1# Copyright (C) 2016 and later: Unicode, Inc. and others.
2# License & terms of use: http://www.unicode.org/copyright.html
3# Copyright (c) 2002-2016  International Business Machines Corporation and
4# others. All Rights Reserved.
5#
6#  file:  line.txt
7#
8#         Line Breaking Rules
9#         Implement default line breaking as defined by
10#         Unicode Standard Annex #14 Revision 37 for Unicode 9.0
11#         http://www.unicode.org/reports/tr14/
12#
13#         Includes extensions to the handling of emoji ZWJ sequences from
14#         https://goo.gl/cluFCn
15#
16#         TODO:  Rule LB 8 remains as it was in Unicode 5.2
17#         This is only because of a limitation of ICU break engine implementation,
18#         not because the older behavior is desirable.
19#
20#         This corresponds to CSS line-break=strict (BCP47 -u-lb-strict).
21#         It sets characters of class CJ to behave like NS.
22
23#
24#  Character Classes defined by TR 14.
25#
26
27!!chain;
28
29$AI = [:LineBreak =  Ambiguous:];
30$AL = [:LineBreak =  Alphabetic:];
31$BA = [:LineBreak =  Break_After:];
32$BB = [:LineBreak =  Break_Before:];
33$BK = [:LineBreak =  Mandatory_Break:];
34$B2 = [:LineBreak =  Break_Both:];
35$CB = [:LineBreak =  Contingent_Break:];
36$CJ = [:LineBreak =  Conditional_Japanese_Starter:];
37$CL = [:LineBreak =  Close_Punctuation:];
38# $CM = [:LineBreak =  Combining_Mark:];
39$CP = [:LineBreak =  Close_Parenthesis:];
40$CR = [:LineBreak =  Carriage_Return:];
41$EB = [[:LineBreak =  EB:]\U0001F3C2\U0001F3C7\U0001F3CC\U0001F46A-\U0001F46D\U0001F46F\U0001F574\U0001F6CC];
42$EM = [:LineBreak =  EM:];
43$EX = [:LineBreak =  Exclamation:];
44$GL = [:LineBreak =  Glue:];
45$HL = [:LineBreak =  Hebrew_Letter:];
46$HY = [:LineBreak =  Hyphen:];
47$H2 = [:LineBreak =  H2:];
48$H3 = [:LineBreak =  H3:];
49$ID = [:LineBreak =  Ideographic:];
50$IN = [:LineBreak =  Inseperable:];
51$IS = [:LineBreak =  Infix_Numeric:];
52$JL = [:LineBreak =  JL:];
53$JV = [:LineBreak =  JV:];
54$JT = [:LineBreak =  JT:];
55$LF = [:LineBreak =  Line_Feed:];
56$NL = [:LineBreak =  Next_Line:];
57# NS includes CJ for CSS strict line breaking.
58$NS = [[:LineBreak =  Nonstarter:] $CJ];
59$NU = [:LineBreak =  Numeric:];
60$OP = [:LineBreak =  Open_Punctuation:];
61$PO = [:LineBreak =  Postfix_Numeric:];
62$PR = [:LineBreak =  Prefix_Numeric:];
63$QU = [:LineBreak =  Quotation:];
64$RI = [:LineBreak =  Regional_Indicator:];
65$SA = [:LineBreak =  Complex_Context:];
66$SG = [:LineBreak =  Surrogate:];
67$SP = [:LineBreak =  Space:];
68$SY = [:LineBreak =  Break_Symbols:];
69$WJ = [:LineBreak =  Word_Joiner:];
70$XX = [:LineBreak =  Unknown:];
71$ZW = [:LineBreak =  ZWSpace:];
72$ZWJ = [:LineBreak = ZWJ:];
73
74$EmojiNRK = [[\p{Emoji}] - [$RI \u002a\u00230-9©®™〰〽]];
75# Data for Extended Pictographic scraped from CLDR common/properties/ExtendedPictographic.txt, r12773
76$Extended_Pict = [\U0001F774-\U0001F77F\u2700-\u2701\u2703-\u2704\u270E\u2710-\u2711\u2765-\u2767\U0001F030-\U0001F093\U0001F094-\U0001F09F\U0001F10D-\U0001F10F\U0001F12F\U0001F16C-\U0001F16F\U0001F1AD-\U0001F1E5\U0001F203-\U0001F20F\U0001F23C-\U0001F23F\U0001F249-\U0001F24F\U0001F252-\U0001F2FF\U0001F7D5-\U0001F7FF\U0001F000-\U0001F003\U0001F005-\U0001F02B\U0001F02C-\U0001F02F\U0001F322-\U0001F323\U0001F394-\U0001F395\U0001F398\U0001F39C-\U0001F39D\U0001F3F1-\U0001F3F2\U0001F3F6\U0001F4FE\U0001F53E-\U0001F548\U0001F54F\U0001F568-\U0001F56E\U0001F571-\U0001F572\U0001F57B-\U0001F586\U0001F588-\U0001F589\U0001F58E-\U0001F58F\U0001F591-\U0001F594\U0001F597-\U0001F5A3\U0001F5A6-\U0001F5A7\U0001F5A9-\U0001F5B0\U0001F5B3-\U0001F5BB\U0001F5BD-\U0001F5C1\U0001F5C5-\U0001F5D0\U0001F5D4-\U0001F5DB\U0001F5DF-\U0001F5E0\U0001F5E2\U0001F5E4-\U0001F5E7\U0001F5E9-\U0001F5EE\U0001F5F0-\U0001F5F2\U0001F5F4-\U0001F5F9\u2605\u2607-\u260D\u260F-\u2610\u2612\u2616-\u2617\u2619-\u261C\u261E-\u261F\u2621\u2624-\u2625\u2627-\u2629\u262B-\u262D\u2630-\u2637\u263B-\u2647\u2654-\u265F\u2661-\u2662\u2664\u2667\u2669-\u267A\u267C-\u267E\u2680-\u2691\u2695\u2698\u269A\u269D-\u269F\u26A2-\u26A9\u26AC-\u26AF\u26B2-\u26BC\u26BF-\u26C3\u26C6-\u26C7\u26C9-\u26CD\u26D0\u26D2\u26D5-\u26E8\u26EB-\u26EF\u26F6\u26FB-\u26FC\u26FE-\u26FF\u2388\U0001FA00-\U0001FFFD\U0001F0A0-\U0001F0AE\U0001F0B1-\U0001F0BF\U0001F0C1-\U0001F0CF\U0001F0D1-\U0001F0F5\U0001F0AF-\U0001F0B0\U0001F0C0\U0001F0D0\U0001F0F6-\U0001F0FF\U0001F80C-\U0001F80F\U0001F848-\U0001F84F\U0001F85A-\U0001F85F\U0001F888-\U0001F88F\U0001F8AE-\U0001F8FF\U0001F900-\U0001F90F\U0001F91F\U0001F928-\U0001F92F\U0001F931-\U0001F932\U0001F93F\U0001F94C-\U0001F94F\U0001F95F-\U0001F97F\U0001F992-\U0001F9BF\U0001F9C1-\U0001F9FF\U0001F6C6-\U0001F6CA\U0001F6E6-\U0001F6E8\U0001F6EA\U0001F6F1-\U0001F6F2\U0001F6D3-\U0001F6DF\U0001F6ED-\U0001F6EF\U0001F6F7-\U0001F6FF];
77
78# By LB9, a ZWJ also behaves as a CM. Including it in the definition of CM avoids having to explicitly
79#         list it in the numerous rules that use CM.
80# By LB1, SA characters with general categor of Mn or Mc also resolve to CM.
81
82$CM = [[:LineBreak = Combining_Mark:] $ZWJ [$SA & [[:Mn:][:Mc:]]]];
83
84#   Dictionary character set, for triggering language-based break engines. Currently
85#   limited to LineBreak=Complex_Context (SA).
86
87$dictionary = [$SA];
88
89#
90#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
91#                               SA  (Dictionary chars, excluding Mn and Mc)
92#                               SG  (Unpaired Surrogates)
93#                               XX  (Unknown, unassigned)
94#                         as $AL  (Alphabetic)
95#
96$ALPlus = [$AL $AI $SG $XX [$SA-[[:Mn:][:Mc:]]]];
97
98
99## -------------------------------------------------
100
101!!forward;
102
103#
104# CAN_CM  is the set of characters that may combine with CM combining chars.
105#         Note that Linebreak UAX 14's concept of a combining char and the rules
106#         for what they can combine with are _very_ different from the rest of Unicode.
107#
108#         Note that $CM itself is left out of this set.  If CM is needed as a base
109#         it must be listed separately in the rule.
110#
111$CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
112$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs
113
114#
115# AL_FOLLOW  set of chars that can unconditionally follow an AL
116#            Needed in rules where stand-alone $CM s are treated as AL.
117#
118$AL_FOLLOW      = [$BK $CR $LF $NL $ZW $SP $CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $PR $PO $ALPlus];
119
120
121#
122#  Rule LB 4, 5    Mandatory (Hard) breaks.
123#
124$LB4Breaks    = [$BK $CR $LF $NL];
125$LB4NonBreaks = [^$BK $CR $LF $NL $CM];
126$CR $LF {100};
127
128#
129#  LB 6    Do not break before hard line breaks.
130#
131$LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
132$CAN_CM $CM*    $LB4Breaks {100};
133^$CM+           $LB4Breaks {100};
134
135# LB 7         x SP
136#              x ZW
137$LB4NonBreaks [$SP $ZW];
138$CAN_CM $CM*  [$SP $ZW];
139^$CM+         [$SP $ZW];
140
141#
142# LB 8         Break after zero width space
143#              TODO:  ZW SP* <break>
144#              An engine change is required to write the reverse rule for this.
145#              For now, leave the Unicode 5.2 rule, ZW <break>
146#
147$LB8Breaks    = [$LB4Breaks $ZW];
148$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
149
150# LB 8a        ZWJ x (ID | Extended_Pict | EmojiNRK)   Emoji ZWJ sequences.
151#
152$ZWJ ($ID | $Extended_Pict | $EmojiNRK);
153
154# LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
155#                                $CM not covered by the above needs to behave like $AL
156#                                See definition of $CAN_CM.
157
158$CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
159^$CM+;
160
161#
162# LB 11  Do not break before or after WORD JOINER & related characters.
163#
164$CAN_CM $CM*  $WJ;
165$LB8NonBreaks $WJ;
166^$CM+         $WJ;
167
168$WJ $CM* .;
169
170#
171# LB 12  Do not break after NBSP and related characters.
172#         GL  x
173#
174$GL $CM* .;
175
176#
177# LB 12a  Do not break before NBSP and related characters ...
178#            [^SP BA HY] x GL
179#
180[[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GL;
181^$CM+ $GL;
182
183
184
185#
186# LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
187#
188$LB8NonBreaks $CL;
189$CAN_CM $CM*  $CL;
190^$CM+         $CL;              # by rule 10, stand-alone CM behaves as AL
191
192$LB8NonBreaks $CP;
193$CAN_CM $CM*  $CP;
194^$CM+         $CP;              # by rule 10, stand-alone CM behaves as AL
195
196$LB8NonBreaks $EX;
197$CAN_CM $CM*  $EX;
198^$CM+         $EX;              # by rule 10, stand-alone CM behaves as AL
199
200$LB8NonBreaks $IS;
201$CAN_CM $CM*  $IS;
202^$CM+         $IS;              # by rule 10, stand-alone CM behaves as AL
203
204$LB8NonBreaks $SY;
205$CAN_CM $CM*  $SY;
206^$CM+         $SY;              # by rule 10, stand-alone CM behaves as AL
207
208
209#
210# LB 14  Do not break after OP, even after spaces
211#
212$OP $CM* $SP* .;
213
214$OP $CM* $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL
215                                   # by rule 8, CM following a SP is stand-alone.
216
217# LB 15
218$QU $CM* $SP* $OP;
219
220# LB 16
221($CL | $CP) $CM* $SP* $NS;
222
223# LB 17
224$B2 $CM* $SP* $B2;
225
226#
227# LB 18  Break after spaces.
228#
229$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
230$LB18Breaks    = [$LB8Breaks $SP];
231
232
233# LB 19
234#         x QU
235$LB18NonBreaks $CM* $QU;
236^$CM+               $QU;
237
238#         QU  x
239$QU $CM* .;
240
241
242# LB 20
243#        <break>  $CB
244#        $CB   <break>
245
246$LB20NonBreaks = [$LB18NonBreaks - $CB];
247
248# LB 21        x   (BA | HY | NS)
249#           BB x
250#
251$LB20NonBreaks $CM* ($BA | $HY | $NS);
252^$CM+ ($BA | $HY | $NS);
253
254$BB $CM* [^$CB];                                  #  $BB  x
255$BB $CM* $LB20NonBreaks;
256
257# LB 21a Don't break after Hebrew + Hyphen
258#   HL (HY | BA) x
259#
260$HL $CM* ($HY | $BA) $CM* [^$CB]?;
261
262# LB 21b (forward) Don't break between SY and HL
263# (break between HL and SY already disallowed by LB 13 above)
264$SY $CM* $HL;
265
266# LB 22
267($ALPlus | $HL) $CM* $IN;
268^$CM+    $IN;     #  by rule 10, any otherwise unattached CM behaves as AL
269$EX $CM*    $IN;
270($ID | $EB | $EM) $CM*  $IN;
271$IN $CM*    $IN;
272$NU $CM*    $IN;
273
274
275# $LB 23
276#
277($ALPlus | $HL) $CM* $NU;
278^$CM+  $NU;       # Rule 10, any otherwise unattached CM behaves as AL
279$NU $CM* ($ALPlus | $HL);
280
281# LB 23a
282#
283$PR $CM* ($ID | $EB | $EM);
284($ID | $EB | $EM) $CM*  $PO;
285
286
287#
288# LB 24
289#
290($PR | $PO) $CM* ($ALPlus | $HL);
291($ALPlus | $HL) $CM* ($PR | $PO);
292^$CM+ ($PR | $PO);       # Rule 10, any otherwise unattached CM behaves as AL
293
294#
295# LB 25   Numbers.
296#
297(($PR | $PO) $CM*)? (($OP | $HY) $CM*)? $NU ($CM* ($NU | $SY | $IS))*
298    ($CM* ($CL | $CP))? ($CM* ($PR | $PO))?;
299
300# LB 26  Do not break a Korean syllable
301#
302$JL $CM* ($JL | $JV | $H2 | $H3);
303($JV | $H2) $CM* ($JV | $JT);
304($JT | $H3) $CM* $JT;
305
306# LB 27  Treat korean Syllable Block the same as ID  (don't break it)
307($JL | $JV | $JT | $H2 | $H3) $CM* $IN;
308($JL | $JV | $JT | $H2 | $H3) $CM* $PO;
309$PR $CM* ($JL | $JV | $JT | $H2 | $H3);
310
311
312# LB 28   Do not break between alphabetics
313#
314($ALPlus | $HL) $CM* ($ALPlus | $HL);
315^$CM+ ($ALPlus | $HL);      # The $CM+ is from rule 10, an unattached CM is treated as AL
316
317# LB 29
318$IS $CM* ($ALPlus | $HL);
319
320# LB 30
321($ALPlus | $HL | $NU) $CM* $OP;
322^$CM+ $OP;         # The $CM+ is from rule 10, an unattached CM is treated as AL.
323$CP $CM* ($ALPlus | $HL | $NU);
324
325# LB 30a  Do not break between regional indicators. Break after pairs of them.
326#         Tricky interaction with LB8a: ZWJ x ID
327$RI $CM* $RI                / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
328$RI $CM* $RI $CM*  $ZWJ     / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM $ID $Extended_Pict $EmojiNRK] {eof}];
329$RI $CM* $RI $CM* [$CM-$ZWJ] / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
330
331$RI $CM* $RI $CM* [$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS {eof}];
332$RI $CM* $RI $CM* $ZWJ ($ID | $Extended_Pict | $EmojiNRK);
333
334# LB 30b Do not break between an Emoji Base and an Emoji Modifier
335$EB $CM* $EM;
336
337#
338#  Reverse Rules.
339#
340## -------------------------------------------------
341
342!!reverse;
343
344#  LB 9 Combining Marks.
345#  Stick together any combining sequences that don't match other rules.
346
347^$CM+ $CAN_CM?;
348
349#
350#  Sequences of the form  (shown forwards)
351#      [CANT_CM]  <break>  [CM]  [whatever]
352#  The CM needs to behave as an AL
353#
354$AL_FOLLOW $CM+ / (
355          [$BK $CR $LF $NL $ZW {eof}] |
356          $SP+ $CM+ $SP |
357          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
358                                               #  LB14 says    OP SP* x .
359                                               #    becomes    OP SP* x AL
360                                               #    becomes    OP SP* x CM+ AL_FOLLOW
361                                               #
362                                               # Further note:  the $AL in [$AL {eof}] is only to work around
363                                               #                a rule compiler bug which complains about
364                                               #                empty sets otherwise.
365
366
367# LB 4, 5, 6
368
369$LB4Breaks [$LB4NonBreaks-$CM];
370$LB4Breaks $CM+ $CAN_CM;
371$LF $CR;
372
373
374# LB 7         x SP
375#              x ZW
376[$SP $ZW] [$LB4NonBreaks-$CM];
377[$SP $ZW] $CM+ $CAN_CM;
378
379# LB 8 ZW SP* <break>
380#     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
381#           Requires an engine enhancement.
382#   / $SP* $ZW
383
384# LB 8a        ZWJ x (ID | Extended_Pict | EmojiNRK)
385#
386($ID | $Extended_Pict | $EmojiNRK) $ZWJ $CM* $CAN_CM?;
387
388
389# LB 9,10  Combining marks.
390#    X   $CM needs to behave like X, where X is not $SP or controls.
391#    $CM not covered by the above needs to behave like $AL
392# Stick together any combining sequences that don't match other rules.
393^$CM+ $CAN_CM;
394
395
396# LB 11
397#
398$WJ $CM* $CAN_CM;
399$WJ      [$LB8NonBreaks-$CM];
400
401     $CANT_CM $CM* $WJ;
402$CAN_CM  $CM* $WJ;
403
404# LB 12a
405#      [^SP BA HY] x GL
406#
407$GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
408
409# LB 12
410#     GL  x
411#
412$CANT_CM $CM* $GL;
413$CAN_CM $CM* $GL;
414
415
416# LB 13
417$CL $CM+ $CAN_CM;
418$CP $CM+ $CAN_CM;
419$EX $CM+ $CAN_CM;
420$IS $CM+ $CAN_CM;
421$SY $CM+ $CAN_CM;
422
423$CL [$LB8NonBreaks-$CM];
424$CP [$LB8NonBreaks-$CM];
425$EX [$LB8NonBreaks-$CM];
426$IS [$LB8NonBreaks-$CM];
427$SY [$LB8NonBreaks-$CM];
428
429
430# LB 14    OP SP* x
431#
432.   $SP* $CM* $OP;
433$AL_FOLLOW? $CM+ $SP+ $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP+ $CM* $OP
434
435
436# LB 15
437$OP $SP* $CM* $QU;
438
439# LB 16
440$NS $SP* $CM* ($CL | $CP);
441
442# LB 17
443$B2 $SP* $CM* $B2;
444
445# LB 18  break after spaces
446#        Nothing explicit needed here.
447
448
449#
450# LB 19
451#
452$QU $CM* $CAN_CM;                                #   . x QU
453$QU      $LB18NonBreaks;
454
455
456$CAN_CM  $CM* $QU;                               #   QU x .
457     $CANT_CM $CM* $QU;
458
459#
460#  LB 20  Break before and after CB.
461#         nothing needed here.
462#
463
464# LB 21
465($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)
466
467[$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
468[^$CB] $CM* $BB;                                      #
469
470# LB21a
471[^$CB] $CM* ($HY | $BA) $CM* $HL;
472
473# LB21b (reverse)
474$HL $CM* $SY;
475
476# LB 22
477$IN $CM* ($ALPlus | $HL);
478$IN $CM* $EX;
479$IN $CM* ($ID | $EB | $EM);
480$IN $CM* $IN;
481$IN $CM* $NU;
482
483# LB 23
484$NU $CM* ($ALPlus | $HL);
485($ALPlus | $HL) $CM* $NU;
486
487# LB23a
488($ID | $EB | $EM) $CM* $PR;
489$PO $CM* ($ID | $EB | $EM);
490
491# LB 24
492($ALPlus | $HL) $CM* ($PR | $PO);
493($PR | $PO) $CM* ($ALPlus | $HL);
494
495
496# LB 25
497($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
498
499# LB 26
500($H3 | $H2 | $JV | $JL) $CM* $JL;
501($JT | $JV) $CM* ($H2 | $JV);
502$JT $CM* ($H3 | $JT);
503
504# LB 27
505$IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
506$PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
507 ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
508
509# LB 28
510($ALPlus | $HL) $CM* ($ALPlus | $HL);
511
512
513# LB 29
514($ALPlus | $HL) $CM* $IS;
515
516# LB 30
517$OP $CM* ($ALPlus | $HL | $NU);
518($ALPlus | $HL | $NU) $CM* $CP;
519
520# LB 30a
521#    Pairs of Regional Indicators.
522#    The following two rules are nearly identical. The first matches only sequences with an odd number of adjacent RIs,
523#    the second with an even number. Stripping away the cruft they look like
524#         [^RI] RI / (RI RI)+ ^RI;
525#         [^RI] RI RI / (RI RI)+ ^RI;
526#
527[{bof} $NS $HY $BA $QU $CL $CP $EX $IS $SY $WJ $GL $ZW $SP $BK $CR $LF $NL $ZWJ] $CM* $RI / ($CM* $RI $CM* $RI)+ $CM* [{eof}[^$RI $CM]];
528[{bof} $NS $HY $BA $QU $CL $CP $EX $IS $SY $WJ $GL $ZW $SP $BK $CR $LF $NL $ZWJ] $CM* $RI $CM* $RI / ($CM* $RI $CM* $RI)+ $CM* [{eof}[^$RI $CM]];
529
530# In general, adjacent RIs stay together. The hard-break rules, above, overide this, forcing in the boundaries between pairs.
531$RI $CM* $RI;
532
533#    WJ, GL, QU, etc. are classes with rules like "WJ x "   which includes "WJ x RI".
534$RI $CM* ([$WJ $GL $QU $BB] |  (($HY | $BA)$CM* $HL));
535
536
537# LB 30b Do not break between an Emoji Base and an Emoji Modifier
538$EM $CM* $EB;
539
540
541## -------------------------------------------------
542
543!!safe_reverse;
544
545# LB 9
546^$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
547^$CM+ $SP / .;
548
549# LB 14
550$SP+ $CM* $OP;
551
552# LB 15
553$SP+ $CM* $QU;
554
555# LB 16
556$SP+ $CM* ($CL | $CP);
557
558# LB 17
559$SP+ $CM* $B2;
560
561# LB 21
562$CM* ($HY | $BA) $CM* $HL;
563
564# LB 25
565($CM* ($IS | $SY))+ $CM* $NU;
566($CL | $CP) $CM* ($NU | $IS | $SY);
567
568#  LB 30
569($CM* $RI)+;
570
571# For dictionary-based break
572$dictionary $dictionary;
573
574## -------------------------------------------------
575
576!!safe_forward;
577
578# Skip forward over all character classes that are involved in
579#   rules containing patterns with possibly more than one char
580#   of context.
581#
582#  It might be slightly more efficient to have specific rules
583#  instead of one generic one, but only if we could
584#  turn off rule chaining.  We don't want to move more
585#  than necessary.
586#
587^[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $RI $ZWJ $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $RI $ZWJ $dictionary];
588$dictionary $dictionary;
589
590