icuSources/data/brkitr/rules/line_normal.txt

   1 # Copyright (c) 2002-2016  International Business Machines Corporation and
   2 # others. All Rights Reserved.
   3 #
   4 #  file:  line_normal.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 languages other than
  22 #         Chinese & Japanese.
  23 #         It sets characters of class CJ to behave like ID.
  24
  25 #
  26 #  Character Classes defined by TR 14.
  27 #
  28
  29 !!chain;
  30
  31 !!lookAheadHardBreak;
  32 #
  33 #  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
  34 #                          and only used for the line break rules.
  35 #
  36 #           It is used in the implementation of rule LB 10
  37 #           which says to treat any combining mark that is not attached to a base
  38 #           character as if it were of class AL  (alphabetic).
  39 #
  40 #           The problem occurs in the reverse rules.
  41 #
  42 #           Consider a sequence like, with correct breaks as shown
  43 #               LF  ID  CM  AL  AL
  44 #                  ^       ^       ^
  45 #           Then consider the sequence without the initial ID (ideographic)
  46 #                 LF  CM  AL  AL
  47 #                    ^           ^
  48 #           Our CM, which in the first example was attached to the ideograph,
  49 #           is now unattached, becomes an alpha, and joins in with the other
  50 #           alphas.
  51 #
  52 #           When iterating forwards, these sequences do not present any problems
  53 #           When iterating backwards, we need to look ahead when encountering
  54 #           a CM to see whether it attaches to something further on or not.
  55 #           (Look-ahead in a reverse rule is looking towards the start)
  56 #
  57 #           If the CM is unattached, we need to force a break.
  58 #
  59 #           !!lookAheadHardBreak forces the run time state machine to
  60 #           stop immediately when a look ahead rule ( '/' operator) matches,
  61 #           and set the match position to that of the look-ahead operator,
  62 #           no matter what other rules may be in play at the time.
  63 #
  64 #           See rule LB 19 for an example.
  65 #
  66
  67 # Temporary definitions of Emoji Base and Emoji Modifiers, until properties are available.
  68
  69 $EB = [\u261D\u26F9\u270A-\u270D\U0001F385\U0001F3C2-\U0001F3C4\U0001F3C7\U0001F3CA-\U0001F3CC\U0001F442-\U0001F443\U0001F446-\U0001F450\U0001F466-\U0001F478\U0001F47C\U0001F481-\U0001F483\U0001F485-\U0001F487\U0001F4AA\U0001F574-\U0001F575\U0001F57A\U0001F590\U0001F595-\U0001F596\U0001F645-\U0001F647\U0001F64B-\U0001F64F\U0001F6A3\U0001F6B4-\U0001F6B6\U0001F6C0\U0001F6CC\U0001F918-\U0001F91E\U0001F926\U0001F930\U0001F933-\U0001F939\U0001F93C-\U0001F93E];
  70 $EM = [\U0001F3FB-\U0001F3FF];
  71
  72 $AI = [[:LineBreak =  Ambiguous:] - [\u2640\u2642]];
  73 $AL = [[:LineBreak =  Alphabetic:] - [$EM\u2695\u2696\u2764]];
  74 $BA = [:LineBreak =  Break_After:];
  75 $BB = [:LineBreak =  Break_Before:];
  76 $BK = [:LineBreak =  Mandatory_Break:];
  77 $B2 = [:LineBreak =  Break_Both:];
  78 $CB = [:LineBreak =  Contingent_Break:];
  79 $CJ = [:LineBreak =  Conditional_Japanese_Starter:];
  80 $CL = [:LineBreak =  Close_Punctuation:];
  81 $CM = [[:LineBreak =  Combining_Mark:] \u200d];
  82 $CP = [:LineBreak =  Close_Parenthesis:];
  83 $CR = [:LineBreak =  Carriage_Return:];
  84 $EX = [:LineBreak =  Exclamation:];
  85 $GL = [:LineBreak =  Glue:];
  86 $HL = [:LineBreak =  Hebrew_Letter:];
  87 $HY = [:LineBreak =  Hyphen:];
  88 $H2 = [:LineBreak =  H2:];
  89 $H3 = [:LineBreak =  H3:];
  90 $ID = [[:LineBreak =  Ideographic:] $CJ [\u2640\u2642\u2695\u2696\u2764] - $EB];
  91 $IN = [:LineBreak =  Inseperable:];
  92 $IS = [:LineBreak =  Infix_Numeric:];
  93 $JL = [:LineBreak =  JL:];
  94 $JV = [:LineBreak =  JV:];
  95 $JT = [:LineBreak =  JT:];
  96 $LF = [:LineBreak =  Line_Feed:];
  97 $NL = [:LineBreak =  Next_Line:];
  98 $NS = [:LineBreak =  Nonstarter:];
  99 $NU = [:LineBreak =  Numeric:];
 100 $OP = [:LineBreak =  Open_Punctuation:];
 101 $PO = [:LineBreak =  Postfix_Numeric:];
 102 $PR = [:LineBreak =  Prefix_Numeric:];
 103 $QU = [:LineBreak =  Quotation:];
 104 $RI = [:LineBreak =  Regional_Indicator:];
 105 $SA = [:LineBreak =  Complex_Context:];
 106 $SG = [:LineBreak =  Surrogate:];
 107 $SP = [:LineBreak =  Space:];
 108 $SY = [:LineBreak =  Break_Symbols:];
 109 $WJ = [:LineBreak =  Word_Joiner:];
 110 $XX = [:LineBreak =  Unknown:];
 111 $ZW = [:LineBreak =  ZWSpace:];
 112 $ZWJ = [\u200d];
 113
 114 #   Dictionary character set, for triggering language-based break engines. Currently
 115 #   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
 116 #   5.0 or later as the definition of Complex_Context was corrected to include all
 117 #   characters requiring dictionary break.
 118
 119 $dictionary = [:LineBreak = Complex_Context:];
 120
 121 #
 122 #  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
 123 #                               SA  (South East Asian: Thai, Lao, Khmer)
 124 #                               SG  (Unpaired Surrogates)
 125 #                               XX  (Unknown, unassigned)
 126 #                         as $AL  (Alphabetic)
 127 #
 128 $ALPlus = [$AL $AI $SA $SG $XX];
 129
 130 #
 131 #  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
 132 #
 133 $ALcm = $ALPlus $CM*;
 134 $BAcm = $BA $CM*;
 135 $BBcm = $BB $CM*;
 136 $B2cm = $B2 $CM*;
 137 $CLcm = $CL $CM*;
 138 $CPcm = $CP $CM*;
 139 $EXcm = $EX $CM*;
 140 $GLcm = $GL $CM*;
 141 $HLcm = $HL $CM*;
 142 $HYcm = $HY $CM*;
 143 $H2cm = $H2 $CM*;
 144 $H3cm = $H3 $CM*;
 145 $INcm = $IN $CM*;
 146 $IScm = $IS $CM*;
 147 $JLcm = $JL $CM*;
 148 $JVcm = $JV $CM*;
 149 $JTcm = $JT $CM*;
 150 $NScm = $NS $CM*;
 151 $NUcm = $NU $CM*;
 152 $OPcm = $OP $CM*;
 153 $POcm = $PO $CM*;
 154 $PRcm = $PR $CM*;
 155 $QUcm = $QU $CM*;
 156 $RIcm = $RI $CM*;
 157 $SYcm = $SY $CM*;
 158 $WJcm = $WJ $CM*;
 159
 160 ## -------------------------------------------------
 161
 162 !!forward;
 163
 164 #
 165 #  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
 166 #
 167 $ALPlus $CM+;
 168 $BA $CM+;
 169 $BB $CM+;
 170 $B2 $CM+;
 171 $CL $CM+;
 172 $CP $CM+;
 173 $EB $CM+;
 174 $EM $CM+;
 175 $EX $CM+;
 176 $GL $CM+;
 177 $HL $CM+;
 178 $HY $CM+;
 179 $H2 $CM+;
 180 $H3 $CM+;
 181 $ID $CM+;
 182 $IN $CM+;
 183 $IS $CM+;
 184 $JL $CM+;
 185 $JV $CM+;
 186 $JT $CM+;
 187 $NS $CM+;
 188 $NU $CM+;
 189 $OP $CM+;
 190 $PO $CM+;
 191 $PR $CM+;
 192 $QU $CM+;
 193 $RI $CM+;
 194 $SY $CM+;
 195 $WJ $CM+;
 196
 197 #
 198 # CAN_CM  is the set of characters that may combine with CM combining chars.
 199 #         Note that Linebreak UAX 14's concept of a combining char and the rules
 200 #         for what they can combine with are _very_ different from the rest of Unicode.
 201 #
 202 #         Note that $CM itself is left out of this set.  If CM is needed as a base
 203 #         it must be listed separately in the rule.
 204 #
 205 $CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
 206 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
 207
 208 #
 209 # AL_FOLLOW  set of chars that can unconditionally follow an AL
 210 #            Needed in rules where stand-alone $CM s are treated as AL.
 211 #            Chaining is disabled with CM because it causes other failures,
 212 #            so for this one case we need to manually list out longer sequences.
 213 #
 214 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
 215 $AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
 216 $AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
 217
 218
 219 #
 220 #  Rule LB 4, 5    Mandatory (Hard) breaks.
 221 #
 222 $LB4Breaks    = [$BK $CR $LF $NL];
 223 $LB4NonBreaks = [^$BK $CR $LF $NL $CM];
 224 $CR $LF {100};
 225
 226 #
 227 #  LB 6    Do not break before hard line breaks.
 228 #
 229 $LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
 230 $CAN_CM $CM*    $LB4Breaks {100};
 231 ^$CM+           $LB4Breaks {100};
 232
 233 # LB 7         x SP
 234 #              x ZW
 235 $LB4NonBreaks [$SP $ZW];
 236 $CAN_CM $CM*  [$SP $ZW];
 237 ^$CM+         [$SP $ZW];
 238
 239 #
 240 # LB 8         Break after zero width space
 241 #              TODO:  ZW SP* <break>
 242 #              An engine change is required to write the reverse rule for this.
 243 #              For now, leave the Unicode 5.2 rule, ZW <break>
 244 #
 245 $LB8Breaks    = [$LB4Breaks $ZW];
 246 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
 247
 248 # LB 8a        ZWJ x ID    Emoji proposal.
 249 #
 250 $ZWJ ($ID | $EB | $EM);
 251
 252 # LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
 253 #                                $CM not covered by the above needs to behave like $AL
 254 #                                See definition of $CAN_CM.
 255
 256 $CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
 257 ^$CM+;
 258
 259 #
 260 # LB 11  Do not break before or after WORD JOINER & related characters.
 261 #
 262 $CAN_CM $CM*  $WJcm;
 263 $LB8NonBreaks $WJcm;
 264 ^$CM+         $WJcm;
 265
 266 $WJcm $CANT_CM;
 267 $WJcm $CAN_CM $CM*;
 268
 269 #
 270 # LB 12  Do not break after NBSP and related characters.
 271 #         GL  x
 272 #
 273 $GLcm $CAN_CM $CM*;
 274 $GLcm $CANT_CM;
 275
 276 #
 277 # LB 12a  Do not break before NBSP and related characters ...
 278 #            [^SP BA HY] x GL
 279 #
 280 [[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
 281 ^$CM+ $GLcm;
 282
 283
 284
 285 #
 286 # LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
 287 #
 288 $LB8NonBreaks $CL;
 289 $CAN_CM $CM*  $CL;
 290 ^$CM+         $CL;              # by rule 10, stand-alone CM behaves as AL
 291
 292 $LB8NonBreaks $CP;
 293 $CAN_CM $CM*  $CP;
 294 ^$CM+         $CP;              # by rule 10, stand-alone CM behaves as AL
 295
 296 $LB8NonBreaks $EX;
 297 $CAN_CM $CM*  $EX;
 298 ^$CM+         $EX;              # by rule 10, stand-alone CM behaves as AL
 299
 300 $LB8NonBreaks $IS;
 301 $CAN_CM $CM*  $IS;
 302 ^$CM+         $IS;              # by rule 10, stand-alone CM behaves as AL
 303
 304 $LB8NonBreaks $SY;
 305 $CAN_CM $CM*  $SY;
 306 ^$CM+         $SY;              # by rule 10, stand-alone CM behaves as AL
 307
 308
 309 #
 310 # LB 14  Do not break after OP, even after spaces
 311 #
 312 $OPcm $SP* $CAN_CM $CM*;
 313 $OPcm $SP* $CANT_CM;
 314
 315 $OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
 316
 317 # LB 15
 318 $QUcm $SP* $OPcm;
 319
 320 # LB 16
 321 ($CLcm | $CPcm) $SP* $NScm;
 322
 323 # LB 17
 324 $B2cm $SP* $B2cm;
 325
 326 #
 327 # LB 18  Break after spaces.
 328 #
 329 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
 330 $LB18Breaks    = [$LB8Breaks $SP];
 331
 332
 333 # LB 19
 334 #         x QU
 335 $LB18NonBreaks $CM* $QUcm;
 336 ^$CM+               $QUcm;
 337
 338 #         QU  x
 339 $QUcm .?;
 340
 341
 342 # LB 20
 343 #        <break>  $CB
 344 #        $CB   <break>
 345
 346 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 347
 348 # LB 21        x   (BA | HY | NS)
 349 #           BB x
 350 #
 351 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
 352 ^$CM+ ($BAcm | $HYcm | $NScm);
 353
 354 $BBcm [^$CB];                                  #  $BB  x
 355 $BBcm $LB20NonBreaks $CM*;
 356
 357 # LB 21a Don't break after Hebrew + Hyphen
 358 #   HL (HY | BA) x
 359 #
 360 $HLcm ($HYcm | $BAcm) [^$CB]?;
 361
 362 # LB 21b (forward) Don't break between SY and HL
 363 # (break between HL and SY already disallowed by LB 13 above)
 364 $SYcm $HLcm;
 365
 366 # LB 22
 367 ($ALcm | $HLcm) $INcm;
 368 ^$CM+    $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 369 $EXcm    $INcm;
 370 ($ID | $EB | $EM) $CM*  $INcm;
 371 $INcm    $INcm;
 372 $NUcm    $INcm;
 373
 374
 375 # $LB 23
 376 ($ID | $EB | $EM) $CM*  $POcm;
 377 $ALcm  $NUcm;       # includes $LB19
 378 $HLcm  $NUcm;
 379 ^$CM+  $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
 380 $NUcm  $ALcm;
 381 $NUcm  $HLcm;
 382
 383 #
 384 # LB 24
 385 #
 386 $PRcm ($ID | $EB | $EM);
 387 $PRcm ($ALcm | $HLcm);
 388 $POcm ($ALcm | $HLcm);
 389 ($ALcm | $HLcm) ($PRcm | $POcm);
 390 ^$CM+ ($PRcm | $POcm);       # Rule 10, any otherwise unattached CM behaves as AL
 391
 392 #
 393 # LB 25   Numbers.
 394 #
 395 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
 396
 397 # LB 26  Do not break a Korean syllable
 398 #
 399 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 400 ($JVcm | $H2cm) ($JVcm | $JTcm);
 401 ($JTcm | $H3cm) $JTcm;
 402
 403 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 404 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 405 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 406 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 407
 408
 409 # LB 28   Do not break between alphabetics
 410 #
 411 ($ALcm | $HLcm) ($ALcm | $HLcm);
 412 ^$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
 413
 414 # LB 29
 415 $IScm ($ALcm | $HLcm);
 416
 417 # LB 30
 418 ($ALcm | $HLcm | $NUcm) $OPcm;
 419 ^$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
 420 $CPcm ($ALcm | $HLcm | $NUcm);
 421
 422 # LB 30a  Do not break between regional indicators. Break after pairs of them.
 423 #         Tricky interaction with LB8a: ZWJ x ID
 424 $RIcm $RI                / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
 425 $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}];
 426 $RIcm $RI $CM* [$CM-$ZWJ] / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
 427
 428 $RIcm $RIcm [$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS {eof}];
 429 $RIcm $RIcm $ZWJ ($ID | $EB | $EM);
 430
 431 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 432 $EB $CM* $EM;
 433
 434 #
 435 #  Reverse Rules.
 436 #
 437 ## -------------------------------------------------
 438
 439 !!reverse;
 440
 441 ^$CM+ $ALPlus;
 442 ^$CM+ $BA;
 443 ^$CM+ $BB;
 444 ^$CM+ $B2;
 445 ^$CM+ $CL;
 446 ^$CM+ $CP;
 447 ^$CM+ $EB;
 448 ^$CM+ $EM;
 449 ^$CM+ $EX;
 450 ^$CM+ $GL;
 451 ^$CM+ $HL;
 452 ^$CM+ $HY;
 453 ^$CM+ $H2;
 454 ^$CM+ $H3;
 455 ^$CM+ $ID;
 456 ^$CM+ $IN;
 457 ^$CM+ $IS;
 458 ^$CM+ $JL;
 459 ^$CM+ $JV;
 460 ^$CM+ $JT;
 461 ^$CM+ $NS;
 462 ^$CM+ $NU;
 463 ^$CM+ $OP;
 464 ^$CM+ $PO;
 465 ^$CM+ $PR;
 466 ^$CM+ $QU;
 467 ^$CM+ $RI;
 468 ^$CM+ $SY;
 469 ^$CM+ $WJ;
 470 ^$CM+;
 471
 472
 473 #
 474 #  Sequences of the form  (shown forwards)
 475 #      [CANT_CM]  <break>  [CM]  [whatever]
 476 #  The CM needs to behave as an AL
 477 #
 478 $AL_FOLLOW $CM+ / (
 479           [$BK $CR $LF $NL $ZW {eof}] |
 480           $SP+ $CM+ $SP |
 481           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 482                                                #  LB14 says    OP SP* x .
 483                                                #    becomes    OP SP* x AL
 484                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 485                                                #
 486                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 487                                                #                a rule compiler bug which complains about
 488                                                #                empty sets otherwise.
 489
 490 #
 491 #  Sequences of the form  (shown forwards)
 492 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 493 #  The CM needs to behave as an AL
 494 #  This rule is concerned about getting the second of the two <breaks> in place.
 495 #
 496
 497 # Apple early addition, remove this, superseded by LB24
 498 # [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 499
 500
 501
 502 # LB 4, 5, 6
 503
 504 $LB4Breaks [$LB4NonBreaks-$CM];
 505 $LB4Breaks $CM+ $CAN_CM;
 506 $LF $CR;
 507
 508
 509 # LB 7         x SP
 510 #              x ZW
 511 [$SP $ZW] [$LB4NonBreaks-$CM];
 512 [$SP $ZW] $CM+ $CAN_CM;
 513
 514 # LB 8 ZW SP* <break>
 515 #     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
 516 #           Requires an engine enhancement.
 517 #   / $SP* $ZW
 518
 519 # LB 8a        ZWJ x ID    Unicode Emoji proposal L2/16-011R3
 520 #                          The ZWJ will look like a CM to whatever precedes it.
 521 #
 522 ($ID | $EB | $EM) $ZWJ $CM* $CAN_CM?;
 523
 524
 525 # LB 9,10  Combining marks.
 526 #    X   $CM needs to behave like X, where X is not $SP or controls.
 527 #    $CM not covered by the above needs to behave like $AL
 528 # Stick together any combining sequences that don't match other rules.
 529 ^$CM+ $CAN_CM;
 530
 531
 532 # LB 11
 533 #
 534 $WJ $CM* $CAN_CM;
 535 $WJ      [$LB8NonBreaks-$CM];
 536
 537      $CANT_CM $CM* $WJ;
 538 $CAN_CM  $CM* $WJ;
 539
 540 # LB 12a
 541 #      [^SP BA HY] x GL
 542 #
 543 $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
 544
 545 # LB 12
 546 #     GL  x
 547 #
 548 $CANT_CM $CM* $GL;
 549 $CAN_CM $CM* $GL;
 550
 551
 552 # LB 13
 553 $CL $CM+ $CAN_CM;
 554 $CP $CM+ $CAN_CM;
 555 $EX $CM+ $CAN_CM;
 556 $IS $CM+ $CAN_CM;
 557 $SY $CM+ $CAN_CM;
 558
 559 $CL [$LB8NonBreaks-$CM];
 560 $CP [$LB8NonBreaks-$CM];
 561 $EX [$LB8NonBreaks-$CM];
 562 $IS [$LB8NonBreaks-$CM];
 563 $SY [$LB8NonBreaks-$CM];
 564
 565 # Rule 13 & 14 taken together for an edge case.
 566 #   Match this, shown forward
 567 #     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
 568 #   This really wants to chain at the $CM+ (which is acting as an $AL)
 569 #   except for $CM chaining being disabled.
 570 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
 571
 572 # LB 14    OP SP* x
 573 #
 574 $CAN_CM    $SP* $CM* $OP;
 575      $CANT_CM   $SP* $CM* $OP;
 576 $AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
 577
 578      $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
 579 $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 580
 581
 582 # LB 15
 583 $OP $SP* $CM* $QU;
 584
 585 # LB 16
 586 $NS $SP* $CM* ($CL | $CP);
 587
 588 # LB 17
 589 $B2 $SP* $CM* $B2;
 590
 591 # LB 18  break after spaces
 592 #        Nothing explicit needed here.
 593
 594
 595 #
 596 # LB 19
 597 #
 598 $QU $CM* $CAN_CM;                                #   . x QU
 599 $QU      $LB18NonBreaks;
 600
 601
 602 $CAN_CM  $CM* $QU;                               #   QU x .
 603      $CANT_CM $CM* $QU;
 604
 605 #
 606 #  LB 20  Break before and after CB.
 607 #         nothing needed here.
 608 #
 609
 610 # LB 21
 611 ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 612
 613 [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 614 [^$CB] $CM* $BB;                                      #
 615
 616 # LB21a
 617 [^$CB] $CM* ($HY | $BA) $CM* $HL;
 618
 619 # LB21b (reverse)
 620 $HL $CM* $SY;
 621
 622 # LB 22
 623 $IN $CM* ($ALPlus | $HL);
 624 $IN $CM* $EX;
 625 $IN $CM* ($ID | $EB | $EM);
 626 $IN $CM* $IN;
 627 $IN $CM* $NU;
 628
 629 # LB 23
 630 $PO $CM* ($ID | $EB | $EM);
 631 $NU $CM* ($ALPlus | $HL);
 632 ($ALPlus | $HL) $CM* $NU;
 633
 634 # LB 24
 635 ($ID | $EB | $EM) $CM* $PR;
 636 ($ALPlus | $HL) $CM* $PR;
 637 ($ALPlus | $HL) $CM* $PO;
 638 $CM* ($PR | $PO) $CM* ($ALPlus | $HL);
 639 $CM* ($PR | $PO) $CM+ / [$BK $CR $LF $NL $ZW $SP {eof}];
 640
 641 # LB 25
 642 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 643
 644 # LB 26
 645 ($H3 | $H2 | $JV | $JL) $CM* $JL;
 646 ($JT | $JV) $CM* ($H2 | $JV);
 647 $JT $CM* ($H3 | $JT);
 648
 649 # LB 27
 650 $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 651 $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 652  ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 653
 654 # LB 28
 655 ($ALPlus | $HL) $CM* ($ALPlus | $HL);
 656
 657
 658 # LB 29
 659 ($ALPlus | $HL) $CM* $IS;
 660
 661 # LB 30
 662 $OP $CM* ($ALPlus | $HL | $NU);
 663 ($ALPlus | $HL | $NU) $CM* $CP;
 664
 665 # LB 30a
 666 #    Pairs of Regional Indicators.
 667 #    The following two rules are nearly identical. The first matches only sequences with an odd number of adjacent RIs,
 668 #    the second with an even number. Stripping away the cruft they look like
 669 #         [^RI] RI / (RI RI)+ ^RI;
 670 #         [^RI] RI RI / (RI RI)+ ^RI;
 671 #
 672 [{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]];
 673 [{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]];
 674
 675 # In general, adjacent RIs stay together. The hard-break rules, above, overide this, forcing in the boundaries between pairs.
 676 $RI $CM* $RI;
 677
 678 #    WJ, GL, QU, etc. are classes with rules like "WJ x "   which includes "WJ x RI".
 679 $RI $CM* ([$WJ $GL $QU $BB] |  (($HY | $BA)$CM* $HL));
 680
 681
 682 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 683 $EM $CM* $EB;
 684
 685
 686 ## -------------------------------------------------
 687
 688 !!safe_reverse;
 689
 690 # LB 9
 691 ^$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 692 ^$CM+ $SP / .;
 693
 694 # LB 14
 695 $SP+ $CM* $OP;
 696
 697 # LB 15
 698 $SP+ $CM* $QU;
 699
 700 # LB 16
 701 $SP+ $CM* ($CL | $CP);
 702
 703 # LB 17
 704 $SP+ $CM* $B2;
 705
 706 # LB 21
 707 $CM* ($HY | $BA) $CM* $HL;
 708
 709 # LB 25
 710 ($CM* ($IS | $SY))+ $CM* $NU;
 711 ($CL | $CP) $CM* ($NU | $IS | $SY);
 712
 713 #  LB 30
 714 ($CM* $RI)+;
 715
 716 # For dictionary-based break
 717 $dictionary $dictionary;
 718
 719 ## -------------------------------------------------
 720
 721 !!safe_forward;
 722
 723 # Skip forward over all character classes that are involved in
 724 #   rules containing patterns with possibly more than one char
 725 #   of context.
 726 #
 727 #  It might be slightly more efficient to have specific rules
 728 #  instead of one generic one, but only if we could
 729 #  turn off rule chaining.  We don't want to move more
 730 #  than necessary.
 731 #
 732 ^[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $RI $ZWJ $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $RI $ZWJ $dictionary];
 733 $dictionary $dictionary;
 734