• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright (c) 2002-2015  International Business Machines Corporation and
2# others. All Rights Reserved.
3#
4#  file:  line_normal_cj.txt
5#
6#         Line Breaking Rules
7#         Implement default line breaking as defined by
8#         Unicode Standard Annex #14 Revision 34 for Unicode 8.0
9#         http://www.unicode.org/reports/tr14/
10#         tailored as noted in 2nd paragraph below..
11#
12#         TODO:  Rule LB 8 remains as it was in Unicode 5.2
13#         This is only because of a limitation of ICU break engine implementation,
14#         not because the older behavior is desirable.
15#
16#         This tailors the line break behavior to correspond to CSS
17#         line-break=normal (BCP47 -u-lb-normal) as defined for Chinese & Japanese.
18#         It sets characters of class CJ to behave like ID.
19#         In addition, it allows breaks:
20#         * before hyphens 2010 & 2013 (both BA) and 301C, 30A0 (both NS)
21
22#
23#  Character Classes defined by TR 14.
24#
25
26!!chain;
27!!LBCMNoChain;
28
29
30!!lookAheadHardBreak;
31#
32#  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
33#                          and only used for the line break rules.
34#
35#           It is used in the implementation of rule LB 10
36#           which says to treat any combining mark that is not attached to a base
37#           character as if it were of class AL  (alphabetic).
38#
39#           The problem occurs in the reverse rules.
40#
41#           Consider a sequence like, with correct breaks as shown
42#               LF  ID  CM  AL  AL
43#                  ^       ^       ^
44#           Then consider the sequence without the initial ID (ideographic)
45#                 LF  CM  AL  AL
46#                    ^           ^
47#           Our CM, which in the first example was attached to the ideograph,
48#           is now unattached, becomes an alpha, and joins in with the other
49#           alphas.
50#
51#           When iterating forwards, these sequences do not present any problems
52#           When iterating backwards, we need to look ahead when encountering
53#           a CM to see whether it attaches to something further on or not.
54#           (Look-ahead in a reverse rule is looking towards the start)
55#
56#           If the CM is unattached, we need to force a break.
57#
58#           !!lookAheadHardBreak forces the run time state machine to
59#           stop immediately when a look ahead rule ( '/' operator) matches,
60#           and set the match position to that of the look-ahead operator,
61#           no matter what other rules may be in play at the time.
62#
63#           See rule LB 19 for an example.
64#
65
66$AI = [:LineBreak =  Ambiguous:];
67$AL = [:LineBreak =  Alphabetic:];
68$BAX = [\u2010 \u2013];
69$BA = [[:LineBreak =  Break_After:] - $BAX];
70$BB = [:LineBreak =  Break_Before:];
71$BK = [:LineBreak =  Mandatory_Break:];
72$B2 = [:LineBreak =  Break_Both:];
73$CB = [:LineBreak =  Contingent_Break:];
74$CJ = [:LineBreak =  Conditional_Japanese_Starter:];
75$CL = [:LineBreak =  Close_Punctuation:];
76$CM = [:LineBreak =  Combining_Mark:];
77$CP = [:LineBreak =  Close_Parenthesis:];
78$CR = [:LineBreak =  Carriage_Return:];
79$EX = [:LineBreak =  Exclamation:];
80$GL = [:LineBreak =  Glue:];
81$HL = [:LineBreak =  Hebrew_Letter:];
82$HY = [:LineBreak =  Hyphen:];
83$H2 = [:LineBreak =  H2:];
84$H3 = [:LineBreak =  H3:];
85$ID = [[:LineBreak =  Ideographic:] $CJ];
86$IN = [:LineBreak =  Inseperable:];
87$IS = [:LineBreak =  Infix_Numeric:];
88$JL = [:LineBreak =  JL:];
89$JV = [:LineBreak =  JV:];
90$JT = [:LineBreak =  JT:];
91$LF = [:LineBreak =  Line_Feed:];
92$NL = [:LineBreak =  Next_Line:];
93$NSX = [\u301C \u30A0];
94$NS = [[:LineBreak =  Nonstarter:] - $NSX];
95$NU = [:LineBreak =  Numeric:];
96$OP = [:LineBreak =  Open_Punctuation:];
97$PO = [:LineBreak =  Postfix_Numeric:];
98$PR = [:LineBreak =  Prefix_Numeric:];
99$QU = [:LineBreak =  Quotation:];
100$RI = [:LineBreak =  Regional_Indicator:];
101$SA = [:LineBreak =  Complex_Context:];
102$SG = [:LineBreak =  Surrogate:];
103$SP = [:LineBreak =  Space:];
104$SY = [:LineBreak =  Break_Symbols:];
105$WJ = [:LineBreak =  Word_Joiner:];
106$XX = [:LineBreak =  Unknown:];
107$ZW = [:LineBreak =  ZWSpace:];
108
109#   Dictionary character set, for triggering language-based break engines. Currently
110#   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
111#   5.0 or later as the definition of Complex_Context was corrected to include all
112#   characters requiring dictionary break.
113
114$dictionary = [:LineBreak = Complex_Context:];
115
116#
117#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
118#                               SA  (South East Asian: Thai, Lao, Khmer)
119#                               SG  (Unpaired Surrogates)
120#                               XX  (Unknown, unassigned)
121#                         as $AL  (Alphabetic)
122#
123$ALPlus = [$AL $AI $SA $SG $XX];
124
125#
126#  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
127#
128$ALcm = $ALPlus $CM*;
129$BAcm = $BA $CM*;
130$BAXcm = $BAX $CM*;
131$BBcm = $BB $CM*;
132$B2cm = $B2 $CM*;
133$CLcm = $CL $CM*;
134$CPcm = $CP $CM*;
135$EXcm = $EX $CM*;
136$GLcm = $GL $CM*;
137$HLcm = $HL $CM*;
138$HYcm = $HY $CM*;
139$H2cm = $H2 $CM*;
140$H3cm = $H3 $CM*;
141$IDcm = $ID $CM*;
142$INcm = $IN $CM*;
143$IScm = $IS $CM*;
144$JLcm = $JL $CM*;
145$JVcm = $JV $CM*;
146$JTcm = $JT $CM*;
147$NScm = $NS $CM*;
148$NSXcm = $NSX $CM*;
149$NUcm = $NU $CM*;
150$OPcm = $OP $CM*;
151$POcm = $PO $CM*;
152$PRcm = $PR $CM*;
153$QUcm = $QU $CM*;
154$RIcm = $RI $CM*;
155$SYcm = $SY $CM*;
156$WJcm = $WJ $CM*;
157
158## -------------------------------------------------
159
160!!forward;
161
162#
163#  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
164#
165$ALPlus $CM+;
166$BA $CM+;
167$BAX $CM+;
168$BB $CM+;
169$B2 $CM+;
170$CL $CM+;
171$CP $CM+;
172$EX $CM+;
173$GL $CM+;
174$HL $CM+;
175$HY $CM+;
176$H2 $CM+;
177$H3 $CM+;
178$ID $CM+;
179$IN $CM+;
180$IS $CM+;
181$JL $CM+;
182$JV $CM+;
183$JT $CM+;
184$NS $CM+;
185$NSX $CM+;
186$NU $CM+;
187$OP $CM+;
188$PO $CM+;
189$PR $CM+;
190$QU $CM+;
191$RI $CM+;
192$SY $CM+;
193$WJ $CM+;
194
195#
196# CAN_CM  is the set of characters that may combine with CM combining chars.
197#         Note that Linebreak UAX 14's concept of a combining char and the rules
198#         for what they can combine with are _very_ different from the rest of Unicode.
199#
200#         Note that $CM itself is left out of this set.  If CM is needed as a base
201#         it must be listed separately in the rule.
202#
203$CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
204$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
205
206#
207# AL_FOLLOW  set of chars that can unconditionally follow an AL
208#            Needed in rules where stand-alone $CM s are treated as AL.
209#            Chaining is disabled with CM because it causes other failures,
210#            so for this one case we need to manually list out longer sequences.
211#
212$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
213$AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
214$AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
215
216
217#
218#  Rule LB 4, 5    Mandatory (Hard) breaks.
219#
220$LB4Breaks    = [$BK $CR $LF $NL];
221$LB4NonBreaks = [^$BK $CR $LF $NL];
222$CR $LF {100};
223
224#
225#  LB 6    Do not break before hard line breaks.
226#
227$LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
228$CAN_CM $CM*    $LB4Breaks {100};
229$CM+            $LB4Breaks {100};
230
231# LB 7         x SP
232#              x ZW
233$LB4NonBreaks [$SP $ZW];
234$CAN_CM $CM*  [$SP $ZW];
235$CM+          [$SP $ZW];
236
237#
238# LB 8         Break after zero width space
239#              TODO:  ZW SP* <break>
240#              An engine change is required to write the reverse rule for this.
241#              For now, leave the Unicode 5.2 rule, ZW <break>
242#
243$LB8Breaks    = [$LB4Breaks $ZW];
244$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
245
246
247# LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
248#                                $CM not covered by the above needs to behave like $AL
249#                                See definition of $CAN_CM.
250
251$CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
252$CM+;
253
254#
255# LB 11  Do not break before or after WORD JOINER & related characters.
256#
257$CAN_CM $CM*  $WJcm;
258$LB8NonBreaks $WJcm;
259$CM+          $WJcm;
260
261$WJcm $CANT_CM;
262$WJcm $CAN_CM $CM*;
263
264#
265# LB 12  Do not break after NBSP and related characters.
266#         GL  x
267#
268$GLcm $CAN_CM $CM*;
269$GLcm $CANT_CM;
270
271#
272# LB 12a  Do not break before NBSP and related characters ...
273#            [^SP BA HY] x GL
274#
275[[$LB8NonBreaks] - [$SP $BA $BAX $HY]] $CM* $GLcm;
276$CM+ GLcm;
277
278
279
280#
281# LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
282#
283$LB8NonBreaks $CL;
284$CAN_CM $CM*  $CL;
285$CM+          $CL;              # by rule 10, stand-alone CM behaves as AL
286
287$LB8NonBreaks $CP;
288$CAN_CM $CM*  $CP;
289$CM+          $CP;              # by rule 10, stand-alone CM behaves as AL
290
291$LB8NonBreaks $EX;
292$CAN_CM $CM*  $EX;
293$CM+          $EX;              # by rule 10, stand-alone CM behaves as AL
294
295$LB8NonBreaks $IS;
296$CAN_CM $CM*  $IS;
297$CM+          $IS;              # by rule 10, stand-alone CM behaves as AL
298
299$LB8NonBreaks $SY;
300$CAN_CM $CM*  $SY;
301$CM+          $SY;              # by rule 10, stand-alone CM behaves as AL
302
303
304#
305# LB 14  Do not break after OP, even after spaces
306#
307$OPcm $SP* $CAN_CM $CM*;
308$OPcm $SP* $CANT_CM;
309
310$OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
311
312# LB 15
313$QUcm $SP* $OPcm;
314
315# LB 16
316# Do not break between closing punctuation and $NS, even with intervening spaces
317# But DO allow a break between closing punctuation and $NSX, don't include it here
318($CLcm | $CPcm) $SP* $NScm;
319
320# LB 17
321$B2cm $SP* $B2cm;
322
323#
324# LB 18  Break after spaces.
325#
326$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
327$LB18Breaks    = [$LB8Breaks $SP];
328
329
330# LB 19
331#         x QU
332$LB18NonBreaks $CM* $QUcm;
333$CM+                $QUcm;
334
335#         QU  x
336$QUcm .?;
337$QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
338                              #  TODO:  I don't think this rule is needed.
339
340
341# LB 20
342#        <break>  $CB
343#        $CB   <break>
344
345$LB20NonBreaks = [$LB18NonBreaks - $CB];
346
347# LB 21        x   (BA | HY | NS)
348#           BB x
349#
350# DO allow breaks here before $BAXcm and $NSXcm, so don't include them
351$LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
352
353$BBcm [^$CB];                                  #  $BB  x
354$BBcm $LB20NonBreaks $CM*;
355
356# LB 21a Don't break after Hebrew + Hyphen
357#   HL (HY | BA) x
358#
359$HLcm ($HYcm | $BAcm | $BAXcm) [^$CB]?;
360
361# LB 21b (forward) Don't break between SY and HL
362# (break between HL and SY already disallowed by LB 13 above)
363$SYcm $HLcm;
364
365# LB 22
366($ALcm | $HLcm) $INcm;
367$CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
368$EXcm    $INcm;
369$IDcm    $INcm;
370$INcm    $INcm;
371$NUcm    $INcm;
372
373
374# $LB 23
375$IDcm  $POcm;
376$ALcm  $NUcm;       # includes $LB19
377$HLcm  $NUcm;
378$CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
379$NUcm  $ALcm;
380$NUcm  $HLcm;
381
382#
383# LB 24
384#
385$PRcm $IDcm;
386$PRcm ($ALcm | $HLcm);
387$POcm ($ALcm | $HLcm);
388
389#
390# LB 25   Numbers.
391#
392($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
393
394# LB 26  Do not break a Korean syllable
395#
396$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
397($JVcm | $H2cm) ($JVcm | $JTcm);
398($JTcm | $H3cm) $JTcm;
399
400# LB 27  Treat korean Syllable Block the same as ID  (don't break it)
401($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
402($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
403$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
404
405
406# LB 28   Do not break between alphabetics
407#
408($ALcm | $HLcm) ($ALcm | $HLcm);
409$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
410
411# LB 29
412$IScm ($ALcm | $HLcm);
413
414# LB 30
415($ALcm | $HLcm | $NUcm) $OPcm;
416$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
417$CPcm ($ALcm | $HLcm | $NUcm);
418
419# LB 30a  Do not break between regional indicators.
420$RIcm $RIcm;
421
422#
423#  Reverse Rules.
424#
425## -------------------------------------------------
426
427!!reverse;
428
429$CM+ $ALPlus;
430$CM+ $BA;
431$CM+ $BAX;
432$CM+ $BB;
433$CM+ $B2;
434$CM+ $CL;
435$CM+ $CP;
436$CM+ $EX;
437$CM+ $GL;
438$CM+ $HL;
439$CM+ $HY;
440$CM+ $H2;
441$CM+ $H3;
442$CM+ $ID;
443$CM+ $IN;
444$CM+ $IS;
445$CM+ $JL;
446$CM+ $JV;
447$CM+ $JT;
448$CM+ $NS;
449$CM+ $NSX;
450$CM+ $NU;
451$CM+ $OP;
452$CM+ $PO;
453$CM+ $PR;
454$CM+ $QU;
455$CM+ $RI;
456$CM+ $SY;
457$CM+ $WJ;
458$CM+;
459
460
461#
462#  Sequences of the form  (shown forwards)
463#      [CANT_CM]  <break>  [CM]  [whatever]
464#  The CM needs to behave as an AL
465#
466$AL_FOLLOW $CM+ / (
467          [$BK $CR $LF $NL $ZW {eof}] |
468          $SP+ $CM+ $SP |
469          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
470                                               #  LB14 says    OP SP* x .
471                                               #    becomes    OP SP* x AL
472                                               #    becomes    OP SP* x CM+ AL_FOLLOW
473                                               #
474                                               # Further note:  the $AL in [$AL {eof}] is only to work around
475                                               #                a rule compiler bug which complains about
476                                               #                empty sets otherwise.
477
478#
479#  Sequences of the form  (shown forwards)
480#      [CANT_CM]  <break> [CM]  <break>  [PR]
481#  The CM needs to behave as an AL
482#  This rule is concerned about getting the second of the two <breaks> in place.
483#
484
485[$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
486
487
488
489# LB 4, 5, 5
490
491$LB4Breaks [$LB4NonBreaks-$CM];
492$LB4Breaks $CM+ $CAN_CM;
493$LF $CR;
494
495
496# LB 7         x SP
497#              x ZW
498[$SP $ZW] [$LB4NonBreaks-$CM];
499[$SP $ZW] $CM+ $CAN_CM;
500
501# LB 8 ZW SP* <break>
502#     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
503#           Requires an engine enhancement.
504#   / $SP* $ZW
505
506# LB 9,10  Combining marks.
507#    X   $CM needs to behave like X, where X is not $SP or controls.
508#    $CM not covered by the above needs to behave like $AL
509# Stick together any combining sequences that don't match other rules.
510$CM+ $CAN_CM;
511
512
513# LB 11
514$CM* $WJ $CM* $CAN_CM;
515$CM* $WJ      [$LB8NonBreaks-$CM];
516
517     $CANT_CM $CM* $WJ;
518$CM* $CAN_CM  $CM* $WJ;
519
520# LB 12a
521#      [^SP BA HY] x GL
522#
523$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $BAX $HY]];
524
525# LB 12
526#     GL  x
527#
528$CANT_CM $CM* $GL;
529$CM* $CAN_CM $CM* $GL;
530
531
532# LB 13
533$CL $CM+ $CAN_CM;
534$CP $CM+ $CAN_CM;
535$EX $CM+ $CAN_CM;
536$IS $CM+ $CAN_CM;
537$SY $CM+ $CAN_CM;
538
539$CL [$LB8NonBreaks-$CM];
540$CP [$LB8NonBreaks-$CM];
541$EX [$LB8NonBreaks-$CM];
542$IS [$LB8NonBreaks-$CM];
543$SY [$LB8NonBreaks-$CM];
544
545# Rule 13 & 14 taken together for an edge case.
546#   Match this, shown forward
547#     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
548#   This really wants to chain at the $CM+ (which is acting as an $AL)
549#   except for $CM chaining being disabled.
550[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
551
552# LB 14    OP SP* x
553#
554$CM* $CAN_CM    $SP* $CM* $OP;
555     $CANT_CM   $SP* $CM* $OP;
556$AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
557
558     $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
559$CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
560$SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
561
562
563
564# LB 15
565$CM* $OP $SP* $CM* $QU;
566
567# LB 16
568# Don't include $NSX here
569$CM* $NS $SP* $CM* ($CL | $CP);
570
571# LB 17
572$CM* $B2 $SP* $CM* $B2;
573
574# LB 18  break after spaces
575#        Nothing explicit needed here.
576
577
578#
579# LB 19
580#
581$CM* $QU $CM* $CAN_CM;                                #   . x QU
582$CM* $QU      $LB18NonBreaks;
583
584
585$CM* $CAN_CM  $CM* $QU;                               #   QU x .
586     $CANT_CM $CM* $QU;
587
588#
589#  LB 20  Break before and after CB.
590#         nothing needed here.
591#
592
593# LB 21
594# Don't include $BAX or $NSX here
595$CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
596
597$CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
598[^$CB] $CM* $BB;                                      #
599
600# LB21a
601[^$CB]? $CM* ($HY | $BA | $BAX) $CM* $HL;
602
603# LB21b (reverse)
604$CM* $HL $CM* $SY;
605
606# LB 22
607$CM* $IN $CM* ($ALPlus | $HL);
608$CM* $IN $CM* $EX;
609$CM* $IN $CM* $ID;
610$CM* $IN $CM* $IN;
611$CM* $IN $CM* $NU;
612
613# LB 23
614$CM* $PO $CM* $ID;
615$CM* $NU $CM* ($ALPlus | $HL);
616$CM* ($ALPlus | $HL) $CM* $NU;
617
618# LB 24
619$CM* $ID $CM* $PR;
620$CM* ($ALPlus | $HL) $CM* $PR;
621$CM* ($ALPlus | $HL) $CM* $PO;
622
623
624# LB 25
625($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
626
627# LB 26
628$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
629$CM* ($JT | $JV) $CM* ($H2 | $JV);
630$CM* $JT $CM* ($H3 | $JT);
631
632# LB 27
633$CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
634$CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
635$CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
636
637# LB 28
638$CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
639
640
641# LB 29
642$CM* ($ALPlus | $HL) $CM* $IS;
643
644# LB 30
645$CM* $OP $CM* ($ALPlus | $HL | $NU);
646$CM* ($ALPlus | $HL | $NU) $CM* $CP;
647
648# LB 30a
649$CM* $RI $CM* $RI;
650
651## -------------------------------------------------
652
653!!safe_reverse;
654
655# LB 9
656$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
657$CM+ $SP / .;
658
659# LB 14
660$SP+ $CM* $OP;
661
662# LB 15
663$SP+ $CM* $QU;
664
665# LB 16
666$SP+ $CM* ($CL | $CP);
667
668# LB 17
669$SP+ $CM* $B2;
670
671# LB 21
672$CM* ($HY | $BA | $BAX) $CM* $HL;
673
674# LB 25
675($CM* ($IS | $SY))+ $CM* $NU;
676($CL | $CP) $CM* ($NU | $IS | $SY);
677
678# For dictionary-based break
679$dictionary $dictionary;
680
681## -------------------------------------------------
682
683!!safe_forward;
684
685# Skip forward over all character classes that are involved in
686#   rules containing patterns with possibly more than one char
687#   of context.
688#
689#  It might be slightly more efficient to have specific rules
690#  instead of one generic one, but only if we could
691#  turn off rule chaining.  We don't want to move more
692#  than necessary.
693#
694[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $BAX $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $BAX $dictionary];
695$dictionary $dictionary;
696
697