1 # Copyright (c) 2002-2009 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 24 for Unicode 5.2 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 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 $CP = [:LineBreak = Close_Parenthesis:]; 67 $CR = [:LineBreak = Carriage_Return:]; 68 $EX = [:LineBreak = Exclamation:]; 69 $GL = [:LineBreak = Glue:]; 70 $HY = [:LineBreak = Hyphen:]; 71 $H2 = [:LineBreak = H2:]; 72 $H3 = [:LineBreak = H3:]; 73 $ID = [:LineBreak = Ideographic:]; 74 $IN = [:LineBreak = Inseperable:]; 75 $IS = [:LineBreak = Infix_Numeric:]; 76 $JL = [:LineBreak = JL:]; 77 $JV = [:LineBreak = JV:]; 78 $JT = [:LineBreak = JT:]; 79 $LF = [:LineBreak = Line_Feed:]; 80 $NL = [:LineBreak = Next_Line:]; 81 $NS = [:LineBreak = Nonstarter:]; 82 $NU = [:LineBreak = Numeric:]; 83 $OP = [:LineBreak = Open_Punctuation:]; 84 $PO = [:LineBreak = Postfix_Numeric:]; 85 $PR = [:LineBreak = Prefix_Numeric:]; 86 $QU = [:LineBreak = Quotation:]; 87 $SA = [:LineBreak = Complex_Context:]; 88 $SG = [:LineBreak = Surrogate:]; 89 $SP = [:LineBreak = Space:]; 90 $SY = [:LineBreak = Break_Symbols:]; 91 $WJ = [:LineBreak = Word_Joiner:]; 92 $XX = [:LineBreak = Unknown:]; 93 $ZW = [:LineBreak = ZWSpace:]; 94 95 # Dictionary character set, for triggering language-based break engines. Currently 96 # limited to LineBreak=Complex_Context. Note that this set only works in Unicode 97 # 5.0 or later as the definition of Complex_Context was corrected to include all 98 # characters requiring dictionary break. 99 100 $dictionary = [:LineBreak = Complex_Context:]; 101 102 # 103 # Rule LB1. By default, treat AI (characters with ambiguous east Asian width), 104 # SA (South East Asian: Thai, Lao, Khmer) 105 # SG (Unpaired Surrogates) 106 # XX (Unknown, unassigned) 107 # as $AL (Alphabetic) 108 # 109 $ALPlus = [$AL $AI $SA $SG $XX]; 110 111 # 112 # Combining Marks. X $CM* behaves as if it were X. Rule LB6. 113 # 114 $ALcm = $ALPlus $CM*; 115 $BAcm = $BA $CM*; 116 $BBcm = $BB $CM*; 117 $B2cm = $B2 $CM*; 118 $CLcm = $CL $CM*; 119 $CPcm = $CP $CM*; 120 $EXcm = $EX $CM*; 121 $GLcm = $GL $CM*; 122 $HYcm = $HY $CM*; 123 $H2cm = $H2 $CM*; 124 $H3cm = $H3 $CM*; 125 $IDcm = $ID $CM*; 126 $INcm = $IN $CM*; 127 $IScm = $IS $CM*; 128 $JLcm = $JL $CM*; 129 $JVcm = $JV $CM*; 130 $JTcm = $JT $CM*; 131 $NScm = $NS $CM*; 132 $NUcm = $NU $CM*; 133 $OPcm = $OP $CM*; 134 $POcm = $PO $CM*; 135 $PRcm = $PR $CM*; 136 $QUcm = $QU $CM*; 137 $SYcm = $SY $CM*; 138 $WJcm = $WJ $CM*; 139 140 ## ------------------------------------------------- 141 142 !!forward; 143 144 # 145 # Each class of character can stand by itself as an unbroken token, with trailing combining stuff 146 # 147 $ALPlus $CM+; 148 $BA $CM+; 149 $BB $CM+; 150 $B2 $CM+; 151 $CL $CM+; 152 $CP $CM+; 153 $EX $CM+; 154 $GL $CM+; 155 $HY $CM+; 156 $H2 $CM+; 157 $H3 $CM+; 158 $ID $CM+; 159 $IN $CM+; 160 $IS $CM+; 161 $JL $CM+; 162 $JV $CM+; 163 $JT $CM+; 164 $NS $CM+; 165 $NU $CM+; 166 $OP $CM+; 167 $PO $CM+; 168 $PR $CM+; 169 $QU $CM+; 170 $SY $CM+; 171 $WJ $CM+; 172 173 # 174 # CAN_CM is the set of characters that may combine with CM combining chars. 175 # Note that Linebreak UAX 14's concept of a combining char and the rules 176 # for what they can combine with are _very_ different from the rest of Unicode. 177 # 178 # Note that $CM itself is left out of this set. If CM is needed as a base 179 # it must be listed separately in the rule. 180 # 181 $CAN_CM = [^$SP $BK $CR $LF $NL $ZW $CM]; # Bases that can take CMs 182 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM]; # Bases that can't take CMs 183 184 # 185 # AL_FOLLOW set of chars that can unconditionally follow an AL 186 # Needed in rules where stand-alone $CM s are treated as AL. 187 # Chaining is disabled with CM because it causes other failures, 188 # so for this one case we need to manually list out longer sequences. 189 # 190 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP]; 191 $AL_FOLLOW_CM = [$CL $CP $EX $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus]; 192 $AL_FOLLOW = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM]; 193 194 195 # 196 # Rule LB 4, 5 Mandatory (Hard) breaks. 197 # 198 $LB4Breaks = [$BK $CR $LF $NL]; 199 $LB4NonBreaks = [^$BK $CR $LF $NL]; 200 $CR $LF {100}; 201 202 # 203 # LB 6 Do not break before hard line breaks. 204 # 205 $LB4NonBreaks? $LB4Breaks {100}; # LB 5 do not break before hard breaks. 206 $CAN_CM $CM* $LB4Breaks {100}; 207 $CM+ $LB4Breaks {100}; 208 209 # LB 7 x SP 210 # x ZW 211 $LB4NonBreaks [$SP $ZW]; 212 $CAN_CM $CM* [$SP $ZW]; 213 $CM+ [$SP $ZW]; 214 215 # 216 # LB 8 Break after zero width space 217 # 218 $LB8Breaks = [$LB4Breaks $ZW]; 219 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]]; 220 221 222 # LB 9 Combining marks. X $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL 223 # $CM not covered by the above needs to behave like $AL 224 # See definition of $CAN_CM. 225 226 $CAN_CM $CM+; # Stick together any combining sequences that don't match other rules. 227 $CM+; 228 229 # 230 # LB 11 Do not break before or after WORD JOINER & related characters. 231 # 232 $CAN_CM $CM* $WJcm; 233 $LB8NonBreaks $WJcm; 234 $CM+ $WJcm; 235 236 $WJcm $CANT_CM; 237 $WJcm $CAN_CM $CM*; 238 239 # 240 # LB 12 Do not break after NBSP and related characters. 241 # GL x 242 # 243 $GLcm $CAN_CM $CM*; 244 $GLcm $CANT_CM; 245 246 # 247 # LB 12a Do not break before NBSP and related characters ... 248 # [^SP BA HY] x GL 249 # 250 [[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm; 251 $CM+ GLcm; 252 253 254 255 # 256 # LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces. 257 # 258 $LB8NonBreaks $CL; 259 $CAN_CM $CM* $CL; 260 $CM+ $CL; # by rule 10, stand-alone CM behaves as AL 261 262 $LB8NonBreaks $CP; 263 $CAN_CM $CM* $CP; 264 $CM+ $CP; # by rule 10, stand-alone CM behaves as AL 265 266 $LB8NonBreaks $EX; 267 $CAN_CM $CM* $EX; 268 $CM+ $EX; # by rule 10, stand-alone CM behaves as AL 269 270 $LB8NonBreaks $IS; 271 $CAN_CM $CM* $IS; 272 $CM+ $IS; # by rule 10, stand-alone CM behaves as AL 273 274 $LB8NonBreaks $SY; 275 $CAN_CM $CM* $SY; 276 $CM+ $SY; # by rule 10, stand-alone CM behaves as AL 277 278 279 # 280 # LB 14 Do not break after OP, even after spaces 281 # 282 $OPcm $SP* $CAN_CM $CM*; 283 $OPcm $SP* $CANT_CM; 284 285 $OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL 286 287 # LB 15 288 $QUcm $SP* $OPcm; 289 290 # LB 16 291 ($CLcm | $CPcm) $SP* $NScm; 292 293 # LB 17 294 $B2cm $SP* $B2cm; 295 296 # 297 # LB 18 Break after spaces. 298 # 299 $LB18NonBreaks = [$LB8NonBreaks - [$SP]]; 300 $LB18Breaks = [$LB8Breaks $SP]; 301 302 303 # LB 19 304 # x QU 305 $LB18NonBreaks $CM* $QUcm; 306 $CM+ $QUcm; 307 308 # QU x 309 $QUcm .?; 310 $QUcm $LB18NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc. 311 # TODO: I don't think this rule is needed. 312 313 314 # LB 20 315 # <break> $CB 316 # $CB <break> 317 318 $LB20NonBreaks = [$LB18NonBreaks - $CB]; 319 320 # LB 21 x (BA | HY | NS) 321 # BB x 322 # 323 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm); 324 325 $BBcm [^$CB]; # $BB x 326 $BBcm $LB20NonBreaks $CM*; 327 328 # LB 22 329 $ALcm $INcm; 330 $CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL 331 $IDcm $INcm; 332 $INcm $INcm; 333 $NUcm $INcm; 334 335 336 # $LB 23 337 $IDcm $POcm; 338 $ALcm $NUcm; # includes $LB19 339 $CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL 340 $NUcm $ALcm; 341 342 # 343 # LB 24 344 # 345 $PRcm $IDcm; 346 $PRcm $ALcm; 347 $POcm $ALcm; 348 349 # 350 # LB 25 Numbers. 351 # 352 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?; 353 354 # LB 26 Do not break a Korean syllable 355 # 356 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm); 357 ($JVcm | $H2cm) ($JVcm | $JTcm); 358 ($JTcm | $H3cm) $JTcm; 359 360 # LB 27 Treat korean Syllable Block the same as ID (don't break it) 361 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm; 362 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm; 363 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm); 364 365 366 # LB 28 Do not break between alphabetics 367 # 368 $ALcm $ALcm; 369 $CM+ $ALcm; # The $CM+ is from rule 10, an unattached CM is treated as AL 370 371 # LB 29 372 $IScm $ALcm; 373 374 # LB 30 375 ($ALcm | $NUcm) $OPcm; 376 $CM+ $OPcm; # The $CM+ is from rule 10, an unattached CM is treated as AL. 377 $CPcm ($ALcm | $NUcm); 378 379 380 # 381 # Reverse Rules. 382 # 383 ## ------------------------------------------------- 384 385 !!reverse; 386 387 $CM+ $ALPlus; 388 $CM+ $BA; 389 $CM+ $BB; 390 $CM+ $B2; 391 $CM+ $CL; 392 $CM+ $CP; 393 $CM+ $EX; 394 $CM+ $GL; 395 $CM+ $HY; 396 $CM+ $H2; 397 $CM+ $H3; 398 $CM+ $ID; 399 $CM+ $IN; 400 $CM+ $IS; 401 $CM+ $JL; 402 $CM+ $JV; 403 $CM+ $JT; 404 $CM+ $NS; 405 $CM+ $NU; 406 $CM+ $OP; 407 $CM+ $PO; 408 $CM+ $PR; 409 $CM+ $QU; 410 $CM+ $SY; 411 $CM+ $WJ; 412 $CM+; 413 414 415 # 416 # Sequences of the form (shown forwards) 417 # [CANT_CM] <break> [CM] [whatever] 418 # The CM needs to behave as an AL 419 # 420 $AL_FOLLOW $CM+ / ( 421 [$BK $CR $LF $NL $ZW {eof}] | 422 $SP+ $CM+ $SP | 423 $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}])); # if LB 14 will match, need to surpress this break. 424 # LB14 says OP SP* x . 425 # becomes OP SP* x AL 426 # becomes OP SP* x CM+ AL_FOLLOW 427 # 428 # Further note: the $AL in [$AL {eof}] is only to work around 429 # a rule compiler bug which complains about 430 # empty sets otherwise. 431 432 # 433 # Sequences of the form (shown forwards) 434 # [CANT_CM] <break> [CM] <break> [PR] 435 # The CM needs to behave as an AL 436 # This rule is concerned about getting the second of the two <breaks> in place. 437 # 438 439 [$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}]; 440 441 442 443 # LB 4, 5, 5 444 445 $LB4Breaks [$LB4NonBreaks-$CM]; 446 $LB4Breaks $CM+ $CAN_CM; 447 $LF $CR; 448 449 450 # LB 7 x SP 451 # x ZW 452 [$SP $ZW] [$LB4NonBreaks-$CM]; 453 [$SP $ZW] $CM+ $CAN_CM; 454 455 # LB 8 Break after zero width space 456 457 458 # LB 9,10 Combining marks. 459 # X $CM needs to behave like X, where X is not $SP or controls. 460 # $CM not covered by the above needs to behave like $AL 461 # Stick together any combining sequences that don't match other rules. 462 $CM+ $CAN_CM; 463 464 465 # LB 11 466 $CM* $WJ $CM* $CAN_CM; 467 $CM* $WJ [$LB8NonBreaks-$CM]; 468 469 $CANT_CM $CM* $WJ; 470 $CM* $CAN_CM $CM* $WJ; 471 472 # LB 12a 473 # [^SP BA HY] x GL 474 # 475 $CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]]; 476 477 # LB 12 478 # GL x 479 # 480 $CANT_CM $CM* $GL; 481 $CM* $CAN_CM $CM* $GL; 482 483 484 # LB 13 485 $CL $CM+ $CAN_CM; 486 $CP $CM+ $CAN_CM; 487 $EX $CM+ $CAN_CM; 488 $IS $CM+ $CAN_CM; 489 $SY $CM+ $CAN_CM; 490 491 $CL [$LB8NonBreaks-$CM]; 492 $CP [$LB8NonBreaks-$CM]; 493 $EX [$LB8NonBreaks-$CM]; 494 $IS [$LB8NonBreaks-$CM]; 495 $SY [$LB8NonBreaks-$CM]; 496 497 # Rule 13 & 14 taken together for an edge case. 498 # Match this, shown forward 499 # OP SP+ ($CM+ behaving as $AL) (CL | CP | EX | IS | IY) 500 # This really wants to chain at the $CM+ (which is acting as an $AL) 501 # except for $CM chaining being disabled. 502 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP; 503 504 # LB 14 OP SP* x 505 # 506 $CM* $CAN_CM $SP* $CM* $OP; 507 $CANT_CM $SP* $CM* $OP; 508 $AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP 509 510 $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP; 511 $CM* $AL_FOLLOW_CM $CM+ $SP+ $CM* $OP; 512 $SY $CM $SP+ $OP; # TODO: Experiment. Remove. 513 514 515 516 # LB 15 517 $CM* $OP $SP* $CM* $QU; 518 519 # LB 16 520 $CM* $NS $SP* $CM* ($CL | $CP); 521 522 # LB 17 523 $CM* $B2 $SP* $CM* $B2; 524 525 # LB 18 break after spaces 526 # Nothing explicit needed here. 527 528 529 # 530 # LB 19 531 # 532 $CM* $QU $CM* $CAN_CM; # . x QU 533 $CM* $QU $LB18NonBreaks; 534 535 536 $CM* $CAN_CM $CM* $QU; # QU x . 537 $CANT_CM $CM* $QU; 538 539 # 540 # LB 20 Break before and after CB. 541 # nothing needed here. 542 # 543 544 # LB 21 545 $CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA | HY | NS) 546 547 $CM* [$LB20NonBreaks-$CM] $CM* $BB; # BB x . 548 [^$CB] $CM* $BB; # 549 550 551 552 # LB 22 553 $CM* $IN $CM* $ALPlus; 554 $CM* $IN $CM* $ID; 555 $CM* $IN $CM* $IN; 556 $CM* $IN $CM* $NU; 557 558 # LB 23 559 $CM* $PO $CM* $ID; 560 $CM* $NU $CM* $ALPlus; 561 $CM* $ALPlus $CM* $NU; 562 563 # LB 24 564 $CM* $ID $CM* $PR; 565 $CM* $ALPlus $CM* $PR; 566 $CM* $ALPlus $CM* $PO; 567 568 569 # LB 25 570 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?; 571 572 # LB 26 573 $CM* ($H3 | $H2 | $JV | $JL) $CM* $JL; 574 $CM* ($JT | $JV) $CM* ($H2 | $JV); 575 $CM* $JT $CM* ($H3 | $JT); 576 577 # LB 27 578 $CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL); 579 $CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL); 580 $CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR; 581 582 # LB 28 583 $CM* $ALPlus $CM* $ALPlus; 584 585 586 # LB 29 587 $CM* $ALPlus $CM* $IS; 588 589 # LB 30 590 $CM* $OP $CM* ($ALPlus | $NU); 591 $CM* ($ALPlus | $NU) $CM* $CP; 592 593 594 ## ------------------------------------------------- 595 596 !!safe_reverse; 597 598 # LB 9 599 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP]; 600 $CM+ $SP / .; 601 602 # LB 14 603 $SP+ $CM* $OP; 604 605 # LB 15 606 $SP+ $CM* $QU; 607 608 # LB 16 609 $SP+ $CM* ($CL | $CP); 610 611 # LB 17 612 $SP+ $CM* $B2; 613 614 # LB 25 615 ($CM* ($IS | $SY))+ $CM* $NU; 616 ($CL | $CP) $CM* ($NU | $IS | $SY); 617 618 # For dictionary-based break 619 $dictionary $dictionary; 620 621 ## ------------------------------------------------- 622 623 !!safe_forward; 624 625 # Skip forward over all character classes that are involved in 626 # rules containing patterns with possibly more than one char 627 # of context. 628 # 629 # It might be slightly more efficient to have specific rules 630 # instead of one generic one, but only if we could 631 # turn off rule chaining. We don't want to move more 632 # than necessary. 633 # 634 [$CM $OP $QU $CL $CP $B2 $PR $HY $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $dictionary]; 635 $dictionary $dictionary; 636 637