• 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_loose.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=loose (BCP47 -u-lb-loose) as defined for languages other than
18#         Chinese & Japanese.
19#         It sets characters of class CJ to behave like ID.
20#         In addition, it allows breaks:
21#         * before iteration marks 3005, 303B, 309D, 309E, 30FD, 30FE (all NS)
22#         * between characters of LineBreak class IN
23
24#
25#  Character Classes defined by TR 14.
26#
27
28!!chain;
29!!LBCMNoChain;
30
31
32!!lookAheadHardBreak;
33#
34#  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
35#                          and only used for the line break rules.
36#
37#           It is used in the implementation of rule LB 10
38#           which says to treat any combining mark that is not attached to a base
39#           character as if it were of class AL  (alphabetic).
40#
41#           The problem occurs in the reverse rules.
42#
43#           Consider a sequence like, with correct breaks as shown
44#               LF  ID  CM  AL  AL
45#                  ^       ^       ^
46#           Then consider the sequence without the initial ID (ideographic)
47#                 LF  CM  AL  AL
48#                    ^           ^
49#           Our CM, which in the first example was attached to the ideograph,
50#           is now unattached, becomes an alpha, and joins in with the other
51#           alphas.
52#
53#           When iterating forwards, these sequences do not present any problems
54#           When iterating backwards, we need to look ahead when encountering
55#           a CM to see whether it attaches to something further on or not.
56#           (Look-ahead in a reverse rule is looking towards the start)
57#
58#           If the CM is unattached, we need to force a break.
59#
60#           !!lookAheadHardBreak forces the run time state machine to
61#           stop immediately when a look ahead rule ( '/' operator) matches,
62#           and set the match position to that of the look-ahead operator,
63#           no matter what other rules may be in play at the time.
64#
65#           See rule LB 19 for an example.
66#
67
68$AI = [:LineBreak =  Ambiguous:];
69$AL = [:LineBreak =  Alphabetic:];
70$BA = [:LineBreak =  Break_After:];
71$BB = [:LineBreak =  Break_Before:];
72$BK = [:LineBreak =  Mandatory_Break:];
73$B2 = [:LineBreak =  Break_Both:];
74$CB = [:LineBreak =  Contingent_Break:];
75$CJ = [:LineBreak =  Conditional_Japanese_Starter:];
76$CL = [:LineBreak =  Close_Punctuation:];
77$CM = [:LineBreak =  Combining_Mark:];
78$CP = [:LineBreak =  Close_Parenthesis:];
79$CR = [:LineBreak =  Carriage_Return:];
80$EX = [:LineBreak =  Exclamation:];
81$GL = [:LineBreak =  Glue:];
82$HL = [:LineBreak =  Hebrew_Letter:];
83$HY = [:LineBreak =  Hyphen:];
84$H2 = [:LineBreak =  H2:];
85$H3 = [:LineBreak =  H3:];
86$ID = [[:LineBreak =  Ideographic:] $CJ];
87$IN = [:LineBreak =  Inseperable:];
88$IS = [:LineBreak =  Infix_Numeric:];
89$JL = [:LineBreak =  JL:];
90$JV = [:LineBreak =  JV:];
91$JT = [:LineBreak =  JT:];
92$LF = [:LineBreak =  Line_Feed:];
93$NL = [:LineBreak =  Next_Line:];
94$NSX = [\u3005 \u303B \u309D \u309E \u30FD \u30FE];
95$NS = [[:LineBreak =  Nonstarter:] - $NSX];
96$NU = [:LineBreak =  Numeric:];
97$OP = [:LineBreak =  Open_Punctuation:];
98$PO = [:LineBreak =  Postfix_Numeric:];
99$PR = [:LineBreak =  Prefix_Numeric:];
100$QU = [:LineBreak =  Quotation:];
101$RI = [:LineBreak =  Regional_Indicator:];
102$SA = [:LineBreak =  Complex_Context:];
103$SG = [:LineBreak =  Surrogate:];
104$SP = [:LineBreak =  Space:];
105$SY = [:LineBreak =  Break_Symbols:];
106$WJ = [:LineBreak =  Word_Joiner:];
107$XX = [:LineBreak =  Unknown:];
108$ZW = [:LineBreak =  ZWSpace:];
109
110#   Dictionary character set, for triggering language-based break engines. Currently
111#   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
112#   5.0 or later as the definition of Complex_Context was corrected to include all
113#   characters requiring dictionary break.
114
115$dictionary = [:LineBreak = Complex_Context:];
116
117#
118#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
119#                               SA  (South East Asian: Thai, Lao, Khmer)
120#                               SG  (Unpaired Surrogates)
121#                               XX  (Unknown, unassigned)
122#                         as $AL  (Alphabetic)
123#
124$ALPlus = [$AL $AI $SA $SG $XX];
125
126#
127#  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
128#
129$ALcm = $ALPlus $CM*;
130$BAcm = $BA $CM*;
131$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 $CM+;
167$BB $CM+;
168$B2 $CM+;
169$CL $CM+;
170$CP $CM+;
171$EX $CM+;
172$GL $CM+;
173$HL $CM+;
174$HY $CM+;
175$H2 $CM+;
176$H3 $CM+;
177$ID $CM+;
178$IN $CM+;
179$IS $CM+;
180$JL $CM+;
181$JV $CM+;
182$JT $CM+;
183$NS $CM+;
184$NSX $CM+;
185$NU $CM+;
186$OP $CM+;
187$PO $CM+;
188$PR $CM+;
189$QU $CM+;
190$RI $CM+;
191$SY $CM+;
192$WJ $CM+;
193
194#
195# CAN_CM  is the set of characters that may combine with CM combining chars.
196#         Note that Linebreak UAX 14's concept of a combining char and the rules
197#         for what they can combine with are _very_ different from the rest of Unicode.
198#
199#         Note that $CM itself is left out of this set.  If CM is needed as a base
200#         it must be listed separately in the rule.
201#
202$CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
203$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
204
205#
206# AL_FOLLOW  set of chars that can unconditionally follow an AL
207#            Needed in rules where stand-alone $CM s are treated as AL.
208#            Chaining is disabled with CM because it causes other failures,
209#            so for this one case we need to manually list out longer sequences.
210#
211$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
212$AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
213$AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
214
215
216#
217#  Rule LB 4, 5    Mandatory (Hard) breaks.
218#
219$LB4Breaks    = [$BK $CR $LF $NL];
220$LB4NonBreaks = [^$BK $CR $LF $NL];
221$CR $LF {100};
222
223#
224#  LB 6    Do not break before hard line breaks.
225#
226$LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
227$CAN_CM $CM*    $LB4Breaks {100};
228$CM+            $LB4Breaks {100};
229
230# LB 7         x SP
231#              x ZW
232$LB4NonBreaks [$SP $ZW];
233$CAN_CM $CM*  [$SP $ZW];
234$CM+          [$SP $ZW];
235
236#
237# LB 8         Break after zero width space
238#              TODO:  ZW SP* <break>
239#              An engine change is required to write the reverse rule for this.
240#              For now, leave the Unicode 5.2 rule, ZW <break>
241#
242$LB8Breaks    = [$LB4Breaks $ZW];
243$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
244
245
246# LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
247#                                $CM not covered by the above needs to behave like $AL
248#                                See definition of $CAN_CM.
249
250$CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
251$CM+;
252
253#
254# LB 11  Do not break before or after WORD JOINER & related characters.
255#
256$CAN_CM $CM*  $WJcm;
257$LB8NonBreaks $WJcm;
258$CM+          $WJcm;
259
260$WJcm $CANT_CM;
261$WJcm $CAN_CM $CM*;
262
263#
264# LB 12  Do not break after NBSP and related characters.
265#         GL  x
266#
267$GLcm $CAN_CM $CM*;
268$GLcm $CANT_CM;
269
270#
271# LB 12a  Do not break before NBSP and related characters ...
272#            [^SP BA HY] x GL
273#
274[[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
275$CM+ GLcm;
276
277
278
279#
280# LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
281#
282$LB8NonBreaks $CL;
283$CAN_CM $CM*  $CL;
284$CM+          $CL;              # by rule 10, stand-alone CM behaves as AL
285
286$LB8NonBreaks $CP;
287$CAN_CM $CM*  $CP;
288$CM+          $CP;              # by rule 10, stand-alone CM behaves as AL
289
290$LB8NonBreaks $EX;
291$CAN_CM $CM*  $EX;
292$CM+          $EX;              # by rule 10, stand-alone CM behaves as AL
293
294$LB8NonBreaks $IS;
295$CAN_CM $CM*  $IS;
296$CM+          $IS;              # by rule 10, stand-alone CM behaves as AL
297
298$LB8NonBreaks $SY;
299$CAN_CM $CM*  $SY;
300$CM+          $SY;              # by rule 10, stand-alone CM behaves as AL
301
302
303#
304# LB 14  Do not break after OP, even after spaces
305#
306$OPcm $SP* $CAN_CM $CM*;
307$OPcm $SP* $CANT_CM;
308
309$OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
310
311# LB 15
312$QUcm $SP* $OPcm;
313
314# LB 16
315# Do not break between closing punctuation and $NS, even with intervening spaces
316# But DO allow a break between closing punctuation and $NSX, don't include it here
317($CLcm | $CPcm) $SP* $NScm;
318
319# LB 17
320$B2cm $SP* $B2cm;
321
322#
323# LB 18  Break after spaces.
324#
325$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
326$LB18Breaks    = [$LB8Breaks $SP];
327
328
329# LB 19
330#         x QU
331$LB18NonBreaks $CM* $QUcm;
332$CM+                $QUcm;
333
334#         QU  x
335$QUcm .?;
336$QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
337                              #  TODO:  I don't think this rule is needed.
338
339
340# LB 20
341#        <break>  $CB
342#        $CB   <break>
343
344$LB20NonBreaks = [$LB18NonBreaks - $CB];
345
346# LB 21        x   (BA | HY | NS)
347#           BB x
348#
349# DO allow breaks here before NSXcm, so don't include it
350$LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
351
352$BBcm [^$CB];                                  #  $BB  x
353$BBcm $LB20NonBreaks $CM*;
354
355# LB 21a Don't break after Hebrew + Hyphen
356#   HL (HY | BA) x
357#
358$HLcm ($HYcm | $BAcm) [^$CB]?;
359
360# LB 21b (forward) Don't break between SY and HL
361# (break between HL and SY already disallowed by LB 13 above)
362$SYcm $HLcm;
363
364# LB 22
365($ALcm | $HLcm) $INcm;
366$CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
367$EXcm    $INcm;
368$IDcm    $INcm;
369# $INcm  $INcm; # delete this rule for CSS loose
370$NUcm    $INcm;
371
372
373# $LB 23
374$IDcm  $POcm;
375$ALcm  $NUcm;       # includes $LB19
376$HLcm  $NUcm;
377$CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
378$NUcm  $ALcm;
379$NUcm  $HLcm;
380
381#
382# LB 24
383#
384$PRcm $IDcm;
385$PRcm ($ALcm | $HLcm);
386$POcm ($ALcm | $HLcm);
387
388#
389# LB 25   Numbers.
390#
391($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
392
393# LB 26  Do not break a Korean syllable
394#
395$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
396($JVcm | $H2cm) ($JVcm | $JTcm);
397($JTcm | $H3cm) $JTcm;
398
399# LB 27  Treat korean Syllable Block the same as ID  (don't break it)
400($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
401($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
402$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
403
404
405# LB 28   Do not break between alphabetics
406#
407($ALcm | $HLcm) ($ALcm | $HLcm);
408$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
409
410# LB 29
411$IScm ($ALcm | $HLcm);
412
413# LB 30
414($ALcm | $HLcm | $NUcm) $OPcm;
415$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
416$CPcm ($ALcm | $HLcm | $NUcm);
417
418# LB 30a  Do not break between regional indicators.
419$RIcm $RIcm;
420
421#
422#  Reverse Rules.
423#
424## -------------------------------------------------
425
426!!reverse;
427
428$CM+ $ALPlus;
429$CM+ $BA;
430$CM+ $BB;
431$CM+ $B2;
432$CM+ $CL;
433$CM+ $CP;
434$CM+ $EX;
435$CM+ $GL;
436$CM+ $HL;
437$CM+ $HY;
438$CM+ $H2;
439$CM+ $H3;
440$CM+ $ID;
441$CM+ $IN;
442$CM+ $IS;
443$CM+ $JL;
444$CM+ $JV;
445$CM+ $JT;
446$CM+ $NS;
447$CM+ $NSX;
448$CM+ $NU;
449$CM+ $OP;
450$CM+ $PO;
451$CM+ $PR;
452$CM+ $QU;
453$CM+ $RI;
454$CM+ $SY;
455$CM+ $WJ;
456$CM+;
457
458
459#
460#  Sequences of the form  (shown forwards)
461#      [CANT_CM]  <break>  [CM]  [whatever]
462#  The CM needs to behave as an AL
463#
464$AL_FOLLOW $CM+ / (
465          [$BK $CR $LF $NL $ZW {eof}] |
466          $SP+ $CM+ $SP |
467          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
468                                               #  LB14 says    OP SP* x .
469                                               #    becomes    OP SP* x AL
470                                               #    becomes    OP SP* x CM+ AL_FOLLOW
471                                               #
472                                               # Further note:  the $AL in [$AL {eof}] is only to work around
473                                               #                a rule compiler bug which complains about
474                                               #                empty sets otherwise.
475
476#
477#  Sequences of the form  (shown forwards)
478#      [CANT_CM]  <break> [CM]  <break>  [PR]
479#  The CM needs to behave as an AL
480#  This rule is concerned about getting the second of the two <breaks> in place.
481#
482
483[$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
484
485
486
487# LB 4, 5, 5
488
489$LB4Breaks [$LB4NonBreaks-$CM];
490$LB4Breaks $CM+ $CAN_CM;
491$LF $CR;
492
493
494# LB 7         x SP
495#              x ZW
496[$SP $ZW] [$LB4NonBreaks-$CM];
497[$SP $ZW] $CM+ $CAN_CM;
498
499# LB 8 ZW SP* <break>
500#     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
501#           Requires an engine enhancement.
502#   / $SP* $ZW
503
504# LB 9,10  Combining marks.
505#    X   $CM needs to behave like X, where X is not $SP or controls.
506#    $CM not covered by the above needs to behave like $AL
507# Stick together any combining sequences that don't match other rules.
508$CM+ $CAN_CM;
509
510
511# LB 11
512$CM* $WJ $CM* $CAN_CM;
513$CM* $WJ      [$LB8NonBreaks-$CM];
514
515     $CANT_CM $CM* $WJ;
516$CM* $CAN_CM  $CM* $WJ;
517
518# LB 12a
519#      [^SP BA HY] x GL
520#
521$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
522
523# LB 12
524#     GL  x
525#
526$CANT_CM $CM* $GL;
527$CM* $CAN_CM $CM* $GL;
528
529
530# LB 13
531$CL $CM+ $CAN_CM;
532$CP $CM+ $CAN_CM;
533$EX $CM+ $CAN_CM;
534$IS $CM+ $CAN_CM;
535$SY $CM+ $CAN_CM;
536
537$CL [$LB8NonBreaks-$CM];
538$CP [$LB8NonBreaks-$CM];
539$EX [$LB8NonBreaks-$CM];
540$IS [$LB8NonBreaks-$CM];
541$SY [$LB8NonBreaks-$CM];
542
543# Rule 13 & 14 taken together for an edge case.
544#   Match this, shown forward
545#     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
546#   This really wants to chain at the $CM+ (which is acting as an $AL)
547#   except for $CM chaining being disabled.
548[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
549
550# LB 14    OP SP* x
551#
552$CM* $CAN_CM    $SP* $CM* $OP;
553     $CANT_CM   $SP* $CM* $OP;
554$AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
555
556     $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
557$CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
558$SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
559
560
561
562# LB 15
563$CM* $OP $SP* $CM* $QU;
564
565# LB 16
566# Don't include $NSX here
567$CM* $NS $SP* $CM* ($CL | $CP);
568
569# LB 17
570$CM* $B2 $SP* $CM* $B2;
571
572# LB 18  break after spaces
573#        Nothing explicit needed here.
574
575
576#
577# LB 19
578#
579$CM* $QU $CM* $CAN_CM;                                #   . x QU
580$CM* $QU      $LB18NonBreaks;
581
582
583$CM* $CAN_CM  $CM* $QU;                               #   QU x .
584     $CANT_CM $CM* $QU;
585
586#
587#  LB 20  Break before and after CB.
588#         nothing needed here.
589#
590
591# LB 21
592# Don't include $NSX here
593$CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
594
595$CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
596[^$CB] $CM* $BB;                                      #
597
598# LB21a
599[^$CB] $CM* ($HY | $BA) $CM* $HL;
600
601# LB21b (reverse)
602$CM* $HL $CM* $SY;
603
604# LB 22
605$CM* $IN $CM* ($ALPlus | $HL);
606$CM* $IN $CM* $EX;
607$CM* $IN $CM* $ID;
608# $CM* $IN $CM* $IN; # delete this rule for CSS loose
609$CM* $IN $CM* $NU;
610
611# LB 23
612$CM* $PO $CM* $ID;
613$CM* $NU $CM* ($ALPlus | $HL);
614$CM* ($ALPlus | $HL) $CM* $NU;
615
616# LB 24
617$CM* $ID $CM* $PR;
618$CM* ($ALPlus | $HL) $CM* $PR;
619$CM* ($ALPlus | $HL) $CM* $PO;
620
621
622# LB 25
623($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
624
625# LB 26
626$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
627$CM* ($JT | $JV) $CM* ($H2 | $JV);
628$CM* $JT $CM* ($H3 | $JT);
629
630# LB 27
631$CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
632$CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
633$CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
634
635# LB 28
636$CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
637
638
639# LB 29
640$CM* ($ALPlus | $HL) $CM* $IS;
641
642# LB 30
643$CM* $OP $CM* ($ALPlus | $HL | $NU);
644$CM* ($ALPlus | $HL | $NU) $CM* $CP;
645
646# LB 30a
647$CM* $RI $CM* $RI;
648
649## -------------------------------------------------
650
651!!safe_reverse;
652
653# LB 9
654$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
655$CM+ $SP / .;
656
657# LB 14
658$SP+ $CM* $OP;
659
660# LB 15
661$SP+ $CM* $QU;
662
663# LB 16
664$SP+ $CM* ($CL | $CP);
665
666# LB 17
667$SP+ $CM* $B2;
668
669# LB 21
670$CM* ($HY | $BA) $CM* $HL;
671
672# LB 25
673($CM* ($IS | $SY))+ $CM* $NU;
674($CL | $CP) $CM* ($NU | $IS | $SY);
675
676# For dictionary-based break
677$dictionary $dictionary;
678
679## -------------------------------------------------
680
681!!safe_forward;
682
683# Skip forward over all character classes that are involved in
684#   rules containing patterns with possibly more than one char
685#   of context.
686#
687#  It might be slightly more efficient to have specific rules
688#  instead of one generic one, but only if we could
689#  turn off rule chaining.  We don't want to move more
690#  than necessary.
691#
692[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
693$dictionary $dictionary;
694
695