1# 2# Copyright (C) 2002-2006, International Business Machines Corporation and others. 3# All Rights Reserved. 4# 5# file: sent.txt 6# 7# ICU Sentence Break Rules 8# See Unicode Standard Annex #29. 9# These rules are based on SA 29 version 5.0.0 10# Includes post 5.0 changes to treat Japanese half width voicing marks 11# as Grapheme Extend. 12# 13 14 15$VoiceMarks = [\uff9e\uff9f]; 16$Thai = [:Script = Thai:]; 17 18# 19# Character categories as defined in TR 29 20# 21$Sep = [\p{Sentence_Break = Sep}]; 22$Format = [\p{Sentence_Break = Format}]; 23$Sp = [\p{Sentence_Break = Sp}]; 24$Lower = [\p{Sentence_Break = Lower}]; 25$Upper = [\p{Sentence_Break = Upper}]; 26$OLetter = [\p{Sentence_Break = OLetter}-$VoiceMarks]; 27$Numeric = [\p{Sentence_Break = Numeric}]; 28$ATerm = [\p{Sentence_Break = ATerm}]; 29$STerm = [\p{Sentence_Break = STerm}]; 30$Close = [\p{Sentence_Break = Close}]; 31 32# 33# Define extended forms of the character classes, 34# incorporate grapheme cluster + format chars. 35# Rules 4 and 5. 36 37 38$CR = \u000d; 39$LF = \u000a; 40$Extend = [[:Grapheme_Extend = TRUE:]$VoiceMarks]; 41 42$SpEx = $Sp ($Extend | $Format)*; 43$LowerEx = $Lower ($Extend | $Format)*; 44$UpperEx = $Upper ($Extend | $Format)*; 45$OLetterEx = $OLetter ($Extend | $Format)*; 46$NumericEx = $Numeric ($Extend | $Format)*; 47$ATermEx = $ATerm ($Extend | $Format)*; 48$STermEx = $STerm ($Extend | $Format)*; 49$CloseEx = $Close ($Extend | $Format)*; 50 51 52## ------------------------------------------------- 53 54!!chain; 55!!forward; 56 57# Rule 3 - break after separators. Keep CR/LF together. 58# 59$CR $LF; 60 61$LettersEx = [$OLetter $Upper $Lower $Numeric $Close $STerm] ($Extend | $Format)*; 62$LettersEx* $Thai $LettersEx* ($ATermEx | $SpEx)*; 63 64# Rule 4 - Break after $Sep. 65# Rule 5 - Ignore $Format and $Extend 66# 67[^$Sep]? ($Extend | $Format)*; 68 69 70# Rule 6 71$ATermEx $NumericEx; 72 73# Rule 7 74$UpperEx $ATermEx $UpperEx; 75 76#Rule 8 77# Note: follows errata for Unicode 5.0 boundary rules. 78$NotLettersEx = [^$OLetter $Upper $Lower $Sep $ATerm $STerm] ($Extend | $Format)*; 79$ATermEx $CloseEx* $SpEx* $NotLettersEx* $Lower; 80 81# Rule 8a 82($STermEx | $ATermEx) $CloseEx* $SpEx* ($STermEx | $ATermEx); 83 84#Rule 9, 10, 11 85($STermEx | $ATermEx) $CloseEx* $SpEx* $Sep?; 86 87#Rule 12 88[[^$STerm $ATerm $Close $Sp $Sep $Format $Extend $Thai]{bof}] ($Extend | $Format | $Close | $Sp)* [^$Thai]; 89[[^$STerm $ATerm $Close $Sp $Sep $Format $Extend]{bof}] ($Extend | $Format | $Close | $Sp)* ([$Sep{eof}] | $CR $LF){100}; 90 91## ------------------------------------------------- 92 93!!reverse; 94 95$SpEx_R = ($Extend | $Format)* $Sp; 96$ATermEx_R = ($Extend | $Format)* $ATerm; 97$STermEx_R = ($Extend | $Format)* $STerm; 98$CloseEx_R = ($Extend | $Format)* $Close; 99 100# 101# Reverse rules. 102# For now, use the old style inexact reverse rules, which are easier 103# to write, but less efficient. 104# TODO: exact reverse rules. It appears that exact reverse rules 105# may require improving support for look-ahead breaks in the 106# builder. Needs more investigation. 107# 108 109[{bof}] (.? | $LF $CR) [^$Sep]* [$Sep {eof}] ($SpEx_R* $CloseEx_R* ($STermEx_R | $ATermEx_R))*; 110#.*; 111 112# Explanation for this rule: 113# 114# It needs to back over 115# The $Sep at which we probably begin 116# All of the non $Sep chars leading to the preceding $Sep 117# The preceding $Sep, which will be the second one that the rule matches. 118# Any immediately preceding STerm or ATerm sequences. We need to see these 119# to get the correct rule status when moving forwards again. 120# 121# [{bof}] inhibit rule chaining. Without this, rule would loop on itself and match 122# the entire string. 123# 124# (.? | $LF $CR) Match one $Sep instance. Use .? rather than $Sep because position might be 125# at the beginning of the string at this point, and we don't want to fail. 126# Can only use {eof} once, and it is used later. 127# 128 129