/* * Copyright (c) 1999, 2007, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. Oracle designates this * particular file as subject to the "Classpath" exception as provided * by Oracle in the LICENSE file that accompanied this code. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. */ /* */ /* * Licensed Materials - Property of IBM * * (C) Copyright IBM Corp. 1999 All Rights Reserved. * (C) IBM Corp. 1997-1998. All Rights Reserved. * * The program is provided "as is" without any warranty express or * implied, including the warranty of non-infringement and the implied * warranties of merchantibility and fitness for a particular purpose. * IBM will not be liable for any damages suffered by you as a result * of using the Program. In no event will IBM be liable for any * special, indirect or consequential damages or lost profits even if * IBM has been advised of the possibility of their occurrence. IBM * will not be liable for any third party claims against you. */ package sun.text.resources; import java.util.ListResourceBundle; /** * Default break-iterator rules. These rules are more or less general for * all locales, although there are probably a few we're missing. The * behavior currently mimics the behavior of BreakIterator in JDK 1.2. * There are known deficiencies in this behavior, including the fact that * the logic for handling CJK characters works for Japanese but not for * Chinese, and that we don't currently have an appropriate locale for * Thai. The resources will eventually be updated to fix these problems. */ /* Modified for Hindi 3/1/99. */ /* * Since JDK 1.5.0, this file no longer goes to runtime and is used at J2SE * build phase in order to create [Character|Word|Line|Sentence]BreakIteratorData * files which are used on runtime instead. */ public class BreakIteratorRules extends ListResourceBundle { protected final Object[][] getContents() { return new Object[][] { // rules describing how to break between logical characters { "CharacterBreakRules", // ignore non-spacing marks and enclosing marks (since we never // put a break before ignore characters, this keeps combining // accents with the base characters they modify) "=[:Mn::Me:];" // other category definitions + "=[\u1100-\u115f];" + "=[\u1160-\u11a7];" + "=[\u11a8-\u11ff];" + "=[\ud800-\udbff];" + "=[\udc00-\udfff];" // break after every character, except as follows: + ".;" // keep base and combining characters togethers + "=[^^[:Cc::Cf::Zl::Zp:]];" + "*;" // keep CRLF sequences together + "\r\n;" // keep surrogate pairs together + ";" // keep Hangul syllables spelled out using conjoining jamo together + "***;" // various additions for Hindi support + "=[\u093c];" + "=[\u0964\u0965];" + "=[\u094d];" + "=[\u093e-\u094c\u0962\u0963];" + "=[\u0915-\u0939];" + "=[\u0958-\u095f];" + "=[\u0902\u0903\u0951-\u0954];" + "=({});" + "=(|);" + "=[\u200d];" + "=({{}});" + "{}{};" + ";" }, // default rules for finding word boundaries { "WordBreakRules", // ignore non-spacing marks, enclosing marks, and format characters, // all of which should not influence the algorithm //"=[:Mn::Me::Cf:];" "=[:Cf:];" + "=[:Mn::Me:];" // Hindi phrase separator, kanji, katakana, hiragana, CJK diacriticals, // other letters, and digits + "=[\u0964\u0965];" + "=[\u3005\u4e00-\u9fa5\uf900-\ufa2d];" + "=[\u30a1-\u30fa\u30fd\u30fe];" + "=[\u3041-\u3094\u309d\u309e];" + "=[\u3099-\u309c\u30fb\u30fc];" + "=[:L::Mc:^[]];" + "=(*);" + "=[:N:];" + "=(*);" // punctuation that can occur in the middle of a word: currently // dashes, apostrophes, quotation marks, and periods + "=[:Pd::Pc:\u00ad\u2027\\\"\\\'\\.];" // punctuation that can occur in the middle of a number: currently // apostrophes, qoutation marks, periods, commas, and the Arabic // decimal point + "=[\\\"\\\'\\,\u066b\\.];" // punctuation that can occur at the beginning of a number: currently // the period, the number sign, and all currency symbols except the cents sign + "=[:Sc:\\#\\.^\u00a2];" // punctuation that can occur at the end of a number: currently // the percent, per-thousand, per-ten-thousand, and Arabic percent // signs, the cents sign, and the ampersand + "=[\\%\\&\u00a2\u066a\u2030\u2031];" // line separators: currently LF, FF, PS, and LS + "=[\n\u000c\u2028\u2029];" // whitespace: all space separators and the tab character + "=[:Zs:\t];" + "=(*);" // a word is a sequence of letters that may contain internal // punctuation, as long as it begins and ends with a letter and // never contains two punctuation marks in a row + "=((*(*)*){});" // a number is a sequence of digits that may contain internal // punctuation, as long as it begins and ends with a digit and // never contains two punctuation marks in a row. + "=(*(*)*);" // break after every character, with the following exceptions // (this will cause punctuation marks that aren't considered // part of words or numbers to be treated as words unto themselves) + ".;" // keep together any sequence of contiguous words and numbers // (including just one of either), plus an optional trailing // number-suffix character + "{}()*{{}};" // keep together and sequence of contiguous words and numbers // that starts with a number-prefix character and a number, // and may end with a number-suffix character + "()*{{}};" // keep together runs of whitespace (optionally with a single trailing // line separator or CRLF sequence) + "*{\r}{};" // keep together runs of Katakana and CJK diacritical marks + "[]*;" // keep together runs of Hiragana and CJK diacritical marks + "[]*;" // keep together runs of Kanji + "*;" // keep together anything else and an enclosing mark + "=[^^[:Cc::Cf::Zl::Zp:]];" + "*;" }, // default rules for determining legal line-breaking positions { "LineBreakRules", // characters that always cause a break: ETX, tab, LF, FF, LS, and PS "=[\u0003\t\n\f\u2028\u2029];" // ignore format characters and control characters EXCEPT for breaking chars + "=[:Cf:[:Cc:^[\r]]];" // enclosing marks + "=[:Mn::Me:];" // Hindi phrase separators + "=[\u0964\u0965];" // characters that always prevent a break: the non-breaking space // and similar characters + "=[\u00a0\u0f0c\u2007\u2011\u202f\ufeff];" // whitespace: space separators and control characters, except for // CR and the other characters mentioned above + "=[:Zs::Cc:^[\r]];" // dashes: dash punctuation and the discretionary hyphen, except for // non-breaking hyphens + "=[:Pd:\u00ad^];" // characters that stick to a word if they precede it: currency symbols // (except the cents sign) and starting punctuation + "=[:Sc::Ps::Pi:^[\u00a2]\\\"\\\'];" // characters that stick to a word if they follow it: ending punctuation, // other punctuation that usually occurs at the end of a sentence, // small Kana characters, some CJK diacritics, etc. + "=[\\\":Pe::Pf:\\!\\%\\.\\,\\:\\;\\?\u00a2\u00b0\u066a\u2030-\u2034\u2103" + "\u2105\u2109\u3001\u3002\u3005\u3041\u3043\u3045\u3047\u3049\u3063" + "\u3083\u3085\u3087\u308e\u3099-\u309e\u30a1\u30a3\u30a5\u30a7\u30a9" + "\u30c3\u30e3\u30e5\u30e7\u30ee\u30f5\u30f6\u30fc-\u30fe\uff01\uff05" + "\uff0c\uff0e\uff1a\uff1b\uff1f];" // Kanji: actually includes Kanji,Kana and Hangul syllables, // except for small Kana and CJK diacritics + "=[\u4e00-\u9fa5\uac00-\ud7a3\uf900-\ufa2d\ufa30-\ufa6a\u3041-\u3094\u30a1-\u30fa^[]];" // digits + "=[:Nd::No:];" // punctuation that can occur in the middle of a number: periods and commas + "=[\\.\\,];" // everything not mentioned above + "=[^[\r]];" // a "number" is a run of prefix characters and dashes, followed by one or // more digits with isolated number-punctuation characters interspersed + "=([]**(*)*);" // the basic core of a word can be either a "number" as defined above, a single // "Kanji" character, or a run of any number of not-explicitly-mentioned // characters (this includes Latin letters) + "=(*||);" // a word may end with an optional suffix that be either a run of one or // more dashes or a run of word-suffix characters + "=((*|*));" // a word, thus, is an optional run of word-prefix characters, followed by // a word core and a word suffix (the syntax of and // actually allows either of them to match the empty string, putting a break // between things like ")(" or "aaa(aaa" + "=(*);" + "=[\\(];" + "=[\\)];" + "=[\\$\\'];" // finally, the rule that does the work: Keep together any run of words that // are joined by runs of one of more non-spacing mark. Also keep a trailing // line-break character or CRLF combination with the word. (line separators // "win" over nbsp's) + "(((**{})|))**{*}{*}{*}{\r}{};" + "\r;" }, // default rules for finding sentence boundaries { "SentenceBreakRules", // ignore non-spacing marks, enclosing marks, and format characters "=[:Mn::Me::Cf:];" // letters + "=[:L:];" // lowercase letters + "=[:Ll:];" // uppercase letters + "=[:Lu:];" // NOT lowercase letters + "=[^];" // whitespace (line separators are treated as whitespace) + "=[\t\r\f\n\u2028:Zs:];" // punctuation which may occur at the beginning of a sentence: "starting // punctuation" and quotation marks + "=[:Ps::Pi:\\\"\\\'];" // punctuation with may occur at the end of a sentence: "ending punctuation" // and quotation marks + "=[:Pe::Pf:\\\"\\\'];" // digits + "=[:N:];" // characters that unambiguously signal the end of a sentence + "=[\\!\\?\u3002\uff01\uff1f];" // periods, which MAY signal the end of a sentence + "=[\\.\uff0e];" // characters that may occur at the beginning of a sentence: basically anything // not mentioned above (letters and digits are specifically excluded) + "=[^[:L:\u2029]];" // Hindi phrase separator + "=[\u0964\u0965];" // always break sentences after paragraph separators + ".*?{\u2029};" // always break after a danda, if it's followed by whitespace + ".*?*;" // if you see a period, skip over additional periods and ending punctuation // and if the next character is a paragraph separator, break after the // paragraph separator //+ ".*?[]**\u2029;" //+ ".*?[]**\u2029;" // if you see a period, skip over additional periods and ending punctuation, // followed by optional whitespace, followed by optional starting punctuation, // and if the next character is something that can start a sentence // (basically, a capital letter), then put the sentence break between the // whitespace and the opening punctuation + ".*?[]**/;" + ".*?[]**/[][]*;" // if you see a sentence-terminating character, skip over any additional // terminators, periods, or ending punctuation, followed by any whitespace, // followed by a SINGLE optional paragraph separator, and put the break there + ".*?[]**{\u2029};" // The following rules are here to aid in backwards iteration. The automatically // generated backwards state table will rewind to the beginning of the // paragraph all the time (or all the way to the beginning of the document // if the document doesn't use the Unicode PS character) because the only // unambiguous character pairs are those involving paragraph separators. // These specify a few more unambiguous breaking situations. // if you see a sentence-starting character, followed by starting punctuation // (remember, we're iterating backwards), followed by an optional run of // whitespace, followed by an optional run of ending punctuation, followed // by a period, this is a safe place to turn around + "!***;" // if you see a letter or a digit, followed by an optional run of // starting punctuation, followed by an optional run of whitespace, // followed by an optional run of ending punctuation, followed by // a sentence terminator, this is a safe place to turn around + "![]***;" } }; } }