569 lines
14 KiB
Plaintext
569 lines
14 KiB
Plaintext
# Copyright (c) 2002-2005 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.
|
|
#
|
|
|
|
|
|
#
|
|
# 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 the incredibly annoying rule LB 7c
|
|
# 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 interating forwards, these sequences do not present any problems
|
|
# When interating 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:];
|
|
$CL = [:LineBreak = Close_Punctuation:];
|
|
$CM = [:LineBreak = Combining_Mark:];
|
|
$CR = [:LineBreak = Carriage_Return:];
|
|
$EX = [:LineBreak = Exclamation:];
|
|
$GL = [:LineBreak = Glue:];
|
|
$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:];
|
|
$NU = [:LineBreak = Numeric:];
|
|
$OP = [:LineBreak = Open_Punctuation:];
|
|
$PO = [:LineBreak = Postfix_Numeric:];
|
|
$PR = [:LineBreak = Prefix_Numeric:];
|
|
$QU = [:LineBreak = Quotation:];
|
|
$SA = [:LineBreak = Complex_Context:];
|
|
$SG = [:LineBreak = Surrogate:];
|
|
$SP = [:LineBreak = Space:];
|
|
$SY = [:LineBreak = Break_Symbols:];
|
|
$WJ = [:LineBreak = Word_Joiner:];
|
|
$XX = [:LineBreak = Unknown:];
|
|
$ZW = [:LineBreak = ZWSpace:];
|
|
|
|
|
|
#
|
|
# 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*;
|
|
$EXcm = $EX $CM*;
|
|
$GLcm = $GL $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*;
|
|
$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+;
|
|
$EX $CM+;
|
|
$GL $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+;
|
|
$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.
|
|
#
|
|
$CAN_CM = [^$BK $CR $LF $NL $ZW $SP $CM]; # Bases that can take CMs
|
|
$CANT_CM = [ $BK $CR $LF $NL $ZW $SP $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 $EX $IS $SY $WJ $GL $QU $BA $HY $NS $IN $NU $ALPlus];
|
|
$AL_FOLLOW = [$AL_FOLLOW_NOCM $AL_FOLLOW_CM];
|
|
|
|
|
|
#
|
|
# Rule LB 3 Mandatory (Hard) breaks.
|
|
#
|
|
$LB3Breaks = [$BK $CR $LF $NL];
|
|
$LB3NonBreaks = [^$BK $CR $LF $NL];
|
|
|
|
$LB3NonBreaks? $LB3Breaks {100}; # LB 3c do not break before hard breaks.
|
|
$CAN_CM $CM* $LB3Breaks {100};
|
|
$CM+ $LB3Breaks {100};
|
|
$CR $LF {100};
|
|
|
|
# LB 4 x SP
|
|
# x ZW
|
|
$LB3NonBreaks [$SP $ZW];
|
|
$CAN_CM $CM* [$SP $ZW];
|
|
$CM+ [$SP $ZW];
|
|
|
|
# LB 5 Break after zero width space
|
|
$LB5Breaks = [$LB3Breaks $ZW];
|
|
$LB5NonBreaks = [[$LB3NonBreaks] - [$ZW]];
|
|
|
|
|
|
# LB 7 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 8
|
|
#
|
|
$LB5NonBreaks $CL;
|
|
$CAN_CM $CM* $CL;
|
|
$CM+ $CL; # by rule 7c, stand-alone CM behaves as AL
|
|
|
|
$LB5NonBreaks $EX;
|
|
$CAN_CM $CM* $EX;
|
|
$CM+ $EX; # by rule 7c, stand-alone CM behaves as AL
|
|
|
|
$LB5NonBreaks $IS;
|
|
$CAN_CM $CM* $IS;
|
|
$CM+ $IS; # by rule 7c, stand-alone CM behaves as AL
|
|
|
|
$LB5NonBreaks $SY;
|
|
$CAN_CM $CM* $SY;
|
|
$CM+ $SY; # by rule 7c, stand-alone CM behaves as AL
|
|
|
|
|
|
#
|
|
# LB 9
|
|
#
|
|
$OPcm $SP* $CAN_CM $CM*;
|
|
$OPcm $SP* $CANT_CM;
|
|
|
|
$OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 7c, stand-alone CM behaves as AL
|
|
|
|
# LB 10
|
|
$QUcm $SP* $OPcm;
|
|
|
|
# LB 11
|
|
$CLcm $SP* $NScm;
|
|
|
|
# LB 11a
|
|
$B2cm $SP* $B2cm;
|
|
|
|
# LB 11b Word Joiner
|
|
#
|
|
$CAN_CM $CM* $WJcm;
|
|
$LB5NonBreaks $WJcm;
|
|
$CM+ $WJcm;
|
|
|
|
$WJcm [^$CAN_CM];
|
|
$WJcm $CAN_CM $CM*;
|
|
|
|
# LB 12
|
|
$LB12NonBreaks = [$LB5NonBreaks - [$SP]];
|
|
$LB12Breaks = [$LB5Breaks $SP];
|
|
|
|
# LB 13
|
|
# x GL
|
|
$LB12NonBreaks $CM* $GLcm;
|
|
$CM+ $GLcm;
|
|
|
|
|
|
# GL x
|
|
#
|
|
$GLcm $LB12Breaks;
|
|
$GLcm $LB12NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc.
|
|
# TODO: I don't think we need this rule.
|
|
# All but $CM will chain off of preceding rule.
|
|
# $GLcm will pick up the CM case by itself.
|
|
|
|
# LB 14
|
|
# x QU
|
|
$LB12NonBreaks $CM* $QUcm;
|
|
$CM+ $QUcm;
|
|
|
|
# QU x
|
|
$QUcm .?;
|
|
$QUcm $LB12NonBreaks $CM*; # Don't let a combining mark go onto $CR, $BK, etc.
|
|
# TODO: I don't think this rule is needed.
|
|
|
|
|
|
# LB 14a
|
|
# <break> $CB
|
|
# $CB <break>
|
|
|
|
$LB14NonBreaks = [$LB12NonBreaks - $CB];
|
|
|
|
# LB 15
|
|
$LB14NonBreaks $CM* ($BAcm | $HYcm | $NScm);
|
|
|
|
$BBcm [^$CB]; # $BB x
|
|
$BBcm $LB14NonBreaks $CM*;
|
|
|
|
# LB 16
|
|
$ALcm $INcm;
|
|
$CM+ $INcm; # by rule 7c, any otherwise unattached CM behaves as AL
|
|
$IDcm $INcm;
|
|
$INcm $INcm;
|
|
$NUcm $INcm;
|
|
|
|
|
|
# $LB 17
|
|
$IDcm $POcm;
|
|
$ALcm $NUcm; # includes $LB19
|
|
$CM+ $NUcm; # Rule 7c, any otherwise unattached CM behaves as AL
|
|
$NUcm $ALcm;
|
|
|
|
# LB 18
|
|
$PRcm? ($OPcm | $HYcm)? $NUcm ($NUcm | $SYcm | $IScm)* $CLcm? $POcm?;
|
|
$PRcm $ALcm;
|
|
$PRcm $IDcm;
|
|
|
|
# LB 18b Do not break a Korean syllable
|
|
$JLcm ($JLcm | $JVcm | $H2cm | $H3cm);
|
|
($JVcm | $H2cm) ($JVcm | $JTcm);
|
|
($JTcm | $H3cm) $JTcm;
|
|
|
|
# LB 18c Treat korean Syllable Block the same as ID (don't break it)
|
|
($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $INcm;
|
|
($JLcm | $JVcm | $JTcm | $H2cm | $H3cm) $POcm;
|
|
$PRcm ($JLcm | $JVcm | $JTcm | $H2cm | $H3cm);
|
|
|
|
|
|
# LB 19
|
|
$ALcm $ALcm;
|
|
$CM+ $ALcm; # The $CM+ is from rule 7C, and unattached CM is treated as AL
|
|
|
|
# LB 19b
|
|
$IScm $ALcm;
|
|
|
|
#
|
|
# Reverse Rules.
|
|
#
|
|
## -------------------------------------------------
|
|
|
|
!!reverse;
|
|
|
|
$CM+ $ALPlus;
|
|
$CM+ $BA;
|
|
$CM+ $BB;
|
|
$CM+ $B2;
|
|
$CM+ $CL;
|
|
$CM+ $EX;
|
|
$CM+ $GL;
|
|
$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+ $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 LB9 will match, need to surpress this break.
|
|
# LB9 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 3
|
|
|
|
$LB3Breaks [$LB3NonBreaks-$CM];
|
|
$LB3Breaks $CM+ $CAN_CM;
|
|
$LF $CR;
|
|
|
|
|
|
# LB 4 x SP
|
|
# x ZW
|
|
[$SP $ZW] [$LB3NonBreaks-$CM];
|
|
[$SP $ZW] $CM+ $CAN_CM;
|
|
|
|
# LB 5 Break after zero width space
|
|
|
|
|
|
# LB 7 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 8
|
|
$CL $CM+ $CAN_CM;
|
|
$EX $CM+ $CAN_CM;
|
|
$IS $CM+ $CAN_CM;
|
|
$SY $CM+ $CAN_CM;
|
|
|
|
$CL [$LB5NonBreaks-$CM];
|
|
$EX [$LB5NonBreaks-$CM];
|
|
$IS [$LB5NonBreaks-$CM];
|
|
$SY [$LB5NonBreaks-$CM];
|
|
|
|
# Rule 9 & 8 together.
|
|
# This really wants to chain at the $CM+ (which is acting as an $AL)
|
|
# except for $CM chaining being disabled.
|
|
[$CL $EX $IS $SY] $CM+ $SP+ $CM* $OP;
|
|
|
|
# LB 9 OP SP* x
|
|
#
|
|
$CM* $CAN_CM $SP* $CM* $OP;
|
|
$CANT_CM $SP* $CM* $OP;
|
|
$AL_FOLLOW? $CM+ $SP $SP* $CM* $OP; # by LB 7, 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; # Experiment. Remove.
|
|
|
|
|
|
|
|
# LB 10
|
|
$CM* $OP $SP* $CM* $QU;
|
|
|
|
# LB 11
|
|
$CM* $NS $SP* $CM* $CL;
|
|
|
|
# LB 11a
|
|
$CM* $B2 $SP* $CM* $B2;
|
|
|
|
# LB 11b
|
|
$CM* $WJ $CM* $CAN_CM;
|
|
$CM* $WJ [$LB5NonBreaks-$CM];
|
|
|
|
$CANT_CM $CM* $WJ;
|
|
$CM* $CAN_CM $CM* $WJ;
|
|
|
|
# LB 12
|
|
|
|
# LB 13
|
|
# x GL
|
|
#
|
|
$CM* $GL $CM* [$LB12NonBreaks-$CM];
|
|
|
|
#
|
|
# GL x
|
|
#
|
|
$CANT_CM $CM* $GL;
|
|
$CM* $CAN_CM $CM* $GL;
|
|
|
|
|
|
|
|
#
|
|
# LB 14
|
|
#
|
|
$CM* $QU $CM* $CAN_CM; # . x QU
|
|
$CM* $QU $LB12NonBreaks;
|
|
|
|
|
|
$CM* $CAN_CM $CM* $QU; # QU x .
|
|
$CANT_CM $CM* $QU;
|
|
|
|
# LB 15
|
|
$CM* ($BA | $HY | $NS) $CM* [$LB14NonBreaks-$CM]; # . x (BA | HY | NS)
|
|
|
|
$CM* [$LB14NonBreaks-$CM] $CM* $BB; # BB x .
|
|
[^$CB] $CM* $BB; #
|
|
|
|
|
|
|
|
# LB 16
|
|
$CM* $IN $CM* $ALPlus;
|
|
$CM* $IN $CM* $ID;
|
|
$CM* $IN $CM* $IN;
|
|
$CM* $IN $CM* $NU;
|
|
|
|
# $LB 17
|
|
$CM* $PO $CM* $ID;
|
|
$CM* $NU $CM* $ALPlus;
|
|
$CM* $ALPlus $CM* $NU;
|
|
|
|
# LB 18
|
|
($CM* $PO)? ($CM* $CL)? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* $PR)?;
|
|
$CM* $ALPlus $CM* $PR;
|
|
$CM* $ID $CM* $PR;
|
|
|
|
# LB 18b
|
|
$CM* ($H3 | $H2 | $JV | $JL) $CM* $JL;
|
|
$CM* ($JT | $JV) $CM* ($H2 | $JV);
|
|
$CM* $JT $CM* ($H3 | $JT);
|
|
|
|
# LB 18c
|
|
$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 19
|
|
$CM* $ALPlus $CM* $ALPlus;
|
|
|
|
|
|
# LB 19b
|
|
$CM* $ALPlus $CM* $IS;
|
|
|
|
|
|
|
|
|
|
## -------------------------------------------------
|
|
|
|
!!safe_reverse;
|
|
|
|
# LB 7
|
|
$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
|
|
$CM+ $SP / .;
|
|
|
|
# LB 9
|
|
$SP+ $CM* $OP;
|
|
|
|
# LB 10
|
|
$SP+ $CM* $QU;
|
|
|
|
# LB 11
|
|
$SP+ $CM* $CL;
|
|
$SP+ $CM* $B2;
|
|
|
|
# LB 18
|
|
($CM* ($IS | $SY))+ $CM* $NU;
|
|
$CL $CM* ($NU | $IS | $SY);
|
|
|
|
## -------------------------------------------------
|
|
|
|
!!safe_forward;
|
|
|
|
# Skip forward over all character classes that are involved in
|
|
# rules containing patterns with possibly more than one char
|
|
# of context.
|
|
#
|
|
# It might be slightly more efficient to have specific rules
|
|
# instead of one generic one, but only if we could
|
|
# turn off rule chaining. We don't want to move more
|
|
# than necessary.
|
|
#
|
|
[$CM $OP $QU $CL $B2 $PR $HY $SP]+ [^$CM $OP $QU $CL $B2 $PR $HY];
|
|
|