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

# Copyright (c) 2002-2015  International Business Machines Corporation and
# others. All Rights Reserved.
#
#  file:  line.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/
#
#         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 corresponds to CSS line-break=strict (BCP47 -u-lb-strict).
#         It sets characters of class CJ to behave like 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:];
$BA = [:LineBreak =  Break_After:];
$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:];
$IN = [:LineBreak =  Inseperable:];
$IS = [:LineBreak =  Infix_Numeric:];
$JL = [:LineBreak =  JL:];
$JV = [:LineBreak =  JV:];
$JT = [:LineBreak =  JT:];
$LF = [:LineBreak =  Line_Feed:];
$NL = [:LineBreak =  Next_Line:];
$NS = [[:LineBreak =  Nonstarter:] $CJ];
$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 \U0001F441 \U0001F466-\U0001F469 \U0001F48B \U0001F5E8];
$EmojiForMods = [\u261D \u26F9 \u270A-\u270D \U0001F385 \U0001F3C3-\U0001F3C4 \U0001F3CA-\U0001F3CB \U0001F442-\U0001F443 \U0001F446-\U0001F450 \U0001F466-\U0001F469 \U0001F46E-\U0001F478 \U0001F47C \U0001F481-\U0001F483 \U0001F485-\U0001F487 \U0001F4AA \U0001F575 \U0001F590 \U0001F595 \U0001F596 \U0001F645-\U0001F647 \U0001F64B-\U0001F64F \U0001F6A3 \U0001F6B4-\U0001F6B6 \U0001F6C0 \U0001F918];
$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*;
$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*;
$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+;
$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+;
$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 $HY $NS $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 $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
($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
#
$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 | $BAcm) [^$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+ $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+ $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 $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
$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
$CM* ($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) $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	1	# Copyright (c) 2002-2015 International Business Machines Corporation and
46f4442e A	2	# others. All Rights Reserved.
	3	#
	4	# file: line.txt
	5	#
	6	# Line Breaking Rules
	7	# Implement default line breaking as defined by
51004dcb	8	# Unicode Standard Annex #14 Revision 29 for Unicode 6.2
46f4442e	9	# http://www.unicode.org/reports/tr14/
729e4ab9 A	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.
08b89b0a A	14	#
	15	# This corresponds to CSS line-break=strict (BCP47 -u-lb-strict).
	16	# It sets characters of class CJ to behave like NS.
46f4442e A	17
	18	#
	19	# Character Classes defined by TR 14.
	20	#
	21
	22	!!chain;
	23	!!LBCMNoChain;
	24
	25
	26	!!lookAheadHardBreak;
	27	#
	28	# !!lookAheadHardBreak Described here because it is (as yet) undocumented elsewhere
	29	# and only used for the line break rules.
	30	#
729e4ab9	31	# It is used in the implementation of rule LB 10
46f4442e A	32	# which says to treat any combining mark that is not attached to a base
	33	# character as if it were of class AL (alphabetic).
	34	#
	35	# The problem occurs in the reverse rules.
	36	#
	37	# Consider a sequence like, with correct breaks as shown
	38	# LF ID CM AL AL
	39	# ^ ^ ^
	40	# Then consider the sequence without the initial ID (ideographic)
	41	# LF CM AL AL
	42	# ^ ^
	43	# Our CM, which in the first example was attached to the ideograph,
	44	# is now unattached, becomes an alpha, and joins in with the other
	45	# alphas.
	46	#
	47	# When iterating forwards, these sequences do not present any problems
	48	# When iterating backwards, we need to look ahead when encountering
	49	# a CM to see whether it attaches to something further on or not.
	50	# (Look-ahead in a reverse rule is looking towards the start)
	51	#
	52	# If the CM is unattached, we need to force a break.
	53	#
	54	# !!lookAheadHardBreak forces the run time state machine to
	55	# stop immediately when a look ahead rule ( '/' operator) matches,
	56	# and set the match position to that of the look-ahead operator,
	57	# no matter what other rules may be in play at the time.
	58	#
	59	# See rule LB 19 for an example.
	60	#
	61
	62	$AI = [:LineBreak = Ambiguous:];
	63	$AL = [:LineBreak = Alphabetic:];
	64	$BA = [:LineBreak = Break_After:];
	65	$BB = [:LineBreak = Break_Before:];
	66	$BK = [:LineBreak = Mandatory_Break:];
	67	$B2 = [:LineBreak = Break_Both:];
	68	$CB = [:LineBreak = Contingent_Break:];
4388f060	69	$CJ = [:LineBreak = Conditional_Japanese_Starter:];
46f4442e A	70	$CL = [:LineBreak = Close_Punctuation:];
46f4442e A	71	$CM = [:LineBreak = Combining_Mark:];
729e4ab9	72	$CP = [:LineBreak = Close_Parenthesis:];
46f4442e A	73	$CR = [:LineBreak = Carriage_Return:];
	74	$EX = [:LineBreak = Exclamation:];
	75	$GL = [:LineBreak = Glue:];
4388f060	76	$HL = [:LineBreak = Hebrew_Letter:];
46f4442e A	77	$HY = [:LineBreak = Hyphen:];
	78	$H2 = [:LineBreak = H2:];
	79	$H3 = [:LineBreak = H3:];
	80	$ID = [:LineBreak = Ideographic:];
	81	$IN = [:LineBreak = Inseperable:];
	82	$IS = [:LineBreak = Infix_Numeric:];
	83	$JL = [:LineBreak = JL:];
	84	$JV = [:LineBreak = JV:];
	85	$JT = [:LineBreak = JT:];
	86	$LF = [:LineBreak = Line_Feed:];
	87	$NL = [:LineBreak = Next_Line:];
4388f060	88	$NS = [[:LineBreak = Nonstarter:] $CJ];
46f4442e A	89	$NU = [:LineBreak = Numeric:];
	90	$OP = [:LineBreak = Open_Punctuation:];
	91	$PO = [:LineBreak = Postfix_Numeric:];
	92	$PR = [:LineBreak = Prefix_Numeric:];
	93	$QU = [:LineBreak = Quotation:];
51004dcb	94	$RI = [:LineBreak = Regional_Indicator:];
46f4442e A	95	$SA = [:LineBreak = Complex_Context:];
	96	$SG = [:LineBreak = Surrogate:];
	97	$SP = [:LineBreak = Space:];
	98	$SY = [:LineBreak = Break_Symbols:];
	99	$WJ = [:LineBreak = Word_Joiner:];
	100	$XX = [:LineBreak = Unknown:];
	101	$ZW = [:LineBreak = ZWSpace:];
	102
08b89b0a A	103	# Special character classes for people & body part emoji:
	104	# Subsets of $CM:
	105	$ZWJ = \u200D;
	106	$EmojiVar = \uFE0F;
	107	# The following are subsets of $ID
b801cf36	108	$EmojiForSeqs = [\u2764 \U0001F441 \U0001F466-\U0001F469 \U0001F48B \U0001F5E8];
a961784b	109	$EmojiForMods = [\u261D \u26F9 \u270A-\u270D \U0001F385 \U0001F3C3-\U0001F3C4 \U0001F3CA-\U0001F3CB \U0001F442-\U0001F443 \U0001F446-\U0001F450 \U0001F466-\U0001F469 \U0001F46E-\U0001F478 \U0001F47C \U0001F481-\U0001F483 \U0001F485-\U0001F487 \U0001F4AA \U0001F575 \U0001F590 \U0001F595 \U0001F596 \U0001F645-\U0001F647 \U0001F64B-\U0001F64F \U0001F6A3 \U0001F6B4-\U0001F6B6 \U0001F6C0 \U0001F918];
08b89b0a A	110	$EmojiMods = [\U0001F3FB-\U0001F3FF];
08b89b0a A	111
46f4442e A	112	# Dictionary character set, for triggering language-based break engines. Currently
	113	# limited to LineBreak=Complex_Context. Note that this set only works in Unicode
	114	# 5.0 or later as the definition of Complex_Context was corrected to include all
	115	# characters requiring dictionary break.
	116
	117	$dictionary = [:LineBreak = Complex_Context:];
	118
	119	#
	120	# Rule LB1. By default, treat AI (characters with ambiguous east Asian width),
	121	# SA (South East Asian: Thai, Lao, Khmer)
	122	# SG (Unpaired Surrogates)
	123	# XX (Unknown, unassigned)
	124	# as $AL (Alphabetic)
	125	#
	126	$ALPlus = [$AL $AI $SA $SG $XX];
	127
	128	#
	129	# Combining Marks. X $CM* behaves as if it were X. Rule LB6.
	130	#
	131	$ALcm = $ALPlus $CM*;
	132	$BAcm = $BA $CM*;
	133	$BBcm = $BB $CM*;
	134	$B2cm = $B2 $CM*;
	135	$CLcm = $CL $CM*;
729e4ab9	136	$CPcm = $CP $CM*;
46f4442e A	137	$EXcm = $EX $CM*;
46f4442e A	138	$GLcm = $GL $CM*;
4388f060	139	$HLcm = $HL $CM*;
46f4442e A	140	$HYcm = $HY $CM*;
	141	$H2cm = $H2 $CM*;
	142	$H3cm = $H3 $CM*;
	143	$IDcm = $ID $CM*;
	144	$INcm = $IN $CM*;
	145	$IScm = $IS $CM*;
	146	$JLcm = $JL $CM*;
	147	$JVcm = $JV $CM*;
	148	$JTcm = $JT $CM*;
	149	$NScm = $NS $CM*;
	150	$NUcm = $NU $CM*;
	151	$OPcm = $OP $CM*;
	152	$POcm = $PO $CM*;
	153	$PRcm = $PR $CM*;
	154	$QUcm = $QU $CM*;
51004dcb	155	$RIcm = $RI $CM*;
46f4442e A	156	$SYcm = $SY $CM*;
	157	$WJcm = $WJ $CM*;
	158
	159	## -------------------------------------------------
	160
	161	!!forward;
	162
	163	#
	164	# Each class of character can stand by itself as an unbroken token, with trailing combining stuff
	165	#
	166	$ALPlus $CM+;
	167	$BA $CM+;
	168	$BB $CM+;
	169	$B2 $CM+;
	170	$CL $CM+;
729e4ab9	171	$CP $CM+;
46f4442e A	172	$EX $CM+;
46f4442e A	173	$GL $CM+;
4388f060	174	$HL $CM+;
46f4442e A	175	$HY $CM+;
	176	$H2 $CM+;
	177	$H3 $CM+;
	178	$ID $CM+;
	179	$IN $CM+;
	180	$IS $CM+;
	181	$JL $CM+;
	182	$JV $CM+;
	183	$JT $CM+;
	184	$NS $CM+;
	185	$NU $CM+;
	186	$OP $CM+;
	187	$PO $CM+;
	188	$PR $CM+;
	189	$QU $CM+;
51004dcb	190	$RI $CM+;
46f4442e A	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];
4388f060	212	$AL_FOLLOW_CM = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HY $NS $IN $NU $ALPlus];
46f4442e A	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
729e4ab9 A	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>
46f4442e A	241	#
	242	$LB8Breaks = [$LB4Breaks $ZW];
	243	$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
	244
	245
08b89b0a A	246	# Special forward rule for people & body part emoji:
	247	# don't break $ZWJ from subsequent $EmojiForSeqs
	248	$EmojiForSeqs $EmojiVar? $EmojiMods? $ZWJ $EmojiForSeqs;
	249
46f4442e A	250	# LB 9 Combining marks. X $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
	251	# $CM not covered by the above needs to behave like $AL
	252	# See definition of $CAN_CM.
	253
	254	$CAN_CM $CM+; # Stick together any combining sequences that don't match other rules.
	255	$CM+;
	256
	257	#
	258	# LB 11 Do not break before or after WORD JOINER & related characters.
	259	#
	260	$CAN_CM $CM* $WJcm;
	261	$LB8NonBreaks $WJcm;
	262	$CM+ $WJcm;
	263
	264	$WJcm $CANT_CM;
	265	$WJcm $CAN_CM $CM*;
	266
	267	#
	268	# LB 12 Do not break after NBSP and related characters.
	269	# GL x
	270	#
	271	$GLcm $CAN_CM $CM*;
	272	$GLcm $CANT_CM;
	273
	274	#
	275	# LB 12a Do not break before NBSP and related characters ...
	276	# [^SP BA HY] x GL
	277	#
	278	[[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
	279	$CM+ GLcm;
	280
	281
	282
	283	#
	284	# LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces.
	285	#
	286	$LB8NonBreaks $CL;
	287	$CAN_CM $CM* $CL;
	288	$CM+ $CL; # by rule 10, stand-alone CM behaves as AL
	289
729e4ab9 A	290	$LB8NonBreaks $CP;
	291	$CAN_CM $CM* $CP;
	292	$CM+ $CP; # by rule 10, stand-alone CM behaves as AL
	293
46f4442e A	294	$LB8NonBreaks $EX;
	295	$CAN_CM $CM* $EX;
	296	$CM+ $EX; # by rule 10, stand-alone CM behaves as AL
	297
	298	$LB8NonBreaks $IS;
	299	$CAN_CM $CM* $IS;
	300	$CM+ $IS; # by rule 10, stand-alone CM behaves as AL
	301
	302	$LB8NonBreaks $SY;
	303	$CAN_CM $CM* $SY;
	304	$CM+ $SY; # by rule 10, stand-alone CM behaves as AL
	305
	306
	307	#
	308	# LB 14 Do not break after OP, even after spaces
	309	#
	310	$OPcm $SP* $CAN_CM $CM*;
	311	$OPcm $SP* $CANT_CM;
	312
	313	$OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL
	314
	315	# LB 15
	316	$QUcm $SP* $OPcm;
	317
	318	# LB 16
729e4ab9	319	($CLcm \| $CPcm) $SP* $NScm;
46f4442e A	320
	321	# LB 17
	322	$B2cm $SP* $B2cm;
	323
	324	#
	325	# LB 18 Break after spaces.
	326	#
	327	$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
	328	$LB18Breaks = [$LB8Breaks $SP];
	329
	330
	331	# LB 19
	332	# x QU
	333	$LB18NonBreaks $CM* $QUcm;
	334	$CM+ $QUcm;
	335
	336	# QU x
	337	$QUcm .?;
	338	$QUcm $LB18NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc.
	339	# TODO: I don't think this rule is needed.
	340
	341
	342	# LB 20
	343	# <break> $CB
	344	# $CB <break>
	345
	346	$LB20NonBreaks = [$LB18NonBreaks - $CB];
	347
	348	# LB 21 x (BA \| HY \| NS)
	349	# BB x
	350	#
	351	$LB20NonBreaks $CM* ($BAcm \| $HYcm \| $NScm);
	352
	353	$BBcm [^$CB]; # $BB x
	354	$BBcm $LB20NonBreaks $CM*;
	355
4388f060 A	356	# LB 21a Don't break after Hebrew + Hyphen
	357	# HL (HY \| BA) x
	358	#
	359	$HLcm ($HYcm \| $BAcm) [^$CB]?;
	360
51004dcb A	361	# LB 21b (forward) Don't break between SY and HL
	362	# (break between HL and SY already disallowed by LB 13 above)
	363	$SYcm $HLcm;
	364
46f4442e	365	# LB 22
4388f060	366	($ALcm \| $HLcm) $INcm;
46f4442e A	367	$CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL
	368	$IDcm $INcm;
	369	$INcm $INcm;
	370	$NUcm $INcm;
	371
	372
	373	# $LB 23
	374	$IDcm $POcm;
	375	$ALcm $NUcm; # includes $LB19
4388f060	376	$HLcm $NUcm;
46f4442e A	377	$CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL
46f4442e A	378	$NUcm $ALcm;
4388f060	379	$NUcm $HLcm;
46f4442e A	380
	381	#
	382	# LB 24
	383	#
	384	$PRcm $IDcm;
4388f060 A	385	$PRcm ($ALcm \| $HLcm);
4388f060 A	386	$POcm ($ALcm \| $HLcm);
46f4442e A	387
	388	#
	389	# LB 25 Numbers.
	390	#
729e4ab9	391	($PRcm \| $POcm)? ($OPcm \| $HYcm)? $NUcm ($NUcm \| $SYcm \| $IScm)* ($CLcm \| $CPcm)? ($PRcm \| $POcm)?;
46f4442e A	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 \| $H3cm);
	403
	404
	405	# LB 28 Do not break between alphabetics
	406	#
4388f060 A	407	($ALcm \| $HLcm) ($ALcm \| $HLcm);
4388f060 A	408	$CM+ ($ALcm \| $HLcm); # The $CM+ is from rule 10, an unattached CM is treated as AL
46f4442e A	409
46f4442e A	410	# LB 29
4388f060	411	$IScm ($ALcm \| $HLcm);
46f4442e	412
729e4ab9	413	# LB 30
4388f060	414	($ALcm \| $HLcm \| $NUcm) $OPcm;
729e4ab9	415	$CM+ $OPcm; # The $CM+ is from rule 10, an unattached CM is treated as AL.
4388f060	416	$CPcm ($ALcm \| $HLcm \| $NUcm);
729e4ab9	417
51004dcb A	418	# LB 30a Do not break between regional indicators.
51004dcb A	419	$RIcm $RIcm;
46f4442e	420
08b89b0a A	421	# Special forward rule for people & body part emoji:
	422	# don't break between relevant emoji and $EmojiMods
	423	$EmojiForMods $EmojiVar? $EmojiMods;
	424
46f4442e A	425	#
	426	# Reverse Rules.
	427	#
	428	## -------------------------------------------------
	429
	430	!!reverse;
	431
	432	$CM+ $ALPlus;
	433	$CM+ $BA;
	434	$CM+ $BB;
	435	$CM+ $B2;
	436	$CM+ $CL;
729e4ab9	437	$CM+ $CP;
46f4442e A	438	$CM+ $EX;
46f4442e A	439	$CM+ $GL;
4388f060	440	$CM+ $HL;
46f4442e A	441	$CM+ $HY;
	442	$CM+ $H2;
	443	$CM+ $H3;
	444	$CM+ $ID;
	445	$CM+ $IN;
	446	$CM+ $IS;
	447	$CM+ $JL;
	448	$CM+ $JV;
	449	$CM+ $JT;
	450	$CM+ $NS;
	451	$CM+ $NU;
	452	$CM+ $OP;
	453	$CM+ $PO;
	454	$CM+ $PR;
	455	$CM+ $QU;
51004dcb	456	$CM+ $RI;
46f4442e A	457	$CM+ $SY;
	458	$CM+ $WJ;
	459	$CM+;
	460
	461
	462	#
	463	# Sequences of the form (shown forwards)
	464	# [CANT_CM] <break> [CM] [whatever]
	465	# The CM needs to behave as an AL
	466	#
	467	$AL_FOLLOW $CM+ / (
	468	[$BK $CR $LF $NL $ZW {eof}] \|
	469	$SP+ $CM+ $SP \|
	470	$SP+ $CM* ([^$OP $CM $SP] \| [$AL {eof}])); # if LB 14 will match, need to surpress this break.
	471	# LB14 says OP SP* x .
	472	# becomes OP SP* x AL
	473	# becomes OP SP* x CM+ AL_FOLLOW
	474	#
	475	# Further note: the $AL in [$AL {eof}] is only to work around
	476	# a rule compiler bug which complains about
	477	# empty sets otherwise.
	478
	479	#
	480	# Sequences of the form (shown forwards)
	481	# [CANT_CM] <break> [CM] <break> [PR]
	482	# The CM needs to behave as an AL
	483	# This rule is concerned about getting the second of the two <breaks> in place.
	484	#
	485
	486	[$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
	487
	488
	489
	490	# LB 4, 5, 5
	491
	492	$LB4Breaks [$LB4NonBreaks-$CM];
	493	$LB4Breaks $CM+ $CAN_CM;
	494	$LF $CR;
	495
	496
	497	# LB 7 x SP
	498	# x ZW
	499	[$SP $ZW] [$LB4NonBreaks-$CM];
	500	[$SP $ZW] $CM+ $CAN_CM;
	501
729e4ab9 A	502	# LB 8 ZW SP* <break>
	503	# TODO: to implement this, we need more than one look-ahead hard break in play at a time.
	504	# Requires an engine enhancement.
	505	# / $SP* $ZW
46f4442e	506
08b89b0a A	507	# Special reverse rule for people & body part emoji:
	508	# don't break $ZWJ from subsequent $EmojiForSeqs
	509	$EmojiForSeqs $ZWJ $EmojiMods? $EmojiVar? $EmojiForSeqs;
	510
46f4442e A	511	# LB 9,10 Combining marks.
	512	# X $CM needs to behave like X, where X is not $SP or controls.
	513	# $CM not covered by the above needs to behave like $AL
	514	# Stick together any combining sequences that don't match other rules.
	515	$CM+ $CAN_CM;
	516
	517
	518	# LB 11
	519	$CM* $WJ $CM* $CAN_CM;
	520	$CM* $WJ [$LB8NonBreaks-$CM];
	521
	522	$CANT_CM $CM* $WJ;
	523	$CM* $CAN_CM $CM* $WJ;
	524
	525	# LB 12a
	526	# [^SP BA HY] x GL
	527	#
	528	$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
	529
	530	# LB 12
	531	# GL x
	532	#
	533	$CANT_CM $CM* $GL;
	534	$CM* $CAN_CM $CM* $GL;
	535
	536
	537	# LB 13
	538	$CL $CM+ $CAN_CM;
729e4ab9	539	$CP $CM+ $CAN_CM;
46f4442e A	540	$EX $CM+ $CAN_CM;
	541	$IS $CM+ $CAN_CM;
	542	$SY $CM+ $CAN_CM;
	543
	544	$CL [$LB8NonBreaks-$CM];
729e4ab9	545	$CP [$LB8NonBreaks-$CM];
46f4442e A	546	$EX [$LB8NonBreaks-$CM];
	547	$IS [$LB8NonBreaks-$CM];
	548	$SY [$LB8NonBreaks-$CM];
	549
	550	# Rule 13 & 14 taken together for an edge case.
	551	# Match this, shown forward
729e4ab9	552	# OP SP+ ($CM+ behaving as $AL) (CL \| CP \| EX \| IS \| IY)
46f4442e A	553	# This really wants to chain at the $CM+ (which is acting as an $AL)
46f4442e A	554	# except for $CM chaining being disabled.
729e4ab9	555	[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
46f4442e A	556
	557	# LB 14 OP SP* x
	558	#
	559	$CM* $CAN_CM $SP* $CM* $OP;
	560	$CANT_CM $SP* $CM* $OP;
	561	$AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
	562
	563	$AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
	564	$CM* $AL_FOLLOW_CM $CM+ $SP+ $CM* $OP;
	565	$SY $CM $SP+ $OP; # TODO: Experiment. Remove.
	566
	567
	568
	569	# LB 15
	570	$CM* $OP $SP* $CM* $QU;
	571
	572	# LB 16
729e4ab9	573	$CM* $NS $SP* $CM* ($CL \| $CP);
46f4442e A	574
	575	# LB 17
	576	$CM* $B2 $SP* $CM* $B2;
	577
	578	# LB 18 break after spaces
	579	# Nothing explicit needed here.
	580
	581
	582	#
	583	# LB 19
	584	#
	585	$CM* $QU $CM* $CAN_CM; # . x QU
	586	$CM* $QU $LB18NonBreaks;
	587
	588
	589	$CM* $CAN_CM $CM* $QU; # QU x .
	590	$CANT_CM $CM* $QU;
	591
	592	#
	593	# LB 20 Break before and after CB.
	594	# nothing needed here.
	595	#
	596
	597	# LB 21
	598	$CM* ($BA \| $HY \| $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA \| HY \| NS)
	599
	600	$CM* [$LB20NonBreaks-$CM] $CM* $BB; # BB x .
	601	[^$CB] $CM* $BB; #
	602
4388f060 A	603	# LB21a
4388f060 A	604	[^$CB] $CM* ($HY \| $BA) $CM* $HL;
46f4442e	605
51004dcb A	606	# LB21b (reverse)
51004dcb A	607	$CM* $HL $CM* $SY;
46f4442e A	608
46f4442e A	609	# LB 22
4388f060	610	$CM* $IN $CM* ($ALPlus \| $HL);
46f4442e A	611	$CM* $IN $CM* $ID;
	612	$CM* $IN $CM* $IN;
	613	$CM* $IN $CM* $NU;
	614
	615	# LB 23
	616	$CM* $PO $CM* $ID;
4388f060 A	617	$CM* $NU $CM* ($ALPlus \| $HL);
4388f060 A	618	$CM* ($ALPlus \| $HL) $CM* $NU;
46f4442e A	619
	620	# LB 24
	621	$CM* $ID $CM* $PR;
4388f060 A	622	$CM* ($ALPlus \| $HL) $CM* $PR;
4388f060 A	623	$CM* ($ALPlus \| $HL) $CM* $PO;
46f4442e A	624
	625
	626	# LB 25
729e4ab9	627	($CM* ($PR \| $PO))? ($CM* ($CL \| $CP))? ($CM* ($NU \| $IS \| $SY))* $CM* $NU ($CM* ($OP \| $HY))? ($CM* ($PR \| $PO))?;
46f4442e A	628
	629	# LB 26
	630	$CM* ($H3 \| $H2 \| $JV \| $JL) $CM* $JL;
	631	$CM* ($JT \| $JV) $CM* ($H2 \| $JV);
	632	$CM* $JT $CM* ($H3 \| $JT);
	633
	634	# LB 27
	635	$CM* $IN $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);
	636	$CM* $PO $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);
	637	$CM* ($H3 \| $H2 \| $JT \| $JV \| $JL) $CM* $PR;
	638
	639	# LB 28
4388f060	640	$CM* ($ALPlus \| $HL) $CM* ($ALPlus \| $HL);
46f4442e A	641
	642
	643	# LB 29
4388f060	644	$CM* ($ALPlus \| $HL) $CM* $IS;
46f4442e	645
729e4ab9	646	# LB 30
4388f060 A	647	$CM* $OP $CM* ($ALPlus \| $HL \| $NU);
4388f060 A	648	$CM* ($ALPlus \| $HL \| $NU) $CM* $CP;
729e4ab9	649
51004dcb A	650	# LB 30a
51004dcb A	651	$CM* $RI $CM* $RI;
46f4442e	652
08b89b0a A	653	# Special reverse rule for people & body part emoji:
	654	# don't break between relevant emoji and $EmojiMods
	655	$EmojiMods $EmojiVar? $EmojiForMods;
	656
46f4442e A	657	## -------------------------------------------------
	658
	659	!!safe_reverse;
	660
	661	# LB 9
	662	$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
	663	$CM+ $SP / .;
	664
	665	# LB 14
	666	$SP+ $CM* $OP;
	667
	668	# LB 15
	669	$SP+ $CM* $QU;
	670
	671	# LB 16
729e4ab9	672	$SP+ $CM* ($CL \| $CP);
46f4442e A	673
	674	# LB 17
	675	$SP+ $CM* $B2;
	676
4388f060 A	677	# LB 21
	678	$CM* ($HY \| $BA) $CM* $HL;
	679
46f4442e A	680	# LB 25
46f4442e A	681	($CM* ($IS \| $SY))+ $CM* $NU;
729e4ab9	682	($CL \| $CP) $CM* ($NU \| $IS \| $SY);
46f4442e A	683
	684	# For dictionary-based break
	685	$dictionary $dictionary;
	686
	687	## -------------------------------------------------
	688
	689	!!safe_forward;
	690
	691	# Skip forward over all character classes that are involved in
	692	# rules containing patterns with possibly more than one char
	693	# of context.
	694	#
	695	# It might be slightly more efficient to have specific rules
	696	# instead of one generic one, but only if we could
	697	# turn off rule chaining. We don't want to move more
	698	# than necessary.
	699	#
4388f060	700	[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
46f4442e A	701	$dictionary $dictionary;
46f4442e A	702