icuSources/data/brkitr/line_normal.txt

   1 # Copyright (c) 2002-2015  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 29 for Unicode 6.2
   9 #         http://www.unicode.org/reports/tr14/
  10 #         tailored as noted in 2nd paragraph below..
  11 #
  12 #         TODO:  Rule LB 8 remains as it was in Unicode 5.2
  13 #         This is only because of a limitation of ICU break engine implementation,
  14 #         not because the older behavior is desirable.
  15 #
  16 #         This tailors the line break behavior to correspond to CSS
  17 #         line-break=normal (BCP47 -u-lb-normal) as defined for languages other than
  18 #         Chinese & Japanese.
  19 #         It sets characters of class CJ to behave like ID.
  20
  21 #
  22 #  Character Classes defined by TR 14.
  23 #
  24
  25 !!chain;
  26 !!LBCMNoChain;
  27
  28
  29 !!lookAheadHardBreak;
  30 #
  31 #  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
  32 #                          and only used for the line break rules.
  33 #
  34 #           It is used in the implementation of rule LB 10
  35 #           which says to treat any combining mark that is not attached to a base
  36 #           character as if it were of class AL  (alphabetic).
  37 #
  38 #           The problem occurs in the reverse rules.
  39 #
  40 #           Consider a sequence like, with correct breaks as shown
  41 #               LF  ID  CM  AL  AL
  42 #                  ^       ^       ^
  43 #           Then consider the sequence without the initial ID (ideographic)
  44 #                 LF  CM  AL  AL
  45 #                    ^           ^
  46 #           Our CM, which in the first example was attached to the ideograph,
  47 #           is now unattached, becomes an alpha, and joins in with the other
  48 #           alphas.
  49 #
  50 #           When iterating forwards, these sequences do not present any problems
  51 #           When iterating backwards, we need to look ahead when encountering
  52 #           a CM to see whether it attaches to something further on or not.
  53 #           (Look-ahead in a reverse rule is looking towards the start)
  54 #
  55 #           If the CM is unattached, we need to force a break.
  56 #
  57 #           !!lookAheadHardBreak forces the run time state machine to
  58 #           stop immediately when a look ahead rule ( '/' operator) matches,
  59 #           and set the match position to that of the look-ahead operator,
  60 #           no matter what other rules may be in play at the time.
  61 #
  62 #           See rule LB 19 for an example.
  63 #
  64
  65 $AI = [:LineBreak =  Ambiguous:];
  66 $AL = [:LineBreak =  Alphabetic:];
  67 $BA = [:LineBreak =  Break_After:];
  68 $BB = [:LineBreak =  Break_Before:];
  69 $BK = [:LineBreak =  Mandatory_Break:];
  70 $B2 = [:LineBreak =  Break_Both:];
  71 $CB = [:LineBreak =  Contingent_Break:];
  72 $CJ = [:LineBreak =  Conditional_Japanese_Starter:];
  73 $CL = [:LineBreak =  Close_Punctuation:];
  74 $CM = [:LineBreak =  Combining_Mark:];
  75 $CP = [:LineBreak =  Close_Parenthesis:];
  76 $CR = [:LineBreak =  Carriage_Return:];
  77 $EX = [:LineBreak =  Exclamation:];
  78 $GL = [:LineBreak =  Glue:];
  79 $HL = [:LineBreak =  Hebrew_Letter:];
  80 $HY = [:LineBreak =  Hyphen:];
  81 $H2 = [:LineBreak =  H2:];
  82 $H3 = [:LineBreak =  H3:];
  83 $ID = [[:LineBreak =  Ideographic:] $CJ];
  84 $IN = [:LineBreak =  Inseperable:];
  85 $IS = [:LineBreak =  Infix_Numeric:];
  86 $JL = [:LineBreak =  JL:];
  87 $JV = [:LineBreak =  JV:];
  88 $JT = [:LineBreak =  JT:];
  89 $LF = [:LineBreak =  Line_Feed:];
  90 $NL = [:LineBreak =  Next_Line:];
  91 $NS = [:LineBreak =  Nonstarter:];
  92 $NU = [:LineBreak =  Numeric:];
  93 $OP = [:LineBreak =  Open_Punctuation:];
  94 $PO = [:LineBreak =  Postfix_Numeric:];
  95 $PR = [:LineBreak =  Prefix_Numeric:];
  96 $QU = [:LineBreak =  Quotation:];
  97 $RI = [:LineBreak =  Regional_Indicator:];
  98 $SA = [:LineBreak =  Complex_Context:];
  99 $SG = [:LineBreak =  Surrogate:];
 100 $SP = [:LineBreak =  Space:];
 101 $SY = [:LineBreak =  Break_Symbols:];
 102 $WJ = [:LineBreak =  Word_Joiner:];
 103 $XX = [:LineBreak =  Unknown:];
 104 $ZW = [:LineBreak =  ZWSpace:];
 105
 106 # Special character classes for people & body part emoji:
 107 # Subsets of $CM:
 108 $ZWJ = \u200D;
 109 $EmojiVar = \uFE0F;
 110 # The following are subsets of $ID
 111 $EmojiForSeqs = [\u2764 \U0001F466-\U0001F469 \U0001F48B];
 112 $EmojiForMods = [\u261D \u270A-\u270C \U0001F385 \U0001F3C3-\U0001F3C4 \U0001F3C7 \U0001F3CA \U0001F442-\U0001F443 \U0001F446-\U0001F450 \U0001F466-\U0001F469 \U0001F46E-\U0001F478 \U0001F47C \U0001F481-\U0001F483 \U0001F485-\U0001F487 \U0001F4AA \U0001F596 \U0001F645-\U0001F647 \U0001F64B-\U0001F64F \U0001F6A3 \U0001F6B4-\U0001F6B6 \U0001F6C0];
 113 $EmojiMods = [\U0001F3FB-\U0001F3FF];
 114
 115 #   Dictionary character set, for triggering language-based break engines. Currently
 116 #   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
 117 #   5.0 or later as the definition of Complex_Context was corrected to include all
 118 #   characters requiring dictionary break.
 119
 120 $dictionary = [:LineBreak = Complex_Context:];
 121
 122 #
 123 #  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
 124 #                               SA  (South East Asian: Thai, Lao, Khmer)
 125 #                               SG  (Unpaired Surrogates)
 126 #                               XX  (Unknown, unassigned)
 127 #                         as $AL  (Alphabetic)
 128 #
 129 $ALPlus = [$AL $AI $SA $SG $XX];
 130
 131 #
 132 #  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
 133 #
 134 $ALcm = $ALPlus $CM*;
 135 $BAcm = $BA $CM*;
 136 $BBcm = $BB $CM*;
 137 $B2cm = $B2 $CM*;
 138 $CLcm = $CL $CM*;
 139 $CPcm = $CP $CM*;
 140 $EXcm = $EX $CM*;
 141 $GLcm = $GL $CM*;
 142 $HLcm = $HL $CM*;
 143 $HYcm = $HY $CM*;
 144 $H2cm = $H2 $CM*;
 145 $H3cm = $H3 $CM*;
 146 $IDcm = $ID $CM*;
 147 $INcm = $IN $CM*;
 148 $IScm = $IS $CM*;
 149 $JLcm = $JL $CM*;
 150 $JVcm = $JV $CM*;
 151 $JTcm = $JT $CM*;
 152 $NScm = $NS $CM*;
 153 $NUcm = $NU $CM*;
 154 $OPcm = $OP $CM*;
 155 $POcm = $PO $CM*;
 156 $PRcm = $PR $CM*;
 157 $QUcm = $QU $CM*;
 158 $RIcm = $RI $CM*;
 159 $SYcm = $SY $CM*;
 160 $WJcm = $WJ $CM*;
 161
 162 ## -------------------------------------------------
 163
 164 !!forward;
 165
 166 #
 167 #  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
 168 #
 169 $ALPlus $CM+;
 170 $BA $CM+;
 171 $BB $CM+;
 172 $B2 $CM+;
 173 $CL $CM+;
 174 $CP $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];
 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
 249 # Special forward rule for people & body part emoji:
 250 # don't break $ZWJ from subsequent $EmojiForSeqs
 251 $EmojiForSeqs $EmojiVar? $EmojiMods? $ZWJ $EmojiForSeqs;
 252
 253 # LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
 254 #                                $CM not covered by the above needs to behave like $AL
 255 #                                See definition of $CAN_CM.
 256
 257 $CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
 258 $CM+;
 259
 260 #
 261 # LB 11  Do not break before or after WORD JOINER & related characters.
 262 #
 263 $CAN_CM $CM*  $WJcm;
 264 $LB8NonBreaks $WJcm;
 265 $CM+          $WJcm;
 266
 267 $WJcm $CANT_CM;
 268 $WJcm $CAN_CM $CM*;
 269
 270 #
 271 # LB 12  Do not break after NBSP and related characters.
 272 #         GL  x
 273 #
 274 $GLcm $CAN_CM $CM*;
 275 $GLcm $CANT_CM;
 276
 277 #
 278 # LB 12a  Do not break before NBSP and related characters ...
 279 #            [^SP BA HY] x GL
 280 #
 281 [[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
 282 $CM+ GLcm;
 283
 284
 285
 286 #
 287 # LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
 288 #
 289 $LB8NonBreaks $CL;
 290 $CAN_CM $CM*  $CL;
 291 $CM+          $CL;              # by rule 10, stand-alone CM behaves as AL
 292
 293 $LB8NonBreaks $CP;
 294 $CAN_CM $CM*  $CP;
 295 $CM+          $CP;              # by rule 10, stand-alone CM behaves as AL
 296
 297 $LB8NonBreaks $EX;
 298 $CAN_CM $CM*  $EX;
 299 $CM+          $EX;              # by rule 10, stand-alone CM behaves as AL
 300
 301 $LB8NonBreaks $IS;
 302 $CAN_CM $CM*  $IS;
 303 $CM+          $IS;              # by rule 10, stand-alone CM behaves as AL
 304
 305 $LB8NonBreaks $SY;
 306 $CAN_CM $CM*  $SY;
 307 $CM+          $SY;              # by rule 10, stand-alone CM behaves as AL
 308
 309
 310 #
 311 # LB 14  Do not break after OP, even after spaces
 312 #
 313 $OPcm $SP* $CAN_CM $CM*;
 314 $OPcm $SP* $CANT_CM;
 315
 316 $OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
 317
 318 # LB 15
 319 $QUcm $SP* $OPcm;
 320
 321 # LB 16
 322 ($CLcm | $CPcm) $SP* $NScm;
 323
 324 # LB 17
 325 $B2cm $SP* $B2cm;
 326
 327 #
 328 # LB 18  Break after spaces.
 329 #
 330 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
 331 $LB18Breaks    = [$LB8Breaks $SP];
 332
 333
 334 # LB 19
 335 #         x QU
 336 $LB18NonBreaks $CM* $QUcm;
 337 $CM+                $QUcm;
 338
 339 #         QU  x
 340 $QUcm .?;
 341 $QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
 342                               #  TODO:  I don't think this rule is needed.
 343
 344
 345 # LB 20
 346 #        <break>  $CB
 347 #        $CB   <break>
 348
 349 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 350
 351 # LB 21        x   (BA | HY | NS)
 352 #           BB x
 353 #
 354 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
 355
 356 $BBcm [^$CB];                                  #  $BB  x
 357 $BBcm $LB20NonBreaks $CM*;
 358
 359 # LB 21a Don't break after Hebrew + Hyphen
 360 #   HL (HY | BA) x
 361 #
 362 $HLcm ($HYcm | $BAcm) [^$CB]?;
 363
 364 # LB 21b (forward) Don't break between SY and HL
 365 # (break between HL and SY already disallowed by LB 13 above)
 366 $SYcm $HLcm;
 367
 368 # LB 22
 369 ($ALcm | $HLcm) $INcm;
 370 $CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 371 $IDcm    $INcm;
 372 $INcm    $INcm;
 373 $NUcm    $INcm;
 374
 375
 376 # $LB 23
 377 $IDcm  $POcm;
 378 $ALcm  $NUcm;       # includes $LB19
 379 $HLcm  $NUcm;
 380 $CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
 381 $NUcm  $ALcm;
 382 $NUcm  $HLcm;
 383
 384 #
 385 # LB 24
 386 #
 387 $PRcm $IDcm;
 388 $PRcm ($ALcm | $HLcm);
 389 $POcm ($ALcm | $HLcm);
 390
 391 #
 392 # LB 25   Numbers.
 393 #
 394 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
 395
 396 # LB 26  Do not break a Korean syllable
 397 #
 398 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 399 ($JVcm | $H2cm) ($JVcm | $JTcm);
 400 ($JTcm | $H3cm) $JTcm;
 401
 402 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 403 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 404 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 405 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 406
 407
 408 # LB 28   Do not break between alphabetics
 409 #
 410 ($ALcm | $HLcm) ($ALcm | $HLcm);
 411 $CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
 412
 413 # LB 29
 414 $IScm ($ALcm | $HLcm);
 415
 416 # LB 30
 417 ($ALcm | $HLcm | $NUcm) $OPcm;
 418 $CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
 419 $CPcm ($ALcm | $HLcm | $NUcm);
 420
 421 # LB 30a  Do not break between regional indicators.
 422 $RIcm $RIcm;
 423
 424 # Special forward rule for people & body part emoji:
 425 # don't break between relevant emoji and $EmojiMods
 426 $EmojiForMods $EmojiVar? $EmojiMods;
 427
 428 #
 429 #  Reverse Rules.
 430 #
 431 ## -------------------------------------------------
 432
 433 !!reverse;
 434
 435 $CM+ $ALPlus;
 436 $CM+ $BA;
 437 $CM+ $BB;
 438 $CM+ $B2;
 439 $CM+ $CL;
 440 $CM+ $CP;
 441 $CM+ $EX;
 442 $CM+ $GL;
 443 $CM+ $HL;
 444 $CM+ $HY;
 445 $CM+ $H2;
 446 $CM+ $H3;
 447 $CM+ $ID;
 448 $CM+ $IN;
 449 $CM+ $IS;
 450 $CM+ $JL;
 451 $CM+ $JV;
 452 $CM+ $JT;
 453 $CM+ $NS;
 454 $CM+ $NU;
 455 $CM+ $OP;
 456 $CM+ $PO;
 457 $CM+ $PR;
 458 $CM+ $QU;
 459 $CM+ $RI;
 460 $CM+ $SY;
 461 $CM+ $WJ;
 462 $CM+;
 463
 464
 465 #
 466 #  Sequences of the form  (shown forwards)
 467 #      [CANT_CM]  <break>  [CM]  [whatever]
 468 #  The CM needs to behave as an AL
 469 #
 470 $AL_FOLLOW $CM+ / (
 471           [$BK $CR $LF $NL $ZW {eof}] |
 472           $SP+ $CM+ $SP |
 473           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 474                                                #  LB14 says    OP SP* x .
 475                                                #    becomes    OP SP* x AL
 476                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 477                                                #
 478                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 479                                                #                a rule compiler bug which complains about
 480                                                #                empty sets otherwise.
 481
 482 #
 483 #  Sequences of the form  (shown forwards)
 484 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 485 #  The CM needs to behave as an AL
 486 #  This rule is concerned about getting the second of the two <breaks> in place.
 487 #
 488
 489 [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 490
 491
 492
 493 # LB 4, 5, 5
 494
 495 $LB4Breaks [$LB4NonBreaks-$CM];
 496 $LB4Breaks $CM+ $CAN_CM;
 497 $LF $CR;
 498
 499
 500 # LB 7         x SP
 501 #              x ZW
 502 [$SP $ZW] [$LB4NonBreaks-$CM];
 503 [$SP $ZW] $CM+ $CAN_CM;
 504
 505 # LB 8 ZW SP* <break>
 506 #     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
 507 #           Requires an engine enhancement.
 508 #   / $SP* $ZW
 509
 510 # Special reverse rule for people & body part emoji:
 511 # don't break $ZWJ from subsequent $EmojiForSeqs
 512 $EmojiForSeqs $ZWJ $EmojiMods? $EmojiVar? $EmojiForSeqs;
 513
 514 # LB 9,10  Combining marks.
 515 #    X   $CM needs to behave like X, where X is not $SP or controls.
 516 #    $CM not covered by the above needs to behave like $AL
 517 # Stick together any combining sequences that don't match other rules.
 518 $CM+ $CAN_CM;
 519
 520
 521 # LB 11
 522 $CM* $WJ $CM* $CAN_CM;
 523 $CM* $WJ      [$LB8NonBreaks-$CM];
 524
 525      $CANT_CM $CM* $WJ;
 526 $CM* $CAN_CM  $CM* $WJ;
 527
 528 # LB 12a
 529 #      [^SP BA HY] x GL
 530 #
 531 $CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
 532
 533 # LB 12
 534 #     GL  x
 535 #
 536 $CANT_CM $CM* $GL;
 537 $CM* $CAN_CM $CM* $GL;
 538
 539
 540 # LB 13
 541 $CL $CM+ $CAN_CM;
 542 $CP $CM+ $CAN_CM;
 543 $EX $CM+ $CAN_CM;
 544 $IS $CM+ $CAN_CM;
 545 $SY $CM+ $CAN_CM;
 546
 547 $CL [$LB8NonBreaks-$CM];
 548 $CP [$LB8NonBreaks-$CM];
 549 $EX [$LB8NonBreaks-$CM];
 550 $IS [$LB8NonBreaks-$CM];
 551 $SY [$LB8NonBreaks-$CM];
 552
 553 # Rule 13 & 14 taken together for an edge case.
 554 #   Match this, shown forward
 555 #     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
 556 #   This really wants to chain at the $CM+ (which is acting as an $AL)
 557 #   except for $CM chaining being disabled.
 558 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
 559
 560 # LB 14    OP SP* x
 561 #
 562 $CM* $CAN_CM    $SP* $CM* $OP;
 563      $CANT_CM   $SP* $CM* $OP;
 564 $AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
 565
 566      $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
 567 $CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 568 $SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
 569
 570
 571
 572 # LB 15
 573 $CM* $OP $SP* $CM* $QU;
 574
 575 # LB 16
 576 $CM* $NS $SP* $CM* ($CL | $CP);
 577
 578 # LB 17
 579 $CM* $B2 $SP* $CM* $B2;
 580
 581 # LB 18  break after spaces
 582 #        Nothing explicit needed here.
 583
 584
 585 #
 586 # LB 19
 587 #
 588 $CM* $QU $CM* $CAN_CM;                                #   . x QU
 589 $CM* $QU      $LB18NonBreaks;
 590
 591
 592 $CM* $CAN_CM  $CM* $QU;                               #   QU x .
 593      $CANT_CM $CM* $QU;
 594
 595 #
 596 #  LB 20  Break before and after CB.
 597 #         nothing needed here.
 598 #
 599
 600 # LB 21
 601 $CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 602
 603 $CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 604 [^$CB] $CM* $BB;                                      #
 605
 606 # LB21a
 607 [^$CB] $CM* ($HY | $BA) $CM* $HL;
 608
 609 # LB21b (reverse)
 610 $CM* $HL $CM* $SY;
 611
 612 # LB 22
 613 $CM* $IN $CM* ($ALPlus | $HL);
 614 $CM* $IN $CM* $ID;
 615 $CM* $IN $CM* $IN;
 616 $CM* $IN $CM* $NU;
 617
 618 # LB 23
 619 $CM* $PO $CM* $ID;
 620 $CM* $NU $CM* ($ALPlus | $HL);
 621 $CM* ($ALPlus | $HL) $CM* $NU;
 622
 623 # LB 24
 624 $CM* $ID $CM* $PR;
 625 $CM* ($ALPlus | $HL) $CM* $PR;
 626 $CM* ($ALPlus | $HL) $CM* $PO;
 627
 628
 629 # LB 25
 630 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 631
 632 # LB 26
 633 $CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
 634 $CM* ($JT | $JV) $CM* ($H2 | $JV);
 635 $CM* $JT $CM* ($H3 | $JT);
 636
 637 # LB 27
 638 $CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 639 $CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 640 $CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 641
 642 # LB 28
 643 $CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
 644
 645
 646 # LB 29
 647 $CM* ($ALPlus | $HL) $CM* $IS;
 648
 649 # LB 30
 650 $CM* $OP $CM* ($ALPlus | $HL | $NU);
 651 $CM* ($ALPlus | $HL | $NU) $CM* $CP;
 652
 653 # LB 30a
 654 $CM* $RI $CM* $RI;
 655
 656 # Special reverse rule for people & body part emoji:
 657 # don't break between relevant emoji and $EmojiMods
 658 $EmojiMods $EmojiVar? $EmojiForMods;
 659
 660 ## -------------------------------------------------
 661
 662 !!safe_reverse;
 663
 664 # LB 9
 665 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 666 $CM+ $SP / .;
 667
 668 # LB 14
 669 $SP+ $CM* $OP;
 670
 671 # LB 15
 672 $SP+ $CM* $QU;
 673
 674 # LB 16
 675 $SP+ $CM* ($CL | $CP);
 676
 677 # LB 17
 678 $SP+ $CM* $B2;
 679
 680 # LB 21
 681 $CM* ($HY | $BA) $CM* $HL;
 682
 683 # LB 25
 684 ($CM* ($IS | $SY))+ $CM* $NU;
 685 ($CL | $CP) $CM* ($NU | $IS | $SY);
 686
 687 # For dictionary-based break
 688 $dictionary $dictionary;
 689
 690 ## -------------------------------------------------
 691
 692 !!safe_forward;
 693
 694 # Skip forward over all character classes that are involved in
 695 #   rules containing patterns with possibly more than one char
 696 #   of context.
 697 #
 698 #  It might be slightly more efficient to have specific rules
 699 #  instead of one generic one, but only if we could
 700 #  turn off rule chaining.  We don't want to move more
 701 #  than necessary.
 702 #
 703 [$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
 704 $dictionary $dictionary;
 705