [$SP $ZW];
 245
 246 #
 247 # LB 8         Break after zero width space
 248 #              TODO:  ZW SP* <break>
 249 #              An engine change is required to write the reverse rule for this.
 250 #              For now, leave the Unicode 5.2 rule, ZW <break>
 251 #
 252 $LB8Breaks    = [$LB4Breaks $ZW];
 253 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
 254
 255 # LB 8a        ZWJ x ID    Emoji proposal.
 256 #
 257 $ZWJ ($ID | $EB | $EM);
 258
 259 # LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
 260 #                                $CM not covered by the above needs to behave like $AL
 261 #                                See definition of $CAN_CM.
 262
 263 $CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
 264 ^$CM+;
 265
 266 #
 267 # LB 11  Do not break before or after WORD JOINER & related characters.
 268 #
 269 $CAN_CM $CM*  $WJcm;
 270 $LB8NonBreaks $WJcm;
 271 ^$CM+         $WJcm;
 272
 273 $WJcm $CANT_CM;
 274 $WJcm $CAN_CM $CM*;
 275
 276 #
 277 # LB 12  Do not break after NBSP and related characters.
 278 #         GL  x
 279 #
 280 $GLcm $CAN_CM $CM*;
 281 $GLcm $CANT_CM;
 282
 283 #
 284 # LB 12a  Do not break before NBSP and related characters ...
 285 #            [^SP BA HY] x GL
 286 #
 287 [[$LB8NonBreaks] - [$SP $BA $BAX $HY]] $CM* $GLcm;
 288 ^$CM+ $GLcm;
 289
 290
 291
 292 #
 293 # LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
 294 #
 295 $LB8NonBreaks $CL;
 296 $CAN_CM $CM*  $CL;
 297 ^$CM+         $CL;              # by rule 10, stand-alone CM behaves as AL
 298
 299 $LB8NonBreaks $CP;
 300 $CAN_CM $CM*  $CP;
 301 ^$CM+         $CP;              # by rule 10, stand-alone CM behaves as AL
 302
 303 $LB8NonBreaks $EX;
 304 $CAN_CM $CM*  $EX;
 305 ^$CM+         $EX;              # by rule 10, stand-alone CM behaves as AL
 306
 307 $LB8NonBreaks $IS;
 308 $CAN_CM $CM*  $IS;
 309 ^$CM+         $IS;              # by rule 10, stand-alone CM behaves as AL
 310
 311 $LB8NonBreaks $SY;
 312 $CAN_CM $CM*  $SY;
 313 ^$CM+         $SY;              # by rule 10, stand-alone CM behaves as AL
 314
 315
 316 #
 317 # LB 14  Do not break after OP, even after spaces
 318 #
 319 $OPcm $SP* $CAN_CM $CM*;
 320 $OPcm $SP* $CANT_CM;
 321
 322 $OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
 323
 324 # LB 15
 325 $QUcm $SP* $OPcm;
 326
 327 # LB 16
 328 # Do not break between closing punctuation and $NS, even with intervening spaces
 329 # But DO allow a break between closing punctuation and $NSX, don't include it here
 330 ($CLcm | $CPcm) $SP* $NScm;
 331
 332 # LB 17
 333 $B2cm $SP* $B2cm;
 334
 335 #
 336 # LB 18  Break after spaces.
 337 #
 338 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
 339 $LB18Breaks    = [$LB8Breaks $SP];
 340
 341
 342 # LB 19
 343 #         x QU
 344 $LB18NonBreaks $CM* $QUcm;
 345 ^$CM+               $QUcm;
 346
 347 #         QU  x
 348 $QUcm .?;
 349
 350
 351 # LB 20
 352 #        <break>  $CB
 353 #        $CB   <break>
 354
 355 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 356
 357 # LB 21        x   (BA | HY | NS)
 358 #           BB x
 359 #
 360 # DO allow breaks here before $BAXcm and $NSXcm, so don't include them
 361 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
 362 ^$CM+ ($BAcm | $HYcm | $NScm);
 363
 364 $BBcm [^$CB];                                  #  $BB  x
 365 $BBcm $LB20NonBreaks $CM*;
 366
 367 # LB 21a Don't break after Hebrew + Hyphen
 368 #   HL (HY | BA) x
 369 #
 370 $HLcm ($HYcm | $BAcm | $BAXcm) [^$CB]?;
 371
 372 # LB 21b (forward) Don't break between SY and HL
 373 # (break between HL and SY already disallowed by LB 13 above)
 374 $SYcm $HLcm;
 375
 376 # LB 22
 377 ($ALcm | $HLcm) $INcm;
 378 ^$CM+    $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 379 $EXcm    $INcm;
 380 ($ID | $EB | $EM) $CM*  $INcm;
 381 $INcm    $INcm;
 382 $NUcm    $INcm;
 383
 384
 385 # $LB 23
 386 ($ID | $EB | $EM) $CM*  $POcm;
 387 $ALcm  $NUcm;       # includes $LB19
 388 $HLcm  $NUcm;
 389 ^$CM+  $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
 390 $NUcm  $ALcm;
 391 $NUcm  $HLcm;
 392
 393 #
 394 # LB 24
 395 #
 396 $PRcm ($ID | $EB | $EM);
 397 $PRcm ($ALcm | $HLcm);
 398 $POcm ($ALcm | $HLcm);
 399 ($ALcm | $HLcm) ($PRcm | $POcm);
 400 ^$CM+ ($PRcm | $POcm);       # Rule 10, any otherwise unattached CM behaves as AL
 401
 402 #
 403 # LB 25   Numbers.
 404 #
 405 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
 406
 407 # LB 26  Do not break a Korean syllable
 408 #
 409 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 410 ($JVcm | $H2cm) ($JVcm | $JTcm);
 411 ($JTcm | $H3cm) $JTcm;
 412
 413 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 414 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 415 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 416 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 417
 418
 419 # LB 28   Do not break between alphabetics
 420 #
 421 ($ALcm | $HLcm) ($ALcm | $HLcm);
 422 ^$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
 423
 424 # LB 29
 425 $IScm ($ALcm | $HLcm);
 426
 427 # LB 30
 428 ($ALcm | $HLcm | $NUcm) $OPcm;
 429 ^$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
 430 $CPcm ($ALcm | $HLcm | $NUcm);
 431
 432 # LB 30a  Do not break between regional indicators. Break after pairs of them.
 433 #         Tricky interaction with LB8a: ZWJ x ID
 434 $RIcm $RI                / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
 435 $RIcm $RI $CM*  $ZWJ     / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM $ID $EB $EM] {eof}];
 436 $RIcm $RI $CM* [$CM-$ZWJ] / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
 437
 438 $RIcm $RIcm [$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS {eof}];
 439 $RIcm $RIcm $ZWJ ($ID | $EB | $EM);
 440
 441 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 442 $EB $CM* $EM;
 443
 444 #
 445 #  Reverse Rules.
 446 #
 447 ## -------------------------------------------------
 448
 449 !!reverse;
 450
 451 ^$CM+ $ALPlus;
 452 ^$CM+ $BA;
 453 ^$CM+ $BAX;
 454 ^$CM+ $BB;
 455 ^$CM+ $B2;
 456 ^$CM+ $CL;
 457 ^$CM+ $CP;
 458 ^$CM+ $EB;
 459 ^$CM+ $EM;
 460 ^$CM+ $EX;
 461 ^$CM+ $GL;
 462 ^$CM+ $HL;
 463 ^$CM+ $HY;
 464 ^$CM+ $H2;
 465 ^$CM+ $H3;
 466 ^$CM+ $ID;
 467 ^$CM+ $IN;
 468 ^$CM+ $IS;
 469 ^$CM+ $JL;
 470 ^$CM+ $JV;
 471 ^$CM+ $JT;
 472 ^$CM+ $NS;
 473 ^$CM+ $NSX;
 474 ^$CM+ $NU;
 475 ^$CM+ $OP;
 476 ^$CM+ $PO;
 477 ^$CM+ $PR;
 478 ^$CM+ $QU;
 479 ^$CM+ $RI;
 480 ^$CM+ $SY;
 481 ^$CM+ $WJ;
 482 ^$CM+;
 483
 484
 485 #
 486 #  Sequences of the form  (shown forwards)
 487 #      [CANT_CM]  <break>  [CM]  [whatever]
 488 #  The CM needs to behave as an AL
 489 #
 490 $AL_FOLLOW $CM+ / (
 491           [$BK $CR $LF $NL $ZW {eof}] |
 492           $SP+ $CM+ $SP |
 493           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 494                                                #  LB14 says    OP SP* x .
 495                                                #    becomes    OP SP* x AL
 496                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 497                                                #
 498                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 499                                                #                a rule compiler bug which complains about
 500                                                #                empty sets otherwise.
 501
 502 #
 503 #  Sequences of the form  (shown forwards)
 504 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 505 #  The CM needs to behave as an AL
 506 #  This rule is concerned about getting the second of the two <breaks> in place.
 507 #
 508
 509 # Apple early addition, remove this, superseded by LB24
 510 # [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 511
 512
 513
 514 # LB 4, 5, 6
 515
 516 $LB4Breaks [$LB4NonBreaks-$CM];
 517 $LB4Breaks $CM+ $CAN_CM;
 518 $LF $CR;
 519
 520
 521 # LB 7         x SP
 522 #              x ZW
 523 [$SP $ZW] [$LB4NonBreaks-$CM];
 524 [$SP $ZW] $CM+ $CAN_CM;
 525
 526 # LB 8 ZW SP* <break>
 527 #     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
 528 #           Requires an engine enhancement.
 529 #   / $SP* $ZW
 530
 531 # LB 8a        ZWJ x ID    Unicode Emoji proposal L2/16-011R3
 532 #                          The ZWJ will look like a CM to whatever precedes it.
 533 #
 534 ($ID | $EB | $EM) $ZWJ $CM* $CAN_CM?;
 535
 536
 537 # LB 9,10  Combining marks.
 538 #    X   $CM needs to behave like X, where X is not $SP or controls.
 539 #    $CM not covered by the above needs to behave like $AL
 540 # Stick together any combining sequences that don't match other rules.
 541 ^$CM+ $CAN_CM;
 542
 543
 544 # LB 11
 545 #
 546 $WJ $CM* $CAN_CM;
 547 $WJ      [$LB8NonBreaks-$CM];
 548
 549      $CANT_CM $CM* $WJ;
 550 $CAN_CM  $CM* $WJ;
 551
 552 # LB 12a
 553 #      [^SP BA HY] x GL
 554 #
 555 $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $BAX $HY]];
 556
 557 # LB 12
 558 #     GL  x
 559 #
 560 $CANT_CM $CM* $GL;
 561 $CAN_CM $CM* $GL;
 562
 563
 564 # LB 13
 565 $CL $CM+ $CAN_CM;
 566 $CP $CM+ $CAN_CM;
 567 $EX $CM+ $CAN_CM;
 568 $IS $CM+ $CAN_CM;
 569 $SY $CM+ $CAN_CM;
 570
 571 $CL [$LB8NonBreaks-$CM];
 572 $CP [$LB8NonBreaks-$CM];
 573 $EX [$LB8NonBreaks-$CM];
 574 $IS [$LB8NonBreaks-$CM];
 575 $SY [$LB8NonBreaks-$CM];
 576
 577 # Rule 13 & 14 taken together for an edge case.
 578 #   Match this, shown forward
 579 #     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
 580 #   This really wants to chain at the $CM+ (which is acting as an $AL)
 581 #   except for $CM chaining being disabled.
 582 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
 583
 584 # LB 14    OP SP* x
 585 #
 586 $CAN_CM    $SP* $CM* $OP;
 587      $CANT_CM   $SP* $CM* $OP;
 588 $AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
 589
 590      $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
 591 $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 592
 593
 594 # LB 15
 595 $OP $SP* $CM* $QU;
 596
 597 # LB 16
 598 # Don't include $NSX here
 599 $NS $SP* $CM* ($CL | $CP);
 600
 601 # LB 17
 602 $B2 $SP* $CM* $B2;
 603
 604 # LB 18  break after spaces
 605 #        Nothing explicit needed here.
 606
 607
 608 #
 609 # LB 19
 610 #
 611 $QU $CM* $CAN_CM;                                #   . x QU
 612 $QU      $LB18NonBreaks;
 613
 614
 615 $CAN_CM  $CM* $QU;                               #   QU x .
 616      $CANT_CM $CM* $QU;
 617
 618 #
 619 #  LB 20  Break before and after CB.
 620 #         nothing needed here.
 621 #
 622
 623 # LB 21
 624 # Don't include $BAX or $NSX here
 625 ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 626
 627 [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 628 [^$CB] $CM* $BB;                                      #
 629
 630 # LB21a  Don't break after Hebrew + Hyphen.
 631 ([^$CB] $CM*)? ($HY | $BA | $BAX) $CM* $HL;
 632
 633 # LB21b (reverse)
 634 $HL $CM* $SY;
 635
 636 # LB 22
 637 $IN $CM* ($ALPlus | $HL);
 638 $IN $CM* $EX;
 639 $IN $CM* ($ID | $EB | $EM);
 640 $IN $CM* $IN;
 641 $IN $CM* $NU;
 642
 643 # LB 23
 644 $PO $CM* ($ID | $EB | $EM);
 645 $NU $CM* ($ALPlus | $HL);
 646 ($ALPlus | $HL) $CM* $NU;
 647
 648 # LB 24
 649 ($ID | $EB | $EM) $CM* $PR;
 650 ($ALPlus | $HL) $CM* $PR;
 651 ($ALPlus | $HL) $CM* $PO;
 652 $CM* ($PR | $PO) $CM* ($ALPlus | $HL);
 653 $CM* ($PR | $PO) $CM+ / [$BK $CR $LF $NL $ZW $SP {eof}];
 654
 655 # LB 25
 656 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 657
 658 # LB 26
 659 ($H3 | $H2 | $JV | $JL) $CM* $JL;
 660 ($JT | $JV) $CM* ($H2 | $JV);
 661 $JT $CM* ($H3 | $JT);
 662
 663 # LB 27
 664 $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 665 $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 666  ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 667
 668 # LB 28
 669 ($ALPlus | $HL) $CM* ($ALPlus | $HL);
 670
 671
 672 # LB 29
 673 ($ALPlus | $HL) $CM* $IS;
 674
 675 # LB 30
 676 $OP $CM* ($ALPlus | $HL | $NU);
 677 ($ALPlus | $HL | $NU) $CM* $CP;
 678
 679 # LB 30a
 680 #    Pairs of Regional Indicators.
 681 #    The following two rules are nearly identical. The first matches only sequences with an odd number of adjacent RIs,
 682 #    the second with an even number. Stripping away the cruft they look like
 683 #         [^RI] RI / (RI RI)+ ^RI;
 684 #         [^RI] RI RI / (RI RI)+ ^RI;
 685 #
 686 [{bof} $NS $HY $BA $QU $CL $CP $EX $IS $SY $WJ $GL $ZW $SP $BK $CR $LF $NL $ZWJ] $CM* $RI / ($CM* $RI $CM* $RI)+ $CM* [{eof}[^$RI $CM]];
 687 [{bof} $NS $HY $BA $QU $CL $CP $EX $IS $SY $WJ $GL $ZW $SP $BK $CR $LF $NL $ZWJ] $CM* $RI $CM* $RI / ($CM* $RI $CM* $RI)+ $CM* [{eof}[^$RI $CM]];
 688
 689 # In general, adjacent RIs stay together. The hard-break rules, above, overide this, forcing in the boundaries between pairs.
 690 $RI $CM* $RI;
 691
 692 #    WJ, GL, QU, etc. are classes with rules like "WJ x "   which includes "WJ x RI".
 693 $RI $CM* ([$WJ $GL $QU $BB] |  (($HY | $BA)$CM* $HL));
 694
 695
 696 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 697 $EM $CM* $EB;
 698
 699
 700 ## -------------------------------------------------
 701
 702 !!safe_reverse;
 703
 704 # LB 9
 705 ^$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 706 ^$CM+ $SP / .;
 707
 708 # LB 14
 709 $SP+ $CM* $OP;
 710
 711 # LB 15
 712 $SP+ $CM* $QU;
 713
 714 # LB 16
 715 $SP+ $CM* ($CL | $CP);
 716
 717 # LB 17
 718 $SP+ $CM* $B2;
 719
 720 # LB 21
 721 $CM* ($HY | $BA | $BAX) $CM* $HL;
 722
 723 # LB 25
 724 ($CM* ($IS | $SY))+ $CM* $NU;
 725 ($CL | $CP) $CM* ($NU | $IS | $SY);
 726
 727 #  LB 30
 728 ($CM* $RI)+;
 729
 730 # For dictionary-based break
 731 $dictionary $dictionary;
 732
 733 ## -------------------------------------------------
 734
 735 !!safe_forward;
 736
 737 # Skip forward over all character classes that are involved in
 738 #   rules containing patterns with possibly more than one char
 739 #   of context.
 740 #
 741 #  It might be slightly more efficient to have specific rules
 742 #  instead of one generic one, but only if we could
 743 #  turn off rule chaining.  We don't want to move more
 744 #  than necessary.
 745 #
 746 ^[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $BAX $SP $RI $ZWJ $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $BAX $RI $ZWJ $dictionary];
 747 $dictionary $dictionary;
 748