+# Copyright (C) 2016 and later: Unicode, Inc. and others.
+# License & terms of use: http://www.unicode.org/copyright.html
# Copyright (c) 2002-2016 International Business Machines Corporation and
# others. All Rights Reserved.
#
#
# Line Breaking Rules
# Implement default line breaking as defined by
-# Unicode Standard Annex #14 Revision 35 for Unicode 8.0
-# http://www.unicode.org/reports/tr14/
+# Unicode Standard Annex #14 Revision 44 for Unicode 13.0
+# http://www.unicode.org/reports/tr14/, with the following modification:
#
-# Includes the Emoji breaking proposals from Unicode L2/16-011R3.
-# http://www.unicode.org/L2/L2016/16011r3-break-prop-emoji.pdf
-#
-# tailored as noted in 2nd paragraph below.
-#
-# 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.
+# Boundaries between hyphens and following letters are suppressed when
+# there is a boundary preceding the hyphen. See rule 20.9
#
# This tailors the line break behavior to correspond to CSS
-# line-break=normal (BCP47 -u-lb-normal) as defined for languages other than
+# line-break=normal (BCP47 -u-lb-normal) as defined for languages other than
# Chinese & Japanese.
# It sets characters of class CJ to behave like ID.
#
!!chain;
+!!quoted_literals_only;
-!!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.
-#
-
-# Temporary definitions of Emoji Base and Emoji Modifiers, until properties are available.
-
-$EB = [\u261D\u26F9\u270A-\u270D\U0001F385\U0001F3C3-\U0001F3C4\U0001F3CA-\U0001F3CB\U0001F442-\U0001F443\U0001F446-\U0001F450\U0001F466-\U0001F469\U0001F46E\U0001F470-\U0001F478\U0001F47C\U0001F481-\U0001F483\U0001F485-\U0001F487\U0001F4AA\U0001F575\U0001F590\U0001F595-\U0001F596\U0001F645-\U0001F647\U0001F64B-\U0001F64F\U0001F6A3\U0001F6B4-\U0001F6B6\U0001F6C0\U0001F918];
-$EM = [\U0001F3FB-\U0001F3FF];
-
-$AI = [[:LineBreak = Ambiguous:] - [\u2640\u2642]];
-$AL = [[:LineBreak = Alphabetic:] - [$EM\u2764]];
+$AI = [:LineBreak = Ambiguous:];
+$AL = [:LineBreak = Alphabetic:];
$BA = [:LineBreak = Break_After:];
+$HH = [\u2010]; # \u2010 is HYPHEN, default line break is BA.
$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:] \u200d];
+# $CM = [:LineBreak = Combining_Mark:];
$CP = [:LineBreak = Close_Parenthesis:];
$CR = [:LineBreak = Carriage_Return:];
+$EB = [:LineBreak = EB:];
+$EM = [:LineBreak = EM:];
$EX = [:LineBreak = Exclamation:];
$GL = [:LineBreak = Glue:];
$HL = [:LineBreak = Hebrew_Letter:];
$HY = [:LineBreak = Hyphen:];
$H2 = [:LineBreak = H2:];
$H3 = [:LineBreak = H3:];
-$ID = [[:LineBreak = Ideographic:] $CJ [\u2640\u2642\u2764] - $EB];
+# CSS Normal tailoring: CJ resolves to ID
+$ID = [[:LineBreak = Ideographic:] $CJ];
$IN = [:LineBreak = Inseperable:];
$IS = [:LineBreak = Infix_Numeric:];
$JL = [:LineBreak = JL:];
$WJ = [:LineBreak = Word_Joiner:];
$XX = [:LineBreak = Unknown:];
$ZW = [:LineBreak = ZWSpace:];
-$ZWJ = [\u200d];
+$ZWJ = [:LineBreak = ZWJ:];
+
+# OP30 and CP30 are variants of OP and CP that appear in-line in rule LB30 from UAX 14,
+# without a formal name. Because ICU rules require multiple uses of the expressions,
+# give them a single definition with a name
+
+$OP30 = [$OP - [\p{ea=F}\p{ea=W}\p{ea=H}]];
+$CP30 = [$CP - [\p{ea=F}\p{ea=W}\p{ea=H}]];
+
+# By LB9, a ZWJ also behaves as a CM. Including it in the definition of CM avoids having to explicitly
+# list it in the numerous rules that use CM.
+# By LB1, SA characters with general categor of Mn or Mc also resolve to CM.
+
+$CM = [[:LineBreak = Combining_Mark:] $ZWJ [$SA & [[:Mn:][:Mc:]]]];
+$CMX = [[$CM] - [$ZWJ]];
# Dictionary character set, for triggering language-based break engines. Currently
-# limited to LineBreak=Complex_Context. Note that this set only works in Unicode
-# 5.0 or later as the definition of Complex_Context was corrected to include all
-# characters requiring dictionary break.
+# limited to LineBreak=Complex_Context (SA).
-$dictionary = [:LineBreak = Complex_Context:];
+$dictionary = [$SA];
#
# Rule LB1. By default, treat AI (characters with ambiguous east Asian width),
-# SA (South East Asian: Thai, Lao, Khmer)
+# SA (Dictionary chars, excluding Mn and Mc)
# SG (Unpaired Surrogates)
# XX (Unknown, unassigned)
# as $AL (Alphabetic)
#
-$ALPlus = [$AL $AI $SA $SG $XX];
-
-#
-# Combining Marks. X $CM* behaves as if it were X. Rule LB6.
-#
-$ALcm = $ALPlus $CM*;
-$BAcm = $BA $CM*;
-$BBcm = $BB $CM*;
-$B2cm = $B2 $CM*;
-$CLcm = $CL $CM*;
-$CPcm = $CP $CM*;
-$EXcm = $EX $CM*;
-$GLcm = $GL $CM*;
-$HLcm = $HL $CM*;
-$HYcm = $HY $CM*;
-$H2cm = $H2 $CM*;
-$H3cm = $H3 $CM*;
-$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 $CM*;
+$ALPlus = [$AL $AI $SG $XX [$SA-[[:Mn:][:Mc:]]]];
-## -------------------------------------------------
-!!forward;
-
-#
-# Each class of character can stand by itself as an unbroken token, with trailing combining stuff
-#
-$ALPlus $CM+;
-$BA $CM+;
-$BB $CM+;
-$B2 $CM+;
-$CL $CM+;
-$CP $CM+;
-$EB $CM+;
-$EM $CM+;
-$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.
#
# 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 $AL_FOLLOW_CM];
+$AL_FOLLOW = [$BK $CR $LF $NL $ZW $SP $CL $CP $EX $HL $IS $SY $WJ $GL $OP30 $QU $BA $HY $NS $IN $NU $PR $PO $ALPlus];
#
#
# 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>
+# ZW SP* รท
#
$LB8Breaks = [$LB4Breaks $ZW];
$LB8NonBreaks = [[$LB4NonBreaks] - [$ZW]];
+$ZW $SP* / [^$SP $ZW $LB4Breaks];
-# LB 8a ZWJ x ID Emoji proposal.
+# LB 8a ZWJ x Do not break Emoji ZWJ sequences.
#
-$ZWJ ($ID | $EB | $EM);
+$ZWJ [^$CM];
# 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 $AL
#
# LB 11 Do not break before or after WORD JOINER & related characters.
#
-$CAN_CM $CM* $WJcm;
-$LB8NonBreaks $WJcm;
-^$CM+ $WJcm;
+$CAN_CM $CM* $WJ;
+$LB8NonBreaks $WJ;
+^$CM+ $WJ;
-$WJcm $CANT_CM;
-$WJcm $CAN_CM $CM*;
+$WJ $CM* .;
#
# LB 12 Do not break after NBSP and related characters.
# GL x
#
-$GLcm $CAN_CM $CM*;
-$GLcm $CANT_CM;
+$GL $CM* .;
#
# LB 12a Do not break before NBSP and related characters ...
# [^SP BA HY] x GL
#
-[[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GLcm;
-^$CM+ $GLcm;
+[[$LB8NonBreaks] - [$SP $BA $HY]] $CM* $GL;
+^$CM+ $GL;
-#
-# LB 13 Don't break before ']' or '!' or ';' or '/', even after spaces.
+
+# LB 13 Don't break before ']' or '!' or or '/', even after spaces.
#
$LB8NonBreaks $CL;
$CAN_CM $CM* $CL;
$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+ $SY; # by rule 10, stand-alone CM behaves as AL
#
# LB 14 Do not break after OP, even after spaces
+# Note subtle interaction with "SP IS /" rules in LB14a.
+# This rule consumes the SP, chaining happens on the IS, effectivley overriding the SP IS rules,
+# which is the desired behavior.
+#
+$OP $CM* $SP* .;
+
+$OP $CM* $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL
+ # by rule 8, CM following a SP is stand-alone.
+
+
+# LB 14a Force a break before start of a number with a leading decimal pt, e.g. " .23"
+# Note: would be simpler to express as "$SP / $IS $CM* $NU;", but ICU rules have limitations.
+# See issue ICU-20303
+
+
+$CanFollowIS = [$BK $CR $LF $NL $SP $ZW $WJ $GL $CL $CP $EX $IS $SY $QU $BA $HY $NS $ALPlus $HL $IN];
+$SP $IS / [^ $CanFollowIS $NU $CM];
+$SP $IS $CM* $CMX / [^ $CanFollowIS $NU $CM];
+
#
-$OPcm $SP* $CAN_CM $CM*;
-$OPcm $SP* $CANT_CM;
+# LB 14b Do not break before numeric separators (IS), even after spaces.
+
+[$LB8NonBreaks - $SP] $IS;
+$SP $IS $CM* [$CanFollowIS {eof}];
+$SP $IS $CM* $ZWJ [^$CM $NU];
+
+$CAN_CM $CM* $IS;
+^$CM+ $IS; # by rule 10, stand-alone CM behaves as AL
-$OPcm $SP+ $CM+ $AL_FOLLOW?; # by rule 10, stand-alone CM behaves as AL
# LB 15
-$QUcm $SP* $OPcm;
+$QU $CM* $SP* $OP;
# LB 16
-($CLcm | $CPcm) $SP* $NScm;
+($CL | $CP) $CM* $SP* $NS;
# LB 17
-$B2cm $SP* $B2cm;
+$B2 $CM* $SP* $B2;
#
# LB 18 Break after spaces.
# LB 19
# x QU
-$LB18NonBreaks $CM* $QUcm;
-^$CM+ $QUcm;
+$LB18NonBreaks $CM* $QU;
+^$CM+ $QU;
# QU x
-$QUcm .?;
-
+$QU $CM* .;
# LB 20
# <break> $CB
# $CB <break>
-
+#
$LB20NonBreaks = [$LB18NonBreaks - $CB];
+# LB 20.09 Don't break between Hyphens and Letters when there is a break preceding the hyphen.
+# Originally added as a Finnish tailoring, now promoted to default ICU behavior.
+# Note: this is not default UAX-14 behaviour. See issue ICU-8151.
+#
+^($HY | $HH) $CM* $ALPlus;
+
# LB 21 x (BA | HY | NS)
# BB x
#
-$LB20NonBreaks $CM* ($BAcm | $HYcm | $NScm);
-^$CM+ ($BAcm | $HYcm | $NScm);
+$LB20NonBreaks $CM* ($BA | $HY | $NS);
+
+
+^$CM+ ($BA | $HY | $NS);
-$BBcm [^$CB]; # $BB x
-$BBcm $LB20NonBreaks $CM*;
+$BB $CM* [^$CB]; # $BB x
+$BB $CM* $LB20NonBreaks;
# LB 21a Don't break after Hebrew + Hyphen
# HL (HY | BA) x
#
-$HLcm ($HYcm | $BAcm) [^$CB]?;
+$HL $CM* ($HY | $BA) $CM* [^$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+ $INcm; # by rule 10, any otherwise unattached CM behaves as AL
-$EXcm $INcm;
-($ID | $EB | $EM) $CM* $INcm;
-$INcm $INcm;
-$NUcm $INcm;
-
-
-# $LB 23
-($ID | $EB | $EM) $CM* $POcm;
-$ALcm $NUcm; # includes $LB19
-$HLcm $NUcm;
-^$CM+ $NUcm; # Rule 10, any otherwise unattached CM behaves as AL
-$NUcm $ALcm;
-$NUcm $HLcm;
+$SY $CM* $HL;
+# LB 22 Do not break before ellipses
#
-# LB 24
-#
-$PRcm ($ID | $EB | $EM);
-$PRcm ($ALcm | $HLcm);
-$POcm ($ALcm | $HLcm);
-($ALcm | $HLcm) ($PRcm | $POcm);
-^$CM+ ($PRcm | $POcm); # Rule 10, any otherwise unattached CM behaves as AL
-
-#
-# LB 25 Numbers.
-#
-($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 | $NUcm);
-
-# LB 30a Do not break between regional indicators. Break after pairs of them.
-# Tricky interaction with LB8a: ZWJ x ID
-$RIcm $RI / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
-$RIcm $RI $CM* $ZWJ / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM $ID $EB $EM] {eof}];
-$RIcm $RI $CM* [$CM-$ZWJ] / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $CM] {eof}];
-
-$RIcm $RIcm [$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS {eof}];
-$RIcm $RIcm $ZWJ ($ID | $EB | $EM);
-
-# LB 30b Do not break between an Emoji Base and an Emoji Modifier
-$EB $CM* $EM;
-
-#
-# Reverse Rules.
-#
-## -------------------------------------------------
-
-!!reverse;
-
-^$CM+ $ALPlus;
-^$CM+ $BA;
-^$CM+ $BB;
-^$CM+ $B2;
-^$CM+ $CL;
-^$CM+ $CP;
-^$CM+ $EB;
-^$CM+ $EM;
-^$CM+ $EX;
-^$CM+ $GL;
-^$CM+ $HL;
-^$CM+ $HY;
-^$CM+ $H2;
-^$CM+ $H3;
-^$CM+ $ID;
+$LB20NonBreaks $CM* $IN;
^$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.
+# LB 23
#
-# 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.
-#
-
-# Apple early addition, remove this, superseded by LB24
-# [$PR ] / $CM+ [$BK $CR $LF $NL $ZW $SP {eof}];
-
-
-
-# LB 4, 5, 6
-
-$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* $ZW
+($ALPlus | $HL) $CM* $NU;
+^$CM+ $NU; # Rule 10, any otherwise unattached CM behaves as AL
+$NU $CM* ($ALPlus | $HL);
-# LB 8a ZWJ x ID Unicode Emoji proposal L2/16-011R3
-# The ZWJ will look like a CM to whatever precedes it.
+# LB 23a
#
-($ID | $EB | $EM) $ZWJ $CM* $CAN_CM?;
-
-
-# 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+ $CAN_CM;
+$PR $CM* ($ID | $EB | $EM);
+($ID | $EB | $EM) $CM* $PO;
-# LB 11
#
-$WJ $CM* $CAN_CM;
-$WJ [$LB8NonBreaks-$CM];
-
- $CANT_CM $CM* $WJ;
-$CAN_CM $CM* $WJ;
-
-# LB 12a
-# [^SP BA HY] x GL
+# LB 24
#
-$GL $CM* [$LB8NonBreaks-[$CM $SP $BA $HY]];
+($PR | $PO) $CM* ($ALPlus | $HL);
+($ALPlus | $HL) $CM* ($PR | $PO);
+^$CM+ ($PR | $PO); # Rule 10, any otherwise unattached CM behaves as AL
-# LB 12
-# GL x
#
-$CANT_CM $CM* $GL;
-$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* $OP;
-
-# LB 14 OP SP* x
+# LB 25 Numbers.
#
-$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* $OP;
-$AL_FOLLOW_CM $CM+ $SP+ $CM* $OP;
-
-
-# LB 15
-$OP $SP* $CM* $QU;
-
-# LB 16
-$NS $SP* $CM* ($CL | $CP);
+(($PR | $PO) $CM*)? (($OP | $HY) $CM*)? ($IS $CM*)? $NU ($CM* ($NU | $SY | $IS))*
+ ($CM* ($CL | $CP))? ($CM* ($PR | $PO))?;
-# LB 17
-$B2 $SP* $CM* $B2;
-
-# LB 18 break after spaces
-# Nothing explicit needed here.
-
-
-#
-# LB 19
+# LB 26 Do not break a Korean syllable
#
-$QU $CM* $CAN_CM; # . x QU
-$QU $LB18NonBreaks;
+$JL $CM* ($JL | $JV | $H2 | $H3);
+($JV | $H2) $CM* ($JV | $JT);
+($JT | $H3) $CM* $JT;
+# LB 27 Treat korean Syllable Block the same as ID (don't break it)
+($JL | $JV | $JT | $H2 | $H3) $CM* $IN;
+($JL | $JV | $JT | $H2 | $H3) $CM* $PO;
+$PR $CM* ($JL | $JV | $JT | $H2 | $H3);
-$CAN_CM $CM* $QU; # QU x .
- $CANT_CM $CM* $QU;
+# LB 28 Do not break between alphabetics
#
-# LB 20 Break before and after CB.
-# nothing needed here.
-#
-
-# LB 21
-($BA | $HY | $NS) $CM* [$LB20NonBreaks-$CM]; # . x (BA | HY | NS)
-
-[$LB20NonBreaks-$CM] $CM* $BB; # BB x .
-[^$CB] $CM* $BB; #
-
-# LB21a
-[^$CB] $CM* ($HY | $BA) $CM* $HL;
-
-# LB21b (reverse)
-$HL $CM* $SY;
-
-# LB 22
-$IN $CM* ($ALPlus | $HL);
-$IN $CM* $EX;
-$IN $CM* ($ID | $EB | $EM);
-$IN $CM* $IN;
-$IN $CM* $NU;
-
-# LB 23
-$PO $CM* ($ID | $EB | $EM);
-$NU $CM* ($ALPlus | $HL);
-($ALPlus | $HL) $CM* $NU;
-
-# LB 24
-($ID | $EB | $EM) $CM* $PR;
-($ALPlus | $HL) $CM* $PR;
-($ALPlus | $HL) $CM* $PO;
-$CM* ($PR | $PO) $CM* ($ALPlus | $HL);
-$CM* ($PR | $PO) $CM+ / [$BK $CR $LF $NL $ZW $SP {eof}];
-
-# LB 25
-($CM* ($PR | $PO))? ($CM* ($CL | $CP))? ($CM* ($NU | $IS | $SY))* $CM* $NU ($CM* ($OP | $HY))? ($CM* ($PR | $PO))?;
-
-# LB 26
-($H3 | $H2 | $JV | $JL) $CM* $JL;
-($JT | $JV) $CM* ($H2 | $JV);
-$JT $CM* ($H3 | $JT);
-
-# LB 27
-$IN $CM* ($H3 | $H2 | $JT | $JV | $JL);
-$PO $CM* ($H3 | $H2 | $JT | $JV | $JL);
- ($H3 | $H2 | $JT | $JV | $JL) $CM* $PR;
-
-# LB 28
($ALPlus | $HL) $CM* ($ALPlus | $HL);
-
+^$CM+ ($ALPlus | $HL); # The $CM+ is from rule 10, an unattached CM is treated as AL
# LB 29
-($ALPlus | $HL) $CM* $IS;
+$IS $CM* ($ALPlus | $HL);
# LB 30
-$OP $CM* ($ALPlus | $HL | $NU);
-($ALPlus | $HL | $NU) $CM* $CP;
-
-# LB 30a
-# Pairs of Regional Indicators.
-# The following two rules are nearly identical. The first matches only sequences with an odd number of adjacent RIs,
-# the second with an even number. Stripping away the cruft they look like
-# [^RI] RI / (RI RI)+ ^RI;
-# [^RI] RI RI / (RI RI)+ ^RI;
-#
-[{bof} $NS $HY $BA $QU $CL $CP $EX $IS $SY $WJ $GL $ZW $SP $BK $CR $LF $NL $ZWJ] $CM* $RI / ($CM* $RI $CM* $RI)+ $CM* [{eof}[^$RI $CM]];
-[{bof} $NS $HY $BA $QU $CL $CP $EX $IS $SY $WJ $GL $ZW $SP $BK $CR $LF $NL $ZWJ] $CM* $RI $CM* $RI / ($CM* $RI $CM* $RI)+ $CM* [{eof}[^$RI $CM]];
-
-# In general, adjacent RIs stay together. The hard-break rules, above, overide this, forcing in the boundaries between pairs.
-$RI $CM* $RI;
-
-# WJ, GL, QU, etc. are classes with rules like "WJ x " which includes "WJ x RI".
-$RI $CM* ([$WJ $GL $QU $BB] | (($HY | $BA)$CM* $HL));
+($ALPlus | $HL | $NU) $CM* $OP30;
+^$CM+ $OP30; # The $CM+ is from rule 10, an unattached CM is treated as AL.
+$CP30 $CM* ($ALPlus | $HL | $NU);
+# LB 30a Do not break between regional indicators. Break after pairs of them.
+# Tricky interaction with LB8a: ZWJ x . together with ZWJ acting like a CM.
+$RI $CM* $RI / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $IN $CM]];
+$RI $CM* $RI $CM* [$CM-$ZWJ] / [[^$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $IN $CM]];
+$RI $CM* $RI $CM* [$BK $CR $LF $NL $SP $ZW $WJ $CL $CP $EX $IS $SY $GL $QU $BA $HY $NS $IN $ZWJ {eof}];
+# note: the preceding rule includes {eof} rather than having the last [set] term qualified with '?'
+# because of the chain-out behavior difference. The rule must chain out only from the [set characters],
+# not from the preceding $RI or $CM, which it would be able to do if the set were optional.
# LB 30b Do not break between an Emoji Base and an Emoji Modifier
-$EM $CM* $EB;
-
-
-## -------------------------------------------------
-
-!!safe_reverse;
-
-# LB 9
-^$CM+ [^$CM $BK $CR $LF $NL $ZW $SP];
-^$CM+ $SP / .;
-
-# LB 14
-$SP+ $CM* $OP;
-
-# LB 15
-$SP+ $CM* $QU;
-
-# LB 16
-$SP+ $CM* ($CL | $CP);
-
-# LB 17
-$SP+ $CM* $B2;
-
-# LB 21
-$CM* ($HY | $BA) $CM* $HL;
-
-# LB 25
-($CM* ($IS | $SY))+ $CM* $NU;
-($CL | $CP) $CM* ($NU | $IS | $SY);
-
-# LB 30
-($CM* $RI)+;
-
-# For dictionary-based break
-$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 $RI $ZWJ $dictionary]+ [^$CM $OP $QU $CL $CP $B2 $PR $HY $BA $RI $ZWJ $dictionary];
-$dictionary $dictionary;
+$EB $CM* $EM;
+# LB 31 Break everywhere else.
+# Match a single code point if no other rule applies.
+.;
projects / apple / icu.git / blobdiff