1 # Copyright (c) 2002-2010 International Business Machines Corporation and 2 # others. All Rights Reserved. 3 # 4 # file: line_fi.txt 5 # 6 # Line Breaking Rules 7 # Implement default line breaking as defined by 8 # Unicode Standard Annex #14 Revision 24 for Unicode 6.0 9 # http://www.unicode.org/reports/tr14/ 10 # 11 # TODO: Rule LB 8 remains as it was in Unicode 5.2 12 # This is only because of a limitation of ICU break engine implementation, 13 # not because the older behavior is desirable. 14 15 # 16 # Character Classes defined by TR 14. 17 # 18 19 !!chain; 20 !!LBCMNoChain; 21 22 23 !!lookAheadHardBreak; 24 # 25 # !!lookAheadHardBreak Described here because it is (as yet) undocumented elsewhere 26 # and only used for the line break rules. 27 # 28 # It is used in the implementation of rule LB 10 29 # which says to treat any combining mark that is not attached to a base 30 # character as if it were of class AL (alphabetic). 31 # 32 # The problem occurs in the reverse rules. 33 # 34 # Consider a sequence like, with correct breaks as shown 35 # LF ID CM AL AL 36 # ^ ^ ^ 37 # Then consider the sequence without the initial ID (ideographic) 38 # LF CM AL AL 39 # ^ ^ 40 # Our CM, which in the first example was attached to the ideograph, 41 # is now unattached, becomes an alpha, and joins in with the other 42 # alphas. 43 # 44 # When iterating forwards, these sequences do not present any problems 45 # When iterating backwards, we need to look ahead when encountering 46 # a CM to see whether it attaches to something further on or not. 47 # (Look-ahead in a reverse rule is looking towards the start) 48 # 49 # If the CM is unattached, we need to force a break. 50 # 51 # !!lookAheadHardBreak forces the run time state machine to 52 # stop immediately when a look ahead rule ( '/' operator) matches, 53 # and set the match position to that of the look-ahead operator, 54 # no matter what other rules may be in play at the time. 55 # 56 # See rule LB 19 for an example. 57 # 58 59 $AI = [:LineBreak = Ambiguous:]; 60 $AL = [:LineBreak = Alphabetic:]; 61 $BA = [[:LineBreak = Break_After:] - [\u2010]]; 62 $HH = [\u2010]; 63 $BB = [:LineBreak = Break_Before:]; 64 $BK = [:LineBreak = Mandatory_Break:]; 65 $B2 = [:LineBreak = Break_Both:]; 66 $CB = [:LineBreak = Contingent_Break:]; 67 $CL = [:LineBreak = Close_Punctuation:]; 68 $CM = [:LineBreak = Combining_Mark:]; 69 $CP = [:LineBreak = Close_Parenthesis:]; 70 $CR = [:LineBreak = Carriage_Return:]; 71 $EX = [:LineBreak = Exclamation:]; 72 $GL = [:LineBreak = Glue:]; 73 $HY = [:LineBreak = Hyphen:]; 74 $H2 = [:LineBreak = H2:]; 75 $H3 = [:LineBreak = H3:]; 76 $ID = [:LineBreak = Ideographic:]; 77 $IN = [:LineBreak = Inseperable:]; 78 $IS = [:LineBreak = Infix_Numeric:]; 79 $JL = [:LineBreak = JL:]; 80 $JV = [:LineBreak = JV:]; 81 $JT = [:LineBreak = JT:]; 82 $LF = [:LineBreak = Line_Feed:]; 83 $NL = [:LineBreak = Next_Line:]; 84 $NS = [:LineBreak = Nonstarter:]; 85 $NU = [:LineBreak = Numeric:]; 86 $OP = [:LineBreak = Open_Punctuation:]; 87 $PO = [:LineBreak = Postfix_Numeric:]; 88 $PR = [:LineBreak = Prefix_Numeric:]; 89 $QU = [:LineBreak = Quotation:]; 90 $SA = [:LineBreak = Complex_Context:]; 91 $SG = [:LineBreak = Surrogate:]; 92 $SP = [:LineBreak = Space:]; 93 $SY = [:LineBreak = Break_Symbols:]; 94 $WJ = [:LineBreak = Word_Joiner:]; 95 $XX = [:LineBreak = Unknown:]; 96 $ZW = [:LineBreak = ZWSpace:]; 97 98 # Dictionary character set, for triggering language-based break engines. Currently 99 # limited to LineBreak=Complex_Context. Note that this set only works in Unicode 100 # 5.0 or later as the definition of Complex_Context was corrected to include all 101 # characters requiring dictionary break. 102 103 $dictionary = [:LineBreak = Complex_Context:]; 104 105 # 106 # Rule LB1. By default, treat AI (characters with ambiguous east Asian width), 107 # SA (South East Asian: Thai, Lao, Khmer) 108 # SG (Unpaired Surrogates) 109 # XX (Unknown, unassigned) 110 # as $AL (Alphabetic) 111 # 112 $ALPlus = [$AL $AI $SA $SG $XX]; 113 114 # 115 # Combining Marks. X $CM* behaves as if it were X. Rule LB6. 116 # 117 $ALcm = $ALPlus $CM*; 118 $BAcm = $BA $CM*; 119 $HHcm = $HH $CM*; 120 $BBcm = $BB $CM*; 121 $B2cm = $B2 $CM*; 122 $CLcm = $CL $CM*; 123 $CPcm = $CP $CM*; 124 $EXcm = $EX $CM*; 125 $GLcm = $GL $CM*; 126 $HYcm = $HY $CM*; 127 $H2cm = $H2 $CM*; 128 $H3cm = $H3 $CM*; 129 $IDcm = $ID $CM*; 130 $INcm = $IN $CM*; 131 $IScm = $IS $CM*; 132 $JLcm = $JL $CM*; 133 $JVcm = $JV $CM*; 134 $JTcm = $JT $CM*; 135 $NScm = $NS $CM*; 136 $NUcm = $NU $CM*; 137 $OPcm = $OP $CM*; 138 $POcm = $PO $CM*; 139 $PRcm = $PR $CM*; 140 $QUcm = $QU $CM*; 141 $SYcm = $SY $CM*; 142 $WJcm = $WJ $CM*; 143 144 ## ------------------------------------------------- 145 146 !!forward; 147 148 # 149 # Each class of character can stand by itself as an unbroken token, with trailing combining stuff 150 # 151 $ALPlus $CM+; 152 $BA $CM+; 153 $HH $CM+; 154 $BB $CM+; 155 $B2 $CM+; 156 $CL $CM+; 157 $CP $CM+; 158 $EX $CM+; 159 $GL $CM+; 160 $HY $CM+; 161 $H2 $CM+; 162 $H3 $CM+; 163 $ID $CM+; 164 $IN $CM+; 165 $IS $CM+; 166 $JL $CM+; 167 $JV $CM+; 168 $JT $CM+; 169 $NS $CM+; 170 $NU $CM+; 171 $OP $CM+; 172 $PO $CM+; 173 $PR $CM+; 174 $QU $CM+; 175 $SY $CM+; 176 $WJ $CM+; 177 178 # 179 # CAN_CM is the set of characters that may combine with CM combining chars. 180 # Note that Linebreak UAX 14's concept of a combining char and the rules 181 # for what they can combine with are _very_ different from the rest of Unicode. 182 # 183 # Note that $CM itself is left out of this set. If CM is needed as a base 184 # it must be listed separately in the rule. 185 # 186 $CAN_CM = [^$SP $BK $CR $LF $NL $ZW $CM]; # Bases that can take CMs 187 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM]; # Bases that can't take CMs 188 189 # 190 # AL_FOLLOW set of chars that can unconditionally follow an AL 191 # Needed in rules where stand-alone $CM s are treated as AL. 192 # Chaining is disabled with CM because it causes other failures, 193 # so for this one case we need to manually list out longer sequences. 194 # 195 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP]; 196 $AL_FOLLOW_CM = [$CL $CP $EX $IS $SY $WJ $GL $OP $QU $BA $HH $HY $NS $IN $NU $ALPlus]; 197 $AL_FOLLOW = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM]; 198 199 200 # 201 # Rule LB 4, 5 Mandatory (Hard) breaks. 202 # 203 $LB4Breaks = [$BK $CR $LF $NL]; 204 $LB4NonBreaks = [^$BK $CR $LF $NL]; 205 $CR $LF {100}; 206 207 # 208 # LB 6 Do not break before hard line breaks. 209 # 210 $LB4NonBreaks? $LB4Breaks {100}; # LB 5 do not break before hard breaks. 211 $CAN_CM $CM* $LB4Breaks {100}; 212 $CM+ $LB4Breaks {100}; 213 214 # LB 7 x SP 215 # x ZW 216 $LB4NonBreaks [$SP $ZW]; 217 $CAN_CM $CM* [$SP $ZW]; 218 $CM+ [$SP $ZW]; 219 220 # 221 # LB 8 Break after zero width space 222 # TODO: ZW SP* <break> 223 # An engine change is required to write the reverse rule for this. 224 # For now, leave the Unicode 5.2 rule, ZW <break> 225 # 226 $LB8Breaks = [$LB4Breaks $ZW]; 227 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]]; 228 229 230 # LB 9 Combining marks. X $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL 231 # $CM not covered by the above needs to behave like $AL 232 # See definition of $CAN_CM. 233 234 $CAN_CM $CM+; # Stick together any combining sequences that don't match other rules. 235 $CM+; 236 237 # 238 # LB 11 Do not break before or after WORD JOINER & related characters. 239 # 240 $CAN_CM $CM* $WJcm; 241 $LB8NonBreaks $WJcm; 242 $CM+ $WJcm; 243 244 $WJcm $CANT_CM; 245 $WJcm $CAN_CM $CM*; 246 247 # 248 # LB 12 Do not break after NBSP and related characters. 249 # GL x 250 # 251 $GLcm $CAN_CM $CM*; 252 $GLcm $CANT_CM; 253 254 # 255 # LB 12a Do not break before NBSP and related characters ... 256 # [^SP BA HY] x GL 257 # 258 [[$LB8NonBreaks] - [$SP $BA $HH $HY]] $CM* $GLcm; 259 $CM+ GLcm; 260 261 262 263 # 264 # LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces. 265 # 266 $LB8NonBreaks $CL; 267 $CAN_CM $CM* $CL; 268 $CM+ $CL; # by rule 10, stand-alone CM behaves as AL 269 270 $LB8NonBreaks $CP; 271 $CAN_CM $CM* $CP; 272 $CM+ $CP; # by rule 10, stand-alone CM behaves as AL 273 274 $LB8NonBreaks $EX; 275 $CAN_CM $CM* $EX; 276 $CM+ $EX; # by rule 10, stand-alone CM behaves as AL 277 278 $LB8NonBreaks $IS; 279 $CAN_CM $CM* $IS; 280 $CM+ $IS; # by rule 10, stand-alone CM behaves as AL 281 282 $LB8NonBreaks $SY; 283 $CAN_CM $CM* $SY; 284 $CM+ $SY; # by rule 10, stand-alone CM behaves as AL 285 286 287 # 288 # LB 14 Do not break after OP, even after spaces 289 # 290 $OPcm $SP* $CAN_CM $CM*; 291 $OPcm $SP* $CANT_CM; 292 293 $OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL 294 295 # LB 15 296 $QUcm $SP* $OPcm; 297 298 # LB 16 299 ($CLcm | $CPcm) $SP* $NScm; 300 301 # LB 17 302 $B2cm $SP* $B2cm; 303 304 # 305 # LB 18 Break after spaces. 306 # 307 $LB18NonBreaks = [$LB8NonBreaks - [$SP]]; 308 $LB18Breaks = [$LB8Breaks $SP]; 309 310 311 # LB 19 312 # x QU 313 $LB18NonBreaks $CM* $QUcm; 314 $CM+ $QUcm; 315 316 # QU x 317 $QUcm .?; 318 $QUcm $LB18NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc. 319 # TODO: I don't think this rule is needed. 320 321 322 # LB 20 323 # <break> $CB 324 # $CB <break> 325 326 $LB20NonBreaks = [$LB18NonBreaks - $CB]; 327 328 # LB 20.09 added rule for Finnish tailoring 329 # LB 21 x (BA | HY | NS) 330 # BB x 331 # 332 $LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm) / $AL; 333 $LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm); 334 ($HY | $HH) $AL; 335 336 $BBcm [^$CB]; # $BB x 337 $BBcm $LB20NonBreaks $CM*; 338 339 # LB 22 340 $ALcm $INcm; 341 $CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL 342 $IDcm $INcm; 343 $INcm $INcm; 344 $NUcm $INcm; 345 346 347 # $LB 23 348 $IDcm $POcm; 349 $ALcm $NUcm; # includes $LB19 350 $CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL 351 $NUcm $ALcm; 352 353 # 354 # LB 24 355 # 356 $PRcm $IDcm; 357 $PRcm $ALcm; 358 $POcm $ALcm; 359 360 # 361 # LB 25 Numbers. 362 # 363 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?; 364 365 # LB 26 Do not break a Korean syllable 366 # 367 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm); 368 ($JVcm | $H2cm) ($JVcm | $JTcm); 369 ($JTcm | $H3cm) $JTcm; 370 371 # LB 27 Treat korean Syllable Block the same as ID (don't break it) 372 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm; 373 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm; 374 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm); 375 376 377 # LB 28 Do not break between alphabetics 378 # 379 $ALcm $ALcm; 380 $CM+ $ALcm; # The $CM+ is from rule 10, an unattached CM is treated as AL 381 382 # LB 29 383 $IScm $ALcm; 384 385 # LB 30 386 ($ALcm | $NUcm) $OPcm; 387 $CM+ $OPcm; # The $CM+ is from rule 10, an unattached CM is treated as AL. 388 $CPcm ($ALcm | $NUcm); 389 390 391 # 392 # Reverse Rules. 393 # 394 ## ------------------------------------------------- 395 396 !!reverse; 397 398 $CM+ $ALPlus; 399 $CM+ $BA; 400 $CM+ $HH; 401 $CM+ $BB; 402 $CM+ $B2; 403 $CM+ $CL; 404 $CM+ $CP; 405 $CM+ $EX; 406 $CM+ $GL; 407 $CM+ $HY; 408 $CM+ $H2; 409 $CM+ $H3; 410 $CM+ $ID; 411 $CM+ $IN; 412 $CM+ $IS; 413 $CM+ $JL; 414 $CM+ $JV; 415 $CM+ $JT; 416 $CM+ $NS; 417 $CM+ $NU; 418 $CM+ $OP; 419 $CM+ $PO; 420 $CM+ $PR; 421 $CM+ $QU; 422 $CM+ $SY; 423 $CM+ $WJ; 424 $CM+; 425 426 427 # 428 # Sequences of the form (shown forwards) 429 # [CANT_CM] <break> [CM] [whatever] 430 # The CM needs to behave as an AL 431 # 432 $AL_FOLLOW $CM+ / ( 433 [$BK $CR $LF $NL $ZW {eof}] | 434 $SP+ $CM+ $SP | 435 $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}])); # if LB 14 will match, need to surpress this break. 436 # LB14 says OP SP* x . 437 # becomes OP SP* x AL 438 # becomes OP SP* x CM+ AL_FOLLOW 439 # 440 # Further note: the $AL in [$AL {eof}] is only to work around 441 # a rule compiler bug which complains about 442 # empty sets otherwise. 443 444 # 445 # Sequences of the form (shown forwards) 446 # [CANT_CM] <break> [CM] <break> [PR] 447 # The CM needs to behave as an AL 448 # This rule is concerned about getting the second of the two <breaks> in place. 449 # 450 451 [$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}]; 452 453 454 455 # LB 4, 5, 5 456 457 $LB4Breaks [$LB4NonBreaks-$CM]; 458 $LB4Breaks $CM+ $CAN_CM; 459 $LF $CR; 460 461 462 # LB 7 x SP 463 # x ZW 464 [$SP $ZW] [$LB4NonBreaks-$CM]; 465 [$SP $ZW] $CM+ $CAN_CM; 466 467 # LB 8 ZW SP* <break> 468 # TODO: to implement this, we need more than one look-ahead hard break in play at a time. 469 # Requires an engine enhancement. 470 # / $SP* $ZW 471 472 # LB 9,10 Combining marks. 473 # X $CM needs to behave like X, where X is not $SP or controls. 474 # $CM not covered by the above needs to behave like $AL 475 # Stick together any combining sequences that don't match other rules. 476 $CM+ $CAN_CM; 477 478 479 # LB 11 480 $CM* $WJ $CM* $CAN_CM; 481 $CM* $WJ [$LB8NonBreaks-$CM]; 482 483 $CANT_CM $CM* $WJ; 484 $CM* $CAN_CM $CM* $WJ; 485 486 # LB 12a 487 # [^SP BA HY] x GL 488 # 489 $CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HH $HY]]; 490 491 # LB 12 492 # GL x 493 # 494 $CANT_CM $CM* $GL; 495 $CM* $CAN_CM $CM* $GL; 496 497 498 # LB 13 499 $CL $CM+ $CAN_CM; 500 $CP $CM+ $CAN_CM; 501 $EX $CM+ $CAN_CM; 502 $IS $CM+ $CAN_CM; 503 $SY $CM+ $CAN_CM; 504 505 $CL [$LB8NonBreaks-$CM]; 506 $CP [$LB8NonBreaks-$CM]; 507 $EX [$LB8NonBreaks-$CM]; 508 $IS [$LB8NonBreaks-$CM]; 509 $SY [$LB8NonBreaks-$CM]; 510 511 # Rule 13 & 14 taken together for an edge case. 512 # Match this, shown forward 513 # OP SP+ ($CM+ behaving as $AL) (CL | CP | EX | IS | IY) 514 # This really wants to chain at the $CM+ (which is acting as an $AL) 515 # except for $CM chaining being disabled. 516 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP; 517 518 # LB 14 OP SP* x 519 # 520 $CM* $CAN_CM $SP* $CM* $OP; 521 $CANT_CM $SP* $CM* $OP; 522 $AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP 523 524 $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP; 525 $CM* $AL_FOLLOW_CM $CM+ $SP+ $CM* $OP; 526 $SY $CM $SP+ $OP; # TODO: Experiment. Remove. 527 528 529 530 # LB 15 531 $CM* $OP $SP* $CM* $QU; 532 533 # LB 16 534 $CM* $NS $SP* $CM* ($CL | $CP); 535 536 # LB 17 537 $CM* $B2 $SP* $CM* $B2; 538 539 # LB 18 break after spaces 540 # Nothing explicit needed here. 541 542 543 # 544 # LB 19 545 # 546 $CM* $QU $CM* $CAN_CM; # . x QU 547 $CM* $QU $LB18NonBreaks; 548 549 550 $CM* $CAN_CM $CM* $QU; # QU x . 551 $CANT_CM $CM* $QU; 552 553 # 554 # LB 20 Break before and after CB. 555 # nothing needed here. 556 # 557 558 # LB 20.09 added rule for Finnish tailoring 559 $AL ($HY | $HH) / $SP; 560 561 # LB 21 562 $CM* ($BA | $HH | $HY | $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA | HY | NS) 563 564 $CM* [$LB20NonBreaks-$CM] $CM* $BB; # BB x . 565 [^$CB] $CM* $BB; # 566 567 568 569 # LB 22 570 $CM* $IN $CM* $ALPlus; 571 $CM* $IN $CM* $ID; 572 $CM* $IN $CM* $IN; 573 $CM* $IN $CM* $NU; 574 575 # LB 23 576 $CM* $PO $CM* $ID; 577 $CM* $NU $CM* $ALPlus; 578 $CM* $ALPlus $CM* $NU; 579 580 # LB 24 581 $CM* $ID $CM* $PR; 582 $CM* $ALPlus $CM* $PR; 583 $CM* $ALPlus $CM* $PO; 584 585 586 # LB 25 587 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?; 588 589 # LB 26 590 $CM* ($H3 | $H2 | $JV | $JL) $CM* $JL; 591 $CM* ($JT | $JV) $CM* ($H2 | $JV); 592 $CM* $JT $CM* ($H3 | $JT); 593 594 # LB 27 595 $CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL); 596 $CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL); 597 $CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR; 598 599 # LB 28 600 $CM* $ALPlus $CM* $ALPlus; 601 602 603 # LB 29 604 $CM* $ALPlus $CM* $IS; 605 606 # LB 30 607 $CM* $OP $CM* ($ALPlus | $NU); 608 $CM* ($ALPlus | $NU) $CM* $CP; 609 610 611 ## ------------------------------------------------- 612 613 !!safe_reverse; 614 615 # LB 9 616 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP]; 617 $CM+ $SP / .; 618 619 # LB 14 620 $SP+ $CM* $OP; 621 622 # LB 15 623 $SP+ $CM* $QU; 624 625 # LB 16 626 $SP+ $CM* ($CL | $CP); 627 628 # LB 17 629 $SP+ $CM* $B2; 630 631 # LB 25 632 ($CM* ($IS | $SY))+ $CM* $NU; 633 ($CL | $CP) $CM* ($NU | $IS | $SY); 634 635 # For dictionary-based break 636 $dictionary $dictionary; 637 638 ## ------------------------------------------------- 639 640 !!safe_forward; 641 642 # Skip forward over all character classes that are involved in 643 # rules containing patterns with possibly more than one char 644 # of context. 645 # 646 # It might be slightly more efficient to have specific rules 647 # instead of one generic one, but only if we could 648 # turn off rule chaining. We don't want to move more 649 # than necessary. 650 # 651 [$CM $OP $QU $CL $CP $B2 $PR $HY $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $dictionary]; 652 $dictionary $dictionary; 653 654
