icuSources/data/brkitr/rules/line_normal_fi.txt

   1 # Copyright (c) 2002-2016  International Business Machines Corporation and
   2 # others. All Rights Reserved.
   3 #
   4 #  file:  line_normal_fi.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 for Finnish, and 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:]-[\u2010]];
  75 $HH = [\u2010];
  76 $BB = [:LineBreak =  Break_Before:];
  77 $BK = [:LineBreak =  Mandatory_Break:];
  78 $B2 = [:LineBreak =  Break_Both:];
  79 $CB = [:LineBreak =  Contingent_Break:];
  80 $CJ = [:LineBreak =  Conditional_Japanese_Starter:];
  81 $CL = [:LineBreak =  Close_Punctuation:];
  82 $CM = [[:LineBreak =  Combining_Mark:] \u200d];
  83 $CP = [:LineBreak =  Close_Parenthesis:];
  84 $CR = [:LineBreak =  Carriage_Return:];
  85 $EX = [:LineBreak =  Exclamation:];
  86 $GL = [:LineBreak =  Glue:];
  87 $HL = [:LineBreak =  Hebrew_Letter:];
  88 $HY = [:LineBreak =  Hyphen:];
  89 $H2 = [:LineBreak =  H2:];
  90 $H3 = [:LineBreak =  H3:];
  91 $ID = [[:LineBreak =  Ideographic:] $CJ [\u2640\u2642\u2695\u2696\u2764] - $EB];
  92 $IN = [:LineBreak =  Inseperable:];
  93 $IS = [:LineBreak =  Infix_Numeric:];
  94 $JL = [:LineBreak =  JL:];
  95 $JV = [:LineBreak =  JV:];
  96 $JT = [:LineBreak =  JT:];
  97 $LF = [:LineBreak =  Line_Feed:];
  98 $NL = [:LineBreak =  Next_Line:];
  99 $NS = [:LineBreak =  Nonstarter:];
 100 $NU = [:LineBreak =  Numeric:];
 101 $OP = [:LineBreak =  Open_Punctuation:];
 102 $PO = [:LineBreak =  Postfix_Numeric:];
 103 $PR = [:LineBreak =  Prefix_Numeric:];
 104 $QU = [:LineBreak =  Quotation:];
 105 $RI = [:LineBreak =  Regional_Indicator:];
 106 $SA = [:LineBreak =  Complex_Context:];
 107 $SG = [:LineBreak =  Surrogate:];
 108 $SP = [:LineBreak =  Space:];
 109 $SY = [:LineBreak =  Break_Symbols:];
 110 $WJ = [:LineBreak =  Word_Joiner:];
 111 $XX = [:LineBreak =  Unknown:];
 112 $ZW = [:LineBreak =  ZWSpace:];
 113 $ZWJ = [\u200d];
 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 $HHcm = $HH $CM*;
 137 $BBcm = $BB $CM*;
 138 $B2cm = $B2 $CM*;
 139 $CLcm = $CL $CM*;
 140 $CPcm = $CP $CM*;
 141 $EXcm = $EX $CM*;
 142 $GLcm = $GL $CM*;
 143 $HLcm = $HL $CM*;
 144 $HYcm = $HY $CM*;
 145 $H2cm = $H2 $CM*;
 146 $H3cm = $H3 $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 $HH $CM+;
 172 $BB $CM+;
 173 $B2 $CM+;
 174 $CL $CM+;
 175 $CP $CM+;
 176 $EB $CM+;
 177 $EM $CM+;
 178 $EX $CM+;
 179 $GL $CM+;
 180 $HL $CM+;
 181 $HY $CM+;
 182 $H2 $CM+;
 183 $H3 $CM+;
 184 $ID $CM+;
 185 $IN $CM+;
 186 $IS $CM+;
 187 $JL $CM+;
 188 $JV $CM+;
 189 $JT $CM+;
 190 $NS $CM+;
 191 $NU $CM+;
 192 $OP $CM+;
 193 $PO $CM+;
 194 $PR $CM+;
 195 $QU $CM+;
 196 $RI $CM+;
 197 $SY $CM+;
 198 $WJ $CM+;
 199
 200 #
 201 # CAN_CM  is the set of characters that may combine with CM combining chars.
 202 #         Note that Linebreak UAX 14's concept of a combining char and the rules
 203 #         for what they can combine with are _very_ different from the rest of Unicode.
 204 #
 205 #         Note that $CM itself is left out of this set.  If CM is needed as a base
 206 #         it must be listed separately in the rule.
 207 #
 208 $CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
 209 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
 210
 211 #
 212 # AL_FOLLOW  set of chars that can unconditionally follow an AL
 213 #            Needed in rules where stand-alone $CM s are treated as AL.
 214 #            Chaining is disabled with CM because it causes other failures,
 215 #            so for this one case we need to manually list out longer sequences.
 216 #
 217 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
 218 $AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HH $HY $NS $IN $NU $ALPlus];
 219 $AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
 220
 221
 222 #
 223 #  Rule LB 4, 5    Mandatory (Hard) breaks.
 224 #
 225 $LB4Breaks    = [$BK $CR $LF $NL];
 226 $LB4NonBreaks = [^$BK $CR $LF $NL $CM];
 227 $CR $LF {100};
 228
 229 #
 230 #  LB 6    Do not break before hard line breaks.
 231 #
 232 $LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
 233 $CAN_CM $CM*    $LB4Breaks {100};
 234 ^$CM+           $LB4Breaks {100};
 235
 236 # LB 7         x SP
 237 #              x ZW
 238 $LB4NonBreaks [$SP $ZW];
 239 $CAN_CM $CM*  [$SP $ZW];
 240 ^$CM+         [$SP $ZW];
 241
 242 #
 243 # LB 8         Break after zero width space
 244 #              TODO:  ZW SP* <break>
 245 #              An engine change is required to write the reverse rule for this.
 246 #              For now, leave the Unicode 5.2 rule, ZW <break>
 247 #
 248 $LB8Breaks    = [$LB4Breaks $ZW];
 249 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
 250
 251 # LB 8a        ZWJ x ID    Emoji proposal.
 252 #
 253 $ZWJ ($ID | $EB | $EM);
 254
 255 # LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
 256 #                                $CM not covered by the above needs to behave like $AL
 257 #                                See definition of $CAN_CM.
 258
 259 $CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
 260 ^$CM+;
 261
 262 #
 263 # LB 11  Do not break before or after WORD JOINER & related characters.
 264 #
 265 $CAN_CM $CM*  $WJcm;
 266 $LB8NonBreaks $WJcm;
 267 ^$CM+         $WJcm;
 268
 269 $WJcm $CANT_CM;
 270 $WJcm $CAN_CM $CM*;
 271
 272 #
 273 # LB 12  Do not break after NBSP and related characters.
 274 #         GL  x
 275 #
 276 $GLcm $CAN_CM $CM*;
 277 $GLcm $CANT_CM;
 278
 279 #
 280 # LB 12a  Do not break before NBSP and related characters ...
 281 #            [^SP BA HY] x GL
 282 #
 283 [[$LB8NonBreaks] - [$SP $BA $HH $HY]] $CM* $GLcm;
 284 ^$CM+ $GLcm;
 285
 286
 287
 288 #
 289 # LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
 290 #
 291 $LB8NonBreaks $CL;
 292 $CAN_CM $CM*  $CL;
 293 ^$CM+         $CL;              # by rule 10, stand-alone CM behaves as AL
 294
 295 $LB8NonBreaks $CP;
 296 $CAN_CM $CM*  $CP;
 297 ^$CM+         $CP;              # by rule 10, stand-alone CM behaves as AL
 298
 299 $LB8NonBreaks $EX;
 300 $CAN_CM $CM*  $EX;
 301 ^$CM+         $EX;              # by rule 10, stand-alone CM behaves as AL
 302
 303 $LB8NonBreaks $IS;
 304 $CAN_CM $CM*  $IS;
 305 ^$CM+         $IS;              # by rule 10, stand-alone CM behaves as AL
 306
 307 $LB8NonBreaks $SY;
 308 $CAN_CM $CM*  $SY;
 309 ^$CM+         $SY;              # by rule 10, stand-alone CM behaves as AL
 310
 311
 312 #
 313 # LB 14  Do not break after OP, even after spaces
 314 #
 315 $OPcm $SP* $CAN_CM $CM*;
 316 $OPcm $SP* $CANT_CM;
 317
 318 $OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
 319
 320 # LB 15
 321 $QUcm $SP* $OPcm;
 322
 323 # LB 16
 324 ($CLcm | $CPcm) $SP* $NScm;
 325
 326 # LB 17
 327 $B2cm $SP* $B2cm;
 328
 329 #
 330 # LB 18  Break after spaces.
 331 #
 332 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
 333 $LB18Breaks    = [$LB8Breaks $SP];
 334
 335
 336 # LB 19
 337 #         x QU
 338 $LB18NonBreaks $CM* $QUcm;
 339 ^$CM+               $QUcm;
 340
 341 #         QU  x
 342 $QUcm .?;
 343
 344
 345 # LB 20
 346 #        <break>  $CB
 347 #        $CB   <break>
 348
 349 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 350
 351 # LB 20.09 added rule for Finnish tailoring
 352 # LB 21        x   (BA | HY | NS)
 353 #           BB x
 354 #
 355 $LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm) / $AL;
 356 $LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm);
 357 ($HY | $HH) $AL;
 358 ^$CM+ ($BAcm | $HHcm | $HYcm | $NScm);
 359
 360 $BBcm [^$CB];                                  #  $BB  x
 361 $BBcm $LB20NonBreaks $CM*;
 362
 363 # LB 21a Don't break after Hebrew + Hyphen
 364 #   HL (HY | BA) x
 365 #
 366 $HLcm ($HYcm | $BAcm | $HHcm) [^$CB]?;
 367
 368 # LB 21b (forward) Don't break between SY and HL
 369 # (break between HL and SY already disallowed by LB 13 above)
 370 $SYcm $HLcm;
 371
 372 # LB 22
 373 ($ALcm | $HLcm) $INcm;
 374 ^$CM+    $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 375 $EXcm    $INcm;
 376 ($ID | $EB | $EM) $CM*  $INcm;
 377 $INcm    $INcm;
 378 $NUcm    $INcm;
 379
 380
 381 # $LB 23
 382 ($ID | $EB | $EM) $CM*  $POcm;
 383 $ALcm  $NUcm;       # includes $LB19
 384 $HLcm  $NUcm;
 385 ^$CM+  $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
 386 $NUcm  $ALcm;
 387 $NUcm  $HLcm;
 388
 389 #
 390 # LB 24
 391 #
 392 $PRcm ($ID | $EB | $EM);
 393 $PRcm ($ALcm | $HLcm);
 394 $POcm ($ALcm | $HLcm);
 395 ($ALcm | $HLcm) ($PRcm | $POcm);
 396 ^$CM+ ($PRcm | $POcm);       # Rule 10, any otherwise unattached CM behaves as AL
 397
 398 #
 399 # LB 25   Numbers.
 400 #
 401 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
 402
 403 # LB 26  Do not break a Korean syllable
 404 #
 405 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 406 ($JVcm | $H2cm) ($JVcm | $JTcm);
 407 ($JTcm | $H3cm) $JTcm;
 408
 409 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 410 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 411 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 412 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 413
 414
 415 # LB 28   Do not break between alphabetics
 416 #
 417 ($ALcm | $HLcm) ($ALcm | $HLcm);
 418 ^$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
 419
 420 # LB 29
 421 $IScm ($ALcm | $HLcm);
 422
 423 # LB 30
 424 ($ALcm | $HLcm | $NUcm) $OPcm;
 425 ^$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
 426 $CPcm ($ALcm | $HLcm | $NUcm);
 427
 428 # LB 30a  Do not break between regional indicators. Break after pairs of them.
 429 #         Tricky interaction with LB8a: ZWJ x ID
 430 $RIcm $RI                / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HH $HY $NS $CM] {eof}];
 431 $RIcm $RI $CM*  $ZWJ     / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HH $HY $NS $CM $ID $EB $EM] {eof}];
 432 $RIcm $RI $CM* [$CM-$ZWJ] / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HH $HY $NS $CM] {eof}];
 433
 434 $RIcm $RIcm [$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HH $HY $NS {eof}];
 435 $RIcm $RIcm $ZWJ ($ID | $EB | $EM);
 436
 437 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 438 $EB $CM* $EM;
 439
 440 #
 441 #  Reverse Rules.
 442 #
 443 ## -------------------------------------------------
 444
 445 !!reverse;
 446
 447 ^$CM+ $ALPlus;
 448 ^$CM+ $BA;
 449 ^$CM+ $HH;
 450 ^$CM+ $BB;
 451 ^$CM+ $B2;
 452 ^$CM+ $CL;
 453 ^$CM+ $CP;
 454 ^$CM+ $EB;
 455 ^$CM+ $EM;
 456 ^$CM+ $EX;
 457 ^$CM+ $GL;
 458 ^$CM+ $HL;
 459 ^$CM+ $HY;
 460 ^$CM+ $H2;
 461 ^$CM+ $H3;
 462 ^$CM+ $ID;
 463 ^$CM+ $IN;
 464 ^$CM+ $IS;
 465 ^$CM+ $JL;
 466 ^$CM+ $JV;
 467 ^$CM+ $JT;
 468 ^$CM+ $NS;
 469 ^$CM+ $NU;
 470 ^$CM+ $OP;
 471 ^$CM+ $PO;
 472 ^$CM+ $PR;
 473 ^$CM+ $QU;
 474 ^$CM+ $RI;
 475 ^$CM+ $SY;
 476 ^$CM+ $WJ;
 477 ^$CM+;
 478
 479
 480 #
 481 #  Sequences of the form  (shown forwards)
 482 #      [CANT_CM]  <break>  [CM]  [whatever]
 483 #  The CM needs to behave as an AL
 484 #
 485 $AL_FOLLOW $CM+ / (
 486           [$BK $CR $LF $NL $ZW {eof}] |
 487           $SP+ $CM+ $SP |
 488           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 489                                                #  LB14 says    OP SP* x .
 490                                                #    becomes    OP SP* x AL
 491                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 492                                                #
 493                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 494                                                #                a rule compiler bug which complains about
 495                                                #                empty sets otherwise.
 496
 497 #
 498 #  Sequences of the form  (shown forwards)
 499 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 500 #  The CM needs to behave as an AL
 501 #  This rule is concerned about getting the second of the two <breaks> in place.
 502 #
 503
 504 # Apple early addition, remove this, superseded by LB24
 505 # [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 506
 507
 508
 509 # LB 4, 5, 6
 510
 511 $LB4Breaks [$LB4NonBreaks-$CM];
 512 $LB4Breaks $CM+ $CAN_CM;
 513 $LF $CR;
 514
 515
 516 # LB 7         x SP
 517 #              x ZW
 518 [$SP $ZW] [$LB4NonBreaks-$CM];
 519 [$SP $ZW] $CM+ $CAN_CM;
 520
 521 # LB 8 ZW SP* <break>
 522 #     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
 523 #           Requires an engine enhancement.
 524 #   / $SP* $ZW
 525
 526 # LB 8a        ZWJ x ID    Unicode Emoji proposal L2/16-011R3
 527 #                          The ZWJ will look like a CM to whatever precedes it.
 528 #
 529 ($ID | $EB | $EM) $ZWJ $CM* $CAN_CM?;
 530
 531
 532 # LB 9,10  Combining marks.
 533 #    X   $CM needs to behave like X, where X is not $SP or controls.
 534 #    $CM not covered by the above needs to behave like $AL
 535 # Stick together any combining sequences that don't match other rules.
 536 ^$CM+ $CAN_CM;
 537
 538
 539 # LB 11
 540 #
 541 $WJ $CM* $CAN_CM;
 542 $WJ      [$LB8NonBreaks-$CM];
 543
 544      $CANT_CM $CM* $WJ;
 545 $CAN_CM  $CM* $WJ;
 546
 547 # LB 12a
 548 #      [^SP BA HY] x GL
 549 #
 550 $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HH $HY]];
 551
 552 # LB 12
 553 #     GL  x
 554 #
 555 $CANT_CM $CM* $GL;
 556 $CAN_CM $CM* $GL;
 557
 558
 559 # LB 13
 560 $CL $CM+ $CAN_CM;
 561 $CP $CM+ $CAN_CM;
 562 $EX $CM+ $CAN_CM;
 563 $IS $CM+ $CAN_CM;
 564 $SY $CM+ $CAN_CM;
 565
 566 $CL [$LB8NonBreaks-$CM];
 567 $CP [$LB8NonBreaks-$CM];
 568 $EX [$LB8NonBreaks-$CM];
 569 $IS [$LB8NonBreaks-$CM];
 570 $SY [$LB8NonBreaks-$CM];
 571
 572 # Rule 13 & 14 taken together for an edge case.
 573 #   Match this, shown forward
 574 #     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
 575 #   This really wants to chain at the $CM+ (which is acting as an $AL)
 576 #   except for $CM chaining being disabled.
 577 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
 578
 579 # LB 14    OP SP* x
 580 #
 581 $CAN_CM    $SP* $CM* $OP;
 582      $CANT_CM   $SP* $CM* $OP;
 583 $AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
 584
 585      $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
 586 $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 587
 588
 589 # LB 15
 590 $OP $SP* $CM* $QU;
 591
 592 # LB 16
 593 $NS $SP* $CM* ($CL | $CP);
 594
 595 # LB 17
 596 $B2 $SP* $CM* $B2;
 597
 598 # LB 18  break after spaces
 599 #        Nothing explicit needed here.
 600
 601
 602 #
 603 # LB 19
 604 #
 605 $QU $CM* $CAN_CM;                                #   . x QU
 606 $QU      $LB18NonBreaks;
 607
 608
 609 $CAN_CM  $CM* $QU;                               #   QU x .
 610      $CANT_CM $CM* $QU;
 611
 612 #
 613 #  LB 20  Break before and after CB.
 614 #         nothing needed here.
 615 #
 616
 617 # LB 20.09 added rule for Finnish tailoring
 618 $AL ($HY | $HH) / $SP;
 619
 620 # LB 21
 621 ($BA | $HH | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 622
 623 [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 624 [^$CB] $CM* $BB;                                      #
 625
 626 # LB21a
 627 [^$CB] $CM* ($HY | $BA | $HH) $CM* $HL;
 628
 629 # LB21b (reverse)
 630 $HL $CM* $SY;
 631
 632 # LB 22
 633 $IN $CM* ($ALPlus | $HL);
 634 $IN $CM* $EX;
 635 $IN $CM* ($ID | $EB | $EM);
 636 $IN $CM* $IN;
 637 $IN $CM* $NU;
 638
 639 # LB 23
 640 $PO $CM* ($ID | $EB | $EM);
 641 $NU $CM* ($ALPlus | $HL);
 642 ($ALPlus | $HL) $CM* $NU;
 643
 644 # LB 24
 645 ($ID | $EB | $EM) $CM* $PR;
 646 ($ALPlus | $HL) $CM* $PR;
 647 ($ALPlus | $HL) $CM* $PO;
 648 $CM* ($PR | $PO) $CM* ($ALPlus | $HL);
 649 $CM* ($PR | $PO) $CM+ / [$BK $CR $LF $NL $ZW $SP {eof}];
 650
 651 # LB 25
 652 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 653
 654 # LB 26
 655 ($H3 | $H2 | $JV | $JL) $CM* $JL;
 656 ($JT | $JV) $CM* ($H2 | $JV);
 657 $JT $CM* ($H3 | $JT);
 658
 659 # LB 27
 660 $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 661 $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 662  ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 663
 664 # LB 28
 665 ($ALPlus | $HL) $CM* ($ALPlus | $HL);
 666
 667
 668 # LB 29
 669 ($ALPlus | $HL) $CM* $IS;
 670
 671 # LB 30
 672 $OP $CM* ($ALPlus | $HL | $NU);
 673 ($ALPlus | $HL | $NU) $CM* $CP;
 674
 675 # LB 30a
 676 #    Pairs of Regional Indicators.
 677 #    The following two rules are nearly identical. The first matches only sequences with an odd number of adjacent RIs,
 678 #    the second with an even number. Stripping away the cruft they look like
 679 #         [^RI] RI / (RI RI)+ ^RI;
 680 #         [^RI] RI RI / (RI RI)+ ^RI;
 681 #
 682 [{bof} $NS $HY $BA $HH $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]];
 683 [{bof} $NS $HY $BA $HH $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]];
 684
 685 # In general, adjacent RIs stay together. The hard-break rules, above, overide this, forcing in the boundaries between pairs.
 686 $RI $CM* $RI;
 687
 688 #    WJ, GL, QU, etc. are classes with rules like "WJ x "   which includes "WJ x RI".
 689 $RI $CM* ([$WJ $GL $QU $BB] |  (($HY | $BA | $HH)$CM* $HL));
 690
 691
 692 # LB 30b Do not break between an Emoji Base and an Emoji Modifier
 693 $EM $CM* $EB;
 694
 695
 696 ## -------------------------------------------------
 697
 698 !!safe_reverse;
 699
 700 # LB 9
 701 ^$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 702 ^$CM+ $SP / .;
 703
 704 # LB 14
 705 $SP+ $CM* $OP;
 706
 707 # LB 15
 708 $SP+ $CM* $QU;
 709
 710 # LB 16
 711 $SP+ $CM* ($CL | $CP);
 712
 713 # LB 17
 714 $SP+ $CM* $B2;
 715
 716 # LB 21
 717 $CM* ($HY | $BA | $HH) $CM* $HL;
 718
 719 # LB 25
 720 ($CM* ($IS | $SY))+ $CM* $NU;
 721 ($CL | $CP) $CM* ($NU | $IS | $SY);
 722
 723 #  LB 30
 724 ($CM* $RI)+;
 725
 726 # For dictionary-based break
 727 $dictionary $dictionary;
 728
 729 ## -------------------------------------------------
 730
 731 !!safe_forward;
 732
 733 # Skip forward over all character classes that are involved in
 734 #   rules containing patterns with possibly more than one char
 735 #   of context.
 736 #
 737 #  It might be slightly more efficient to have specific rules
 738 #  instead of one generic one, but only if we could
 739 #  turn off rule chaining.  We don't want to move more
 740 #  than necessary.
 741 #
 742 ^[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $HH $SP $RI $ZWJ $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $HH $RI $ZWJ $dictionary];
 743 $dictionary $dictionary;
 744