1# Copyright (c) 2002-2015  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 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#         This corresponds to CSS line-break=strict (BCP47 -u-lb-strict).
16#         It sets characters of class CJ to behave like NS.
17
18#
19#  Character Classes defined by TR 14.
20#
21
22!!chain;
23!!LBCMNoChain;
24
25
26!!lookAheadHardBreak;
27#
28#  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
29#                          and only used for the line break rules.
30#
31#           It is used in the implementation of rule LB 10
32#           which says to treat any combining mark that is not attached to a base
33#           character as if it were of class AL  (alphabetic).
34#
35#           The problem occurs in the reverse rules.
36#
37#           Consider a sequence like, with correct breaks as shown
38#               LF  ID  CM  AL  AL
39#                  ^       ^       ^
40#           Then consider the sequence without the initial ID (ideographic)
41#                 LF  CM  AL  AL
42#                    ^           ^
43#           Our CM, which in the first example was attached to the ideograph,
44#           is now unattached, becomes an alpha, and joins in with the other
45#           alphas.
46#
47#           When iterating forwards, these sequences do not present any problems
48#           When iterating backwards, we need to look ahead when encountering
49#           a CM to see whether it attaches to something further on or not.
50#           (Look-ahead in a reverse rule is looking towards the start)
51#
52#           If the CM is unattached, we need to force a break.
53#
54#           !!lookAheadHardBreak forces the run time state machine to
55#           stop immediately when a look ahead rule ( '/' operator) matches,
56#           and set the match position to that of the look-ahead operator,
57#           no matter what other rules may be in play at the time.
58#
59#           See rule LB 19 for an example.
60#
61
62$AI = [:LineBreak =  Ambiguous:];
63$AL = [:LineBreak =  Alphabetic:];
64$BA = [:LineBreak =  Break_After:];
65$BB = [:LineBreak =  Break_Before:];
66$BK = [:LineBreak =  Mandatory_Break:];
67$B2 = [:LineBreak =  Break_Both:];
68$CB = [:LineBreak =  Contingent_Break:];
69$CJ = [:LineBreak =  Conditional_Japanese_Starter:];
70$CL = [:LineBreak =  Close_Punctuation:];
71$CM = [:LineBreak =  Combining_Mark:];
72$CP = [:LineBreak =  Close_Parenthesis:];
73$CR = [:LineBreak =  Carriage_Return:];
74$EX = [:LineBreak =  Exclamation:];
75$GL = [:LineBreak =  Glue:];
76$HL = [:LineBreak =  Hebrew_Letter:];
77$HY = [:LineBreak =  Hyphen:];
78$H2 = [:LineBreak =  H2:];
79$H3 = [:LineBreak =  H3:];
80$ID = [:LineBreak =  Ideographic:];
81$IN = [:LineBreak =  Inseperable:];
82$IS = [:LineBreak =  Infix_Numeric:];
83$JL = [:LineBreak =  JL:];
84$JV = [:LineBreak =  JV:];
85$JT = [:LineBreak =  JT:];
86$LF = [:LineBreak =  Line_Feed:];
87$NL = [:LineBreak =  Next_Line:];
88$NS = [[:LineBreak =  Nonstarter:] $CJ];
89$NU = [:LineBreak =  Numeric:];
90$OP = [:LineBreak =  Open_Punctuation:];
91$PO = [:LineBreak =  Postfix_Numeric:];
92$PR = [:LineBreak =  Prefix_Numeric:];
93$QU = [:LineBreak =  Quotation:];
94$RI = [:LineBreak =  Regional_Indicator:];
95$SA = [:LineBreak =  Complex_Context:];
96$SG = [:LineBreak =  Surrogate:];
97$SP = [:LineBreak =  Space:];
98$SY = [:LineBreak =  Break_Symbols:];
99$WJ = [:LineBreak =  Word_Joiner:];
100$XX = [:LineBreak =  Unknown:];
101$ZW = [:LineBreak =  ZWSpace:];
102
103#   Dictionary character set, for triggering language-based break engines. Currently
104#   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
105#   5.0 or later as the definition of Complex_Context was corrected to include all
106#   characters requiring dictionary break.
107
108$dictionary = [:LineBreak = Complex_Context:];
109
110#
111#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
112#                               SA  (South East Asian: Thai, Lao, Khmer)
113#                               SG  (Unpaired Surrogates)
114#                               XX  (Unknown, unassigned)
115#                         as $AL  (Alphabetic)
116#
117$ALPlus = [$AL $AI $SA $SG $XX];
118
119#
120#  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
121#
122$ALcm = $ALPlus $CM*;
123$BAcm = $BA $CM*;
124$BBcm = $BB $CM*;
125$B2cm = $B2 $CM*;
126$CLcm = $CL $CM*;
127$CPcm = $CP $CM*;
128$EXcm = $EX $CM*;
129$GLcm = $GL $CM*;
130$HLcm = $HL $CM*;
131$HYcm = $HY $CM*;
132$H2cm = $H2 $CM*;
133$H3cm = $H3 $CM*;
134$IDcm = $ID $CM*;
135$INcm = $IN $CM*;
136$IScm = $IS $CM*;
137$JLcm = $JL $CM*;
138$JVcm = $JV $CM*;
139$JTcm = $JT $CM*;
140$NScm = $NS $CM*;
141$NUcm = $NU $CM*;
142$OPcm = $OP $CM*;
143$POcm = $PO $CM*;
144$PRcm = $PR $CM*;
145$QUcm = $QU $CM*;
146$RIcm = $RI $CM*;
147$SYcm = $SY $CM*;
148$WJcm = $WJ $CM*;
149
150## -------------------------------------------------
151
152!!forward;
153
154#
155#  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
156#
157$ALPlus $CM+;
158$BA $CM+;
159$BB $CM+;
160$B2 $CM+;
161$CL $CM+;
162$CP $CM+;
163$EX $CM+;
164$GL $CM+;
165$HL $CM+;
166$HY $CM+;
167$H2 $CM+;
168$H3 $CM+;
169$ID $CM+;
170$IN $CM+;
171$IS $CM+;
172$JL $CM+;
173$JV $CM+;
174$JT $CM+;
175$NS $CM+;
176$NU $CM+;
177$OP $CM+;
178$PO $CM+;
179$PR $CM+;
180$QU $CM+;
181$RI $CM+;
182$SY $CM+;
183$WJ $CM+;
184
185#
186# CAN_CM  is the set of characters that may combine with CM combining chars.
187#         Note that Linebreak UAX 14's concept of a combining char and the rules
188#         for what they can combine with are _very_ different from the rest of Unicode.
189#
190#         Note that $CM itself is left out of this set.  If CM is needed as a base
191#         it must be listed separately in the rule.
192#
193$CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
194$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
195
196#
197# AL_FOLLOW  set of chars that can unconditionally follow an AL
198#            Needed in rules where stand-alone $CM s are treated as AL.
199#            Chaining is disabled with CM because it causes other failures,
200#            so for this one case we need to manually list out longer sequences.
201#
202$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
203$AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
204$AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
205
206
207#
208#  Rule LB 4, 5    Mandatory (Hard) breaks.
209#
210$LB4Breaks    = [$BK $CR $LF $NL];
211$LB4NonBreaks = [^$BK $CR $LF $NL];
212$CR $LF {100};
213
214#
215#  LB 6    Do not break before hard line breaks.
216#
217$LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
218$CAN_CM $CM*    $LB4Breaks {100};
219$CM+            $LB4Breaks {100};
220
221# LB 7         x SP
222#              x ZW
223$LB4NonBreaks [$SP $ZW];
224$CAN_CM $CM*  [$SP $ZW];
225$CM+          [$SP $ZW];
226
227#
228# LB 8         Break after zero width space
229#              TODO:  ZW SP* <break>
230#              An engine change is required to write the reverse rule for this.
231#              For now, leave the Unicode 5.2 rule, ZW <break>
232#
233$LB8Breaks    = [$LB4Breaks $ZW];
234$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
235
236
237# LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL 
238#                                $CM not covered by the above needs to behave like $AL   
239#                                See definition of $CAN_CM.
240
241$CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
242$CM+;
243
244#
245# LB 11  Do not break before or after WORD JOINER & related characters.
246#
247$CAN_CM $CM*  $WJcm;
248$LB8NonBreaks $WJcm;
249$CM+          $WJcm;
250
251$WJcm $CANT_CM;
252$WJcm $CAN_CM $CM*;
253
254#
255# LB 12  Do not break after NBSP and related characters.
256#         GL  x
257#
258$GLcm $CAN_CM $CM*;
259$GLcm $CANT_CM;
260 
261#
262# LB 12a  Do not break before NBSP and related characters ...
263#            [^SP BA HY] x GL
264#
265[[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
266$CM+ GLcm;
267
268
269
270#
271# LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
272#
273$LB8NonBreaks $CL;
274$CAN_CM $CM*  $CL;
275$CM+          $CL;              # by rule 10, stand-alone CM behaves as AL
276
277$LB8NonBreaks $CP;
278$CAN_CM $CM*  $CP;
279$CM+          $CP;              # by rule 10, stand-alone CM behaves as AL
280
281$LB8NonBreaks $EX;
282$CAN_CM $CM*  $EX;
283$CM+          $EX;              # by rule 10, stand-alone CM behaves as AL
284
285$LB8NonBreaks $IS;
286$CAN_CM $CM*  $IS;
287$CM+          $IS;              # by rule 10, stand-alone CM behaves as AL
288
289$LB8NonBreaks $SY;
290$CAN_CM $CM*  $SY;
291$CM+          $SY;              # by rule 10, stand-alone CM behaves as AL
292
293
294#
295# LB 14  Do not break after OP, even after spaces
296#
297$OPcm $SP* $CAN_CM $CM*;
298$OPcm $SP* $CANT_CM;
299
300$OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
301
302# LB 15
303$QUcm $SP* $OPcm;
304
305# LB 16
306($CLcm | $CPcm) $SP* $NScm;
307
308# LB 17
309$B2cm $SP* $B2cm;
310
311#
312# LB 18  Break after spaces.
313#
314$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
315$LB18Breaks    = [$LB8Breaks $SP];
316
317
318# LB 19
319#         x QU
320$LB18NonBreaks $CM* $QUcm;
321$CM+                $QUcm;
322
323#         QU  x
324$QUcm .?;
325$QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
326                              #  TODO:  I don't think this rule is needed.
327
328
329# LB 20
330#        <break>  $CB
331#        $CB   <break>
332
333$LB20NonBreaks = [$LB18NonBreaks - $CB];
334
335# LB 21        x   (BA | HY | NS)
336#           BB x
337#
338$LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm); 
339
340$BBcm [^$CB];                                  #  $BB  x
341$BBcm $LB20NonBreaks $CM*;
342
343# LB 21a Don't break after Hebrew + Hyphen
344#   HL (HY | BA) x
345#  
346$HLcm ($HYcm | $BAcm) [^$CB]?;
347
348# LB 21b (forward) Don't break between SY and HL
349# (break between HL and SY already disallowed by LB 13 above)
350$SYcm $HLcm;
351
352# LB 22
353($ALcm | $HLcm) $INcm;
354$CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
355$IDcm    $INcm;
356$INcm    $INcm;
357$NUcm    $INcm;
358
359
360# $LB 23
361$IDcm  $POcm;
362$ALcm  $NUcm;       # includes $LB19
363$HLcm  $NUcm;
364$CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
365$NUcm  $ALcm;
366$NUcm  $HLcm;
367
368#
369# LB 24
370#
371$PRcm $IDcm;
372$PRcm ($ALcm | $HLcm);
373$POcm ($ALcm | $HLcm);
374
375#
376# LB 25   Numbers.
377#
378($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
379
380# LB 26  Do not break a Korean syllable
381#
382$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
383($JVcm | $H2cm) ($JVcm | $JTcm);
384($JTcm | $H3cm) $JTcm;
385
386# LB 27  Treat korean Syllable Block the same as ID  (don't break it)
387($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
388($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
389$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
390
391
392# LB 28   Do not break between alphabetics
393#
394($ALcm | $HLcm) ($ALcm | $HLcm);
395$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
396
397# LB 29
398$IScm ($ALcm | $HLcm);
399
400# LB 30
401($ALcm | $HLcm | $NUcm) $OPcm;
402$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.          
403$CPcm ($ALcm | $HLcm | $NUcm);
404
405# LB 30a  Do not break between regional indicators.
406$RIcm $RIcm;
407
408#
409#  Reverse Rules.
410#
411## -------------------------------------------------
412
413!!reverse;
414
415$CM+ $ALPlus;
416$CM+ $BA;
417$CM+ $BB;
418$CM+ $B2;
419$CM+ $CL;
420$CM+ $CP;
421$CM+ $EX;
422$CM+ $GL;
423$CM+ $HL;
424$CM+ $HY;
425$CM+ $H2;
426$CM+ $H3;
427$CM+ $ID;
428$CM+ $IN;
429$CM+ $IS;
430$CM+ $JL;
431$CM+ $JV;
432$CM+ $JT;
433$CM+ $NS;
434$CM+ $NU;
435$CM+ $OP;
436$CM+ $PO;
437$CM+ $PR;
438$CM+ $QU;
439$CM+ $RI;
440$CM+ $SY;
441$CM+ $WJ;
442$CM+;
443
444
445#
446#  Sequences of the form  (shown forwards)
447#      [CANT_CM]  <break>  [CM]  [whatever]
448#  The CM needs to behave as an AL
449#
450$AL_FOLLOW $CM+ / (
451          [$BK $CR $LF $NL $ZW {eof}] |
452          $SP+ $CM+ $SP |
453          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
454                                               #  LB14 says    OP SP* x .        
455                                               #    becomes    OP SP* x AL
456                                               #    becomes    OP SP* x CM+ AL_FOLLOW
457                                               #
458                                               # Further note:  the $AL in [$AL {eof}] is only to work around
459                                               #                a rule compiler bug which complains about
460                                               #                empty sets otherwise.
461          
462#
463#  Sequences of the form  (shown forwards)
464#      [CANT_CM]  <break> [CM]  <break>  [PR]
465#  The CM needs to behave as an AL
466#  This rule is concerned about getting the second of the two <breaks> in place.
467#
468
469[$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
470
471
472
473# LB 4, 5, 5
474
475$LB4Breaks [$LB4NonBreaks-$CM];
476$LB4Breaks $CM+ $CAN_CM;
477$LF $CR;
478
479
480# LB 7         x SP
481#              x ZW
482[$SP $ZW] [$LB4NonBreaks-$CM];
483[$SP $ZW] $CM+ $CAN_CM;
484
485# LB 8 ZW SP* <break>
486#     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
487#           Requires an engine enhancement.
488#   / $SP* $ZW
489
490# LB 9,10  Combining marks.
491#    X   $CM needs to behave like X, where X is not $SP or controls.
492#    $CM not covered by the above needs to behave like $AL
493# Stick together any combining sequences that don't match other rules.
494$CM+ $CAN_CM;
495
496
497# LB 11
498$CM* $WJ $CM* $CAN_CM;
499$CM* $WJ      [$LB8NonBreaks-$CM];
500
501     $CANT_CM $CM* $WJ;
502$CM* $CAN_CM  $CM* $WJ;
503
504# LB 12a
505#      [^SP BA HY] x GL
506#
507$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
508
509# LB 12
510#     GL  x
511#
512$CANT_CM $CM* $GL;
513$CM* $CAN_CM $CM* $GL;
514
515
516# LB 13
517$CL $CM+ $CAN_CM;
518$CP $CM+ $CAN_CM;
519$EX $CM+ $CAN_CM;
520$IS $CM+ $CAN_CM;
521$SY $CM+ $CAN_CM;
522
523$CL [$LB8NonBreaks-$CM];
524$CP [$LB8NonBreaks-$CM];
525$EX [$LB8NonBreaks-$CM];
526$IS [$LB8NonBreaks-$CM];
527$SY [$LB8NonBreaks-$CM];
528
529# Rule 13 & 14 taken together for an edge case.
530#   Match this, shown forward
531#     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
532#   This really wants to chain at the $CM+ (which is acting as an $AL)
533#   except for $CM chaining being disabled.
534[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;  
535
536# LB 14    OP SP* x
537#
538$CM* $CAN_CM    $SP* $CM* $OP;
539     $CANT_CM   $SP* $CM* $OP;
540$AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
541     
542     $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
543$CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
544$SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
545
546
547
548# LB 15
549$CM* $OP $SP* $CM* $QU;
550
551# LB 16
552$CM* $NS $SP* $CM* ($CL | $CP);
553
554# LB 17
555$CM* $B2 $SP* $CM* $B2;
556
557# LB 18  break after spaces
558#        Nothing explicit needed here.
559
560
561#
562# LB 19
563#
564$CM* $QU $CM* $CAN_CM;                                #   . x QU
565$CM* $QU      $LB18NonBreaks;
566
567
568$CM* $CAN_CM  $CM* $QU;                               #   QU x .
569     $CANT_CM $CM* $QU;
570     
571#
572#  LB 20  Break before and after CB.
573#         nothing needed here.
574#
575
576# LB 21
577$CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
578
579$CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
580[^$CB] $CM* $BB;                                      # 
581
582# LB21a
583[^$CB] $CM* ($HY | $BA) $CM* $HL;
584
585# LB21b (reverse)
586$CM* $HL $CM* $SY;
587
588# LB 22
589$CM* $IN $CM* ($ALPlus | $HL);
590$CM* $IN $CM* $ID;
591$CM* $IN $CM* $IN;
592$CM* $IN $CM* $NU;
593
594# LB 23
595$CM* $PO $CM* $ID;
596$CM* $NU $CM* ($ALPlus | $HL);
597$CM* ($ALPlus | $HL) $CM* $NU;
598
599# LB 24
600$CM* $ID $CM* $PR;
601$CM* ($ALPlus | $HL) $CM* $PR;
602$CM* ($ALPlus | $HL) $CM* $PO;
603
604
605# LB 25
606($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
607
608# LB 26
609$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
610$CM* ($JT | $JV) $CM* ($H2 | $JV);
611$CM* $JT $CM* ($H3 | $JT);
612
613# LB 27
614$CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
615$CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
616$CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
617
618# LB 28
619$CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
620
621
622# LB 29
623$CM* ($ALPlus | $HL) $CM* $IS;
624
625# LB 30
626$CM* $OP $CM* ($ALPlus | $HL | $NU);
627$CM* ($ALPlus | $HL | $NU) $CM* $CP;
628
629# LB 30a
630$CM* $RI $CM* $RI;
631
632## -------------------------------------------------
633
634!!safe_reverse;
635
636# LB 9
637$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
638$CM+ $SP / .;
639
640# LB 14
641$SP+ $CM* $OP;
642
643# LB 15
644$SP+ $CM* $QU;
645
646# LB 16
647$SP+ $CM* ($CL | $CP);
648
649# LB 17
650$SP+ $CM* $B2;
651
652# LB 21
653$CM* ($HY | $BA) $CM* $HL;
654
655# LB 25
656($CM* ($IS | $SY))+ $CM* $NU;
657($CL | $CP) $CM* ($NU | $IS | $SY);
658
659# For dictionary-based break
660$dictionary $dictionary;
661
662## -------------------------------------------------
663
664!!safe_forward;
665
666# Skip forward over all character classes that are involved in
667#   rules containing patterns with possibly more than one char
668#   of context.
669#
670#  It might be slightly more efficient to have specific rules
671#  instead of one generic one, but only if we could
672#  turn off rule chaining.  We don't want to move more
673#  than necessary.
674#
675[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
676$dictionary $dictionary;
677
678