icuSources/data/brkitr/line_ja.txt

   1 # Copyright (c) 2002-2015  International Business Machines Corporation and
   2 # others. All Rights Reserved.
   3 #
   4 #  file:  line_ja.txt
   5 #
   6 #         Line Breaking Rules
   7 #         Implement default line breaking as defined by
   8 #         Unicode Standard Annex #14 Revision 29 for Unicode 6.2
   9 #         http://www.unicode.org/reports/tr14/
  10 #
  11 #         TODO:  Rule LB 8 remains as it was in Unicode 5.2
  12 #         This is only because of a limitation of ICU break engine implementation,
  13 #         not because the older behavior is desirable.
  14
  15 #
  16 #  Character Classes defined by TR 14.
  17 #
  18
  19 !!chain;
  20 !!LBCMNoChain;
  21
  22
  23 !!lookAheadHardBreak;
  24 #
  25 #  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
  26 #                          and only used for the line break rules.
  27 #
  28 #           It is used in the implementation of rule LB 10
  29 #           which says to treat any combining mark that is not attached to a base
  30 #           character as if it were of class AL  (alphabetic).
  31 #
  32 #           The problem occurs in the reverse rules.
  33 #
  34 #           Consider a sequence like, with correct breaks as shown
  35 #               LF  ID  CM  AL  AL
  36 #                  ^       ^       ^
  37 #           Then consider the sequence without the initial ID (ideographic)
  38 #                 LF  CM  AL  AL
  39 #                    ^           ^
  40 #           Our CM, which in the first example was attached to the ideograph,
  41 #           is now unattached, becomes an alpha, and joins in with the other
  42 #           alphas.
  43 #
  44 #           When iterating forwards, these sequences do not present any problems
  45 #           When iterating backwards, we need to look ahead when encountering
  46 #           a CM to see whether it attaches to something further on or not.
  47 #           (Look-ahead in a reverse rule is looking towards the start)
  48 #
  49 #           If the CM is unattached, we need to force a break.
  50 #
  51 #           !!lookAheadHardBreak forces the run time state machine to
  52 #           stop immediately when a look ahead rule ( '/' operator) matches,
  53 #           and set the match position to that of the look-ahead operator,
  54 #           no matter what other rules may be in play at the time.
  55 #
  56 #           See rule LB 19 for an example.
  57 #
  58
  59 $AI = [:LineBreak =  Ambiguous:];
  60 $AL = [:LineBreak =  Alphabetic:];
  61 $BA = [:LineBreak =  Break_After:];
  62 $BB = [:LineBreak =  Break_Before:];
  63 $BK = [:LineBreak =  Mandatory_Break:];
  64 $B2 = [:LineBreak =  Break_Both:];
  65 $CB = [:LineBreak =  Contingent_Break:];
  66 $CJ = [:LineBreak =  Conditional_Japanese_Starter:];
  67 $CL = [:LineBreak =  Close_Punctuation:];
  68 $CM = [:LineBreak =  Combining_Mark:];
  69 $CP = [:LineBreak =  Close_Parenthesis:];
  70 $CR = [:LineBreak =  Carriage_Return:];
  71 $EX = [:LineBreak =  Exclamation:];
  72 $GL = [:LineBreak =  Glue:];
  73 $HL = [:LineBreak =  Hebrew_Letter:];
  74 $HY = [:LineBreak =  Hyphen:];
  75 $H2 = [:LineBreak =  H2:];
  76 $H3 = [:LineBreak =  H3:];
  77 $ID = [[:LineBreak =  Ideographic:] $CJ];
  78 $IN = [:LineBreak =  Inseperable:];
  79 $IS = [:LineBreak =  Infix_Numeric:];
  80 $JL = [:LineBreak =  JL:];
  81 $JV = [:LineBreak =  JV:];
  82 $JT = [:LineBreak =  JT:];
  83 $LF = [:LineBreak =  Line_Feed:];
  84 $NL = [:LineBreak =  Next_Line:];
  85 $NS = [:LineBreak =  Nonstarter:];
  86 $NU = [:LineBreak =  Numeric:];
  87 $OP = [:LineBreak =  Open_Punctuation:];
  88 $PO = [:LineBreak =  Postfix_Numeric:];
  89 $PR = [:LineBreak =  Prefix_Numeric:];
  90 $QU = [:LineBreak =  Quotation:];
  91 $RI = [:LineBreak =  Regional_Indicator:];
  92 $SA = [:LineBreak =  Complex_Context:];
  93 $SG = [:LineBreak =  Surrogate:];
  94 $SP = [:LineBreak =  Space:];
  95 $SY = [:LineBreak =  Break_Symbols:];
  96 $WJ = [:LineBreak =  Word_Joiner:];
  97 $XX = [:LineBreak =  Unknown:];
  98 $ZW = [:LineBreak =  ZWSpace:];
  99
 100 # Special character classes for people & body part emoji:
 101 # Subsets of $CM:
 102 $ZWJ = \u200D;
 103 $EmojiVar = \uFE0F;
 104 # The following are subsets of $ID
 105 $EmojiForSeqs = [\u2764 \U0001F441 \U0001F466-\U0001F469 \U0001F48B \U0001F5E8];
 106 $EmojiForMods = [\u261D \u26F9 \u270A-\u270D \U0001F385 \U0001F3C3-\U0001F3C4 \U0001F3C7 \U0001F3CA-\U0001F3CB \U0001F442-\U0001F443 \U0001F446-\U0001F450 \U0001F466-\U0001F469 \U0001F46E-\U0001F478 \U0001F47C \U0001F481-\U0001F483 \U0001F485-\U0001F487 \U0001F4AA \U0001F590 \U0001F595 \U0001F596 \U0001F645-\U0001F647 \U0001F64B-\U0001F64F \U0001F6A3 \U0001F6B4-\U0001F6B6 \U0001F6C0 \U0001F918];
 107 $EmojiMods = [\U0001F3FB-\U0001F3FF];
 108
 109 #   Dictionary character set, for triggering language-based break engines. Currently
 110 #   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
 111 #   5.0 or later as the definition of Complex_Context was corrected to include all
 112 #   characters requiring dictionary break.
 113
 114 $dictionary = [:LineBreak = Complex_Context:];
 115
 116 #
 117 #  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
 118 #                               SA  (South East Asian: Thai, Lao, Khmer)
 119 #                               SG  (Unpaired Surrogates)
 120 #                               XX  (Unknown, unassigned)
 121 #                         as $AL  (Alphabetic)
 122 #
 123 $ALPlus = [$AL $AI $SA $SG $XX];
 124
 125 #
 126 #  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
 127 #
 128 $ALcm = $ALPlus $CM*;
 129 $BAcm = $BA $CM*;
 130 $BBcm = $BB $CM*;
 131 $B2cm = $B2 $CM*;
 132 $CLcm = $CL $CM*;
 133 $CPcm = $CP $CM*;
 134 $EXcm = $EX $CM*;
 135 $GLcm = $GL $CM*;
 136 $HLcm = $HL $CM*;
 137 $HYcm = $HY $CM*;
 138 $H2cm = $H2 $CM*;
 139 $H3cm = $H3 $CM*;
 140 $IDcm = $ID $CM*;
 141 $INcm = $IN $CM*;
 142 $IScm = $IS $CM*;
 143 $JLcm = $JL $CM*;
 144 $JVcm = $JV $CM*;
 145 $JTcm = $JT $CM*;
 146 $NScm = $NS $CM*;
 147 $NUcm = $NU $CM*;
 148 $OPcm = $OP $CM*;
 149 $POcm = $PO $CM*;
 150 $PRcm = $PR $CM*;
 151 $QUcm = $QU $CM*;
 152 $RIcm = $RI $CM*;
 153 $SYcm = $SY $CM*;
 154 $WJcm = $WJ $CM*;
 155
 156 ## -------------------------------------------------
 157
 158 !!forward;
 159
 160 #
 161 #  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
 162 #
 163 $ALPlus $CM+;
 164 $BA $CM+;
 165 $BB $CM+;
 166 $B2 $CM+;
 167 $CL $CM+;
 168 $CP $CM+;
 169 $EX $CM+;
 170 $GL $CM+;
 171 $HL $CM+;
 172 $HY $CM+;
 173 $H2 $CM+;
 174 $H3 $CM+;
 175 $ID $CM+;
 176 $IN $CM+;
 177 $IS $CM+;
 178 $JL $CM+;
 179 $JV $CM+;
 180 $JT $CM+;
 181 $NS $CM+;
 182 $NU $CM+;
 183 $OP $CM+;
 184 $PO $CM+;
 185 $PR $CM+;
 186 $QU $CM+;
 187 $RI $CM+;
 188 $SY $CM+;
 189 $WJ $CM+;
 190
 191 #
 192 # CAN_CM  is the set of characters that may combine with CM combining chars.
 193 #         Note that Linebreak UAX 14's concept of a combining char and the rules
 194 #         for what they can combine with are _very_ different from the rest of Unicode.
 195 #
 196 #         Note that $CM itself is left out of this set.  If CM is needed as a base
 197 #         it must be listed separately in the rule.
 198 #
 199 $CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
 200 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
 201
 202 #
 203 # AL_FOLLOW  set of chars that can unconditionally follow an AL
 204 #            Needed in rules where stand-alone $CM s are treated as AL.
 205 #            Chaining is disabled with CM because it causes other failures,
 206 #            so for this one case we need to manually list out longer sequences.
 207 #
 208 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
 209 $AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
 210 $AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
 211
 212
 213 #
 214 #  Rule LB 4, 5    Mandatory (Hard) breaks.
 215 #
 216 $LB4Breaks    = [$BK $CR $LF $NL];
 217 $LB4NonBreaks = [^$BK $CR $LF $NL];
 218 $CR $LF {100};
 219
 220 #
 221 #  LB 6    Do not break before hard line breaks.
 222 #
 223 $LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
 224 $CAN_CM $CM*    $LB4Breaks {100};
 225 $CM+            $LB4Breaks {100};
 226
 227 # LB 7         x SP
 228 #              x ZW
 229 $LB4NonBreaks [$SP $ZW];
 230 $CAN_CM $CM*  [$SP $ZW];
 231 $CM+          [$SP $ZW];
 232
 233 #
 234 # LB 8         Break after zero width space
 235 #              TODO:  ZW SP* <break>
 236 #              An engine change is required to write the reverse rule for this.
 237 #              For now, leave the Unicode 5.2 rule, ZW <break>
 238 #
 239 $LB8Breaks    = [$LB4Breaks $ZW];
 240 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
 241
 242
 243 # Special forward rule for people & body part emoji:
 244 # don't break $ZWJ from subsequent $EmojiForSeqs
 245 $EmojiForSeqs $EmojiVar? $EmojiMods? $ZWJ $EmojiForSeqs;
 246
 247 # LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
 248 #                                $CM not covered by the above needs to behave like $AL
 249 #                                See definition of $CAN_CM.
 250
 251 $CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
 252 $CM+;
 253
 254 #
 255 # LB 11  Do not break before or after WORD JOINER & related characters.
 256 #
 257 $CAN_CM $CM*  $WJcm;
 258 $LB8NonBreaks $WJcm;
 259 $CM+          $WJcm;
 260
 261 $WJcm $CANT_CM;
 262 $WJcm $CAN_CM $CM*;
 263
 264 #
 265 # LB 12  Do not break after NBSP and related characters.
 266 #         GL  x
 267 #
 268 $GLcm $CAN_CM $CM*;
 269 $GLcm $CANT_CM;
 270
 271 #
 272 # LB 12a  Do not break before NBSP and related characters ...
 273 #            [^SP BA HY] x GL
 274 #
 275 [[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
 276 $CM+ GLcm;
 277
 278
 279
 280 #
 281 # LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
 282 #
 283 $LB8NonBreaks $CL;
 284 $CAN_CM $CM*  $CL;
 285 $CM+          $CL;              # by rule 10, stand-alone CM behaves as AL
 286
 287 $LB8NonBreaks $CP;
 288 $CAN_CM $CM*  $CP;
 289 $CM+          $CP;              # by rule 10, stand-alone CM behaves as AL
 290
 291 $LB8NonBreaks $EX;
 292 $CAN_CM $CM*  $EX;
 293 $CM+          $EX;              # by rule 10, stand-alone CM behaves as AL
 294
 295 $LB8NonBreaks $IS;
 296 $CAN_CM $CM*  $IS;
 297 $CM+          $IS;              # by rule 10, stand-alone CM behaves as AL
 298
 299 $LB8NonBreaks $SY;
 300 $CAN_CM $CM*  $SY;
 301 $CM+          $SY;              # by rule 10, stand-alone CM behaves as AL
 302
 303
 304 #
 305 # LB 14  Do not break after OP, even after spaces
 306 #
 307 $OPcm $SP* $CAN_CM $CM*;
 308 $OPcm $SP* $CANT_CM;
 309
 310 $OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
 311
 312 # LB 15
 313 $QUcm $SP* $OPcm;
 314
 315 # LB 16
 316 ($CLcm | $CPcm) $SP* $NScm;
 317
 318 # LB 17
 319 $B2cm $SP* $B2cm;
 320
 321 #
 322 # LB 18  Break after spaces.
 323 #
 324 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
 325 $LB18Breaks    = [$LB8Breaks $SP];
 326
 327
 328 # LB 19
 329 #         x QU
 330 $LB18NonBreaks $CM* $QUcm;
 331 $CM+                $QUcm;
 332
 333 #         QU  x
 334 $QUcm .?;
 335 $QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
 336                               #  TODO:  I don't think this rule is needed.
 337
 338
 339 # LB 20
 340 #        <break>  $CB
 341 #        $CB   <break>
 342
 343 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 344
 345 # LB 21        x   (BA | HY | NS)
 346 #           BB x
 347 #
 348 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
 349
 350 $BBcm [^$CB];                                  #  $BB  x
 351 $BBcm $LB20NonBreaks $CM*;
 352
 353 # LB 21a Don't break after Hebrew + Hyphen
 354 #   HL (HY | BA) x
 355 #
 356 $HLcm ($HYcm | $BAcm) [^$CB]?;
 357
 358 # LB 21b (forward) Don't break between SY and HL
 359 # (break between HL and SY already disallowed by LB 13 above)
 360 $SYcm $HLcm;
 361
 362 # LB 22
 363 ($ALcm | $HLcm) $INcm;
 364 $CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 365 $IDcm    $INcm;
 366 $INcm    $INcm;
 367 $NUcm    $INcm;
 368
 369
 370 # $LB 23
 371 $IDcm  $POcm;
 372 $ALcm  $NUcm;       # includes $LB19
 373 $HLcm  $NUcm;
 374 $CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
 375 $NUcm  $ALcm;
 376 $NUcm  $HLcm;
 377
 378 #
 379 # LB 24
 380 #
 381 $PRcm $IDcm;
 382 $PRcm ($ALcm | $HLcm);
 383 $POcm ($ALcm | $HLcm);
 384
 385 #
 386 # LB 25   Numbers.
 387 #
 388 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
 389
 390 # LB 26  Do not break a Korean syllable
 391 #
 392 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 393 ($JVcm | $H2cm) ($JVcm | $JTcm);
 394 ($JTcm | $H3cm) $JTcm;
 395
 396 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 397 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 398 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 399 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 400
 401
 402 # LB 28   Do not break between alphabetics
 403 #
 404 ($ALcm | $HLcm) ($ALcm | $HLcm);
 405 $CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
 406
 407 # LB 29
 408 $IScm ($ALcm | $HLcm);
 409
 410 # LB 30
 411 ($ALcm | $HLcm | $NUcm) $OPcm;
 412 $CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
 413 $CPcm ($ALcm | $HLcm | $NUcm);
 414
 415 # LB 30a  Do not break between regional indicators.
 416 $RIcm $RIcm;
 417
 418 # Special forward rule for people & body part emoji:
 419 # don't break between relevant emoji and $EmojiMods
 420 $EmojiForMods $EmojiVar? $EmojiMods;
 421
 422 #
 423 #  Reverse Rules.
 424 #
 425 ## -------------------------------------------------
 426
 427 !!reverse;
 428
 429 $CM+ $ALPlus;
 430 $CM+ $BA;
 431 $CM+ $BB;
 432 $CM+ $B2;
 433 $CM+ $CL;
 434 $CM+ $CP;
 435 $CM+ $EX;
 436 $CM+ $GL;
 437 $CM+ $HL;
 438 $CM+ $HY;
 439 $CM+ $H2;
 440 $CM+ $H3;
 441 $CM+ $ID;
 442 $CM+ $IN;
 443 $CM+ $IS;
 444 $CM+ $JL;
 445 $CM+ $JV;
 446 $CM+ $JT;
 447 $CM+ $NS;
 448 $CM+ $NU;
 449 $CM+ $OP;
 450 $CM+ $PO;
 451 $CM+ $PR;
 452 $CM+ $QU;
 453 $CM+ $RI;
 454 $CM+ $SY;
 455 $CM+ $WJ;
 456 $CM+;
 457
 458
 459 #
 460 #  Sequences of the form  (shown forwards)
 461 #      [CANT_CM]  <break>  [CM]  [whatever]
 462 #  The CM needs to behave as an AL
 463 #
 464 $AL_FOLLOW $CM+ / (
 465           [$BK $CR $LF $NL $ZW {eof}] |
 466           $SP+ $CM+ $SP |
 467           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 468                                                #  LB14 says    OP SP* x .
 469                                                #    becomes    OP SP* x AL
 470                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 471                                                #
 472                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 473                                                #                a rule compiler bug which complains about
 474                                                #                empty sets otherwise.
 475
 476 #
 477 #  Sequences of the form  (shown forwards)
 478 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 479 #  The CM needs to behave as an AL
 480 #  This rule is concerned about getting the second of the two <breaks> in place.
 481 #
 482
 483 [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 484
 485
 486
 487 # LB 4, 5, 5
 488
 489 $LB4Breaks [$LB4NonBreaks-$CM];
 490 $LB4Breaks $CM+ $CAN_CM;
 491 $LF $CR;
 492
 493
 494 # LB 7         x SP
 495 #              x ZW
 496 [$SP $ZW] [$LB4NonBreaks-$CM];
 497 [$SP $ZW] $CM+ $CAN_CM;
 498
 499 # LB 8 ZW SP* <break>
 500 #     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
 501 #           Requires an engine enhancement.
 502 #   / $SP* $ZW
 503
 504 # Special reverse rule for people & body part emoji:
 505 # don't break $ZWJ from subsequent $EmojiForSeqs
 506 $EmojiForSeqs $ZWJ $EmojiMods? $EmojiVar? $EmojiForSeqs;
 507
 508 # LB 9,10  Combining marks.
 509 #    X   $CM needs to behave like X, where X is not $SP or controls.
 510 #    $CM not covered by the above needs to behave like $AL
 511 # Stick together any combining sequences that don't match other rules.
 512 $CM+ $CAN_CM;
 513
 514
 515 # LB 11
 516 $CM* $WJ $CM* $CAN_CM;
 517 $CM* $WJ      [$LB8NonBreaks-$CM];
 518
 519      $CANT_CM $CM* $WJ;
 520 $CM* $CAN_CM  $CM* $WJ;
 521
 522 # LB 12a
 523 #      [^SP BA HY] x GL
 524 #
 525 $CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
 526
 527 # LB 12
 528 #     GL  x
 529 #
 530 $CANT_CM $CM* $GL;
 531 $CM* $CAN_CM $CM* $GL;
 532
 533
 534 # LB 13
 535 $CL $CM+ $CAN_CM;
 536 $CP $CM+ $CAN_CM;
 537 $EX $CM+ $CAN_CM;
 538 $IS $CM+ $CAN_CM;
 539 $SY $CM+ $CAN_CM;
 540
 541 $CL [$LB8NonBreaks-$CM];
 542 $CP [$LB8NonBreaks-$CM];
 543 $EX [$LB8NonBreaks-$CM];
 544 $IS [$LB8NonBreaks-$CM];
 545 $SY [$LB8NonBreaks-$CM];
 546
 547 # Rule 13 & 14 taken together for an edge case.
 548 #   Match this, shown forward
 549 #     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
 550 #   This really wants to chain at the $CM+ (which is acting as an $AL)
 551 #   except for $CM chaining being disabled.
 552 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
 553
 554 # LB 14    OP SP* x
 555 #
 556 $CM* $CAN_CM    $SP* $CM* $OP;
 557      $CANT_CM   $SP* $CM* $OP;
 558 $AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
 559
 560      $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
 561 $CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 562 $SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
 563
 564
 565
 566 # LB 15
 567 $CM* $OP $SP* $CM* $QU;
 568
 569 # LB 16
 570 $CM* $NS $SP* $CM* ($CL | $CP);
 571
 572 # LB 17
 573 $CM* $B2 $SP* $CM* $B2;
 574
 575 # LB 18  break after spaces
 576 #        Nothing explicit needed here.
 577
 578
 579 #
 580 # LB 19
 581 #
 582 $CM* $QU $CM* $CAN_CM;                                #   . x QU
 583 $CM* $QU      $LB18NonBreaks;
 584
 585
 586 $CM* $CAN_CM  $CM* $QU;                               #   QU x .
 587      $CANT_CM $CM* $QU;
 588
 589 #
 590 #  LB 20  Break before and after CB.
 591 #         nothing needed here.
 592 #
 593
 594 # LB 21
 595 $CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 596
 597 $CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 598 [^$CB] $CM* $BB;                                      #
 599
 600 # LB21a
 601 [^$CB] $CM* ($HY | $BA) $CM* $HL;
 602
 603 # LB21b (reverse)
 604 $CM* $HL $CM* $SY;
 605
 606 # LB 22
 607 $CM* $IN $CM* ($ALPlus | $HL);
 608 $CM* $IN $CM* $ID;
 609 $CM* $IN $CM* $IN;
 610 $CM* $IN $CM* $NU;
 611
 612 # LB 23
 613 $CM* $PO $CM* $ID;
 614 $CM* $NU $CM* ($ALPlus | $HL);
 615 $CM* ($ALPlus | $HL) $CM* $NU;
 616
 617 # LB 24
 618 $CM* $ID $CM* $PR;
 619 $CM* ($ALPlus | $HL) $CM* $PR;
 620 $CM* ($ALPlus | $HL) $CM* $PO;
 621
 622
 623 # LB 25
 624 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 625
 626 # LB 26
 627 $CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
 628 $CM* ($JT | $JV) $CM* ($H2 | $JV);
 629 $CM* $JT $CM* ($H3 | $JT);
 630
 631 # LB 27
 632 $CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 633 $CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 634 $CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 635
 636 # LB 28
 637 $CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
 638
 639
 640 # LB 29
 641 $CM* ($ALPlus | $HL) $CM* $IS;
 642
 643 # LB 30
 644 $CM* $OP $CM* ($ALPlus | $HL | $NU);
 645 $CM* ($ALPlus | $HL | $NU) $CM* $CP;
 646
 647 # LB 30a
 648 $CM* $RI $CM* $RI;
 649
 650 # Special reverse rule for people & body part emoji:
 651 # don't break between relevant emoji and $EmojiMods
 652 $EmojiMods $EmojiVar? $EmojiForMods;
 653
 654 ## -------------------------------------------------
 655
 656 !!safe_reverse;
 657
 658 # LB 9
 659 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 660 $CM+ $SP / .;
 661
 662 # LB 14
 663 $SP+ $CM* $OP;
 664
 665 # LB 15
 666 $SP+ $CM* $QU;
 667
 668 # LB 16
 669 $SP+ $CM* ($CL | $CP);
 670
 671 # LB 17
 672 $SP+ $CM* $B2;
 673
 674 # LB 21
 675 $CM* ($HY | $BA) $CM* $HL;
 676
 677 # LB 25
 678 ($CM* ($IS | $SY))+ $CM* $NU;
 679 ($CL | $CP) $CM* ($NU | $IS | $SY);
 680
 681 # For dictionary-based break
 682 $dictionary $dictionary;
 683
 684 ## -------------------------------------------------
 685
 686 !!safe_forward;
 687
 688 # Skip forward over all character classes that are involved in
 689 #   rules containing patterns with possibly more than one char
 690 #   of context.
 691 #
 692 #  It might be slightly more efficient to have specific rules
 693 #  instead of one generic one, but only if we could
 694 #  turn off rule chaining.  We don't want to move more
 695 #  than necessary.
 696 #
 697 [$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
 698 $dictionary $dictionary;
 699