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

# Copyright (c) 2002-2013  International Business Machines Corporation and
# others. All Rights Reserved.
#
#  file:  line_fi.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.

#
#  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:] - [\u2010]];
$HH = [\u2010];
$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:];

#   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*;
$HHcm = $HH $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+;
$HH $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 $HH $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]];


# 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 $HH $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 20.09 added rule for Finnish tailoring
# LB 21        x   (BA | HY | NS)
#           BB x
#
$LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm) / $AL; 
$LB20NonBreaks $CM* ($BAcm | $HHcm | $HYcm | $NScm); 
($HY | $HH) $AL;

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

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

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

!!reverse;

$CM+ $ALPlus;
$CM+ $BA;
$CM+ $HH;
$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

# 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 $HH $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 20.09 added rule for Finnish tailoring
$AL ($HY | $HH) / $SP;

# LB 21
$CM* ($BA | $HH | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)

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

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

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

!!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 | $HH) $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
57a6839d	1	# Copyright (c) 2002-2013 International Business Machines Corporation and
729e4ab9 A	2	# others. All Rights Reserved.
	3	#
	4	# file: line_fi.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
729e4ab9 A	9	# http://www.unicode.org/reports/tr14/
	10	#
	11	# TODO: Rule LB 8 remains as it was in Unicode 5.2
	12	# This is only because of a limitation of ICU break engine implementation,
	13	# not because the older behavior is desirable.
	14
	15	#
	16	# Character Classes defined by TR 14.
	17	#
	18
	19	!!chain;
	20	!!LBCMNoChain;
	21
	22
	23	!!lookAheadHardBreak;
	24	#
	25	# !!lookAheadHardBreak Described here because it is (as yet) undocumented elsewhere
	26	# and only used for the line break rules.
	27	#
	28	# It is used in the implementation of rule LB 10
	29	# which says to treat any combining mark that is not attached to a base
	30	# character as if it were of class AL (alphabetic).
	31	#
	32	# The problem occurs in the reverse rules.
	33	#
	34	# Consider a sequence like, with correct breaks as shown
	35	# LF ID CM AL AL
	36	# ^ ^ ^
	37	# Then consider the sequence without the initial ID (ideographic)
	38	# LF CM AL AL
	39	# ^ ^
	40	# Our CM, which in the first example was attached to the ideograph,
	41	# is now unattached, becomes an alpha, and joins in with the other
	42	# alphas.
	43	#
	44	# When iterating forwards, these sequences do not present any problems
	45	# When iterating backwards, we need to look ahead when encountering
	46	# a CM to see whether it attaches to something further on or not.
	47	# (Look-ahead in a reverse rule is looking towards the start)
	48	#
	49	# If the CM is unattached, we need to force a break.
	50	#
	51	# !!lookAheadHardBreak forces the run time state machine to
	52	# stop immediately when a look ahead rule ( '/' operator) matches,
	53	# and set the match position to that of the look-ahead operator,
	54	# no matter what other rules may be in play at the time.
	55	#
	56	# See rule LB 19 for an example.
	57	#
	58
	59	$AI = [:LineBreak = Ambiguous:];
	60	$AL = [:LineBreak = Alphabetic:];
	61	$BA = [[:LineBreak = Break_After:] - [\u2010]];
	62	$HH = [\u2010];
	63	$BB = [:LineBreak = Break_Before:];
	64	$BK = [:LineBreak = Mandatory_Break:];
	65	$B2 = [:LineBreak = Break_Both:];
	66	$CB = [:LineBreak = Contingent_Break:];
4388f060	67	$CJ = [:LineBreak = Conditional_Japanese_Starter:];
729e4ab9 A	68	$CL = [:LineBreak = Close_Punctuation:];
	69	$CM = [:LineBreak = Combining_Mark:];
	70	$CP = [:LineBreak = Close_Parenthesis:];
	71	$CR = [:LineBreak = Carriage_Return:];
	72	$EX = [:LineBreak = Exclamation:];
	73	$GL = [:LineBreak = Glue:];
4388f060	74	$HL = [:LineBreak = Hebrew_Letter:];
729e4ab9 A	75	$HY = [:LineBreak = Hyphen:];
	76	$H2 = [:LineBreak = H2:];
	77	$H3 = [:LineBreak = H3:];
	78	$ID = [:LineBreak = Ideographic:];
	79	$IN = [:LineBreak = Inseperable:];
	80	$IS = [:LineBreak = Infix_Numeric:];
	81	$JL = [:LineBreak = JL:];
	82	$JV = [:LineBreak = JV:];
	83	$JT = [:LineBreak = JT:];
	84	$LF = [:LineBreak = Line_Feed:];
	85	$NL = [:LineBreak = Next_Line:];
4388f060	86	$NS = [[:LineBreak = Nonstarter:] $CJ];
729e4ab9 A	87	$NU = [:LineBreak = Numeric:];
	88	$OP = [:LineBreak = Open_Punctuation:];
	89	$PO = [:LineBreak = Postfix_Numeric:];
	90	$PR = [:LineBreak = Prefix_Numeric:];
	91	$QU = [:LineBreak = Quotation:];
51004dcb	92	$RI = [:LineBreak = Regional_Indicator:];
729e4ab9 A	93	$SA = [:LineBreak = Complex_Context:];
	94	$SG = [:LineBreak = Surrogate:];
	95	$SP = [:LineBreak = Space:];
	96	$SY = [:LineBreak = Break_Symbols:];
	97	$WJ = [:LineBreak = Word_Joiner:];
	98	$XX = [:LineBreak = Unknown:];
	99	$ZW = [:LineBreak = ZWSpace:];
	100
	101	# Dictionary character set, for triggering language-based break engines. Currently
	102	# limited to LineBreak=Complex_Context. Note that this set only works in Unicode
	103	# 5.0 or later as the definition of Complex_Context was corrected to include all
	104	# characters requiring dictionary break.
	105
	106	$dictionary = [:LineBreak = Complex_Context:];
	107
	108	#
	109	# Rule LB1. By default, treat AI (characters with ambiguous east Asian width),
	110	# SA (South East Asian: Thai, Lao, Khmer)
	111	# SG (Unpaired Surrogates)
	112	# XX (Unknown, unassigned)
	113	# as $AL (Alphabetic)
	114	#
	115	$ALPlus = [$AL $AI $SA $SG $XX];
	116
	117	#
	118	# Combining Marks. X $CM* behaves as if it were X. Rule LB6.
	119	#
	120	$ALcm = $ALPlus $CM*;
	121	$BAcm = $BA $CM*;
	122	$HHcm = $HH $CM*;
	123	$BBcm = $BB $CM*;
	124	$B2cm = $B2 $CM*;
	125	$CLcm = $CL $CM*;
	126	$CPcm = $CP $CM*;
	127	$EXcm = $EX $CM*;
	128	$GLcm = $GL $CM*;
4388f060	129	$HLcm = $HL $CM*;
729e4ab9 A	130	$HYcm = $HY $CM*;
	131	$H2cm = $H2 $CM*;
	132	$H3cm = $H3 $CM*;
	133	$IDcm = $ID $CM*;
	134	$INcm = $IN $CM*;
	135	$IScm = $IS $CM*;
	136	$JLcm = $JL $CM*;
	137	$JVcm = $JV $CM*;
	138	$JTcm = $JT $CM*;
	139	$NScm = $NS $CM*;
	140	$NUcm = $NU $CM*;
	141	$OPcm = $OP $CM*;
	142	$POcm = $PO $CM*;
	143	$PRcm = $PR $CM*;
	144	$QUcm = $QU $CM*;
51004dcb	145	$RIcm = $RI $CM*;
729e4ab9 A	146	$SYcm = $SY $CM*;
	147	$WJcm = $WJ $CM*;
	148
	149	## -------------------------------------------------
	150
	151	!!forward;
	152
	153	#
	154	# Each class of character can stand by itself as an unbroken token, with trailing combining stuff
	155	#
	156	$ALPlus $CM+;
	157	$BA $CM+;
	158	$HH $CM+;
	159	$BB $CM+;
	160	$B2 $CM+;
	161	$CL $CM+;
	162	$CP $CM+;
	163	$EX $CM+;
	164	$GL $CM+;
4388f060	165	$HL $CM+;
729e4ab9 A	166	$HY $CM+;
	167	$H2 $CM+;
	168	$H3 $CM+;
	169	$ID $CM+;
	170	$IN $CM+;
	171	$IS $CM+;
	172	$JL $CM+;
	173	$JV $CM+;
	174	$JT $CM+;
	175	$NS $CM+;
	176	$NU $CM+;
	177	$OP $CM+;
	178	$PO $CM+;
	179	$PR $CM+;
	180	$QU $CM+;
51004dcb	181	$RI $CM+;
729e4ab9 A	182	$SY $CM+;
	183	$WJ $CM+;
	184
	185	#
	186	# CAN_CM is the set of characters that may combine with CM combining chars.
	187	# Note that Linebreak UAX 14's concept of a combining char and the rules
	188	# for what they can combine with are _very_ different from the rest of Unicode.
	189	#
	190	# Note that $CM itself is left out of this set. If CM is needed as a base
	191	# it must be listed separately in the rule.
	192	#
	193	$CAN_CM = [^$SP $BK $CR $LF $NL $ZW $CM]; # Bases that can take CMs
	194	$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM]; # Bases that can't take CMs
	195
	196	#
	197	# AL_FOLLOW set of chars that can unconditionally follow an AL
	198	# Needed in rules where stand-alone $CM s are treated as AL.
	199	# Chaining is disabled with CM because it causes other failures,
	200	# so for this one case we need to manually list out longer sequences.
	201	#
	202	$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];
4388f060	203	$AL_FOLLOW_CM = [$CL $CP $EX $HL $IS $SY $WJ $GL $OP $QU $BA $HH $HY $NS $IN $NU $ALPlus];
729e4ab9 A	204	$AL_FOLLOW = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
	205
	206
	207	#
	208	# Rule LB 4, 5 Mandatory (Hard) breaks.
	209	#
	210	$LB4Breaks = [$BK $CR $LF $NL];
	211	$LB4NonBreaks = [^$BK $CR $LF $NL];
	212	$CR $LF {100};
	213
	214	#
	215	# LB 6 Do not break before hard line breaks.
	216	#
	217	$LB4NonBreaks? $LB4Breaks {100}; # LB 5 do not break before hard breaks.
	218	$CAN_CM $CM* $LB4Breaks {100};
	219	$CM+ $LB4Breaks {100};
	220
	221	# LB 7 x SP
	222	# x ZW
	223	$LB4NonBreaks [$SP $ZW];
	224	$CAN_CM $CM* [$SP $ZW];
	225	$CM+ [$SP $ZW];
	226
	227	#
	228	# LB 8 Break after zero width space
	229	# TODO: ZW SP* <break>
	230	# An engine change is required to write the reverse rule for this.
	231	# For now, leave the Unicode 5.2 rule, ZW <break>
	232	#
	233	$LB8Breaks = [$LB4Breaks $ZW];
	234	$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
	235
	236
	237	# LB 9 Combining marks. X $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL
	238	# $CM not covered by the above needs to behave like $AL
	239	# See definition of $CAN_CM.
	240
	241	$CAN_CM $CM+; # Stick together any combining sequences that don't match other rules.
	242	$CM+;
	243
	244	#
	245	# LB 11 Do not break before or after WORD JOINER & related characters.
	246	#
	247	$CAN_CM $CM* $WJcm;
	248	$LB8NonBreaks $WJcm;
	249	$CM+ $WJcm;
	250
	251	$WJcm $CANT_CM;
	252	$WJcm $CAN_CM $CM*;
	253
	254	#
	255	# LB 12 Do not break after NBSP and related characters.
	256	# GL x
	257	#
	258	$GLcm $CAN_CM $CM*;
	259	$GLcm $CANT_CM;
	260
	261	#
	262	# LB 12a Do not break before NBSP and related characters ...
	263	# [^SP BA HY] x GL
	264	#
	265	[[$LB8NonBreaks] - [$SP $BA $HH $HY]] $CM* $GLcm;
	266	$CM+ GLcm;
	267
268
269
270	#
271	# LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces.
272	#
273	$LB8NonBreaks $CL;
274	$CAN_CM $CM* $CL;
275	$CM+ $CL; # by rule 10, stand-alone CM behaves as AL
276
277	$LB8NonBreaks $CP;
278	$CAN_CM $CM* $CP;
279	$CM+ $CP; # by rule 10, stand-alone CM behaves as AL
280
281	$LB8NonBreaks $EX;
282	$CAN_CM $CM* $EX;
283	$CM+ $EX; # by rule 10, stand-alone CM behaves as AL
284
285	$LB8NonBreaks $IS;
286	$CAN_CM $CM* $IS;
287	$CM+ $IS; # by rule 10, stand-alone CM behaves as AL
288
289	$LB8NonBreaks $SY;
290	$CAN_CM $CM* $SY;
291	$CM+ $SY; # by rule 10, stand-alone CM behaves as AL
292
293
294	#
295	# LB 14 Do not break after OP, even after spaces
296	#
297	$OPcm $SP* $CAN_CM $CM*;
298	$OPcm $SP* $CANT_CM;
299
300	$OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL
301
302	# LB 15
303	$QUcm $SP* $OPcm;
304
305	# LB 16
306	($CLcm \| $CPcm) $SP* $NScm;
307
308	# LB 17
309	$B2cm $SP* $B2cm;
310
311	#
312	# LB 18 Break after spaces.
313	#
314	$LB18NonBreaks = [$LB8NonBreaks - [$SP]];
315	$LB18Breaks = [$LB8Breaks $SP];
316
317
318	# LB 19
319	# x QU
320	$LB18NonBreaks $CM* $QUcm;
321	$CM+ $QUcm;
322
323	# QU x
324	$QUcm .?;
325	$QUcm $LB18NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc.
326	# TODO: I don't think this rule is needed.
327
328
329	# LB 20
330	# <break> $CB
331	# $CB <break>
332
333	$LB20NonBreaks = [$LB18NonBreaks - $CB];
334
335	# LB 20.09 added rule for Finnish tailoring
336	# LB 21 x (BA \| HY \| NS)
337	# BB x
338	#
339	$LB20NonBreaks $CM* ($BAcm \| $HHcm \| $HYcm \| $NScm) / $AL;
340	$LB20NonBreaks $CM* ($BAcm \| $HHcm \| $HYcm \| $NScm);
341	($HY \| $HH) $AL;
342
343	$BBcm [^$CB]; # $BB x
344	$BBcm $LB20NonBreaks $CM*;
345
4388f060 A	346	# LB 21a Don't break after Hebrew + Hyphen
	347	# HL (HY \| BA) x
	348	#
	349	$HLcm ($HYcm \| $BAcm \| $HHcm) [^$CB]?;
	350
57a6839d A	351	# LB 21b (forward) Don't break between SY and HL
	352	# (break between HL and SY already disallowed by LB 13 above)
	353	$SYcm $HLcm;
	354
729e4ab9	355	# LB 22
4388f060	356	($ALcm \| $HLcm) $INcm;
729e4ab9 A	357	$CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL
	358	$IDcm $INcm;
	359	$INcm $INcm;
	360	$NUcm $INcm;
	361
	362
	363	# $LB 23
	364	$IDcm $POcm;
	365	$ALcm $NUcm; # includes $LB19
4388f060	366	$HLcm $NUcm;
729e4ab9 A	367	$CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL
729e4ab9 A	368	$NUcm $ALcm;
4388f060	369	$NUcm $HLcm;
729e4ab9 A	370
	371	#
	372	# LB 24
	373	#
	374	$PRcm $IDcm;
4388f060 A	375	$PRcm ($ALcm \| $HLcm);
4388f060 A	376	$POcm ($ALcm \| $HLcm);
729e4ab9 A	377
	378	#
	379	# LB 25 Numbers.
	380	#
	381	($PRcm \| $POcm)? ($OPcm \| $HYcm)? $NUcm ($NUcm \| $SYcm \| $IScm)* ($CLcm \| $CPcm)? ($PRcm \| $POcm)?;
	382
	383	# LB 26 Do not break a Korean syllable
	384	#
	385	$JLcm ($JLcm \| $JVcm \| $H2cm \| $H3cm);
	386	($JVcm \| $H2cm) ($JVcm \| $JTcm);
	387	($JTcm \| $H3cm) $JTcm;
	388
	389	# LB 27 Treat korean Syllable Block the same as ID (don't break it)
	390	($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm) $INcm;
	391	($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm) $POcm;
	392	$PRcm ($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm);
	393
	394
	395	# LB 28 Do not break between alphabetics
	396	#
4388f060 A	397	($ALcm \| $HLcm) ($ALcm \| $HLcm);
4388f060 A	398	$CM+ ($ALcm \| $HLcm); # The $CM+ is from rule 10, an unattached CM is treated as AL
729e4ab9 A	399
729e4ab9 A	400	# LB 29
4388f060	401	$IScm ($ALcm \| $HLcm);
729e4ab9 A	402
729e4ab9 A	403	# LB 30
4388f060	404	($ALcm \| $HLcm \| $NUcm) $OPcm;
729e4ab9	405	$CM+ $OPcm; # The $CM+ is from rule 10, an unattached CM is treated as AL.
4388f060	406	$CPcm ($ALcm \| $HLcm \| $NUcm);
729e4ab9	407
51004dcb A	408	# LB 30a Do not break between regional indicators.
51004dcb A	409	$RIcm $RIcm;
729e4ab9 A	410
	411	#
	412	# Reverse Rules.
	413	#
	414	## -------------------------------------------------
	415
	416	!!reverse;
	417
	418	$CM+ $ALPlus;
	419	$CM+ $BA;
	420	$CM+ $HH;
	421	$CM+ $BB;
	422	$CM+ $B2;
	423	$CM+ $CL;
	424	$CM+ $CP;
	425	$CM+ $EX;
	426	$CM+ $GL;
4388f060	427	$CM+ $HL;
729e4ab9 A	428	$CM+ $HY;
	429	$CM+ $H2;
	430	$CM+ $H3;
	431	$CM+ $ID;
	432	$CM+ $IN;
	433	$CM+ $IS;
	434	$CM+ $JL;
	435	$CM+ $JV;
	436	$CM+ $JT;
	437	$CM+ $NS;
	438	$CM+ $NU;
	439	$CM+ $OP;
	440	$CM+ $PO;
	441	$CM+ $PR;
	442	$CM+ $QU;
51004dcb	443	$CM+ $RI;
729e4ab9 A	444	$CM+ $SY;
	445	$CM+ $WJ;
	446	$CM+;
	447
	448
	449	#
	450	# Sequences of the form (shown forwards)
	451	# [CANT_CM] <break> [CM] [whatever]
	452	# The CM needs to behave as an AL
	453	#
	454	$AL_FOLLOW $CM+ / (
	455	[$BK $CR $LF $NL $ZW {eof}] \|
	456	$SP+ $CM+ $SP \|
	457	$SP+ $CM* ([^$OP $CM $SP] \| [$AL {eof}])); # if LB 14 will match, need to surpress this break.
	458	# LB14 says OP SP* x .
	459	# becomes OP SP* x AL
	460	# becomes OP SP* x CM+ AL_FOLLOW
	461	#
	462	# Further note: the $AL in [$AL {eof}] is only to work around
	463	# a rule compiler bug which complains about
	464	# empty sets otherwise.
	465
	466	#
	467	# Sequences of the form (shown forwards)
	468	# [CANT_CM] <break> [CM] <break> [PR]
	469	# The CM needs to behave as an AL
	470	# This rule is concerned about getting the second of the two <breaks> in place.
	471	#
	472
	473	[$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
	474
	475
	476
	477	# LB 4, 5, 5
	478
	479	$LB4Breaks [$LB4NonBreaks-$CM];
	480	$LB4Breaks $CM+ $CAN_CM;
	481	$LF $CR;
	482
	483
	484	# LB 7 x SP
	485	# x ZW
	486	[$SP $ZW] [$LB4NonBreaks-$CM];
	487	[$SP $ZW] $CM+ $CAN_CM;
	488
	489	# LB 8 ZW SP* <break>
	490	# TODO: to implement this, we need more than one look-ahead hard break in play at a time.
	491	# Requires an engine enhancement.
	492	# / $SP* $ZW
	493
	494	# LB 9,10 Combining marks.
	495	# X $CM needs to behave like X, where X is not $SP or controls.
	496	# $CM not covered by the above needs to behave like $AL
	497	# Stick together any combining sequences that don't match other rules.
	498	$CM+ $CAN_CM;
	499
	500
	501	# LB 11
	502	$CM* $WJ $CM* $CAN_CM;
	503	$CM* $WJ [$LB8NonBreaks-$CM];
	504
	505	$CANT_CM $CM* $WJ;
	506	$CM* $CAN_CM $CM* $WJ;
	507
508	# LB 12a
509	# [^SP BA HY] x GL
510	#
511	$CM* $GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HH $HY]];
512
513	# LB 12
514	# GL x
515	#
516	$CANT_CM $CM* $GL;
517	$CM* $CAN_CM $CM* $GL;
518
519
520	# LB 13
521	$CL $CM+ $CAN_CM;
522	$CP $CM+ $CAN_CM;
523	$EX $CM+ $CAN_CM;
524	$IS $CM+ $CAN_CM;
525	$SY $CM+ $CAN_CM;
526
527	$CL [$LB8NonBreaks-$CM];
528	$CP [$LB8NonBreaks-$CM];
529	$EX [$LB8NonBreaks-$CM];
530	$IS [$LB8NonBreaks-$CM];
531	$SY [$LB8NonBreaks-$CM];
532
533	# Rule 13 & 14 taken together for an edge case.
534	# Match this, shown forward
535	# OP SP+ ($CM+ behaving as $AL) (CL \| CP \| EX \| IS \| IY)
536	# This really wants to chain at the $CM+ (which is acting as an $AL)
537	# except for $CM chaining being disabled.
538	[$CL $CP $EX $IS $SY] $CM+ $SP+ $CM* $OP;
539
540	# LB 14 OP SP* x
541	#
542	$CM* $CAN_CM $SP* $CM* $OP;
543	$CANT_CM $SP* $CM* $OP;
544	$AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP
545
546	$AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;
547	$CM* $AL_FOLLOW_CM $CM+ $SP+ $CM* $OP;
548	$SY $CM $SP+ $OP; # TODO: Experiment. Remove.
549
550
551
552	# LB 15
553	$CM* $OP $SP* $CM* $QU;
554
555	# LB 16
556	$CM* $NS $SP* $CM* ($CL \| $CP);
557
558	# LB 17
559	$CM* $B2 $SP* $CM* $B2;
560
561	# LB 18 break after spaces
562	# Nothing explicit needed here.
563
564
565	#
566	# LB 19
567	#
568	$CM* $QU $CM* $CAN_CM; # . x QU
569	$CM* $QU $LB18NonBreaks;
570
571
572	$CM* $CAN_CM $CM* $QU; # QU x .
573	$CANT_CM $CM* $QU;
574
575	#
576	# LB 20 Break before and after CB.
577	# nothing needed here.
578	#
579
580	# LB 20.09 added rule for Finnish tailoring
581	$AL ($HY \| $HH) / $SP;
582
583	# LB 21
584	$CM* ($BA \| $HH \| $HY \| $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA \| HY \| NS)
585
586	$CM* [$LB20NonBreaks-$CM] $CM* $BB; # BB x .
587	[^$CB] $CM* $BB; #
588
4388f060 A	589	# LB21a
4388f060 A	590	[^$CB] $CM* ($HY \| $BA \| $HH) $CM* $HL;
729e4ab9	591
57a6839d A	592	# LB21b (reverse)
57a6839d A	593	$CM* $HL $CM* $SY;
729e4ab9 A	594
729e4ab9 A	595	# LB 22
4388f060	596	$CM* $IN $CM* ($ALPlus \| $HL);
729e4ab9 A	597	$CM* $IN $CM* $ID;
	598	$CM* $IN $CM* $IN;
	599	$CM* $IN $CM* $NU;
	600
	601	# LB 23
	602	$CM* $PO $CM* $ID;
4388f060 A	603	$CM* $NU $CM* ($ALPlus \| $HL);
4388f060 A	604	$CM* ($ALPlus \| $HL) $CM* $NU;
729e4ab9 A	605
	606	# LB 24
	607	$CM* $ID $CM* $PR;
4388f060 A	608	$CM* ($ALPlus \| $HL) $CM* $PR;
4388f060 A	609	$CM* ($ALPlus \| $HL) $CM* $PO;
729e4ab9 A	610
	611
	612	# LB 25
	613	($CM* ($PR \| $PO))? ($CM* ($CL \| $CP))? ($CM* ($NU \| $IS \| $SY))* $CM* $NU ($CM* ($OP \| $HY))? ($CM* ($PR \| $PO))?;
	614
	615	# LB 26
	616	$CM* ($H3 \| $H2 \| $JV \| $JL) $CM* $JL;
	617	$CM* ($JT \| $JV) $CM* ($H2 \| $JV);
	618	$CM* $JT $CM* ($H3 \| $JT);
	619
	620	# LB 27
	621	$CM* $IN $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);
	622	$CM* $PO $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);
	623	$CM* ($H3 \| $H2 \| $JT \| $JV \| $JL) $CM* $PR;
	624
	625	# LB 28
4388f060	626	$CM* ($ALPlus \| $HL) $CM* ($ALPlus \| $HL);
729e4ab9 A	627
	628
	629	# LB 29
4388f060	630	$CM* ($ALPlus \| $HL) $CM* $IS;
729e4ab9 A	631
729e4ab9 A	632	# LB 30
4388f060 A	633	$CM* $OP $CM* ($ALPlus \| $HL \| $NU);
4388f060 A	634	$CM* ($ALPlus \| $HL \| $NU) $CM* $CP;
729e4ab9	635
51004dcb A	636	# LB 30a
51004dcb A	637	$CM* $RI $CM* $RI;
729e4ab9 A	638
	639	## -------------------------------------------------
	640
	641	!!safe_reverse;
	642
	643	# LB 9
	644	$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
	645	$CM+ $SP / .;
	646
	647	# LB 14
	648	$SP+ $CM* $OP;
	649
	650	# LB 15
	651	$SP+ $CM* $QU;
	652
	653	# LB 16
	654	$SP+ $CM* ($CL \| $CP);
	655
	656	# LB 17
	657	$SP+ $CM* $B2;
	658
4388f060 A	659	# LB 21
	660	$CM* ($HY \| $BA \| $HH) $CM* $HL;
	661
729e4ab9 A	662	# LB 25
	663	($CM* ($IS \| $SY))+ $CM* $NU;
	664	($CL \| $CP) $CM* ($NU \| $IS \| $SY);
	665
	666	# For dictionary-based break
	667	$dictionary $dictionary;
	668
	669	## -------------------------------------------------
	670
	671	!!safe_forward;
	672
	673	# Skip forward over all character classes that are involved in
	674	# rules containing patterns with possibly more than one char
	675	# of context.
	676	#
	677	# It might be slightly more efficient to have specific rules
	678	# instead of one generic one, but only if we could
	679	# turn off rule chaining. We don't want to move more
	680	# than necessary.
	681	#
4388f060	682	[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary];
729e4ab9 A	683	$dictionary $dictionary;
729e4ab9 A	684