icuSources/data/brkitr/line_loose.txt

   1 # Copyright (c) 2002-2015  International Business Machines Corporation and
   2 # others. All Rights Reserved.
   3 #
   4 #  file:  line_loose.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=loose (BCP47 -u-lb-loose) as defined for languages other than
  18 #         Chinese & Japanese.
  19 #         It sets characters of class CJ to behave like ID.
  20 #         In addition, it allows breaks:
  21 #         * before iteration marks 3005, 303B, 309D, 309E, 30FD, 30FE (all NS)
  22 #         * between characters of LineBreak class IN
  23
  24 #
  25 #  Character Classes defined by TR 14.
  26 #
  27
  28 !!chain;
  29 !!LBCMNoChain;
  30
  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 $AI = [:LineBreak =  Ambiguous:];
  69 $AL = [:LineBreak =  Alphabetic:];
  70 $BA = [:LineBreak =  Break_After:];
  71 $BB = [:LineBreak =  Break_Before:];
  72 $BK = [:LineBreak =  Mandatory_Break:];
  73 $B2 = [:LineBreak =  Break_Both:];
  74 $CB = [:LineBreak =  Contingent_Break:];
  75 $CJ = [:LineBreak =  Conditional_Japanese_Starter:];
  76 $CL = [:LineBreak =  Close_Punctuation:];
  77 $CM = [:LineBreak =  Combining_Mark:];
  78 $CP = [:LineBreak =  Close_Parenthesis:];
  79 $CR = [:LineBreak =  Carriage_Return:];
  80 $EX = [:LineBreak =  Exclamation:];
  81 $GL = [:LineBreak =  Glue:];
  82 $HL = [:LineBreak =  Hebrew_Letter:];
  83 $HY = [:LineBreak =  Hyphen:];
  84 $H2 = [:LineBreak =  H2:];
  85 $H3 = [:LineBreak =  H3:];
  86 $ID = [[:LineBreak =  Ideographic:] $CJ];
  87 $IN = [:LineBreak =  Inseperable:];
  88 $IS = [:LineBreak =  Infix_Numeric:];
  89 $JL = [:LineBreak =  JL:];
  90 $JV = [:LineBreak =  JV:];
  91 $JT = [:LineBreak =  JT:];
  92 $LF = [:LineBreak =  Line_Feed:];
  93 $NL = [:LineBreak =  Next_Line:];
  94 $NSX = [\u3005 \u303B \u309D \u309E \u30FD \u30FE];
  95 $NS = [[:LineBreak =  Nonstarter:] - $NSX];
  96 $NU = [:LineBreak =  Numeric:];
  97 $OP = [:LineBreak =  Open_Punctuation:];
  98 $PO = [:LineBreak =  Postfix_Numeric:];
  99 $PR = [:LineBreak =  Prefix_Numeric:];
 100 $QU = [:LineBreak =  Quotation:];
 101 $RI = [:LineBreak =  Regional_Indicator:];
 102 $SA = [:LineBreak =  Complex_Context:];
 103 $SG = [:LineBreak =  Surrogate:];
 104 $SP = [:LineBreak =  Space:];
 105 $SY = [:LineBreak =  Break_Symbols:];
 106 $WJ = [:LineBreak =  Word_Joiner:];
 107 $XX = [:LineBreak =  Unknown:];
 108 $ZW = [:LineBreak =  ZWSpace:];
 109
 110 # Special character classes for people & body part emoji:
 111 # Subsets of $CM:
 112 $ZWJ = \u200D;
 113 $EmojiVar = \uFE0F;
 114 # The following are subsets of $ID
 115 $EmojiForSeqs = [\u2764 \U0001F466-\U0001F469 \U0001F48B];
 116 $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];
 117 $EmojiMods = [\U0001F3FB-\U0001F3FF];
 118
 119 #   Dictionary character set, for triggering language-based break engines. Currently
 120 #   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
 121 #   5.0 or later as the definition of Complex_Context was corrected to include all
 122 #   characters requiring dictionary break.
 123
 124 $dictionary = [:LineBreak = Complex_Context:];
 125
 126 #
 127 #  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
 128 #                               SA  (South East Asian: Thai, Lao, Khmer)
 129 #                               SG  (Unpaired Surrogates)
 130 #                               XX  (Unknown, unassigned)
 131 #                         as $AL  (Alphabetic)
 132 #
 133 $ALPlus = [$AL $AI $SA $SG $XX];
 134
 135 #
 136 #  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
 137 #
 138 $ALcm = $ALPlus $CM*;
 139 $BAcm = $BA $CM*;
 140 $BBcm = $BB $CM*;
 141 $B2cm = $B2 $CM*;
 142 $CLcm = $CL $CM*;
 143 $CPcm = $CP $CM*;
 144 $EXcm = $EX $CM*;
 145 $GLcm = $GL $CM*;
 146 $HLcm = $HL $CM*;
 147 $HYcm = $HY $CM*;
 148 $H2cm = $H2 $CM*;
 149 $H3cm = $H3 $CM*;
 150 $IDcm = $ID $CM*;
 151 $INcm = $IN $CM*;
 152 $IScm = $IS $CM*;
 153 $JLcm = $JL $CM*;
 154 $JVcm = $JV $CM*;
 155 $JTcm = $JT $CM*;
 156 $NScm = $NS $CM*;
 157 $NSXcm = $NSX $CM*;
 158 $NUcm = $NU $CM*;
 159 $OPcm = $OP $CM*;
 160 $POcm = $PO $CM*;
 161 $PRcm = $PR $CM*;
 162 $QUcm = $QU $CM*;
 163 $RIcm = $RI $CM*;
 164 $SYcm = $SY $CM*;
 165 $WJcm = $WJ $CM*;
 166
 167 ## -------------------------------------------------
 168
 169 !!forward;
 170
 171 #
 172 #  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
 173 #
 174 $ALPlus $CM+;
 175 $BA $CM+;
 176 $BB $CM+;
 177 $B2 $CM+;
 178 $CL $CM+;
 179 $CP $CM+;
 180 $EX $CM+;
 181 $GL $CM+;
 182 $HL $CM+;
 183 $HY $CM+;
 184 $H2 $CM+;
 185 $H3 $CM+;
 186 $ID $CM+;
 187 $IN $CM+;
 188 $IS $CM+;
 189 $JL $CM+;
 190 $JV $CM+;
 191 $JT $CM+;
 192 $NS $CM+;
 193 $NSX $CM+;
 194 $NU $CM+;
 195 $OP $CM+;
 196 $PO $CM+;
 197 $PR $CM+;
 198 $QU $CM+;
 199 $RI $CM+;
 200 $SY $CM+;
 201 $WJ $CM+;
 202
 203 #
 204 # CAN_CM  is the set of characters that may combine with CM combining chars.
 205 #         Note that Linebreak UAX 14's concept of a combining char and the rules
 206 #         for what they can combine with are _very_ different from the rest of Unicode.
 207 #
 208 #         Note that $CM itself is left out of this set.  If CM is needed as a base
 209 #         it must be listed separately in the rule.
 210 #
 211 $CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
 212 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
 213
 214 #
 215 # AL_FOLLOW  set of chars that can unconditionally follow an AL
 216 #            Needed in rules where stand-alone $CM s are treated as AL.
 217 #            Chaining is disabled with CM because it causes other failures,
 218 #            so for this one case we need to manually list out longer sequences.
 219 #
 220 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
 221 $AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $NSX $IN $NU $ALPlus];
 222 $AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
 223
 224
 225 #
 226 #  Rule LB 4, 5    Mandatory (Hard) breaks.
 227 #
 228 $LB4Breaks    = [$BK $CR $LF $NL];
 229 $LB4NonBreaks = [^$BK $CR $LF $NL];
 230 $CR $LF {100};
 231
 232 #
 233 #  LB 6    Do not break before hard line breaks.
 234 #
 235 $LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
 236 $CAN_CM $CM*    $LB4Breaks {100};
 237 $CM+            $LB4Breaks {100};
 238
 239 # LB 7         x SP
 240 #              x ZW
 241 $LB4NonBreaks [$SP $ZW];
 242 $CAN_CM $CM*  [$SP $ZW];
 243 $CM+          [$SP $ZW];
 244
 245 #
 246 # LB 8         Break after zero width space
 247 #              TODO:  ZW SP* <break>
 248 #              An engine change is required to write the reverse rule for this.
 249 #              For now, leave the Unicode 5.2 rule, ZW <break>
 250 #
 251 $LB8Breaks    = [$LB4Breaks $ZW];
 252 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
 253
 254
 255 # Special forward rule for people & body part emoji:
 256 # don't break $ZWJ from subsequent $EmojiForSeqs
 257 $EmojiForSeqs $EmojiVar? $EmojiMods? $ZWJ $EmojiForSeqs;
 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 $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 $QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
 350                               #  TODO:  I don't think this rule is needed.
 351
 352
 353 # LB 20
 354 #        <break>  $CB
 355 #        $CB   <break>
 356
 357 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 358
 359 # LB 21        x   (BA | HY | NS)
 360 #           BB x
 361 #
 362 # DO allow breaks here before NSXcm, so don't include it
 363 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
 364
 365 $BBcm [^$CB];                                  #  $BB  x
 366 $BBcm $LB20NonBreaks $CM*;
 367
 368 # LB 21a Don't break after Hebrew + Hyphen
 369 #   HL (HY | BA) x
 370 #
 371 $HLcm ($HYcm | $BAcm) [^$CB]?;
 372
 373 # LB 21b (forward) Don't break between SY and HL
 374 # (break between HL and SY already disallowed by LB 13 above)
 375 $SYcm $HLcm;
 376
 377 # LB 22
 378 ($ALcm | $HLcm) $INcm;
 379 $CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 380 $IDcm    $INcm;
 381 # $INcm  $INcm; # delete this rule for CSS loose
 382 $NUcm    $INcm;
 383
 384
 385 # $LB 23
 386 $IDcm  $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 $IDcm;
 397 $PRcm ($ALcm | $HLcm);
 398 $POcm ($ALcm | $HLcm);
 399
 400 #
 401 # LB 25   Numbers.
 402 #
 403 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
 404
 405 # LB 26  Do not break a Korean syllable
 406 #
 407 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 408 ($JVcm | $H2cm) ($JVcm | $JTcm);
 409 ($JTcm | $H3cm) $JTcm;
 410
 411 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 412 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 413 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 414 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 415
 416
 417 # LB 28   Do not break between alphabetics
 418 #
 419 ($ALcm | $HLcm) ($ALcm | $HLcm);
 420 $CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
 421
 422 # LB 29
 423 $IScm ($ALcm | $HLcm);
 424
 425 # LB 30
 426 ($ALcm | $HLcm | $NUcm) $OPcm;
 427 $CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
 428 $CPcm ($ALcm | $HLcm | $NUcm);
 429
 430 # LB 30a  Do not break between regional indicators.
 431 $RIcm $RIcm;
 432
 433 # Special forward rule for people & body part emoji:
 434 # don't break between relevant emoji and $EmojiMods
 435 $EmojiForMods $EmojiVar? $EmojiMods;
 436
 437 #
 438 #  Reverse Rules.
 439 #
 440 ## -------------------------------------------------
 441
 442 !!reverse;
 443
 444 $CM+ $ALPlus;
 445 $CM+ $BA;
 446 $CM+ $BB;
 447 $CM+ $B2;
 448 $CM+ $CL;
 449 $CM+ $CP;
 450 $CM+ $EX;
 451 $CM+ $GL;
 452 $CM+ $HL;
 453 $CM+ $HY;
 454 $CM+ $H2;
 455 $CM+ $H3;
 456 $CM+ $ID;
 457 $CM+ $IN;
 458 $CM+ $IS;
 459 $CM+ $JL;
 460 $CM+ $JV;
 461 $CM+ $JT;
 462 $CM+ $NS;
 463 $CM+ $NSX;
 464 $CM+ $NU;
 465 $CM+ $OP;
 466 $CM+ $PO;
 467 $CM+ $PR;
 468 $CM+ $QU;
 469 $CM+ $RI;
 470 $CM+ $SY;
 471 $CM+ $WJ;
 472 $CM+;
 473
 474
 475 #
 476 #  Sequences of the form  (shown forwards)
 477 #      [CANT_CM]  <break>  [CM]  [whatever]
 478 #  The CM needs to behave as an AL
 479 #
 480 $AL_FOLLOW $CM+ / (
 481           [$BK $CR $LF $NL $ZW {eof}] |
 482           $SP+ $CM+ $SP |
 483           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 484                                                #  LB14 says    OP SP* x .
 485                                                #    becomes    OP SP* x AL
 486                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 487                                                #
 488                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 489                                                #                a rule compiler bug which complains about
 490                                                #                empty sets otherwise.
 491
 492 #
 493 #  Sequences of the form  (shown forwards)
 494 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 495 #  The CM needs to behave as an AL
 496 #  This rule is concerned about getting the second of the two <breaks> in place.
 497 #
 498
 499 [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 500
 501
 502
 503 # LB 4, 5, 5
 504
 505 $LB4Breaks [$LB4NonBreaks-$CM];
 506 $LB4Breaks $CM+ $CAN_CM;
 507 $LF $CR;
 508
 509
 510 # LB 7         x SP
 511 #              x ZW
 512 [$SP $ZW] [$LB4NonBreaks-$CM];
 513 [$SP $ZW] $CM+ $CAN_CM;
 514
 515 # LB 8 ZW SP* <break>
 516 #     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
 517 #           Requires an engine enhancement.
 518 #   / $SP* $ZW
 519
 520 # Special reverse rule for people & body part emoji:
 521 # don't break $ZWJ from subsequent $EmojiForSeqs
 522 $EmojiForSeqs $ZWJ $EmojiMods? $EmojiVar? $EmojiForSeqs;
 523
 524 # LB 9,10  Combining marks.
 525 #    X   $CM needs to behave like X, where X is not $SP or controls.
 526 #    $CM not covered by the above needs to behave like $AL
 527 # Stick together any combining sequences that don't match other rules.
 528 $CM+ $CAN_CM;
 529
 530
 531 # LB 11
 532 $CM* $WJ $CM* $CAN_CM;
 533 $CM* $WJ      [$LB8NonBreaks-$CM];
 534
 535      $CANT_CM $CM* $WJ;
 536 $CM* $CAN_CM  $CM* $WJ;
 537
 538 # LB 12a
 539 #      [^SP BA HY] x GL
 540 #
 541 $CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
 542
 543 # LB 12
 544 #     GL  x
 545 #
 546 $CANT_CM $CM* $GL;
 547 $CM* $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 $CM* $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 $CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 578 $SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
 579
 580
 581
 582 # LB 15
 583 $CM* $OP $SP* $CM* $QU;
 584
 585 # LB 16
 586 # Don't include $NSX here
 587 $CM* $NS $SP* $CM* ($CL | $CP);
 588
 589 # LB 17
 590 $CM* $B2 $SP* $CM* $B2;
 591
 592 # LB 18  break after spaces
 593 #        Nothing explicit needed here.
 594
 595
 596 #
 597 # LB 19
 598 #
 599 $CM* $QU $CM* $CAN_CM;                                #   . x QU
 600 $CM* $QU      $LB18NonBreaks;
 601
 602
 603 $CM* $CAN_CM  $CM* $QU;                               #   QU x .
 604      $CANT_CM $CM* $QU;
 605
 606 #
 607 #  LB 20  Break before and after CB.
 608 #         nothing needed here.
 609 #
 610
 611 # LB 21
 612 # Don't include $NSX here
 613 $CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 614
 615 $CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 616 [^$CB] $CM* $BB;                                      #
 617
 618 # LB21a
 619 [^$CB] $CM* ($HY | $BA) $CM* $HL;
 620
 621 # LB21b (reverse)
 622 $CM* $HL $CM* $SY;
 623
 624 # LB 22
 625 $CM* $IN $CM* ($ALPlus | $HL);
 626 $CM* $IN $CM* $ID;
 627 # $CM* $IN $CM* $IN; # delete this rule for CSS loose
 628 $CM* $IN $CM* $NU;
 629
 630 # LB 23
 631 $CM* $PO $CM* $ID;
 632 $CM* $NU $CM* ($ALPlus | $HL);
 633 $CM* ($ALPlus | $HL) $CM* $NU;
 634
 635 # LB 24
 636 $CM* $ID $CM* $PR;
 637 $CM* ($ALPlus | $HL) $CM* $PR;
 638 $CM* ($ALPlus | $HL) $CM* $PO;
 639
 640
 641 # LB 25
 642 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 643
 644 # LB 26
 645 $CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
 646 $CM* ($JT | $JV) $CM* ($H2 | $JV);
 647 $CM* $JT $CM* ($H3 | $JT);
 648
 649 # LB 27
 650 $CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 651 $CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 652 $CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 653
 654 # LB 28
 655 $CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
 656
 657
 658 # LB 29
 659 $CM* ($ALPlus | $HL) $CM* $IS;
 660
 661 # LB 30
 662 $CM* $OP $CM* ($ALPlus | $HL | $NU);
 663 $CM* ($ALPlus | $HL | $NU) $CM* $CP;
 664
 665 # LB 30a
 666 $CM* $RI $CM* $RI;
 667
 668 # Special reverse rule for people & body part emoji:
 669 # don't break between relevant emoji and $EmojiMods
 670 $EmojiMods $EmojiVar? $EmojiForMods;
 671
 672 ## -------------------------------------------------
 673
 674 !!safe_reverse;
 675
 676 # LB 9
 677 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 678 $CM+ $SP / .;
 679
 680 # LB 14
 681 $SP+ $CM* $OP;
 682
 683 # LB 15
 684 $SP+ $CM* $QU;
 685
 686 # LB 16
 687 $SP+ $CM* ($CL | $CP);
 688
 689 # LB 17
 690 $SP+ $CM* $B2;
 691
 692 # LB 21
 693 $CM* ($HY | $BA) $CM* $HL;
 694
 695 # LB 25
 696 ($CM* ($IS | $SY))+ $CM* $NU;
 697 ($CL | $CP) $CM* ($NU | $IS | $SY);
 698
 699 # For dictionary-based break
 700 $dictionary $dictionary;
 701
 702 ## -------------------------------------------------
 703
 704 !!safe_forward;
 705
 706 # Skip forward over all character classes that are involved in
 707 #   rules containing patterns with possibly more than one char
 708 #   of context.
 709 #
 710 #  It might be slightly more efficient to have specific rules
 711 #  instead of one generic one, but only if we could
 712 #  turn off rule chaining.  We don't want to move more
 713 #  than necessary.
 714 #
 715 [$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
 716 $dictionary $dictionary;
 717