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

# Copyright (c) 2002-2006  International Business Machines Corporation and\r
# others. All Rights Reserved.\r
#\r
#  file:  line.txt\r
#\r
#         Line Breaking Rules\r
#         Implement default line breaking as defined by Unicode Standard Annex #14 version 5.0.0\r
#         http://www.unicode.org/reports/tr14/\r
\r
\r
\r
#\r
#  Character Classes defined by TR 14.\r
#\r
\r
!!chain;\r
!!LBCMNoChain;\r
\r
\r
!!lookAheadHardBreak;\r
#\r
#  !!lookAheadHardBreak    Described here because it is (as yet) undocumented elsewhere\r
#                          and only used for the line break rules.\r
#\r
#           It is used in the implementation of the incredibly annoying rule LB 10\r
#           which says to treat any combining mark that is not attached to a base\r
#           character as if it were of class AL  (alphabetic).\r
#\r
#           The problem occurs in the reverse rules.\r
#\r
#           Consider a sequence like, with correct breaks as shown\r
#               LF  ID  CM  AL  AL\r
#                  ^       ^       ^\r
#           Then consider the sequence without the initial ID (ideographic)\r
#                 LF  CM  AL  AL\r
#                    ^           ^\r
#           Our CM, which in the first example was attached to the ideograph,\r
#           is now unattached, becomes an alpha, and joins in with the other\r
#           alphas.\r
#\r
#           When iterating forwards, these sequences do not present any problems\r
#           When iterating backwards, we need to look ahead when encountering\r
#           a CM to see whether it attaches to something further on or not.\r
#           (Look-ahead in a reverse rule is looking towards the start)\r
#\r
#           If the CM is unattached, we need to force a break.\r
#\r
#           !!lookAheadHardBreak forces the run time state machine to\r
#           stop immediately when a look ahead rule ( '/' operator) matches,\r
#           and set the match position to that of the look-ahead operator,\r
#           no matter what other rules may be in play at the time.\r
#\r
#           See rule LB 19 for an example.\r
#\r
\r
$AI = [:LineBreak =  Ambiguous:];\r
$AL = [:LineBreak =  Alphabetic:];\r
$BA = [:LineBreak =  Break_After:];\r
$BB = [:LineBreak =  Break_Before:];\r
$BK = [:LineBreak =  Mandatory_Break:];\r
$B2 = [:LineBreak =  Break_Both:];\r
$CB = [:LineBreak =  Contingent_Break:];\r
$CL = [:LineBreak =  Close_Punctuation:];\r
$CM = [:LineBreak =  Combining_Mark:];\r
$CR = [:LineBreak =  Carriage_Return:];\r
$EX = [:LineBreak =  Exclamation:];\r
$GL = [:LineBreak =  Glue:];\r
$HY = [:LineBreak =  Hyphen:];\r
$H2 = [:LineBreak =  H2:];\r
$H3 = [:LineBreak =  H3:];\r
$ID = [:LineBreak =  Ideographic:];\r
$IN = [:LineBreak =  Inseperable:];\r
$IS = [:LineBreak =  Infix_Numeric:];\r
$JL = [:LineBreak =  JL:];\r
$JV = [:LineBreak =  JV:];\r
$JT = [:LineBreak =  JT:];\r
$LF = [:LineBreak =  Line_Feed:];\r
$NL = [:LineBreak =  Next_Line:];\r
$NS = [:LineBreak =  Nonstarter:];\r
$NU = [:LineBreak =  Numeric:];\r
$OP = [:LineBreak =  Open_Punctuation:];\r
$PO = [:LineBreak =  Postfix_Numeric:];\r
$PR = [:LineBreak =  Prefix_Numeric:];\r
$QU = [:LineBreak =  Quotation:];\r
$SA = [:LineBreak =  Complex_Context:];\r
$SG = [:LineBreak =  Surrogate:];\r
$SP = [:LineBreak =  Space:];\r
$SY = [:LineBreak =  Break_Symbols:];\r
$WJ = [:LineBreak =  Word_Joiner:];\r
$XX = [:LineBreak =  Unknown:];\r
$ZW = [:LineBreak =  ZWSpace:];\r
\r
#   Dictionary character set, for triggering language-based break engines. Currently\r
#   limited to LineBreak=Complex_Context. Note that this set only works in Unicode\r
#   5.0 or later as the definition of Complex_Context was corrected to include all\r
#   characters requiring dictionary break.\r
\r
$dictionary = [:LineBreak = Complex_Context:];\r
\r
#\r
#  Rule LB1.  By default, treat AI  (characters with ambiguous east Asian width),\r
#                               SA  (South East Asian: Thai, Lao, Khmer)\r
#                               SG  (Unpaired Surrogates)\r
#                               XX  (Unknown, unassigned)\r
#                         as $AL  (Alphabetic)\r
#\r
$ALPlus = [$AL $AI $SA $SG $XX];\r
\r
#\r
#  Combining Marks.   X $CM*  behaves as if it were X.  Rule LB6.\r
#\r
$ALcm = $ALPlus $CM*;\r
$BAcm = $BA $CM*;\r
$BBcm = $BB $CM*;\r
$B2cm = $B2 $CM*;\r
$CLcm = $CL $CM*;\r
$EXcm = $EX $CM*;\r
$GLcm = $GL $CM*;\r
$HYcm = $HY $CM*;\r
$H2cm = $H2 $CM*;\r
$H3cm = $H3 $CM*;\r
$IDcm = $ID $CM*;\r
$INcm = $IN $CM*;\r
$IScm = $IS $CM*;\r
$JLcm = $JL $CM*;\r
$JVcm = $JV $CM*;\r
$JTcm = $JT $CM*;\r
$NScm = $NS $CM*;\r
$NUcm = $NU $CM*;\r
$OPcm = $OP $CM*;\r
$POcm = $PO $CM*;\r
$PRcm = $PR $CM*;\r
$QUcm = $QU $CM*;\r
$SYcm = $SY $CM*;\r
$WJcm = $WJ $CM*;\r
\r
## -------------------------------------------------\r
\r
!!forward;\r
\r
#\r
#  Each class of character can stand by itself as an unbroken token, with trailing combining stuff\r
#\r
$ALPlus $CM+;\r
$BA $CM+;\r
$BB $CM+;\r
$B2 $CM+;\r
$CL $CM+;\r
$EX $CM+;\r
$GL $CM+;\r
$HY $CM+;\r
$H2 $CM+;\r
$H3 $CM+;\r
$ID $CM+;\r
$IN $CM+;\r
$IS $CM+;\r
$JL $CM+;\r
$JV $CM+;\r
$JT $CM+;\r
$NS $CM+;\r
$NU $CM+;\r
$OP $CM+;\r
$PO $CM+;\r
$PR $CM+;\r
$QU $CM+;\r
$SY $CM+;\r
$WJ $CM+;\r
\r
#\r
# CAN_CM  is the set of characters that may combine with CM combining chars.\r
#         Note that Linebreak UAX 14's concept of a combining char and the rules\r
#         for what they can combine with are _very_ different from the rest of Unicode.\r
#\r
#         Note that $CM itself is left out of this set.  If CM is needed as a base\r
#         it must be listed separately in the rule.\r
#\r
$CAN_CM  = [^$SP $BK $CR $LF $NL $ZW $CM];       # Bases that can   take CMs\r
$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM];       # Bases that can't take CMs\r
\r
#\r
# AL_FOLLOW  set of chars that can unconditionally follow an AL\r
#            Needed in rules where stand-alone $CM s are treated as AL.\r
#            Chaining is disabled with CM because it causes other failures,\r
#            so for this one case we need to manually list out longer sequences.\r
#\r
$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];\r
$AL_FOLLOW_CM   = [$CL $EX $IS $SY $WJ $GL $QU $BA $HY $NS $IN $NU $ALPlus $OP];\r
$AL_FOLLOW      = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];\r
\r
\r
#\r
#  Rule LB 4, 5    Mandatory (Hard) breaks.\r
#\r
$LB4Breaks    = [$BK $CR $LF $NL];\r
$LB4NonBreaks = [^$BK $CR $LF $NL];\r
$CR $LF {100};\r
\r
#\r
#  LB 6    Do not break before hard line breaks.\r
#\r
$LB4NonBreaks?  $LB4Breaks {100};    # LB 5  do not break before hard breaks.\r
$CAN_CM $CM*    $LB4Breaks {100};\r
$CM+            $LB4Breaks {100};\r
\r
# LB 7         x SP\r
#              x ZW\r
$LB4NonBreaks [$SP $ZW];\r
$CAN_CM $CM*  [$SP $ZW];\r
$CM+          [$SP $ZW];\r
\r
#\r
# LB 8         Break after zero width space\r
#\r
$LB8Breaks    = [$LB4Breaks $ZW];\r
$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];\r
\r
\r
# LB 9     Combining marks.      X   $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL \r
#                                $CM not covered by the above needs to behave like $AL   \r
#                                See definition of $CAN_CM.\r
\r
$CAN_CM $CM+;                   #  Stick together any combining sequences that don't match other rules.\r
$CM+;\r
\r
#\r
# LB 11  Do not break before or after WORD JOINER & related characters.\r
#\r
$CAN_CM $CM*  $WJcm;\r
$LB8NonBreaks $WJcm;\r
$CM+          $WJcm;\r
\r
$WJcm [^$CAN_CM];\r
$WJcm $CAN_CM $CM*;\r
\r
#\r
# LB 12  Do not break before or after NBSP and related characters.\r
#\r
#         (!SP) x GL\r
[$LB8NonBreaks-$SP] $CM* $GLcm;\r
$CM+               $GLcm;\r
\r
#         GL  x\r
$GLcm ($LB8Breaks | $SP);\r
$GLcm [$LB8NonBreaks-$SP] $CM*;     # Don't let a combining mark go onto $CR, $BK, etc.\r
                              #  TODO:  I don't think we need this rule.\r
                              #         All but $CM will chain off of preceding rule.\r
                              #         $GLcm will pick up the CM case by itself.\r
\r
\r
\r
\r
#\r
# LB 13   Don't break before ']' or '!' or ';' or '/', even after spaces.\r
#\r
$LB8NonBreaks $CL;\r
$CAN_CM $CM*  $CL;\r
$CM+          $CL;              # by rule 10, stand-alone CM behaves as AL\r
\r
$LB8NonBreaks $EX;\r
$CAN_CM $CM*  $EX;\r
$CM+          $EX;              # by rule 10, stand-alone CM behaves as AL\r
\r
$LB8NonBreaks $IS;\r
$CAN_CM $CM*  $IS;\r
$CM+          $IS;              # by rule 10, stand-alone CM behaves as AL\r
\r
$LB8NonBreaks $SY;\r
$CAN_CM $CM*  $SY;\r
$CM+          $SY;              # by rule 10, stand-alone CM behaves as AL\r
\r
\r
#\r
# LB 14  Do not break after OP, even after spaced\r
#\r
$OPcm $SP* $CAN_CM $CM*;\r
$OPcm $SP* $CANT_CM;\r
\r
$OPcm $SP+ $CM+ $AL_FOLLOW?;    # by rule 10, stand-alone CM behaves as AL\r
\r
# LB 15\r
$QUcm $SP* $OPcm;\r
\r
# LB 16\r
$CLcm $SP* $NScm;\r
\r
# LB 17\r
$B2cm $SP* $B2cm;\r
\r
#\r
# LB 18  Break after spaces.\r
#\r
$LB18NonBreaks = [$LB8NonBreaks - [$SP]];\r
$LB18Breaks    = [$LB8Breaks $SP];\r
\r
\r
# LB 19\r
#         x QU\r
$LB18NonBreaks $CM* $QUcm;\r
$CM+                $QUcm;\r
\r
#         QU  x\r
$QUcm .?;\r
$QUcm $LB18NonBreaks $CM*;    # Don't let a combining mark go onto $CR, $BK, etc.\r
                              #  TODO:  I don't think this rule is needed.\r
\r
\r
# LB 20\r
#        <break>  $CB\r
#        $CB   <break>\r
\r
$LB20NonBreaks = [$LB18NonBreaks - $CB];\r
\r
# LB 21        x   (BA | HY | NS)\r
#           BB x\r
#\r
$LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm); \r
\r
$BBcm [^$CB];                                  #  $BB  x\r
$BBcm $LB20NonBreaks $CM*;\r
\r
# LB 22\r
$ALcm    $INcm;\r
$CM+     $INcm;     #  by rule 10, any otherwise unattached CM behaves as AL\r
$IDcm    $INcm;\r
$INcm    $INcm;\r
$NUcm    $INcm;\r
\r
\r
# $LB 23\r
$IDcm  $POcm;\r
$ALcm  $NUcm;       # includes $LB19\r
$CM+   $NUcm;       # Rule 10, any otherwise unattached CM behaves as AL\r
$NUcm  $ALcm;\r
\r
#\r
# LB 24\r
#\r
$PRcm $IDcm;\r
$PRcm $ALcm;\r
$POcm $ALcm;\r
\r
#\r
# LB 25   Numbers.\r
#\r
($PRcm | $POcm)? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* $CLcm? ($PRcm | $POcm)?;\r
\r
# LB 26  Do not break a Korean syllable\r
#\r
$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);\r
($JVcm | $H2cm) ($JVcm | $JTcm);\r
($JTcm | $H3cm) $JTcm;\r
\r
# LB 27  Treat korean Syllable Block the same as ID  (don't break it)\r
($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;\r
($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;\r
$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);\r
\r
\r
# LB 28   Do not break between alphabetics\r
#\r
$ALcm $ALcm;\r
$CM+ $ALcm;      # The $CM+ is from rule 10, and unattached CM is treated as AL\r
\r
# LB 29\r
$IScm $ALcm;\r
\r
#\r
# Rule 30   Do not break between letters, numbers or ordinary symbols\r
#           and opening or closing punctuation\r
#\r
($ALcm | $NUcm) $OPcm;\r
$CM+ $OPcm;\r
$CLcm ($ALcm | $NUcm);\r
\r
\r
\r
#\r
#  Reverse Rules.\r
#\r
## -------------------------------------------------\r
\r
!!reverse;\r
\r
$CM+ $ALPlus;\r
$CM+ $BA;\r
$CM+ $BB;\r
$CM+ $B2;\r
$CM+ $CL;\r
$CM+ $EX;\r
$CM+ $GL;\r
$CM+ $HY;\r
$CM+ $H2;\r
$CM+ $H3;\r
$CM+ $ID;\r
$CM+ $IN;\r
$CM+ $IS;\r
$CM+ $JL;\r
$CM+ $JV;\r
$CM+ $JT;\r
$CM+ $NS;\r
$CM+ $NU;\r
$CM+ $OP;\r
$CM+ $PO;\r
$CM+ $PR;\r
$CM+ $QU;\r
$CM+ $SY;\r
$CM+ $WJ;\r
$CM+;\r
\r
\r
#\r
#  Sequences of the form  (shown forwards)\r
#      [CANT_CM]  <break>  [CM]  [whatever]\r
#  The CM needs to behave as an AL\r
#\r
$AL_FOLLOW $CM+ / (\r
          [$BK $CR $LF $NL $ZW {eof}] |\r
          $SP+ $CM+ $SP |\r
          $SP+ $CM* ([^$OP $CM $SP] | [$AL {eof}]));   # if LB 14 will match, need to surpress this break.\r
                                               #  LB14 says    OP SP* x .        \r
                                               #    becomes    OP SP* x AL\r
                                               #    becomes    OP SP* x CM+ AL_FOLLOW\r
                                               #\r
                                               # Further note:  the $AL in [$AL {eof}] is only to work around\r
                                               #                a rule compiler bug which complains about\r
                                               #                empty sets otherwise.\r
          \r
#\r
#  Sequences of the form  (shown forwards)\r
#      [CANT_CM]  <break> [CM]  <break>  [PR]\r
#  The CM needs to behave as an AL\r
#  This rule is concerned about getting the second of the two <breaks> in place.\r
#\r
\r
[$PR   ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];\r
\r
\r
\r
# LB 4, 5, 5\r
\r
$LB4Breaks [$LB4NonBreaks-$CM];\r
$LB4Breaks $CM+ $CAN_CM;\r
$LF $CR;\r
\r
\r
# LB 7         x SP\r
#              x ZW\r
[$SP $ZW] [$LB4NonBreaks-$CM];\r
[$SP $ZW] $CM+ $CAN_CM;\r
\r
# LB 8 Break after zero width space\r
\r
\r
# LB 9,10  Combining marks.\r
#    X   $CM needs to behave like X, where X is not $SP or controls.\r
#    $CM not covered by the above needs to behave like $AL\r
# Stick together any combining sequences that don't match other rules.\r
$CM+ $CAN_CM;\r
\r
\r
# LB 11\r
$CM* $WJ $CM* $CAN_CM;\r
$CM* $WJ      [$LB8NonBreaks-$CM];\r
\r
     $CANT_CM $CM* $WJ;\r
$CM* $CAN_CM  $CM* $WJ;\r
\r
# LB 12\r
#         x GL\r
#\r
$CM* $GL $CM* [$LB8NonBreaks-$CM-$SP];\r
\r
#\r
#     GL  x\r
#\r
$CANT_CM $CM* $GL;\r
$CM* $CAN_CM $CM* $GL;\r
\r
\r
# LB 13\r
$CL $CM+ $CAN_CM;\r
$EX $CM+ $CAN_CM;\r
$IS $CM+ $CAN_CM;\r
$SY $CM+ $CAN_CM;\r
\r
$CL [$LB8NonBreaks-$CM];\r
$EX [$LB8NonBreaks-$CM];\r
$IS [$LB8NonBreaks-$CM];\r
$SY [$LB8NonBreaks-$CM];\r
\r
# Rule 13 & 14 taken together for an edge case.\r
#   Match this, shown forward\r
#     OP SP+  ($CM+ behaving as $AL) (CL | EX | IS | IY)\r
#   This really wants to chain at the $CM+ (which is acting as an $AL)\r
#   except for $CM chaining being disabled.\r
[$CL $EX $IS $SY] $CM+ $SP+ $CM* $OP;  \r
\r
# LB 14    OP SP* x\r
#\r
$CM* $CAN_CM    $SP* $CM* $OP;\r
     $CANT_CM   $SP* $CM* $OP;\r
$AL_FOLLOW? $CM+  $SP $SP* $CM* $OP;     #  by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP\r
     \r
     $AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;\r
$CM* $AL_FOLLOW_CM   $CM+ $SP+ $CM* $OP;\r
$SY $CM $SP+ $OP;   # TODO:  Experiment.  Remove.\r
\r
\r
\r
# LB 15\r
$CM* $OP $SP* $CM* $QU;\r
\r
# LB 16\r
$CM* $NS $SP* $CM* $CL;\r
\r
# LB 17\r
$CM* $B2 $SP* $CM* $B2;\r
\r
# LB 18  break after spaces\r
#        Nothing explicit needed here.\r
\r
\r
#\r
# LB 19\r
#\r
$CM* $QU $CM* $CAN_CM;                                #   . x QU\r
$CM* $QU      $LB18NonBreaks;\r
\r
\r
$CM* $CAN_CM  $CM* $QU;                               #   QU x .\r
     $CANT_CM $CM* $QU;\r
     \r
#\r
#  LB 20  Break before and after CB.\r
#         nothing needed here.\r
#\r
\r
# LB 21\r
$CM* ($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM];     #  . x (BA | HY | NS)\r
\r
$CM* [$LB20NonBreaks-$CM] $CM* $BB;                   #  BB x .\r
[^$CB] $CM* $BB;                                      # \r
\r
\r
\r
# LB 22\r
$CM* $IN $CM* $ALPlus;\r
$CM* $IN $CM* $ID;\r
$CM* $IN $CM* $IN;\r
$CM* $IN $CM* $NU;\r
\r
# LB 23\r
$CM* $PO $CM* $ID;\r
$CM* $NU $CM* $ALPlus;\r
$CM* $ALPlus $CM* $NU;\r
\r
# LB 24\r
$CM* $ID $CM* $PR;\r
$CM* $ALPlus $CM* $PR;\r
$CM* $ALPlus $CM* $PO;\r
\r
\r
# LB 25\r
($CM* ($PR | $PO))? ($CM* $CL)? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;\r
\r
# LB 26\r
$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;\r
$CM* ($JT | $JV) $CM* ($H2 | $JV);\r
$CM* $JT $CM* ($H3 | $JT);\r
\r
# LB 27\r
$CM* $IN $CM* ($H3 | $H2 | $JT | $JV | $JL);\r
$CM* $PO $CM* ($H3 | $H2 | $JT | $JV | $JL);\r
$CM* ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;\r
\r
# LB 28\r
$CM* $ALPlus $CM* $ALPlus;\r
\r
\r
# LB 29\r
$CM* $ALPlus $CM* $IS;\r
\r
# LB 30\r
$CM* $OP $CM* ($NU | $ALPlus);\r
$CM* ($NU | $ALPlus) $CM* $CL;\r
\r
\r
## -------------------------------------------------\r
\r
!!safe_reverse;\r
\r
# LB 7\r
$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];\r
$CM+ $SP / .;\r
\r
# LB 9\r
$SP+ $CM* $OP;\r
\r
# LB 10\r
$SP+ $CM* $QU;\r
\r
# LB 11\r
$SP+ $CM* $CL;\r
$SP+ $CM* $B2;\r
\r
# LB 18\r
($CM* ($IS | $SY))+ $CM* $NU;\r
$CL $CM* ($NU | $IS | $SY);\r
\r
# For dictionary-based break\r
$dictionary $dictionary;\r
\r
## -------------------------------------------------\r
\r
!!safe_forward;\r
\r
# Skip forward over all character classes that are involved in\r
#   rules containing patterns with possibly more than one char\r
#   of context.\r
#\r
#  It might be slightly more efficient to have specific rules\r
#  instead of one generic one, but only if we could\r
#  turn off rule chaining.  We don't want to move more\r
#  than necessary.\r
#\r
[$CM $OP $QU $CL $B2 $PR $HY $SP $dictionary]+ [^$CM $OP $QU $CL $B2 $PR $HY $dictionary];\r
$dictionary $dictionary;\r
\r
Commit	Line	Data
73c04bcf A	1	# Copyright (c) 2002-2006 International Business Machines Corporation and\r
	2	# others. All Rights Reserved.\r
	3	#\r
	4	# file: line.txt\r
	5	#\r
	6	# Line Breaking Rules\r
	7	# Implement default line breaking as defined by Unicode Standard Annex #14 version 5.0.0\r
	8	# http://www.unicode.org/reports/tr14/\r
	9	\r
	10	\r
	11	\r
	12	#\r
	13	# Character Classes defined by TR 14.\r
	14	#\r
	15	\r
	16	!!chain;\r
	17	!!LBCMNoChain;\r
	18	\r
	19	\r
	20	!!lookAheadHardBreak;\r
	21	#\r
	22	# !!lookAheadHardBreak Described here because it is (as yet) undocumented elsewhere\r
	23	# and only used for the line break rules.\r
	24	#\r
	25	# It is used in the implementation of the incredibly annoying rule LB 10\r
	26	# which says to treat any combining mark that is not attached to a base\r
	27	# character as if it were of class AL (alphabetic).\r
	28	#\r
	29	# The problem occurs in the reverse rules.\r
	30	#\r
	31	# Consider a sequence like, with correct breaks as shown\r
	32	# LF ID CM AL AL\r
	33	# ^ ^ ^\r
	34	# Then consider the sequence without the initial ID (ideographic)\r
	35	# LF CM AL AL\r
	36	# ^ ^\r
	37	# Our CM, which in the first example was attached to the ideograph,\r
	38	# is now unattached, becomes an alpha, and joins in with the other\r
	39	# alphas.\r
	40	#\r
	41	# When iterating forwards, these sequences do not present any problems\r
	42	# When iterating backwards, we need to look ahead when encountering\r
	43	# a CM to see whether it attaches to something further on or not.\r
	44	# (Look-ahead in a reverse rule is looking towards the start)\r
	45	#\r
	46	# If the CM is unattached, we need to force a break.\r
	47	#\r
	48	# !!lookAheadHardBreak forces the run time state machine to\r
	49	# stop immediately when a look ahead rule ( '/' operator) matches,\r
	50	# and set the match position to that of the look-ahead operator,\r
	51	# no matter what other rules may be in play at the time.\r
	52	#\r
	53	# See rule LB 19 for an example.\r
	54	#\r
	55	\r
	56	$AI = [:LineBreak = Ambiguous:];\r
	57	$AL = [:LineBreak = Alphabetic:];\r
	58	$BA = [:LineBreak = Break_After:];\r
	59	$BB = [:LineBreak = Break_Before:];\r
	60	$BK = [:LineBreak = Mandatory_Break:];\r
	61	$B2 = [:LineBreak = Break_Both:];\r
	62	$CB = [:LineBreak = Contingent_Break:];\r
	63	$CL = [:LineBreak = Close_Punctuation:];\r
	64	$CM = [:LineBreak = Combining_Mark:];\r
65	$CR = [:LineBreak = Carriage_Return:];\r
66	$EX = [:LineBreak = Exclamation:];\r
67	$GL = [:LineBreak = Glue:];\r
68	$HY = [:LineBreak = Hyphen:];\r
69	$H2 = [:LineBreak = H2:];\r
70	$H3 = [:LineBreak = H3:];\r
71	$ID = [:LineBreak = Ideographic:];\r
72	$IN = [:LineBreak = Inseperable:];\r
73	$IS = [:LineBreak = Infix_Numeric:];\r
74	$JL = [:LineBreak = JL:];\r
75	$JV = [:LineBreak = JV:];\r
76	$JT = [:LineBreak = JT:];\r
77	$LF = [:LineBreak = Line_Feed:];\r
78	$NL = [:LineBreak = Next_Line:];\r
79	$NS = [:LineBreak = Nonstarter:];\r
80	$NU = [:LineBreak = Numeric:];\r
81	$OP = [:LineBreak = Open_Punctuation:];\r
82	$PO = [:LineBreak = Postfix_Numeric:];\r
83	$PR = [:LineBreak = Prefix_Numeric:];\r
84	$QU = [:LineBreak = Quotation:];\r
85	$SA = [:LineBreak = Complex_Context:];\r
86	$SG = [:LineBreak = Surrogate:];\r
87	$SP = [:LineBreak = Space:];\r
88	$SY = [:LineBreak = Break_Symbols:];\r
89	$WJ = [:LineBreak = Word_Joiner:];\r
90	$XX = [:LineBreak = Unknown:];\r
91	$ZW = [:LineBreak = ZWSpace:];\r
92	\r
93	# Dictionary character set, for triggering language-based break engines. Currently\r
94	# limited to LineBreak=Complex_Context. Note that this set only works in Unicode\r
95	# 5.0 or later as the definition of Complex_Context was corrected to include all\r
96	# characters requiring dictionary break.\r
97	\r
98	$dictionary = [:LineBreak = Complex_Context:];\r
99	\r
100	#\r
101	# Rule LB1. By default, treat AI (characters with ambiguous east Asian width),\r
102	# SA (South East Asian: Thai, Lao, Khmer)\r
103	# SG (Unpaired Surrogates)\r
104	# XX (Unknown, unassigned)\r
105	# as $AL (Alphabetic)\r
106	#\r
107	$ALPlus = [$AL $AI $SA $SG $XX];\r
108	\r
109	#\r
110	# Combining Marks. X $CM* behaves as if it were X. Rule LB6.\r
111	#\r
112	$ALcm = $ALPlus $CM*;\r
113	$BAcm = $BA $CM*;\r
114	$BBcm = $BB $CM*;\r
115	$B2cm = $B2 $CM*;\r
116	$CLcm = $CL $CM*;\r
117	$EXcm = $EX $CM*;\r
118	$GLcm = $GL $CM*;\r
119	$HYcm = $HY $CM*;\r
120	$H2cm = $H2 $CM*;\r
121	$H3cm = $H3 $CM*;\r
122	$IDcm = $ID $CM*;\r
123	$INcm = $IN $CM*;\r
124	$IScm = $IS $CM*;\r
125	$JLcm = $JL $CM*;\r
126	$JVcm = $JV $CM*;\r
127	$JTcm = $JT $CM*;\r
128	$NScm = $NS $CM*;\r
129	$NUcm = $NU $CM*;\r
130	$OPcm = $OP $CM*;\r
131	$POcm = $PO $CM*;\r
132	$PRcm = $PR $CM*;\r
133	$QUcm = $QU $CM*;\r
134	$SYcm = $SY $CM*;\r
135	$WJcm = $WJ $CM*;\r
136	\r
137	## -------------------------------------------------\r
138	\r
139	!!forward;\r
140	\r
141	#\r
142	# Each class of character can stand by itself as an unbroken token, with trailing combining stuff\r
143	#\r
144	$ALPlus $CM+;\r
145	$BA $CM+;\r
146	$BB $CM+;\r
147	$B2 $CM+;\r
148	$CL $CM+;\r
149	$EX $CM+;\r
150	$GL $CM+;\r
151	$HY $CM+;\r
152	$H2 $CM+;\r
153	$H3 $CM+;\r
154	$ID $CM+;\r
155	$IN $CM+;\r
156	$IS $CM+;\r
157	$JL $CM+;\r
158	$JV $CM+;\r
159	$JT $CM+;\r
160	$NS $CM+;\r
161	$NU $CM+;\r
162	$OP $CM+;\r
163	$PO $CM+;\r
164	$PR $CM+;\r
165	$QU $CM+;\r
166	$SY $CM+;\r
167	$WJ $CM+;\r
168	\r
169	#\r
170	# CAN_CM is the set of characters that may combine with CM combining chars.\r
171	# Note that Linebreak UAX 14's concept of a combining char and the rules\r
172	# for what they can combine with are _very_ different from the rest of Unicode.\r
173	#\r
174	# Note that $CM itself is left out of this set. If CM is needed as a base\r
175	# it must be listed separately in the rule.\r
176	#\r
177	$CAN_CM = [^$SP $BK $CR $LF $NL $ZW $CM]; # Bases that can take CMs\r
178	$CANT_CM = [ $SP $BK $CR $LF $NL $ZW $CM]; # Bases that can't take CMs\r
179	\r
180	#\r
181	# AL_FOLLOW set of chars that can unconditionally follow an AL\r
182	# Needed in rules where stand-alone $CM s are treated as AL.\r
183	# Chaining is disabled with CM because it causes other failures,\r
184	# so for this one case we need to manually list out longer sequences.\r
185	#\r
186	$AL_FOLLOW_NOCM = [$BK $CR $LF $NL $ZW $SP];\r
187	$AL_FOLLOW_CM = [$CL $EX $IS $SY $WJ $GL $QU $BA $HY $NS $IN $NU $ALPlus $OP];\r
188	$AL_FOLLOW = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];\r
189	\r
190	\r
191	#\r
192	# Rule LB 4, 5 Mandatory (Hard) breaks.\r
193	#\r
194	$LB4Breaks = [$BK $CR $LF $NL];\r
195	$LB4NonBreaks = [^$BK $CR $LF $NL];\r
196	$CR $LF {100};\r
197	\r
198	#\r
199	# LB 6 Do not break before hard line breaks.\r
200	#\r
201	$LB4NonBreaks? $LB4Breaks {100}; # LB 5 do not break before hard breaks.\r
202	$CAN_CM $CM* $LB4Breaks {100};\r
203	$CM+ $LB4Breaks {100};\r
204	\r
205	# LB 7 x SP\r
206	# x ZW\r
207	$LB4NonBreaks [$SP $ZW];\r
208	$CAN_CM $CM* [$SP $ZW];\r
209	$CM+ [$SP $ZW];\r
210	\r
211	#\r
212	# LB 8 Break after zero width space\r
213	#\r
214	$LB8Breaks = [$LB4Breaks $ZW];\r
215	$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];\r
216	\r
217	\r
218	# LB 9 Combining marks. X $CM needs to behave like X, where X is not $SP, $BK $CR $LF $NL \r
219	# $CM not covered by the above needs to behave like $AL \r
220	# See definition of $CAN_CM.\r
221	\r
222	$CAN_CM $CM+; # Stick together any combining sequences that don't match other rules.\r
223	$CM+;\r
224	\r
225	#\r
226	# LB 11 Do not break before or after WORD JOINER & related characters.\r
227	#\r
228	$CAN_CM $CM* $WJcm;\r
229	$LB8NonBreaks $WJcm;\r
230	$CM+ $WJcm;\r
231	\r
232	$WJcm [^$CAN_CM];\r
233	$WJcm $CAN_CM $CM*;\r
234	\r
235	#\r
236	# LB 12 Do not break before or after NBSP and related characters.\r
237	#\r
238	# (!SP) x GL\r
239	[$LB8NonBreaks-$SP] $CM* $GLcm;\r
240	$CM+ $GLcm;\r
241	\r
242	# GL x\r
243	$GLcm ($LB8Breaks \| $SP);\r
244	$GLcm [$LB8NonBreaks-$SP] $CM*; # Don't let a combining mark go onto $CR, $BK, etc.\r
245	# TODO: I don't think we need this rule.\r
246	# All but $CM will chain off of preceding rule.\r
247	# $GLcm will pick up the CM case by itself.\r
248	\r
249	\r
250	\r
251	\r
252	#\r
253	# LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces.\r
254	#\r
255	$LB8NonBreaks $CL;\r
256	$CAN_CM $CM* $CL;\r
257	$CM+ $CL; # by rule 10, stand-alone CM behaves as AL\r
258	\r
259	$LB8NonBreaks $EX;\r
260	$CAN_CM $CM* $EX;\r
261	$CM+ $EX; # by rule 10, stand-alone CM behaves as AL\r
262	\r
263	$LB8NonBreaks $IS;\r
264	$CAN_CM $CM* $IS;\r
265	$CM+ $IS; # by rule 10, stand-alone CM behaves as AL\r
266	\r
267	$LB8NonBreaks $SY;\r
268	$CAN_CM $CM* $SY;\r
269	$CM+ $SY; # by rule 10, stand-alone CM behaves as AL\r
270	\r
271	\r
272	#\r
273	# LB 14 Do not break after OP, even after spaced\r
274	#\r
275	$OPcm $SP* $CAN_CM $CM*;\r
276	$OPcm $SP* $CANT_CM;\r
277	\r
278	$OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL\r
279	\r
280	# LB 15\r
281	$QUcm $SP* $OPcm;\r
282	\r
283	# LB 16\r
284	$CLcm $SP* $NScm;\r
285	\r
286	# LB 17\r
287	$B2cm $SP* $B2cm;\r
288	\r
289	#\r
290	# LB 18 Break after spaces.\r
291	#\r
292	$LB18NonBreaks = [$LB8NonBreaks - [$SP]];\r
293	$LB18Breaks = [$LB8Breaks $SP];\r
294	\r
295	\r
296	# LB 19\r
297	# x QU\r
298	$LB18NonBreaks $CM* $QUcm;\r
299	$CM+ $QUcm;\r
300	\r
301	# QU x\r
302	$QUcm .?;\r
303	$QUcm $LB18NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc.\r
304	# TODO: I don't think this rule is needed.\r
305	\r
306	\r
307	# LB 20\r
308	# <break> $CB\r
309	# $CB <break>\r
310	\r
311	$LB20NonBreaks = [$LB18NonBreaks - $CB];\r
312	\r
313	# LB 21 x (BA \| HY \| NS)\r
314	# BB x\r
315	#\r
316	$LB20NonBreaks $CM* ($BAcm \| $HYcm \| $NScm); \r
317	\r
318	$BBcm [^$CB]; # $BB x\r
319	$BBcm $LB20NonBreaks $CM*;\r
320	\r
321	# LB 22\r
322	$ALcm $INcm;\r
323	$CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL\r
324	$IDcm $INcm;\r
325	$INcm $INcm;\r
326	$NUcm $INcm;\r
327	\r
328	\r
329	# $LB 23\r
330	$IDcm $POcm;\r
331	$ALcm $NUcm; # includes $LB19\r
332	$CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL\r
333	$NUcm $ALcm;\r
334	\r
335	#\r
336	# LB 24\r
337	#\r
338	$PRcm $IDcm;\r
339	$PRcm $ALcm;\r
340	$POcm $ALcm;\r
341	\r
342	#\r
343	# LB 25 Numbers.\r
344	#\r
345	($PRcm \| $POcm)? ($OPcm \| $HYcm)? $NUcm ($NUcm \| $SYcm \| $IScm)* $CLcm? ($PRcm \| $POcm)?;\r
346	\r
347	# LB 26 Do not break a Korean syllable\r
348	#\r
349	$JLcm ($JLcm \| $JVcm \| $H2cm \| $H3cm);\r
350	($JVcm \| $H2cm) ($JVcm \| $JTcm);\r
351	($JTcm \| $H3cm) $JTcm;\r
352	\r
353	# LB 27 Treat korean Syllable Block the same as ID (don't break it)\r
354	($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm) $INcm;\r
355	($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm) $POcm;\r
356	$PRcm ($JLcm \| $JVcm \| $JTcm \| $H2cm \| $H3cm);\r
357	\r
358	\r
359	# LB 28 Do not break between alphabetics\r
360	#\r
361	$ALcm $ALcm;\r
362	$CM+ $ALcm; # The $CM+ is from rule 10, and unattached CM is treated as AL\r
363	\r
364	# LB 29\r
365	$IScm $ALcm;\r
366	\r
367	#\r
368	# Rule 30 Do not break between letters, numbers or ordinary symbols\r
369	# and opening or closing punctuation\r
370	#\r
371	($ALcm \| $NUcm) $OPcm;\r
372	$CM+ $OPcm;\r
373	$CLcm ($ALcm \| $NUcm);\r
374	\r
375	\r
376	\r
377	#\r
378	# Reverse Rules.\r
379	#\r
380	## -------------------------------------------------\r
381	\r
382	!!reverse;\r
383	\r
384	$CM+ $ALPlus;\r
385	$CM+ $BA;\r
386	$CM+ $BB;\r
387	$CM+ $B2;\r
388	$CM+ $CL;\r
389	$CM+ $EX;\r
390	$CM+ $GL;\r
391	$CM+ $HY;\r
392	$CM+ $H2;\r
393	$CM+ $H3;\r
394	$CM+ $ID;\r
395	$CM+ $IN;\r
396	$CM+ $IS;\r
397	$CM+ $JL;\r
398	$CM+ $JV;\r
399	$CM+ $JT;\r
400	$CM+ $NS;\r
401	$CM+ $NU;\r
402	$CM+ $OP;\r
403	$CM+ $PO;\r
404	$CM+ $PR;\r
405	$CM+ $QU;\r
406	$CM+ $SY;\r
407	$CM+ $WJ;\r
408	$CM+;\r
409	\r
410	\r
411	#\r
412	# Sequences of the form (shown forwards)\r
413	# [CANT_CM] <break> [CM] [whatever]\r
414	# The CM needs to behave as an AL\r
415	#\r
416	$AL_FOLLOW $CM+ / (\r
417	[$BK $CR $LF $NL $ZW {eof}] \|\r
418	$SP+ $CM+ $SP \|\r
419	$SP+ $CM* ([^$OP $CM $SP] \| [$AL {eof}])); # if LB 14 will match, need to surpress this break.\r
420	# LB14 says OP SP* x . \r
421	# becomes OP SP* x AL\r
422	# becomes OP SP* x CM+ AL_FOLLOW\r
423	#\r
424	# Further note: the $AL in [$AL {eof}] is only to work around\r
425	# a rule compiler bug which complains about\r
426	# empty sets otherwise.\r
427	\r
428	#\r
429	# Sequences of the form (shown forwards)\r
430	# [CANT_CM] <break> [CM] <break> [PR]\r
431	# The CM needs to behave as an AL\r
432	# This rule is concerned about getting the second of the two <breaks> in place.\r
433	#\r
434	\r
435	[$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];\r
436	\r
437	\r
438	\r
439	# LB 4, 5, 5\r
440	\r
441	$LB4Breaks [$LB4NonBreaks-$CM];\r
442	$LB4Breaks $CM+ $CAN_CM;\r
443	$LF $CR;\r
444	\r
445	\r
446	# LB 7 x SP\r
447	# x ZW\r
448	[$SP $ZW] [$LB4NonBreaks-$CM];\r
449	[$SP $ZW] $CM+ $CAN_CM;\r
450	\r
451	# LB 8 Break after zero width space\r
452	\r
453	\r
454	# LB 9,10 Combining marks.\r
455	# X $CM needs to behave like X, where X is not $SP or controls.\r
456	# $CM not covered by the above needs to behave like $AL\r
457	# Stick together any combining sequences that don't match other rules.\r
458	$CM+ $CAN_CM;\r
459	\r
460	\r
461	# LB 11\r
462	$CM* $WJ $CM* $CAN_CM;\r
463	$CM* $WJ [$LB8NonBreaks-$CM];\r
464	\r
465	$CANT_CM $CM* $WJ;\r
466	$CM* $CAN_CM $CM* $WJ;\r
467	\r
468	# LB 12\r
469	# x GL\r
470	#\r
471	$CM* $GL $CM* [$LB8NonBreaks-$CM-$SP];\r
472	\r
473	#\r
474	# GL x\r
475	#\r
476	$CANT_CM $CM* $GL;\r
477	$CM* $CAN_CM $CM* $GL;\r
478	\r
479	\r
480	# LB 13\r
481	$CL $CM+ $CAN_CM;\r
482	$EX $CM+ $CAN_CM;\r
483	$IS $CM+ $CAN_CM;\r
484	$SY $CM+ $CAN_CM;\r
485	\r
486	$CL [$LB8NonBreaks-$CM];\r
487	$EX [$LB8NonBreaks-$CM];\r
488	$IS [$LB8NonBreaks-$CM];\r
489	$SY [$LB8NonBreaks-$CM];\r
490	\r
491	# Rule 13 & 14 taken together for an edge case.\r
492	# Match this, shown forward\r
493	# OP SP+ ($CM+ behaving as $AL) (CL \| EX \| IS \| IY)\r
494	# This really wants to chain at the $CM+ (which is acting as an $AL)\r
495	# except for $CM chaining being disabled.\r
496	[$CL $EX $IS $SY] $CM+ $SP+ $CM* $OP; \r
497	\r
498	# LB 14 OP SP* x\r
499	#\r
500	$CM* $CAN_CM $SP* $CM* $OP;\r
501	$CANT_CM $SP* $CM* $OP;\r
502	$AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 10, behaves like $AL_FOLLOW? $AL $SP* $CM* $OP\r
503	\r
504	$AL_FOLLOW_NOCM $CM+ $SP+ $CM* $OP;\r
505	$CM* $AL_FOLLOW_CM $CM+ $SP+ $CM* $OP;\r
506	$SY $CM $SP+ $OP; # TODO: Experiment. Remove.\r
507	\r
508	\r
509	\r
510	# LB 15\r
511	$CM* $OP $SP* $CM* $QU;\r
512	\r
513	# LB 16\r
514	$CM* $NS $SP* $CM* $CL;\r
515	\r
516	# LB 17\r
517	$CM* $B2 $SP* $CM* $B2;\r
518	\r
519	# LB 18 break after spaces\r
520	# Nothing explicit needed here.\r
521	\r
522	\r
523	#\r
524	# LB 19\r
525	#\r
526	$CM* $QU $CM* $CAN_CM; # . x QU\r
527	$CM* $QU $LB18NonBreaks;\r
528	\r
529	\r
530	$CM* $CAN_CM $CM* $QU; # QU x .\r
531	$CANT_CM $CM* $QU;\r
532	\r
533	#\r
534	# LB 20 Break before and after CB.\r
535	# nothing needed here.\r
536	#\r
537	\r
538	# LB 21\r
539	$CM* ($BA \| $HY \| $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA \| HY \| NS)\r
540	\r
541	$CM* [$LB20NonBreaks-$CM] $CM* $BB; # BB x .\r
542	[^$CB] $CM* $BB; # \r
543	\r
544	\r
545	\r
546	# LB 22\r
547	$CM* $IN $CM* $ALPlus;\r
548	$CM* $IN $CM* $ID;\r
549	$CM* $IN $CM* $IN;\r
550	$CM* $IN $CM* $NU;\r
551	\r
552	# LB 23\r
553	$CM* $PO $CM* $ID;\r
554	$CM* $NU $CM* $ALPlus;\r
555	$CM* $ALPlus $CM* $NU;\r
556	\r
557	# LB 24\r
558	$CM* $ID $CM* $PR;\r
559	$CM* $ALPlus $CM* $PR;\r
560	$CM* $ALPlus $CM* $PO;\r
561	\r
562	\r
563	# LB 25\r
564	($CM* ($PR \| $PO))? ($CM* $CL)? ($CM* ($NU \| $IS \| $SY))* $CM* $NU ($CM* ($OP \| $HY))? ($CM* ($PR \| $PO))?;\r
565	\r
566	# LB 26\r
567	$CM* ($H3 \| $H2 \| $JV \| $JL) $CM* $JL;\r
568	$CM* ($JT \| $JV) $CM* ($H2 \| $JV);\r
569	$CM* $JT $CM* ($H3 \| $JT);\r
570	\r
571	# LB 27\r
572	$CM* $IN $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);\r
573	$CM* $PO $CM* ($H3 \| $H2 \| $JT \| $JV \| $JL);\r
574	$CM* ($H3 \| $H2 \| $JT \| $JV \| $JL) $CM* $PR;\r
575	\r
576	# LB 28\r
577	$CM* $ALPlus $CM* $ALPlus;\r
578	\r
579	\r
580	# LB 29\r
581	$CM* $ALPlus $CM* $IS;\r
582	\r
583	# LB 30\r
584	$CM* $OP $CM* ($NU \| $ALPlus);\r
585	$CM* ($NU \| $ALPlus) $CM* $CL;\r
586	\r
587	\r
588	## -------------------------------------------------\r
589	\r
590	!!safe_reverse;\r
591	\r
592	# LB 7\r
593	$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];\r
594	$CM+ $SP / .;\r
595	\r
596	# LB 9\r
597	$SP+ $CM* $OP;\r
598	\r
599	# LB 10\r
600	$SP+ $CM* $QU;\r
601	\r
602	# LB 11\r
603	$SP+ $CM* $CL;\r
604	$SP+ $CM* $B2;\r
605	\r
606	# LB 18\r
607	($CM* ($IS \| $SY))+ $CM* $NU;\r
608	$CL $CM* ($NU \| $IS \| $SY);\r
609	\r
610	# For dictionary-based break\r
611	$dictionary $dictionary;\r
612	\r
613	## -------------------------------------------------\r
614	\r
615	!!safe_forward;\r
616	\r
617	# Skip forward over all character classes that are involved in\r
618	# rules containing patterns with possibly more than one char\r
619	# of context.\r
620	#\r
621	# It might be slightly more efficient to have specific rules\r
622	# instead of one generic one, but only if we could\r
623	# turn off rule chaining. We don't want to move more\r
624	# than necessary.\r
625	#\r
626	[$CM $OP $QU $CL $B2 $PR $HY $SP $dictionary]+ [^$CM $OP $QU $CL $B2 $PR $HY $dictionary];\r
627	$dictionary $dictionary;\r
628	\r