1# Copyright (c) 2002-2013  International Business Machines Corporation and
2# others. All Rights Reserved.
3#
4#  file:  line_fi.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$BA = [[:LineBreak =  Break_After:] - [\u2010]];
62$HH = [\u2010];
63$BB = [:LineBreak =  Break_Before:];
64$BK = [:LineBreak =  Mandatory_Break:];
65$B2 = [:LineBreak =  Break_Both:];
66$CB = [:LineBreak =  Contingent_Break:];
67$CJ = [:LineBreak =  Conditional_Japanese_Starter:];
68$CL = [:LineBreak =  Close_Punctuation:];
69$CM = [:LineBreak =  Combining_Mark:];
70$CP = [:LineBreak =  Close_Parenthesis:];
71$CR = [:LineBreak =  Carriage_Return:];
72$EX = [:LineBreak =  Exclamation:];
73$GL = [:LineBreak =  Glue:];
74$HL = [:LineBreak =  Hebrew_Letter:];
75$HY = [:LineBreak =  Hyphen:];
76$H2 = [:LineBreak =  H2:];
77$H3 = [:LineBreak =  H3:];
78$ID = [:LineBreak =  Ideographic:];
79$IN = [:LineBreak =  Inseperable:];
80$IS = [:LineBreak =  Infix_Numeric:];
81$JL = [:LineBreak =  JL:];
82$JV = [:LineBreak =  JV:];
83$JT = [:LineBreak =  JT:];
84$LF = [:LineBreak =  Line_Feed:];
85$NL = [:LineBreak =  Next_Line:];
86$NS = [[:LineBreak =  Nonstarter:] $CJ];
87$NU = [:LineBreak =  Numeric:];
88$OP = [:LineBreak =  Open_Punctuation:];
89$PO = [:LineBreak =  Postfix_Numeric:];
90$PR = [:LineBreak =  Prefix_Numeric:];
91$QU = [:LineBreak =  Quotation:];
92$RI = [:LineBreak =  Regional_Indicator:];
93$SA = [:LineBreak =  Complex_Context:];
94$SG = [:LineBreak =  Surrogate:];
95$SP = [:LineBreak =  Space:];
96$SY = [:LineBreak =  Break_Symbols:];
97$WJ = [:LineBreak =  Word_Joiner:];
98$XX = [:LineBreak =  Unknown:];
99$ZW = [:LineBreak =  ZWSpace:];
100
101#   Dictionary character set, for triggering language-based break engines. Currently
102#   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
103#   5.0 or later as the definition of Complex_Context was corrected to include all
104#   characters requiring dictionary break.
105
106$dictionary = [:LineBreak = Complex_Context:];
107
108#
109#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
110#                               SA  (South East Asian: Thai, Lao, Khmer)
111#                               SG  (Unpaired Surrogates)
112#                               XX  (Unknown, unassigned)
113#                         as $AL  (Alphabetic)
114#
115$ALPlus = [$AL $AI $SA $SG $XX];
116
117#
118#  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
119#
120$ALcm = $ALPlus $CM*;
121$BAcm = $BA $CM*;
122$HHcm = $HH $CM*;
123$BBcm = $BB $CM*;
124$B2cm = $B2 $CM*;
125$CLcm = $CL $CM*;
126$CPcm = $CP $CM*;
127$EXcm = $EX $CM*;
128$GLcm = $GL $CM*;
129$HLcm = $HL $CM*;
130$HYcm = $HY $CM*;
131$H2cm = $H2 $CM*;
132$H3cm = $H3 $CM*;
133$IDcm = $ID $CM*;
134$INcm = $IN $CM*;
135$IScm = $IS $CM*;
136$JLcm = $JL $CM*;
137$JVcm = $JV $CM*;
138$JTcm = $JT $CM*;
139$NScm = $NS $CM*;
140$NUcm = $NU $CM*;
141$OPcm = $OP $CM*;
142$POcm = $PO $CM*;
143$PRcm = $PR $CM*;
144$QUcm = $QU $CM*;
145$RIcm = $RI $CM*;
146$SYcm = $SY $CM*;
147$WJcm = $WJ $CM*;
148
149## -------------------------------------------------
150
151!!forward;
152
153#
154#  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
155#
156$ALPlus $CM+;
157$BA $CM+;
158$HH $CM+;
159$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 $HH $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 $HH $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 - $CB];
334
335# LB 20.09 added rule for Finnish tailoring
336# LB 21        x   (BA | HY | NS)
337#           BB x
338#
339$LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm) / $AL; 
340$LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm); 
341($HY | $HH) $AL;
342
343$BBcm [^$CB];                                  #  $BB  x
344$BBcm $LB20NonBreaks $CM*;
345
346# LB 21a Don't break after Hebrew + Hyphen
347#   HL (HY | BA) x
348#  
349$HLcm ($HYcm | $BAcm | $HHcm) [^$CB]?;
350
351# LB 21b (forward) Don't break between SY and HL
352# (break between HL and SY already disallowed by LB 13 above)
353$SYcm $HLcm;
354
355# LB 22
356($ALcm | $HLcm) $INcm;
357$CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
358$IDcm    $INcm;
359$INcm    $INcm;
360$NUcm    $INcm;
361
362
363# $LB 23
364$IDcm  $POcm;
365$ALcm  $NUcm;       # includes $LB19
366$HLcm  $NUcm;
367$CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
368$NUcm  $ALcm;
369$NUcm  $HLcm;
370
371#
372# LB 24
373#
374$PRcm $IDcm;
375$PRcm ($ALcm | $HLcm);
376$POcm ($ALcm | $HLcm);
377
378#
379# LB 25   Numbers.
380#
381($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;
382
383# LB 26  Do not break a Korean syllable
384#
385$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
386($JVcm | $H2cm) ($JVcm | $JTcm);
387($JTcm | $H3cm) $JTcm;
388
389# LB 27  Treat korean Syllable Block the same as ID  (don't break it)
390($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
391($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
392$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
393
394
395# LB 28   Do not break between alphabetics
396#
397($ALcm | $HLcm) ($ALcm | $HLcm);
398$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL
399
400# LB 29
401$IScm ($ALcm | $HLcm);
402
403# LB 30
404($ALcm | $HLcm | $NUcm) $OPcm;
405$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.          
406$CPcm ($ALcm | $HLcm | $NUcm);
407
408# LB 30a  Do not break between regional indicators.
409$RIcm $RIcm;
410
411#
412#  Reverse Rules.
413#
414## -------------------------------------------------
415
416!!reverse;
417
418$CM+ $ALPlus;
419$CM+ $BA;
420$CM+ $HH;
421$CM+ $BB;
422$CM+ $B2;
423$CM+ $CL;
424$CM+ $CP;
425$CM+ $EX;
426$CM+ $GL;
427$CM+ $HL;
428$CM+ $HY;
429$CM+ $H2;
430$CM+ $H3;
431$CM+ $ID;
432$CM+ $IN;
433$CM+ $IS;
434$CM+ $JL;
435$CM+ $JV;
436$CM+ $JT;
437$CM+ $NS;
438$CM+ $NU;
439$CM+ $OP;
440$CM+ $PO;
441$CM+ $PR;
442$CM+ $QU;
443$CM+ $RI;
444$CM+ $SY;
445$CM+ $WJ;
446$CM+;
447
448
449#
450#  Sequences of the form  (shown forwards)
451#      [CANT_CM]  <break>  [CM]  [whatever]
452#  The CM needs to behave as an AL
453#
454$AL_FOLLOW $CM+ / (
455          [$BK $CR $LF $NL $ZW {eof}] |
456          $SP+ $CM+ $SP |
457          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
458                                               #  LB14 says    OP SP* x .        
459                                               #    becomes    OP SP* x AL
460                                               #    becomes    OP SP* x CM+ AL_FOLLOW
461                                               #
462                                               # Further note:  the $AL in [$AL {eof}] is only to work around
463                                               #                a rule compiler bug which complains about
464                                               #                empty sets otherwise.
465          
466#
467#  Sequences of the form  (shown forwards)
468#      [CANT_CM]  <break> [CM]  <break>  [PR]
469#  The CM needs to behave as an AL
470#  This rule is concerned about getting the second of the two <breaks> in place.
471#
472
473[$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
474
475
476
477# LB 4, 5, 5
478
479$LB4Breaks [$LB4NonBreaks-$CM];
480$LB4Breaks $CM+ $CAN_CM;
481$LF $CR;
482
483
484# LB 7         x SP
485#              x ZW
486[$SP $ZW] [$LB4NonBreaks-$CM];
487[$SP $ZW] $CM+ $CAN_CM;
488
489# LB 8 ZW SP* <break>
490#     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
491#           Requires an engine enhancement.
492#   / $SP* $ZW
493
494# LB 9,10  Combining marks.
495#    X   $CM needs to behave like X, where X is not $SP or controls.
496#    $CM not covered by the above needs to behave like $AL
497# Stick together any combining sequences that don't match other rules.
498$CM+ $CAN_CM;
499
500
501# LB 11
502$CM* $WJ $CM* $CAN_CM;
503$CM* $WJ      [$LB8NonBreaks-$CM];
504
505     $CANT_CM $CM* $WJ;
506$CM* $CAN_CM  $CM* $WJ;
507
508# LB 12a
509#      [^SP BA HY] x GL
510#
511$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HH $HY]];
512
513# LB 12
514#     GL  x
515#
516$CANT_CM $CM* $GL;
517$CM* $CAN_CM $CM* $GL;
518
519
520# LB 13
521$CL $CM+ $CAN_CM;
522$CP $CM+ $CAN_CM;
523$EX $CM+ $CAN_CM;
524$IS $CM+ $CAN_CM;
525$SY $CM+ $CAN_CM;
526
527$CL [$LB8NonBreaks-$CM];
528$CP [$LB8NonBreaks-$CM];
529$EX [$LB8NonBreaks-$CM];
530$IS [$LB8NonBreaks-$CM];
531$SY [$LB8NonBreaks-$CM];
532
533# Rule 13 & 14 taken together for an edge case.
534#   Match this, shown forward
535#     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
536#   This really wants to chain at the $CM+ (which is acting as an $AL)
537#   except for $CM chaining being disabled.
538[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;  
539
540# LB 14    OP SP* x
541#
542$CM* $CAN_CM    $SP* $CM* $OP;
543     $CANT_CM   $SP* $CM* $OP;
544$AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
545     
546     $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
547$CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
548$SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.
549
550
551
552# LB 15
553$CM* $OP $SP* $CM* $QU;
554
555# LB 16
556$CM* $NS $SP* $CM* ($CL | $CP);
557
558# LB 17
559$CM* $B2 $SP* $CM* $B2;
560
561# LB 18  break after spaces
562#        Nothing explicit needed here.
563
564
565#
566# LB 19
567#
568$CM* $QU $CM* $CAN_CM;                                #   . x QU
569$CM* $QU      $LB18NonBreaks;
570
571
572$CM* $CAN_CM  $CM* $QU;                               #   QU x .
573     $CANT_CM $CM* $QU;
574     
575#
576#  LB 20  Break before and after CB.
577#         nothing needed here.
578#
579
580# LB 20.09 added rule for Finnish tailoring
581$AL ($HY | $HH) / $SP;
582
583# LB 21
584$CM* ($BA | $HH | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
585
586$CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
587[^$CB] $CM* $BB;                                      # 
588
589# LB21a
590[^$CB] $CM* ($HY | $BA | $HH) $CM* $HL;
591
592# LB21b (reverse)
593$CM* $HL $CM* $SY;
594
595# LB 22
596$CM* $IN $CM* ($ALPlus | $HL);
597$CM* $IN $CM* $ID;
598$CM* $IN $CM* $IN;
599$CM* $IN $CM* $NU;
600
601# LB 23
602$CM* $PO $CM* $ID;
603$CM* $NU $CM* ($ALPlus | $HL);
604$CM* ($ALPlus | $HL) $CM* $NU;
605
606# LB 24
607$CM* $ID $CM* $PR;
608$CM* ($ALPlus | $HL) $CM* $PR;
609$CM* ($ALPlus | $HL) $CM* $PO;
610
611
612# LB 25
613($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
614
615# LB 26
616$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
617$CM* ($JT | $JV) $CM* ($H2 | $JV);
618$CM* $JT $CM* ($H3 | $JT);
619
620# LB 27
621$CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
622$CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
623$CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
624
625# LB 28
626$CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);
627
628
629# LB 29
630$CM* ($ALPlus | $HL) $CM* $IS;
631
632# LB 30
633$CM* $OP $CM* ($ALPlus | $HL | $NU);
634$CM* ($ALPlus | $HL | $NU) $CM* $CP;
635
636# LB 30a
637$CM* $RI $CM* $RI;
638
639## -------------------------------------------------
640
641!!safe_reverse;
642
643# LB 9
644$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
645$CM+ $SP / .;
646
647# LB 14
648$SP+ $CM* $OP;
649
650# LB 15
651$SP+ $CM* $QU;
652
653# LB 16
654$SP+ $CM* ($CL | $CP);
655
656# LB 17
657$SP+ $CM* $B2;
658
659# LB 21
660$CM* ($HY | $BA | $HH) $CM* $HL;
661
662# LB 25
663($CM* ($IS | $SY))+ $CM* $NU;
664($CL | $CP) $CM* ($NU | $IS | $SY);
665
666# For dictionary-based break
667$dictionary $dictionary;
668
669## -------------------------------------------------
670
671!!safe_forward;
672
673# Skip forward over all character classes that are involved in
674#   rules containing patterns with possibly more than one char
675#   of context.
676#
677#  It might be slightly more efficient to have specific rules
678#  instead of one generic one, but only if we could
679#  turn off rule chaining.  We don't want to move more
680#  than necessary.
681#
682[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
683$dictionary $dictionary;
684
685