line_ja.txt revision c73f511526464f8e56c242df80552e9b0d94ae3d 
1# Copyright (c) 2002-2013  International Business Machines Corporation and
2# others. All Rights Reserved.
3#
4#  file:  line_ja.txt
5#
6#         Line Breaking Rules
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/
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$BA = [:LineBreak =  Break_After:];
62$BB = [:LineBreak =  Break_Before:];
63$BK = [:LineBreak =  Mandatory_Break:];
64$B2 = [:LineBreak =  Break_Both:];
65$CB = [:LineBreak =  Contingent_Break:];
66$CJ = [:LineBreak =  Conditional_Japanese_Starter:];
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$HL = [:LineBreak =  Hebrew_Letter:];
74$HY = [:LineBreak =  Hyphen:];
75$H2 = [:LineBreak =  H2:];
76$H3 = [:LineBreak =  H3:];
77$ID = [[:LineBreak =  Ideographic:] $CJ];
78$IN = [:LineBreak =  Inseperable:];
79$IS = [:LineBreak =  Infix_Numeric:];
80$JL = [:LineBreak =  JL:];
81$JV = [:LineBreak =  JV:];
82$JT = [:LineBreak =  JT:];
83$LF = [:LineBreak =  Line_Feed:];
84$NL = [:LineBreak =  Next_Line:];
85$NS = [:LineBreak =  Nonstarter:];
86$NU = [:LineBreak =  Numeric:];
87$OP = [:LineBreak =  Open_Punctuation:];
88$PO = [:LineBreak =  Postfix_Numeric:];
89$PR = [:LineBreak =  Prefix_Numeric:];
90$QU = [:LineBreak =  Quotation:];
91$RI = [:LineBreak =  Regional_Indicator:];
92$SA = [:LineBreak =  Complex_Context:];
93$SG = [:LineBreak =  Surrogate:];
94$SP = [:LineBreak =  Space:];
95$SY = [:LineBreak =  Break_Symbols:];
96$WJ = [:LineBreak =  Word_Joiner:];
97$XX = [:LineBreak =  Unknown:];
98$ZW = [:LineBreak =  ZWSpace:];
99
100#   Dictionary character set, for triggering language-based break engines. Currently
101#   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
102#   5.0 or later as the definition of Complex_Context was corrected to include all
103#   characters requiring dictionary break.
104
105$dictionary = [:LineBreak = Complex_Context:];
106
107#
108#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
109#                               SA  (South East Asian: Thai, Lao, Khmer)
110#                               SG  (Unpaired Surrogates)
111#                               XX  (Unknown, unassigned)
112#                         as $AL  (Alphabetic)
113#
114$ALPlus = [$AL $AI $SA $SG $XX];
115
116#
117#  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
118#
119$ALcm = $ALPlus $CM*;
120$BAcm = $BA $CM*;
121$BBcm = $BB $CM*;
122$B2cm = $B2 $CM*;
123$CLcm = $CL $CM*;
124$CPcm = $CP $CM*;
125$EXcm = $EX $CM*;
126$GLcm = $GL $CM*;
127$HLcm = $HL $CM*;
128$HYcm = $HY $CM*;
129$H2cm = $H2 $CM*;
130$H3cm = $H3 $CM*;
131$IDcm = $ID $CM*;
132$INcm = $IN $CM*;
133$IScm = $IS $CM*;
134$JLcm = $JL $CM*;
135$JVcm = $JV $CM*;
136$JTcm = $JT $CM*;
137$NScm = $NS $CM*;
138$NUcm = $NU $CM*;
139$OPcm = $OP $CM*;
140$POcm = $PO $CM*;
141$PRcm = $PR $CM*;
142$QUcm = $QU $CM*;
143$RIcm = $RI $CM*;
144$SYcm = $SY $CM*;
145$WJcm = $WJ $CM*;
146
147## -------------------------------------------------
148
149!!forward;
150
151#
152#  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
153#
154$ALPlus $CM+;
155$BA $CM+;
156$BB $CM+;
157$B2 $CM+;
158$CL $CM+;
159$CP $CM+;
160$EX $CM+;
161$GL $CM+;
162$HL $CM+;
163$HY $CM+;
164$H2 $CM+;
165$H3 $CM+;
166$ID $CM+;
167$IN $CM+;
168$IS $CM+;
169$JL $CM+;
170$JV $CM+;
171$JT $CM+;
172$NS $CM+;
173$NU $CM+;
174$OP $CM+;
175$PO $CM+;
176$PR $CM+;
177$QU $CM+;
178$RI $CM+;
179$SY $CM+;
180$WJ $CM+;
181
182#
183# CAN_CM  is the set of characters that may combine with CM combining chars.
184#         Note that Linebreak UAX 14's concept of a combining char and the rules
185#         for what they can combine with are _very_ different from the rest of Unicode.
186#
187#         Note that $CM itself is left out of this set.  If CM is needed as a base
188#         it must be listed separately in the rule.
189#
190$CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
191$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
192
193#
194# AL_FOLLOW  set of chars that can unconditionally follow an AL
195#            Needed in rules where stand-alone $CM s are treated as AL.
196#            Chaining is disabled with CM because it causes other failures,
197#            so for this one case we need to manually list out longer sequences.
198#
199$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
200$AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
201$AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
202
203
204#
205#  Rule LB 4, 5    Mandatory (Hard) breaks.
206#
207$LB4Breaks    = [$BK $CR $LF $NL];
208$LB4NonBreaks = [^$BK $CR $LF $NL];
209$CR $LF {100};
210
211#
212#  LB 6    Do not break before hard line breaks.
213#
214$LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
215$CAN_CM $CM*    $LB4Breaks {100};
216$CM+            $LB4Breaks {100};
217
218# LB 7         x SP
219#              x ZW
220$LB4NonBreaks [$SP $ZW];
221$CAN_CM $CM*  [$SP $ZW];
222$CM+          [$SP $ZW];
223
224#
225# LB 8         Break after zero width space
226#              TODO:  ZW SP* <break>
227#              An engine change is required to write the reverse rule for this.
228#              For now, leave the Unicode 5.2 rule, ZW <break>
229#
230$LB8Breaks    = [$LB4Breaks $ZW];
231$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
232
233
234# LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL 
235#                                $CM not covered by the above needs to behave like $AL   
236#                                See definition of $CAN_CM.
237
238$CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
239$CM+;
240
241#
242# LB 11  Do not break before or after WORD JOINER & related characters.
243#
244$CAN_CM $CM*  $WJcm;
245$LB8NonBreaks $WJcm;
246$CM+          $WJcm;
247
248$WJcm $CANT_CM;
249$WJcm $CAN_CM $CM*;
250
251#
252# LB 12  Do not break after NBSP and related characters.
253#         GL  x
254#
255$GLcm $CAN_CM $CM*;
256$GLcm $CANT_CM;
257 
258#
259# LB 12a  Do not break before NBSP and related characters ...
260#            [^SP BA HY] x GL
261#
262[[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
263$CM+ GLcm;
264
265
266
267#
268# LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
269#
270$LB8NonBreaks $CL;
271$CAN_CM $CM*  $CL;
272$CM+          $CL;              # by rule 10, stand-alone CM behaves as AL
273
274$LB8NonBreaks $CP;
275$CAN_CM $CM*  $CP;
276$CM+          $CP;              # by rule 10, stand-alone CM behaves as AL
277
278$LB8NonBreaks $EX;
279$CAN_CM $CM*  $EX;
280$CM+          $EX;              # by rule 10, stand-alone CM behaves as AL
281
282$LB8NonBreaks $IS;
283$CAN_CM $CM*  $IS;
284$CM+          $IS;              # by rule 10, stand-alone CM behaves as AL
285
286$LB8NonBreaks $SY;
287$CAN_CM $CM*  $SY;
288$CM+          $SY;              # by rule 10, stand-alone CM behaves as AL
289
290
291#
292# LB 14  Do not break after OP, even after spaces
293#
294$OPcm $SP* $CAN_CM $CM*;
295$OPcm $SP* $CANT_CM;
296
297$OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
298
299# LB 15
300$QUcm $SP* $OPcm;
301
302# LB 16
303($CLcm | $CPcm) $SP* $NScm;
304
305# LB 17
306$B2cm $SP* $B2cm;
307
308#
309# LB 18  Break after spaces.
310#
311$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
312$LB18Breaks    = [$LB8Breaks $SP];
313
314
315# LB 19
316#         x QU
317$LB18NonBreaks $CM* $QUcm;
318$CM+                $QUcm;
319
320#         QU  x
321$QUcm .?;
322$QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
323                              #  TODO:  I don't think this rule is needed.
324
325
326# LB 20
327#        <break>  $CB
328#        $CB   <break>
329
330$LB20NonBreaks = [$LB18NonBreaks - $CB];
331
332# LB 21        x   (BA | HY | NS)
333#           BB x
334#
335$LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm); 
336
337$BBcm [^$CB];                                  #  $BB  x
338$BBcm $LB20NonBreaks $CM*;
339
340# LB 21a Don't break after Hebrew + Hyphen
341#   HL (HY | BA) x
342#  
343$HLcm ($HYcm | $BAcm) [^$CB]?;
344
345# LB 21b (forward) Don't break between SY and HL
346# (break between HL and SY already disallowed by LB 13 above)
347$SYcm $HLcm;
348
349# LB 22
350($ALcm | $HLcm) $INcm;
351$CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
352$IDcm    $INcm;
353$INcm    $INcm;
354$NUcm    $INcm;
355
356
357# $LB 23
358$IDcm  $POcm;
359$ALcm  $NUcm;       # includes $LB19
360$HLcm  $NUcm;
361$CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
362$NUcm  $ALcm;
363$NUcm  $HLcm;
364
365#
366# LB 24
367#
368$PRcm $IDcm;
369$PRcm ($ALcm | $HLcm);
370$POcm ($ALcm | $HLcm);
371
372#
373# LB 25   Numbers.
374#
375($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
376
377# LB 26  Do not break a Korean syllable
378#
379$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
380($JVcm | $H2cm) ($JVcm | $JTcm);
381($JTcm | $H3cm) $JTcm;
382
383# LB 27  Treat korean Syllable Block the same as ID  (don't break it)
384($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
385($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
386$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
387
388
389# LB 28   Do not break between alphabetics
390#
391($ALcm | $HLcm) ($ALcm | $HLcm);
392$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
393
394# LB 29
395$IScm ($ALcm | $HLcm);
396
397# LB 30
398($ALcm | $HLcm | $NUcm) $OPcm;
399$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.          
400$CPcm ($ALcm | $HLcm | $NUcm);
401
402# LB 30a  Do not break between regional indicators.
403$RIcm $RIcm;
404
405#
406#  Reverse Rules.
407#
408## -------------------------------------------------
409
410!!reverse;
411
412$CM+ $ALPlus;
413$CM+ $BA;
414$CM+ $BB;
415$CM+ $B2;
416$CM+ $CL;
417$CM+ $CP;
418$CM+ $EX;
419$CM+ $GL;
420$CM+ $HL;
421$CM+ $HY;
422$CM+ $H2;
423$CM+ $H3;
424$CM+ $ID;
425$CM+ $IN;
426$CM+ $IS;
427$CM+ $JL;
428$CM+ $JV;
429$CM+ $JT;
430$CM+ $NS;
431$CM+ $NU;
432$CM+ $OP;
433$CM+ $PO;
434$CM+ $PR;
435$CM+ $QU;
436$CM+ $RI;
437$CM+ $SY;
438$CM+ $WJ;
439$CM+;
440
441
442#
443#  Sequences of the form  (shown forwards)
444#      [CANT_CM]  <break>  [CM]  [whatever]
445#  The CM needs to behave as an AL
446#
447$AL_FOLLOW $CM+ / (
448          [$BK $CR $LF $NL $ZW {eof}] |
449          $SP+ $CM+ $SP |
450          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
451                                               #  LB14 says    OP SP* x .        
452                                               #    becomes    OP SP* x AL
453                                               #    becomes    OP SP* x CM+ AL_FOLLOW
454                                               #
455                                               # Further note:  the $AL in [$AL {eof}] is only to work around
456                                               #                a rule compiler bug which complains about
457                                               #                empty sets otherwise.
458          
459#
460#  Sequences of the form  (shown forwards)
461#      [CANT_CM]  <break> [CM]  <break>  [PR]
462#  The CM needs to behave as an AL
463#  This rule is concerned about getting the second of the two <breaks> in place.
464#
465
466[$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
467
468
469
470# LB 4, 5, 5
471
472$LB4Breaks [$LB4NonBreaks-$CM];
473$LB4Breaks $CM+ $CAN_CM;
474$LF $CR;
475
476
477# LB 7         x SP
478#              x ZW
479[$SP $ZW] [$LB4NonBreaks-$CM];
480[$SP $ZW] $CM+ $CAN_CM;
481
482# LB 8 ZW SP* <break>
483#     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
484#           Requires an engine enhancement.
485#   / $SP* $ZW
486
487# LB 9,10  Combining marks.
488#    X   $CM needs to behave like X, where X is not $SP or controls.
489#    $CM not covered by the above needs to behave like $AL
490# Stick together any combining sequences that don't match other rules.
491$CM+ $CAN_CM;
492
493
494# LB 11
495$CM* $WJ $CM* $CAN_CM;
496$CM* $WJ      [$LB8NonBreaks-$CM];
497
498     $CANT_CM $CM* $WJ;
499$CM* $CAN_CM  $CM* $WJ;
500
501# LB 12a
502#      [^SP BA HY] x GL
503#
504$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
505
506# LB 12
507#     GL  x
508#
509$CANT_CM $CM* $GL;
510$CM* $CAN_CM $CM* $GL;
511
512
513# LB 13
514$CL $CM+ $CAN_CM;
515$CP $CM+ $CAN_CM;
516$EX $CM+ $CAN_CM;
517$IS $CM+ $CAN_CM;
518$SY $CM+ $CAN_CM;
519
520$CL [$LB8NonBreaks-$CM];
521$CP [$LB8NonBreaks-$CM];
522$EX [$LB8NonBreaks-$CM];
523$IS [$LB8NonBreaks-$CM];
524$SY [$LB8NonBreaks-$CM];
525
526# Rule 13 & 14 taken together for an edge case.
527#   Match this, shown forward
528#     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
529#   This really wants to chain at the $CM+ (which is acting as an $AL)
530#   except for $CM chaining being disabled.
531[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;  
532
533# LB 14    OP SP* x
534#
535$CM* $CAN_CM    $SP* $CM* $OP;
536     $CANT_CM   $SP* $CM* $OP;
537$AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
538     
539     $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
540$CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
541$SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
542
543
544
545# LB 15
546$CM* $OP $SP* $CM* $QU;
547
548# LB 16
549$CM* $NS $SP* $CM* ($CL | $CP);
550
551# LB 17
552$CM* $B2 $SP* $CM* $B2;
553
554# LB 18  break after spaces
555#        Nothing explicit needed here.
556
557
558#
559# LB 19
560#
561$CM* $QU $CM* $CAN_CM;                                #   . x QU
562$CM* $QU      $LB18NonBreaks;
563
564
565$CM* $CAN_CM  $CM* $QU;                               #   QU x .
566     $CANT_CM $CM* $QU;
567     
568#
569#  LB 20  Break before and after CB.
570#         nothing needed here.
571#
572
573# LB 21
574$CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
575
576$CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
577[^$CB] $CM* $BB;                                      # 
578
579# LB21a
580[^$CB] $CM* ($HY | $BA) $CM* $HL;
581
582# LB21b (reverse)
583$CM* $HL $CM* $SY;
584
585# LB 22
586$CM* $IN $CM* ($ALPlus | $HL);
587$CM* $IN $CM* $ID;
588$CM* $IN $CM* $IN;
589$CM* $IN $CM* $NU;
590
591# LB 23
592$CM* $PO $CM* $ID;
593$CM* $NU $CM* ($ALPlus | $HL);
594$CM* ($ALPlus | $HL) $CM* $NU;
595
596# LB 24
597$CM* $ID $CM* $PR;
598$CM* ($ALPlus | $HL) $CM* $PR;
599$CM* ($ALPlus | $HL) $CM* $PO;
600
601
602# LB 25
603($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
604
605# LB 26
606$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
607$CM* ($JT | $JV) $CM* ($H2 | $JV);
608$CM* $JT $CM* ($H3 | $JT);
609
610# LB 27
611$CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
612$CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
613$CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
614
615# LB 28
616$CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
617
618
619# LB 29
620$CM* ($ALPlus | $HL) $CM* $IS;
621
622# LB 30
623$CM* $OP $CM* ($ALPlus | $HL | $NU);
624$CM* ($ALPlus | $HL | $NU) $CM* $CP;
625
626# LB 30a
627$CM* $RI $CM* $RI;
628
629## -------------------------------------------------
630
631!!safe_reverse;
632
633# LB 9
634$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
635$CM+ $SP / .;
636
637# LB 14
638$SP+ $CM* $OP;
639
640# LB 15
641$SP+ $CM* $QU;
642
643# LB 16
644$SP+ $CM* ($CL | $CP);
645
646# LB 17
647$SP+ $CM* $B2;
648
649# LB 21
650$CM* ($HY | $BA) $CM* $HL;
651
652# LB 25
653($CM* ($IS | $SY))+ $CM* $NU;
654($CL | $CP) $CM* ($NU | $IS | $SY);
655
656# For dictionary-based break
657$dictionary $dictionary;
658
659## -------------------------------------------------
660
661!!safe_forward;
662
663# Skip forward over all character classes that are involved in
664#   rules containing patterns with possibly more than one char
665#   of context.
666#
667#  It might be slightly more efficient to have specific rules
668#  instead of one generic one, but only if we could
669#  turn off rule chaining.  We don't want to move more
670#  than necessary.
671#
672[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
673$dictionary $dictionary;
674
675