1 /* 2 * Copyright (c) 1999, 2019, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. Oracle designates this 8 * particular file as subject to the "Classpath" exception as provided 9 * by Oracle in the LICENSE file that accompanied this code. 10 * 11 * This code is distributed in the hope that it will be useful, but WITHOUT 12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 14 * version 2 for more details (a copy is included in the LICENSE file that 15 * accompanied this code). 16 * 17 * You should have received a copy of the GNU General Public License version 18 * 2 along with this work; if not, write to the Free Software Foundation, 19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 20 * 21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 22 * or visit www.oracle.com if you need additional information or have any 23 * questions. 24 */ 25 26 package java.util.regex; 27 28 import java.text.Normalizer; 29 import java.text.Normalizer.Form; 30 import java.util.Locale; 31 import java.util.Iterator; 32 import java.util.Map; 33 import java.util.ArrayList; 34 import java.util.HashMap; 35 import java.util.LinkedHashSet; 36 import java.util.List; 37 import java.util.Set; 38 import java.util.Arrays; 39 import java.util.NoSuchElementException; 40 import java.util.Spliterator; 41 import java.util.Spliterators; 42 import java.util.function.Predicate; 43 import java.util.stream.Stream; 44 import java.util.stream.StreamSupport; 45 46 import jdk.internal.util.ArraysSupport; 47 48 /** 49 * A compiled representation of a regular expression. 50 * 51 * <p> A regular expression, specified as a string, must first be compiled into 52 * an instance of this class. The resulting pattern can then be used to create 53 * a {@link Matcher} object that can match arbitrary {@linkplain 54 * java.lang.CharSequence character sequences} against the regular 55 * expression. All of the state involved in performing a match resides in the 56 * matcher, so many matchers can share the same pattern. 57 * 58 * <p> A typical invocation sequence is thus 59 * 60 * <blockquote><pre> 61 * Pattern p = Pattern.{@link #compile compile}("a*b"); 62 * Matcher m = p.{@link #matcher matcher}("aaaaab"); 63 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote> 64 * 65 * <p> A {@link #matches matches} method is defined by this class as a 66 * convenience for when a regular expression is used just once. This method 67 * compiles an expression and matches an input sequence against it in a single 68 * invocation. The statement 69 * 70 * <blockquote><pre> 71 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote> 72 * 73 * is equivalent to the three statements above, though for repeated matches it 74 * is less efficient since it does not allow the compiled pattern to be reused. 75 * 76 * <p> Instances of this class are immutable and are safe for use by multiple 77 * concurrent threads. Instances of the {@link Matcher} class are not safe for 78 * such use. 79 * 80 * 81 * <h2><a id="sum">Summary of regular-expression constructs</a></h2> 82 * 83 * <table class="borderless"> 84 * <caption style="display:none">Regular expression constructs, and what they match</caption> 85 * <thead style="text-align:left"> 86 * <tr> 87 * <th id="construct">Construct</th> 88 * <th id="matches">Matches</th> 89 * </tr> 90 * </thead> 91 * <tbody style="text-align:left"> 92 * 93 * <tr><th colspan="2" style="padding-top:20px" id="characters">Characters</th></tr> 94 * 95 * <tr><th style="vertical-align:top; font-weight: normal" id="x"><i>x</i></th> 96 * <td headers="matches characters x">The character <i>x</i></td></tr> 97 * <tr><th style="vertical-align:top; font-weight: normal" id="backslash">{@code \\}</th> 98 * <td headers="matches characters backslash">The backslash character</td></tr> 99 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_n">{@code \0}<i>n</i></th> 100 * <td headers="matches characters octal_n">The character with octal value {@code 0}<i>n</i> 101 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 102 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nn">{@code \0}<i>nn</i></th> 103 * <td headers="matches characters octal_nn">The character with octal value {@code 0}<i>nn</i> 104 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 105 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nnn">{@code \0}<i>mnn</i></th> 106 * <td headers="matches characters octal_nnn">The character with octal value {@code 0}<i>mnn</i> 107 * (0 {@code <=} <i>m</i> {@code <=} 3, 108 * 0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 109 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hh">{@code \x}<i>hh</i></th> 110 * <td headers="matches characters hex_hh">The character with hexadecimal value {@code 0x}<i>hh</i></td></tr> 111 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hhhh"><code>\u</code><i>hhhh</i></th> 112 * <td headers="matches characters hex_hhhh">The character with hexadecimal value {@code 0x}<i>hhhh</i></td></tr> 113 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_h_h"><code>\x</code><i>{h...h}</i></th> 114 * <td headers="matches characters hex_h_h">The character with hexadecimal value {@code 0x}<i>h...h</i> 115 * ({@link java.lang.Character#MIN_CODE_POINT Character.MIN_CODE_POINT} 116 * <= {@code 0x}<i>h...h</i> <= 117 * {@link java.lang.Character#MAX_CODE_POINT Character.MAX_CODE_POINT})</td></tr> 118 * <tr><th style="vertical-align:top; font-weight: normal" id="unicode_name"><code>\N{</code><i>name</i><code>}</code></th> 119 * <td headers="matches characters unicode_name">The character with Unicode character name <i>'name'</i></td></tr> 120 * <tr><th style="vertical-align:top; font-weight:normal" id="tab">{@code \t}</th> 121 * <td headers="matches characters tab">The tab character (<code>'\u0009'</code>)</td></tr> 122 * <tr><th style="vertical-align:top; font-weight:normal" id="newline">{@code \n}</th> 123 * <td headers="matches characters newline">The newline (line feed) character (<code>'\u000A'</code>)</td></tr> 124 * <tr><th style="vertical-align:top; font-weight:normal" id="return">{@code \r}</th> 125 * <td headers="matches characters return">The carriage-return character (<code>'\u000D'</code>)</td></tr> 126 * <tr><th style="vertical-align:top; font-weight:normal" id="form_feed">{@code \f}</th> 127 * <td headers="matches characters form_feed">The form-feed character (<code>'\u000C'</code>)</td></tr> 128 * <tr><th style="vertical-align:top; font-weight:normal" id="bell">{@code \a}</th> 129 * <td headers="matches characters bell">The alert (bell) character (<code>'\u0007'</code>)</td></tr> 130 * <tr><th style="vertical-align:top; font-weight:normal" id="escape">{@code \e}</th> 131 * <td headers="matches characters escape">The escape character (<code>'\u001B'</code>)</td></tr> 132 * <tr><th style="vertical-align:top; font-weight:normal" id="ctrl_x">{@code \c}<i>x</i></th> 133 * <td headers="matches characters ctrl_x">The control character corresponding to <i>x</i></td></tr> 134 * 135 * <tr><th colspan="2" style="padding-top:20px" id="classes">Character classes</th></tr> 136 * 137 * <tr><th style="vertical-align:top; font-weight:normal" id="simple">{@code [abc]}</th> 138 * <td headers="matches classes simple">{@code a}, {@code b}, or {@code c} (simple class)</td></tr> 139 * <tr><th style="vertical-align:top; font-weight:normal" id="negation">{@code [^abc]}</th> 140 * <td headers="matches classes negation">Any character except {@code a}, {@code b}, or {@code c} (negation)</td></tr> 141 * <tr><th style="vertical-align:top; font-weight:normal" id="range">{@code [a-zA-Z]}</th> 142 * <td headers="matches classes range">{@code a} through {@code z} 143 * or {@code A} through {@code Z}, inclusive (range)</td></tr> 144 * <tr><th style="vertical-align:top; font-weight:normal" id="union">{@code [a-d[m-p]]}</th> 145 * <td headers="matches classes union">{@code a} through {@code d}, 146 * or {@code m} through {@code p}: {@code [a-dm-p]} (union)</td></tr> 147 * <tr><th style="vertical-align:top; font-weight:normal" id="intersection">{@code [a-z&&[def]]}</th> 148 * <td headers="matches classes intersection">{@code d}, {@code e}, or {@code f} (intersection)</tr> 149 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction1">{@code [a-z&&[^bc]]}</th> 150 * <td headers="matches classes subtraction1">{@code a} through {@code z}, 151 * except for {@code b} and {@code c}: {@code [ad-z]} (subtraction)</td></tr> 152 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction2">{@code [a-z&&[^m-p]]}</th> 153 * <td headers="matches classes subtraction2">{@code a} through {@code z}, 154 * and not {@code m} through {@code p}: {@code [a-lq-z]}(subtraction)</td></tr> 155 * 156 * <tr><th colspan="2" style="padding-top:20px" id="predef">Predefined character classes</th></tr> 157 * 158 * <tr><th style="vertical-align:top; font-weight:normal" id="any">{@code .}</th> 159 * <td headers="matches predef any">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr> 160 * <tr><th style="vertical-align:top; font-weight:normal" id="digit">{@code \d}</th> 161 * <td headers="matches predef digit">A digit: {@code [0-9]}</td></tr> 162 * <tr><th style="vertical-align:top; font-weight:normal" id="non_digit">{@code \D}</th> 163 * <td headers="matches predef non_digit">A non-digit: {@code [^0-9]}</td></tr> 164 * <tr><th style="vertical-align:top; font-weight:normal" id="horiz_white">{@code \h}</th> 165 * <td headers="matches predef horiz_white">A horizontal whitespace character: 166 * <code>[ \t\xA0\u1680\u180e\u2000-\u200a\u202f\u205f\u3000]</code></td></tr> 167 * <tr><th style="vertical-align:top; font-weight:normal" id="non_horiz_white">{@code \H}</th> 168 * <td headers="matches predef non_horiz_white">A non-horizontal whitespace character: {@code [^\h]}</td></tr> 169 * <tr><th style="vertical-align:top; font-weight:normal" id="white">{@code \s}</th> 170 * <td headers="matches predef white">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 171 * <tr><th style="vertical-align:top; font-weight:normal" id="non_white">{@code \S}</th> 172 * <td headers="matches predef non_white">A non-whitespace character: {@code [^\s]}</td></tr> 173 * <tr><th style="vertical-align:top; font-weight:normal" id="vert_white">{@code \v}</th> 174 * <td headers="matches predef vert_white">A vertical whitespace character: <code>[\n\x0B\f\r\x85\u2028\u2029]</code> 175 * </td></tr> 176 * <tr><th style="vertical-align:top; font-weight:normal" id="non_vert_white">{@code \V}</th> 177 * <td headers="matches predef non_vert_white">A non-vertical whitespace character: {@code [^\v]}</td></tr> 178 * <tr><th style="vertical-align:top; font-weight:normal" id="word">{@code \w}</th> 179 * <td headers="matches predef word">A word character: {@code [a-zA-Z_0-9]}</td></tr> 180 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word">{@code \W}</th> 181 * <td headers="matches predef non_word">A non-word character: {@code [^\w]}</td></tr> 182 * 183 * <tr><th colspan="2" style="padding-top:20px" id="posix"><b>POSIX character classes (US-ASCII only)</b></th></tr> 184 * 185 * <tr><th style="vertical-align:top; font-weight:normal" id="Lower">{@code \p{Lower}}</th> 186 * <td headers="matches posix Lower">A lower-case alphabetic character: {@code [a-z]}</td></tr> 187 * <tr><th style="vertical-align:top; font-weight:normal" id="Upper">{@code \p{Upper}}</th> 188 * <td headers="matches posix Upper">An upper-case alphabetic character:{@code [A-Z]}</td></tr> 189 * <tr><th style="vertical-align:top; font-weight:normal" id="ASCII">{@code \p{ASCII}}</th> 190 * <td headers="matches posix ASCII">All ASCII:{@code [\x00-\x7F]}</td></tr> 191 * <tr><th style="vertical-align:top; font-weight:normal" id="Alpha">{@code \p{Alpha}}</th> 192 * <td headers="matches posix Alpha">An alphabetic character:{@code [\p{Lower}\p{Upper}]}</td></tr> 193 * <tr><th style="vertical-align:top; font-weight:normal" id="Digit">{@code \p{Digit}}</th> 194 * <td headers="matches posix Digit">A decimal digit: {@code [0-9]}</td></tr> 195 * <tr><th style="vertical-align:top; font-weight:normal" id="Alnum">{@code \p{Alnum}}</th> 196 * <td headers="matches posix Alnum">An alphanumeric character:{@code [\p{Alpha}\p{Digit}]}</td></tr> 197 * <tr><th style="vertical-align:top; font-weight:normal" id="Punct">{@code \p{Punct}}</th> 198 * <td headers="matches posix Punct">Punctuation: One of {@code !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~}</td></tr> 199 * <!-- {@code [\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]} 200 * {@code [\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]} --> 201 * <tr><th style="vertical-align:top; font-weight:normal" id="Graph">{@code \p{Graph}}</th> 202 * <td headers="matches posix Graph">A visible character: {@code [\p{Alnum}\p{Punct}]}</td></tr> 203 * <tr><th style="vertical-align:top; font-weight:normal" id="Print">{@code \p{Print}}</th> 204 * <td headers="matches posix Print">A printable character: {@code [\p{Graph}\x20]}</td></tr> 205 * <tr><th style="vertical-align:top; font-weight:normal" id="Blank">{@code \p{Blank}}</th> 206 * <td headers="matches posix Blank">A space or a tab: {@code [ \t]}</td></tr> 207 * <tr><th style="vertical-align:top; font-weight:normal" id="Cntrl">{@code \p{Cntrl}}</th> 208 * <td headers="matches posix Cntrl">A control character: {@code [\x00-\x1F\x7F]}</td></tr> 209 * <tr><th style="vertical-align:top; font-weight:normal" id="XDigit">{@code \p{XDigit}}</th> 210 * <td headers="matches posix XDigit">A hexadecimal digit: {@code [0-9a-fA-F]}</td></tr> 211 * <tr><th style="vertical-align:top; font-weight:normal" id="Space">{@code \p{Space}}</th> 212 * <td headers="matches posix Space">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 213 * 214 * <tr><th colspan="2" style="padding-top:20px" id="java">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr> 215 * 216 * <tr><th style="vertical-align:top; font-weight:normal" id="javaLowerCase">{@code \p{javaLowerCase}}</th> 217 * <td headers="matches java javaLowerCase">Equivalent to java.lang.Character.isLowerCase()</td></tr> 218 * <tr><th style="vertical-align:top; font-weight:normal" id="javaUpperCase">{@code \p{javaUpperCase}}</th> 219 * <td headers="matches java javaUpperCase">Equivalent to java.lang.Character.isUpperCase()</td></tr> 220 * <tr><th style="vertical-align:top; font-weight:normal" id="javaWhitespace">{@code \p{javaWhitespace}}</th> 221 * <td headers="matches java javaWhitespace">Equivalent to java.lang.Character.isWhitespace()</td></tr> 222 * <tr><th style="vertical-align:top; font-weight:normal" id="javaMirrored">{@code \p{javaMirrored}}</th> 223 * <td headers="matches java javaMirrored">Equivalent to java.lang.Character.isMirrored()</td></tr> 224 * 225 * <tr><th colspan="2" style="padding-top:20px" id="unicode">Classes for Unicode scripts, blocks, categories and binary properties</th></tr> 226 * 227 * <tr><th style="vertical-align:top; font-weight:normal" id="IsLatin">{@code \p{IsLatin}}</th> 228 * <td headers="matches unicode IsLatin">A Latin script character (<a href="#usc">script</a>)</td></tr> 229 * <tr><th style="vertical-align:top; font-weight:normal" id="InGreek">{@code \p{InGreek}}</th> 230 * <td headers="matches unicode InGreek">A character in the Greek block (<a href="#ubc">block</a>)</td></tr> 231 * <tr><th style="vertical-align:top; font-weight:normal" id="Lu">{@code \p{Lu}}</th> 232 * <td headers="matches unicode Lu">An uppercase letter (<a href="#ucc">category</a>)</td></tr> 233 * <tr><th style="vertical-align:top; font-weight:normal" id="IsAlphabetic">{@code \p{IsAlphabetic}}</th> 234 * <td headers="matches unicode IsAlphabetic">An alphabetic character (<a href="#ubpc">binary property</a>)</td></tr> 235 * <tr><th style="vertical-align:top; font-weight:normal" id="Sc">{@code \p{Sc}}</th> 236 * <td headers="matches unicode Sc">A currency symbol</td></tr> 237 * <tr><th style="vertical-align:top; font-weight:normal" id="not_InGreek">{@code \P{InGreek}}</th> 238 * <td headers="matches unicode not_InGreek">Any character except one in the Greek block (negation)</td></tr> 239 * <tr><th style="vertical-align:top; font-weight:normal" id="not_uppercase">{@code [\p{L}&&[^\p{Lu}]]}</th> 240 * <td headers="matches unicode not_uppercase">Any letter except an uppercase letter (subtraction)</td></tr> 241 * 242 * <tr><th colspan="2" style="padding-top:20px" id="bounds">Boundary matchers</th></tr> 243 * 244 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_line">{@code ^}</th> 245 * <td headers="matches bounds begin_line">The beginning of a line</td></tr> 246 * <tr><th style="vertical-align:top; font-weight:normal" id="end_line">{@code $}</th> 247 * <td headers="matches bounds end_line">The end of a line</td></tr> 248 * <tr><th style="vertical-align:top; font-weight:normal" id="word_boundary">{@code \b}</th> 249 * <td headers="matches bounds word_boundary">A word boundary</td></tr> 250 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_cluster_boundary">{@code \b{g}}</th> 251 * <td headers="matches bounds grapheme_cluster_boundary">A Unicode extended grapheme cluster boundary</td></tr> 252 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word_boundary">{@code \B}</th> 253 * <td headers="matches bounds non_word_boundary">A non-word boundary</td></tr> 254 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_input">{@code \A}</th> 255 * <td headers="matches bounds begin_input">The beginning of the input</td></tr> 256 * <tr><th style="vertical-align:top; font-weight:normal" id="end_prev_match">{@code \G}</th> 257 * <td headers="matches bounds end_prev_match">The end of the previous match</td></tr> 258 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input_except_term">{@code \Z}</th> 259 * <td headers="matches bounds end_input_except_term">The end of the input but for the final 260 * <a href="#lt">terminator</a>, if any</td></tr> 261 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input">{@code \z}</th> 262 * <td headers="matches bounds end_input">The end of the input</td></tr> 263 * 264 * <tr><th colspan="2" style="padding-top:20px" id="linebreak">Linebreak matcher</th></tr> 265 * 266 * <tr><th style="vertical-align:top; font-weight:normal" id="any_unicode_linebreak">{@code \R}</th> 267 * <td headers="matches linebreak any_unicode_linebreak">Any Unicode linebreak sequence, is equivalent to 268 * <code>\u000D\u000A|[\u000A\u000B\u000C\u000D\u0085\u2028\u2029] 269 * </code></td></tr> 270 * 271 * <tr><th colspan="2" style="padding-top:20px" id="grapheme">Unicode Extended Grapheme matcher</th></tr> 272 * 273 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_any">{@code \X}</th> 274 * <td headers="matches grapheme grapheme_any">Any Unicode extended grapheme cluster</td></tr> 275 * 276 * <tr><th colspan="2" style="padding-top:20px" id="greedy">Greedy quantifiers</th></tr> 277 * 278 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_once_or_not"><i>X</i>{@code ?}</th> 279 * <td headers="matches greedy greedy_once_or_not"><i>X</i>, once or not at all</td></tr> 280 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_zero_or_more"><i>X</i>{@code *}</th> 281 * <td headers="matches greedy greedy_zero_or_more"><i>X</i>, zero or more times</td></tr> 282 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_one_or_more"><i>X</i>{@code +}</th> 283 * <td headers="matches greedy greedy_one_or_more"><i>X</i>, one or more times</td></tr> 284 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_exactly"><i>X</i><code>{</code><i>n</i><code>}</code></th> 285 * <td headers="matches greedy greedy_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 286 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least"><i>X</i><code>{</code><i>n</i>{@code ,}}</th> 287 * <td headers="matches greedy greedy_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 288 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}</code></th> 289 * <td headers="matches greedy greedy_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 290 * 291 * <tr><th colspan="2" style="padding-top:20px" id="reluc">Reluctant quantifiers</th></tr> 292 * 293 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_once_or_not"><i>X</i>{@code ??}</th> 294 * <td headers="matches reluc reluc_once_or_not"><i>X</i>, once or not at all</td></tr> 295 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_zero_or_more"><i>X</i>{@code *?}</th> 296 * <td headers="matches reluc reluc_zero_or_more"><i>X</i>, zero or more times</td></tr> 297 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_one_or_more"><i>X</i>{@code +?}</th> 298 * <td headers="matches reluc reluc_one_or_more"><i>X</i>, one or more times</td></tr> 299 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_exactly"><i>X</i><code>{</code><i>n</i><code>}?</code></th> 300 * <td headers="matches reluc reluc_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 301 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least"><i>X</i><code>{</code><i>n</i><code>,}?</code></th> 302 * <td headers="matches reluc reluc_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 303 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}?</code></th> 304 * <td headers="matches reluc reluc_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 305 * 306 * <tr><th colspan="2" style="padding-top:20px" id="poss">Possessive quantifiers</th></tr> 307 * 308 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_once_or_not"><i>X</i>{@code ?+}</th> 309 * <td headers="matches poss poss_once_or_not"><i>X</i>, once or not at all</td></tr> 310 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_zero_or_more"><i>X</i>{@code *+}</th> 311 * <td headers="matches poss poss_zero_or_more"><i>X</i>, zero or more times</td></tr> 312 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_one_or_more"><i>X</i>{@code ++}</th> 313 * <td headers="matches poss poss_one_or_more"><i>X</i>, one or more times</td></tr> 314 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_exactly"><i>X</i><code>{</code><i>n</i><code>}+</code></th> 315 * <td headers="matches poss poss_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 316 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least"><i>X</i><code>{</code><i>n</i><code>,}+</code></th> 317 * <td headers="matches poss poss_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 318 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}+</code></th> 319 * <td headers="matches poss poss_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 320 * 321 * <tr><th colspan="2" style="padding-top:20px" id="logical">Logical operators</th></tr> 322 * 323 * <tr><th style="vertical-align:top; font-weight:normal" id="concat"><i>XY</i></th> 324 * <td headers="matches logical concat"><i>X</i> followed by <i>Y</i></td></tr> 325 * <tr><th style="vertical-align:top; font-weight:normal" id="alternate"><i>X</i>{@code |}<i>Y</i></th> 326 * <td headers="matches logical alternate">Either <i>X</i> or <i>Y</i></td></tr> 327 * <tr><th style="vertical-align:top; font-weight:normal" id="group">{@code (}<i>X</i>{@code )}</th> 328 * <td headers="matches logical group">X, as a <a href="#cg">capturing group</a></td></tr> 329 * 330 * <tr><th colspan="2" style="padding-top:20px" id="backref">Back references</th></tr> 331 * 332 * <tr><th style="vertical-align:top; font-weight:normal" id="back_nth">{@code \}<i>n</i></th> 333 * <td headers="matches backref back_nth">Whatever the <i>n</i><sup>th</sup> 334 * <a href="#cg">capturing group</a> matched</td></tr> 335 * <tr><th style="vertical-align:top; font-weight:normal" id="back_named">{@code \}<i>k</i><<i>name</i>></th> 336 * <td headers="matches backref back_named">Whatever the 337 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr> 338 * 339 * <tr><th colspan="2" style="padding-top:20px" id="quote">Quotation</th></tr> 340 * 341 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_follow">{@code \}</th> 342 * <td headers="matches quote quote_follow">Nothing, but quotes the following character</td></tr> 343 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_begin">{@code \Q}</th> 344 * <td headers="matches quote quote_begin">Nothing, but quotes all characters until {@code \E}</td></tr> 345 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_end">{@code \E}</th> 346 * <td headers="matches quote quote_end">Nothing, but ends quoting started by {@code \Q}</td></tr> 347 * <!-- Metachars: !$()*+.<>?[\]^{|} --> 348 * 349 * <tr><th colspan="2" style="padding-top:20px" id="special">Special constructs (named-capturing and non-capturing)</th></tr> 350 * 351 * <tr><th style="vertical-align:top; font-weight:normal" id="named_group"><code>(?<<a href="#groupname">name</a>></code><i>X</i>{@code )}</th> 352 * <td headers="matches special named_group"><i>X</i>, as a named-capturing group</td></tr> 353 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group">{@code (?:}<i>X</i>{@code )}</th> 354 * <td headers="matches special non_capture_group"><i>X</i>, as a non-capturing group</td></tr> 355 * <tr><th style="vertical-align:top; font-weight:normal" id="flags"><code>(?idmsuxU-idmsuxU) </code></th> 356 * <td headers="matches special flags">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a> 357 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> 358 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a> 359 * on - off</td></tr> 360 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group_flags"><code>(?idmsuxU-idmsuxU:</code><i>X</i>{@code )} </th> 361 * <td headers="matches special non_capture_group_flags"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the 362 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a> 363 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a > 364 * <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a> on - off</td></tr> 365 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookahead">{@code (?=}<i>X</i>{@code )}</th> 366 * <td headers="matches special pos_lookahead"><i>X</i>, via zero-width positive lookahead</td></tr> 367 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookahead">{@code (?!}<i>X</i>{@code )}</th> 368 * <td headers="matches special neg_lookahead"><i>X</i>, via zero-width negative lookahead</td></tr> 369 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookbehind">{@code (?<=}<i>X</i>{@code )}</th> 370 * <td headers="matches special pos_lookbehind"><i>X</i>, via zero-width positive lookbehind</td></tr> 371 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookbehind">{@code (?<!}<i>X</i>{@code )}</th> 372 * <td headers="matches special neg_lookbehind"><i>X</i>, via zero-width negative lookbehind</td></tr> 373 * <tr><th style="vertical-align:top; font-weight:normal" id="indep_non_capture_group">{@code (?>}<i>X</i>{@code )}</th> 374 * <td headers="matches special indep_non_capture_group"><i>X</i>, as an independent, non-capturing group</td></tr> 375 * 376 * </tbody> 377 * </table> 378 * 379 * <hr> 380 * 381 * 382 * <h2><a id="bs">Backslashes, escapes, and quoting</a></h2> 383 * 384 * <p> The backslash character ({@code '\'}) serves to introduce escaped 385 * constructs, as defined in the table above, as well as to quote characters 386 * that otherwise would be interpreted as unescaped constructs. Thus the 387 * expression {@code \\} matches a single backslash and <code>\{</code> matches a 388 * left brace. 389 * 390 * <p> It is an error to use a backslash prior to any alphabetic character that 391 * does not denote an escaped construct; these are reserved for future 392 * extensions to the regular-expression language. A backslash may be used 393 * prior to a non-alphabetic character regardless of whether that character is 394 * part of an unescaped construct. 395 * 396 * <p> Backslashes within string literals in Java source code are interpreted 397 * as required by 398 * <cite>The Java™ Language Specification</cite> 399 * as either Unicode escapes (section 3.3) or other character escapes (section 3.10.6) 400 * It is therefore necessary to double backslashes in string 401 * literals that represent regular expressions to protect them from 402 * interpretation by the Java bytecode compiler. The string literal 403 * <code>"\b"</code>, for example, matches a single backspace character when 404 * interpreted as a regular expression, while {@code "\\b"} matches a 405 * word boundary. The string literal {@code "\(hello\)"} is illegal 406 * and leads to a compile-time error; in order to match the string 407 * {@code (hello)} the string literal {@code "\\(hello\\)"} 408 * must be used. 409 * 410 * <h2><a id="cc">Character Classes</a></h2> 411 * 412 * <p> Character classes may appear within other character classes, and 413 * may be composed by the union operator (implicit) and the intersection 414 * operator ({@code &&}). 415 * The union operator denotes a class that contains every character that is 416 * in at least one of its operand classes. The intersection operator 417 * denotes a class that contains every character that is in both of its 418 * operand classes. 419 * 420 * <p> The precedence of character-class operators is as follows, from 421 * highest to lowest: 422 * 423 * <table class="striped" style="margin-left: 2em;"> 424 * <caption style="display:none">Precedence of character class operators.</caption> 425 * <thead> 426 * <tr><th scope="col">Precedence<th scope="col">Name<th scope="col">Example 427 * </thead> 428 * <tbody> 429 * <tr><th scope="row">1</th> 430 * <td>Literal escape </td> 431 * <td>{@code \x}</td></tr> 432 * <tr><th scope="row">2</th> 433 * <td>Grouping</td> 434 * <td>{@code [...]}</td></tr> 435 * <tr><th scope="row">3</th> 436 * <td>Range</td> 437 * <td>{@code a-z}</td></tr> 438 * <tr><th scope="row">4</th> 439 * <td>Union</td> 440 * <td>{@code [a-e][i-u]}</td></tr> 441 * <tr><th scope="row">5</th> 442 * <td>Intersection</td> 443 * <td>{@code [a-z&&[aeiou]]}</td></tr> 444 * </tbody> 445 * </table> 446 * 447 * <p> Note that a different set of metacharacters are in effect inside 448 * a character class than outside a character class. For instance, the 449 * regular expression {@code .} loses its special meaning inside a 450 * character class, while the expression {@code -} becomes a range 451 * forming metacharacter. 452 * 453 * <h2><a id="lt">Line terminators</a></h2> 454 * 455 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks 456 * the end of a line of the input character sequence. The following are 457 * recognized as line terminators: 458 * 459 * <ul> 460 * 461 * <li> A newline (line feed) character ({@code '\n'}), 462 * 463 * <li> A carriage-return character followed immediately by a newline 464 * character ({@code "\r\n"}), 465 * 466 * <li> A standalone carriage-return character ({@code '\r'}), 467 * 468 * <li> A next-line character (<code>'\u0085'</code>), 469 * 470 * <li> A line-separator character (<code>'\u2028'</code>), or 471 * 472 * <li> A paragraph-separator character (<code>'\u2029'</code>). 473 * 474 * </ul> 475 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators 476 * recognized are newline characters. 477 * 478 * <p> The regular expression {@code .} matches any character except a line 479 * terminator unless the {@link #DOTALL} flag is specified. 480 * 481 * <p> By default, the regular expressions {@code ^} and {@code $} ignore 482 * line terminators and only match at the beginning and the end, respectively, 483 * of the entire input sequence. If {@link #MULTILINE} mode is activated then 484 * {@code ^} matches at the beginning of input and after any line terminator 485 * except at the end of input. When in {@link #MULTILINE} mode {@code $} 486 * matches just before a line terminator or the end of the input sequence. 487 * 488 * <h2><a id="cg">Groups and capturing</a></h2> 489 * 490 * <h3><a id="gnumber">Group number</a></h3> 491 * <p> Capturing groups are numbered by counting their opening parentheses from 492 * left to right. In the expression {@code ((A)(B(C)))}, for example, there 493 * are four such groups: </p> 494 * 495 * <ol style="margin-left:2em;"> 496 * <li> {@code ((A)(B(C)))} 497 * <li> {@code (A)} 498 * <li> {@code (B(C))} 499 * <li> {@code (C)} 500 * </ol> 501 * 502 * <p> Group zero always stands for the entire expression. 503 * 504 * <p> Capturing groups are so named because, during a match, each subsequence 505 * of the input sequence that matches such a group is saved. The captured 506 * subsequence may be used later in the expression, via a back reference, and 507 * may also be retrieved from the matcher once the match operation is complete. 508 * 509 * <h3><a id="groupname">Group name</a></h3> 510 * <p>A capturing group can also be assigned a "name", a {@code named-capturing group}, 511 * and then be back-referenced later by the "name". Group names are composed of 512 * the following characters. The first character must be a {@code letter}. 513 * 514 * <ul> 515 * <li> The uppercase letters {@code 'A'} through {@code 'Z'} 516 * (<code>'\u0041'</code> through <code>'\u005a'</code>), 517 * <li> The lowercase letters {@code 'a'} through {@code 'z'} 518 * (<code>'\u0061'</code> through <code>'\u007a'</code>), 519 * <li> The digits {@code '0'} through {@code '9'} 520 * (<code>'\u0030'</code> through <code>'\u0039'</code>), 521 * </ul> 522 * 523 * <p> A {@code named-capturing group} is still numbered as described in 524 * <a href="#gnumber">Group number</a>. 525 * 526 * <p> The captured input associated with a group is always the subsequence 527 * that the group most recently matched. If a group is evaluated a second time 528 * because of quantification then its previously-captured value, if any, will 529 * be retained if the second evaluation fails. Matching the string 530 * {@code "aba"} against the expression {@code (a(b)?)+}, for example, leaves 531 * group two set to {@code "b"}. All captured input is discarded at the 532 * beginning of each match. 533 * 534 * <p> Groups beginning with {@code (?} are either pure, <i>non-capturing</i> groups 535 * that do not capture text and do not count towards the group total, or 536 * <i>named-capturing</i> group. 537 * 538 * <h2> Unicode support </h2> 539 * 540 * <p> This class is in conformance with Level 1 of <a 541 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 542 * Standard #18: Unicode Regular Expression</i></a>, plus RL2.1 543 * Canonical Equivalents and RL2.2 Extended Grapheme Clusters. 544 * <p> 545 * <b>Unicode escape sequences</b> such as <code>\u2014</code> in Java source code 546 * are processed as described in section 3.3 of 547 * <cite>The Java™ Language Specification</cite>. 548 * Such escape sequences are also implemented directly by the regular-expression 549 * parser so that Unicode escapes can be used in expressions that are read from 550 * files or from the keyboard. Thus the strings <code>"\u2014"</code> and 551 * {@code "\\u2014"}, while not equal, compile into the same pattern, which 552 * matches the character with hexadecimal value {@code 0x2014}. 553 * <p> 554 * A Unicode character can also be represented by using its <b>Hex notation</b> 555 * (hexadecimal code point value) directly as described in construct 556 * <code>\x{...}</code>, for example a supplementary character U+2011F can be 557 * specified as <code>\x{2011F}</code>, instead of two consecutive Unicode escape 558 * sequences of the surrogate pair <code>\uD840</code><code>\uDD1F</code>. 559 * <p> 560 * <b>Unicode character names</b> are supported by the named character construct 561 * <code>\N{</code>...<code>}</code>, for example, <code>\N{WHITE SMILING FACE}</code> 562 * specifies character <code>\u263A</code>. The character names supported 563 * by this class are the valid Unicode character names matched by 564 * {@link java.lang.Character#codePointOf(String) Character.codePointOf(name)}. 565 * <p> 566 * <a href="http://www.unicode.org/reports/tr18/#Default_Grapheme_Clusters"> 567 * <b>Unicode extended grapheme clusters</b></a> are supported by the grapheme 568 * cluster matcher {@code \X} and the corresponding boundary matcher {@code \b{g}}. 569 * <p> 570 * Unicode scripts, blocks, categories and binary properties are written with 571 * the {@code \p} and {@code \P} constructs as in Perl. 572 * <code>\p{</code><i>prop</i><code>}</code> matches if 573 * the input has the property <i>prop</i>, while <code>\P{</code><i>prop</i><code>}</code> 574 * does not match if the input has that property. 575 * <p> 576 * Scripts, blocks, categories and binary properties can be used both inside 577 * and outside of a character class. 578 * 579 * <p> 580 * <b><a id="usc">Scripts</a></b> are specified either with the prefix {@code Is}, as in 581 * {@code IsHiragana}, or by using the {@code script} keyword (or its short 582 * form {@code sc}) as in {@code script=Hiragana} or {@code sc=Hiragana}. 583 * <p> 584 * The script names supported by {@code Pattern} are the valid script names 585 * accepted and defined by 586 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}. 587 * 588 * <p> 589 * <b><a id="ubc">Blocks</a></b> are specified with the prefix {@code In}, as in 590 * {@code InMongolian}, or by using the keyword {@code block} (or its short 591 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}. 592 * <p> 593 * The block names supported by {@code Pattern} are the valid block names 594 * accepted and defined by 595 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}. 596 * <p> 597 * 598 * <b><a id="ucc">Categories</a></b> may be specified with the optional prefix {@code Is}: 599 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode 600 * letters. Same as scripts and blocks, categories can also be specified 601 * by using the keyword {@code general_category} (or its short form 602 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}. 603 * <p> 604 * The supported categories are those of 605 * <a href="http://www.unicode.org/unicode/standard/standard.html"> 606 * <i>The Unicode Standard</i></a> in the version specified by the 607 * {@link java.lang.Character Character} class. The category names are those 608 * defined in the Standard, both normative and informative. 609 * <p> 610 * 611 * <b><a id="ubpc">Binary properties</a></b> are specified with the prefix {@code Is}, as in 612 * {@code IsAlphabetic}. The supported binary properties by {@code Pattern} 613 * are 614 * <ul> 615 * <li> Alphabetic 616 * <li> Ideographic 617 * <li> Letter 618 * <li> Lowercase 619 * <li> Uppercase 620 * <li> Titlecase 621 * <li> Punctuation 622 * <Li> Control 623 * <li> White_Space 624 * <li> Digit 625 * <li> Hex_Digit 626 * <li> Join_Control 627 * <li> Noncharacter_Code_Point 628 * <li> Assigned 629 * </ul> 630 * <p> 631 * The following <b>Predefined Character classes</b> and <b>POSIX character classes</b> 632 * are in conformance with the recommendation of <i>Annex C: Compatibility Properties</i> 633 * of <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Regular Expression 634 * </i></a>, when {@link #UNICODE_CHARACTER_CLASS} flag is specified. 635 * 636 * <table class="striped"> 637 * <caption style="display:none">predefined and posix character classes in Unicode mode</caption> 638 * <thead> 639 * <tr> 640 * <th scope="col" id="predef_classes">Classes</th> 641 * <th scope="col" id="predef_matches">Matches</th> 642 * </tr> 643 * </thead> 644 * <tbody> 645 * <tr><th scope="row">{@code \p{Lower}}</th> 646 * <td>A lowercase character:{@code \p{IsLowercase}}</td></tr> 647 * <tr><th scope="row">{@code \p{Upper}}</th> 648 * <td>An uppercase character:{@code \p{IsUppercase}}</td></tr> 649 * <tr><th scope="row">{@code \p{ASCII}}</th> 650 * <td>All ASCII:{@code [\x00-\x7F]}</td></tr> 651 * <tr><th scope="row">{@code \p{Alpha}}</th> 652 * <td>An alphabetic character:{@code \p{IsAlphabetic}}</td></tr> 653 * <tr><th scope="row">{@code \p{Digit}}</th> 654 * <td>A decimal digit character:{@code \p{IsDigit}}</td></tr> 655 * <tr><th scope="row">{@code \p{Alnum}}</th> 656 * <td>An alphanumeric character:{@code [\p{IsAlphabetic}\p{IsDigit}]}</td></tr> 657 * <tr><th scope="row">{@code \p{Punct}}</th> 658 * <td>A punctuation character:{@code \p{IsPunctuation}}</td></tr> 659 * <tr><th scope="row">{@code \p{Graph}}</th> 660 * <td>A visible character: {@code [^\p{IsWhite_Space}\p{gc=Cc}\p{gc=Cs}\p{gc=Cn}]}</td></tr> 661 * <tr><th scope="row">{@code \p{Print}}</th> 662 * <td>A printable character: {@code [\p{Graph}\p{Blank}&&[^\p{Cntrl}]]}</td></tr> 663 * <tr><th scope="row">{@code \p{Blank}}</th> 664 * <td>A space or a tab: {@code [\p{IsWhite_Space}&&[^\p{gc=Zl}\p{gc=Zp}\x0a\x0b\x0c\x0d\x85]]}</td></tr> 665 * <tr><th scope="row">{@code \p{Cntrl}}</th> 666 * <td>A control character: {@code \p{gc=Cc}}</td></tr> 667 * <tr><th scope="row">{@code \p{XDigit}}</th> 668 * <td>A hexadecimal digit: {@code [\p{gc=Nd}\p{IsHex_Digit}]}</td></tr> 669 * <tr><th scope="row">{@code \p{Space}}</th> 670 * <td>A whitespace character:{@code \p{IsWhite_Space}}</td></tr> 671 * <tr><th scope="row">{@code \d}</th> 672 * <td>A digit: {@code \p{IsDigit}}</td></tr> 673 * <tr><th scope="row">{@code \D}</th> 674 * <td>A non-digit: {@code [^\d]}</td></tr> 675 * <tr><th scope="row">{@code \s}</th> 676 * <td>A whitespace character: {@code \p{IsWhite_Space}}</td></tr> 677 * <tr><th scope="row">{@code \S}</th> 678 * <td>A non-whitespace character: {@code [^\s]}</td></tr> 679 * <tr><th scope="row">{@code \w}</th> 680 * <td>A word character: {@code [\p{Alpha}\p{gc=Mn}\p{gc=Me}\p{gc=Mc}\p{Digit}\p{gc=Pc}\p{IsJoin_Control}]}</td></tr> 681 * <tr><th scope="row">{@code \W}</th> 682 * <td>A non-word character: {@code [^\w]}</td></tr> 683 * </tbody> 684 * </table> 685 * <p> 686 * <a id="jcc"> 687 * Categories that behave like the java.lang.Character 688 * boolean is<i>methodname</i> methods (except for the deprecated ones) are 689 * available through the same <code>\p{</code><i>prop</i><code>}</code> syntax where 690 * the specified property has the name <code>java<i>methodname</i></code></a>. 691 * 692 * <h2> Comparison to Perl 5 </h2> 693 * 694 * <p>The {@code Pattern} engine performs traditional NFA-based matching 695 * with ordered alternation as occurs in Perl 5. 696 * 697 * <p> Perl constructs not supported by this class: </p> 698 * 699 * <ul> 700 * <li><p> The backreference constructs, <code>\g{</code><i>n</i><code>}</code> for 701 * the <i>n</i><sup>th</sup><a href="#cg">capturing group</a> and 702 * <code>\g{</code><i>name</i><code>}</code> for 703 * <a href="#groupname">named-capturing group</a>. 704 * </p></li> 705 * 706 * <li><p> The conditional constructs 707 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code )} and 708 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code |}<i>Y</i>{@code )}, 709 * </p></li> 710 * 711 * <li><p> The embedded code constructs <code>(?{</code><i>code</i><code>})</code> 712 * and <code>(??{</code><i>code</i><code>})</code>,</p></li> 713 * 714 * <li><p> The embedded comment syntax {@code (?#comment)}, and </p></li> 715 * 716 * <li><p> The preprocessing operations {@code \l} <code>\u</code>, 717 * {@code \L}, and {@code \U}. </p></li> 718 * 719 * </ul> 720 * 721 * <p> Constructs supported by this class but not by Perl: </p> 722 * 723 * <ul> 724 * 725 * <li><p> Character-class union and intersection as described 726 * <a href="#cc">above</a>.</p></li> 727 * 728 * </ul> 729 * 730 * <p> Notable differences from Perl: </p> 731 * 732 * <ul> 733 * 734 * <li><p> In Perl, {@code \1} through {@code \9} are always interpreted 735 * as back references; a backslash-escaped number greater than {@code 9} is 736 * treated as a back reference if at least that many subexpressions exist, 737 * otherwise it is interpreted, if possible, as an octal escape. In this 738 * class octal escapes must always begin with a zero. In this class, 739 * {@code \1} through {@code \9} are always interpreted as back 740 * references, and a larger number is accepted as a back reference if at 741 * least that many subexpressions exist at that point in the regular 742 * expression, otherwise the parser will drop digits until the number is 743 * smaller or equal to the existing number of groups or it is one digit. 744 * </p></li> 745 * 746 * <li><p> Perl uses the {@code g} flag to request a match that resumes 747 * where the last match left off. This functionality is provided implicitly 748 * by the {@link Matcher} class: Repeated invocations of the {@link 749 * Matcher#find find} method will resume where the last match left off, 750 * unless the matcher is reset. </p></li> 751 * 752 * <li><p> In Perl, embedded flags at the top level of an expression affect 753 * the whole expression. In this class, embedded flags always take effect 754 * at the point at which they appear, whether they are at the top level or 755 * within a group; in the latter case, flags are restored at the end of the 756 * group just as in Perl. </p></li> 757 * 758 * </ul> 759 * 760 * 761 * <p> For a more precise description of the behavior of regular expression 762 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/"> 763 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl, 764 * O'Reilly and Associates, 2006.</a> 765 * </p> 766 * 767 * @see java.lang.String#split(String, int) 768 * @see java.lang.String#split(String) 769 * 770 * @author Mike McCloskey 771 * @author Mark Reinhold 772 * @author JSR-51 Expert Group 773 * @since 1.4 774 * @spec JSR-51 775 */ 776 777 public final class Pattern 778 implements java.io.Serializable 779 { 780 781 /** 782 * Regular expression modifier values. Instead of being passed as 783 * arguments, they can also be passed as inline modifiers. 784 * For example, the following statements have the same effect. 785 * <pre> 786 * Pattern p1 = Pattern.compile("abc", Pattern.CASE_INSENSITIVE|Pattern.MULTILINE); 787 * Pattern p2 = Pattern.compile("(?im)abc", 0); 788 * </pre> 789 */ 790 791 /** 792 * Enables Unix lines mode. 793 * 794 * <p> In this mode, only the {@code '\n'} line terminator is recognized 795 * in the behavior of {@code .}, {@code ^}, and {@code $}. 796 * 797 * <p> Unix lines mode can also be enabled via the embedded flag 798 * expression {@code (?d)}. 799 */ 800 public static final int UNIX_LINES = 0x01; 801 802 /** 803 * Enables case-insensitive matching. 804 * 805 * <p> By default, case-insensitive matching assumes that only characters 806 * in the US-ASCII charset are being matched. Unicode-aware 807 * case-insensitive matching can be enabled by specifying the {@link 808 * #UNICODE_CASE} flag in conjunction with this flag. 809 * 810 * <p> Case-insensitive matching can also be enabled via the embedded flag 811 * expression {@code (?i)}. 812 * 813 * <p> Specifying this flag may impose a slight performance penalty. </p> 814 */ 815 public static final int CASE_INSENSITIVE = 0x02; 816 817 /** 818 * Permits whitespace and comments in pattern. 819 * 820 * <p> In this mode, whitespace is ignored, and embedded comments starting 821 * with {@code #} are ignored until the end of a line. 822 * 823 * <p> Comments mode can also be enabled via the embedded flag 824 * expression {@code (?x)}. 825 */ 826 public static final int COMMENTS = 0x04; 827 828 /** 829 * Enables multiline mode. 830 * 831 * <p> In multiline mode the expressions {@code ^} and {@code $} match 832 * just after or just before, respectively, a line terminator or the end of 833 * the input sequence. By default these expressions only match at the 834 * beginning and the end of the entire input sequence. 835 * 836 * <p> Multiline mode can also be enabled via the embedded flag 837 * expression {@code (?m)}. </p> 838 */ 839 public static final int MULTILINE = 0x08; 840 841 /** 842 * Enables literal parsing of the pattern. 843 * 844 * <p> When this flag is specified then the input string that specifies 845 * the pattern is treated as a sequence of literal characters. 846 * Metacharacters or escape sequences in the input sequence will be 847 * given no special meaning. 848 * 849 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on 850 * matching when used in conjunction with this flag. The other flags 851 * become superfluous. 852 * 853 * <p> There is no embedded flag character for enabling literal parsing. 854 * @since 1.5 855 */ 856 public static final int LITERAL = 0x10; 857 858 /** 859 * Enables dotall mode. 860 * 861 * <p> In dotall mode, the expression {@code .} matches any character, 862 * including a line terminator. By default this expression does not match 863 * line terminators. 864 * 865 * <p> Dotall mode can also be enabled via the embedded flag 866 * expression {@code (?s)}. (The {@code s} is a mnemonic for 867 * "single-line" mode, which is what this is called in Perl.) </p> 868 */ 869 public static final int DOTALL = 0x20; 870 871 /** 872 * Enables Unicode-aware case folding. 873 * 874 * <p> When this flag is specified then case-insensitive matching, when 875 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner 876 * consistent with the Unicode Standard. By default, case-insensitive 877 * matching assumes that only characters in the US-ASCII charset are being 878 * matched. 879 * 880 * <p> Unicode-aware case folding can also be enabled via the embedded flag 881 * expression {@code (?u)}. 882 * 883 * <p> Specifying this flag may impose a performance penalty. </p> 884 */ 885 public static final int UNICODE_CASE = 0x40; 886 887 /** 888 * Enables canonical equivalence. 889 * 890 * <p> When this flag is specified then two characters will be considered 891 * to match if, and only if, their full canonical decompositions match. 892 * The expression <code>"a\u030A"</code>, for example, will match the 893 * string <code>"\u00E5"</code> when this flag is specified. By default, 894 * matching does not take canonical equivalence into account. 895 * 896 * <p> There is no embedded flag character for enabling canonical 897 * equivalence. 898 * 899 * <p> Specifying this flag may impose a performance penalty. </p> 900 */ 901 public static final int CANON_EQ = 0x80; 902 903 /** 904 * Enables the Unicode version of <i>Predefined character classes</i> and 905 * <i>POSIX character classes</i>. 906 * 907 * <p> When this flag is specified then the (US-ASCII only) 908 * <i>Predefined character classes</i> and <i>POSIX character classes</i> 909 * are in conformance with 910 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 911 * Standard #18: Unicode Regular Expression</i></a> 912 * <i>Annex C: Compatibility Properties</i>. 913 * <p> 914 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded 915 * flag expression {@code (?U)}. 916 * <p> 917 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case 918 * folding. 919 * <p> 920 * Specifying this flag may impose a performance penalty. </p> 921 * @since 1.7 922 */ 923 public static final int UNICODE_CHARACTER_CLASS = 0x100; 924 925 /** 926 * Contains all possible flags for compile(regex, flags). 927 */ 928 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE | 929 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL | 930 UNICODE_CHARACTER_CLASS | COMMENTS; 931 932 /* Pattern has only two serialized components: The pattern string 933 * and the flags, which are all that is needed to recompile the pattern 934 * when it is deserialized. 935 */ 936 937 /** use serialVersionUID from Merlin b59 for interoperability */ 938 private static final long serialVersionUID = 5073258162644648461L; 939 940 /** 941 * The original regular-expression pattern string. 942 * 943 * @serial 944 */ 945 private String pattern; 946 947 /** 948 * The original pattern flags. 949 * 950 * @serial 951 */ 952 private int flags; 953 954 /** 955 * The temporary pattern flags used during compiling. The flags might be turn 956 * on and off by embedded flag. 957 */ 958 private transient int flags0; 959 960 /** 961 * Boolean indicating this Pattern is compiled; this is necessary in order 962 * to lazily compile deserialized Patterns. 963 */ 964 private transient volatile boolean compiled; 965 966 /** 967 * The normalized pattern string. 968 */ 969 private transient String normalizedPattern; 970 971 /** 972 * The starting point of state machine for the find operation. This allows 973 * a match to start anywhere in the input. 974 */ 975 transient Node root; 976 977 /** 978 * The root of object tree for a match operation. The pattern is matched 979 * at the beginning. This may include a find that uses BnM or a First 980 * node. 981 */ 982 transient Node matchRoot; 983 984 /** 985 * Temporary storage used by parsing pattern slice. 986 */ 987 transient int[] buffer; 988 989 /** 990 * A temporary storage used for predicate for double return. 991 */ 992 transient CharPredicate predicate; 993 994 /** 995 * Map the "name" of the "named capturing group" to its group id 996 * node. 997 */ 998 transient volatile Map<String, Integer> namedGroups; 999 1000 /** 1001 * Temporary storage used while parsing group references. 1002 */ 1003 transient GroupHead[] groupNodes; 1004 1005 /** 1006 * Temporary storage used to store the top level closure nodes. 1007 */ 1008 transient List<Node> topClosureNodes; 1009 1010 /** 1011 * The number of top greedy closure nodes in this Pattern. Used by 1012 * matchers to allocate storage needed for a IntHashSet to keep the 1013 * beginning pos {@code i} of all failed match. 1014 */ 1015 transient int localTCNCount; 1016 1017 /* 1018 * Turn off the stop-exponential-backtracking optimization if there 1019 * is a group ref in the pattern. 1020 */ 1021 transient boolean hasGroupRef; 1022 1023 /** 1024 * Temporary null terminated code point array used by pattern compiling. 1025 */ 1026 private transient int[] temp; 1027 1028 /** 1029 * The number of capturing groups in this Pattern. Used by matchers to 1030 * allocate storage needed to perform a match. 1031 */ 1032 transient int capturingGroupCount; 1033 1034 /** 1035 * The local variable count used by parsing tree. Used by matchers to 1036 * allocate storage needed to perform a match. 1037 */ 1038 transient int localCount; 1039 1040 /** 1041 * Index into the pattern string that keeps track of how much has been 1042 * parsed. 1043 */ 1044 private transient int cursor; 1045 1046 /** 1047 * Holds the length of the pattern string. 1048 */ 1049 private transient int patternLength; 1050 1051 /** 1052 * If the Start node might possibly match supplementary characters. 1053 * It is set to true during compiling if 1054 * (1) There is supplementary char in pattern, or 1055 * (2) There is complement node of a "family" CharProperty 1056 */ 1057 private transient boolean hasSupplementary; 1058 1059 /** 1060 * Compiles the given regular expression into a pattern. 1061 * 1062 * @param regex 1063 * The expression to be compiled 1064 * @return the given regular expression compiled into a pattern 1065 * @throws PatternSyntaxException 1066 * If the expression's syntax is invalid 1067 */ 1068 public static Pattern compile(String regex) { 1069 return new Pattern(regex, 0); 1070 } 1071 1072 /** 1073 * Compiles the given regular expression into a pattern with the given 1074 * flags. 1075 * 1076 * @param regex 1077 * The expression to be compiled 1078 * 1079 * @param flags 1080 * Match flags, a bit mask that may include 1081 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL}, 1082 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES}, 1083 * {@link #LITERAL}, {@link #UNICODE_CHARACTER_CLASS} 1084 * and {@link #COMMENTS} 1085 * 1086 * @return the given regular expression compiled into a pattern with the given flags 1087 * @throws IllegalArgumentException 1088 * If bit values other than those corresponding to the defined 1089 * match flags are set in {@code flags} 1090 * 1091 * @throws PatternSyntaxException 1092 * If the expression's syntax is invalid 1093 */ 1094 public static Pattern compile(String regex, int flags) { 1095 return new Pattern(regex, flags); 1096 } 1097 1098 /** 1099 * Returns the regular expression from which this pattern was compiled. 1100 * 1101 * @return The source of this pattern 1102 */ 1103 public String pattern() { 1104 return pattern; 1105 } 1106 1107 /** 1108 * <p>Returns the string representation of this pattern. This 1109 * is the regular expression from which this pattern was 1110 * compiled.</p> 1111 * 1112 * @return The string representation of this pattern 1113 * @since 1.5 1114 */ 1115 public String toString() { 1116 return pattern; 1117 } 1118 1119 /** 1120 * Creates a matcher that will match the given input against this pattern. 1121 * 1122 * @param input 1123 * The character sequence to be matched 1124 * 1125 * @return A new matcher for this pattern 1126 */ 1127 public Matcher matcher(CharSequence input) { 1128 if (!compiled) { 1129 synchronized(this) { 1130 if (!compiled) 1131 compile(); 1132 } 1133 } 1134 Matcher m = new Matcher(this, input); 1135 return m; 1136 } 1137 1138 /** 1139 * Returns this pattern's match flags. 1140 * 1141 * @return The match flags specified when this pattern was compiled 1142 */ 1143 public int flags() { 1144 return flags0; 1145 } 1146 1147 /** 1148 * Compiles the given regular expression and attempts to match the given 1149 * input against it. 1150 * 1151 * <p> An invocation of this convenience method of the form 1152 * 1153 * <blockquote><pre> 1154 * Pattern.matches(regex, input);</pre></blockquote> 1155 * 1156 * behaves in exactly the same way as the expression 1157 * 1158 * <blockquote><pre> 1159 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote> 1160 * 1161 * <p> If a pattern is to be used multiple times, compiling it once and reusing 1162 * it will be more efficient than invoking this method each time. </p> 1163 * 1164 * @param regex 1165 * The expression to be compiled 1166 * 1167 * @param input 1168 * The character sequence to be matched 1169 * @return whether or not the regular expression matches on the input 1170 * @throws PatternSyntaxException 1171 * If the expression's syntax is invalid 1172 */ 1173 public static boolean matches(String regex, CharSequence input) { 1174 Pattern p = Pattern.compile(regex); 1175 Matcher m = p.matcher(input); 1176 return m.matches(); 1177 } 1178 1179 /** 1180 * Splits the given input sequence around matches of this pattern. 1181 * 1182 * <p> The array returned by this method contains each substring of the 1183 * input sequence that is terminated by another subsequence that matches 1184 * this pattern or is terminated by the end of the input sequence. The 1185 * substrings in the array are in the order in which they occur in the 1186 * input. If this pattern does not match any subsequence of the input then 1187 * the resulting array has just one element, namely the input sequence in 1188 * string form. 1189 * 1190 * <p> When there is a positive-width match at the beginning of the input 1191 * sequence then an empty leading substring is included at the beginning 1192 * of the resulting array. A zero-width match at the beginning however 1193 * never produces such empty leading substring. 1194 * 1195 * <p> The {@code limit} parameter controls the number of times the 1196 * pattern is applied and therefore affects the length of the resulting 1197 * array. 1198 * <ul> 1199 * <li><p> 1200 * If the <i>limit</i> is positive then the pattern will be applied 1201 * at most <i>limit</i> - 1 times, the array's length will be 1202 * no greater than <i>limit</i>, and the array's last entry will contain 1203 * all input beyond the last matched delimiter.</p></li> 1204 * 1205 * <li><p> 1206 * If the <i>limit</i> is zero then the pattern will be applied as 1207 * many times as possible, the array can have any length, and trailing 1208 * empty strings will be discarded.</p></li> 1209 * 1210 * <li><p> 1211 * If the <i>limit</i> is negative then the pattern will be applied 1212 * as many times as possible and the array can have any length.</p></li> 1213 * </ul> 1214 * 1215 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1216 * results with these parameters: 1217 * 1218 * <table class="plain" style="margin-left:2em;"> 1219 * <caption style="display:none">Split example showing regex, limit, and result</caption> 1220 * <thead> 1221 * <tr> 1222 * <th scope="col">Regex</th> 1223 * <th scope="col">Limit</th> 1224 * <th scope="col">Result</th> 1225 * </tr> 1226 * </thead> 1227 * <tbody> 1228 * <tr><th scope="row" rowspan="3" style="font-weight:normal">:</th> 1229 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">2</th> 1230 * <td>{@code { "boo", "and:foo" }}</td></tr> 1231 * <tr><!-- : --> 1232 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1233 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1234 * <tr><!-- : --> 1235 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1236 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1237 * <tr><th scope="row" rowspan="3" style="font-weight:normal">o</th> 1238 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1239 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1240 * <tr><!-- o --> 1241 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1242 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1243 * <tr><!-- o --> 1244 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">0</th> 1245 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1246 * </tbody> 1247 * </table> 1248 * 1249 * @param input 1250 * The character sequence to be split 1251 * 1252 * @param limit 1253 * The result threshold, as described above 1254 * 1255 * @return The array of strings computed by splitting the input 1256 * around matches of this pattern 1257 */ 1258 public String[] split(CharSequence input, int limit) { 1259 int index = 0; 1260 boolean matchLimited = limit > 0; 1261 ArrayList<String> matchList = new ArrayList<>(); 1262 Matcher m = matcher(input); 1263 1264 // Add segments before each match found 1265 while(m.find()) { 1266 if (!matchLimited || matchList.size() < limit - 1) { 1267 if (index == 0 && index == m.start() && m.start() == m.end()) { 1268 // no empty leading substring included for zero-width match 1269 // at the beginning of the input char sequence. 1270 continue; 1271 } 1272 String match = input.subSequence(index, m.start()).toString(); 1273 matchList.add(match); 1274 index = m.end(); 1275 } else if (matchList.size() == limit - 1) { // last one 1276 String match = input.subSequence(index, 1277 input.length()).toString(); 1278 matchList.add(match); 1279 index = m.end(); 1280 } 1281 } 1282 1283 // If no match was found, return this 1284 if (index == 0) 1285 return new String[] {input.toString()}; 1286 1287 // Add remaining segment 1288 if (!matchLimited || matchList.size() < limit) 1289 matchList.add(input.subSequence(index, input.length()).toString()); 1290 1291 // Construct result 1292 int resultSize = matchList.size(); 1293 if (limit == 0) 1294 while (resultSize > 0 && matchList.get(resultSize-1).isEmpty()) 1295 resultSize--; 1296 String[] result = new String[resultSize]; 1297 return matchList.subList(0, resultSize).toArray(result); 1298 } 1299 1300 /** 1301 * Splits the given input sequence around matches of this pattern. 1302 * 1303 * <p> This method works as if by invoking the two-argument {@link 1304 * #split(java.lang.CharSequence, int) split} method with the given input 1305 * sequence and a limit argument of zero. Trailing empty strings are 1306 * therefore not included in the resulting array. </p> 1307 * 1308 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1309 * results with these expressions: 1310 * 1311 * <table class="plain" style="margin-left:2em"> 1312 * <caption style="display:none">Split examples showing regex and result</caption> 1313 * <thead> 1314 * <tr> 1315 * <th scope="col">Regex</th> 1316 * <th scope="col">Result</th> 1317 * </tr> 1318 * </thead> 1319 * <tbody> 1320 * <tr><th scope="row" style="text-weight:normal">:</th> 1321 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1322 * <tr><th scope="row" style="text-weight:normal">o</th> 1323 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1324 * </tbody> 1325 * </table> 1326 * 1327 * 1328 * @param input 1329 * The character sequence to be split 1330 * 1331 * @return The array of strings computed by splitting the input 1332 * around matches of this pattern 1333 */ 1334 public String[] split(CharSequence input) { 1335 return split(input, 0); 1336 } 1337 1338 /** 1339 * Returns a literal pattern {@code String} for the specified 1340 * {@code String}. 1341 * 1342 * <p>This method produces a {@code String} that can be used to 1343 * create a {@code Pattern} that would match the string 1344 * {@code s} as if it were a literal pattern.</p> Metacharacters 1345 * or escape sequences in the input sequence will be given no special 1346 * meaning. 1347 * 1348 * @param s The string to be literalized 1349 * @return A literal string replacement 1350 * @since 1.5 1351 */ 1352 public static String quote(String s) { 1353 int slashEIndex = s.indexOf("\\E"); 1354 if (slashEIndex == -1) 1355 return "\\Q" + s + "\\E"; 1356 1357 int lenHint = s.length(); 1358 lenHint = (lenHint < Integer.MAX_VALUE - 8 - lenHint) ? 1359 (lenHint << 1) : (Integer.MAX_VALUE - 8); 1360 1361 StringBuilder sb = new StringBuilder(lenHint); 1362 sb.append("\\Q"); 1363 int current = 0; 1364 do { 1365 sb.append(s, current, slashEIndex) 1366 .append("\\E\\\\E\\Q"); 1367 current = slashEIndex + 2; 1368 } while ((slashEIndex = s.indexOf("\\E", current)) != -1); 1369 1370 return sb.append(s, current, s.length()) 1371 .append("\\E") 1372 .toString(); 1373 } 1374 1375 /** 1376 * Recompile the Pattern instance from a stream. The original pattern 1377 * string is read in and the object tree is recompiled from it. 1378 */ 1379 private void readObject(java.io.ObjectInputStream s) 1380 throws java.io.IOException, ClassNotFoundException { 1381 1382 // Read in all fields 1383 s.defaultReadObject(); 1384 1385 // reset the flags 1386 flags0 = flags; 1387 1388 // Initialize counts 1389 capturingGroupCount = 1; 1390 localCount = 0; 1391 localTCNCount = 0; 1392 1393 // if length > 0, the Pattern is lazily compiled 1394 if (pattern.isEmpty()) { 1395 root = new Start(lastAccept); 1396 matchRoot = lastAccept; 1397 compiled = true; 1398 } 1399 } 1400 1401 /** 1402 * This private constructor is used to create all Patterns. The pattern 1403 * string and match flags are all that is needed to completely describe 1404 * a Pattern. An empty pattern string results in an object tree with 1405 * only a Start node and a LastNode node. 1406 */ 1407 private Pattern(String p, int f) { 1408 if ((f & ~ALL_FLAGS) != 0) { 1409 throw new IllegalArgumentException("Unknown flag 0x" 1410 + Integer.toHexString(f)); 1411 } 1412 pattern = p; 1413 flags = f; 1414 1415 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present 1416 if ((flags & UNICODE_CHARACTER_CLASS) != 0) 1417 flags |= UNICODE_CASE; 1418 1419 // 'flags' for compiling 1420 flags0 = flags; 1421 1422 // Reset group index count 1423 capturingGroupCount = 1; 1424 localCount = 0; 1425 localTCNCount = 0; 1426 1427 if (!pattern.isEmpty()) { 1428 compile(); 1429 } else { 1430 root = new Start(lastAccept); 1431 matchRoot = lastAccept; 1432 } 1433 } 1434 1435 /** 1436 * The pattern is converted to normalized form ({@link 1437 * java.text.Normalizer.Form#NFC NFC}, canonical decomposition, 1438 * followed by canonical composition for the character class 1439 * part, and {@link java.text.Normalizer.Form#NFD NFD}, 1440 * canonical decomposition for the rest), and then a pure 1441 * group is constructed to match canonical equivalences of the 1442 * characters. 1443 */ 1444 private static String normalize(String pattern) { 1445 int plen = pattern.length(); 1446 StringBuilder pbuf = new StringBuilder(plen); 1447 char last = 0; 1448 int lastStart = 0; 1449 char cc = 0; 1450 for (int i = 0; i < plen;) { 1451 char c = pattern.charAt(i); 1452 if (cc == 0 && // top level 1453 c == '\\' && i + 1 < plen && pattern.charAt(i + 1) == '\\') { 1454 i += 2; last = 0; 1455 continue; 1456 } 1457 if (c == '[' && last != '\\') { 1458 if (cc == 0) { 1459 if (lastStart < i) 1460 normalizeSlice(pattern, lastStart, i, pbuf); 1461 lastStart = i; 1462 } 1463 cc++; 1464 } else if (c == ']' && last != '\\') { 1465 cc--; 1466 if (cc == 0) { 1467 normalizeClazz(pattern, lastStart, i + 1, pbuf); 1468 lastStart = i + 1; 1469 } 1470 } 1471 last = c; 1472 i++; 1473 } 1474 assert (cc == 0); 1475 if (lastStart < plen) 1476 normalizeSlice(pattern, lastStart, plen, pbuf); 1477 return pbuf.toString(); 1478 } 1479 1480 private static void normalizeSlice(String src, int off, int limit, 1481 StringBuilder dst) 1482 { 1483 int len = src.length(); 1484 int off0 = off; 1485 while (off < limit && ASCII.isAscii(src.charAt(off))) { 1486 off++; 1487 } 1488 if (off == limit) { 1489 dst.append(src, off0, limit); 1490 return; 1491 } 1492 off--; 1493 if (off < off0) 1494 off = off0; 1495 else 1496 dst.append(src, off0, off); 1497 while (off < limit) { 1498 int ch0 = src.codePointAt(off); 1499 if (".$|()[]{}^?*+\\".indexOf(ch0) != -1) { 1500 dst.append((char)ch0); 1501 off++; 1502 continue; 1503 } 1504 int j = Grapheme.nextBoundary(src, off, limit); 1505 int ch1; 1506 String seq = src.substring(off, j); 1507 String nfd = Normalizer.normalize(seq, Normalizer.Form.NFD); 1508 off = j; 1509 if (nfd.codePointCount(0, nfd.length()) > 1) { 1510 ch0 = nfd.codePointAt(0); 1511 ch1 = nfd.codePointAt(Character.charCount(ch0)); 1512 if (Character.getType(ch1) == Character.NON_SPACING_MARK) { 1513 Set<String> altns = new LinkedHashSet<>(); 1514 altns.add(seq); 1515 produceEquivalentAlternation(nfd, altns); 1516 dst.append("(?:"); 1517 altns.forEach( s -> dst.append(s).append('|')); 1518 dst.delete(dst.length() - 1, dst.length()); 1519 dst.append(")"); 1520 continue; 1521 } 1522 } 1523 String nfc = Normalizer.normalize(seq, Normalizer.Form.NFC); 1524 if (!seq.equals(nfc) && !nfd.equals(nfc)) 1525 dst.append("(?:" + seq + "|" + nfd + "|" + nfc + ")"); 1526 else if (!seq.equals(nfd)) 1527 dst.append("(?:" + seq + "|" + nfd + ")"); 1528 else 1529 dst.append(seq); 1530 } 1531 } 1532 1533 private static void normalizeClazz(String src, int off, int limit, 1534 StringBuilder dst) 1535 { 1536 dst.append(Normalizer.normalize(src.substring(off, limit), Form.NFC)); 1537 } 1538 1539 /** 1540 * Given a specific sequence composed of a regular character and 1541 * combining marks that follow it, produce the alternation that will 1542 * match all canonical equivalences of that sequence. 1543 */ 1544 private static void produceEquivalentAlternation(String src, 1545 Set<String> dst) 1546 { 1547 int len = countChars(src, 0, 1); 1548 if (src.length() == len) { 1549 dst.add(src); // source has one character. 1550 return; 1551 } 1552 String base = src.substring(0,len); 1553 String combiningMarks = src.substring(len); 1554 String[] perms = producePermutations(combiningMarks); 1555 // Add combined permutations 1556 for(int x = 0; x < perms.length; x++) { 1557 String next = base + perms[x]; 1558 dst.add(next); 1559 next = composeOneStep(next); 1560 if (next != null) { 1561 produceEquivalentAlternation(next, dst); 1562 } 1563 } 1564 } 1565 1566 /** 1567 * Returns an array of strings that have all the possible 1568 * permutations of the characters in the input string. 1569 * This is used to get a list of all possible orderings 1570 * of a set of combining marks. Note that some of the permutations 1571 * are invalid because of combining class collisions, and these 1572 * possibilities must be removed because they are not canonically 1573 * equivalent. 1574 */ 1575 private static String[] producePermutations(String input) { 1576 if (input.length() == countChars(input, 0, 1)) 1577 return new String[] {input}; 1578 1579 if (input.length() == countChars(input, 0, 2)) { 1580 int c0 = Character.codePointAt(input, 0); 1581 int c1 = Character.codePointAt(input, Character.charCount(c0)); 1582 if (getClass(c1) == getClass(c0)) { 1583 return new String[] {input}; 1584 } 1585 String[] result = new String[2]; 1586 result[0] = input; 1587 StringBuilder sb = new StringBuilder(2); 1588 sb.appendCodePoint(c1); 1589 sb.appendCodePoint(c0); 1590 result[1] = sb.toString(); 1591 return result; 1592 } 1593 1594 int length = 1; 1595 int nCodePoints = countCodePoints(input); 1596 for(int x=1; x<nCodePoints; x++) 1597 length = length * (x+1); 1598 1599 String[] temp = new String[length]; 1600 1601 int combClass[] = new int[nCodePoints]; 1602 for(int x=0, i=0; x<nCodePoints; x++) { 1603 int c = Character.codePointAt(input, i); 1604 combClass[x] = getClass(c); 1605 i += Character.charCount(c); 1606 } 1607 1608 // For each char, take it out and add the permutations 1609 // of the remaining chars 1610 int index = 0; 1611 int len; 1612 // offset maintains the index in code units. 1613 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) { 1614 len = countChars(input, offset, 1); 1615 for(int y=x-1; y>=0; y--) { 1616 if (combClass[y] == combClass[x]) { 1617 continue loop; 1618 } 1619 } 1620 StringBuilder sb = new StringBuilder(input); 1621 String otherChars = sb.delete(offset, offset+len).toString(); 1622 String[] subResult = producePermutations(otherChars); 1623 1624 String prefix = input.substring(offset, offset+len); 1625 for (String sre : subResult) 1626 temp[index++] = prefix + sre; 1627 } 1628 String[] result = new String[index]; 1629 System.arraycopy(temp, 0, result, 0, index); 1630 return result; 1631 } 1632 1633 private static int getClass(int c) { 1634 return sun.text.Normalizer.getCombiningClass(c); 1635 } 1636 1637 /** 1638 * Attempts to compose input by combining the first character 1639 * with the first combining mark following it. Returns a String 1640 * that is the composition of the leading character with its first 1641 * combining mark followed by the remaining combining marks. Returns 1642 * null if the first two characters cannot be further composed. 1643 */ 1644 private static String composeOneStep(String input) { 1645 int len = countChars(input, 0, 2); 1646 String firstTwoCharacters = input.substring(0, len); 1647 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC); 1648 if (result.equals(firstTwoCharacters)) 1649 return null; 1650 else { 1651 String remainder = input.substring(len); 1652 return result + remainder; 1653 } 1654 } 1655 1656 /** 1657 * Preprocess any \Q...\E sequences in `temp', meta-quoting them. 1658 * See the description of `quotemeta' in perlfunc(1). 1659 */ 1660 private void RemoveQEQuoting() { 1661 final int pLen = patternLength; 1662 int i = 0; 1663 while (i < pLen-1) { 1664 if (temp[i] != '\\') 1665 i += 1; 1666 else if (temp[i + 1] != 'Q') 1667 i += 2; 1668 else 1669 break; 1670 } 1671 if (i >= pLen - 1) // No \Q sequence found 1672 return; 1673 int j = i; 1674 i += 2; 1675 int newTempLen; 1676 try { 1677 newTempLen = Math.addExact(j + 2, Math.multiplyExact(3, pLen - i)); 1678 } catch (ArithmeticException ae) { 1679 throw new OutOfMemoryError(); 1680 } 1681 int[] newtemp = new int[newTempLen]; 1682 System.arraycopy(temp, 0, newtemp, 0, j); 1683 1684 boolean inQuote = true; 1685 boolean beginQuote = true; 1686 while (i < pLen) { 1687 int c = temp[i++]; 1688 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) { 1689 newtemp[j++] = c; 1690 } else if (ASCII.isDigit(c)) { 1691 if (beginQuote) { 1692 /* 1693 * A unicode escape \[0xu] could be before this quote, 1694 * and we don't want this numeric char to processed as 1695 * part of the escape. 1696 */ 1697 newtemp[j++] = '\\'; 1698 newtemp[j++] = 'x'; 1699 newtemp[j++] = '3'; 1700 } 1701 newtemp[j++] = c; 1702 } else if (c != '\\') { 1703 if (inQuote) newtemp[j++] = '\\'; 1704 newtemp[j++] = c; 1705 } else if (inQuote) { 1706 if (temp[i] == 'E') { 1707 i++; 1708 inQuote = false; 1709 } else { 1710 newtemp[j++] = '\\'; 1711 newtemp[j++] = '\\'; 1712 } 1713 } else { 1714 if (temp[i] == 'Q') { 1715 i++; 1716 inQuote = true; 1717 beginQuote = true; 1718 continue; 1719 } else { 1720 newtemp[j++] = c; 1721 if (i != pLen) 1722 newtemp[j++] = temp[i++]; 1723 } 1724 } 1725 1726 beginQuote = false; 1727 } 1728 1729 patternLength = j; 1730 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination 1731 } 1732 1733 /** 1734 * Copies regular expression to an int array and invokes the parsing 1735 * of the expression which will create the object tree. 1736 */ 1737 private void compile() { 1738 // Handle canonical equivalences 1739 if (has(CANON_EQ) && !has(LITERAL)) { 1740 normalizedPattern = normalize(pattern); 1741 } else { 1742 normalizedPattern = pattern; 1743 } 1744 patternLength = normalizedPattern.length(); 1745 1746 // Copy pattern to int array for convenience 1747 // Use double zero to terminate pattern 1748 temp = new int[patternLength + 2]; 1749 1750 hasSupplementary = false; 1751 int c, count = 0; 1752 // Convert all chars into code points 1753 for (int x = 0; x < patternLength; x += Character.charCount(c)) { 1754 c = normalizedPattern.codePointAt(x); 1755 if (isSupplementary(c)) { 1756 hasSupplementary = true; 1757 } 1758 temp[count++] = c; 1759 } 1760 1761 patternLength = count; // patternLength now in code points 1762 1763 if (! has(LITERAL)) 1764 RemoveQEQuoting(); 1765 1766 // Allocate all temporary objects here. 1767 buffer = new int[32]; 1768 groupNodes = new GroupHead[10]; 1769 namedGroups = null; 1770 topClosureNodes = new ArrayList<>(10); 1771 1772 if (has(LITERAL)) { 1773 // Literal pattern handling 1774 matchRoot = newSlice(temp, patternLength, hasSupplementary); 1775 matchRoot.next = lastAccept; 1776 } else { 1777 // Start recursive descent parsing 1778 matchRoot = expr(lastAccept); 1779 // Check extra pattern characters 1780 if (patternLength != cursor) { 1781 if (peek() == ')') { 1782 throw error("Unmatched closing ')'"); 1783 } else { 1784 throw error("Unexpected internal error"); 1785 } 1786 } 1787 } 1788 1789 // Peephole optimization 1790 if (matchRoot instanceof Slice) { 1791 root = BnM.optimize(matchRoot); 1792 if (root == matchRoot) { 1793 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1794 } 1795 } else if (matchRoot instanceof Begin || matchRoot instanceof First) { 1796 root = matchRoot; 1797 } else { 1798 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1799 } 1800 1801 // Optimize the greedy Loop to prevent exponential backtracking, IF there 1802 // is no group ref in this pattern. With a non-negative localTCNCount value, 1803 // the greedy type Loop, Curly will skip the backtracking for any starting 1804 // position "i" that failed in the past. 1805 if (!hasGroupRef) { 1806 for (Node node : topClosureNodes) { 1807 if (node instanceof Loop) { 1808 // non-deterministic-greedy-group 1809 ((Loop)node).posIndex = localTCNCount++; 1810 } 1811 } 1812 } 1813 1814 // Release temporary storage 1815 temp = null; 1816 buffer = null; 1817 groupNodes = null; 1818 patternLength = 0; 1819 compiled = true; 1820 topClosureNodes = null; 1821 } 1822 1823 Map<String, Integer> namedGroups() { 1824 Map<String, Integer> groups = namedGroups; 1825 if (groups == null) { 1826 namedGroups = groups = new HashMap<>(2); 1827 } 1828 return groups; 1829 } 1830 1831 /** 1832 * Used to accumulate information about a subtree of the object graph 1833 * so that optimizations can be applied to the subtree. 1834 */ 1835 static final class TreeInfo { 1836 int minLength; 1837 int maxLength; 1838 boolean maxValid; 1839 boolean deterministic; 1840 1841 TreeInfo() { 1842 reset(); 1843 } 1844 void reset() { 1845 minLength = 0; 1846 maxLength = 0; 1847 maxValid = true; 1848 deterministic = true; 1849 } 1850 } 1851 1852 /* 1853 * The following private methods are mainly used to improve the 1854 * readability of the code. In order to let the Java compiler easily 1855 * inline them, we should not put many assertions or error checks in them. 1856 */ 1857 1858 /** 1859 * Indicates whether a particular flag is set or not. 1860 */ 1861 private boolean has(int f) { 1862 return (flags0 & f) != 0; 1863 } 1864 1865 /** 1866 * Match next character, signal error if failed. 1867 */ 1868 private void accept(int ch, String s) { 1869 int testChar = temp[cursor++]; 1870 if (has(COMMENTS)) 1871 testChar = parsePastWhitespace(testChar); 1872 if (ch != testChar) { 1873 throw error(s); 1874 } 1875 } 1876 1877 /** 1878 * Mark the end of pattern with a specific character. 1879 */ 1880 private void mark(int c) { 1881 temp[patternLength] = c; 1882 } 1883 1884 /** 1885 * Peek the next character, and do not advance the cursor. 1886 */ 1887 private int peek() { 1888 int ch = temp[cursor]; 1889 if (has(COMMENTS)) 1890 ch = peekPastWhitespace(ch); 1891 return ch; 1892 } 1893 1894 /** 1895 * Read the next character, and advance the cursor by one. 1896 */ 1897 private int read() { 1898 int ch = temp[cursor++]; 1899 if (has(COMMENTS)) 1900 ch = parsePastWhitespace(ch); 1901 return ch; 1902 } 1903 1904 /** 1905 * Read the next character, and advance the cursor by one, 1906 * ignoring the COMMENTS setting 1907 */ 1908 private int readEscaped() { 1909 int ch = temp[cursor++]; 1910 return ch; 1911 } 1912 1913 /** 1914 * Advance the cursor by one, and peek the next character. 1915 */ 1916 private int next() { 1917 int ch = temp[++cursor]; 1918 if (has(COMMENTS)) 1919 ch = peekPastWhitespace(ch); 1920 return ch; 1921 } 1922 1923 /** 1924 * Advance the cursor by one, and peek the next character, 1925 * ignoring the COMMENTS setting 1926 */ 1927 private int nextEscaped() { 1928 int ch = temp[++cursor]; 1929 return ch; 1930 } 1931 1932 /** 1933 * If in xmode peek past whitespace and comments. 1934 */ 1935 private int peekPastWhitespace(int ch) { 1936 while (ASCII.isSpace(ch) || ch == '#') { 1937 while (ASCII.isSpace(ch)) 1938 ch = temp[++cursor]; 1939 if (ch == '#') { 1940 ch = peekPastLine(); 1941 } 1942 } 1943 return ch; 1944 } 1945 1946 /** 1947 * If in xmode parse past whitespace and comments. 1948 */ 1949 private int parsePastWhitespace(int ch) { 1950 while (ASCII.isSpace(ch) || ch == '#') { 1951 while (ASCII.isSpace(ch)) 1952 ch = temp[cursor++]; 1953 if (ch == '#') 1954 ch = parsePastLine(); 1955 } 1956 return ch; 1957 } 1958 1959 /** 1960 * xmode parse past comment to end of line. 1961 */ 1962 private int parsePastLine() { 1963 int ch = temp[cursor++]; 1964 while (ch != 0 && !isLineSeparator(ch)) 1965 ch = temp[cursor++]; 1966 return ch; 1967 } 1968 1969 /** 1970 * xmode peek past comment to end of line. 1971 */ 1972 private int peekPastLine() { 1973 int ch = temp[++cursor]; 1974 while (ch != 0 && !isLineSeparator(ch)) 1975 ch = temp[++cursor]; 1976 return ch; 1977 } 1978 1979 /** 1980 * Determines if character is a line separator in the current mode 1981 */ 1982 private boolean isLineSeparator(int ch) { 1983 if (has(UNIX_LINES)) { 1984 return ch == '\n'; 1985 } else { 1986 return (ch == '\n' || 1987 ch == '\r' || 1988 (ch|1) == '\u2029' || 1989 ch == '\u0085'); 1990 } 1991 } 1992 1993 /** 1994 * Read the character after the next one, and advance the cursor by two. 1995 */ 1996 private int skip() { 1997 int i = cursor; 1998 int ch = temp[i+1]; 1999 cursor = i + 2; 2000 return ch; 2001 } 2002 2003 /** 2004 * Unread one next character, and retreat cursor by one. 2005 */ 2006 private void unread() { 2007 cursor--; 2008 } 2009 2010 /** 2011 * Internal method used for handling all syntax errors. The pattern is 2012 * displayed with a pointer to aid in locating the syntax error. 2013 */ 2014 private PatternSyntaxException error(String s) { 2015 return new PatternSyntaxException(s, normalizedPattern, cursor - 1); 2016 } 2017 2018 /** 2019 * Determines if there is any supplementary character or unpaired 2020 * surrogate in the specified range. 2021 */ 2022 private boolean findSupplementary(int start, int end) { 2023 for (int i = start; i < end; i++) { 2024 if (isSupplementary(temp[i])) 2025 return true; 2026 } 2027 return false; 2028 } 2029 2030 /** 2031 * Determines if the specified code point is a supplementary 2032 * character or unpaired surrogate. 2033 */ 2034 private static final boolean isSupplementary(int ch) { 2035 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || 2036 Character.isSurrogate((char)ch); 2037 } 2038 2039 /** 2040 * The following methods handle the main parsing. They are sorted 2041 * according to their precedence order, the lowest one first. 2042 */ 2043 2044 /** 2045 * The expression is parsed with branch nodes added for alternations. 2046 * This may be called recursively to parse sub expressions that may 2047 * contain alternations. 2048 */ 2049 private Node expr(Node end) { 2050 Node prev = null; 2051 Node firstTail = null; 2052 Branch branch = null; 2053 Node branchConn = null; 2054 2055 for (;;) { 2056 Node node = sequence(end); 2057 Node nodeTail = root; //double return 2058 if (prev == null) { 2059 prev = node; 2060 firstTail = nodeTail; 2061 } else { 2062 // Branch 2063 if (branchConn == null) { 2064 branchConn = new BranchConn(); 2065 branchConn.next = end; 2066 } 2067 if (node == end) { 2068 // if the node returned from sequence() is "end" 2069 // we have an empty expr, set a null atom into 2070 // the branch to indicate to go "next" directly. 2071 node = null; 2072 } else { 2073 // the "tail.next" of each atom goes to branchConn 2074 nodeTail.next = branchConn; 2075 } 2076 if (prev == branch) { 2077 branch.add(node); 2078 } else { 2079 if (prev == end) { 2080 prev = null; 2081 } else { 2082 // replace the "end" with "branchConn" at its tail.next 2083 // when put the "prev" into the branch as the first atom. 2084 firstTail.next = branchConn; 2085 } 2086 prev = branch = new Branch(prev, node, branchConn); 2087 } 2088 } 2089 if (peek() != '|') { 2090 return prev; 2091 } 2092 next(); 2093 } 2094 } 2095 2096 @SuppressWarnings("fallthrough") 2097 /** 2098 * Parsing of sequences between alternations. 2099 */ 2100 private Node sequence(Node end) { 2101 Node head = null; 2102 Node tail = null; 2103 Node node; 2104 LOOP: 2105 for (;;) { 2106 int ch = peek(); 2107 switch (ch) { 2108 case '(': 2109 // Because group handles its own closure, 2110 // we need to treat it differently 2111 node = group0(); 2112 // Check for comment or flag group 2113 if (node == null) 2114 continue; 2115 if (head == null) 2116 head = node; 2117 else 2118 tail.next = node; 2119 // Double return: Tail was returned in root 2120 tail = root; 2121 continue; 2122 case '[': 2123 if (has(CANON_EQ) && !has(LITERAL)) 2124 node = new NFCCharProperty(clazz(true)); 2125 else 2126 node = newCharProperty(clazz(true)); 2127 break; 2128 case '\\': 2129 ch = nextEscaped(); 2130 if (ch == 'p' || ch == 'P') { 2131 boolean oneLetter = true; 2132 boolean comp = (ch == 'P'); 2133 ch = next(); // Consume { if present 2134 if (ch != '{') { 2135 unread(); 2136 } else { 2137 oneLetter = false; 2138 } 2139 // node = newCharProperty(family(oneLetter, comp)); 2140 if (has(CANON_EQ) && !has(LITERAL)) 2141 node = new NFCCharProperty(family(oneLetter, comp)); 2142 else 2143 node = newCharProperty(family(oneLetter, comp)); 2144 } else { 2145 unread(); 2146 node = atom(); 2147 } 2148 break; 2149 case '^': 2150 next(); 2151 if (has(MULTILINE)) { 2152 if (has(UNIX_LINES)) 2153 node = new UnixCaret(); 2154 else 2155 node = new Caret(); 2156 } else { 2157 node = new Begin(); 2158 } 2159 break; 2160 case '$': 2161 next(); 2162 if (has(UNIX_LINES)) 2163 node = new UnixDollar(has(MULTILINE)); 2164 else 2165 node = new Dollar(has(MULTILINE)); 2166 break; 2167 case '.': 2168 next(); 2169 if (has(DOTALL)) { 2170 node = new CharProperty(ALL()); 2171 } else { 2172 if (has(UNIX_LINES)) { 2173 node = new CharProperty(UNIXDOT()); 2174 } else { 2175 node = new CharProperty(DOT()); 2176 } 2177 } 2178 break; 2179 case '|': 2180 case ')': 2181 break LOOP; 2182 case ']': // Now interpreting dangling ] and } as literals 2183 case '}': 2184 node = atom(); 2185 break; 2186 case '?': 2187 case '*': 2188 case '+': 2189 next(); 2190 throw error("Dangling meta character '" + ((char)ch) + "'"); 2191 case 0: 2192 if (cursor >= patternLength) { 2193 break LOOP; 2194 } 2195 // Fall through 2196 default: 2197 node = atom(); 2198 break; 2199 } 2200 2201 node = closure(node); 2202 /* save the top dot-greedy nodes (.*, .+) as well 2203 if (node instanceof GreedyCharProperty && 2204 ((GreedyCharProperty)node).cp instanceof Dot) { 2205 topClosureNodes.add(node); 2206 } 2207 */ 2208 if (head == null) { 2209 head = tail = node; 2210 } else { 2211 tail.next = node; 2212 tail = node; 2213 } 2214 } 2215 if (head == null) { 2216 return end; 2217 } 2218 tail.next = end; 2219 root = tail; //double return 2220 return head; 2221 } 2222 2223 @SuppressWarnings("fallthrough") 2224 /** 2225 * Parse and add a new Single or Slice. 2226 */ 2227 private Node atom() { 2228 int first = 0; 2229 int prev = -1; 2230 boolean hasSupplementary = false; 2231 int ch = peek(); 2232 for (;;) { 2233 switch (ch) { 2234 case '*': 2235 case '+': 2236 case '?': 2237 case '{': 2238 if (first > 1) { 2239 cursor = prev; // Unwind one character 2240 first--; 2241 } 2242 break; 2243 case '$': 2244 case '.': 2245 case '^': 2246 case '(': 2247 case '[': 2248 case '|': 2249 case ')': 2250 break; 2251 case '\\': 2252 ch = nextEscaped(); 2253 if (ch == 'p' || ch == 'P') { // Property 2254 if (first > 0) { // Slice is waiting; handle it first 2255 unread(); 2256 break; 2257 } else { // No slice; just return the family node 2258 boolean comp = (ch == 'P'); 2259 boolean oneLetter = true; 2260 ch = next(); // Consume { if present 2261 if (ch != '{') 2262 unread(); 2263 else 2264 oneLetter = false; 2265 if (has(CANON_EQ) && !has(LITERAL)) 2266 return new NFCCharProperty(family(oneLetter, comp)); 2267 else 2268 return newCharProperty(family(oneLetter, comp)); 2269 } 2270 } 2271 unread(); 2272 prev = cursor; 2273 ch = escape(false, first == 0, false); 2274 if (ch >= 0) { 2275 append(ch, first); 2276 first++; 2277 if (isSupplementary(ch)) { 2278 hasSupplementary = true; 2279 } 2280 ch = peek(); 2281 continue; 2282 } else if (first == 0) { 2283 return root; 2284 } 2285 // Unwind meta escape sequence 2286 cursor = prev; 2287 break; 2288 case 0: 2289 if (cursor >= patternLength) { 2290 break; 2291 } 2292 // Fall through 2293 default: 2294 prev = cursor; 2295 append(ch, first); 2296 first++; 2297 if (isSupplementary(ch)) { 2298 hasSupplementary = true; 2299 } 2300 ch = next(); 2301 continue; 2302 } 2303 break; 2304 } 2305 if (first == 1) { 2306 return newCharProperty(single(buffer[0])); 2307 } else { 2308 return newSlice(buffer, first, hasSupplementary); 2309 } 2310 } 2311 2312 private void append(int ch, int index) { 2313 int len = buffer.length; 2314 if (index - len >= 0) { 2315 len = ArraysSupport.newLength(len, 2316 1 + index - len, /* minimum growth */ 2317 len /* preferred growth */); 2318 buffer = Arrays.copyOf(buffer, len); 2319 } 2320 buffer[index] = ch; 2321 } 2322 2323 /** 2324 * Parses a backref greedily, taking as many numbers as it 2325 * can. The first digit is always treated as a backref, but 2326 * multi digit numbers are only treated as a backref if at 2327 * least that many backrefs exist at this point in the regex. 2328 */ 2329 private Node ref(int refNum) { 2330 boolean done = false; 2331 while(!done) { 2332 int ch = peek(); 2333 switch(ch) { 2334 case '0': 2335 case '1': 2336 case '2': 2337 case '3': 2338 case '4': 2339 case '5': 2340 case '6': 2341 case '7': 2342 case '8': 2343 case '9': 2344 int newRefNum = (refNum * 10) + (ch - '0'); 2345 // Add another number if it doesn't make a group 2346 // that doesn't exist 2347 if (capturingGroupCount - 1 < newRefNum) { 2348 done = true; 2349 break; 2350 } 2351 refNum = newRefNum; 2352 read(); 2353 break; 2354 default: 2355 done = true; 2356 break; 2357 } 2358 } 2359 hasGroupRef = true; 2360 if (has(CASE_INSENSITIVE)) 2361 return new CIBackRef(refNum, has(UNICODE_CASE)); 2362 else 2363 return new BackRef(refNum); 2364 } 2365 2366 /** 2367 * Parses an escape sequence to determine the actual value that needs 2368 * to be matched. 2369 * If -1 is returned and create was true a new object was added to the tree 2370 * to handle the escape sequence. 2371 * If the returned value is greater than zero, it is the value that 2372 * matches the escape sequence. 2373 */ 2374 private int escape(boolean inclass, boolean create, boolean isrange) { 2375 int ch = skip(); 2376 switch (ch) { 2377 case '0': 2378 return o(); 2379 case '1': 2380 case '2': 2381 case '3': 2382 case '4': 2383 case '5': 2384 case '6': 2385 case '7': 2386 case '8': 2387 case '9': 2388 if (inclass) break; 2389 if (create) { 2390 root = ref((ch - '0')); 2391 } 2392 return -1; 2393 case 'A': 2394 if (inclass) break; 2395 if (create) root = new Begin(); 2396 return -1; 2397 case 'B': 2398 if (inclass) break; 2399 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS)); 2400 return -1; 2401 case 'C': 2402 break; 2403 case 'D': 2404 if (create) { 2405 predicate = has(UNICODE_CHARACTER_CLASS) ? 2406 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2407 predicate = predicate.negate(); 2408 if (!inclass) 2409 root = newCharProperty(predicate); 2410 } 2411 return -1; 2412 case 'E': 2413 case 'F': 2414 break; 2415 case 'G': 2416 if (inclass) break; 2417 if (create) root = new LastMatch(); 2418 return -1; 2419 case 'H': 2420 if (create) { 2421 predicate = HorizWS().negate(); 2422 if (!inclass) 2423 root = newCharProperty(predicate); 2424 } 2425 return -1; 2426 case 'I': 2427 case 'J': 2428 case 'K': 2429 case 'L': 2430 case 'M': 2431 break; 2432 case 'N': 2433 return N(); 2434 case 'O': 2435 case 'P': 2436 case 'Q': 2437 break; 2438 case 'R': 2439 if (inclass) break; 2440 if (create) root = new LineEnding(); 2441 return -1; 2442 case 'S': 2443 if (create) { 2444 predicate = has(UNICODE_CHARACTER_CLASS) ? 2445 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2446 predicate = predicate.negate(); 2447 if (!inclass) 2448 root = newCharProperty(predicate); 2449 } 2450 return -1; 2451 case 'T': 2452 case 'U': 2453 break; 2454 case 'V': 2455 if (create) { 2456 predicate = VertWS().negate(); 2457 if (!inclass) 2458 root = newCharProperty(predicate); 2459 } 2460 return -1; 2461 case 'W': 2462 if (create) { 2463 predicate = has(UNICODE_CHARACTER_CLASS) ? 2464 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2465 predicate = predicate.negate(); 2466 if (!inclass) 2467 root = newCharProperty(predicate); 2468 } 2469 return -1; 2470 case 'X': 2471 if (inclass) break; 2472 if (create) { 2473 root = new XGrapheme(); 2474 } 2475 return -1; 2476 case 'Y': 2477 break; 2478 case 'Z': 2479 if (inclass) break; 2480 if (create) { 2481 if (has(UNIX_LINES)) 2482 root = new UnixDollar(false); 2483 else 2484 root = new Dollar(false); 2485 } 2486 return -1; 2487 case 'a': 2488 return '\007'; 2489 case 'b': 2490 if (inclass) break; 2491 if (create) { 2492 if (peek() == '{') { 2493 if (skip() == 'g') { 2494 if (read() == '}') { 2495 root = new GraphemeBound(); 2496 return -1; 2497 } 2498 break; // error missing trailing } 2499 } 2500 unread(); unread(); 2501 } 2502 root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS)); 2503 } 2504 return -1; 2505 case 'c': 2506 return c(); 2507 case 'd': 2508 if (create) { 2509 predicate = has(UNICODE_CHARACTER_CLASS) ? 2510 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2511 if (!inclass) 2512 root = newCharProperty(predicate); 2513 } 2514 return -1; 2515 case 'e': 2516 return '\033'; 2517 case 'f': 2518 return '\f'; 2519 case 'g': 2520 break; 2521 case 'h': 2522 if (create) { 2523 predicate = HorizWS(); 2524 if (!inclass) 2525 root = newCharProperty(predicate); 2526 } 2527 return -1; 2528 case 'i': 2529 case 'j': 2530 break; 2531 case 'k': 2532 if (inclass) 2533 break; 2534 if (read() != '<') 2535 throw error("\\k is not followed by '<' for named capturing group"); 2536 String name = groupname(read()); 2537 if (!namedGroups().containsKey(name)) 2538 throw error("named capturing group <" + name + "> does not exist"); 2539 if (create) { 2540 hasGroupRef = true; 2541 if (has(CASE_INSENSITIVE)) 2542 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE)); 2543 else 2544 root = new BackRef(namedGroups().get(name)); 2545 } 2546 return -1; 2547 case 'l': 2548 case 'm': 2549 break; 2550 case 'n': 2551 return '\n'; 2552 case 'o': 2553 case 'p': 2554 case 'q': 2555 break; 2556 case 'r': 2557 return '\r'; 2558 case 's': 2559 if (create) { 2560 predicate = has(UNICODE_CHARACTER_CLASS) ? 2561 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2562 if (!inclass) 2563 root = newCharProperty(predicate); 2564 } 2565 return -1; 2566 case 't': 2567 return '\t'; 2568 case 'u': 2569 return u(); 2570 case 'v': 2571 // '\v' was implemented as VT/0x0B in releases < 1.8 (though 2572 // undocumented). In JDK8 '\v' is specified as a predefined 2573 // character class for all vertical whitespace characters. 2574 // So [-1, root=VertWS node] pair is returned (instead of a 2575 // single 0x0B). This breaks the range if '\v' is used as 2576 // the start or end value, such as [\v-...] or [...-\v], in 2577 // which a single definite value (0x0B) is expected. For 2578 // compatibility concern '\013'/0x0B is returned if isrange. 2579 if (isrange) 2580 return '\013'; 2581 if (create) { 2582 predicate = VertWS(); 2583 if (!inclass) 2584 root = newCharProperty(predicate); 2585 } 2586 return -1; 2587 case 'w': 2588 if (create) { 2589 predicate = has(UNICODE_CHARACTER_CLASS) ? 2590 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2591 if (!inclass) 2592 root = newCharProperty(predicate); 2593 } 2594 return -1; 2595 case 'x': 2596 return x(); 2597 case 'y': 2598 break; 2599 case 'z': 2600 if (inclass) break; 2601 if (create) root = new End(); 2602 return -1; 2603 default: 2604 return ch; 2605 } 2606 throw error("Illegal/unsupported escape sequence"); 2607 } 2608 2609 /** 2610 * Parse a character class, and return the node that matches it. 2611 * 2612 * Consumes a ] on the way out if consume is true. Usually consume 2613 * is true except for the case of [abc&&def] where def is a separate 2614 * right hand node with "understood" brackets. 2615 */ 2616 private CharPredicate clazz(boolean consume) { 2617 CharPredicate prev = null; 2618 CharPredicate curr = null; 2619 BitClass bits = new BitClass(); 2620 2621 boolean isNeg = false; 2622 boolean hasBits = false; 2623 int ch = next(); 2624 2625 // Negates if first char in a class, otherwise literal 2626 if (ch == '^' && temp[cursor-1] == '[') { 2627 ch = next(); 2628 isNeg = true; 2629 } 2630 for (;;) { 2631 switch (ch) { 2632 case '[': 2633 curr = clazz(true); 2634 if (prev == null) 2635 prev = curr; 2636 else 2637 prev = prev.union(curr); 2638 ch = peek(); 2639 continue; 2640 case '&': 2641 ch = next(); 2642 if (ch == '&') { 2643 ch = next(); 2644 CharPredicate right = null; 2645 while (ch != ']' && ch != '&') { 2646 if (ch == '[') { 2647 if (right == null) 2648 right = clazz(true); 2649 else 2650 right = right.union(clazz(true)); 2651 } else { // abc&&def 2652 unread(); 2653 right = clazz(false); 2654 } 2655 ch = peek(); 2656 } 2657 if (hasBits) { 2658 // bits used, union has high precedence 2659 if (prev == null) { 2660 prev = curr = bits; 2661 } else { 2662 prev = prev.union(bits); 2663 } 2664 hasBits = false; 2665 } 2666 if (right != null) 2667 curr = right; 2668 if (prev == null) { 2669 if (right == null) 2670 throw error("Bad class syntax"); 2671 else 2672 prev = right; 2673 } else { 2674 prev = prev.and(curr); 2675 } 2676 } else { 2677 // treat as a literal & 2678 unread(); 2679 break; 2680 } 2681 continue; 2682 case 0: 2683 if (cursor >= patternLength) 2684 throw error("Unclosed character class"); 2685 break; 2686 case ']': 2687 if (prev != null || hasBits) { 2688 if (consume) 2689 next(); 2690 if (prev == null) 2691 prev = bits; 2692 else if (hasBits) 2693 prev = prev.union(bits); 2694 if (isNeg) 2695 return prev.negate(); 2696 return prev; 2697 } 2698 break; 2699 default: 2700 break; 2701 } 2702 curr = range(bits); 2703 if (curr == null) { // the bits used 2704 hasBits = true; 2705 } else { 2706 if (prev == null) 2707 prev = curr; 2708 else if (prev != curr) 2709 prev = prev.union(curr); 2710 } 2711 ch = peek(); 2712 } 2713 } 2714 2715 private CharPredicate bitsOrSingle(BitClass bits, int ch) { 2716 /* Bits can only handle codepoints in [u+0000-u+00ff] range. 2717 Use "single" node instead of bits when dealing with unicode 2718 case folding for codepoints listed below. 2719 (1)Uppercase out of range: u+00ff, u+00b5 2720 toUpperCase(u+00ff) -> u+0178 2721 toUpperCase(u+00b5) -> u+039c 2722 (2)LatinSmallLetterLongS u+17f 2723 toUpperCase(u+017f) -> u+0053 2724 (3)LatinSmallLetterDotlessI u+131 2725 toUpperCase(u+0131) -> u+0049 2726 (4)LatinCapitalLetterIWithDotAbove u+0130 2727 toLowerCase(u+0130) -> u+0069 2728 (5)KelvinSign u+212a 2729 toLowerCase(u+212a) ==> u+006B 2730 (6)AngstromSign u+212b 2731 toLowerCase(u+212b) ==> u+00e5 2732 */ 2733 if (ch < 256 && 2734 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) && 2735 (ch == 0xff || ch == 0xb5 || 2736 ch == 0x49 || ch == 0x69 || //I and i 2737 ch == 0x53 || ch == 0x73 || //S and s 2738 ch == 0x4b || ch == 0x6b || //K and k 2739 ch == 0xc5 || ch == 0xe5))) { //A+ring 2740 bits.add(ch, flags0); 2741 return null; 2742 } 2743 return single(ch); 2744 } 2745 2746 /** 2747 * Returns a suitably optimized, single character predicate 2748 */ 2749 private CharPredicate single(final int ch) { 2750 if (has(CASE_INSENSITIVE)) { 2751 int lower, upper; 2752 if (has(UNICODE_CASE)) { 2753 upper = Character.toUpperCase(ch); 2754 lower = Character.toLowerCase(upper); 2755 // Unicode case insensitive matches 2756 if (upper != lower) 2757 return SingleU(lower); 2758 } else if (ASCII.isAscii(ch)) { 2759 lower = ASCII.toLower(ch); 2760 upper = ASCII.toUpper(ch); 2761 // Case insensitive matches a given BMP character 2762 if (lower != upper) 2763 return SingleI(lower, upper); 2764 } 2765 } 2766 if (isSupplementary(ch)) 2767 return SingleS(ch); 2768 return Single(ch); // Match a given BMP character 2769 } 2770 2771 /** 2772 * Parse a single character or a character range in a character class 2773 * and return its representative node. 2774 */ 2775 private CharPredicate range(BitClass bits) { 2776 int ch = peek(); 2777 if (ch == '\\') { 2778 ch = nextEscaped(); 2779 if (ch == 'p' || ch == 'P') { // A property 2780 boolean comp = (ch == 'P'); 2781 boolean oneLetter = true; 2782 // Consume { if present 2783 ch = next(); 2784 if (ch != '{') 2785 unread(); 2786 else 2787 oneLetter = false; 2788 return family(oneLetter, comp); 2789 } else { // ordinary escape 2790 boolean isrange = temp[cursor+1] == '-'; 2791 unread(); 2792 ch = escape(true, true, isrange); 2793 if (ch == -1) 2794 return predicate; 2795 } 2796 } else { 2797 next(); 2798 } 2799 if (ch >= 0) { 2800 if (peek() == '-') { 2801 int endRange = temp[cursor+1]; 2802 if (endRange == '[') { 2803 return bitsOrSingle(bits, ch); 2804 } 2805 if (endRange != ']') { 2806 next(); 2807 int m = peek(); 2808 if (m == '\\') { 2809 m = escape(true, false, true); 2810 } else { 2811 next(); 2812 } 2813 if (m < ch) { 2814 throw error("Illegal character range"); 2815 } 2816 if (has(CASE_INSENSITIVE)) { 2817 if (has(UNICODE_CASE)) 2818 return CIRangeU(ch, m); 2819 return CIRange(ch, m); 2820 } else { 2821 return Range(ch, m); 2822 } 2823 } 2824 } 2825 return bitsOrSingle(bits, ch); 2826 } 2827 throw error("Unexpected character '"+((char)ch)+"'"); 2828 } 2829 2830 /** 2831 * Parses a Unicode character family and returns its representative node. 2832 */ 2833 private CharPredicate family(boolean singleLetter, boolean isComplement) { 2834 next(); 2835 String name; 2836 CharPredicate p = null; 2837 2838 if (singleLetter) { 2839 int c = temp[cursor]; 2840 if (!Character.isSupplementaryCodePoint(c)) { 2841 name = String.valueOf((char)c); 2842 } else { 2843 name = new String(temp, cursor, 1); 2844 } 2845 read(); 2846 } else { 2847 int i = cursor; 2848 mark('}'); 2849 while(read() != '}') { 2850 } 2851 mark('\000'); 2852 int j = cursor; 2853 if (j > patternLength) 2854 throw error("Unclosed character family"); 2855 if (i + 1 >= j) 2856 throw error("Empty character family"); 2857 name = new String(temp, i, j-i-1); 2858 } 2859 2860 int i = name.indexOf('='); 2861 if (i != -1) { 2862 // property construct \p{name=value} 2863 String value = name.substring(i + 1); 2864 name = name.substring(0, i).toLowerCase(Locale.ENGLISH); 2865 switch (name) { 2866 case "sc": 2867 case "script": 2868 p = CharPredicates.forUnicodeScript(value); 2869 break; 2870 case "blk": 2871 case "block": 2872 p = CharPredicates.forUnicodeBlock(value); 2873 break; 2874 case "gc": 2875 case "general_category": 2876 p = CharPredicates.forProperty(value); 2877 break; 2878 default: 2879 break; 2880 } 2881 if (p == null) 2882 throw error("Unknown Unicode property {name=<" + name + ">, " 2883 + "value=<" + value + ">}"); 2884 2885 } else { 2886 if (name.startsWith("In")) { 2887 // \p{InBlockName} 2888 p = CharPredicates.forUnicodeBlock(name.substring(2)); 2889 } else if (name.startsWith("Is")) { 2890 // \p{IsGeneralCategory} and \p{IsScriptName} 2891 name = name.substring(2); 2892 p = CharPredicates.forUnicodeProperty(name); 2893 if (p == null) 2894 p = CharPredicates.forProperty(name); 2895 if (p == null) 2896 p = CharPredicates.forUnicodeScript(name); 2897 } else { 2898 if (has(UNICODE_CHARACTER_CLASS)) { 2899 p = CharPredicates.forPOSIXName(name); 2900 } 2901 if (p == null) 2902 p = CharPredicates.forProperty(name); 2903 } 2904 if (p == null) 2905 throw error("Unknown character property name {In/Is" + name + "}"); 2906 } 2907 if (isComplement) { 2908 // it might be too expensive to detect if a complement of 2909 // CharProperty can match "certain" supplementary. So just 2910 // go with StartS. 2911 hasSupplementary = true; 2912 p = p.negate(); 2913 } 2914 return p; 2915 } 2916 2917 private CharProperty newCharProperty(CharPredicate p) { 2918 if (p == null) 2919 return null; 2920 if (p instanceof BmpCharPredicate) 2921 return new BmpCharProperty((BmpCharPredicate)p); 2922 else 2923 return new CharProperty(p); 2924 } 2925 2926 /** 2927 * Parses and returns the name of a "named capturing group", the trailing 2928 * ">" is consumed after parsing. 2929 */ 2930 private String groupname(int ch) { 2931 StringBuilder sb = new StringBuilder(); 2932 if (!ASCII.isAlpha(ch)) 2933 throw error("capturing group name does not start with a Latin letter"); 2934 do { 2935 sb.append((char) ch); 2936 } while (ASCII.isAlnum(ch=read())); 2937 if (ch != '>') 2938 throw error("named capturing group is missing trailing '>'"); 2939 return sb.toString(); 2940 } 2941 2942 /** 2943 * Parses a group and returns the head node of a set of nodes that process 2944 * the group. Sometimes a double return system is used where the tail is 2945 * returned in root. 2946 */ 2947 private Node group0() { 2948 boolean capturingGroup = false; 2949 Node head; 2950 Node tail; 2951 int save = flags0; 2952 int saveTCNCount = topClosureNodes.size(); 2953 root = null; 2954 int ch = next(); 2955 if (ch == '?') { 2956 ch = skip(); 2957 switch (ch) { 2958 case ':': // (?:xxx) pure group 2959 head = createGroup(true); 2960 tail = root; 2961 head.next = expr(tail); 2962 break; 2963 case '=': // (?=xxx) and (?!xxx) lookahead 2964 case '!': 2965 head = createGroup(true); 2966 tail = root; 2967 head.next = expr(tail); 2968 if (ch == '=') { 2969 head = tail = new Pos(head); 2970 } else { 2971 head = tail = new Neg(head); 2972 } 2973 break; 2974 case '>': // (?>xxx) independent group 2975 head = createGroup(true); 2976 tail = root; 2977 head.next = expr(tail); 2978 head = tail = new Ques(head, Qtype.INDEPENDENT); 2979 break; 2980 case '<': // (?<xxx) look behind 2981 ch = read(); 2982 if (ch != '=' && ch != '!') { 2983 // named captured group 2984 String name = groupname(ch); 2985 if (namedGroups().containsKey(name)) 2986 throw error("Named capturing group <" + name 2987 + "> is already defined"); 2988 capturingGroup = true; 2989 head = createGroup(false); 2990 tail = root; 2991 namedGroups().put(name, capturingGroupCount-1); 2992 head.next = expr(tail); 2993 break; 2994 } 2995 int start = cursor; 2996 head = createGroup(true); 2997 tail = root; 2998 head.next = expr(tail); 2999 tail.next = LookBehindEndNode.INSTANCE; 3000 TreeInfo info = new TreeInfo(); 3001 head.study(info); 3002 if (info.maxValid == false) { 3003 throw error("Look-behind group does not have " 3004 + "an obvious maximum length"); 3005 } 3006 boolean hasSupplementary = findSupplementary(start, patternLength); 3007 if (ch == '=') { 3008 head = tail = (hasSupplementary ? 3009 new BehindS(head, info.maxLength, 3010 info.minLength) : 3011 new Behind(head, info.maxLength, 3012 info.minLength)); 3013 } else { // if (ch == '!') 3014 head = tail = (hasSupplementary ? 3015 new NotBehindS(head, info.maxLength, 3016 info.minLength) : 3017 new NotBehind(head, info.maxLength, 3018 info.minLength)); 3019 } 3020 // clear all top-closure-nodes inside lookbehind 3021 if (saveTCNCount < topClosureNodes.size()) 3022 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3023 break; 3024 case '$': 3025 case '@': 3026 throw error("Unknown group type"); 3027 default: // (?xxx:) inlined match flags 3028 unread(); 3029 addFlag(); 3030 ch = read(); 3031 if (ch == ')') { 3032 return null; // Inline modifier only 3033 } 3034 if (ch != ':') { 3035 throw error("Unknown inline modifier"); 3036 } 3037 head = createGroup(true); 3038 tail = root; 3039 head.next = expr(tail); 3040 break; 3041 } 3042 } else { // (xxx) a regular group 3043 capturingGroup = true; 3044 head = createGroup(false); 3045 tail = root; 3046 head.next = expr(tail); 3047 } 3048 3049 accept(')', "Unclosed group"); 3050 flags0 = save; 3051 3052 // Check for quantifiers 3053 Node node = closure(head); 3054 if (node == head) { // No closure 3055 root = tail; 3056 return node; // Dual return 3057 } 3058 if (head == tail) { // Zero length assertion 3059 root = node; 3060 return node; // Dual return 3061 } 3062 3063 // have group closure, clear all inner closure nodes from the 3064 // top list (no backtracking stopper optimization for inner 3065 if (saveTCNCount < topClosureNodes.size()) 3066 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3067 3068 if (node instanceof Ques) { 3069 Ques ques = (Ques) node; 3070 if (ques.type == Qtype.POSSESSIVE) { 3071 root = node; 3072 return node; 3073 } 3074 tail.next = new BranchConn(); 3075 tail = tail.next; 3076 if (ques.type == Qtype.GREEDY) { 3077 head = new Branch(head, null, tail); 3078 } else { // Reluctant quantifier 3079 head = new Branch(null, head, tail); 3080 } 3081 root = tail; 3082 return head; 3083 } else if (node instanceof Curly) { 3084 Curly curly = (Curly) node; 3085 if (curly.type == Qtype.POSSESSIVE) { 3086 root = node; 3087 return node; 3088 } 3089 // Discover if the group is deterministic 3090 TreeInfo info = new TreeInfo(); 3091 if (head.study(info)) { // Deterministic 3092 GroupTail temp = (GroupTail) tail; 3093 head = root = new GroupCurly(head.next, curly.cmin, 3094 curly.cmax, curly.type, 3095 ((GroupTail)tail).localIndex, 3096 ((GroupTail)tail).groupIndex, 3097 capturingGroup); 3098 return head; 3099 } else { // Non-deterministic 3100 int temp = ((GroupHead) head).localIndex; 3101 Loop loop; 3102 if (curly.type == Qtype.GREEDY) { 3103 loop = new Loop(this.localCount, temp); 3104 // add the max_reps greedy to the top-closure-node list 3105 if (curly.cmax == MAX_REPS) 3106 topClosureNodes.add(loop); 3107 } else { // Reluctant Curly 3108 loop = new LazyLoop(this.localCount, temp); 3109 } 3110 Prolog prolog = new Prolog(loop); 3111 this.localCount += 1; 3112 loop.cmin = curly.cmin; 3113 loop.cmax = curly.cmax; 3114 loop.body = head; 3115 tail.next = loop; 3116 root = loop; 3117 return prolog; // Dual return 3118 } 3119 } 3120 throw error("Internal logic error"); 3121 } 3122 3123 /** 3124 * Create group head and tail nodes using double return. If the group is 3125 * created with anonymous true then it is a pure group and should not 3126 * affect group counting. 3127 */ 3128 private Node createGroup(boolean anonymous) { 3129 int localIndex = localCount++; 3130 int groupIndex = 0; 3131 if (!anonymous) 3132 groupIndex = capturingGroupCount++; 3133 GroupHead head = new GroupHead(localIndex); 3134 root = new GroupTail(localIndex, groupIndex); 3135 3136 // for debug/print only, head.match does NOT need the "tail" info 3137 head.tail = (GroupTail)root; 3138 3139 if (!anonymous && groupIndex < 10) 3140 groupNodes[groupIndex] = head; 3141 return head; 3142 } 3143 3144 @SuppressWarnings("fallthrough") 3145 /** 3146 * Parses inlined match flags and set them appropriately. 3147 */ 3148 private void addFlag() { 3149 int ch = peek(); 3150 for (;;) { 3151 switch (ch) { 3152 case 'i': 3153 flags0 |= CASE_INSENSITIVE; 3154 break; 3155 case 'm': 3156 flags0 |= MULTILINE; 3157 break; 3158 case 's': 3159 flags0 |= DOTALL; 3160 break; 3161 case 'd': 3162 flags0 |= UNIX_LINES; 3163 break; 3164 case 'u': 3165 flags0 |= UNICODE_CASE; 3166 break; 3167 case 'c': 3168 flags0 |= CANON_EQ; 3169 break; 3170 case 'x': 3171 flags0 |= COMMENTS; 3172 break; 3173 case 'U': 3174 flags0 |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3175 break; 3176 case '-': // subFlag then fall through 3177 ch = next(); 3178 subFlag(); 3179 default: 3180 return; 3181 } 3182 ch = next(); 3183 } 3184 } 3185 3186 @SuppressWarnings("fallthrough") 3187 /** 3188 * Parses the second part of inlined match flags and turns off 3189 * flags appropriately. 3190 */ 3191 private void subFlag() { 3192 int ch = peek(); 3193 for (;;) { 3194 switch (ch) { 3195 case 'i': 3196 flags0 &= ~CASE_INSENSITIVE; 3197 break; 3198 case 'm': 3199 flags0 &= ~MULTILINE; 3200 break; 3201 case 's': 3202 flags0 &= ~DOTALL; 3203 break; 3204 case 'd': 3205 flags0 &= ~UNIX_LINES; 3206 break; 3207 case 'u': 3208 flags0 &= ~UNICODE_CASE; 3209 break; 3210 case 'c': 3211 flags0 &= ~CANON_EQ; 3212 break; 3213 case 'x': 3214 flags0 &= ~COMMENTS; 3215 break; 3216 case 'U': 3217 flags0 &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3218 break; 3219 default: 3220 return; 3221 } 3222 ch = next(); 3223 } 3224 } 3225 3226 static final int MAX_REPS = 0x7FFFFFFF; 3227 3228 static enum Qtype { 3229 GREEDY, LAZY, POSSESSIVE, INDEPENDENT 3230 } 3231 3232 private Node curly(Node prev, int cmin) { 3233 int ch = next(); 3234 if (ch == '?') { 3235 next(); 3236 return new Curly(prev, cmin, MAX_REPS, Qtype.LAZY); 3237 } else if (ch == '+') { 3238 next(); 3239 return new Curly(prev, cmin, MAX_REPS, Qtype.POSSESSIVE); 3240 } 3241 if (prev instanceof BmpCharProperty) { 3242 return new BmpCharPropertyGreedy((BmpCharProperty)prev, cmin); 3243 } else if (prev instanceof CharProperty) { 3244 return new CharPropertyGreedy((CharProperty)prev, cmin); 3245 } 3246 return new Curly(prev, cmin, MAX_REPS, Qtype.GREEDY); 3247 } 3248 3249 /** 3250 * Processes repetition. If the next character peeked is a quantifier 3251 * then new nodes must be appended to handle the repetition. 3252 * Prev could be a single or a group, so it could be a chain of nodes. 3253 */ 3254 private Node closure(Node prev) { 3255 int ch = peek(); 3256 switch (ch) { 3257 case '?': 3258 ch = next(); 3259 if (ch == '?') { 3260 next(); 3261 return new Ques(prev, Qtype.LAZY); 3262 } else if (ch == '+') { 3263 next(); 3264 return new Ques(prev, Qtype.POSSESSIVE); 3265 } 3266 return new Ques(prev, Qtype.GREEDY); 3267 case '*': 3268 return curly(prev, 0); 3269 case '+': 3270 return curly(prev, 1); 3271 case '{': 3272 ch = skip(); 3273 if (ASCII.isDigit(ch)) { 3274 int cmin = 0, cmax; 3275 try { 3276 do { 3277 cmin = Math.addExact(Math.multiplyExact(cmin, 10), 3278 ch - '0'); 3279 } while (ASCII.isDigit(ch = read())); 3280 if (ch == ',') { 3281 ch = read(); 3282 if (ch == '}') { 3283 unread(); 3284 return curly(prev, cmin); 3285 } else { 3286 cmax = 0; 3287 while (ASCII.isDigit(ch)) { 3288 cmax = Math.addExact(Math.multiplyExact(cmax, 10), 3289 ch - '0'); 3290 ch = read(); 3291 } 3292 } 3293 } else { 3294 cmax = cmin; 3295 } 3296 } catch (ArithmeticException ae) { 3297 throw error("Illegal repetition range"); 3298 } 3299 if (ch != '}') 3300 throw error("Unclosed counted closure"); 3301 if (cmax < cmin) 3302 throw error("Illegal repetition range"); 3303 ch = peek(); 3304 if (ch == '?') { 3305 next(); 3306 return new Curly(prev, cmin, cmax, Qtype.LAZY); 3307 } else if (ch == '+') { 3308 next(); 3309 return new Curly(prev, cmin, cmax, Qtype.POSSESSIVE); 3310 } else { 3311 return new Curly(prev, cmin, cmax, Qtype.GREEDY); 3312 } 3313 } else { 3314 throw error("Illegal repetition"); 3315 } 3316 default: 3317 return prev; 3318 } 3319 } 3320 3321 /** 3322 * Utility method for parsing control escape sequences. 3323 */ 3324 private int c() { 3325 if (cursor < patternLength) { 3326 return read() ^ 64; 3327 } 3328 throw error("Illegal control escape sequence"); 3329 } 3330 3331 /** 3332 * Utility method for parsing octal escape sequences. 3333 */ 3334 private int o() { 3335 int n = read(); 3336 if (((n-'0')|('7'-n)) >= 0) { 3337 int m = read(); 3338 if (((m-'0')|('7'-m)) >= 0) { 3339 int o = read(); 3340 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) { 3341 return (n - '0') * 64 + (m - '0') * 8 + (o - '0'); 3342 } 3343 unread(); 3344 return (n - '0') * 8 + (m - '0'); 3345 } 3346 unread(); 3347 return (n - '0'); 3348 } 3349 throw error("Illegal octal escape sequence"); 3350 } 3351 3352 /** 3353 * Utility method for parsing hexadecimal escape sequences. 3354 */ 3355 private int x() { 3356 int n = read(); 3357 if (ASCII.isHexDigit(n)) { 3358 int m = read(); 3359 if (ASCII.isHexDigit(m)) { 3360 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m); 3361 } 3362 } else if (n == '{' && ASCII.isHexDigit(peek())) { 3363 int ch = 0; 3364 while (ASCII.isHexDigit(n = read())) { 3365 ch = (ch << 4) + ASCII.toDigit(n); 3366 if (ch > Character.MAX_CODE_POINT) 3367 throw error("Hexadecimal codepoint is too big"); 3368 } 3369 if (n != '}') 3370 throw error("Unclosed hexadecimal escape sequence"); 3371 return ch; 3372 } 3373 throw error("Illegal hexadecimal escape sequence"); 3374 } 3375 3376 /** 3377 * Utility method for parsing unicode escape sequences. 3378 */ 3379 private int cursor() { 3380 return cursor; 3381 } 3382 3383 private void setcursor(int pos) { 3384 cursor = pos; 3385 } 3386 3387 private int uxxxx() { 3388 int n = 0; 3389 for (int i = 0; i < 4; i++) { 3390 int ch = read(); 3391 if (!ASCII.isHexDigit(ch)) { 3392 throw error("Illegal Unicode escape sequence"); 3393 } 3394 n = n * 16 + ASCII.toDigit(ch); 3395 } 3396 return n; 3397 } 3398 3399 private int u() { 3400 int n = uxxxx(); 3401 if (Character.isHighSurrogate((char)n)) { 3402 int cur = cursor(); 3403 if (read() == '\\' && read() == 'u') { 3404 int n2 = uxxxx(); 3405 if (Character.isLowSurrogate((char)n2)) 3406 return Character.toCodePoint((char)n, (char)n2); 3407 } 3408 setcursor(cur); 3409 } 3410 return n; 3411 } 3412 3413 private int N() { 3414 if (read() == '{') { 3415 int i = cursor; 3416 while (cursor < patternLength && read() != '}') {} 3417 if (cursor > patternLength) 3418 throw error("Unclosed character name escape sequence"); 3419 String name = new String(temp, i, cursor - i - 1); 3420 try { 3421 return Character.codePointOf(name); 3422 } catch (IllegalArgumentException x) { 3423 throw error("Unknown character name [" + name + "]"); 3424 } 3425 } 3426 throw error("Illegal character name escape sequence"); 3427 } 3428 3429 // 3430 // Utility methods for code point support 3431 // 3432 private static final int countChars(CharSequence seq, int index, 3433 int lengthInCodePoints) { 3434 // optimization 3435 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) { 3436 assert (index >= 0 && index < seq.length()); 3437 return 1; 3438 } 3439 int length = seq.length(); 3440 int x = index; 3441 if (lengthInCodePoints >= 0) { 3442 assert (index >= 0 && index < length); 3443 for (int i = 0; x < length && i < lengthInCodePoints; i++) { 3444 if (Character.isHighSurrogate(seq.charAt(x++))) { 3445 if (x < length && Character.isLowSurrogate(seq.charAt(x))) { 3446 x++; 3447 } 3448 } 3449 } 3450 return x - index; 3451 } 3452 3453 assert (index >= 0 && index <= length); 3454 if (index == 0) { 3455 return 0; 3456 } 3457 int len = -lengthInCodePoints; 3458 for (int i = 0; x > 0 && i < len; i++) { 3459 if (Character.isLowSurrogate(seq.charAt(--x))) { 3460 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) { 3461 x--; 3462 } 3463 } 3464 } 3465 return index - x; 3466 } 3467 3468 private static final int countCodePoints(CharSequence seq) { 3469 int length = seq.length(); 3470 int n = 0; 3471 for (int i = 0; i < length; ) { 3472 n++; 3473 if (Character.isHighSurrogate(seq.charAt(i++))) { 3474 if (i < length && Character.isLowSurrogate(seq.charAt(i))) { 3475 i++; 3476 } 3477 } 3478 } 3479 return n; 3480 } 3481 3482 /** 3483 * Creates a bit vector for matching Latin-1 values. A normal BitClass 3484 * never matches values above Latin-1, and a complemented BitClass always 3485 * matches values above Latin-1. 3486 */ 3487 static final class BitClass implements BmpCharPredicate { 3488 final boolean[] bits; 3489 BitClass() { 3490 bits = new boolean[256]; 3491 } 3492 BitClass add(int c, int flags) { 3493 assert c >= 0 && c <= 255; 3494 if ((flags & CASE_INSENSITIVE) != 0) { 3495 if (ASCII.isAscii(c)) { 3496 bits[ASCII.toUpper(c)] = true; 3497 bits[ASCII.toLower(c)] = true; 3498 } else if ((flags & UNICODE_CASE) != 0) { 3499 bits[Character.toLowerCase(c)] = true; 3500 bits[Character.toUpperCase(c)] = true; 3501 } 3502 } 3503 bits[c] = true; 3504 return this; 3505 } 3506 public boolean is(int ch) { 3507 return ch < 256 && bits[ch]; 3508 } 3509 } 3510 3511 3512 /** 3513 * Utility method for creating a string slice matcher. 3514 */ 3515 private Node newSlice(int[] buf, int count, boolean hasSupplementary) { 3516 int[] tmp = new int[count]; 3517 if (has(CASE_INSENSITIVE)) { 3518 if (has(UNICODE_CASE)) { 3519 for (int i = 0; i < count; i++) { 3520 tmp[i] = Character.toLowerCase( 3521 Character.toUpperCase(buf[i])); 3522 } 3523 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp); 3524 } 3525 for (int i = 0; i < count; i++) { 3526 tmp[i] = ASCII.toLower(buf[i]); 3527 } 3528 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp); 3529 } 3530 for (int i = 0; i < count; i++) { 3531 tmp[i] = buf[i]; 3532 } 3533 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp); 3534 } 3535 3536 /** 3537 * The following classes are the building components of the object 3538 * tree that represents a compiled regular expression. The object tree 3539 * is made of individual elements that handle constructs in the Pattern. 3540 * Each type of object knows how to match its equivalent construct with 3541 * the match() method. 3542 */ 3543 3544 /** 3545 * Base class for all node classes. Subclasses should override the match() 3546 * method as appropriate. This class is an accepting node, so its match() 3547 * always returns true. 3548 */ 3549 static class Node extends Object { 3550 Node next; 3551 Node() { 3552 next = Pattern.accept; 3553 } 3554 /** 3555 * This method implements the classic accept node. 3556 */ 3557 boolean match(Matcher matcher, int i, CharSequence seq) { 3558 matcher.last = i; 3559 matcher.groups[0] = matcher.first; 3560 matcher.groups[1] = matcher.last; 3561 return true; 3562 } 3563 /** 3564 * This method is good for all zero length assertions. 3565 */ 3566 boolean study(TreeInfo info) { 3567 if (next != null) { 3568 return next.study(info); 3569 } else { 3570 return info.deterministic; 3571 } 3572 } 3573 } 3574 3575 static class LastNode extends Node { 3576 /** 3577 * This method implements the classic accept node with 3578 * the addition of a check to see if the match occurred 3579 * using all of the input. 3580 */ 3581 boolean match(Matcher matcher, int i, CharSequence seq) { 3582 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to) 3583 return false; 3584 matcher.last = i; 3585 matcher.groups[0] = matcher.first; 3586 matcher.groups[1] = matcher.last; 3587 return true; 3588 } 3589 } 3590 3591 /** 3592 * Used for REs that can start anywhere within the input string. 3593 * This basically tries to match repeatedly at each spot in the 3594 * input string, moving forward after each try. An anchored search 3595 * or a BnM will bypass this node completely. 3596 */ 3597 static class Start extends Node { 3598 int minLength; 3599 Start(Node node) { 3600 this.next = node; 3601 TreeInfo info = new TreeInfo(); 3602 next.study(info); 3603 minLength = info.minLength; 3604 } 3605 boolean match(Matcher matcher, int i, CharSequence seq) { 3606 if (i > matcher.to - minLength) { 3607 matcher.hitEnd = true; 3608 return false; 3609 } 3610 int guard = matcher.to - minLength; 3611 for (; i <= guard; i++) { 3612 if (next.match(matcher, i, seq)) { 3613 matcher.first = i; 3614 matcher.groups[0] = matcher.first; 3615 matcher.groups[1] = matcher.last; 3616 return true; 3617 } 3618 } 3619 matcher.hitEnd = true; 3620 return false; 3621 } 3622 boolean study(TreeInfo info) { 3623 next.study(info); 3624 info.maxValid = false; 3625 info.deterministic = false; 3626 return false; 3627 } 3628 } 3629 3630 /* 3631 * StartS supports supplementary characters, including unpaired surrogates. 3632 */ 3633 static final class StartS extends Start { 3634 StartS(Node node) { 3635 super(node); 3636 } 3637 boolean match(Matcher matcher, int i, CharSequence seq) { 3638 if (i > matcher.to - minLength) { 3639 matcher.hitEnd = true; 3640 return false; 3641 } 3642 int guard = matcher.to - minLength; 3643 while (i <= guard) { 3644 //if ((ret = next.match(matcher, i, seq)) || i == guard) 3645 if (next.match(matcher, i, seq)) { 3646 matcher.first = i; 3647 matcher.groups[0] = matcher.first; 3648 matcher.groups[1] = matcher.last; 3649 return true; 3650 } 3651 if (i == guard) 3652 break; 3653 // Optimization to move to the next character. This is 3654 // faster than countChars(seq, i, 1). 3655 if (Character.isHighSurrogate(seq.charAt(i++))) { 3656 if (i < seq.length() && 3657 Character.isLowSurrogate(seq.charAt(i))) { 3658 i++; 3659 } 3660 } 3661 } 3662 matcher.hitEnd = true; 3663 return false; 3664 } 3665 } 3666 3667 /** 3668 * Node to anchor at the beginning of input. This object implements the 3669 * match for a \A sequence, and the caret anchor will use this if not in 3670 * multiline mode. 3671 */ 3672 static final class Begin extends Node { 3673 boolean match(Matcher matcher, int i, CharSequence seq) { 3674 int fromIndex = (matcher.anchoringBounds) ? 3675 matcher.from : 0; 3676 if (i == fromIndex && next.match(matcher, i, seq)) { 3677 matcher.first = i; 3678 matcher.groups[0] = i; 3679 matcher.groups[1] = matcher.last; 3680 return true; 3681 } else { 3682 return false; 3683 } 3684 } 3685 } 3686 3687 /** 3688 * Node to anchor at the end of input. This is the absolute end, so this 3689 * should not match at the last newline before the end as $ will. 3690 */ 3691 static final class End extends Node { 3692 boolean match(Matcher matcher, int i, CharSequence seq) { 3693 int endIndex = (matcher.anchoringBounds) ? 3694 matcher.to : matcher.getTextLength(); 3695 if (i == endIndex) { 3696 matcher.hitEnd = true; 3697 return next.match(matcher, i, seq); 3698 } 3699 return false; 3700 } 3701 } 3702 3703 /** 3704 * Node to anchor at the beginning of a line. This is essentially the 3705 * object to match for the multiline ^. 3706 */ 3707 static final class Caret extends Node { 3708 boolean match(Matcher matcher, int i, CharSequence seq) { 3709 int startIndex = matcher.from; 3710 int endIndex = matcher.to; 3711 if (!matcher.anchoringBounds) { 3712 startIndex = 0; 3713 endIndex = matcher.getTextLength(); 3714 } 3715 // Perl does not match ^ at end of input even after newline 3716 if (i == endIndex) { 3717 matcher.hitEnd = true; 3718 return false; 3719 } 3720 if (i > startIndex) { 3721 char ch = seq.charAt(i-1); 3722 if (ch != '\n' && ch != '\r' 3723 && (ch|1) != '\u2029' 3724 && ch != '\u0085' ) { 3725 return false; 3726 } 3727 // Should treat /r/n as one newline 3728 if (ch == '\r' && seq.charAt(i) == '\n') 3729 return false; 3730 } 3731 return next.match(matcher, i, seq); 3732 } 3733 } 3734 3735 /** 3736 * Node to anchor at the beginning of a line when in unixdot mode. 3737 */ 3738 static final class UnixCaret extends Node { 3739 boolean match(Matcher matcher, int i, CharSequence seq) { 3740 int startIndex = matcher.from; 3741 int endIndex = matcher.to; 3742 if (!matcher.anchoringBounds) { 3743 startIndex = 0; 3744 endIndex = matcher.getTextLength(); 3745 } 3746 // Perl does not match ^ at end of input even after newline 3747 if (i == endIndex) { 3748 matcher.hitEnd = true; 3749 return false; 3750 } 3751 if (i > startIndex) { 3752 char ch = seq.charAt(i-1); 3753 if (ch != '\n') { 3754 return false; 3755 } 3756 } 3757 return next.match(matcher, i, seq); 3758 } 3759 } 3760 3761 /** 3762 * Node to match the location where the last match ended. 3763 * This is used for the \G construct. 3764 */ 3765 static final class LastMatch extends Node { 3766 boolean match(Matcher matcher, int i, CharSequence seq) { 3767 if (i != matcher.oldLast) 3768 return false; 3769 return next.match(matcher, i, seq); 3770 } 3771 } 3772 3773 /** 3774 * Node to anchor at the end of a line or the end of input based on the 3775 * multiline mode. 3776 * 3777 * When not in multiline mode, the $ can only match at the very end 3778 * of the input, unless the input ends in a line terminator in which 3779 * it matches right before the last line terminator. 3780 * 3781 * Note that \r\n is considered an atomic line terminator. 3782 * 3783 * Like ^ the $ operator matches at a position, it does not match the 3784 * line terminators themselves. 3785 */ 3786 static final class Dollar extends Node { 3787 boolean multiline; 3788 Dollar(boolean mul) { 3789 multiline = mul; 3790 } 3791 boolean match(Matcher matcher, int i, CharSequence seq) { 3792 int endIndex = (matcher.anchoringBounds) ? 3793 matcher.to : matcher.getTextLength(); 3794 if (!multiline) { 3795 if (i < endIndex - 2) 3796 return false; 3797 if (i == endIndex - 2) { 3798 char ch = seq.charAt(i); 3799 if (ch != '\r') 3800 return false; 3801 ch = seq.charAt(i + 1); 3802 if (ch != '\n') 3803 return false; 3804 } 3805 } 3806 // Matches before any line terminator; also matches at the 3807 // end of input 3808 // Before line terminator: 3809 // If multiline, we match here no matter what 3810 // If not multiline, fall through so that the end 3811 // is marked as hit; this must be a /r/n or a /n 3812 // at the very end so the end was hit; more input 3813 // could make this not match here 3814 if (i < endIndex) { 3815 char ch = seq.charAt(i); 3816 if (ch == '\n') { 3817 // No match between \r\n 3818 if (i > 0 && seq.charAt(i-1) == '\r') 3819 return false; 3820 if (multiline) 3821 return next.match(matcher, i, seq); 3822 } else if (ch == '\r' || ch == '\u0085' || 3823 (ch|1) == '\u2029') { 3824 if (multiline) 3825 return next.match(matcher, i, seq); 3826 } else { // No line terminator, no match 3827 return false; 3828 } 3829 } 3830 // Matched at current end so hit end 3831 matcher.hitEnd = true; 3832 // If a $ matches because of end of input, then more input 3833 // could cause it to fail! 3834 matcher.requireEnd = true; 3835 return next.match(matcher, i, seq); 3836 } 3837 boolean study(TreeInfo info) { 3838 next.study(info); 3839 return info.deterministic; 3840 } 3841 } 3842 3843 /** 3844 * Node to anchor at the end of a line or the end of input based on the 3845 * multiline mode when in unix lines mode. 3846 */ 3847 static final class UnixDollar extends Node { 3848 boolean multiline; 3849 UnixDollar(boolean mul) { 3850 multiline = mul; 3851 } 3852 boolean match(Matcher matcher, int i, CharSequence seq) { 3853 int endIndex = (matcher.anchoringBounds) ? 3854 matcher.to : matcher.getTextLength(); 3855 if (i < endIndex) { 3856 char ch = seq.charAt(i); 3857 if (ch == '\n') { 3858 // If not multiline, then only possible to 3859 // match at very end or one before end 3860 if (multiline == false && i != endIndex - 1) 3861 return false; 3862 // If multiline return next.match without setting 3863 // matcher.hitEnd 3864 if (multiline) 3865 return next.match(matcher, i, seq); 3866 } else { 3867 return false; 3868 } 3869 } 3870 // Matching because at the end or 1 before the end; 3871 // more input could change this so set hitEnd 3872 matcher.hitEnd = true; 3873 // If a $ matches because of end of input, then more input 3874 // could cause it to fail! 3875 matcher.requireEnd = true; 3876 return next.match(matcher, i, seq); 3877 } 3878 boolean study(TreeInfo info) { 3879 next.study(info); 3880 return info.deterministic; 3881 } 3882 } 3883 3884 /** 3885 * Node class that matches a Unicode line ending '\R' 3886 */ 3887 static final class LineEnding extends Node { 3888 boolean match(Matcher matcher, int i, CharSequence seq) { 3889 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029]) 3890 if (i < matcher.to) { 3891 int ch = seq.charAt(i); 3892 if (ch == 0x0A || ch == 0x0B || ch == 0x0C || 3893 ch == 0x85 || ch == 0x2028 || ch == 0x2029) 3894 return next.match(matcher, i + 1, seq); 3895 if (ch == 0x0D) { 3896 i++; 3897 if (i < matcher.to) { 3898 if (seq.charAt(i) == 0x0A && 3899 next.match(matcher, i + 1, seq)) { 3900 return true; 3901 } 3902 } else { 3903 matcher.hitEnd = true; 3904 } 3905 return next.match(matcher, i, seq); 3906 } 3907 } else { 3908 matcher.hitEnd = true; 3909 } 3910 return false; 3911 } 3912 boolean study(TreeInfo info) { 3913 info.minLength++; 3914 info.maxLength += 2; 3915 return next.study(info); 3916 } 3917 } 3918 3919 /** 3920 * Abstract node class to match one character satisfying some 3921 * boolean property. 3922 */ 3923 static class CharProperty extends Node { 3924 final CharPredicate predicate; 3925 3926 CharProperty (CharPredicate predicate) { 3927 this.predicate = predicate; 3928 } 3929 boolean match(Matcher matcher, int i, CharSequence seq) { 3930 if (i < matcher.to) { 3931 int ch = Character.codePointAt(seq, i); 3932 return predicate.is(ch) && 3933 next.match(matcher, i + Character.charCount(ch), seq); 3934 } else { 3935 matcher.hitEnd = true; 3936 return false; 3937 } 3938 } 3939 boolean study(TreeInfo info) { 3940 info.minLength++; 3941 info.maxLength++; 3942 return next.study(info); 3943 } 3944 } 3945 3946 /** 3947 * Optimized version of CharProperty that works only for 3948 * properties never satisfied by Supplementary characters. 3949 */ 3950 private static class BmpCharProperty extends CharProperty { 3951 BmpCharProperty (BmpCharPredicate predicate) { 3952 super(predicate); 3953 } 3954 boolean match(Matcher matcher, int i, CharSequence seq) { 3955 if (i < matcher.to) { 3956 return predicate.is(seq.charAt(i)) && 3957 next.match(matcher, i + 1, seq); 3958 } else { 3959 matcher.hitEnd = true; 3960 return false; 3961 } 3962 } 3963 } 3964 3965 private static class NFCCharProperty extends Node { 3966 CharPredicate predicate; 3967 NFCCharProperty (CharPredicate predicate) { 3968 this.predicate = predicate; 3969 } 3970 3971 boolean match(Matcher matcher, int i, CharSequence seq) { 3972 if (i < matcher.to) { 3973 int ch0 = Character.codePointAt(seq, i); 3974 int n = Character.charCount(ch0); 3975 int j = i + n; 3976 // Fast check if it's necessary to call Normalizer; 3977 // testing Grapheme.isBoundary is enough for this case 3978 while (j < matcher.to) { 3979 int ch1 = Character.codePointAt(seq, j); 3980 if (Grapheme.isBoundary(ch0, ch1)) 3981 break; 3982 ch0 = ch1; 3983 j += Character.charCount(ch1); 3984 } 3985 if (i + n == j) { // single, assume nfc cp 3986 if (predicate.is(ch0)) 3987 return next.match(matcher, j, seq); 3988 } else { 3989 while (i + n < j) { 3990 String nfc = Normalizer.normalize( 3991 seq.toString().substring(i, j), Normalizer.Form.NFC); 3992 if (nfc.codePointCount(0, nfc.length()) == 1) { 3993 if (predicate.is(nfc.codePointAt(0)) && 3994 next.match(matcher, j, seq)) { 3995 return true; 3996 } 3997 } 3998 3999 ch0 = Character.codePointBefore(seq, j); 4000 j -= Character.charCount(ch0); 4001 } 4002 } 4003 if (j < matcher.to) 4004 return false; 4005 } 4006 matcher.hitEnd = true; 4007 return false; 4008 } 4009 4010 boolean study(TreeInfo info) { 4011 info.minLength++; 4012 info.deterministic = false; 4013 return next.study(info); 4014 } 4015 } 4016 4017 /** 4018 * Node class that matches an unicode extended grapheme cluster 4019 */ 4020 static class XGrapheme extends Node { 4021 boolean match(Matcher matcher, int i, CharSequence seq) { 4022 if (i < matcher.to) { 4023 i = Grapheme.nextBoundary(seq, i, matcher.to); 4024 return next.match(matcher, i, seq); 4025 } 4026 matcher.hitEnd = true; 4027 return false; 4028 } 4029 4030 boolean study(TreeInfo info) { 4031 info.minLength++; 4032 info.deterministic = false; 4033 return next.study(info); 4034 } 4035 } 4036 4037 /** 4038 * Node class that handles grapheme boundaries 4039 */ 4040 static class GraphemeBound extends Node { 4041 boolean match(Matcher matcher, int i, CharSequence seq) { 4042 int startIndex = matcher.from; 4043 int endIndex = matcher.to; 4044 if (matcher.transparentBounds) { 4045 startIndex = 0; 4046 endIndex = matcher.getTextLength(); 4047 } 4048 if (i == startIndex) { 4049 return next.match(matcher, i, seq); 4050 } 4051 if (i < endIndex) { 4052 if (Character.isSurrogatePair(seq.charAt(i-1), seq.charAt(i)) || 4053 Grapheme.nextBoundary(seq, 4054 i - Character.charCount(Character.codePointBefore(seq, i)), 4055 i + Character.charCount(Character.codePointAt(seq, i))) > i) { 4056 return false; 4057 } 4058 } else { 4059 matcher.hitEnd = true; 4060 matcher.requireEnd = true; 4061 } 4062 return next.match(matcher, i, seq); 4063 } 4064 } 4065 4066 /** 4067 * Base class for all Slice nodes 4068 */ 4069 static class SliceNode extends Node { 4070 int[] buffer; 4071 SliceNode(int[] buf) { 4072 buffer = buf; 4073 } 4074 boolean study(TreeInfo info) { 4075 info.minLength += buffer.length; 4076 info.maxLength += buffer.length; 4077 return next.study(info); 4078 } 4079 } 4080 4081 /** 4082 * Node class for a case sensitive/BMP-only sequence of literal 4083 * characters. 4084 */ 4085 static class Slice extends SliceNode { 4086 Slice(int[] buf) { 4087 super(buf); 4088 } 4089 boolean match(Matcher matcher, int i, CharSequence seq) { 4090 int[] buf = buffer; 4091 int len = buf.length; 4092 for (int j=0; j<len; j++) { 4093 if ((i+j) >= matcher.to) { 4094 matcher.hitEnd = true; 4095 return false; 4096 } 4097 if (buf[j] != seq.charAt(i+j)) 4098 return false; 4099 } 4100 return next.match(matcher, i+len, seq); 4101 } 4102 } 4103 4104 /** 4105 * Node class for a case_insensitive/BMP-only sequence of literal 4106 * characters. 4107 */ 4108 static class SliceI extends SliceNode { 4109 SliceI(int[] buf) { 4110 super(buf); 4111 } 4112 boolean match(Matcher matcher, int i, CharSequence seq) { 4113 int[] buf = buffer; 4114 int len = buf.length; 4115 for (int j=0; j<len; j++) { 4116 if ((i+j) >= matcher.to) { 4117 matcher.hitEnd = true; 4118 return false; 4119 } 4120 int c = seq.charAt(i+j); 4121 if (buf[j] != c && 4122 buf[j] != ASCII.toLower(c)) 4123 return false; 4124 } 4125 return next.match(matcher, i+len, seq); 4126 } 4127 } 4128 4129 /** 4130 * Node class for a unicode_case_insensitive/BMP-only sequence of 4131 * literal characters. Uses unicode case folding. 4132 */ 4133 static final class SliceU extends SliceNode { 4134 SliceU(int[] buf) { 4135 super(buf); 4136 } 4137 boolean match(Matcher matcher, int i, CharSequence seq) { 4138 int[] buf = buffer; 4139 int len = buf.length; 4140 for (int j=0; j<len; j++) { 4141 if ((i+j) >= matcher.to) { 4142 matcher.hitEnd = true; 4143 return false; 4144 } 4145 int c = seq.charAt(i+j); 4146 if (buf[j] != c && 4147 buf[j] != Character.toLowerCase(Character.toUpperCase(c))) 4148 return false; 4149 } 4150 return next.match(matcher, i+len, seq); 4151 } 4152 } 4153 4154 /** 4155 * Node class for a case sensitive sequence of literal characters 4156 * including supplementary characters. 4157 */ 4158 static final class SliceS extends Slice { 4159 SliceS(int[] buf) { 4160 super(buf); 4161 } 4162 boolean match(Matcher matcher, int i, CharSequence seq) { 4163 int[] buf = buffer; 4164 int x = i; 4165 for (int j = 0; j < buf.length; j++) { 4166 if (x >= matcher.to) { 4167 matcher.hitEnd = true; 4168 return false; 4169 } 4170 int c = Character.codePointAt(seq, x); 4171 if (buf[j] != c) 4172 return false; 4173 x += Character.charCount(c); 4174 if (x > matcher.to) { 4175 matcher.hitEnd = true; 4176 return false; 4177 } 4178 } 4179 return next.match(matcher, x, seq); 4180 } 4181 } 4182 4183 /** 4184 * Node class for a case insensitive sequence of literal characters 4185 * including supplementary characters. 4186 */ 4187 static class SliceIS extends SliceNode { 4188 SliceIS(int[] buf) { 4189 super(buf); 4190 } 4191 int toLower(int c) { 4192 return ASCII.toLower(c); 4193 } 4194 boolean match(Matcher matcher, int i, CharSequence seq) { 4195 int[] buf = buffer; 4196 int x = i; 4197 for (int j = 0; j < buf.length; j++) { 4198 if (x >= matcher.to) { 4199 matcher.hitEnd = true; 4200 return false; 4201 } 4202 int c = Character.codePointAt(seq, x); 4203 if (buf[j] != c && buf[j] != toLower(c)) 4204 return false; 4205 x += Character.charCount(c); 4206 if (x > matcher.to) { 4207 matcher.hitEnd = true; 4208 return false; 4209 } 4210 } 4211 return next.match(matcher, x, seq); 4212 } 4213 } 4214 4215 /** 4216 * Node class for a case insensitive sequence of literal characters. 4217 * Uses unicode case folding. 4218 */ 4219 static final class SliceUS extends SliceIS { 4220 SliceUS(int[] buf) { 4221 super(buf); 4222 } 4223 int toLower(int c) { 4224 return Character.toLowerCase(Character.toUpperCase(c)); 4225 } 4226 } 4227 4228 /** 4229 * The 0 or 1 quantifier. This one class implements all three types. 4230 */ 4231 static final class Ques extends Node { 4232 Node atom; 4233 Qtype type; 4234 Ques(Node node, Qtype type) { 4235 this.atom = node; 4236 this.type = type; 4237 } 4238 boolean match(Matcher matcher, int i, CharSequence seq) { 4239 switch (type) { 4240 case GREEDY: 4241 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)) 4242 || next.match(matcher, i, seq); 4243 case LAZY: 4244 return next.match(matcher, i, seq) 4245 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)); 4246 case POSSESSIVE: 4247 if (atom.match(matcher, i, seq)) i = matcher.last; 4248 return next.match(matcher, i, seq); 4249 default: 4250 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq); 4251 } 4252 } 4253 boolean study(TreeInfo info) { 4254 if (type != Qtype.INDEPENDENT) { 4255 int minL = info.minLength; 4256 atom.study(info); 4257 info.minLength = minL; 4258 info.deterministic = false; 4259 return next.study(info); 4260 } else { 4261 atom.study(info); 4262 return next.study(info); 4263 } 4264 } 4265 } 4266 4267 /** 4268 * Handles the greedy style repetition with the specified minimum 4269 * and the maximum equal to MAX_REPS, for *, + and {N,} quantifiers. 4270 */ 4271 static class CharPropertyGreedy extends Node { 4272 final CharPredicate predicate; 4273 final int cmin; 4274 4275 CharPropertyGreedy(CharProperty cp, int cmin) { 4276 this.predicate = cp.predicate; 4277 this.cmin = cmin; 4278 } 4279 boolean match(Matcher matcher, int i, CharSequence seq) { 4280 int n = 0; 4281 int to = matcher.to; 4282 // greedy, all the way down 4283 while (i < to) { 4284 int ch = Character.codePointAt(seq, i); 4285 if (!predicate.is(ch)) 4286 break; 4287 i += Character.charCount(ch); 4288 n++; 4289 } 4290 if (i >= to) { 4291 matcher.hitEnd = true; 4292 } 4293 while (n >= cmin) { 4294 if (next.match(matcher, i, seq)) 4295 return true; 4296 if (n == cmin) 4297 return false; 4298 // backing off if match fails 4299 int ch = Character.codePointBefore(seq, i); 4300 i -= Character.charCount(ch); 4301 n--; 4302 } 4303 return false; 4304 } 4305 4306 boolean study(TreeInfo info) { 4307 info.minLength += cmin; 4308 if (info.maxValid) { 4309 info.maxLength += MAX_REPS; 4310 } 4311 info.deterministic = false; 4312 return next.study(info); 4313 } 4314 } 4315 4316 static final class BmpCharPropertyGreedy extends CharPropertyGreedy { 4317 4318 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) { 4319 super(bcp, cmin); 4320 } 4321 4322 boolean match(Matcher matcher, int i, CharSequence seq) { 4323 int n = 0; 4324 int to = matcher.to; 4325 while (i < to && predicate.is(seq.charAt(i))) { 4326 i++; n++; 4327 } 4328 if (i >= to) { 4329 matcher.hitEnd = true; 4330 } 4331 while (n >= cmin) { 4332 if (next.match(matcher, i, seq)) 4333 return true; 4334 i--; n--; // backing off if match fails 4335 } 4336 return false; 4337 } 4338 } 4339 4340 /** 4341 * Handles the curly-brace style repetition with a specified minimum and 4342 * maximum occurrences. The * quantifier is handled as a special case. 4343 * This class handles the three types. 4344 */ 4345 static final class Curly extends Node { 4346 Node atom; 4347 Qtype type; 4348 int cmin; 4349 int cmax; 4350 4351 Curly(Node node, int cmin, int cmax, Qtype type) { 4352 this.atom = node; 4353 this.type = type; 4354 this.cmin = cmin; 4355 this.cmax = cmax; 4356 } 4357 boolean match(Matcher matcher, int i, CharSequence seq) { 4358 int j; 4359 for (j = 0; j < cmin; j++) { 4360 if (atom.match(matcher, i, seq)) { 4361 i = matcher.last; 4362 continue; 4363 } 4364 return false; 4365 } 4366 if (type == Qtype.GREEDY) 4367 return match0(matcher, i, j, seq); 4368 else if (type == Qtype.LAZY) 4369 return match1(matcher, i, j, seq); 4370 else 4371 return match2(matcher, i, j, seq); 4372 } 4373 // Greedy match. 4374 // i is the index to start matching at 4375 // j is the number of atoms that have matched 4376 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4377 if (j >= cmax) { 4378 // We have matched the maximum... continue with the rest of 4379 // the regular expression 4380 return next.match(matcher, i, seq); 4381 } 4382 int backLimit = j; 4383 while (atom.match(matcher, i, seq)) { 4384 // k is the length of this match 4385 int k = matcher.last - i; 4386 if (k == 0) // Zero length match 4387 break; 4388 // Move up index and number matched 4389 i = matcher.last; 4390 j++; 4391 // We are greedy so match as many as we can 4392 while (j < cmax) { 4393 if (!atom.match(matcher, i, seq)) 4394 break; 4395 if (i + k != matcher.last) { 4396 if (match0(matcher, matcher.last, j+1, seq)) 4397 return true; 4398 break; 4399 } 4400 i += k; 4401 j++; 4402 } 4403 // Handle backing off if match fails 4404 while (j >= backLimit) { 4405 if (next.match(matcher, i, seq)) 4406 return true; 4407 i -= k; 4408 j--; 4409 } 4410 return false; 4411 } 4412 return next.match(matcher, i, seq); 4413 } 4414 // Reluctant match. At this point, the minimum has been satisfied. 4415 // i is the index to start matching at 4416 // j is the number of atoms that have matched 4417 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4418 for (;;) { 4419 // Try finishing match without consuming any more 4420 if (next.match(matcher, i, seq)) 4421 return true; 4422 // At the maximum, no match found 4423 if (j >= cmax) 4424 return false; 4425 // Okay, must try one more atom 4426 if (!atom.match(matcher, i, seq)) 4427 return false; 4428 // If we haven't moved forward then must break out 4429 if (i == matcher.last) 4430 return false; 4431 // Move up index and number matched 4432 i = matcher.last; 4433 j++; 4434 } 4435 } 4436 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4437 for (; j < cmax; j++) { 4438 if (!atom.match(matcher, i, seq)) 4439 break; 4440 if (i == matcher.last) 4441 break; 4442 i = matcher.last; 4443 } 4444 return next.match(matcher, i, seq); 4445 } 4446 boolean study(TreeInfo info) { 4447 // Save original info 4448 int minL = info.minLength; 4449 int maxL = info.maxLength; 4450 boolean maxV = info.maxValid; 4451 boolean detm = info.deterministic; 4452 info.reset(); 4453 4454 atom.study(info); 4455 4456 int temp = info.minLength * cmin + minL; 4457 if (temp < minL) { 4458 temp = 0xFFFFFFF; // arbitrary large number 4459 } 4460 info.minLength = temp; 4461 4462 if (maxV & info.maxValid) { 4463 temp = info.maxLength * cmax + maxL; 4464 info.maxLength = temp; 4465 if (temp < maxL) { 4466 info.maxValid = false; 4467 } 4468 } else { 4469 info.maxValid = false; 4470 } 4471 4472 if (info.deterministic && cmin == cmax) 4473 info.deterministic = detm; 4474 else 4475 info.deterministic = false; 4476 return next.study(info); 4477 } 4478 } 4479 4480 /** 4481 * Handles the curly-brace style repetition with a specified minimum and 4482 * maximum occurrences in deterministic cases. This is an iterative 4483 * optimization over the Prolog and Loop system which would handle this 4484 * in a recursive way. The * quantifier is handled as a special case. 4485 * If capture is true then this class saves group settings and ensures 4486 * that groups are unset when backing off of a group match. 4487 */ 4488 static final class GroupCurly extends Node { 4489 Node atom; 4490 Qtype type; 4491 int cmin; 4492 int cmax; 4493 int localIndex; 4494 int groupIndex; 4495 boolean capture; 4496 4497 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local, 4498 int group, boolean capture) { 4499 this.atom = node; 4500 this.type = type; 4501 this.cmin = cmin; 4502 this.cmax = cmax; 4503 this.localIndex = local; 4504 this.groupIndex = group; 4505 this.capture = capture; 4506 } 4507 boolean match(Matcher matcher, int i, CharSequence seq) { 4508 int[] groups = matcher.groups; 4509 int[] locals = matcher.locals; 4510 int save0 = locals[localIndex]; 4511 int save1 = 0; 4512 int save2 = 0; 4513 4514 if (capture) { 4515 save1 = groups[groupIndex]; 4516 save2 = groups[groupIndex+1]; 4517 } 4518 4519 // Notify GroupTail there is no need to setup group info 4520 // because it will be set here 4521 locals[localIndex] = -1; 4522 4523 boolean ret = true; 4524 for (int j = 0; j < cmin; j++) { 4525 if (atom.match(matcher, i, seq)) { 4526 if (capture) { 4527 groups[groupIndex] = i; 4528 groups[groupIndex+1] = matcher.last; 4529 } 4530 i = matcher.last; 4531 } else { 4532 ret = false; 4533 break; 4534 } 4535 } 4536 if (ret) { 4537 if (type == Qtype.GREEDY) { 4538 ret = match0(matcher, i, cmin, seq); 4539 } else if (type == Qtype.LAZY) { 4540 ret = match1(matcher, i, cmin, seq); 4541 } else { 4542 ret = match2(matcher, i, cmin, seq); 4543 } 4544 } 4545 if (!ret) { 4546 locals[localIndex] = save0; 4547 if (capture) { 4548 groups[groupIndex] = save1; 4549 groups[groupIndex+1] = save2; 4550 } 4551 } 4552 return ret; 4553 } 4554 // Aggressive group match 4555 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4556 // don't back off passing the starting "j" 4557 int min = j; 4558 int[] groups = matcher.groups; 4559 int save0 = 0; 4560 int save1 = 0; 4561 if (capture) { 4562 save0 = groups[groupIndex]; 4563 save1 = groups[groupIndex+1]; 4564 } 4565 for (;;) { 4566 if (j >= cmax) 4567 break; 4568 if (!atom.match(matcher, i, seq)) 4569 break; 4570 int k = matcher.last - i; 4571 if (k <= 0) { 4572 if (capture) { 4573 groups[groupIndex] = i; 4574 groups[groupIndex+1] = i + k; 4575 } 4576 i = i + k; 4577 break; 4578 } 4579 for (;;) { 4580 if (capture) { 4581 groups[groupIndex] = i; 4582 groups[groupIndex+1] = i + k; 4583 } 4584 i = i + k; 4585 if (++j >= cmax) 4586 break; 4587 if (!atom.match(matcher, i, seq)) 4588 break; 4589 if (i + k != matcher.last) { 4590 if (match0(matcher, i, j, seq)) 4591 return true; 4592 break; 4593 } 4594 } 4595 while (j > min) { 4596 if (next.match(matcher, i, seq)) { 4597 if (capture) { 4598 groups[groupIndex+1] = i; 4599 groups[groupIndex] = i - k; 4600 } 4601 return true; 4602 } 4603 // backing off 4604 i = i - k; 4605 if (capture) { 4606 groups[groupIndex+1] = i; 4607 groups[groupIndex] = i - k; 4608 } 4609 j--; 4610 4611 } 4612 break; 4613 } 4614 if (capture) { 4615 groups[groupIndex] = save0; 4616 groups[groupIndex+1] = save1; 4617 } 4618 return next.match(matcher, i, seq); 4619 } 4620 // Reluctant matching 4621 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4622 for (;;) { 4623 if (next.match(matcher, i, seq)) 4624 return true; 4625 if (j >= cmax) 4626 return false; 4627 if (!atom.match(matcher, i, seq)) 4628 return false; 4629 if (i == matcher.last) 4630 return false; 4631 if (capture) { 4632 matcher.groups[groupIndex] = i; 4633 matcher.groups[groupIndex+1] = matcher.last; 4634 } 4635 i = matcher.last; 4636 j++; 4637 } 4638 } 4639 // Possessive matching 4640 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4641 for (; j < cmax; j++) { 4642 if (!atom.match(matcher, i, seq)) { 4643 break; 4644 } 4645 if (capture) { 4646 matcher.groups[groupIndex] = i; 4647 matcher.groups[groupIndex+1] = matcher.last; 4648 } 4649 if (i == matcher.last) { 4650 break; 4651 } 4652 i = matcher.last; 4653 } 4654 return next.match(matcher, i, seq); 4655 } 4656 boolean study(TreeInfo info) { 4657 // Save original info 4658 int minL = info.minLength; 4659 int maxL = info.maxLength; 4660 boolean maxV = info.maxValid; 4661 boolean detm = info.deterministic; 4662 info.reset(); 4663 4664 atom.study(info); 4665 4666 int temp = info.minLength * cmin + minL; 4667 if (temp < minL) { 4668 temp = 0xFFFFFFF; // Arbitrary large number 4669 } 4670 info.minLength = temp; 4671 4672 if (maxV & info.maxValid) { 4673 temp = info.maxLength * cmax + maxL; 4674 info.maxLength = temp; 4675 if (temp < maxL) { 4676 info.maxValid = false; 4677 } 4678 } else { 4679 info.maxValid = false; 4680 } 4681 4682 if (info.deterministic && cmin == cmax) { 4683 info.deterministic = detm; 4684 } else { 4685 info.deterministic = false; 4686 } 4687 return next.study(info); 4688 } 4689 } 4690 4691 /** 4692 * A Guard node at the end of each atom node in a Branch. It 4693 * serves the purpose of chaining the "match" operation to 4694 * "next" but not the "study", so we can collect the TreeInfo 4695 * of each atom node without including the TreeInfo of the 4696 * "next". 4697 */ 4698 static final class BranchConn extends Node { 4699 BranchConn() {} 4700 boolean match(Matcher matcher, int i, CharSequence seq) { 4701 return next.match(matcher, i, seq); 4702 } 4703 boolean study(TreeInfo info) { 4704 return info.deterministic; 4705 } 4706 } 4707 4708 /** 4709 * Handles the branching of alternations. Note this is also used for 4710 * the ? quantifier to branch between the case where it matches once 4711 * and where it does not occur. 4712 */ 4713 static final class Branch extends Node { 4714 Node[] atoms = new Node[2]; 4715 int size = 2; 4716 Node conn; 4717 Branch(Node first, Node second, Node branchConn) { 4718 conn = branchConn; 4719 atoms[0] = first; 4720 atoms[1] = second; 4721 } 4722 4723 void add(Node node) { 4724 if (size >= atoms.length) { 4725 Node[] tmp = new Node[atoms.length*2]; 4726 System.arraycopy(atoms, 0, tmp, 0, atoms.length); 4727 atoms = tmp; 4728 } 4729 atoms[size++] = node; 4730 } 4731 4732 boolean match(Matcher matcher, int i, CharSequence seq) { 4733 for (int n = 0; n < size; n++) { 4734 if (atoms[n] == null) { 4735 if (conn.next.match(matcher, i, seq)) 4736 return true; 4737 } else if (atoms[n].match(matcher, i, seq)) { 4738 return true; 4739 } 4740 } 4741 return false; 4742 } 4743 4744 boolean study(TreeInfo info) { 4745 int minL = info.minLength; 4746 int maxL = info.maxLength; 4747 boolean maxV = info.maxValid; 4748 4749 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num 4750 int maxL2 = -1; 4751 for (int n = 0; n < size; n++) { 4752 info.reset(); 4753 if (atoms[n] != null) 4754 atoms[n].study(info); 4755 minL2 = Math.min(minL2, info.minLength); 4756 maxL2 = Math.max(maxL2, info.maxLength); 4757 maxV = (maxV & info.maxValid); 4758 } 4759 4760 minL += minL2; 4761 maxL += maxL2; 4762 4763 info.reset(); 4764 conn.next.study(info); 4765 4766 info.minLength += minL; 4767 info.maxLength += maxL; 4768 info.maxValid &= maxV; 4769 info.deterministic = false; 4770 return false; 4771 } 4772 } 4773 4774 /** 4775 * The GroupHead saves the location where the group begins in the locals 4776 * and restores them when the match is done. 4777 * 4778 * The matchRef is used when a reference to this group is accessed later 4779 * in the expression. The locals will have a negative value in them to 4780 * indicate that we do not want to unset the group if the reference 4781 * doesn't match. 4782 */ 4783 static final class GroupHead extends Node { 4784 int localIndex; 4785 GroupTail tail; // for debug/print only, match does not need to know 4786 GroupHead(int localCount) { 4787 localIndex = localCount; 4788 } 4789 boolean match(Matcher matcher, int i, CharSequence seq) { 4790 int save = matcher.locals[localIndex]; 4791 matcher.locals[localIndex] = i; 4792 boolean ret = next.match(matcher, i, seq); 4793 matcher.locals[localIndex] = save; 4794 return ret; 4795 } 4796 } 4797 4798 /** 4799 * The GroupTail handles the setting of group beginning and ending 4800 * locations when groups are successfully matched. It must also be able to 4801 * unset groups that have to be backed off of. 4802 * 4803 * The GroupTail node is also used when a previous group is referenced, 4804 * and in that case no group information needs to be set. 4805 */ 4806 static final class GroupTail extends Node { 4807 int localIndex; 4808 int groupIndex; 4809 GroupTail(int localCount, int groupCount) { 4810 localIndex = localCount; 4811 groupIndex = groupCount + groupCount; 4812 } 4813 boolean match(Matcher matcher, int i, CharSequence seq) { 4814 int tmp = matcher.locals[localIndex]; 4815 if (tmp >= 0) { // This is the normal group case. 4816 // Save the group so we can unset it if it 4817 // backs off of a match. 4818 int groupStart = matcher.groups[groupIndex]; 4819 int groupEnd = matcher.groups[groupIndex+1]; 4820 4821 matcher.groups[groupIndex] = tmp; 4822 matcher.groups[groupIndex+1] = i; 4823 if (next.match(matcher, i, seq)) { 4824 return true; 4825 } 4826 matcher.groups[groupIndex] = groupStart; 4827 matcher.groups[groupIndex+1] = groupEnd; 4828 return false; 4829 } else { 4830 // This is a group reference case. We don't need to save any 4831 // group info because it isn't really a group. 4832 matcher.last = i; 4833 return true; 4834 } 4835 } 4836 } 4837 4838 /** 4839 * This sets up a loop to handle a recursive quantifier structure. 4840 */ 4841 static final class Prolog extends Node { 4842 Loop loop; 4843 Prolog(Loop loop) { 4844 this.loop = loop; 4845 } 4846 boolean match(Matcher matcher, int i, CharSequence seq) { 4847 return loop.matchInit(matcher, i, seq); 4848 } 4849 boolean study(TreeInfo info) { 4850 return loop.study(info); 4851 } 4852 } 4853 4854 /** 4855 * Handles the repetition count for a greedy Curly. The matchInit 4856 * is called from the Prolog to save the index of where the group 4857 * beginning is stored. A zero length group check occurs in the 4858 * normal match but is skipped in the matchInit. 4859 */ 4860 static class Loop extends Node { 4861 Node body; 4862 int countIndex; // local count index in matcher locals 4863 int beginIndex; // group beginning index 4864 int cmin, cmax; 4865 int posIndex; 4866 Loop(int countIndex, int beginIndex) { 4867 this.countIndex = countIndex; 4868 this.beginIndex = beginIndex; 4869 this.posIndex = -1; 4870 } 4871 boolean match(Matcher matcher, int i, CharSequence seq) { 4872 // Avoid infinite loop in zero-length case. 4873 if (i > matcher.locals[beginIndex]) { 4874 int count = matcher.locals[countIndex]; 4875 4876 // This block is for before we reach the minimum 4877 // iterations required for the loop to match 4878 if (count < cmin) { 4879 matcher.locals[countIndex] = count + 1; 4880 boolean b = body.match(matcher, i, seq); 4881 // If match failed we must backtrack, so 4882 // the loop count should NOT be incremented 4883 if (!b) 4884 matcher.locals[countIndex] = count; 4885 // Return success or failure since we are under 4886 // minimum 4887 return b; 4888 } 4889 // This block is for after we have the minimum 4890 // iterations required for the loop to match 4891 if (count < cmax) { 4892 // Let's check if we have already tried and failed 4893 // at this starting position "i" in the past. 4894 // If yes, then just return false wihtout trying 4895 // again, to stop the exponential backtracking. 4896 if (posIndex != -1 && 4897 matcher.localsPos[posIndex].contains(i)) { 4898 return next.match(matcher, i, seq); 4899 } 4900 matcher.locals[countIndex] = count + 1; 4901 boolean b = body.match(matcher, i, seq); 4902 // If match failed we must backtrack, so 4903 // the loop count should NOT be incremented 4904 if (b) 4905 return true; 4906 matcher.locals[countIndex] = count; 4907 // save the failed position 4908 if (posIndex != -1) { 4909 matcher.localsPos[posIndex].add(i); 4910 } 4911 } 4912 } 4913 return next.match(matcher, i, seq); 4914 } 4915 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4916 int save = matcher.locals[countIndex]; 4917 boolean ret; 4918 if (posIndex != -1 && matcher.localsPos[posIndex] == null) { 4919 matcher.localsPos[posIndex] = new IntHashSet(); 4920 } 4921 if (0 < cmin) { 4922 matcher.locals[countIndex] = 1; 4923 ret = body.match(matcher, i, seq); 4924 } else if (0 < cmax) { 4925 matcher.locals[countIndex] = 1; 4926 ret = body.match(matcher, i, seq); 4927 if (ret == false) 4928 ret = next.match(matcher, i, seq); 4929 } else { 4930 ret = next.match(matcher, i, seq); 4931 } 4932 matcher.locals[countIndex] = save; 4933 return ret; 4934 } 4935 boolean study(TreeInfo info) { 4936 info.maxValid = false; 4937 info.deterministic = false; 4938 return false; 4939 } 4940 } 4941 4942 /** 4943 * Handles the repetition count for a reluctant Curly. The matchInit 4944 * is called from the Prolog to save the index of where the group 4945 * beginning is stored. A zero length group check occurs in the 4946 * normal match but is skipped in the matchInit. 4947 */ 4948 static final class LazyLoop extends Loop { 4949 LazyLoop(int countIndex, int beginIndex) { 4950 super(countIndex, beginIndex); 4951 } 4952 boolean match(Matcher matcher, int i, CharSequence seq) { 4953 // Check for zero length group 4954 if (i > matcher.locals[beginIndex]) { 4955 int count = matcher.locals[countIndex]; 4956 if (count < cmin) { 4957 matcher.locals[countIndex] = count + 1; 4958 boolean result = body.match(matcher, i, seq); 4959 // If match failed we must backtrack, so 4960 // the loop count should NOT be incremented 4961 if (!result) 4962 matcher.locals[countIndex] = count; 4963 return result; 4964 } 4965 if (next.match(matcher, i, seq)) 4966 return true; 4967 if (count < cmax) { 4968 matcher.locals[countIndex] = count + 1; 4969 boolean result = body.match(matcher, i, seq); 4970 // If match failed we must backtrack, so 4971 // the loop count should NOT be incremented 4972 if (!result) 4973 matcher.locals[countIndex] = count; 4974 return result; 4975 } 4976 return false; 4977 } 4978 return next.match(matcher, i, seq); 4979 } 4980 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4981 int save = matcher.locals[countIndex]; 4982 boolean ret = false; 4983 if (0 < cmin) { 4984 matcher.locals[countIndex] = 1; 4985 ret = body.match(matcher, i, seq); 4986 } else if (next.match(matcher, i, seq)) { 4987 ret = true; 4988 } else if (0 < cmax) { 4989 matcher.locals[countIndex] = 1; 4990 ret = body.match(matcher, i, seq); 4991 } 4992 matcher.locals[countIndex] = save; 4993 return ret; 4994 } 4995 boolean study(TreeInfo info) { 4996 info.maxValid = false; 4997 info.deterministic = false; 4998 return false; 4999 } 5000 } 5001 5002 /** 5003 * Refers to a group in the regular expression. Attempts to match 5004 * whatever the group referred to last matched. 5005 */ 5006 static class BackRef extends Node { 5007 int groupIndex; 5008 BackRef(int groupCount) { 5009 super(); 5010 groupIndex = groupCount + groupCount; 5011 } 5012 boolean match(Matcher matcher, int i, CharSequence seq) { 5013 int j = matcher.groups[groupIndex]; 5014 int k = matcher.groups[groupIndex+1]; 5015 5016 int groupSize = k - j; 5017 // If the referenced group didn't match, neither can this 5018 if (j < 0) 5019 return false; 5020 5021 // If there isn't enough input left no match 5022 if (i + groupSize > matcher.to) { 5023 matcher.hitEnd = true; 5024 return false; 5025 } 5026 // Check each new char to make sure it matches what the group 5027 // referenced matched last time around 5028 for (int index=0; index<groupSize; index++) 5029 if (seq.charAt(i+index) != seq.charAt(j+index)) 5030 return false; 5031 5032 return next.match(matcher, i+groupSize, seq); 5033 } 5034 boolean study(TreeInfo info) { 5035 info.maxValid = false; 5036 return next.study(info); 5037 } 5038 } 5039 5040 static class CIBackRef extends Node { 5041 int groupIndex; 5042 boolean doUnicodeCase; 5043 CIBackRef(int groupCount, boolean doUnicodeCase) { 5044 super(); 5045 groupIndex = groupCount + groupCount; 5046 this.doUnicodeCase = doUnicodeCase; 5047 } 5048 boolean match(Matcher matcher, int i, CharSequence seq) { 5049 int j = matcher.groups[groupIndex]; 5050 int k = matcher.groups[groupIndex+1]; 5051 5052 int groupSize = k - j; 5053 5054 // If the referenced group didn't match, neither can this 5055 if (j < 0) 5056 return false; 5057 5058 // If there isn't enough input left no match 5059 if (i + groupSize > matcher.to) { 5060 matcher.hitEnd = true; 5061 return false; 5062 } 5063 5064 // Check each new char to make sure it matches what the group 5065 // referenced matched last time around 5066 int x = i; 5067 for (int index=0; index<groupSize; index++) { 5068 int c1 = Character.codePointAt(seq, x); 5069 int c2 = Character.codePointAt(seq, j); 5070 if (c1 != c2) { 5071 if (doUnicodeCase) { 5072 int cc1 = Character.toUpperCase(c1); 5073 int cc2 = Character.toUpperCase(c2); 5074 if (cc1 != cc2 && 5075 Character.toLowerCase(cc1) != 5076 Character.toLowerCase(cc2)) 5077 return false; 5078 } else { 5079 if (ASCII.toLower(c1) != ASCII.toLower(c2)) 5080 return false; 5081 } 5082 } 5083 x += Character.charCount(c1); 5084 j += Character.charCount(c2); 5085 } 5086 5087 return next.match(matcher, i+groupSize, seq); 5088 } 5089 boolean study(TreeInfo info) { 5090 info.maxValid = false; 5091 return next.study(info); 5092 } 5093 } 5094 5095 /** 5096 * Searches until the next instance of its atom. This is useful for 5097 * finding the atom efficiently without passing an instance of it 5098 * (greedy problem) and without a lot of wasted search time (reluctant 5099 * problem). 5100 */ 5101 static final class First extends Node { 5102 Node atom; 5103 First(Node node) { 5104 this.atom = BnM.optimize(node); 5105 } 5106 boolean match(Matcher matcher, int i, CharSequence seq) { 5107 if (atom instanceof BnM) { 5108 return atom.match(matcher, i, seq) 5109 && next.match(matcher, matcher.last, seq); 5110 } 5111 for (;;) { 5112 if (i > matcher.to) { 5113 matcher.hitEnd = true; 5114 return false; 5115 } 5116 if (atom.match(matcher, i, seq)) { 5117 return next.match(matcher, matcher.last, seq); 5118 } 5119 i += countChars(seq, i, 1); 5120 matcher.first++; 5121 } 5122 } 5123 boolean study(TreeInfo info) { 5124 atom.study(info); 5125 info.maxValid = false; 5126 info.deterministic = false; 5127 return next.study(info); 5128 } 5129 } 5130 5131 /** 5132 * Zero width positive lookahead. 5133 */ 5134 static final class Pos extends Node { 5135 Node cond; 5136 Pos(Node cond) { 5137 this.cond = cond; 5138 } 5139 boolean match(Matcher matcher, int i, CharSequence seq) { 5140 int savedTo = matcher.to; 5141 boolean conditionMatched; 5142 5143 // Relax transparent region boundaries for lookahead 5144 if (matcher.transparentBounds) 5145 matcher.to = matcher.getTextLength(); 5146 try { 5147 conditionMatched = cond.match(matcher, i, seq); 5148 } finally { 5149 // Reinstate region boundaries 5150 matcher.to = savedTo; 5151 } 5152 return conditionMatched && next.match(matcher, i, seq); 5153 } 5154 } 5155 5156 /** 5157 * Zero width negative lookahead. 5158 */ 5159 static final class Neg extends Node { 5160 Node cond; 5161 Neg(Node cond) { 5162 this.cond = cond; 5163 } 5164 boolean match(Matcher matcher, int i, CharSequence seq) { 5165 int savedTo = matcher.to; 5166 boolean conditionMatched; 5167 5168 // Relax transparent region boundaries for lookahead 5169 if (matcher.transparentBounds) 5170 matcher.to = matcher.getTextLength(); 5171 try { 5172 if (i < matcher.to) { 5173 conditionMatched = !cond.match(matcher, i, seq); 5174 } else { 5175 // If a negative lookahead succeeds then more input 5176 // could cause it to fail! 5177 matcher.requireEnd = true; 5178 conditionMatched = !cond.match(matcher, i, seq); 5179 } 5180 } finally { 5181 // Reinstate region boundaries 5182 matcher.to = savedTo; 5183 } 5184 return conditionMatched && next.match(matcher, i, seq); 5185 } 5186 } 5187 5188 /** 5189 * For use with lookbehinds; matches the position where the lookbehind 5190 * was encountered. 5191 */ 5192 static class LookBehindEndNode extends Node { 5193 private LookBehindEndNode() {} // Singleton 5194 5195 static LookBehindEndNode INSTANCE = new LookBehindEndNode(); 5196 5197 boolean match(Matcher matcher, int i, CharSequence seq) { 5198 return i == matcher.lookbehindTo; 5199 } 5200 } 5201 5202 /** 5203 * Zero width positive lookbehind. 5204 */ 5205 static class Behind extends Node { 5206 Node cond; 5207 int rmax, rmin; 5208 Behind(Node cond, int rmax, int rmin) { 5209 this.cond = cond; 5210 this.rmax = rmax; 5211 this.rmin = rmin; 5212 } 5213 5214 boolean match(Matcher matcher, int i, CharSequence seq) { 5215 int savedFrom = matcher.from; 5216 boolean conditionMatched = false; 5217 int startIndex = (!matcher.transparentBounds) ? 5218 matcher.from : 0; 5219 int from = Math.max(i - rmax, startIndex); 5220 // Set end boundary 5221 int savedLBT = matcher.lookbehindTo; 5222 matcher.lookbehindTo = i; 5223 // Relax transparent region boundaries for lookbehind 5224 if (matcher.transparentBounds) 5225 matcher.from = 0; 5226 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5227 conditionMatched = cond.match(matcher, j, seq); 5228 } 5229 matcher.from = savedFrom; 5230 matcher.lookbehindTo = savedLBT; 5231 return conditionMatched && next.match(matcher, i, seq); 5232 } 5233 } 5234 5235 /** 5236 * Zero width positive lookbehind, including supplementary 5237 * characters or unpaired surrogates. 5238 */ 5239 static final class BehindS extends Behind { 5240 BehindS(Node cond, int rmax, int rmin) { 5241 super(cond, rmax, rmin); 5242 } 5243 boolean match(Matcher matcher, int i, CharSequence seq) { 5244 int rmaxChars = countChars(seq, i, -rmax); 5245 int rminChars = countChars(seq, i, -rmin); 5246 int savedFrom = matcher.from; 5247 int startIndex = (!matcher.transparentBounds) ? 5248 matcher.from : 0; 5249 boolean conditionMatched = false; 5250 int from = Math.max(i - rmaxChars, startIndex); 5251 // Set end boundary 5252 int savedLBT = matcher.lookbehindTo; 5253 matcher.lookbehindTo = i; 5254 // Relax transparent region boundaries for lookbehind 5255 if (matcher.transparentBounds) 5256 matcher.from = 0; 5257 5258 for (int j = i - rminChars; 5259 !conditionMatched && j >= from; 5260 j -= j>from ? countChars(seq, j, -1) : 1) { 5261 conditionMatched = cond.match(matcher, j, seq); 5262 } 5263 matcher.from = savedFrom; 5264 matcher.lookbehindTo = savedLBT; 5265 return conditionMatched && next.match(matcher, i, seq); 5266 } 5267 } 5268 5269 /** 5270 * Zero width negative lookbehind. 5271 */ 5272 static class NotBehind extends Node { 5273 Node cond; 5274 int rmax, rmin; 5275 NotBehind(Node cond, int rmax, int rmin) { 5276 this.cond = cond; 5277 this.rmax = rmax; 5278 this.rmin = rmin; 5279 } 5280 5281 boolean match(Matcher matcher, int i, CharSequence seq) { 5282 int savedLBT = matcher.lookbehindTo; 5283 int savedFrom = matcher.from; 5284 boolean conditionMatched = false; 5285 int startIndex = (!matcher.transparentBounds) ? 5286 matcher.from : 0; 5287 int from = Math.max(i - rmax, startIndex); 5288 matcher.lookbehindTo = i; 5289 // Relax transparent region boundaries for lookbehind 5290 if (matcher.transparentBounds) 5291 matcher.from = 0; 5292 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5293 conditionMatched = cond.match(matcher, j, seq); 5294 } 5295 // Reinstate region boundaries 5296 matcher.from = savedFrom; 5297 matcher.lookbehindTo = savedLBT; 5298 return !conditionMatched && next.match(matcher, i, seq); 5299 } 5300 } 5301 5302 /** 5303 * Zero width negative lookbehind, including supplementary 5304 * characters or unpaired surrogates. 5305 */ 5306 static final class NotBehindS extends NotBehind { 5307 NotBehindS(Node cond, int rmax, int rmin) { 5308 super(cond, rmax, rmin); 5309 } 5310 boolean match(Matcher matcher, int i, CharSequence seq) { 5311 int rmaxChars = countChars(seq, i, -rmax); 5312 int rminChars = countChars(seq, i, -rmin); 5313 int savedFrom = matcher.from; 5314 int savedLBT = matcher.lookbehindTo; 5315 boolean conditionMatched = false; 5316 int startIndex = (!matcher.transparentBounds) ? 5317 matcher.from : 0; 5318 int from = Math.max(i - rmaxChars, startIndex); 5319 matcher.lookbehindTo = i; 5320 // Relax transparent region boundaries for lookbehind 5321 if (matcher.transparentBounds) 5322 matcher.from = 0; 5323 for (int j = i - rminChars; 5324 !conditionMatched && j >= from; 5325 j -= j>from ? countChars(seq, j, -1) : 1) { 5326 conditionMatched = cond.match(matcher, j, seq); 5327 } 5328 //Reinstate region boundaries 5329 matcher.from = savedFrom; 5330 matcher.lookbehindTo = savedLBT; 5331 return !conditionMatched && next.match(matcher, i, seq); 5332 } 5333 } 5334 5335 /** 5336 * Handles word boundaries. Includes a field to allow this one class to 5337 * deal with the different types of word boundaries we can match. The word 5338 * characters include underscores, letters, and digits. Non spacing marks 5339 * can are also part of a word if they have a base character, otherwise 5340 * they are ignored for purposes of finding word boundaries. 5341 */ 5342 static final class Bound extends Node { 5343 static int LEFT = 0x1; 5344 static int RIGHT= 0x2; 5345 static int BOTH = 0x3; 5346 static int NONE = 0x4; 5347 int type; 5348 boolean useUWORD; 5349 Bound(int n, boolean useUWORD) { 5350 type = n; 5351 this.useUWORD = useUWORD; 5352 } 5353 5354 boolean isWord(int ch) { 5355 return useUWORD ? CharPredicates.WORD().is(ch) 5356 : (ch == '_' || Character.isLetterOrDigit(ch)); 5357 } 5358 5359 int check(Matcher matcher, int i, CharSequence seq) { 5360 int ch; 5361 boolean left = false; 5362 int startIndex = matcher.from; 5363 int endIndex = matcher.to; 5364 if (matcher.transparentBounds) { 5365 startIndex = 0; 5366 endIndex = matcher.getTextLength(); 5367 } 5368 if (i > startIndex) { 5369 ch = Character.codePointBefore(seq, i); 5370 left = (isWord(ch) || 5371 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5372 && hasBaseCharacter(matcher, i-1, seq))); 5373 } 5374 boolean right = false; 5375 if (i < endIndex) { 5376 ch = Character.codePointAt(seq, i); 5377 right = (isWord(ch) || 5378 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5379 && hasBaseCharacter(matcher, i, seq))); 5380 } else { 5381 // Tried to access char past the end 5382 matcher.hitEnd = true; 5383 // The addition of another char could wreck a boundary 5384 matcher.requireEnd = true; 5385 } 5386 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE); 5387 } 5388 boolean match(Matcher matcher, int i, CharSequence seq) { 5389 return (check(matcher, i, seq) & type) > 0 5390 && next.match(matcher, i, seq); 5391 } 5392 } 5393 5394 /** 5395 * Non spacing marks only count as word characters in bounds calculations 5396 * if they have a base character. 5397 */ 5398 private static boolean hasBaseCharacter(Matcher matcher, int i, 5399 CharSequence seq) 5400 { 5401 int start = (!matcher.transparentBounds) ? 5402 matcher.from : 0; 5403 for (int x=i; x >= start; x--) { 5404 int ch = Character.codePointAt(seq, x); 5405 if (Character.isLetterOrDigit(ch)) 5406 return true; 5407 if (Character.getType(ch) == Character.NON_SPACING_MARK) 5408 continue; 5409 return false; 5410 } 5411 return false; 5412 } 5413 5414 /** 5415 * Attempts to match a slice in the input using the Boyer-Moore string 5416 * matching algorithm. The algorithm is based on the idea that the 5417 * pattern can be shifted farther ahead in the search text if it is 5418 * matched right to left. 5419 * <p> 5420 * The pattern is compared to the input one character at a time, from 5421 * the rightmost character in the pattern to the left. If the characters 5422 * all match the pattern has been found. If a character does not match, 5423 * the pattern is shifted right a distance that is the maximum of two 5424 * functions, the bad character shift and the good suffix shift. This 5425 * shift moves the attempted match position through the input more 5426 * quickly than a naive one position at a time check. 5427 * <p> 5428 * The bad character shift is based on the character from the text that 5429 * did not match. If the character does not appear in the pattern, the 5430 * pattern can be shifted completely beyond the bad character. If the 5431 * character does occur in the pattern, the pattern can be shifted to 5432 * line the pattern up with the next occurrence of that character. 5433 * <p> 5434 * The good suffix shift is based on the idea that some subset on the right 5435 * side of the pattern has matched. When a bad character is found, the 5436 * pattern can be shifted right by the pattern length if the subset does 5437 * not occur again in pattern, or by the amount of distance to the 5438 * next occurrence of the subset in the pattern. 5439 * 5440 * Boyer-Moore search methods adapted from code by Amy Yu. 5441 */ 5442 static class BnM extends Node { 5443 int[] buffer; 5444 int[] lastOcc; 5445 int[] optoSft; 5446 5447 /** 5448 * Pre calculates arrays needed to generate the bad character 5449 * shift and the good suffix shift. Only the last seven bits 5450 * are used to see if chars match; This keeps the tables small 5451 * and covers the heavily used ASCII range, but occasionally 5452 * results in an aliased match for the bad character shift. 5453 */ 5454 static Node optimize(Node node) { 5455 if (!(node instanceof Slice)) { 5456 return node; 5457 } 5458 5459 int[] src = ((Slice) node).buffer; 5460 int patternLength = src.length; 5461 // The BM algorithm requires a bit of overhead; 5462 // If the pattern is short don't use it, since 5463 // a shift larger than the pattern length cannot 5464 // be used anyway. 5465 if (patternLength < 4) { 5466 return node; 5467 } 5468 int i, j; 5469 int[] lastOcc = new int[128]; 5470 int[] optoSft = new int[patternLength]; 5471 // Precalculate part of the bad character shift 5472 // It is a table for where in the pattern each 5473 // lower 7-bit value occurs 5474 for (i = 0; i < patternLength; i++) { 5475 lastOcc[src[i]&0x7F] = i + 1; 5476 } 5477 // Precalculate the good suffix shift 5478 // i is the shift amount being considered 5479 NEXT: for (i = patternLength; i > 0; i--) { 5480 // j is the beginning index of suffix being considered 5481 for (j = patternLength - 1; j >= i; j--) { 5482 // Testing for good suffix 5483 if (src[j] == src[j-i]) { 5484 // src[j..len] is a good suffix 5485 optoSft[j-1] = i; 5486 } else { 5487 // No match. The array has already been 5488 // filled up with correct values before. 5489 continue NEXT; 5490 } 5491 } 5492 // This fills up the remaining of optoSft 5493 // any suffix can not have larger shift amount 5494 // then its sub-suffix. Why??? 5495 while (j > 0) { 5496 optoSft[--j] = i; 5497 } 5498 } 5499 // Set the guard value because of unicode compression 5500 optoSft[patternLength-1] = 1; 5501 if (node instanceof SliceS) 5502 return new BnMS(src, lastOcc, optoSft, node.next); 5503 return new BnM(src, lastOcc, optoSft, node.next); 5504 } 5505 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5506 this.buffer = src; 5507 this.lastOcc = lastOcc; 5508 this.optoSft = optoSft; 5509 this.next = next; 5510 } 5511 boolean match(Matcher matcher, int i, CharSequence seq) { 5512 int[] src = buffer; 5513 int patternLength = src.length; 5514 int last = matcher.to - patternLength; 5515 5516 // Loop over all possible match positions in text 5517 NEXT: while (i <= last) { 5518 // Loop over pattern from right to left 5519 for (int j = patternLength - 1; j >= 0; j--) { 5520 int ch = seq.charAt(i+j); 5521 if (ch != src[j]) { 5522 // Shift search to the right by the maximum of the 5523 // bad character shift and the good suffix shift 5524 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]); 5525 continue NEXT; 5526 } 5527 } 5528 // Entire pattern matched starting at i 5529 matcher.first = i; 5530 boolean ret = next.match(matcher, i + patternLength, seq); 5531 if (ret) { 5532 matcher.first = i; 5533 matcher.groups[0] = matcher.first; 5534 matcher.groups[1] = matcher.last; 5535 return true; 5536 } 5537 i++; 5538 } 5539 // BnM is only used as the leading node in the unanchored case, 5540 // and it replaced its Start() which always searches to the end 5541 // if it doesn't find what it's looking for, so hitEnd is true. 5542 matcher.hitEnd = true; 5543 return false; 5544 } 5545 boolean study(TreeInfo info) { 5546 info.minLength += buffer.length; 5547 info.maxValid = false; 5548 return next.study(info); 5549 } 5550 } 5551 5552 /** 5553 * Supplementary support version of BnM(). Unpaired surrogates are 5554 * also handled by this class. 5555 */ 5556 static final class BnMS extends BnM { 5557 int lengthInChars; 5558 5559 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5560 super(src, lastOcc, optoSft, next); 5561 for (int cp : buffer) { 5562 lengthInChars += Character.charCount(cp); 5563 } 5564 } 5565 boolean match(Matcher matcher, int i, CharSequence seq) { 5566 int[] src = buffer; 5567 int patternLength = src.length; 5568 int last = matcher.to - lengthInChars; 5569 5570 // Loop over all possible match positions in text 5571 NEXT: while (i <= last) { 5572 // Loop over pattern from right to left 5573 int ch; 5574 for (int j = countChars(seq, i, patternLength), x = patternLength - 1; 5575 j > 0; j -= Character.charCount(ch), x--) { 5576 ch = Character.codePointBefore(seq, i+j); 5577 if (ch != src[x]) { 5578 // Shift search to the right by the maximum of the 5579 // bad character shift and the good suffix shift 5580 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]); 5581 i += countChars(seq, i, n); 5582 continue NEXT; 5583 } 5584 } 5585 // Entire pattern matched starting at i 5586 matcher.first = i; 5587 boolean ret = next.match(matcher, i + lengthInChars, seq); 5588 if (ret) { 5589 matcher.first = i; 5590 matcher.groups[0] = matcher.first; 5591 matcher.groups[1] = matcher.last; 5592 return true; 5593 } 5594 i += countChars(seq, i, 1); 5595 } 5596 matcher.hitEnd = true; 5597 return false; 5598 } 5599 } 5600 5601 @FunctionalInterface 5602 static interface CharPredicate { 5603 boolean is(int ch); 5604 5605 default CharPredicate and(CharPredicate p) { 5606 return ch -> is(ch) && p.is(ch); 5607 } 5608 default CharPredicate union(CharPredicate p) { 5609 return ch -> is(ch) || p.is(ch); 5610 } 5611 default CharPredicate union(CharPredicate p1, 5612 CharPredicate p2 ) { 5613 return ch -> is(ch) || p1.is(ch) || p2.is(ch); 5614 } 5615 default CharPredicate negate() { 5616 return ch -> !is(ch); 5617 } 5618 } 5619 5620 static interface BmpCharPredicate extends CharPredicate { 5621 5622 default CharPredicate and(CharPredicate p) { 5623 if (p instanceof BmpCharPredicate) 5624 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch)); 5625 return ch -> is(ch) && p.is(ch); 5626 } 5627 default CharPredicate union(CharPredicate p) { 5628 if (p instanceof BmpCharPredicate) 5629 return (BmpCharPredicate)(ch -> is(ch) || p.is(ch)); 5630 return ch -> is(ch) || p.is(ch); 5631 } 5632 static CharPredicate union(CharPredicate... predicates) { 5633 CharPredicate cp = ch -> { 5634 for (CharPredicate p : predicates) { 5635 if (!p.is(ch)) 5636 return false; 5637 } 5638 return true; 5639 }; 5640 for (CharPredicate p : predicates) { 5641 if (! (p instanceof BmpCharPredicate)) 5642 return cp; 5643 } 5644 return (BmpCharPredicate)cp; 5645 } 5646 } 5647 5648 /** 5649 * matches a Perl vertical whitespace 5650 */ 5651 static BmpCharPredicate VertWS() { 5652 return cp -> (cp >= 0x0A && cp <= 0x0D) || 5653 cp == 0x85 || cp == 0x2028 || cp == 0x2029; 5654 } 5655 5656 /** 5657 * matches a Perl horizontal whitespace 5658 */ 5659 static BmpCharPredicate HorizWS() { 5660 return cp -> 5661 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 || 5662 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f || 5663 cp == 0x205f || cp == 0x3000; 5664 } 5665 5666 /** 5667 * for the Unicode category ALL and the dot metacharacter when 5668 * in dotall mode. 5669 */ 5670 static CharPredicate ALL() { 5671 return ch -> true; 5672 } 5673 5674 /** 5675 * for the dot metacharacter when dotall is not enabled. 5676 */ 5677 static CharPredicate DOT() { 5678 return ch -> 5679 (ch != '\n' && ch != '\r' 5680 && (ch|1) != '\u2029' 5681 && ch != '\u0085'); 5682 } 5683 5684 /** 5685 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled. 5686 */ 5687 static CharPredicate UNIXDOT() { 5688 return ch -> ch != '\n'; 5689 } 5690 5691 /** 5692 * Indicate that matches a Supplementary Unicode character 5693 */ 5694 static CharPredicate SingleS(int c) { 5695 return ch -> ch == c; 5696 } 5697 5698 /** 5699 * A bmp/optimized predicate of single 5700 */ 5701 static BmpCharPredicate Single(int c) { 5702 return ch -> ch == c; 5703 } 5704 5705 /** 5706 * Case insensitive matches a given BMP character 5707 */ 5708 static BmpCharPredicate SingleI(int lower, int upper) { 5709 return ch -> ch == lower || ch == upper; 5710 } 5711 5712 /** 5713 * Unicode case insensitive matches a given Unicode character 5714 */ 5715 static CharPredicate SingleU(int lower) { 5716 return ch -> lower == ch || 5717 lower == Character.toLowerCase(Character.toUpperCase(ch)); 5718 } 5719 5720 private static boolean inRange(int lower, int ch, int upper) { 5721 return lower <= ch && ch <= upper; 5722 } 5723 5724 /** 5725 * Charactrs within a explicit value range 5726 */ 5727 static CharPredicate Range(int lower, int upper) { 5728 if (upper < Character.MIN_HIGH_SURROGATE || 5729 lower > Character.MAX_HIGH_SURROGATE && 5730 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT) 5731 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper)); 5732 return ch -> inRange(lower, ch, upper); 5733 } 5734 5735 /** 5736 * Charactrs within a explicit value range in a case insensitive manner. 5737 */ 5738 static CharPredicate CIRange(int lower, int upper) { 5739 return ch -> inRange(lower, ch, upper) || 5740 ASCII.isAscii(ch) && 5741 (inRange(lower, ASCII.toUpper(ch), upper) || 5742 inRange(lower, ASCII.toLower(ch), upper)); 5743 } 5744 5745 static CharPredicate CIRangeU(int lower, int upper) { 5746 return ch -> { 5747 if (inRange(lower, ch, upper)) 5748 return true; 5749 int up = Character.toUpperCase(ch); 5750 return inRange(lower, up, upper) || 5751 inRange(lower, Character.toLowerCase(up), upper); 5752 }; 5753 } 5754 5755 /** 5756 * This must be the very first initializer. 5757 */ 5758 static final Node accept = new Node(); 5759 5760 static final Node lastAccept = new LastNode(); 5761 5762 /** 5763 * Creates a predicate that tests if this pattern is found in a given input 5764 * string. 5765 * 5766 * @apiNote 5767 * This method creates a predicate that behaves as if it creates a matcher 5768 * from the input sequence and then calls {@code find}, for example a 5769 * predicate of the form: 5770 * <pre>{@code 5771 * s -> matcher(s).find(); 5772 * }</pre> 5773 * 5774 * @return The predicate which can be used for finding a match on a 5775 * subsequence of a string 5776 * @since 1.8 5777 * @see Matcher#find 5778 */ 5779 public Predicate<String> asPredicate() { 5780 return s -> matcher(s).find(); 5781 } 5782 5783 /** 5784 * Creates a predicate that tests if this pattern matches a given input string. 5785 * 5786 * @apiNote 5787 * This method creates a predicate that behaves as if it creates a matcher 5788 * from the input sequence and then calls {@code matches}, for example a 5789 * predicate of the form: 5790 * <pre>{@code 5791 * s -> matcher(s).matches(); 5792 * }</pre> 5793 * 5794 * @return The predicate which can be used for matching an input string 5795 * against this pattern. 5796 * @since 11 5797 * @see Matcher#matches 5798 */ 5799 public Predicate<String> asMatchPredicate() { 5800 return s -> matcher(s).matches(); 5801 } 5802 5803 /** 5804 * Creates a stream from the given input sequence around matches of this 5805 * pattern. 5806 * 5807 * <p> The stream returned by this method contains each substring of the 5808 * input sequence that is terminated by another subsequence that matches 5809 * this pattern or is terminated by the end of the input sequence. The 5810 * substrings in the stream are in the order in which they occur in the 5811 * input. Trailing empty strings will be discarded and not encountered in 5812 * the stream. 5813 * 5814 * <p> If this pattern does not match any subsequence of the input then 5815 * the resulting stream has just one element, namely the input sequence in 5816 * string form. 5817 * 5818 * <p> When there is a positive-width match at the beginning of the input 5819 * sequence then an empty leading substring is included at the beginning 5820 * of the stream. A zero-width match at the beginning however never produces 5821 * such empty leading substring. 5822 * 5823 * <p> If the input sequence is mutable, it must remain constant during the 5824 * execution of the terminal stream operation. Otherwise, the result of the 5825 * terminal stream operation is undefined. 5826 * 5827 * @param input 5828 * The character sequence to be split 5829 * 5830 * @return The stream of strings computed by splitting the input 5831 * around matches of this pattern 5832 * @see #split(CharSequence) 5833 * @since 1.8 5834 */ 5835 public Stream<String> splitAsStream(final CharSequence input) { 5836 class MatcherIterator implements Iterator<String> { 5837 private Matcher matcher; 5838 // The start position of the next sub-sequence of input 5839 // when current == input.length there are no more elements 5840 private int current; 5841 // null if the next element, if any, needs to obtained 5842 private String nextElement; 5843 // > 0 if there are N next empty elements 5844 private int emptyElementCount; 5845 5846 public String next() { 5847 if (!hasNext()) 5848 throw new NoSuchElementException(); 5849 5850 if (emptyElementCount == 0) { 5851 String n = nextElement; 5852 nextElement = null; 5853 return n; 5854 } else { 5855 emptyElementCount--; 5856 return ""; 5857 } 5858 } 5859 5860 public boolean hasNext() { 5861 if (matcher == null) { 5862 matcher = matcher(input); 5863 // If the input is an empty string then the result can only be a 5864 // stream of the input. Induce that by setting the empty 5865 // element count to 1 5866 emptyElementCount = input.length() == 0 ? 1 : 0; 5867 } 5868 if (nextElement != null || emptyElementCount > 0) 5869 return true; 5870 5871 if (current == input.length()) 5872 return false; 5873 5874 // Consume the next matching element 5875 // Count sequence of matching empty elements 5876 while (matcher.find()) { 5877 nextElement = input.subSequence(current, matcher.start()).toString(); 5878 current = matcher.end(); 5879 if (!nextElement.isEmpty()) { 5880 return true; 5881 } else if (current > 0) { // no empty leading substring for zero-width 5882 // match at the beginning of the input 5883 emptyElementCount++; 5884 } 5885 } 5886 5887 // Consume last matching element 5888 nextElement = input.subSequence(current, input.length()).toString(); 5889 current = input.length(); 5890 if (!nextElement.isEmpty()) { 5891 return true; 5892 } else { 5893 // Ignore a terminal sequence of matching empty elements 5894 emptyElementCount = 0; 5895 nextElement = null; 5896 return false; 5897 } 5898 } 5899 } 5900 return StreamSupport.stream(Spliterators.spliteratorUnknownSize( 5901 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false); 5902 } 5903 }