1 # Copyright (c) 2002-2013 International Business Machines Corporation and
2 # others. All Rights Reserved.
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/
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.
15 # CHROME: 1. Use line_ja.txt to apply small kana rules in all locales.
16 # 2. Adjust CL, OP, and IS to handle 'comma-variants' consistently.
24 # !!lookAheadHardBreak Described here because it is (as yet) undocumented elsewhere
25 # and only used for the line break rules.
27 # It is used in the implementation of rule LB 10
28 # which says to treat any combining mark that is not attached to a base
29 # character as if it were of class AL (alphabetic).
31 # The problem occurs in the reverse rules.
33 # Consider a sequence like, with correct breaks as shown
36 # Then consider the sequence without the initial ID (ideographic)
39 # Our CM, which in the first example was attached to the ideograph,
40 # is now unattached, becomes an alpha, and joins in with the other
43 # When iterating forwards, these sequences do not present any problems
44 # When iterating backwards, we need to look ahead when encountering
45 # a CM to see whether it attaches to something further on or not.
46 # (Look-ahead in a reverse rule is looking towards the start)
48 # If the CM is unattached, we need to force a break.
50 # !!lookAheadHardBreak forces the run time state machine to
51 # stop immediately when a look ahead rule ( '/' operator) matches,
52 # and set the match position to that of the look-ahead operator,
53 # no matter what other rules may be in play at the time.
55 # See rule LB 19 for an example.
58 $AI = [:LineBreak = Ambiguous:];
59 $AL = [[:LineBreak = Alphabetic:] - [\u23B4\u23B5]];
60 $BA = [:LineBreak = Break_After:];
61 $BB = [:LineBreak = Break_Before:];
62 $BK = [:LineBreak = Mandatory_Break:];
63 $B2 = [:LineBreak = Break_Both:];
64 $CB = [:LineBreak = Contingent_Break:];
65 $CJ = [:LineBreak = Conditional_Japanese_Starter:];
66 $CL = [[:LineBreak = Close_Punctuation:] [\uFE51\uFE10\u23B5]];
67 $CM = [:LineBreak = Combining_Mark:];
68 $CP = [:LineBreak = Close_Parenthesis:];
69 $CR = [:LineBreak = Carriage_Return:];
70 $EX = [:LineBreak = Exclamation:];
71 $GL = [:LineBreak = Glue:];
72 $HL = [:LineBreak = Hebrew_Letter:];
73 $HY = [:LineBreak = Hyphen:];
74 $H2 = [:LineBreak = H2:];
75 $H3 = [:LineBreak = H3:];
76 $ID = [[[:LineBreak = Ideographic:] $CJ] - [\uFE51]];
77 $IN = [:LineBreak = Inseperable:];
78 $IS = [[:LineBreak = Infix_Numeric:] - [\uFE10]];
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:] \u23B4];
87 $PO = [:LineBreak = Postfix_Numeric:];
88 $PR = [:LineBreak = Prefix_Numeric:];
89 $QU = [:LineBreak = Quotation:];
90 $RI = [:LineBreak = Regional_Indicator:];
91 $SA = [:LineBreak = Complex_Context:];
92 $SG = [:LineBreak = Surrogate:];
93 $SP = [:LineBreak = Space:];
94 $SY = [:LineBreak = Break_Symbols:];
95 $WJ = [:LineBreak = Word_Joiner:];
96 $XX = [:LineBreak = Unknown:];
97 $ZW = [:LineBreak = ZWSpace:];
99 # Dictionary character set, for triggering language-based break engines. Currently
100 # limited to LineBreak=Complex_Context. Note that this set only works in Unicode
101 # 5.0 or later as the definition of Complex_Context was corrected to include all
102 # characters requiring dictionary break.
104 $dictionary = [:LineBreak = Complex_Context:];
107 # Rule LB1. By default, treat AI (characters with ambiguous east Asian width),
108 # SA (South East Asian: Thai, Lao, Khmer)
109 # SG (Unpaired Surrogates)
110 # XX (Unknown, unassigned)
111 # as $AL (Alphabetic)
113 $ALPlus = [$AL $AI $SA $SG $XX];
116 # Combining Marks. X $CM* behaves as if it were X. Rule LB6.
118 $ALcm = $ALPlus $CM*;
146 ## -------------------------------------------------
151 # Each class of character can stand by itself as an unbroken token, with trailing combining stuff
182 # CAN_CM is the set of characters that may combine with CM combining chars.
183 # Note that Linebreak UAX 14's concept of a combining char and the rules
184 # for what they can combine with are _very_ different from the rest of Unicode.
186 # Note that $CM itself is left out of this set. If CM is needed as a base
187 # it must be listed separately in the rule.
189 $CAN_CM = [^$SP $BK $CR $LF $NL $ZW $CM]; # Bases that can take CMs
190 $CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM]; # Bases that can't take CMs
193 # AL_FOLLOW set of chars that can unconditionally follow an AL
194 # Needed in rules where stand-alone $CM s are treated as AL.
195 # Chaining is disabled with CM because it causes other failures,
196 # so for this one case we need to manually list out longer sequences.
198 $AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
199 $AL_FOLLOW_CM = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
200 $AL_FOLLOW = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
204 # Rule LB 4, 5 Mandatory (Hard) breaks.
206 $LB4Breaks = [$BK $CR $LF $NL];
207 $LB4NonBreaks = [^$BK $CR $LF $NL];
211 # LB 6 Do not break before hard line breaks.
213 $LB4NonBreaks? $LB4Breaks {100}; # LB 5 do not break before hard breaks.
214 $CAN_CM $CM* $LB4Breaks {100};
215 $CM+ $LB4Breaks {100};
219 $LB4NonBreaks [$SP $ZW];
220 $CAN_CM $CM* [$SP $ZW];
224 # LB 8 Break after zero width space
225 # TODO: ZW SP* <break>
226 # An engine change is required to write the reverse rule for this.
227 # For now, leave the Unicode 5.2 rule, ZW <break>
229 $LB8Breaks = [$LB4Breaks $ZW];
230 $LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
233 # LB 9 Combining marks. X $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
234 # $CM not covered by the above needs to behave like $AL
235 # See definition of $CAN_CM.
237 $CAN_CM $CM+; # Stick together any combining sequences that don't match other rules.
241 # LB 11 Do not break before or after WORD JOINER & related characters.
251 # LB 12 Do not break after NBSP and related characters.
258 # LB 12a Do not break before NBSP and related characters ...
261 [[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
267 # LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces.
271 $CM+ $CL; # by rule 10, stand-alone CM behaves as AL
275 $CM+ $CP; # by rule 10, stand-alone CM behaves as AL
279 $CM+ $EX; # by rule 10, stand-alone CM behaves as AL
283 $CM+ $IS; # by rule 10, stand-alone CM behaves as AL
287 $CM+ $SY; # by rule 10, stand-alone CM behaves as AL
291 # LB 14 Do not break after OP, even after spaces
293 $OPcm $SP* $CAN_CM $CM*;
296 $OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL
302 ($CLcm | $CPcm) $SP* $NScm;
308 # LB 18 Break after spaces.
310 $LB18NonBreaks = [$LB8NonBreaks - [$SP]];
311 $LB18Breaks = [$LB8Breaks $SP];
316 $LB18NonBreaks $CM* $QUcm;
321 $QUcm $LB18NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc.
322 # TODO: I don't think this rule is needed.
329 $LB20NonBreaks = [$LB18NonBreaks - $CB];
331 # LB 21 x (BA | HY | NS)
334 $LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
336 $BBcm [^$CB]; # $BB x
337 $BBcm $LB20NonBreaks $CM*;
339 # LB 21a Don't break after Hebrew + Hyphen
342 $HLcm ($HYcm | $BAcm) [^$CB]?;
344 # LB 21b (forward) Don't break between SY and HL
345 # (break between HL and SY already disallowed by LB 13 above)
349 ($ALcm | $HLcm) $INcm;
350 $CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL
358 $ALcm $NUcm; # includes $LB19
360 $CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL
368 $PRcm ($ALcm | $HLcm);
369 $POcm ($ALcm | $HLcm);
374 ($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
376 # LB 26 Do not break a Korean syllable
378 $JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
379 ($JVcm | $H2cm) ($JVcm | $JTcm);
380 ($JTcm | $H3cm) $JTcm;
382 # LB 27 Treat korean Syllable Block the same as ID (don't break it)
383 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
384 ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
385 $PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
388 # LB 28 Do not break between alphabetics
390 ($ALcm | $HLcm) ($ALcm | $HLcm);
391 $CM+ ($ALcm | $HLcm); # The $CM+ is from rule 10, an unattached CM is treated as AL
394 $IScm ($ALcm | $HLcm);
397 ($ALcm | $HLcm | $NUcm) $OPcm;
398 $CM+ $OPcm; # The $CM+ is from rule 10, an unattached CM is treated as AL.
399 $CPcm ($ALcm | $HLcm | $NUcm);
401 # LB 30a Do not break between regional indicators.
407 ## -------------------------------------------------
442 # Sequences of the form (shown forwards)
443 # [CANT_CM] <break> [CM] [whatever]
444 # The CM needs to behave as an AL
447 [$BK $CR $LF $NL $ZW {eof}] |
449 $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}])); # if LB 14 will match, need to surpress this break.
450 # LB14 says OP SP* x .
451 # becomes OP SP* x AL
452 # becomes OP SP* x CM+ AL_FOLLOW
454 # Further note: the $AL in [$AL {eof}] is only to work around
455 # a rule compiler bug which complains about
456 # empty sets otherwise.
459 # Sequences of the form (shown forwards)
460 # [CANT_CM] <break> [CM] <break> [PR]
461 # The CM needs to behave as an AL
462 # This rule is concerned about getting the second of the two <breaks> in place.
465 [$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
471 $LB4Breaks [$LB4NonBreaks-$CM];
472 $LB4Breaks $CM+ $CAN_CM;
478 [$SP $ZW] [$LB4NonBreaks-$CM];
479 [$SP $ZW] $CM+ $CAN_CM;
481 # LB 8 ZW SP* <break>
482 # TODO: to implement this, we need more than one look-ahead hard break in play at a time.
483 # Requires an engine enhancement.
486 # LB 9,10 Combining marks.
487 # X $CM needs to behave like X, where X is not $SP or controls.
488 # $CM not covered by the above needs to behave like $AL
489 # Stick together any combining sequences that don't match other rules.
494 $CM* $WJ $CM* $CAN_CM;
495 $CM* $WJ [$LB8NonBreaks-$CM];
498 $CM* $CAN_CM $CM* $WJ;
503 $CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
509 $CM* $CAN_CM $CM* $GL;
519 $CL [$LB8NonBreaks-$CM];
520 $CP [$LB8NonBreaks-$CM];
521 $EX [$LB8NonBreaks-$CM];
522 $IS [$LB8NonBreaks-$CM];
523 $SY [$LB8NonBreaks-$CM];
525 # Rule 13 & 14 taken together for an edge case.
526 # Match this, shown forward
527 # OP SP+ ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
528 # This really wants to chain at the $CM+ (which is acting as an $AL)
529 # except for $CM chaining being disabled.
530 [$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
534 $CM* $CAN_CM $SP* $CM* $OP;
535 $CANT_CM $SP* $CM* $OP;
536 $AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
538 $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
539 $CM* $AL_FOLLOW_CM $CM+ $SP+ $CM* $OP;
540 $SY $CM $SP+ $OP; # TODO: Experiment. Remove.
545 $CM* $OP $SP* $CM* $QU;
548 $CM* $NS $SP* $CM* ($CL | $CP);
551 $CM* $B2 $SP* $CM* $B2;
553 # LB 18 break after spaces
554 # Nothing explicit needed here.
560 $CM* $QU $CM* $CAN_CM; # . x QU
561 $CM* $QU $LB18NonBreaks;
564 $CM* $CAN_CM $CM* $QU; # QU x .
568 # LB 20 Break before and after CB.
569 # nothing needed here.
573 $CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA | HY | NS)
575 $CM* [$LB20NonBreaks-$CM] $CM* $BB; # BB x .
579 [^$CB] $CM* ($HY | $BA) $CM* $HL;
585 $CM* $IN $CM* ($ALPlus | $HL);
592 $CM* $NU $CM* ($ALPlus | $HL);
593 $CM* ($ALPlus | $HL) $CM* $NU;
597 $CM* ($ALPlus | $HL) $CM* $PR;
598 $CM* ($ALPlus | $HL) $CM* $PO;
602 ($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
605 $CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
606 $CM* ($JT | $JV) $CM* ($H2 | $JV);
607 $CM* $JT $CM* ($H3 | $JT);
610 $CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
611 $CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
612 $CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
615 $CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
619 $CM* ($ALPlus | $HL) $CM* $IS;
622 $CM* $OP $CM* ($ALPlus | $HL | $NU);
623 $CM* ($ALPlus | $HL | $NU) $CM* $CP;
628 ## -------------------------------------------------
633 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
643 $SP+ $CM* ($CL | $CP);
649 $CM* ($HY | $BA) $CM* $HL;
652 ($CM* ($IS | $SY))+ $CM* $NU;
653 ($CL | $CP) $CM* ($NU | $IS | $SY);
655 # For dictionary-based break
656 $dictionary $dictionary;
658 ## -------------------------------------------------
662 # Skip forward over all character classes that are involved in
663 # rules containing patterns with possibly more than one char
666 # It might be slightly more efficient to have specific rules
667 # instead of one generic one, but only if we could
668 # turn off rule chaining. We don't want to move more
671 [$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
672 $dictionary $dictionary;