icuSources/data/brkitr/rules/line.txt

   1 # Copyright (c) 2002-2016  International Business Machines Corporation and
   2 # others. All Rights Reserved.
   3 #
   4 #  file:  line.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 #         TODO:  Rule LB 8 remains as it was in Unicode 5.2
  15 #         This is only because of a limitation of ICU break engine implementation,
  16 #         not because the older behavior is desirable.
  17 #
  18 #         This corresponds to CSS line-break=strict (BCP47 -u-lb-strict).
  19 #         It sets characters of class CJ to behave like NS.
  20
  21 #
  22 #  Character Classes defined by TR 14.
  23 #
  24
  25 !!chain;
  26
  27 !!lookAheadHardBreak;
  28 #
  29 #  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
  30 #                          and only used for the line break rules.
  31 #
  32 #           It is used in the implementation of rule LB 10
  33 #           which says to treat any combining mark that is not attached to a base
  34 #           character as if it were of class AL  (alphabetic).
  35 #
  36 #           The problem occurs in the reverse rules.
  37 #
  38 #           Consider a sequence like, with correct breaks as shown
  39 #               LF  ID  CM  AL  AL
  40 #                  ^       ^       ^
  41 #           Then consider the sequence without the initial ID (ideographic)
  42 #                 LF  CM  AL  AL
  43 #                    ^           ^
  44 #           Our CM, which in the first example was attached to the ideograph,
  45 #           is now unattached, becomes an alpha, and joins in with the other
  46 #           alphas.
  47 #
  48 #           When iterating forwards, these sequences do not present any problems
  49 #           When iterating backwards, we need to look ahead when encountering
  50 #           a CM to see whether it attaches to something further on or not.
  51 #           (Look-ahead in a reverse rule is looking towards the start)
  52 #
  53 #           If the CM is unattached, we need to force a break.
  54 #
  55 #           !!lookAheadHardBreak forces the run time state machine to
  56 #           stop immediately when a look ahead rule ( '/' operator) matches,
  57 #           and set the match position to that of the look-ahead operator,
  58 #           no matter what other rules may be in play at the time.
  59 #
  60 #           See rule LB 19 for an example.
  61 #
  62
  63 # Temporary definitions of Emoji Base and Emoji Modifiers, until properties are available.
  64
  65 # Apple $EmojiForMods becomes $EB here (deletes \U0001F46F, should not have been in $EmojiForMods)
  66 $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];
  67 # Apple $EmojiMods becomes $EM here, same set
  68 $EM = [\U0001F3FB-\U0001F3FF];
  69
  70 $AI = [[:LineBreak =  Ambiguous:] - [\u2640\u2642]];
  71 $AL = [[:LineBreak =  Alphabetic:] - [$EM\u2695\u2696\u2764]];
  72 $BA = [:LineBreak =  Break_After:];
  73 $BB = [:LineBreak =  Break_Before:];
  74 $BK = [:LineBreak =  Mandatory_Break:];
  75 $B2 = [:LineBreak =  Break_Both:];
  76 $CB = [:LineBreak =  Contingent_Break:];
  77 $CJ = [:LineBreak =  Conditional_Japanese_Starter:];
  78 $CL = [:LineBreak =  Close_Punctuation:];
  79 $CM = [[:LineBreak =  Combining_Mark:] \u200d];
  80 $CP = [:LineBreak =  Close_Parenthesis:];
  81 $CR = [:LineBreak =  Carriage_Return:];
  82 $EX = [:LineBreak =  Exclamation:];
  83 $GL = [:LineBreak =  Glue:];
  84 $HL = [:LineBreak =  Hebrew_Letter:];
  85 $HY = [:LineBreak =  Hyphen:];
  86 $H2 = [:LineBreak =  H2:];
  87 $H3 = [:LineBreak =  H3:];
  88 $ID = [[:LineBreak =  Ideographic:][\u2640\u2642\u2695\u2696\u2764] - $EB];
  89 $IN = [:LineBreak =  Inseperable:];
  90 $IS = [:LineBreak =  Infix_Numeric:];
  91 $JL = [:LineBreak =  JL:];
  92 $JV = [:LineBreak =  JV:];
  93 $JT = [:LineBreak =  JT:];
  94 $LF = [:LineBreak =  Line_Feed:];
  95 $NL = [:LineBreak =  Next_Line:];
  96 $NS = [[:LineBreak =  Nonstarter:] $CJ];
  97 $NU = [:LineBreak =  Numeric:];
  98 $OP = [:LineBreak =  Open_Punctuation:];
  99 $PO = [:LineBreak =  Postfix_Numeric:];
 100 $PR = [:LineBreak =  Prefix_Numeric:];
 101 $QU = [:LineBreak =  Quotation:];
 102 $RI = [:LineBreak =  Regional_Indicator:];
 103 $SA = [:LineBreak =  Complex_Context:];
 104 $SG = [:LineBreak =  Surrogate:];
 105 $SP = [:LineBreak =  Space:];
 106 $SY = [:LineBreak =  Break_Symbols:];
 107 $WJ = [:LineBreak =  Word_Joiner:];
 108 $XX = [:LineBreak =  Unknown:];
 109 $ZW = [:LineBreak =  ZWSpace:];
 110 $ZWJ = [\u200d];
 111
 112 #   Dictionary character set, for triggering language-based break engines. Currently
 113 #   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
 114 #   5.0 or later as the definition of Complex_Context was corrected to include all
 115 #   characters requiring dictionary break.
 116
 117 $dictionary = [:LineBreak = Complex_Context:];
 118
 119 #
 120 #  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
 121 #                               SA  (South East Asian: Thai, Lao, Khmer)
 122 #                               SG  (Unpaired Surrogates)
 123 #                               XX  (Unknown, unassigned)
 124 #                         as $AL  (Alphabetic)
 125 #
 126 $ALPlus = [$AL $AI $SA $SG $XX];
 127
 128 #
 129 #  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
 130 #
 131 $ALcm = $ALPlus $CM*;
 132 $BAcm = $BA $CM*;
 133 $BBcm = $BB $CM*;
 134 $B2cm = $B2 $CM*;
 135 $CLcm = $CL $CM*;
 136 $CPcm = $CP $CM*;
 137 $EXcm = $EX $CM*;
 138 $GLcm = $GL $CM*;
 139 $HLcm = $HL $CM*;
 140 $HYcm = $HY $CM*;
 141 $H2cm = $H2 $CM*;
 142 $H3cm = $H3 $CM*;
 143 $INcm = $IN $CM*;
 144 $IScm = $IS $CM*;
 145 $JLcm = $JL $CM*;
 146 $JVcm = $JV $CM*;
 147 $JTcm = $JT $CM*;
 148 $NScm = $NS $CM*;
 149 $NUcm = $NU $CM*;
 150 $OPcm = $OP $CM*;
 151 $POcm = $PO $CM*;
 152 $PRcm = $PR $CM*;
 153 $QUcm = $QU $CM*;
 154 $RIcm = $RI $CM*;
 155 $SYcm = $SY $CM*;
 156 $WJcm = $WJ $CM*;
 157
 158 ## -------------------------------------------------
 159
 160 !!forward;
 161
 162 #
 163 #  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
 164 #
 165 $ALPlus $CM+;
 166 $BA $CM+;
 167 $BB $CM+;
 168 $B2 $CM+;
 169 $CL $CM+;
 170 $CP $CM+;
 171 $EB $CM+;
 172 $EM $CM+;
 173 $EX $CM+;
 174 $GL $CM+;
 175 $HL $CM+;
 176 $HY $CM+;
 177 $H2 $CM+;
 178 $H3 $CM+;
 179 $ID $CM+;
 180 $IN $CM+;
 181 $IS $CM+;
 182 $JL $CM+;
 183 $JV $CM+;
 184 $JT $CM+;
 185 $NS $CM+;
 186 $NU $CM+;
 187 $OP $CM+;
 188 $PO $CM+;
 189 $PR $CM+;
 190 $QU $CM+;
 191 $RI $CM+;
 192 $SY $CM+;
 193 $WJ $CM+;
 194
 195 #
 196 # CAN_CM  is the set of characters that may combine with CM combining chars.
 197 #         Note that Linebreak UAX 14's concept of a combining char and the rules
 198 #         for what they can combine with are _very_ different from the rest of Unicode.
 199 #
 200 #         Note that $CM itself is left out of this set.  If CM is needed as a base
 201 #         it must be listed separately in the rule.
 202 #
 203 $CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
 204 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
 205
 206 #
 207 # AL_FOLLOW  set of chars that can unconditionally follow an AL
 208 #            Needed in rules where stand-alone $CM s are treated as AL.
 209 #            Chaining is disabled with CM because it causes other failures,
 210 #            so for this one case we need to manually list out longer sequences.
 211 #
 212 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
 213 $AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
 214 $AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
 215
 216
 217 #
 218 #  Rule LB 4, 5    Mandatory (Hard) breaks.
 219 #
 220 $LB4Breaks    = [$BK $CR $LF $NL];
 221 $LB4NonBreaks = [^$BK $CR $LF $NL $CM];
 222 $CR $LF {100};
 223
 224 #
 225 #  LB 6    Do not break before hard line breaks.
 226 #
 227 $LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
 228 $CAN_CM $CM*    $LB4Breaks {100};
 229 ^$CM+           $LB4Breaks {100};
 230
 231 # LB 7         x SP
 232 #              x ZW
 233 $LB4NonBreaks [$SP $ZW];
 234 $CAN_CM $CM*  [$SP $ZW];
 235 ^$CM+         [$SP $ZW];
 236
 237 #
 238 # LB 8         Break after zero width space
 239 #              TODO:  ZW SP* <break>
 240 #              An engine change is required to write the reverse rule for this.
 241 #              For now, leave the Unicode 5.2 rule, ZW <break>
 242 #
 243 $LB8Breaks    = [$LB4Breaks $ZW];
 244 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
 245
 246 # LB 8a        ZWJ x ID    Emoji proposal.
 247 #
 248 $ZWJ ($ID | $EB | $EM);
 249
 250 # LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
 251 #                                $CM not covered by the above needs to behave like $AL
 252 #                                See definition of $CAN_CM.
 253
 254 $CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
 255 ^$CM+;
 256
 257 #
 258 # LB 11  Do not break before or after WORD JOINER & related characters.
 259 #
 260 $CAN_CM $CM*  $WJcm;
 261 $LB8NonBreaks $WJcm;
 262 ^$CM+         $WJcm;
 263
 264 $WJcm $CANT_CM;
 265 $WJcm $CAN_CM $CM*;
 266
 267 #
 268 # LB 12  Do not break after NBSP and related characters.
 269 #         GL  x
 270 #
 271 $GLcm $CAN_CM $CM*;
 272 $GLcm $CANT_CM;
 273
 274 #
 275 # LB 12a  Do not break before NBSP and related characters ...
 276 #            [^SP BA HY] x GL
 277 #
 278 [[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
 279 ^$CM+ $GLcm;
 280
 281
 282
 283 #
 284 # LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
 285 #
 286 $LB8NonBreaks $CL;
 287 $CAN_CM $CM*  $CL;
 288 ^$CM+         $CL;              # by rule 10, stand-alone CM behaves as AL
 289
 290 $LB8NonBreaks $CP;
 291 $CAN_CM $CM*  $CP;
 292 ^$CM+         $CP;              # by rule 10, stand-alone CM behaves as AL
 293
 294 $LB8NonBreaks $EX;
 295 $CAN_CM $CM*  $EX;
 296 ^$CM+         $EX;              # by rule 10, stand-alone CM behaves as AL
 297
 298 $LB8NonBreaks $IS;
 299 $CAN_CM $CM*  $IS;
 300 ^$CM+         $IS;              # by rule 10, stand-alone CM behaves as AL
 301
 302 $LB8NonBreaks $SY;
 303 $CAN_CM $CM*  $SY;
 304 ^$CM+         $SY;              # by rule 10, stand-alone CM behaves as AL
 305
 306
 307 #
 308 # LB 14  Do not break after OP, even after spaces
 309 #
 310 $OPcm $SP* $CAN_CM $CM*;
 311 $OPcm $SP* $CANT_CM;
 312
 313 $OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
 314
 315 # LB 15
 316 $QUcm $SP* $OPcm;
 317
 318 # LB 16
 319 ($CLcm | $CPcm) $SP* $NScm;
 320
 321 # LB 17
 322 $B2cm $SP* $B2cm;
 323
 324 #
 325 # LB 18  Break after spaces.
 326 #
 327 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
 328 $LB18Breaks    = [$LB8Breaks $SP];
 329
 330
 331 # LB 19
 332 #         x QU
 333 $LB18NonBreaks $CM* $QUcm;
 334 ^$CM+               $QUcm;
 335
 336 #         QU  x
 337 $QUcm .?;
 338
 339
 340 # LB 20
 341 #        <break>  $CB
 342 #        $CB   <break>
 343
 344 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 345
 346 # LB 21        x   (BA | HY | NS)
 347 #           BB x
 348 #
 349 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
 350 ^$CM+ ($BAcm | $HYcm | $NScm);
 351
 352 $BBcm [^$CB];                                  #  $BB  x
 353 $BBcm $LB20NonBreaks $CM*;
 354
 355 # LB 21a Don't break after Hebrew + Hyphen
 356 #   HL (HY | BA) x
 357 #
 358 $HLcm ($HYcm | $BAcm) [^$CB]?;
 359
 360 # LB 21b (forward) Don't break between SY and HL
 361 # (break between HL and SY already disallowed by LB 13 above)
 362 $SYcm $HLcm;
 363
 364 # LB 22
 365 ($ALcm | $HLcm) $INcm;
 366 ^$CM+    $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 367 $EXcm    $INcm;
 368 ($ID | $EB | $EM) $CM*  $INcm;
 369 $INcm    $INcm;
 370 $NUcm    $INcm;
 371
 372
 373 # $LB 23
 374 ($ID | $EB | $EM) $CM*  $POcm;
 375 $ALcm  $NUcm;       # includes $LB19
 376 $HLcm  $NUcm;
 377 ^$CM+  $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
 378 $NUcm  $ALcm;
 379 $NUcm  $HLcm;
 380
 381 #
 382 # LB 24
 383 #
 384 $PRcm ($ID | $EB | $EM);
 385 $PRcm ($ALcm | $HLcm);
 386 $POcm ($ALcm | $HLcm);
 387 ($ALcm | $HLcm) ($PRcm | $POcm);
 388 ^$CM+ ($PRcm | $POcm);       # Rule 10, any otherwise unattached CM behaves as AL
 389
 390 #
 391 # LB 25   Numbers.
 392 #
 393 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
 394
 395 # LB 26  Do not break a Korean syllable
 396 #
 397 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 398 ($JVcm | $H2cm) ($JVcm | $JTcm);
 399 ($JTcm | $H3cm) $JTcm;
 400
 401 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 402 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 403 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 404 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 405
 406
 407 # LB 28   Do not break between alphabetics
 408 #
 409 ($ALcm | $HLcm) ($ALcm | $HLcm);
 410 ^$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
 411
 412 # LB 29
 413 $IScm ($ALcm | $HLcm);
 414
 415 # LB 30
 416 ($ALcm | $HLcm | $NUcm) $OPcm;
 417 ^$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
 418 $CPcm ($ALcm | $HLcm | $NUcm);
 419
 420 # LB 30a  Do not break between regional indicators. Break after pairs of them.
 421 #         Tricky interaction with LB8a: ZWJ x ID
 422 $RIcm $RI                / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
 423 $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}];
 424 $RIcm $RI $CM* [$CM-$ZWJ] / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
 425
 426 $RIcm $RIcm [$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS {eof}];
 427 $RIcm $RIcm $ZWJ ($ID | $EB | $EM);
 428
 429 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 430 $EB $CM* $EM;
 431
 432 #
 433 #  Reverse Rules.
 434 #
 435 ## -------------------------------------------------
 436
 437 !!reverse;
 438
 439 ^$CM+ $ALPlus;
 440 ^$CM+ $BA;
 441 ^$CM+ $BB;
 442 ^$CM+ $B2;
 443 ^$CM+ $CL;
 444 ^$CM+ $CP;
 445 ^$CM+ $EB;
 446 ^$CM+ $EM;
 447 ^$CM+ $EX;
 448 ^$CM+ $GL;
 449 ^$CM+ $HL;
 450 ^$CM+ $HY;
 451 ^$CM+ $H2;
 452 ^$CM+ $H3;
 453 ^$CM+ $ID;
 454 ^$CM+ $IN;
 455 ^$CM+ $IS;
 456 ^$CM+ $JL;
 457 ^$CM+ $JV;
 458 ^$CM+ $JT;
 459 ^$CM+ $NS;
 460 ^$CM+ $NU;
 461 ^$CM+ $OP;
 462 ^$CM+ $PO;
 463 ^$CM+ $PR;
 464 ^$CM+ $QU;
 465 ^$CM+ $RI;
 466 ^$CM+ $SY;
 467 ^$CM+ $WJ;
 468 ^$CM+;
 469
 470
 471 #
 472 #  Sequences of the form  (shown forwards)
 473 #      [CANT_CM]  <break>  [CM]  [whatever]
 474 #  The CM needs to behave as an AL
 475 #
 476 $AL_FOLLOW $CM+ / (
 477           [$BK $CR $LF $NL $ZW {eof}] |
 478           $SP+ $CM+ $SP |
 479           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 480                                                #  LB14 says    OP SP* x .
 481                                                #    becomes    OP SP* x AL
 482                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 483                                                #
 484                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 485                                                #                a rule compiler bug which complains about
 486                                                #                empty sets otherwise.
 487
 488 #
 489 #  Sequences of the form  (shown forwards)
 490 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 491 #  The CM needs to behave as an AL
 492 #  This rule is concerned about getting the second of the two <breaks> in place.
 493 #
 494
 495 # Apple early addition, remove this, superseded by LB24
 496 # [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 497
 498
 499
 500 # LB 4, 5, 6
 501
 502 $LB4Breaks [$LB4NonBreaks-$CM];
 503 $LB4Breaks $CM+ $CAN_CM;
 504 $LF $CR;
 505
 506
 507 # LB 7         x SP
 508 #              x ZW
 509 [$SP $ZW] [$LB4NonBreaks-$CM];
 510 [$SP $ZW] $CM+ $CAN_CM;
 511
 512 # LB 8 ZW SP* <break>
 513 #     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
 514 #           Requires an engine enhancement.
 515 #   / $SP* $ZW
 516
 517 # LB 8a        ZWJ x ID    Unicode Emoji proposal L2/16-011R3
 518 #                          The ZWJ will look like a CM to whatever precedes it.
 519 #
 520 ($ID | $EB | $EM) $ZWJ $CM* $CAN_CM?;
 521
 522
 523 # LB 9,10  Combining marks.
 524 #    X   $CM needs to behave like X, where X is not $SP or controls.
 525 #    $CM not covered by the above needs to behave like $AL
 526 # Stick together any combining sequences that don't match other rules.
 527 ^$CM+ $CAN_CM;
 528
 529
 530 # LB 11
 531 #
 532 $WJ $CM* $CAN_CM;
 533 $WJ      [$LB8NonBreaks-$CM];
 534
 535      $CANT_CM $CM* $WJ;
 536 $CAN_CM  $CM* $WJ;
 537
 538 # LB 12a
 539 #      [^SP BA HY] x GL
 540 #
 541 $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
 542
 543 # LB 12
 544 #     GL  x
 545 #
 546 $CANT_CM $CM* $GL;
 547 $CAN_CM $CM* $GL;
 548
 549
 550 # LB 13
 551 $CL $CM+ $CAN_CM;
 552 $CP $CM+ $CAN_CM;
 553 $EX $CM+ $CAN_CM;
 554 $IS $CM+ $CAN_CM;
 555 $SY $CM+ $CAN_CM;
 556
 557 $CL [$LB8NonBreaks-$CM];
 558 $CP [$LB8NonBreaks-$CM];
 559 $EX [$LB8NonBreaks-$CM];
 560 $IS [$LB8NonBreaks-$CM];
 561 $SY [$LB8NonBreaks-$CM];
 562
 563 # Rule 13 & 14 taken together for an edge case.
 564 #   Match this, shown forward
 565 #     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
 566 #   This really wants to chain at the $CM+ (which is acting as an $AL)
 567 #   except for $CM chaining being disabled.
 568 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
 569
 570 # LB 14    OP SP* x
 571 #
 572 $CAN_CM    $SP* $CM* $OP;
 573      $CANT_CM   $SP* $CM* $OP;
 574 $AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
 575
 576      $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
 577 $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 578
 579
 580 # LB 15
 581 $OP $SP* $CM* $QU;
 582
 583 # LB 16
 584 $NS $SP* $CM* ($CL | $CP);
 585
 586 # LB 17
 587 $B2 $SP* $CM* $B2;
 588
 589 # LB 18  break after spaces
 590 #        Nothing explicit needed here.
 591
 592
 593 #
 594 # LB 19
 595 #
 596 $QU $CM* $CAN_CM;                                #   . x QU
 597 $QU      $LB18NonBreaks;
 598
 599
 600 $CAN_CM  $CM* $QU;                               #   QU x .
 601      $CANT_CM $CM* $QU;
 602
 603 #
 604 #  LB 20  Break before and after CB.
 605 #         nothing needed here.
 606 #
 607
 608 # LB 21
 609 ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 610
 611 [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 612 [^$CB] $CM* $BB;                                      #
 613
 614 # LB21a
 615 [^$CB] $CM* ($HY | $BA) $CM* $HL;
 616
 617 # LB21b (reverse)
 618 $HL $CM* $SY;
 619
 620 # LB 22
 621 $IN $CM* ($ALPlus | $HL);
 622 $IN $CM* $EX;
 623 $IN $CM* ($ID | $EB | $EM);
 624 $IN $CM* $IN;
 625 $IN $CM* $NU;
 626
 627 # LB 23
 628 $PO $CM* ($ID | $EB | $EM);
 629 $NU $CM* ($ALPlus | $HL);
 630 ($ALPlus | $HL) $CM* $NU;
 631
 632 # LB 24
 633 ($ID | $EB | $EM) $CM* $PR;
 634 ($ALPlus | $HL) $CM* $PR;
 635 ($ALPlus | $HL) $CM* $PO;
 636 $CM* ($PR | $PO) $CM* ($ALPlus | $HL);
 637 $CM* ($PR | $PO) $CM+ / [$BK $CR $LF $NL $ZW $SP {eof}];
 638
 639 # LB 25
 640 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 641
 642 # LB 26
 643 ($H3 | $H2 | $JV | $JL) $CM* $JL;
 644 ($JT | $JV) $CM* ($H2 | $JV);
 645 $JT $CM* ($H3 | $JT);
 646
 647 # LB 27
 648 $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 649 $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 650  ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 651
 652 # LB 28
 653 ($ALPlus | $HL) $CM* ($ALPlus | $HL);
 654
 655
 656 # LB 29
 657 ($ALPlus | $HL) $CM* $IS;
 658
 659 # LB 30
 660 $OP $CM* ($ALPlus | $HL | $NU);
 661 ($ALPlus | $HL | $NU) $CM* $CP;
 662
 663 # LB 30a
 664 #    Pairs of Regional Indicators.
 665 #    The following two rules are nearly identical. The first matches only sequences with an odd number of adjacent RIs,
 666 #    the second with an even number. Stripping away the cruft they look like
 667 #         [^RI] RI / (RI RI)+ ^RI;
 668 #         [^RI] RI RI / (RI RI)+ ^RI;
 669 #
 670 [{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]];
 671 [{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]];
 672
 673 # In general, adjacent RIs stay together. The hard-break rules, above, overide this, forcing in the boundaries between pairs.
 674 $RI $CM* $RI;
 675
 676 #    WJ, GL, QU, etc. are classes with rules like "WJ x "   which includes "WJ x RI".
 677 $RI $CM* ([$WJ $GL $QU $BB] |  (($HY | $BA)$CM* $HL));
 678
 679
 680 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 681 $EM $CM* $EB;
 682
 683
 684 ## -------------------------------------------------
 685
 686 !!safe_reverse;
 687
 688 # LB 9
 689 ^$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 690 ^$CM+ $SP / .;
 691
 692 # LB 14
 693 $SP+ $CM* $OP;
 694
 695 # LB 15
 696 $SP+ $CM* $QU;
 697
 698 # LB 16
 699 $SP+ $CM* ($CL | $CP);
 700
 701 # LB 17
 702 $SP+ $CM* $B2;
 703
 704 # LB 21
 705 $CM* ($HY | $BA) $CM* $HL;
 706
 707 # LB 25
 708 ($CM* ($IS | $SY))+ $CM* $NU;
 709 ($CL | $CP) $CM* ($NU | $IS | $SY);
 710
 711 #  LB 30
 712 ($CM* $RI)+;
 713
 714 # For dictionary-based break
 715 $dictionary $dictionary;
 716
 717 ## -------------------------------------------------
 718
 719 !!safe_forward;
 720
 721 # Skip forward over all character classes that are involved in
 722 #   rules containing patterns with possibly more than one char
 723 #   of context.
 724 #
 725 #  It might be slightly more efficient to have specific rules
 726 #  instead of one generic one, but only if we could
 727 #  turn off rule chaining.  We don't want to move more
 728 #  than necessary.
 729 #
 730 ^[$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];
 731 $dictionary $dictionary;
 732