X-Git-Url: https://git.saurik.com/apple/icu.git/blobdiff_plain/374ca955a76ecab1204ca8bfa63ff9238d998416..4388f060552cc537e71e957d32f35e9d75a61233:/icuSources/data/brkitr/line.txt diff --git a/icuSources/data/brkitr/line.txt b/icuSources/data/brkitr/line.txt index 944a1ced..dcad39bd 100644 --- a/icuSources/data/brkitr/line.txt +++ b/icuSources/data/brkitr/line.txt @@ -1,12 +1,16 @@ -# Copyright (c) 2002-2004 International Business Machines Corporation and +# Copyright (c) 2002-2011 International Business Machines Corporation and # others. All Rights Reserved. # # file: line.txt # # Line Breaking Rules -# Implement default line breaking as defined by Unicode TR 14. +# Implement default line breaking as defined by +# Unicode Standard Annex #14 Revision 28 for Unicode 6.1 +# 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. @@ -14,7 +18,43 @@ !!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:]; @@ -23,18 +63,26 @@ $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:]; +$NS = [[:LineBreak = Nonstarter:] $CJ]; $NU = [:LineBreak = Numeric:]; $OP = [:LineBreak = Open_Punctuation:]; $PO = [:LineBreak = Postfix_Numeric:]; @@ -48,26 +96,21 @@ $WJ = [:LineBreak = Word_Joiner:]; $XX = [:LineBreak = Unknown:]; $ZW = [:LineBreak = ZWSpace:]; +# Dictionary character set, for triggering language-based break engines. Currently +# limited to LineBreak=Complex_Context. Note that this set only works in Unicode +# 5.0 or later as the definition of Complex_Context was corrected to include all +# characters requiring dictionary break. -# -# Korean Syllable Definitions -# -$L = [:Hangul_Syllable_Type = L:]; -$V = [:Hangul_Syllable_Type = V:]; -$T = [:Hangul_Syllable_Type = T:]; - -$LV = [:Hangul_Syllable_Type = LV:]; -$LVT = [:Hangul_Syllable_Type = LVT:]; - -$HangulSyllable = $L+ | ($L* ($LV? $V+ | $LV | $LVT) $T*) | $T+; +$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 | $XX; +$ALPlus = [$AL $AI $SA $SG $XX]; # # Combining Marks. X $CM* behaves as if it were X. Rule LB6. @@ -77,22 +120,32 @@ $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 $CM*; -$IDcm = ($ID | $HangulSyllable) $CM*; +$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 $CM*; -$SPcm = $SP $CM*; $SYcm = $SY $CM*; $WJcm = $WJ $CM*; +## ------------------------------------------------- + +!!forward; + # # Each class of character can stand by itself as an unbroken token, with trailing combining stuff # @@ -101,124 +154,244 @@ $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 $CM+; -$SP $CM+; $SY $CM+; $WJ $CM+; -## ------------------------------------------------- - -!!forward; +# +# 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 $CM]; # Bases that can't take CMs # -# Rule LB 3 -$LB3Breaks = [$BK $CR $LF $NL]; -$LB3NonBreaks = [^$BK $CR $LF $NL]; -$LB5NonBreaks = [[$LB3NonBreaks] - [$ZW]]; +# 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 $AL_FOLLOW_CM]; + -$LB3NonBreaks? $LB3Breaks {100}; -$LB5NonBreaks $CM* $LB3Breaks {100}; +# +# Rule LB 4, 5 Mandatory (Hard) breaks. +# +$LB4Breaks = [$BK $CR $LF $NL]; +$LB4NonBreaks = [^$BK $CR $LF $NL]; $CR $LF {100}; -# LB 4 x SP -# x ZW -$ZW [$SP $ZW]; -$LB5NonBreaks $CM* [$SP $ZW]; +# +# LB 6 Do not break before hard line breaks. +# +$LB4NonBreaks? $LB4Breaks {100}; # LB 5 do not break before hard breaks. +$CAN_CM $CM* $LB4Breaks {100}; +$CM+ $LB4Breaks {100}; -# LB 5 Break after zero width space -$LB5Breaks = [$LB3Breaks $ZW]; +# LB 7 x SP +# x ZW +$LB4NonBreaks [$SP $ZW]; +$CAN_CM $CM* [$SP $ZW]; +$CM+ [$SP $ZW]; -# LB 6 # -# Korean Syllable Definitions +# LB 8 Break after zero width space +# TODO: ZW SP* +# An engine change is required to write the reverse rule for this. +# For now, leave the Unicode 5.2 rule, ZW # +$LB8Breaks = [$LB4Breaks $ZW]; +$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]]; -($HangulSyllable) $CM*; -# LB 7 Combining marks. $SP $CM needs to behave like $ID. -# X $CM needs to behave like X, where X is not $SP. +# 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 $AL -# -$LB5NonBreaks $CM+; # Stick together any combining sequences that don't match other rules. +# See definition of $CAN_CM. -# LB 8 -$LB5NonBreaks $CM* $CL; -$LB5NonBreaks $CM* $EX; -$LB5NonBreaks $CM* $IS; -$LB5NonBreaks $CM* $SY; +$CAN_CM $CM+; # Stick together any combining sequences that don't match other rules. +$CM+; -# LB 9 -$OPcm $SP* .?; -$OPcm $SP* $LB5NonBreaks $CM*; +# +# LB 11 Do not break before or after WORD JOINER & related characters. +# +$CAN_CM $CM* $WJcm; +$LB8NonBreaks $WJcm; +$CM+ $WJcm; -# LB 10 -$QUcm $SP* $OPcm; +$WJcm $CANT_CM; +$WJcm $CAN_CM $CM*; -# LB 11 -$CLcm $SP* $NScm; +# +# LB 12 Do not break after NBSP and related characters. +# GL x +# +$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 11a -($B2cm)+; -# LB 11b -$LB5NonBreaks $CM* ($GLcm | $WJcm); -($GLcm | $WJcm) .?; -# LB 12 -$LB12NonBreaks = [$LB5NonBreaks - $SP]; +# +# LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces. +# +$LB8NonBreaks $CL; +$CAN_CM $CM* $CL; +$CM+ $CL; # by rule 10, stand-alone CM behaves as AL -# LB 14 -$LB12NonBreaks $CM* $QUcm+ .?; -$LB12NonBreaks $CM* $QUcm+ $LB5NonBreaks $CM*; -$SP $CM+ $QUcm+ .?; # LB7a SP CM+ behaves as ID -$SP $CM+ $QUcm+ $LB5NonBreaks $CM*; +$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+ $IS; # by rule 10, stand-alone CM behaves as AL -$QUcm $LB3NonBreaks?; -$QUcm $LB5NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc. +$LB8NonBreaks $SY; +$CAN_CM $CM* $SY; +$CM+ $SY; # by rule 10, stand-alone CM behaves as AL -# LB 14a -$LB14NonBreaks = [$LB12NonBreaks - $CB]; -$LB14CanBreakAfter = $LB14NonBreaks $CM* | $SP $CM+; + +# +# LB 14 Do not break after OP, even after spaces +# +$OPcm $SP* $CAN_CM $CM*; +$OPcm $SP* $CANT_CM; + +$OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL # LB 15 -$LB14CanBreakAfter ($BAcm | $HYcm | $NScm); -$BBcm [^$CB]; -$BBcm [^$CB $CR $LF $BK $NL $ZW] $CM*; +$QUcm $SP* $OPcm; # LB 16 -$ALcm $INcm; -$CM+ $INcm; # by rule 7c, any otherwise unattached CM behaves as AL +($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 +# $CB +# $CB + +$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 22 +($ALcm | $HLcm) $INcm; +$CM+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL $IDcm $INcm; -$SP $CM+ $INcm; # by rule 7a, $SP $CM behaves like ID $INcm $INcm; $NUcm $INcm; -# $LB 17 -($IDcm | $SP $CM+) $POcm; -$ALcm+ $NUcm; # includes $LB19 -$CM+ $NUcm; # Rule 7c -$NUcm $ALcm+; +# $LB 23 +$IDcm $POcm; +$ALcm $NUcm; # includes $LB19 +$HLcm $NUcm; +$CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL +$NUcm $ALcm; +$NUcm $HLcm; -# LB 18 -$PRcm? ($OPcm | $HYcm)? $NUcm ($NUcm | $IScm | $SYcm)* $CLcm? $POcm?; -$PRcm $ALcm; +# +# 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 | $NUcm); -# LB 19 -$CM* $ALcm+; # The $CM* is from rule 7C, and unattached CM is treated as AL -$IScm $ALcm; # # Reverse Rules. @@ -232,160 +405,256 @@ $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+ $SP; $CM+ $SY; $CM+ $WJ; +$CM+; -# LB 3 -$LB3Breaks $LB3NonBreaks; -$LB3Breaks $CM* $LB5NonBreaks; -$LF $CR; +# +# Sequences of the form (shown forwards) +# [CANT_CM] [CM] [whatever] +# The CM needs to behave as an AL +# +$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] [CM] [PR] +# The CM needs to behave as an AL +# This rule is concerned about getting the second of the two in place. +# -# LB 4 x SP -# x ZW -[$SP $ZW] $LB3NonBreaks; -[$SP $ZW] $CM* $LB5NonBreaks; +[$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}]; -# LB 5 Break after zero width space -# LB 6 Jamo is treated like an alphabet -$BackHangulSyllable = $L+ | ($T* ($V+$LV? | $LV | $LVT) $L*) | $T+; -$CM* $BackHangulSyllable; +# LB 4, 5, 5 -# LB 7 Combining marks. -# $SP $CM needs to behave like $ID. -# X $CM needs to behave like X, where X is not $SP. -# $CM not covered by the above needs to behave like $AL -# Stick together any combining sequences that don't match other rules. -$CM+ $LB5NonBreaks; +$LB4Breaks [$LB4NonBreaks-$CM]; +$LB4Breaks $CM+ $CAN_CM; +$LF $CR; -# LB 8 -$CL $CM* $LB5NonBreaks; -$EX $CM* $LB5NonBreaks; -$IS $CM* $LB5NonBreaks; -$SY $CM* $LB5NonBreaks; -# LB 9 -$LB5NonBreaks $SP* $CM* $OP; +# LB 7 x SP +# x ZW +[$SP $ZW] [$LB4NonBreaks-$CM]; +[$SP $ZW] $CM+ $CAN_CM; + +# LB 8 ZW SP* +# 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+ $CAN_CM; -# LB 10 -$CM* $OP $SP* $CM* $QU; # LB 11 -$CM* $NS $SP* $CM* $CL; +$CM* $WJ $CM* $CAN_CM; +$CM* $WJ [$LB8NonBreaks-$CM]; -# LB 11a -($CM* $B2)+; + $CANT_CM $CM* $WJ; +$CM* $CAN_CM $CM* $WJ; -# LB 11b -$CM* ($GL | $WJ) $CM* $LB5NonBreaks; -$CM* $LB5NonBreaks $CM* ($GL | $WJ); -. $CM* ($GL | $WJ); +# LB 12a +# [^SP BA HY] x GL +# +$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+ $OP; # TODO: Experiment. Remove. -# LB 14 -$CM* $QU $CM* $LB12NonBreaks; -$CM* $QU $CM+ $SP; -$CM* $LB5NonBreaks $CM* $QU; -# LB 14a -$BackLB14CanBreakAfter = ($CM* [$LB14NonBreaks - $CM]) | ($CM+ $SP); # LB 15 -$CM* ($BA | $HY | $NS) $BackLB14CanBreakAfter; -($CM* ($BA | $HY | $NS))+ $CM+ / $LB5Breaks; -[$CR $LF $BK $NL $ZW] $CM* $BB; -$CM* [^$CB $CR $LF $BK $NL $ZW] $CM* $BB; +$CM* $OP $SP* $CM* $QU; # LB 16 -$CM* $IN $CM* $ALPlus; -# by rule 7c, any otherwise unattached CM behaves as AL -$CM* $IN $CM+ / $LB5Breaks; +$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* $BB; # -$CM* $IN $CM* ($ID | $CM $SP); +# LB21a +[^$CB] $CM* ($HY | $BA) $CM* $HL; + + +# LB 22 +$CM* $IN $CM* ($ALPlus | $HL); +$CM* $IN $CM* $ID; $CM* $IN $CM* $IN; $CM* $IN $CM* $NU; -# $LB 17 -$CM* $PO $CM* ($ID | $CM $SP); -$CM* $NU ($CM* $ALPlus)+; # includes $LB19 -$CM* $NU $CM+ / $LB5Breaks; # Rule 7c +# LB 23 +$CM* $PO $CM* $ID; +$CM* $NU $CM* ($ALPlus | $HL); +$CM* ($ALPlus | $HL) $CM* $NU; -$CM* $ALPlus $CM* $NU; +# LB 24 +$CM* $ID $CM* $PR; +$CM* ($ALPlus | $HL) $CM* $PR; +$CM* ($ALPlus | $HL) $CM* $PO; -# LB 18 -($CM* $PO)? ($CM* $CL)? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* $PR)?; -$CM* $ALPlus $CM* $PR; -$CM* ($ID | $BackHangulSyllable) $CM* $PR; -# LB 19 -$CM* $ALPlus $CM* $ALPlus; -# The $CM* is from rule 7C, and unattached CM is treated as AL -$CM* $ALPlus $CM* $IS; -$CM* $ALPlus $CM+ / $LB5Breaks; +# LB 25 +($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* $CP; -## problem state table can't handle lookahead when it is at the -## start of the string, currently handled in the rbbi code -## todo fix this ## ------------------------------------------------- !!safe_reverse; -# LB 6 -$V+ $L; - -# LB 7 +# LB 9 $CM+ [^$CM $BK $CR $LF $NL $ZW $SP]; $CM+ $SP / .; -# LB 9 +# LB 14 $SP+ $CM* $OP; -# LB 10 +# LB 15 $SP+ $CM* $QU; -# LB 11 -$SP+ $CM* $CL; - -# LB 18 -($CM* ($IS | $SY))+ $CM* $NU; -$CL $CM* ($NU | $IS | $SY); +# LB 16 +$SP+ $CM* ($CL | $CP); -## ------------------------------------------------- +# LB 17 +$SP+ $CM* $B2; -!!safe_forward; +# LB 21 +$CM* ($HY | $BA) $CM* $HL; -# LB 6 -$V+ $T; +# LB 25 +($CM* ($IS | $SY))+ $CM* $NU; +($CL | $CP) $CM* ($NU | $IS | $SY); -# LB 7 -[^$BK $CR $LF $NL $ZW $SP] $CM+; -$SP $CM+ / [^$CM]; +# For dictionary-based break +$dictionary $dictionary; -# LB 9 -$OP $CM* $SP+; +## ------------------------------------------------- -# LB 10 -$QU $CM* $SP+; +!!safe_forward; -# LB 11 -$CL $CM* $SP+; +# Skip forward over all character classes that are involved in +# rules containing patterns with possibly more than one char +# of context. +# +# It might be slightly more efficient to have specific rules +# instead of one generic one, but only if we could +# turn off rule chaining. We don't want to move more +# than necessary. +# +[$CM $OP $QU $CL $CP $B2 $PR $HY $BA $SP $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $dictionary]; +$dictionary $dictionary; -# LB 18 -$CM* $PRcm? ($OPcm | $HYcm)? $NU;