[apple/icu.git] / icuSources / data / brkitr / line_normal_cj.txt

# Copyright (c) 2002-2015  International Business Machines Corporation and
# others. All Rights Reserved.
#
#  file:  line_normal_cj.txt
#
#         Line Breaking Rules
#         Implement default line breaking as defined by 
#         Unicode Standard Annex #14 Revision 29 for Unicode 6.2
#         http://www.unicode.org/reports/tr14/
#         tailored as noted in 2nd paragraph below..
#
#         TODO:  Rule LB 8 remains as it was in Unicode 5.2
#         This is only because of a limitation of ICU break engine implementation,
#         not because the older behavior is desirable.
#
#         This tailors the line break behavior to correspond to CSS
#         line-break=normal (BCP47 -u-lb-normal) as defined for Chinese & Japanese.
#         It sets characters of class CJ to behave like ID.
#         In addition, it allows breaks:
#         * before hyphens 2010 & 2013 (both BA) and 301C, 30A0 (both NS)

#
#  Character Classes defined by TR 14.
#

!!chain;
!!LBCMNoChain;


!!lookAheadHardBreak;
#
#  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere
#                          and only used for the line break rules.
#
#           It is used in the implementation of rule LB 10
#           which says to treat any combining mark that is not attached to a base
#           character as if it were of class AL  (alphabetic).
#
#           The problem occurs in the reverse rules.
#
#           Consider a sequence like, with correct breaks as shown
#               LF  ID  CM  AL  AL
#                  ^       ^       ^
#           Then consider the sequence without the initial ID (ideographic)
#                 LF  CM  AL  AL
#                    ^           ^
#           Our CM, which in the first example was attached to the ideograph,
#           is now unattached, becomes an alpha, and joins in with the other
#           alphas.
#
#           When iterating forwards, these sequences do not present any problems
#           When iterating backwards, we need to look ahead when encountering
#           a CM to see whether it attaches to something further on or not.
#           (Look-ahead in a reverse rule is looking towards the start)
#
#           If the CM is unattached, we need to force a break.
#
#           !!lookAheadHardBreak forces the run time state machine to
#           stop immediately when a look ahead rule ( '/' operator) matches,
#           and set the match position to that of the look-ahead operator,
#           no matter what other rules may be in play at the time.
#
#           See rule LB 19 for an example.
#

$AI = [:LineBreak =  Ambiguous:];
$AL = [:LineBreak =  Alphabetic:];
$BAX = [\u2010 \u2013];
$BA = [[:LineBreak =  Break_After:] - $BAX];
$BB = [:LineBreak =  Break_Before:];
$BK = [:LineBreak =  Mandatory_Break:];
$B2 = [:LineBreak =  Break_Both:];
$CB = [:LineBreak =  Contingent_Break:];
$CJ = [:LineBreak =  Conditional_Japanese_Starter:];
$CL = [:LineBreak =  Close_Punctuation:];
$CM = [:LineBreak =  Combining_Mark:];
$CP = [:LineBreak =  Close_Parenthesis:];
$CR = [:LineBreak =  Carriage_Return:];
$EX = [:LineBreak =  Exclamation:];
$GL = [:LineBreak =  Glue:];
$HL = [:LineBreak =  Hebrew_Letter:];
$HY = [:LineBreak =  Hyphen:];
$H2 = [:LineBreak =  H2:];
$H3 = [:LineBreak =  H3:];
$ID = [[:LineBreak =  Ideographic:] $CJ];
$IN = [:LineBreak =  Inseperable:];
$IS = [:LineBreak =  Infix_Numeric:];
$JL = [:LineBreak =  JL:];
$JV = [:LineBreak =  JV:];
$JT = [:LineBreak =  JT:];
$LF = [:LineBreak =  Line_Feed:];
$NL = [:LineBreak =  Next_Line:];
$NSX = [\u301C \u30A0];
$NS = [[:LineBreak =  Nonstarter:] - $NSX];
$NU = [:LineBreak =  Numeric:];
$OP = [:LineBreak =  Open_Punctuation:];
$PO = [:LineBreak =  Postfix_Numeric:];
$PR = [:LineBreak =  Prefix_Numeric:];
$QU = [:LineBreak =  Quotation:];
$RI = [:LineBreak =  Regional_Indicator:];
$SA = [:LineBreak =  Complex_Context:];
$SG = [:LineBreak =  Surrogate:];
$SP = [:LineBreak =  Space:];
$SY = [:LineBreak =  Break_Symbols:];
$WJ = [:LineBreak =  Word_Joiner:];
$XX = [:LineBreak =  Unknown:];
$ZW = [:LineBreak =  ZWSpace:];

# Special character classes for people & body part emoji:
# Subsets of $CM:
$ZWJ = \u200D;
$EmojiVar = \uFE0F;
# The following are subsets of $ID
$EmojiForSeqs = [\u2764 \U0001F466-\U0001F469 \U0001F48B];
$EmojiForMods = [\u261D \u270A-\u270C \U0001F385 \U0001F3C3-\U0001F3C4 \U0001F3C7 \U0001F3CA \U0001F442-\U0001F443 \U0001F446-\U0001F450 \U0001F466-\U0001F469 \U0001F46E-\U0001F478 \U0001F47C \U0001F481-\U0001F483 \U0001F485-\U0001F487 \U0001F4AA \U0001F596 \U0001F645-\U0001F647 \U0001F64B-\U0001F64F \U0001F6A3 \U0001F6B4-\U0001F6B6 \U0001F6C0];
$EmojiMods = [\U0001F3FB-\U0001F3FF];

#   Dictionary character set, for triggering language-based break engines. Currently
#   limited to LineBreak=Complex_Context. Note that this set only works in Unicode
#   5.0 or later as the definition of Complex_Context was corrected to include all
#   characters requiring dictionary break.

$dictionary = [:LineBreak = Complex_Context:];

#
#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),
#                               SA  (South East Asian: Thai, Lao, Khmer)
#                               SG  (Unpaired Surrogates)
#                               XX  (Unknown, unassigned)
#                         as $AL  (Alphabetic)
#
$ALPlus = [$AL $AI $SA $SG $XX];

#
#  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.
#
$ALcm = $ALPlus $CM*;
$BAcm = $BA $CM*;
$BAXcm = $BAX $CM*;
$BBcm = $BB $CM*;
$B2cm = $B2 $CM*;
$CLcm = $CL $CM*;
$CPcm = $CP $CM*;
$EXcm = $EX $CM*;
$GLcm = $GL $CM*;
$HLcm = $HL $CM*;
$HYcm = $HY $CM*;
$H2cm = $H2 $CM*;
$H3cm = $H3 $CM*;
$IDcm = $ID $CM*;
$INcm = $IN $CM*;
$IScm = $IS $CM*;
$JLcm = $JL $CM*;
$JVcm = $JV $CM*;
$JTcm = $JT $CM*;
$NScm = $NS $CM*;
$NSXcm = $NSX $CM*;
$NUcm = $NU $CM*;
$OPcm = $OP $CM*;
$POcm = $PO $CM*;
$PRcm = $PR $CM*;
$QUcm = $QU $CM*;
$RIcm = $RI $CM*;
$SYcm = $SY $CM*;
$WJcm = $WJ $CM*;

## -------------------------------------------------

!!forward;

#
#  Each class of character can stand by itself as an unbroken token, with trailing combining stuff
#
$ALPlus $CM+;
$BA $CM+;
$BAX $CM+;
$BB $CM+;
$B2 $CM+;
$CL $CM+;
$CP $CM+;
$EX $CM+;
$GL $CM+;
$HL $CM+;
$HY $CM+;
$H2 $CM+;
$H3 $CM+;
$ID $CM+;
$IN $CM+;
$IS $CM+;
$JL $CM+;
$JV $CM+;
$JT $CM+;
$NS $CM+;
$NSX $CM+;
$NU $CM+;
$OP $CM+;
$PO $CM+;
$PR $CM+;
$QU $CM+;
$RI $CM+;
$SY $CM+;
$WJ $CM+;

#
# CAN_CM  is the set of characters that may combine with CM combining chars.
#         Note that Linebreak UAX 14's concept of a combining char and the rules
#         for what they can combine with are _very_ different from the rest of Unicode.
#
#         Note that $CM itself is left out of this set.  If CM is needed as a base
#         it must be listed separately in the rule.
#
$CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs
$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs

#
# AL_FOLLOW  set of chars that can unconditionally follow an AL
#            Needed in rules where stand-alone $CM s are treated as AL.
#            Chaining is disabled with CM because it causes other failures,
#            so for this one case we need to manually list out longer sequences.
#
$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
$AL_FOLLOW_CM   = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $BAX $HY $NS $NSX $IN $NU $ALPlus];
$AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];


#
#  Rule LB 4, 5    Mandatory (Hard) breaks.
#
$LB4Breaks    = [$BK $CR $LF $NL];
$LB4NonBreaks = [^$BK $CR $LF $NL];
$CR $LF {100};

#
#  LB 6    Do not break before hard line breaks.
#
$LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.
$CAN_CM $CM*    $LB4Breaks {100};
$CM+            $LB4Breaks {100};

# LB 7         x SP
#              x ZW
$LB4NonBreaks [$SP $ZW];
$CAN_CM $CM*  [$SP $ZW];
$CM+          [$SP $ZW];

#
# LB 8         Break after zero width space
#              TODO:  ZW SP* <break>
#              An engine change is required to write the reverse rule for this.
#              For now, leave the Unicode 5.2 rule, ZW <break>
#
$LB8Breaks    = [$LB4Breaks $ZW];
$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];


# Special forward rule for people & body part emoji:
# don't break $ZWJ from subsequent $EmojiForSeqs
$EmojiForSeqs $EmojiVar? $EmojiMods? $ZWJ $EmojiForSeqs;

# LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL 
#                                $CM not covered by the above needs to behave like $AL   
#                                See definition of $CAN_CM.

$CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.
$CM+;

#
# LB 11  Do not break before or after WORD JOINER & related characters.
#
$CAN_CM $CM*  $WJcm;
$LB8NonBreaks $WJcm;
$CM+          $WJcm;

$WJcm $CANT_CM;
$WJcm $CAN_CM $CM*;

#
# LB 12  Do not break after NBSP and related characters.
#         GL  x
#
$GLcm $CAN_CM $CM*;
$GLcm $CANT_CM;
 
#
# LB 12a  Do not break before NBSP and related characters ...
#            [^SP BA HY] x GL
#
[[$LB8NonBreaks] - [$SP $BA $BAX $HY]] $CM* $GLcm;
$CM+ GLcm;


#
# LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.
#
$LB8NonBreaks $CL;
$CAN_CM $CM*  $CL;
$CM+          $CL;              # by rule 10, stand-alone CM behaves as AL

$LB8NonBreaks $CP;
$CAN_CM $CM*  $CP;
$CM+          $CP;              # by rule 10, stand-alone CM behaves as AL

$LB8NonBreaks $EX;
$CAN_CM $CM*  $EX;
$CM+          $EX;              # by rule 10, stand-alone CM behaves as AL

$LB8NonBreaks $IS;
$CAN_CM $CM*  $IS;
$CM+          $IS;              # by rule 10, stand-alone CM behaves as AL

$LB8NonBreaks $SY;
$CAN_CM $CM*  $SY;
$CM+          $SY;              # by rule 10, stand-alone CM behaves as AL


#
# LB 14  Do not break after OP, even after spaces
#
$OPcm $SP* $CAN_CM $CM*;
$OPcm $SP* $CANT_CM;

$OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL

# LB 15
$QUcm $SP* $OPcm;

# LB 16
# Do not break between closing punctuation and $NS, even with intervening spaces
# But DO allow a break between closing punctuation and $NSX, don't include it here
($CLcm | $CPcm) $SP* $NScm;

# LB 17
$B2cm $SP* $B2cm;

#
# LB 18  Break after spaces.
#
$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
$LB18Breaks    = [$LB8Breaks $SP];


# LB 19
#         x QU
$LB18NonBreaks $CM* $QUcm;
$CM+                $QUcm;

#         QU  x
$QUcm .?;
$QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.
                              #  TODO:  I don't think this rule is needed.


# LB 20
#        <break>  $CB
#        $CB   <break>

$LB20NonBreaks = [$LB18NonBreaks - $CB];

# LB 21        x   (BA | HY | NS)
#           BB x
#
# DO allow breaks here before $BAXcm and $NSXcm, so don't include them
$LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm); 

$BBcm [^$CB];                                  #  $BB  x
$BBcm $LB20NonBreaks $CM*;

# LB 21a Don't break after Hebrew + Hyphen
#   HL (HY | BA) x
#  
$HLcm ($HYcm | $BAcm | $BAXcm) [^$CB]?;

# LB 21b (forward) Don't break between SY and HL
# (break between HL and SY already disallowed by LB 13 above)
$SYcm $HLcm;

# LB 22
($ALcm | $HLcm) $INcm;
$CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL
$IDcm    $INcm;
$INcm    $INcm;
$NUcm    $INcm;


# $LB 23
$IDcm  $POcm;
$ALcm  $NUcm;       # includes $LB19
$HLcm  $NUcm;
$CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL
$NUcm  $ALcm;
$NUcm  $HLcm;

#
# LB 24
#
$PRcm $IDcm;
$PRcm ($ALcm | $HLcm);
$POcm ($ALcm | $HLcm);

#
# LB 25   Numbers.
#
($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* ($CLcm | $CPcm)? ($PRcm | $POcm)?;

# LB 26  Do not break a Korean syllable
#
$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
($JVcm | $H2cm) ($JVcm | $JTcm);
($JTcm | $H3cm) $JTcm;

# LB 27  Treat korean Syllable Block the same as ID  (don't break it)
($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);


# LB 28   Do not break between alphabetics
#
($ALcm | $HLcm) ($ALcm | $HLcm);
$CM+ ($ALcm | $HLcm);      # The $CM+ is from rule 10, an unattached CM is treated as AL

# LB 29
$IScm ($ALcm | $HLcm);

# LB 30
($ALcm | $HLcm | $NUcm) $OPcm;
$CM+ $OPcm;         # The $CM+ is from rule 10, an unattached CM is treated as AL.          
$CPcm ($ALcm | $HLcm | $NUcm);

# LB 30a  Do not break between regional indicators.
$RIcm $RIcm;

# Special forward rule for people & body part emoji:
# don't break between relevant emoji and $EmojiMods
$EmojiForMods $EmojiVar? $EmojiMods;

#
#  Reverse Rules.
#
## -------------------------------------------------

!!reverse;

$CM+ $ALPlus;
$CM+ $BA;
$CM+ $BAX;
$CM+ $BB;
$CM+ $B2;
$CM+ $CL;
$CM+ $CP;
$CM+ $EX;
$CM+ $GL;
$CM+ $HL;
$CM+ $HY;
$CM+ $H2;
$CM+ $H3;
$CM+ $ID;
$CM+ $IN;
$CM+ $IS;
$CM+ $JL;
$CM+ $JV;
$CM+ $JT;
$CM+ $NS;
$CM+ $NSX;
$CM+ $NU;
$CM+ $OP;
$CM+ $PO;
$CM+ $PR;
$CM+ $QU;
$CM+ $RI;
$CM+ $SY;
$CM+ $WJ;
$CM+;


#
#  Sequences of the form  (shown forwards)
#      [CANT_CM]  <break>  [CM]  [whatever]
#  The CM needs to behave as an AL
#
$AL_FOLLOW $CM+ / (
          [$BK $CR $LF $NL $ZW {eof}] |
          $SP+ $CM+ $SP |
          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.
                                               #  LB14 says    OP SP* x .        
                                               #    becomes    OP SP* x AL
                                               #    becomes    OP SP* x CM+ AL_FOLLOW
                                               #
                                               # Further note:  the $AL in [$AL {eof}] is only to work around
                                               #                a rule compiler bug which complains about
                                               #                empty sets otherwise.
          
#
#  Sequences of the form  (shown forwards)
#      [CANT_CM]  <break> [CM]  <break>  [PR]
#  The CM needs to behave as an AL
#  This rule is concerned about getting the second of the two <breaks> in place.
#

[$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];


# LB 4, 5, 5

$LB4Breaks [$LB4NonBreaks-$CM];
$LB4Breaks $CM+ $CAN_CM;
$LF $CR;


# LB 7         x SP
#              x ZW
[$SP $ZW] [$LB4NonBreaks-$CM];
[$SP $ZW] $CM+ $CAN_CM;

# LB 8 ZW SP* <break>
#     TODO: to implement this, we need more than one look-ahead hard break in play at a time.
#           Requires an engine enhancement.
#   / $SP* $ZW

# Special reverse rule for people & body part emoji:
# don't break $ZWJ from subsequent $EmojiForSeqs
$EmojiForSeqs $ZWJ $EmojiMods? $EmojiVar? $EmojiForSeqs;

# LB 9,10  Combining marks.
#    X   $CM needs to behave like X, where X is not $SP or controls.
#    $CM not covered by the above needs to behave like $AL
# Stick together any combining sequences that don't match other rules.
$CM+ $CAN_CM;


# LB 11
$CM* $WJ $CM* $CAN_CM;
$CM* $WJ      [$LB8NonBreaks-$CM];

     $CANT_CM $CM* $WJ;
$CM* $CAN_CM  $CM* $WJ;

# LB 12a
#      [^SP BA HY] x GL
#
$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $BAX $HY]];

# LB 12
#     GL  x
#
$CANT_CM $CM* $GL;
$CM* $CAN_CM $CM* $GL;


# LB 13
$CL $CM+ $CAN_CM;
$CP $CM+ $CAN_CM;
$EX $CM+ $CAN_CM;
$IS $CM+ $CAN_CM;
$SY $CM+ $CAN_CM;

$CL [$LB8NonBreaks-$CM];
$CP [$LB8NonBreaks-$CM];
$EX [$LB8NonBreaks-$CM];
$IS [$LB8NonBreaks-$CM];
$SY [$LB8NonBreaks-$CM];

# Rule 13 & 14 taken together for an edge case.
#   Match this, shown forward
#     OP SP+  ($CM+ behaving as $AL) (CL | CP | EX | IS | IY)
#   This really wants to chain at the $CM+ (which is acting as an $AL)
#   except for $CM chaining being disabled.
[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;  

# LB 14    OP SP* x
#
$CM* $CAN_CM    $SP* $CM* $OP;
     $CANT_CM   $SP* $CM* $OP;
$AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
     
     $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
$CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;
$SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.


# LB 15
$CM* $OP $SP* $CM* $QU;

# LB 16
# Don't include $NSX here
$CM* $NS $SP* $CM* ($CL | $CP);

# LB 17
$CM* $B2 $SP* $CM* $B2;

# LB 18  break after spaces
#        Nothing explicit needed here.


#
# LB 19
#
$CM* $QU $CM* $CAN_CM;                                #   . x QU
$CM* $QU      $LB18NonBreaks;


$CM* $CAN_CM  $CM* $QU;                               #   QU x .
     $CANT_CM $CM* $QU;
     
#
#  LB 20  Break before and after CB.
#         nothing needed here.
#

# LB 21
# Don't include $BAX or $NSX here
$CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)

$CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .
[^$CB] $CM* $BB;                                      # 

# LB21a
[^$CB] $CM* ($HY | $BA | $BAX) $CM* $HL;

# LB21b (reverse)
$CM* $HL $CM* $SY;

# LB 22
$CM* $IN $CM* ($ALPlus | $HL);
$CM* $IN $CM* $ID;
$CM* $IN $CM* $IN;
$CM* $IN $CM* $NU;

# LB 23
$CM* $PO $CM* $ID;
$CM* $NU $CM* ($ALPlus | $HL);
$CM* ($ALPlus | $HL) $CM* $NU;

# LB 24
$CM* $ID $CM* $PR;
$CM* ($ALPlus | $HL) $CM* $PR;
$CM* ($ALPlus | $HL) $CM* $PO;


# LB 25
($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;

# LB 26
$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
$CM* ($JT | $JV) $CM* ($H2 | $JV);
$CM* $JT $CM* ($H3 | $JT);

# LB 27
$CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
$CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
$CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;

# LB 28
$CM* ($ALPlus | $HL) $CM* ($ALPlus | $HL);


# LB 29
$CM* ($ALPlus | $HL) $CM* $IS;

# LB 30
$CM* $OP $CM* ($ALPlus | $HL | $NU);
$CM* ($ALPlus | $HL | $NU) $CM* $CP;

# LB 30a
$CM* $RI $CM* $RI;

# Special reverse rule for people & body part emoji:
# don't break between relevant emoji and $EmojiMods
$EmojiMods $EmojiVar? $EmojiForMods;

## -------------------------------------------------

!!safe_reverse;

# LB 9
$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
$CM+ $SP / .;

# LB 14
$SP+ $CM* $OP;

# LB 15
$SP+ $CM* $QU;

# LB 16
$SP+ $CM* ($CL | $CP);

# LB 17
$SP+ $CM* $B2;

# LB 21
$CM* ($HY | $BA) $CM* $HL;

# LB 25
($CM* ($IS | $SY))+ $CM* $NU;
($CL | $CP) $CM* ($NU | $IS | $SY);

# For dictionary-based break
$dictionary $dictionary;

## -------------------------------------------------

!!safe_forward;

# Skip forward over all character classes that are involved in
#   rules containing patterns with possibly more than one char
#   of context.
#
#  It might be slightly more efficient to have specific rules
#  instead of one generic one, but only if we could
#  turn off rule chaining.  We don't want to move more
#  than necessary.
#
[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
$dictionary $dictionary;
Commit	Line	Data
08b89b0a A	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 29 for Unicode 6.2
	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	# Special character classes for people & body part emoji:
110	# Subsets of $CM:
111	$ZWJ = \u200D;
112	$EmojiVar = \uFE0F;
113	# The following are subsets of $ID
114	$EmojiForSeqs = [\u2764 \U0001F466-\U0001F469 \U0001F48B];
115	$EmojiForMods = [\u261D \u270A-\u270C \U0001F385 \U0001F3C3-\U0001F3C4 \U0001F3C7 \U0001F3CA \U0001F442-\U0001F443 \U0001F446-\U0001F450 \U0001F466-\U0001F469 \U0001F46E-\U0001F478 \U0001F47C \U0001F481-\U0001F483 \U0001F485-\U0001F487 \U0001F4AA \U0001F596 \U0001F645-\U0001F647 \U0001F64B-\U0001F64F \U0001F6A3 \U0001F6B4-\U0001F6B6 \U0001F6C0];
116	$EmojiMods = [\U0001F3FB-\U0001F3FF];
117
118	# Dictionary character set, for triggering language-based break engines. Currently
119	# limited to LineBreak=Complex_Context. Note that this set only works in Unicode
120	# 5.0 or later as the definition of Complex_Context was corrected to include all
121	# characters requiring dictionary break.
122
123	$dictionary = [:LineBreak = Complex_Context:];
124
125	#
126	# Rule LB1. By default, treat AI (characters with ambiguous east Asian width),
127	# SA (South East Asian: Thai, Lao, Khmer)
128	# SG (Unpaired Surrogates)
129	# XX (Unknown, unassigned)
130	# as $AL (Alphabetic)
131	#
132	$ALPlus = [$AL $AI $SA $SG $XX];
133
134	#
135	# Combining Marks. X $CM* behaves as if it were X. Rule LB6.
136	#
137	$ALcm = $ALPlus $CM*;
138	$BAcm = $BA $CM*;
139	$BAXcm = $BAX $CM*;
140	$BBcm = $BB $CM*;
141	$B2cm = $B2 $CM*;
142	$CLcm = $CL $CM*;
143	$CPcm = $CP $CM*;
144	$EXcm = $EX $CM*;
145	$GLcm = $GL $CM*;
146	$HLcm = $HL $CM*;
147	$HYcm = $HY $CM*;
148	$H2cm = $H2 $CM*;
149	$H3cm = $H3 $CM*;
150	$IDcm = $ID $CM*;
151	$INcm = $IN $CM*;
152	$IScm = $IS $CM*;
153	$JLcm = $JL $CM*;
154	$JVcm = $JV $CM*;
155	$JTcm = $JT $CM*;
156	$NScm = $NS $CM*;
157	$NSXcm = $NSX $CM*;
158	$NUcm = $NU $CM*;
159	$OPcm = $OP $CM*;
160	$POcm = $PO $CM*;
161	$PRcm = $PR $CM*;
162	$QUcm = $QU $CM*;
163	$RIcm = $RI $CM*;
164	$SYcm = $SY $CM*;
165	$WJcm = $WJ $CM*;
166
167	## -------------------------------------------------
168
169	!!forward;
170
171	#
172	# Each class of character can stand by itself as an unbroken token, with trailing combining stuff
173	#
174	$ALPlus $CM+;
175	$BA $CM+;
176	$BAX $CM+;
177	$BB $CM+;
178	$B2 $CM+;
179	$CL $CM+;
180	$CP $CM+;
181	$EX $CM+;
182	$GL $CM+;
183	$HL $CM+;
184	$HY $CM+;
185	$H2 $CM+;
186	$H3 $CM+;
187	$ID $CM+;
188	$IN $CM+;
189	$IS $CM+;
190	$JL $CM+;
191	$JV $CM+;
192	$JT $CM+;
193	$NS $CM+;
194	$NSX $CM+;
195	$NU $CM+;
196	$OP $CM+;
197	$PO $CM+;
198	$PR $CM+;
199	$QU $CM+;
200	$RI $CM+;
201	$SY $CM+;
202	$WJ $CM+;
203
204	#
205	# CAN_CM is the set of characters that may combine with CM combining chars.
206	# Note that Linebreak UAX 14's concept of a combining char and the rules
207	# for what they can combine with are _very_ different from the rest of Unicode.
208	#
209	# Note that $CM itself is left out of this set. If CM is needed as a base
210	# it must be listed separately in the rule.
211	#
212	$CAN_CM = [^$SP $BK $CR $LF $NL $ZW $CM]; # Bases that can take CMs
213	$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM]; # Bases that can't take CMs
214
215	#
216	# AL_FOLLOW set of chars that can unconditionally follow an AL
217	# Needed in rules where stand-alone $CM s are treated as AL.
218	# Chaining is disabled with CM because it causes other failures,
219	# so for this one case we need to manually list out longer sequences.
220	#
221	$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
222	$AL_FOLLOW_CM = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $BAX $HY $NS $NSX $IN $NU $ALPlus];
223	$AL_FOLLOW = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
224
225
226	#
227	# Rule LB 4, 5 Mandatory (Hard) breaks.
228	#
229	$LB4Breaks = [$BK $CR $LF $NL];
230	$LB4NonBreaks = [^$BK $CR $LF $NL];
231	$CR $LF {100};
232
233	#
234	# LB 6 Do not break before hard line breaks.
235	#
236	$LB4NonBreaks? $LB4Breaks {100}; # LB 5 do not break before hard breaks.
237	$CAN_CM $CM* $LB4Breaks {100};
238	$CM+ $LB4Breaks {100};
239
240	# LB 7 x SP
241	# x ZW
242	$LB4NonBreaks [$SP $ZW];
243	$CAN_CM $CM* [$SP $ZW];
244	$CM+ [$SP $ZW];
245
246	#
247	# LB 8 Break after zero width space
248	# TODO: ZW SP* <break>
249	# An engine change is required to write the reverse rule for this.
250	# For now, leave the Unicode 5.2 rule, ZW <break>
251	#
252	$LB8Breaks = [$LB4Breaks $ZW];
253	$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
254
255
256	# Special forward rule for people & body part emoji:
257	# don't break $ZWJ from subsequent $EmojiForSeqs
258	$EmojiForSeqs $EmojiVar? $EmojiMods? $ZWJ $EmojiForSeqs;
259
260	# LB 9 Combining marks. X $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
261	# $CM not covered by the above needs to behave like $AL
262	# See definition of $CAN_CM.
263
264	$CAN_CM $CM+; # Stick together any combining sequences that don't match other rules.
265	$CM+;
266
267	#
268	# LB 11 Do not break before or after WORD JOINER & related characters.
269	#
270	$CAN_CM $CM* $WJcm;
271	$LB8NonBreaks $WJcm;
272	$CM+ $WJcm;
273
274	$WJcm $CANT_CM;
275	$WJcm $CAN_CM $CM*;
276
277	#
278	# LB 12 Do not break after NBSP and related characters.
279	# GL x
280	#
281	$GLcm $CAN_CM $CM*;
282	$GLcm $CANT_CM;
283
284	#
285	# LB 12a Do not break before NBSP and related characters ...
286	# [^SP BA HY] x GL
287	#
288	[[$LB8NonBreaks] - [$SP $BA $BAX $HY]] $CM* $GLcm;
289	$CM+ GLcm;
290
291
292
293	#
294	# LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces.
295	#
296	$LB8NonBreaks $CL;
297	$CAN_CM $CM* $CL;
298	$CM+ $CL; # by rule 10, stand-alone CM behaves as AL
299
300	$LB8NonBreaks $CP;
301	$CAN_CM $CM* $CP;
302	$CM+ $CP; # by rule 10, stand-alone CM behaves as AL
303
304	$LB8NonBreaks $EX;
305	$CAN_CM $CM* $EX;
306	$CM+ $EX; # by rule 10, stand-alone CM behaves as AL
307
308	$LB8NonBreaks $IS;
309	$CAN_CM $CM* $IS;
310	$CM+ $IS; # by rule 10, stand-alone CM behaves as AL
311
312	$LB8NonBreaks $SY;
313	$CAN_CM $CM* $SY;
314	$CM+ $SY; # by rule 10, stand-alone CM behaves as AL
315
316
317	#
318	# LB 14 Do not break after OP, even after spaces
319	#
320	$OPcm $SP* $CAN_CM $CM*;
321	$OPcm $SP* $CANT_CM;
322
323	$OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL
324
325	# LB 15
326	$QUcm $SP* $OPcm;
327
328	# LB 16
329	# Do not break between closing punctuation and $NS, even with intervening spaces
330	# But DO allow a break between closing punctuation and $NSX, don't include it here
331	($CLcm \| $CPcm) $SP* $NScm;
332
333	# LB 17
334	$B2cm $SP* $B2cm;
335
336	#
337	# LB 18 Break after spaces.
338	#
339	$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
340	$LB18Breaks = [$LB8Breaks $SP];
341
342
343	# LB 19
344	# x QU
345	$LB18NonBreaks $CM* $QUcm;
346	$CM+ $QUcm;
347
348	# QU x
349	$QUcm .?;
350	$QUcm $LB18NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc.
351	# TODO: I don't think this rule is needed.
352
353
354	# LB 20
355	# <break> $CB
356	# $CB <break>
357
358	$LB20NonBreaks = [$LB18NonBreaks - $CB];
359
360	# LB 21 x (BA \| HY \| NS)
361	# BB x
362	#
363	# DO allow breaks here before $BAXcm and $NSXcm, so don't include them
364	$LB20NonBreaks $CM* ($BAcm \| $HYcm \| $NScm);
365
366	$BBcm [^$CB]; # $BB x
367	$BBcm $LB20NonBreaks $CM*;
368
369	# LB 21a Don't break after Hebrew + Hyphen
370	# HL (HY \| BA) x
371	#
372	$HLcm ($HYcm \| $BAcm \| $BAXcm) [^$CB]?;
373
374	# LB 21b (forward) Don't break between SY and HL
375	# (break between HL and SY already disallowed by LB 13 above)
376	$SYcm $HLcm;
377
378	# LB 22
379	($ALcm \| $HLcm) $INcm;
380	$CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL
381	$IDcm $INcm;
382	$INcm $INcm;
383	$NUcm $INcm;
384
385
386	# $LB 23
387	$IDcm $POcm;
388	$ALcm $NUcm; # includes $LB19
389	$HLcm $NUcm;
390	$CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL
391	$NUcm $ALcm;
392	$NUcm $HLcm;
393
394	#
395	# LB 24
396	#
397	$PRcm $IDcm;
398	$PRcm ($ALcm \| $HLcm);
399	$POcm ($ALcm \| $HLcm);
400
401	#
402	# LB 25 Numbers.
403	#
404	($PRcm \| $POcm)? ($OPcm \| $HYcm)? $NUcm ($NUcm \| $SYcm \| $IScm)* ($CLcm \| $CPcm)? ($PRcm \| $POcm)?;
405
406	# LB 26 Do not break a Korean syllable
407	#
408	$JLcm ($JLcm \| $JVcm \| $H2cm \| $H3cm);
409	($JVcm \| $H2cm) ($JVcm \| $JTcm);
410	($JTcm \| $H3cm) $JTcm;
411
412	# LB 27 Treat korean Syllable Block the same as ID (don't break it)
413	($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm) $INcm;
414	($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm) $POcm;
415	$PRcm ($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm);
416
417
418	# LB 28 Do not break between alphabetics
419	#
420	($ALcm \| $HLcm) ($ALcm \| $HLcm);
421	$CM+ ($ALcm \| $HLcm); # The $CM+ is from rule 10, an unattached CM is treated as AL
422
423	# LB 29
424	$IScm ($ALcm \| $HLcm);
425
426	# LB 30
427	($ALcm \| $HLcm \| $NUcm) $OPcm;
428	$CM+ $OPcm; # The $CM+ is from rule 10, an unattached CM is treated as AL.
429	$CPcm ($ALcm \| $HLcm \| $NUcm);
430
431	# LB 30a Do not break between regional indicators.
432	$RIcm $RIcm;
433
434	# Special forward rule for people & body part emoji:
435	# don't break between relevant emoji and $EmojiMods
436	$EmojiForMods $EmojiVar? $EmojiMods;
437
438	#
439	# Reverse Rules.
440	#
441	## -------------------------------------------------
442
443	!!reverse;
444
445	$CM+ $ALPlus;
446	$CM+ $BA;
447	$CM+ $BAX;
448	$CM+ $BB;
449	$CM+ $B2;
450	$CM+ $CL;
451	$CM+ $CP;
452	$CM+ $EX;
453	$CM+ $GL;
454	$CM+ $HL;
455	$CM+ $HY;
456	$CM+ $H2;
457	$CM+ $H3;
458	$CM+ $ID;
459	$CM+ $IN;
460	$CM+ $IS;
461	$CM+ $JL;
462	$CM+ $JV;
463	$CM+ $JT;
464	$CM+ $NS;
465	$CM+ $NSX;
466	$CM+ $NU;
467	$CM+ $OP;
468	$CM+ $PO;
469	$CM+ $PR;
470	$CM+ $QU;
471	$CM+ $RI;
472	$CM+ $SY;
473	$CM+ $WJ;
474	$CM+;
475
476
477	#
478	# Sequences of the form (shown forwards)
479	# [CANT_CM] <break> [CM] [whatever]
480	# The CM needs to behave as an AL
481	#
482	$AL_FOLLOW $CM+ / (
483	[$BK $CR $LF $NL $ZW {eof}] \|
484	$SP+ $CM+ $SP \|
485	$SP+ $CM* ([^$OP $CM $SP] \| [$AL {eof}])); # if LB 14 will match, need to surpress this break.
486	# LB14 says OP SP* x .
487	# becomes OP SP* x AL
488	# becomes OP SP* x CM+ AL_FOLLOW
489	#
490	# Further note: the $AL in [$AL {eof}] is only to work around
491	# a rule compiler bug which complains about
492	# empty sets otherwise.
493
494	#
495	# Sequences of the form (shown forwards)
496	# [CANT_CM] <break> [CM] <break> [PR]
497	# The CM needs to behave as an AL
498	# This rule is concerned about getting the second of the two <breaks> in place.
499	#
500
501	[$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
502
503
504
505	# LB 4, 5, 5
506
507	$LB4Breaks [$LB4NonBreaks-$CM];
508	$LB4Breaks $CM+ $CAN_CM;
509	$LF $CR;
510
511
512	# LB 7 x SP
513	# x ZW
514	[$SP $ZW] [$LB4NonBreaks-$CM];
515	[$SP $ZW] $CM+ $CAN_CM;
516
517	# LB 8 ZW SP* <break>
518	# TODO: to implement this, we need more than one look-ahead hard break in play at a time.
519	# Requires an engine enhancement.
520	# / $SP* $ZW
521
522	# Special reverse rule for people & body part emoji:
523	# don't break $ZWJ from subsequent $EmojiForSeqs
524	$EmojiForSeqs $ZWJ $EmojiMods? $EmojiVar? $EmojiForSeqs;
525
526	# LB 9,10 Combining marks.
527	# X $CM needs to behave like X, where X is not $SP or controls.
528	# $CM not covered by the above needs to behave like $AL
529	# Stick together any combining sequences that don't match other rules.
530	$CM+ $CAN_CM;
531
532
533	# LB 11
534	$CM* $WJ $CM* $CAN_CM;
535	$CM* $WJ [$LB8NonBreaks-$CM];
536
537	$CANT_CM $CM* $WJ;
538	$CM* $CAN_CM $CM* $WJ;
539
540	# LB 12a
541	# [^SP BA HY] x GL
542	#
543	$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $BAX $HY]];
544
545	# LB 12
546	# GL x
547	#
548	$CANT_CM $CM* $GL;
549	$CM* $CAN_CM $CM* $GL;
550
551
552	# LB 13
553	$CL $CM+ $CAN_CM;
554	$CP $CM+ $CAN_CM;
555	$EX $CM+ $CAN_CM;
556	$IS $CM+ $CAN_CM;
557	$SY $CM+ $CAN_CM;
558
559	$CL [$LB8NonBreaks-$CM];
560	$CP [$LB8NonBreaks-$CM];
561	$EX [$LB8NonBreaks-$CM];
562	$IS [$LB8NonBreaks-$CM];
563	$SY [$LB8NonBreaks-$CM];
564
565	# Rule 13 & 14 taken together for an edge case.
566	# Match this, shown forward
567	# OP SP+ ($CM+ behaving as $AL) (CL \| CP \| EX \| IS \| IY)
568	# This really wants to chain at the $CM+ (which is acting as an $AL)
569	# except for $CM chaining being disabled.
570	[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
571
572	# LB 14 OP SP* x
573	#
574	$CM* $CAN_CM $SP* $CM* $OP;
575	$CANT_CM $SP* $CM* $OP;
576	$AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
577
578	$AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
579	$CM* $AL_FOLLOW_CM $CM+ $SP+ $CM* $OP;
580	$SY $CM $SP+ $OP; # TODO: Experiment. Remove.
581
582
583
584	# LB 15
585	$CM* $OP $SP* $CM* $QU;
586
587	# LB 16
588	# Don't include $NSX here
589	$CM* $NS $SP* $CM* ($CL \| $CP);
590
591	# LB 17
592	$CM* $B2 $SP* $CM* $B2;
593
594	# LB 18 break after spaces
595	# Nothing explicit needed here.
596
597
598	#
599	# LB 19
600	#
601	$CM* $QU $CM* $CAN_CM; # . x QU
602	$CM* $QU $LB18NonBreaks;
603
604
605	$CM* $CAN_CM $CM* $QU; # QU x .
606	$CANT_CM $CM* $QU;
607
608	#
609	# LB 20 Break before and after CB.
610	# nothing needed here.
611	#
612
613	# LB 21
614	# Don't include $BAX or $NSX here
615	$CM* ($BA \| $HY \| $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA \| HY \| NS)
616
617	$CM* [$LB20NonBreaks-$CM] $CM* $BB; # BB x .
618	[^$CB] $CM* $BB; #
619
620	# LB21a
621	[^$CB] $CM* ($HY \| $BA \| $BAX) $CM* $HL;
622
623	# LB21b (reverse)
624	$CM* $HL $CM* $SY;
625
626	# LB 22
627	$CM* $IN $CM* ($ALPlus \| $HL);
628	$CM* $IN $CM* $ID;
629	$CM* $IN $CM* $IN;
630	$CM* $IN $CM* $NU;
631
632	# LB 23
633	$CM* $PO $CM* $ID;
634	$CM* $NU $CM* ($ALPlus \| $HL);
635	$CM* ($ALPlus \| $HL) $CM* $NU;
636
637	# LB 24
638	$CM* $ID $CM* $PR;
639	$CM* ($ALPlus \| $HL) $CM* $PR;
640	$CM* ($ALPlus \| $HL) $CM* $PO;
641
642
643	# LB 25
644	($CM* ($PR \| $PO))? ($CM* ($CL \| $CP))? ($CM* ($NU \| $IS \| $SY))* $CM* $NU ($CM* ($OP \| $HY))? ($CM* ($PR \| $PO))?;
645
646	# LB 26
647	$CM* ($H3 \| $H2 \| $JV \| $JL) $CM* $JL;
648	$CM* ($JT \| $JV) $CM* ($H2 \| $JV);
649	$CM* $JT $CM* ($H3 \| $JT);
650
651	# LB 27
652	$CM* $IN $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);
653	$CM* $PO $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);
654	$CM* ($H3 \| $H2 \| $JT \| $JV \| $JL) $CM* $PR;
655
656	# LB 28
657	$CM* ($ALPlus \| $HL) $CM* ($ALPlus \| $HL);
658
659
660	# LB 29
661	$CM* ($ALPlus \| $HL) $CM* $IS;
662
663	# LB 30
664	$CM* $OP $CM* ($ALPlus \| $HL \| $NU);
665	$CM* ($ALPlus \| $HL \| $NU) $CM* $CP;
666
667	# LB 30a
668	$CM* $RI $CM* $RI;
669
670	# Special reverse rule for people & body part emoji:
671	# don't break between relevant emoji and $EmojiMods
672	$EmojiMods $EmojiVar? $EmojiForMods;
673
674	## -------------------------------------------------
675
676	!!safe_reverse;
677
678	# LB 9
679	$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
680	$CM+ $SP / .;
681
682	# LB 14
683	$SP+ $CM* $OP;
684
685	# LB 15
686	$SP+ $CM* $QU;
687
688	# LB 16
689	$SP+ $CM* ($CL \| $CP);
690
691	# LB 17
692	$SP+ $CM* $B2;
693
694	# LB 21
695	$CM* ($HY \| $BA) $CM* $HL;
696
697	# LB 25
698	($CM* ($IS \| $SY))+ $CM* $NU;
699	($CL \| $CP) $CM* ($NU \| $IS \| $SY);
700
701	# For dictionary-based break
702	$dictionary $dictionary;
703
704	## -------------------------------------------------
705
706	!!safe_forward;
707
708	# Skip forward over all character classes that are involved in
709	# rules containing patterns with possibly more than one char
710	# of context.
711	#
712	# It might be slightly more efficient to have specific rules
713	# instead of one generic one, but only if we could
714	# turn off rule chaining. We don't want to move more
715	# than necessary.
716	#
717	[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
718	$dictionary $dictionary;
719