icuSources/data/brkitr/line.txt

   1 # Copyright (c) 2002-2008  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 21 for Unicode 5.1
   9 #         http://www.unicode.org/reports/tr14/
  10
  11
  12
  13 #
  14 #  Character Classes defined by TR 14.
  15 #
  16
  17 !!chain;
  18 !!LBCMNoChain;
  19
  20
  21 !!lookAheadHardBreak;
  22 #
  23 #  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
  24 #                          and only used for the line break rules.
  25 #
  26 #           It is used in the implementation of the incredibly annoying rule LB 10
  27 #           which says to treat any combining mark that is not attached to a base
  28 #           character as if it were of class AL  (alphabetic).
  29 #
  30 #           The problem occurs in the reverse rules.
  31 #
  32 #           Consider a sequence like, with correct breaks as shown
  33 #               LF  ID  CM  AL  AL
  34 #                  ^       ^       ^
  35 #           Then consider the sequence without the initial ID (ideographic)
  36 #                 LF  CM  AL  AL
  37 #                    ^           ^
  38 #           Our CM, which in the first example was attached to the ideograph,
  39 #           is now unattached, becomes an alpha, and joins in with the other
  40 #           alphas.
  41 #
  42 #           When iterating forwards, these sequences do not present any problems
  43 #           When iterating backwards, we need to look ahead when encountering
  44 #           a CM to see whether it attaches to something further on or not.
  45 #           (Look-ahead in a reverse rule is looking towards the start)
  46 #
  47 #           If the CM is unattached, we need to force a break.
  48 #
  49 #           !!lookAheadHardBreak forces the run time state machine to
  50 #           stop immediately when a look ahead rule ( '/' operator) matches,
  51 #           and set the match position to that of the look-ahead operator,
  52 #           no matter what other rules may be in play at the time.
  53 #
  54 #           See rule LB 19 for an example.
  55 #
  56
  57 $AI = [:LineBreak =  Ambiguous:];
  58 $AL = [:LineBreak =  Alphabetic:];
  59 $BA = [:LineBreak =  Break_After:];
  60 $BB = [:LineBreak =  Break_Before:];
  61 $BK = [:LineBreak =  Mandatory_Break:];
  62 $B2 = [:LineBreak =  Break_Both:];
  63 $CB = [:LineBreak =  Contingent_Break:];
  64 $CL = [:LineBreak =  Close_Punctuation:];
  65 $CM = [:LineBreak =  Combining_Mark:];
  66 $CR = [:LineBreak =  Carriage_Return:];
  67 $EX = [:LineBreak =  Exclamation:];
  68 $GL = [:LineBreak =  Glue:];
  69 $HY = [:LineBreak =  Hyphen:];
  70 $H2 = [:LineBreak =  H2:];
  71 $H3 = [:LineBreak =  H3:];
  72 $ID = [:LineBreak =  Ideographic:];
  73 $IN = [:LineBreak =  Inseperable:];
  74 $IS = [:LineBreak =  Infix_Numeric:];
  75 $JL = [:LineBreak =  JL:];
  76 $JV = [:LineBreak =  JV:];
  77 $JT = [:LineBreak =  JT:];
  78 $LF = [:LineBreak =  Line_Feed:];
  79 $NL = [:LineBreak =  Next_Line:];
  80 $NS = [:LineBreak =  Nonstarter:];
  81 $NU = [:LineBreak =  Numeric:];
  82 $OP = [:LineBreak =  Open_Punctuation:];
  83 $PO = [:LineBreak =  Postfix_Numeric:];
  84 $PR = [:LineBreak =  Prefix_Numeric:];
  85 $QU = [:LineBreak =  Quotation:];
  86 $SA = [:LineBreak =  Complex_Context:];
  87 $SG = [:LineBreak =  Surrogate:];
  88 $SP = [:LineBreak =  Space:];
  89 $SY = [:LineBreak =  Break_Symbols:];
  90 $WJ = [:LineBreak =  Word_Joiner:];
  91 $XX = [:LineBreak =  Unknown:];
  92 $ZW = [:LineBreak =  ZWSpace:];
  93
  94 #   Dictionary character set, for triggering language-based break engines. Currently
  95 #   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
  96 #   5.0 or later as the definition of Complex_Context was corrected to include all
  97 #   characters requiring dictionary break.
  98
  99 $dictionary = [:LineBreak = Complex_Context:];
 100
 101 #
 102 #  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
 103 #                               SA  (South East Asian: Thai, Lao, Khmer)
 104 #                               SG  (Unpaired Surrogates)
 105 #                               XX  (Unknown, unassigned)
 106 #                         as $AL  (Alphabetic)
 107 #
 108 $ALPlus = [$AL $AI $SA $SG $XX];
 109
 110 #
 111 #  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
 112 #
 113 $ALcm = $ALPlus $CM*;
 114 $BAcm = $BA $CM*;
 115 $BBcm = $BB $CM*;
 116 $B2cm = $B2 $CM*;
 117 $CLcm = $CL $CM*;
 118 $EXcm = $EX $CM*;
 119 $GLcm = $GL $CM*;
 120 $HYcm = $HY $CM*;
 121 $H2cm = $H2 $CM*;
 122 $H3cm = $H3 $CM*;
 123 $IDcm = $ID $CM*;
 124 $INcm = $IN $CM*;
 125 $IScm = $IS $CM*;
 126 $JLcm = $JL $CM*;
 127 $JVcm = $JV $CM*;
 128 $JTcm = $JT $CM*;
 129 $NScm = $NS $CM*;
 130 $NUcm = $NU $CM*;
 131 $OPcm = $OP $CM*;
 132 $POcm = $PO $CM*;
 133 $PRcm = $PR $CM*;
 134 $QUcm = $QU $CM*;
 135 $SYcm = $SY $CM*;
 136 $WJcm = $WJ $CM*;
 137
 138 ## -------------------------------------------------
 139
 140 !!forward;
 141
 142 #
 143 #  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
 144 #
 145 $ALPlus $CM+;
 146 $BA $CM+;
 147 $BB $CM+;
 148 $B2 $CM+;
 149 $CL $CM+;
 150 $EX $CM+;
 151 $GL $CM+;
 152 $HY $CM+;
 153 $H2 $CM+;
 154 $H3 $CM+;
 155 $ID $CM+;
 156 $IN $CM+;
 157 $IS $CM+;
 158 $JL $CM+;
 159 $JV $CM+;
 160 $JT $CM+;
 161 $NS $CM+;
 162 $NU $CM+;
 163 $OP $CM+;
 164 $PO $CM+;
 165 $PR $CM+;
 166 $QU $CM+;
 167 $SY $CM+;
 168 $WJ $CM+;
 169
 170 #
 171 # CAN_CM  is the set of characters that may combine with CM combining chars.
 172 #         Note that Linebreak UAX 14's concept of a combining char and the rules
 173 #         for what they can combine with are _very_ different from the rest of Unicode.
 174 #
 175 #         Note that $CM itself is left out of this set.  If CM is needed as a base
 176 #         it must be listed separately in the rule.
 177 #
 178 $CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
 179 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
 180
 181 #
 182 # AL_FOLLOW  set of chars that can unconditionally follow an AL
 183 #            Needed in rules where stand-alone $CM s are treated as AL.
 184 #            Chaining is disabled with CM because it causes other failures,
 185 #            so for this one case we need to manually list out longer sequences.
 186 #
 187 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
 188 $AL_FOLLOW_CM   = [$CL $EX $IS $SY $WJ $GL $QU $BA $HY $NS $IN $NU $ALPlus];
 189 $AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
 190
 191
 192 #
 193 #  Rule LB 4, 5    Mandatory (Hard) breaks.
 194 #
 195 $LB4Breaks    = [$BK $CR $LF $NL];
 196 $LB4NonBreaks = [^$BK $CR $LF $NL];
 197 $CR $LF {100};
 198
 199 #
 200 #  LB 6    Do not break before hard line breaks.
 201 #
 202 $LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
 203 $CAN_CM $CM*    $LB4Breaks {100};
 204 $CM+            $LB4Breaks {100};
 205
 206 # LB 7         x SP
 207 #              x ZW
 208 $LB4NonBreaks [$SP $ZW];
 209 $CAN_CM $CM*  [$SP $ZW];
 210 $CM+          [$SP $ZW];
 211
 212 #
 213 # LB 8         Break after zero width space
 214 #
 215 $LB8Breaks    = [$LB4Breaks $ZW];
 216 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
 217
 218
 219 # LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
 220 #                                $CM not covered by the above needs to behave like $AL
 221 #                                See definition of $CAN_CM.
 222
 223 $CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
 224 $CM+;
 225
 226 #
 227 # LB 11  Do not break before or after WORD JOINER & related characters.
 228 #
 229 $CAN_CM $CM*  $WJcm;
 230 $LB8NonBreaks $WJcm;
 231 $CM+          $WJcm;
 232
 233 $WJcm $CANT_CM;
 234 $WJcm $CAN_CM $CM*;
 235
 236 #
 237 # LB 12  Do not break after NBSP and related characters.
 238 #         GL  x
 239 #
 240 $GLcm $CAN_CM $CM*;
 241 $GLcm $CANT_CM;
 242
 243 #
 244 # LB 12a  Do not break before NBSP and related characters ...
 245 #            [^SP BA HY] x GL
 246 #
 247 [[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
 248 $CM+ GLcm;
 249
 250
 251
 252 #
 253 # LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
 254 #
 255 $LB8NonBreaks $CL;
 256 $CAN_CM $CM*  $CL;
 257 $CM+          $CL;              # by rule 10, stand-alone CM behaves as AL
 258
 259 $LB8NonBreaks $EX;
 260 $CAN_CM $CM*  $EX;
 261 $CM+          $EX;              # by rule 10, stand-alone CM behaves as AL
 262
 263 $LB8NonBreaks $IS;
 264 $CAN_CM $CM*  $IS;
 265 $CM+          $IS;              # by rule 10, stand-alone CM behaves as AL
 266
 267 $LB8NonBreaks $SY;
 268 $CAN_CM $CM*  $SY;
 269 $CM+          $SY;              # by rule 10, stand-alone CM behaves as AL
 270
 271
 272 #
 273 # LB 14  Do not break after OP, even after spaces
 274 #
 275 $OPcm $SP* $CAN_CM $CM*;
 276 $OPcm $SP* $CANT_CM;
 277
 278 $OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
 279
 280 # LB 15
 281 $QUcm $SP* $OPcm;
 282
 283 # LB 16
 284 $CLcm $SP* $NScm;
 285
 286 # LB 17
 287 $B2cm $SP* $B2cm;
 288
 289 #
 290 # LB 18  Break after spaces.
 291 #
 292 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
 293 $LB18Breaks    = [$LB8Breaks $SP];
 294
 295
 296 # LB 19
 297 #         x QU
 298 $LB18NonBreaks $CM* $QUcm;
 299 $CM+                $QUcm;
 300
 301 #         QU  x
 302 $QUcm .?;
 303 $QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
 304                               #  TODO:  I don't think this rule is needed.
 305
 306
 307 # LB 20
 308 #        <break>  $CB
 309 #        $CB   <break>
 310
 311 $LB20NonBreaks = [$LB18NonBreaks - $CB];
 312
 313 # LB 21        x   (BA | HY | NS)
 314 #           BB x
 315 #
 316 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
 317
 318 $BBcm [^$CB];                                  #  $BB  x
 319 $BBcm $LB20NonBreaks $CM*;
 320
 321 # LB 22
 322 $ALcm    $INcm;
 323 $CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
 324 $IDcm    $INcm;
 325 $INcm    $INcm;
 326 $NUcm    $INcm;
 327
 328
 329 # $LB 23
 330 $IDcm  $POcm;
 331 $ALcm  $NUcm;       # includes $LB19
 332 $CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
 333 $NUcm  $ALcm;
 334
 335 #
 336 # LB 24
 337 #
 338 $PRcm $IDcm;
 339 $PRcm $ALcm;
 340 $POcm $ALcm;
 341
 342 #
 343 # LB 25   Numbers.
 344 #
 345 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* $CLcm? ($PRcm | $POcm)?;
 346
 347 # LB 26  Do not break a Korean syllable
 348 #
 349 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
 350 ($JVcm | $H2cm) ($JVcm | $JTcm);
 351 ($JTcm | $H3cm) $JTcm;
 352
 353 # LB 27  Treat korean Syllable Block the same as ID  (don't break it)
 354 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
 355 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
 356 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
 357
 358
 359 # LB 28   Do not break between alphabetics
 360 #
 361 $ALcm $ALcm;
 362 $CM+ $ALcm;      # The $CM+ is from rule 10, and unattached CM is treated as AL
 363
 364 # LB 29
 365 $IScm $ALcm;
 366
 367
 368 #
 369 #  Reverse Rules.
 370 #
 371 ## -------------------------------------------------
 372
 373 !!reverse;
 374
 375 $CM+ $ALPlus;
 376 $CM+ $BA;
 377 $CM+ $BB;
 378 $CM+ $B2;
 379 $CM+ $CL;
 380 $CM+ $EX;
 381 $CM+ $GL;
 382 $CM+ $HY;
 383 $CM+ $H2;
 384 $CM+ $H3;
 385 $CM+ $ID;
 386 $CM+ $IN;
 387 $CM+ $IS;
 388 $CM+ $JL;
 389 $CM+ $JV;
 390 $CM+ $JT;
 391 $CM+ $NS;
 392 $CM+ $NU;
 393 $CM+ $OP;
 394 $CM+ $PO;
 395 $CM+ $PR;
 396 $CM+ $QU;
 397 $CM+ $SY;
 398 $CM+ $WJ;
 399 $CM+;
 400
 401
 402 #
 403 #  Sequences of the form  (shown forwards)
 404 #      [CANT_CM]  <break>  [CM]  [whatever]
 405 #  The CM needs to behave as an AL
 406 #
 407 $AL_FOLLOW $CM+ / (
 408           [$BK $CR $LF $NL $ZW {eof}] |
 409           $SP+ $CM+ $SP |
 410           $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
 411                                                #  LB14 says    OP SP* x .
 412                                                #    becomes    OP SP* x AL
 413                                                #    becomes    OP SP* x CM+ AL_FOLLOW
 414                                                #
 415                                                # Further note:  the $AL in [$AL {eof}] is only to work around
 416                                                #                a rule compiler bug which complains about
 417                                                #                empty sets otherwise.
 418
 419 #
 420 #  Sequences of the form  (shown forwards)
 421 #      [CANT_CM]  <break> [CM]  <break>  [PR]
 422 #  The CM needs to behave as an AL
 423 #  This rule is concerned about getting the second of the two <breaks> in place.
 424 #
 425
 426 [$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
 427
 428
 429
 430 # LB 4, 5, 5
 431
 432 $LB4Breaks [$LB4NonBreaks-$CM];
 433 $LB4Breaks $CM+ $CAN_CM;
 434 $LF $CR;
 435
 436
 437 # LB 7         x SP
 438 #              x ZW
 439 [$SP $ZW] [$LB4NonBreaks-$CM];
 440 [$SP $ZW] $CM+ $CAN_CM;
 441
 442 # LB 8 Break after zero width space
 443
 444
 445 # LB 9,10  Combining marks.
 446 #    X   $CM needs to behave like X, where X is not $SP or controls.
 447 #    $CM not covered by the above needs to behave like $AL
 448 # Stick together any combining sequences that don't match other rules.
 449 $CM+ $CAN_CM;
 450
 451
 452 # LB 11
 453 $CM* $WJ $CM* $CAN_CM;
 454 $CM* $WJ      [$LB8NonBreaks-$CM];
 455
 456      $CANT_CM $CM* $WJ;
 457 $CM* $CAN_CM  $CM* $WJ;
 458
 459 # LB 12a
 460 #      [^SP BA HY] x GL
 461 #
 462 $CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
 463
 464 # LB 12
 465 #     GL  x
 466 #
 467 $CANT_CM $CM* $GL;
 468 $CM* $CAN_CM $CM* $GL;
 469
 470
 471 # LB 13
 472 $CL $CM+ $CAN_CM;
 473 $EX $CM+ $CAN_CM;
 474 $IS $CM+ $CAN_CM;
 475 $SY $CM+ $CAN_CM;
 476
 477 $CL [$LB8NonBreaks-$CM];
 478 $EX [$LB8NonBreaks-$CM];
 479 $IS [$LB8NonBreaks-$CM];
 480 $SY [$LB8NonBreaks-$CM];
 481
 482 # Rule 13 & 14 taken together for an edge case.
 483 #   Match this, shown forward
 484 #     OP SP+  ($CM+ behaving as $AL) (CL | EX | IS | IY)
 485 #   This really wants to chain at the $CM+ (which is acting as an $AL)
 486 #   except for $CM chaining being disabled.
 487 [$CL $EX $IS $SY] $CM+ $SP+ $CM* $OP;
 488
 489 # LB 14    OP SP* x
 490 #
 491 $CM* $CAN_CM    $SP* $CM* $OP;
 492      $CANT_CM   $SP* $CM* $OP;
 493 $AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
 494
 495      $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
 496 $CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
 497 $SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
 498
 499
 500
 501 # LB 15
 502 $CM* $OP $SP* $CM* $QU;
 503
 504 # LB 16
 505 $CM* $NS $SP* $CM* $CL;
 506
 507 # LB 17
 508 $CM* $B2 $SP* $CM* $B2;
 509
 510 # LB 18  break after spaces
 511 #        Nothing explicit needed here.
 512
 513
 514 #
 515 # LB 19
 516 #
 517 $CM* $QU $CM* $CAN_CM;                                #   . x QU
 518 $CM* $QU      $LB18NonBreaks;
 519
 520
 521 $CM* $CAN_CM  $CM* $QU;                               #   QU x .
 522      $CANT_CM $CM* $QU;
 523
 524 #
 525 #  LB 20  Break before and after CB.
 526 #         nothing needed here.
 527 #
 528
 529 # LB 21
 530 $CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
 531
 532 $CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
 533 [^$CB] $CM* $BB;                                      #
 534
 535
 536
 537 # LB 22
 538 $CM* $IN $CM* $ALPlus;
 539 $CM* $IN $CM* $ID;
 540 $CM* $IN $CM* $IN;
 541 $CM* $IN $CM* $NU;
 542
 543 # LB 23
 544 $CM* $PO $CM* $ID;
 545 $CM* $NU $CM* $ALPlus;
 546 $CM* $ALPlus $CM* $NU;
 547
 548 # LB 24
 549 $CM* $ID $CM* $PR;
 550 $CM* $ALPlus $CM* $PR;
 551 $CM* $ALPlus $CM* $PO;
 552
 553
 554 # LB 25
 555 ($CM* ($PR | $PO))? ($CM* $CL)? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
 556
 557 # LB 26
 558 $CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
 559 $CM* ($JT | $JV) $CM* ($H2 | $JV);
 560 $CM* $JT $CM* ($H3 | $JT);
 561
 562 # LB 27
 563 $CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
 564 $CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
 565 $CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
 566
 567 # LB 28
 568 $CM* $ALPlus $CM* $ALPlus;
 569
 570
 571 # LB 29
 572 $CM* $ALPlus $CM* $IS;
 573
 574
 575 ## -------------------------------------------------
 576
 577 !!safe_reverse;
 578
 579 # LB 9
 580 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
 581 $CM+ $SP / .;
 582
 583 # LB 14
 584 $SP+ $CM* $OP;
 585
 586 # LB 15
 587 $SP+ $CM* $QU;
 588
 589 # LB 16
 590 $SP+ $CM* $CL;
 591
 592 # LB 17
 593 $SP+ $CM* $B2;
 594
 595 # LB 25
 596 ($CM* ($IS | $SY))+ $CM* $NU;
 597 $CL $CM* ($NU | $IS | $SY);
 598
 599 # For dictionary-based break
 600 $dictionary $dictionary;
 601
 602 ## -------------------------------------------------
 603
 604 !!safe_forward;
 605
 606 # Skip forward over all character classes that are involved in
 607 #   rules containing patterns with possibly more than one char
 608 #   of context.
 609 #
 610 #  It might be slightly more efficient to have specific rules
 611 #  instead of one generic one, but only if we could
 612 #  turn off rule chaining.  We don't want to move more
 613 #  than necessary.
 614 #
 615 [$CM $OP $QU $CL $B2 $PR $HY $SP $dictionary]+ [^$CM $OP $QU $CL $B2 $PR $HY $dictionary];
 616 $dictionary $dictionary;
 617