1 /* 2 * Copyright (c) 1999, 2020, 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:}<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, 3rd 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 * 786 * Pattern p1 = Pattern.compile("abc", Pattern.CASE_INSENSITIVE|Pattern.MULTILINE); 787 * Pattern p2 = Pattern.compile("(?im)abc", 0); 788 */ 789 790 /** 791 * Enables Unix lines mode. 792 * 793 * <p> In this mode, only the {@code '\n'} line terminator is recognized 794 * in the behavior of {@code .}, {@code ^}, and {@code $}. 795 * 796 * <p> Unix lines mode can also be enabled via the embedded flag 797 * expression {@code (?d)}. 798 */ 799 public static final int UNIX_LINES = 0x01; 800 801 /** 802 * Enables case-insensitive matching. 803 * 804 * <p> By default, case-insensitive matching assumes that only characters 805 * in the US-ASCII charset are being matched. Unicode-aware 806 * case-insensitive matching can be enabled by specifying the {@link 807 * #UNICODE_CASE} flag in conjunction with this flag. 808 * 809 * <p> Case-insensitive matching can also be enabled via the embedded flag 810 * expression {@code (?i)}. 811 * 812 * <p> Specifying this flag may impose a slight performance penalty. </p> 813 */ 814 public static final int CASE_INSENSITIVE = 0x02; 815 816 /** 817 * Permits whitespace and comments in pattern. 818 * 819 * <p> In this mode, whitespace is ignored, and embedded comments starting 820 * with {@code #} are ignored until the end of a line. 821 * 822 * <p> Comments mode can also be enabled via the embedded flag 823 * expression {@code (?x)}. 824 */ 825 public static final int COMMENTS = 0x04; 826 827 /** 828 * Enables multiline mode. 829 * 830 * <p> In multiline mode the expressions {@code ^} and {@code $} match 831 * just after or just before, respectively, a line terminator or the end of 832 * the input sequence. By default these expressions only match at the 833 * beginning and the end of the entire input sequence. 834 * 835 * <p> Multiline mode can also be enabled via the embedded flag 836 * expression {@code (?m)}. </p> 837 */ 838 public static final int MULTILINE = 0x08; 839 840 /** 841 * Enables literal parsing of the pattern. 842 * 843 * <p> When this flag is specified then the input string that specifies 844 * the pattern is treated as a sequence of literal characters. 845 * Metacharacters or escape sequences in the input sequence will be 846 * given no special meaning. 847 * 848 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on 849 * matching when used in conjunction with this flag. The other flags 850 * become superfluous. 851 * 852 * <p> There is no embedded flag character for enabling literal parsing. 853 * @since 1.5 854 */ 855 public static final int LITERAL = 0x10; 856 857 /** 858 * Enables dotall mode. 859 * 860 * <p> In dotall mode, the expression {@code .} matches any character, 861 * including a line terminator. By default this expression does not match 862 * line terminators. 863 * 864 * <p> Dotall mode can also be enabled via the embedded flag 865 * expression {@code (?s)}. (The {@code s} is a mnemonic for 866 * "single-line" mode, which is what this is called in Perl.) </p> 867 */ 868 public static final int DOTALL = 0x20; 869 870 /** 871 * Enables Unicode-aware case folding. 872 * 873 * <p> When this flag is specified then case-insensitive matching, when 874 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner 875 * consistent with the Unicode Standard. By default, case-insensitive 876 * matching assumes that only characters in the US-ASCII charset are being 877 * matched. 878 * 879 * <p> Unicode-aware case folding can also be enabled via the embedded flag 880 * expression {@code (?u)}. 881 * 882 * <p> Specifying this flag may impose a performance penalty. </p> 883 */ 884 public static final int UNICODE_CASE = 0x40; 885 886 /** 887 * Enables canonical equivalence. 888 * 889 * <p> When this flag is specified then two characters will be considered 890 * to match if, and only if, their full canonical decompositions match. 891 * The expression <code>"a\u030A"</code>, for example, will match the 892 * string <code>"\u00E5"</code> when this flag is specified. By default, 893 * matching does not take canonical equivalence into account. 894 * 895 * <p> There is no embedded flag character for enabling canonical 896 * equivalence. 897 * 898 * <p> Specifying this flag may impose a performance penalty. </p> 899 */ 900 public static final int CANON_EQ = 0x80; 901 902 /** 903 * Enables the Unicode version of <i>Predefined character classes</i> and 904 * <i>POSIX character classes</i>. 905 * 906 * <p> When this flag is specified then the (US-ASCII only) 907 * <i>Predefined character classes</i> and <i>POSIX character classes</i> 908 * are in conformance with 909 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 910 * Standard #18: Unicode Regular Expression</i></a> 911 * <i>Annex C: Compatibility Properties</i>. 912 * <p> 913 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded 914 * flag expression {@code (?U)}. 915 * <p> 916 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case 917 * folding. 918 * <p> 919 * Specifying this flag may impose a performance penalty. </p> 920 * @since 1.7 921 */ 922 public static final int UNICODE_CHARACTER_CLASS = 0x100; 923 924 /** 925 * Contains all possible flags for compile(regex, flags). 926 */ 927 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE | 928 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL | 929 UNICODE_CHARACTER_CLASS | COMMENTS; 930 931 /* Pattern has only two serialized components: The pattern string 932 * and the flags, which are all that is needed to recompile the pattern 933 * when it is deserialized. 934 */ 935 936 /** use serialVersionUID from Merlin b59 for interoperability */ 937 @java.io.Serial 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 @java.io.Serial 1380 private void readObject(java.io.ObjectInputStream s) 1381 throws java.io.IOException, ClassNotFoundException { 1382 1383 // Read in all fields 1384 s.defaultReadObject(); 1385 1386 // reset the flags 1387 flags0 = flags; 1388 1389 // Initialize counts 1390 capturingGroupCount = 1; 1391 localCount = 0; 1392 localTCNCount = 0; 1393 1394 // if length > 0, the Pattern is lazily compiled 1395 if (pattern.isEmpty()) { 1396 root = new Start(lastAccept); 1397 matchRoot = lastAccept; 1398 compiled = true; 1399 } 1400 } 1401 1402 /** 1403 * This private constructor is used to create all Patterns. The pattern 1404 * string and match flags are all that is needed to completely describe 1405 * a Pattern. An empty pattern string results in an object tree with 1406 * only a Start node and a LastNode node. 1407 */ 1408 private Pattern(String p, int f) { 1409 if ((f & ~ALL_FLAGS) != 0) { 1410 throw new IllegalArgumentException("Unknown flag 0x" 1411 + Integer.toHexString(f)); 1412 } 1413 pattern = p; 1414 flags = f; 1415 1416 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present 1417 if ((flags & UNICODE_CHARACTER_CLASS) != 0) 1418 flags |= UNICODE_CASE; 1419 1420 // 'flags' for compiling 1421 flags0 = flags; 1422 1423 // Reset group index count 1424 capturingGroupCount = 1; 1425 localCount = 0; 1426 localTCNCount = 0; 1427 1428 if (!pattern.isEmpty()) { 1429 try { 1430 compile(); 1431 } catch (StackOverflowError soe) { 1432 throw error("Stack overflow during pattern compilation"); 1433 } 1434 } else { 1435 root = new Start(lastAccept); 1436 matchRoot = lastAccept; 1437 } 1438 } 1439 1440 /** 1441 * The pattern is converted to normalized form ({@link 1442 * java.text.Normalizer.Form#NFC NFC}, canonical decomposition, 1443 * followed by canonical composition for the character class 1444 * part, and {@link java.text.Normalizer.Form#NFD NFD}, 1445 * canonical decomposition for the rest), and then a pure 1446 * group is constructed to match canonical equivalences of the 1447 * characters. 1448 */ 1449 private static String normalize(String pattern) { 1450 int plen = pattern.length(); 1451 StringBuilder pbuf = new StringBuilder(plen); 1452 char last = 0; 1453 int lastStart = 0; 1454 char cc = 0; 1455 for (int i = 0; i < plen;) { 1456 char c = pattern.charAt(i); 1457 if (cc == 0 && // top level 1458 c == '\\' && i + 1 < plen && pattern.charAt(i + 1) == '\\') { 1459 i += 2; last = 0; 1460 continue; 1461 } 1462 if (c == '[' && last != '\\') { 1463 if (cc == 0) { 1464 if (lastStart < i) 1465 normalizeSlice(pattern, lastStart, i, pbuf); 1466 lastStart = i; 1467 } 1468 cc++; 1469 } else if (c == ']' && last != '\\') { 1470 cc--; 1471 if (cc == 0) { 1472 normalizeClazz(pattern, lastStart, i + 1, pbuf); 1473 lastStart = i + 1; 1474 } 1475 } 1476 last = c; 1477 i++; 1478 } 1479 assert (cc == 0); 1480 if (lastStart < plen) 1481 normalizeSlice(pattern, lastStart, plen, pbuf); 1482 return pbuf.toString(); 1483 } 1484 1485 private static void normalizeSlice(String src, int off, int limit, 1486 StringBuilder dst) 1487 { 1488 int len = src.length(); 1489 int off0 = off; 1490 while (off < limit && ASCII.isAscii(src.charAt(off))) { 1491 off++; 1492 } 1493 if (off == limit) { 1494 dst.append(src, off0, limit); 1495 return; 1496 } 1497 off--; 1498 if (off < off0) 1499 off = off0; 1500 else 1501 dst.append(src, off0, off); 1502 while (off < limit) { 1503 int ch0 = src.codePointAt(off); 1504 if (".$|()[]{}^?*+\\".indexOf(ch0) != -1) { 1505 dst.append((char)ch0); 1506 off++; 1507 continue; 1508 } 1509 int j = Grapheme.nextBoundary(src, off, limit); 1510 int ch1; 1511 String seq = src.substring(off, j); 1512 String nfd = Normalizer.normalize(seq, Normalizer.Form.NFD); 1513 off = j; 1514 if (nfd.codePointCount(0, nfd.length()) > 1) { 1515 ch0 = nfd.codePointAt(0); 1516 ch1 = nfd.codePointAt(Character.charCount(ch0)); 1517 if (Character.getType(ch1) == Character.NON_SPACING_MARK) { 1518 Set<String> altns = new LinkedHashSet<>(); 1519 altns.add(seq); 1520 produceEquivalentAlternation(nfd, altns); 1521 dst.append("(?:"); 1522 altns.forEach( s -> dst.append(s).append('|')); 1523 dst.delete(dst.length() - 1, dst.length()); 1524 dst.append(")"); 1525 continue; 1526 } 1527 } 1528 String nfc = Normalizer.normalize(seq, Normalizer.Form.NFC); 1529 if (!seq.equals(nfc) && !nfd.equals(nfc)) 1530 dst.append("(?:" + seq + "|" + nfd + "|" + nfc + ")"); 1531 else if (!seq.equals(nfd)) 1532 dst.append("(?:" + seq + "|" + nfd + ")"); 1533 else 1534 dst.append(seq); 1535 } 1536 } 1537 1538 private static void normalizeClazz(String src, int off, int limit, 1539 StringBuilder dst) 1540 { 1541 dst.append(Normalizer.normalize(src.substring(off, limit), Form.NFC)); 1542 } 1543 1544 /** 1545 * Given a specific sequence composed of a regular character and 1546 * combining marks that follow it, produce the alternation that will 1547 * match all canonical equivalences of that sequence. 1548 */ 1549 private static void produceEquivalentAlternation(String src, 1550 Set<String> dst) 1551 { 1552 int len = countChars(src, 0, 1); 1553 if (src.length() == len) { 1554 dst.add(src); // source has one character. 1555 return; 1556 } 1557 String base = src.substring(0,len); 1558 String combiningMarks = src.substring(len); 1559 String[] perms = producePermutations(combiningMarks); 1560 // Add combined permutations 1561 for(int x = 0; x < perms.length; x++) { 1562 String next = base + perms[x]; 1563 dst.add(next); 1564 next = composeOneStep(next); 1565 if (next != null) { 1566 produceEquivalentAlternation(next, dst); 1567 } 1568 } 1569 } 1570 1571 /** 1572 * Returns an array of strings that have all the possible 1573 * permutations of the characters in the input string. 1574 * This is used to get a list of all possible orderings 1575 * of a set of combining marks. Note that some of the permutations 1576 * are invalid because of combining class collisions, and these 1577 * possibilities must be removed because they are not canonically 1578 * equivalent. 1579 */ 1580 private static String[] producePermutations(String input) { 1581 if (input.length() == countChars(input, 0, 1)) 1582 return new String[] {input}; 1583 1584 if (input.length() == countChars(input, 0, 2)) { 1585 int c0 = Character.codePointAt(input, 0); 1586 int c1 = Character.codePointAt(input, Character.charCount(c0)); 1587 if (getClass(c1) == getClass(c0)) { 1588 return new String[] {input}; 1589 } 1590 String[] result = new String[2]; 1591 result[0] = input; 1592 StringBuilder sb = new StringBuilder(2); 1593 sb.appendCodePoint(c1); 1594 sb.appendCodePoint(c0); 1595 result[1] = sb.toString(); 1596 return result; 1597 } 1598 1599 int length = 1; 1600 int nCodePoints = countCodePoints(input); 1601 for(int x=1; x<nCodePoints; x++) 1602 length = length * (x+1); 1603 1604 String[] temp = new String[length]; 1605 1606 int combClass[] = new int[nCodePoints]; 1607 for(int x=0, i=0; x<nCodePoints; x++) { 1608 int c = Character.codePointAt(input, i); 1609 combClass[x] = getClass(c); 1610 i += Character.charCount(c); 1611 } 1612 1613 // For each char, take it out and add the permutations 1614 // of the remaining chars 1615 int index = 0; 1616 int len; 1617 // offset maintains the index in code units. 1618 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) { 1619 len = countChars(input, offset, 1); 1620 for(int y=x-1; y>=0; y--) { 1621 if (combClass[y] == combClass[x]) { 1622 continue loop; 1623 } 1624 } 1625 StringBuilder sb = new StringBuilder(input); 1626 String otherChars = sb.delete(offset, offset+len).toString(); 1627 String[] subResult = producePermutations(otherChars); 1628 1629 String prefix = input.substring(offset, offset+len); 1630 for (String sre : subResult) 1631 temp[index++] = prefix + sre; 1632 } 1633 String[] result = new String[index]; 1634 System.arraycopy(temp, 0, result, 0, index); 1635 return result; 1636 } 1637 1638 private static int getClass(int c) { 1639 return sun.text.Normalizer.getCombiningClass(c); 1640 } 1641 1642 /** 1643 * Attempts to compose input by combining the first character 1644 * with the first combining mark following it. Returns a String 1645 * that is the composition of the leading character with its first 1646 * combining mark followed by the remaining combining marks. Returns 1647 * null if the first two characters cannot be further composed. 1648 */ 1649 private static String composeOneStep(String input) { 1650 int len = countChars(input, 0, 2); 1651 String firstTwoCharacters = input.substring(0, len); 1652 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC); 1653 if (result.equals(firstTwoCharacters)) 1654 return null; 1655 else { 1656 String remainder = input.substring(len); 1657 return result + remainder; 1658 } 1659 } 1660 1661 /** 1662 * Preprocess any \Q...\E sequences in `temp', meta-quoting them. 1663 * See the description of `quotemeta' in perlfunc(1). 1664 */ 1665 private void RemoveQEQuoting() { 1666 final int pLen = patternLength; 1667 int i = 0; 1668 while (i < pLen-1) { 1669 if (temp[i] != '\\') 1670 i += 1; 1671 else if (temp[i + 1] != 'Q') 1672 i += 2; 1673 else 1674 break; 1675 } 1676 if (i >= pLen - 1) // No \Q sequence found 1677 return; 1678 int j = i; 1679 i += 2; 1680 int newTempLen; 1681 try { 1682 newTempLen = Math.addExact(j + 2, Math.multiplyExact(3, pLen - i)); 1683 } catch (ArithmeticException ae) { 1684 throw new OutOfMemoryError("Required pattern length too large"); 1685 } 1686 int[] newtemp = new int[newTempLen]; 1687 System.arraycopy(temp, 0, newtemp, 0, j); 1688 1689 boolean inQuote = true; 1690 boolean beginQuote = true; 1691 while (i < pLen) { 1692 int c = temp[i++]; 1693 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) { 1694 newtemp[j++] = c; 1695 } else if (ASCII.isDigit(c)) { 1696 if (beginQuote) { 1697 /* 1698 * A unicode escape \[0xu] could be before this quote, 1699 * and we don't want this numeric char to processed as 1700 * part of the escape. 1701 */ 1702 newtemp[j++] = '\\'; 1703 newtemp[j++] = 'x'; 1704 newtemp[j++] = '3'; 1705 } 1706 newtemp[j++] = c; 1707 } else if (c != '\\') { 1708 if (inQuote) newtemp[j++] = '\\'; 1709 newtemp[j++] = c; 1710 } else if (inQuote) { 1711 if (temp[i] == 'E') { 1712 i++; 1713 inQuote = false; 1714 } else { 1715 newtemp[j++] = '\\'; 1716 newtemp[j++] = '\\'; 1717 } 1718 } else { 1719 if (temp[i] == 'Q') { 1720 i++; 1721 inQuote = true; 1722 beginQuote = true; 1723 continue; 1724 } else { 1725 newtemp[j++] = c; 1726 if (i != pLen) 1727 newtemp[j++] = temp[i++]; 1728 } 1729 } 1730 1731 beginQuote = false; 1732 } 1733 1734 patternLength = j; 1735 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination 1736 } 1737 1738 /** 1739 * Copies regular expression to an int array and invokes the parsing 1740 * of the expression which will create the object tree. 1741 */ 1742 private void compile() { 1743 // Handle canonical equivalences 1744 if (has(CANON_EQ) && !has(LITERAL)) { 1745 normalizedPattern = normalize(pattern); 1746 } else { 1747 normalizedPattern = pattern; 1748 } 1749 patternLength = normalizedPattern.length(); 1750 1751 // Copy pattern to int array for convenience 1752 // Use double zero to terminate pattern 1753 temp = new int[patternLength + 2]; 1754 1755 hasSupplementary = false; 1756 int c, count = 0; 1757 // Convert all chars into code points 1758 for (int x = 0; x < patternLength; x += Character.charCount(c)) { 1759 c = normalizedPattern.codePointAt(x); 1760 if (isSupplementary(c)) { 1761 hasSupplementary = true; 1762 } 1763 temp[count++] = c; 1764 } 1765 1766 patternLength = count; // patternLength now in code points 1767 1768 if (! has(LITERAL)) 1769 RemoveQEQuoting(); 1770 1771 // Allocate all temporary objects here. 1772 buffer = new int[32]; 1773 groupNodes = new GroupHead[10]; 1774 namedGroups = null; 1775 topClosureNodes = new ArrayList<>(10); 1776 1777 if (has(LITERAL)) { 1778 // Literal pattern handling 1779 matchRoot = newSlice(temp, patternLength, hasSupplementary); 1780 matchRoot.next = lastAccept; 1781 } else { 1782 // Start recursive descent parsing 1783 matchRoot = expr(lastAccept); 1784 // Check extra pattern characters 1785 if (patternLength != cursor) { 1786 if (peek() == ')') { 1787 throw error("Unmatched closing ')'"); 1788 } else { 1789 throw error("Unexpected internal error"); 1790 } 1791 } 1792 } 1793 1794 // Peephole optimization 1795 if (matchRoot instanceof Slice) { 1796 root = BnM.optimize(matchRoot); 1797 if (root == matchRoot) { 1798 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1799 } 1800 } else if (matchRoot instanceof Begin || matchRoot instanceof First) { 1801 root = matchRoot; 1802 } else { 1803 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1804 } 1805 1806 // Optimize the greedy Loop to prevent exponential backtracking, IF there 1807 // is no group ref in this pattern. With a non-negative localTCNCount value, 1808 // the greedy type Loop, Curly will skip the backtracking for any starting 1809 // position "i" that failed in the past. 1810 if (!hasGroupRef) { 1811 for (Node node : topClosureNodes) { 1812 if (node instanceof Loop) { 1813 // non-deterministic-greedy-group 1814 ((Loop)node).posIndex = localTCNCount++; 1815 } 1816 } 1817 } 1818 1819 // Release temporary storage 1820 temp = null; 1821 buffer = null; 1822 groupNodes = null; 1823 patternLength = 0; 1824 compiled = true; 1825 topClosureNodes = null; 1826 } 1827 1828 Map<String, Integer> namedGroups() { 1829 Map<String, Integer> groups = namedGroups; 1830 if (groups == null) { 1831 namedGroups = groups = new HashMap<>(2); 1832 } 1833 return groups; 1834 } 1835 1836 /** 1837 * Used to accumulate information about a subtree of the object graph 1838 * so that optimizations can be applied to the subtree. 1839 */ 1840 static final class TreeInfo { 1841 int minLength; 1842 int maxLength; 1843 boolean maxValid; 1844 boolean deterministic; 1845 1846 TreeInfo() { 1847 reset(); 1848 } 1849 void reset() { 1850 minLength = 0; 1851 maxLength = 0; 1852 maxValid = true; 1853 deterministic = true; 1854 } 1855 } 1856 1857 /* 1858 * The following private methods are mainly used to improve the 1859 * readability of the code. In order to let the Java compiler easily 1860 * inline them, we should not put many assertions or error checks in them. 1861 */ 1862 1863 /** 1864 * Indicates whether a particular flag is set or not. 1865 */ 1866 private boolean has(int f) { 1867 return (flags0 & f) != 0; 1868 } 1869 1870 /** 1871 * Match next character, signal error if failed. 1872 */ 1873 private void accept(int ch, String s) { 1874 int testChar = temp[cursor++]; 1875 if (has(COMMENTS)) 1876 testChar = parsePastWhitespace(testChar); 1877 if (ch != testChar) { 1878 throw error(s); 1879 } 1880 } 1881 1882 /** 1883 * Mark the end of pattern with a specific character. 1884 */ 1885 private void mark(int c) { 1886 temp[patternLength] = c; 1887 } 1888 1889 /** 1890 * Peek the next character, and do not advance the cursor. 1891 */ 1892 private int peek() { 1893 int ch = temp[cursor]; 1894 if (has(COMMENTS)) 1895 ch = peekPastWhitespace(ch); 1896 return ch; 1897 } 1898 1899 /** 1900 * Read the next character, and advance the cursor by one. 1901 */ 1902 private int read() { 1903 int ch = temp[cursor++]; 1904 if (has(COMMENTS)) 1905 ch = parsePastWhitespace(ch); 1906 return ch; 1907 } 1908 1909 /** 1910 * Read the next character, and advance the cursor by one, 1911 * ignoring the COMMENTS setting 1912 */ 1913 private int readEscaped() { 1914 int ch = temp[cursor++]; 1915 return ch; 1916 } 1917 1918 /** 1919 * Advance the cursor by one, and peek the next character. 1920 */ 1921 private int next() { 1922 int ch = temp[++cursor]; 1923 if (has(COMMENTS)) 1924 ch = peekPastWhitespace(ch); 1925 return ch; 1926 } 1927 1928 /** 1929 * Advance the cursor by one, and peek the next character, 1930 * ignoring the COMMENTS setting 1931 */ 1932 private int nextEscaped() { 1933 int ch = temp[++cursor]; 1934 return ch; 1935 } 1936 1937 /** 1938 * If in xmode peek past whitespace and comments. 1939 */ 1940 private int peekPastWhitespace(int ch) { 1941 while (ASCII.isSpace(ch) || ch == '#') { 1942 while (ASCII.isSpace(ch)) 1943 ch = temp[++cursor]; 1944 if (ch == '#') { 1945 ch = peekPastLine(); 1946 } 1947 } 1948 return ch; 1949 } 1950 1951 /** 1952 * If in xmode parse past whitespace and comments. 1953 */ 1954 private int parsePastWhitespace(int ch) { 1955 while (ASCII.isSpace(ch) || ch == '#') { 1956 while (ASCII.isSpace(ch)) 1957 ch = temp[cursor++]; 1958 if (ch == '#') 1959 ch = parsePastLine(); 1960 } 1961 return ch; 1962 } 1963 1964 /** 1965 * xmode parse past comment to end of line. 1966 */ 1967 private int parsePastLine() { 1968 int ch = temp[cursor++]; 1969 while (ch != 0 && !isLineSeparator(ch)) 1970 ch = temp[cursor++]; 1971 if (ch == 0 && cursor > patternLength) { 1972 cursor = patternLength; 1973 ch = temp[cursor++]; 1974 } 1975 return ch; 1976 } 1977 1978 /** 1979 * xmode peek past comment to end of line. 1980 */ 1981 private int peekPastLine() { 1982 int ch = temp[++cursor]; 1983 while (ch != 0 && !isLineSeparator(ch)) 1984 ch = temp[++cursor]; 1985 if (ch == 0 && cursor > patternLength) { 1986 cursor = patternLength; 1987 ch = temp[cursor]; 1988 } 1989 return ch; 1990 } 1991 1992 /** 1993 * Determines if character is a line separator in the current mode 1994 */ 1995 private boolean isLineSeparator(int ch) { 1996 if (has(UNIX_LINES)) { 1997 return ch == '\n'; 1998 } else { 1999 return (ch == '\n' || 2000 ch == '\r' || 2001 (ch|1) == '\u2029' || 2002 ch == '\u0085'); 2003 } 2004 } 2005 2006 /** 2007 * Read the character after the next one, and advance the cursor by two. 2008 */ 2009 private int skip() { 2010 int i = cursor; 2011 int ch = temp[i+1]; 2012 cursor = i + 2; 2013 return ch; 2014 } 2015 2016 /** 2017 * Unread one next character, and retreat cursor by one. 2018 */ 2019 private void unread() { 2020 cursor--; 2021 } 2022 2023 /** 2024 * Internal method used for handling all syntax errors. The pattern is 2025 * displayed with a pointer to aid in locating the syntax error. 2026 */ 2027 private PatternSyntaxException error(String s) { 2028 return new PatternSyntaxException(s, normalizedPattern, cursor - 1); 2029 } 2030 2031 /** 2032 * Determines if there is any supplementary character or unpaired 2033 * surrogate in the specified range. 2034 */ 2035 private boolean findSupplementary(int start, int end) { 2036 for (int i = start; i < end; i++) { 2037 if (isSupplementary(temp[i])) 2038 return true; 2039 } 2040 return false; 2041 } 2042 2043 /** 2044 * Determines if the specified code point is a supplementary 2045 * character or unpaired surrogate. 2046 */ 2047 private static final boolean isSupplementary(int ch) { 2048 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || 2049 Character.isSurrogate((char)ch); 2050 } 2051 2052 /** 2053 * The following methods handle the main parsing. They are sorted 2054 * according to their precedence order, the lowest one first. 2055 */ 2056 2057 /** 2058 * The expression is parsed with branch nodes added for alternations. 2059 * This may be called recursively to parse sub expressions that may 2060 * contain alternations. 2061 */ 2062 private Node expr(Node end) { 2063 Node prev = null; 2064 Node firstTail = null; 2065 Branch branch = null; 2066 BranchConn branchConn = null; 2067 2068 for (;;) { 2069 Node node = sequence(end); 2070 Node nodeTail = root; //double return 2071 if (prev == null) { 2072 prev = node; 2073 firstTail = nodeTail; 2074 } else { 2075 // Branch 2076 if (branchConn == null) { 2077 branchConn = new BranchConn(); 2078 branchConn.next = end; 2079 } 2080 if (node == end) { 2081 // if the node returned from sequence() is "end" 2082 // we have an empty expr, set a null atom into 2083 // the branch to indicate to go "next" directly. 2084 node = null; 2085 } else { 2086 // the "tail.next" of each atom goes to branchConn 2087 nodeTail.next = branchConn; 2088 } 2089 if (prev == branch) { 2090 branch.add(node); 2091 } else { 2092 if (prev == end) { 2093 prev = null; 2094 } else { 2095 // replace the "end" with "branchConn" at its tail.next 2096 // when put the "prev" into the branch as the first atom. 2097 firstTail.next = branchConn; 2098 } 2099 prev = branch = new Branch(prev, node, branchConn); 2100 } 2101 } 2102 if (peek() != '|') { 2103 return prev; 2104 } 2105 next(); 2106 } 2107 } 2108 2109 @SuppressWarnings("fallthrough") 2110 /** 2111 * Parsing of sequences between alternations. 2112 */ 2113 private Node sequence(Node end) { 2114 Node head = null; 2115 Node tail = null; 2116 Node node; 2117 LOOP: 2118 for (;;) { 2119 int ch = peek(); 2120 switch (ch) { 2121 case '(': 2122 // Because group handles its own closure, 2123 // we need to treat it differently 2124 node = group0(); 2125 // Check for comment or flag group 2126 if (node == null) 2127 continue; 2128 if (head == null) 2129 head = node; 2130 else 2131 tail.next = node; 2132 // Double return: Tail was returned in root 2133 tail = root; 2134 continue; 2135 case '[': 2136 if (has(CANON_EQ) && !has(LITERAL)) 2137 node = new NFCCharProperty(clazz(true)); 2138 else 2139 node = newCharProperty(clazz(true)); 2140 break; 2141 case '\\': 2142 ch = nextEscaped(); 2143 if (ch == 'p' || ch == 'P') { 2144 boolean oneLetter = true; 2145 boolean comp = (ch == 'P'); 2146 ch = next(); // Consume { if present 2147 if (ch != '{') { 2148 unread(); 2149 } else { 2150 oneLetter = false; 2151 } 2152 // node = newCharProperty(family(oneLetter, comp)); 2153 if (has(CANON_EQ) && !has(LITERAL)) 2154 node = new NFCCharProperty(family(oneLetter, comp)); 2155 else 2156 node = newCharProperty(family(oneLetter, comp)); 2157 } else { 2158 unread(); 2159 node = atom(); 2160 } 2161 break; 2162 case '^': 2163 next(); 2164 if (has(MULTILINE)) { 2165 if (has(UNIX_LINES)) 2166 node = new UnixCaret(); 2167 else 2168 node = new Caret(); 2169 } else { 2170 node = new Begin(); 2171 } 2172 break; 2173 case '$': 2174 next(); 2175 if (has(UNIX_LINES)) 2176 node = new UnixDollar(has(MULTILINE)); 2177 else 2178 node = new Dollar(has(MULTILINE)); 2179 break; 2180 case '.': 2181 next(); 2182 if (has(DOTALL)) { 2183 node = new CharProperty(ALL()); 2184 } else { 2185 if (has(UNIX_LINES)) { 2186 node = new CharProperty(UNIXDOT()); 2187 } else { 2188 node = new CharProperty(DOT()); 2189 } 2190 } 2191 break; 2192 case '|': 2193 case ')': 2194 break LOOP; 2195 case ']': // Now interpreting dangling ] and } as literals 2196 case '}': 2197 node = atom(); 2198 break; 2199 case '?': 2200 case '*': 2201 case '+': 2202 next(); 2203 throw error("Dangling meta character '" + ((char)ch) + "'"); 2204 case 0: 2205 if (cursor >= patternLength) { 2206 break LOOP; 2207 } 2208 // Fall through 2209 default: 2210 node = atom(); 2211 break; 2212 } 2213 2214 if (node instanceof LineEnding) { 2215 LineEnding le = (LineEnding)node; 2216 node = closureOfLineEnding(le); 2217 2218 if (node != le) { 2219 // LineEnding was replaced with an anonymous group 2220 if (head == null) 2221 head = node; 2222 else 2223 tail.next = node; 2224 // Double return: Tail was returned in root 2225 tail = root; 2226 continue; 2227 } 2228 } else { 2229 node = closure(node); 2230 } 2231 2232 /* save the top dot-greedy nodes (.*, .+) as well 2233 if (node instanceof GreedyCharProperty && 2234 ((GreedyCharProperty)node).cp instanceof Dot) { 2235 topClosureNodes.add(node); 2236 } 2237 */ 2238 if (head == null) { 2239 head = tail = node; 2240 } else { 2241 tail.next = node; 2242 tail = node; 2243 } 2244 } 2245 if (head == null) { 2246 return end; 2247 } 2248 tail.next = end; 2249 root = tail; //double return 2250 return head; 2251 } 2252 2253 @SuppressWarnings("fallthrough") 2254 /** 2255 * Parse and add a new Single or Slice. 2256 */ 2257 private Node atom() { 2258 int first = 0; 2259 int prev = -1; 2260 boolean hasSupplementary = false; 2261 int ch = peek(); 2262 for (;;) { 2263 switch (ch) { 2264 case '*': 2265 case '+': 2266 case '?': 2267 case '{': 2268 if (first > 1) { 2269 cursor = prev; // Unwind one character 2270 first--; 2271 } 2272 break; 2273 case '$': 2274 case '.': 2275 case '^': 2276 case '(': 2277 case '[': 2278 case '|': 2279 case ')': 2280 break; 2281 case '\\': 2282 ch = nextEscaped(); 2283 if (ch == 'p' || ch == 'P') { // Property 2284 if (first > 0) { // Slice is waiting; handle it first 2285 unread(); 2286 break; 2287 } else { // No slice; just return the family node 2288 boolean comp = (ch == 'P'); 2289 boolean oneLetter = true; 2290 ch = next(); // Consume { if present 2291 if (ch != '{') 2292 unread(); 2293 else 2294 oneLetter = false; 2295 if (has(CANON_EQ) && !has(LITERAL)) 2296 return new NFCCharProperty(family(oneLetter, comp)); 2297 else 2298 return newCharProperty(family(oneLetter, comp)); 2299 } 2300 } 2301 unread(); 2302 prev = cursor; 2303 ch = escape(false, first == 0, false); 2304 if (ch >= 0) { 2305 append(ch, first); 2306 first++; 2307 if (isSupplementary(ch)) { 2308 hasSupplementary = true; 2309 } 2310 ch = peek(); 2311 continue; 2312 } else if (first == 0) { 2313 return root; 2314 } 2315 // Unwind meta escape sequence 2316 cursor = prev; 2317 break; 2318 case 0: 2319 if (cursor >= patternLength) { 2320 break; 2321 } 2322 // Fall through 2323 default: 2324 prev = cursor; 2325 append(ch, first); 2326 first++; 2327 if (isSupplementary(ch)) { 2328 hasSupplementary = true; 2329 } 2330 ch = next(); 2331 continue; 2332 } 2333 break; 2334 } 2335 if (first == 1) { 2336 return newCharProperty(single(buffer[0])); 2337 } else { 2338 return newSlice(buffer, first, hasSupplementary); 2339 } 2340 } 2341 2342 private void append(int ch, int index) { 2343 int len = buffer.length; 2344 if (index - len >= 0) { 2345 len = ArraysSupport.newLength(len, 2346 1 + index - len, /* minimum growth */ 2347 len /* preferred growth */); 2348 buffer = Arrays.copyOf(buffer, len); 2349 } 2350 buffer[index] = ch; 2351 } 2352 2353 /** 2354 * Parses a backref greedily, taking as many numbers as it 2355 * can. The first digit is always treated as a backref, but 2356 * multi digit numbers are only treated as a backref if at 2357 * least that many backrefs exist at this point in the regex. 2358 */ 2359 private Node ref(int refNum) { 2360 boolean done = false; 2361 while(!done) { 2362 int ch = peek(); 2363 switch(ch) { 2364 case '0': 2365 case '1': 2366 case '2': 2367 case '3': 2368 case '4': 2369 case '5': 2370 case '6': 2371 case '7': 2372 case '8': 2373 case '9': 2374 int newRefNum = (refNum * 10) + (ch - '0'); 2375 // Add another number if it doesn't make a group 2376 // that doesn't exist 2377 if (capturingGroupCount - 1 < newRefNum) { 2378 done = true; 2379 break; 2380 } 2381 refNum = newRefNum; 2382 read(); 2383 break; 2384 default: 2385 done = true; 2386 break; 2387 } 2388 } 2389 hasGroupRef = true; 2390 if (has(CASE_INSENSITIVE)) 2391 return new CIBackRef(refNum, has(UNICODE_CASE)); 2392 else 2393 return new BackRef(refNum); 2394 } 2395 2396 /** 2397 * Parses an escape sequence to determine the actual value that needs 2398 * to be matched. 2399 * If -1 is returned and create was true a new object was added to the tree 2400 * to handle the escape sequence. 2401 * If the returned value is greater than zero, it is the value that 2402 * matches the escape sequence. 2403 */ 2404 private int escape(boolean inclass, boolean create, boolean isrange) { 2405 int ch = skip(); 2406 switch (ch) { 2407 case '0': 2408 return o(); 2409 case '1': 2410 case '2': 2411 case '3': 2412 case '4': 2413 case '5': 2414 case '6': 2415 case '7': 2416 case '8': 2417 case '9': 2418 if (inclass) break; 2419 if (create) { 2420 root = ref((ch - '0')); 2421 } 2422 return -1; 2423 case 'A': 2424 if (inclass) break; 2425 if (create) root = new Begin(); 2426 return -1; 2427 case 'B': 2428 if (inclass) break; 2429 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS)); 2430 return -1; 2431 case 'C': 2432 break; 2433 case 'D': 2434 if (create) { 2435 predicate = has(UNICODE_CHARACTER_CLASS) ? 2436 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2437 predicate = predicate.negate(); 2438 if (!inclass) 2439 root = newCharProperty(predicate); 2440 } 2441 return -1; 2442 case 'E': 2443 case 'F': 2444 break; 2445 case 'G': 2446 if (inclass) break; 2447 if (create) root = new LastMatch(); 2448 return -1; 2449 case 'H': 2450 if (create) { 2451 predicate = HorizWS().negate(); 2452 if (!inclass) 2453 root = newCharProperty(predicate); 2454 } 2455 return -1; 2456 case 'I': 2457 case 'J': 2458 case 'K': 2459 case 'L': 2460 case 'M': 2461 break; 2462 case 'N': 2463 return N(); 2464 case 'O': 2465 case 'P': 2466 case 'Q': 2467 break; 2468 case 'R': 2469 if (inclass) break; 2470 if (create) root = new LineEnding(); 2471 return -1; 2472 case 'S': 2473 if (create) { 2474 predicate = has(UNICODE_CHARACTER_CLASS) ? 2475 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2476 predicate = predicate.negate(); 2477 if (!inclass) 2478 root = newCharProperty(predicate); 2479 } 2480 return -1; 2481 case 'T': 2482 case 'U': 2483 break; 2484 case 'V': 2485 if (create) { 2486 predicate = VertWS().negate(); 2487 if (!inclass) 2488 root = newCharProperty(predicate); 2489 } 2490 return -1; 2491 case 'W': 2492 if (create) { 2493 predicate = has(UNICODE_CHARACTER_CLASS) ? 2494 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2495 predicate = predicate.negate(); 2496 if (!inclass) 2497 root = newCharProperty(predicate); 2498 } 2499 return -1; 2500 case 'X': 2501 if (inclass) break; 2502 if (create) { 2503 root = new XGrapheme(); 2504 } 2505 return -1; 2506 case 'Y': 2507 break; 2508 case 'Z': 2509 if (inclass) break; 2510 if (create) { 2511 if (has(UNIX_LINES)) 2512 root = new UnixDollar(false); 2513 else 2514 root = new Dollar(false); 2515 } 2516 return -1; 2517 case 'a': 2518 return '\007'; 2519 case 'b': 2520 if (inclass) break; 2521 if (create) { 2522 if (peek() == '{') { 2523 if (skip() == 'g') { 2524 if (read() == '}') { 2525 root = new GraphemeBound(); 2526 return -1; 2527 } 2528 break; // error missing trailing } 2529 } 2530 unread(); unread(); 2531 } 2532 root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS)); 2533 } 2534 return -1; 2535 case 'c': 2536 return c(); 2537 case 'd': 2538 if (create) { 2539 predicate = has(UNICODE_CHARACTER_CLASS) ? 2540 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2541 if (!inclass) 2542 root = newCharProperty(predicate); 2543 } 2544 return -1; 2545 case 'e': 2546 return '\033'; 2547 case 'f': 2548 return '\f'; 2549 case 'g': 2550 break; 2551 case 'h': 2552 if (create) { 2553 predicate = HorizWS(); 2554 if (!inclass) 2555 root = newCharProperty(predicate); 2556 } 2557 return -1; 2558 case 'i': 2559 case 'j': 2560 break; 2561 case 'k': 2562 if (inclass) 2563 break; 2564 if (read() != '<') 2565 throw error("\\k is not followed by '<' for named capturing group"); 2566 String name = groupname(read()); 2567 if (!namedGroups().containsKey(name)) 2568 throw error("named capturing group <" + name + "> does not exist"); 2569 if (create) { 2570 hasGroupRef = true; 2571 if (has(CASE_INSENSITIVE)) 2572 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE)); 2573 else 2574 root = new BackRef(namedGroups().get(name)); 2575 } 2576 return -1; 2577 case 'l': 2578 case 'm': 2579 break; 2580 case 'n': 2581 return '\n'; 2582 case 'o': 2583 case 'p': 2584 case 'q': 2585 break; 2586 case 'r': 2587 return '\r'; 2588 case 's': 2589 if (create) { 2590 predicate = has(UNICODE_CHARACTER_CLASS) ? 2591 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2592 if (!inclass) 2593 root = newCharProperty(predicate); 2594 } 2595 return -1; 2596 case 't': 2597 return '\t'; 2598 case 'u': 2599 return u(); 2600 case 'v': 2601 // '\v' was implemented as VT/0x0B in releases < 1.8 (though 2602 // undocumented). In JDK8 '\v' is specified as a predefined 2603 // character class for all vertical whitespace characters. 2604 // So [-1, root=VertWS node] pair is returned (instead of a 2605 // single 0x0B). This breaks the range if '\v' is used as 2606 // the start or end value, such as [\v-...] or [...-\v], in 2607 // which a single definite value (0x0B) is expected. For 2608 // compatibility concern '\013'/0x0B is returned if isrange. 2609 if (isrange) 2610 return '\013'; 2611 if (create) { 2612 predicate = VertWS(); 2613 if (!inclass) 2614 root = newCharProperty(predicate); 2615 } 2616 return -1; 2617 case 'w': 2618 if (create) { 2619 predicate = has(UNICODE_CHARACTER_CLASS) ? 2620 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2621 if (!inclass) 2622 root = newCharProperty(predicate); 2623 } 2624 return -1; 2625 case 'x': 2626 return x(); 2627 case 'y': 2628 break; 2629 case 'z': 2630 if (inclass) break; 2631 if (create) root = new End(); 2632 return -1; 2633 default: 2634 return ch; 2635 } 2636 throw error("Illegal/unsupported escape sequence"); 2637 } 2638 2639 /** 2640 * Parse a character class, and return the node that matches it. 2641 * 2642 * Consumes a ] on the way out if consume is true. Usually consume 2643 * is true except for the case of [abc&&def] where def is a separate 2644 * right hand node with "understood" brackets. 2645 */ 2646 private CharPredicate clazz(boolean consume) { 2647 CharPredicate prev = null; 2648 CharPredicate curr = null; 2649 BitClass bits = new BitClass(); 2650 2651 boolean isNeg = false; 2652 boolean hasBits = false; 2653 int ch = next(); 2654 2655 // Negates if first char in a class, otherwise literal 2656 if (ch == '^' && temp[cursor-1] == '[') { 2657 ch = next(); 2658 isNeg = true; 2659 } 2660 for (;;) { 2661 switch (ch) { 2662 case '[': 2663 curr = clazz(true); 2664 if (prev == null) 2665 prev = curr; 2666 else 2667 prev = prev.union(curr); 2668 ch = peek(); 2669 continue; 2670 case '&': 2671 ch = next(); 2672 if (ch == '&') { 2673 ch = next(); 2674 CharPredicate right = null; 2675 while (ch != ']' && ch != '&') { 2676 if (ch == '[') { 2677 if (right == null) 2678 right = clazz(true); 2679 else 2680 right = right.union(clazz(true)); 2681 } else { // abc&&def 2682 unread(); 2683 right = clazz(false); 2684 } 2685 ch = peek(); 2686 } 2687 if (hasBits) { 2688 // bits used, union has high precedence 2689 if (prev == null) { 2690 prev = curr = bits; 2691 } else { 2692 prev = prev.union(bits); 2693 } 2694 hasBits = false; 2695 } 2696 if (right != null) 2697 curr = right; 2698 if (prev == null) { 2699 if (right == null) 2700 throw error("Bad class syntax"); 2701 else 2702 prev = right; 2703 } else { 2704 prev = prev.and(curr); 2705 } 2706 } else { 2707 // treat as a literal & 2708 unread(); 2709 break; 2710 } 2711 continue; 2712 case 0: 2713 if (cursor >= patternLength) 2714 throw error("Unclosed character class"); 2715 break; 2716 case ']': 2717 if (prev != null || hasBits) { 2718 if (consume) 2719 next(); 2720 if (prev == null) 2721 prev = bits; 2722 else if (hasBits) 2723 prev = prev.union(bits); 2724 if (isNeg) 2725 return prev.negate(); 2726 return prev; 2727 } 2728 break; 2729 default: 2730 break; 2731 } 2732 curr = range(bits); 2733 if (curr == null) { // the bits used 2734 hasBits = true; 2735 } else { 2736 if (prev == null) 2737 prev = curr; 2738 else if (prev != curr) 2739 prev = prev.union(curr); 2740 } 2741 ch = peek(); 2742 } 2743 } 2744 2745 private CharPredicate bitsOrSingle(BitClass bits, int ch) { 2746 /* Bits can only handle codepoints in [u+0000-u+00ff] range. 2747 Use "single" node instead of bits when dealing with unicode 2748 case folding for codepoints listed below. 2749 (1)Uppercase out of range: u+00ff, u+00b5 2750 toUpperCase(u+00ff) -> u+0178 2751 toUpperCase(u+00b5) -> u+039c 2752 (2)LatinSmallLetterLongS u+17f 2753 toUpperCase(u+017f) -> u+0053 2754 (3)LatinSmallLetterDotlessI u+131 2755 toUpperCase(u+0131) -> u+0049 2756 (4)LatinCapitalLetterIWithDotAbove u+0130 2757 toLowerCase(u+0130) -> u+0069 2758 (5)KelvinSign u+212a 2759 toLowerCase(u+212a) ==> u+006B 2760 (6)AngstromSign u+212b 2761 toLowerCase(u+212b) ==> u+00e5 2762 */ 2763 if (ch < 256 && 2764 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) && 2765 (ch == 0xff || ch == 0xb5 || 2766 ch == 0x49 || ch == 0x69 || //I and i 2767 ch == 0x53 || ch == 0x73 || //S and s 2768 ch == 0x4b || ch == 0x6b || //K and k 2769 ch == 0xc5 || ch == 0xe5))) { //A+ring 2770 bits.add(ch, flags0); 2771 return null; 2772 } 2773 return single(ch); 2774 } 2775 2776 /** 2777 * Returns a suitably optimized, single character predicate 2778 */ 2779 private CharPredicate single(final int ch) { 2780 if (has(CASE_INSENSITIVE)) { 2781 int lower, upper; 2782 if (has(UNICODE_CASE)) { 2783 upper = Character.toUpperCase(ch); 2784 lower = Character.toLowerCase(upper); 2785 // Unicode case insensitive matches 2786 if (upper != lower) 2787 return SingleU(lower); 2788 } else if (ASCII.isAscii(ch)) { 2789 lower = ASCII.toLower(ch); 2790 upper = ASCII.toUpper(ch); 2791 // Case insensitive matches a given BMP character 2792 if (lower != upper) 2793 return SingleI(lower, upper); 2794 } 2795 } 2796 if (isSupplementary(ch)) 2797 return SingleS(ch); 2798 return Single(ch); // Match a given BMP character 2799 } 2800 2801 /** 2802 * Parse a single character or a character range in a character class 2803 * and return its representative node. 2804 */ 2805 private CharPredicate range(BitClass bits) { 2806 int ch = peek(); 2807 if (ch == '\\') { 2808 ch = nextEscaped(); 2809 if (ch == 'p' || ch == 'P') { // A property 2810 boolean comp = (ch == 'P'); 2811 boolean oneLetter = true; 2812 // Consume { if present 2813 ch = next(); 2814 if (ch != '{') 2815 unread(); 2816 else 2817 oneLetter = false; 2818 return family(oneLetter, comp); 2819 } else { // ordinary escape 2820 boolean isrange = temp[cursor+1] == '-'; 2821 unread(); 2822 ch = escape(true, true, isrange); 2823 if (ch == -1) 2824 return predicate; 2825 } 2826 } else { 2827 next(); 2828 } 2829 if (ch >= 0) { 2830 if (peek() == '-') { 2831 int endRange = temp[cursor+1]; 2832 if (endRange == '[') { 2833 return bitsOrSingle(bits, ch); 2834 } 2835 if (endRange != ']') { 2836 next(); 2837 int m = peek(); 2838 if (m == '\\') { 2839 m = escape(true, false, true); 2840 } else { 2841 next(); 2842 } 2843 if (m < ch) { 2844 throw error("Illegal character range"); 2845 } 2846 if (has(CASE_INSENSITIVE)) { 2847 if (has(UNICODE_CASE)) 2848 return CIRangeU(ch, m); 2849 return CIRange(ch, m); 2850 } else { 2851 return Range(ch, m); 2852 } 2853 } 2854 } 2855 return bitsOrSingle(bits, ch); 2856 } 2857 throw error("Unexpected character '"+((char)ch)+"'"); 2858 } 2859 2860 /** 2861 * Parses a Unicode character family and returns its representative node. 2862 */ 2863 private CharPredicate family(boolean singleLetter, boolean isComplement) { 2864 next(); 2865 String name; 2866 CharPredicate p = null; 2867 2868 if (singleLetter) { 2869 int c = temp[cursor]; 2870 if (!Character.isSupplementaryCodePoint(c)) { 2871 name = String.valueOf((char)c); 2872 } else { 2873 name = new String(temp, cursor, 1); 2874 } 2875 read(); 2876 } else { 2877 int i = cursor; 2878 mark('}'); 2879 while(read() != '}') { 2880 } 2881 mark('\000'); 2882 int j = cursor; 2883 if (j > patternLength) 2884 throw error("Unclosed character family"); 2885 if (i + 1 >= j) 2886 throw error("Empty character family"); 2887 name = new String(temp, i, j-i-1); 2888 } 2889 2890 int i = name.indexOf('='); 2891 if (i != -1) { 2892 // property construct \p{name=value} 2893 String value = name.substring(i + 1); 2894 name = name.substring(0, i).toLowerCase(Locale.ENGLISH); 2895 switch (name) { 2896 case "sc": 2897 case "script": 2898 p = CharPredicates.forUnicodeScript(value); 2899 break; 2900 case "blk": 2901 case "block": 2902 p = CharPredicates.forUnicodeBlock(value); 2903 break; 2904 case "gc": 2905 case "general_category": 2906 p = CharPredicates.forProperty(value, has(CASE_INSENSITIVE)); 2907 break; 2908 default: 2909 break; 2910 } 2911 if (p == null) 2912 throw error("Unknown Unicode property {name=<" + name + ">, " 2913 + "value=<" + value + ">}"); 2914 2915 } else { 2916 if (name.startsWith("In")) { 2917 // \p{InBlockName} 2918 p = CharPredicates.forUnicodeBlock(name.substring(2)); 2919 } else if (name.startsWith("Is")) { 2920 // \p{IsGeneralCategory} and \p{IsScriptName} 2921 String shortName = name.substring(2); 2922 p = CharPredicates.forUnicodeProperty(shortName, has(CASE_INSENSITIVE)); 2923 if (p == null) 2924 p = CharPredicates.forProperty(shortName, has(CASE_INSENSITIVE)); 2925 if (p == null) 2926 p = CharPredicates.forUnicodeScript(shortName); 2927 } else { 2928 if (has(UNICODE_CHARACTER_CLASS)) 2929 p = CharPredicates.forPOSIXName(name, has(CASE_INSENSITIVE)); 2930 if (p == null) 2931 p = CharPredicates.forProperty(name, has(CASE_INSENSITIVE)); 2932 } 2933 if (p == null) 2934 throw error("Unknown character property name {" + name + "}"); 2935 } 2936 if (isComplement) { 2937 // it might be too expensive to detect if a complement of 2938 // CharProperty can match "certain" supplementary. So just 2939 // go with StartS. 2940 hasSupplementary = true; 2941 p = p.negate(); 2942 } 2943 return p; 2944 } 2945 2946 private CharProperty newCharProperty(CharPredicate p) { 2947 if (p == null) 2948 return null; 2949 if (p instanceof BmpCharPredicate) 2950 return new BmpCharProperty((BmpCharPredicate)p); 2951 else 2952 return new CharProperty(p); 2953 } 2954 2955 /** 2956 * Parses and returns the name of a "named capturing group", the trailing 2957 * ">" is consumed after parsing. 2958 */ 2959 private String groupname(int ch) { 2960 StringBuilder sb = new StringBuilder(); 2961 if (!ASCII.isAlpha(ch)) 2962 throw error("capturing group name does not start with a Latin letter"); 2963 do { 2964 sb.append((char) ch); 2965 } while (ASCII.isAlnum(ch=read())); 2966 if (ch != '>') 2967 throw error("named capturing group is missing trailing '>'"); 2968 return sb.toString(); 2969 } 2970 2971 /** 2972 * Parses a group and returns the head node of a set of nodes that process 2973 * the group. Sometimes a double return system is used where the tail is 2974 * returned in root. 2975 */ 2976 private Node group0() { 2977 boolean capturingGroup = false; 2978 Node head; 2979 Node tail; 2980 int save = flags0; 2981 int saveTCNCount = topClosureNodes.size(); 2982 root = null; 2983 int ch = next(); 2984 if (ch == '?') { 2985 ch = skip(); 2986 switch (ch) { 2987 case ':': // (?:xxx) pure group 2988 head = createGroup(true); 2989 tail = root; 2990 head.next = expr(tail); 2991 break; 2992 case '=': // (?=xxx) and (?!xxx) lookahead 2993 case '!': 2994 head = createGroup(true); 2995 tail = root; 2996 head.next = expr(tail); 2997 if (ch == '=') { 2998 head = tail = new Pos(head); 2999 } else { 3000 head = tail = new Neg(head); 3001 } 3002 break; 3003 case '>': // (?>xxx) independent group 3004 head = createGroup(true); 3005 tail = root; 3006 head.next = expr(tail); 3007 head = tail = new Ques(head, Qtype.INDEPENDENT); 3008 break; 3009 case '<': // (?<xxx) look behind 3010 ch = read(); 3011 if (ch != '=' && ch != '!') { 3012 // named captured group 3013 String name = groupname(ch); 3014 if (namedGroups().containsKey(name)) 3015 throw error("Named capturing group <" + name 3016 + "> is already defined"); 3017 capturingGroup = true; 3018 head = createGroup(false); 3019 tail = root; 3020 namedGroups().put(name, capturingGroupCount-1); 3021 head.next = expr(tail); 3022 break; 3023 } 3024 int start = cursor; 3025 head = createGroup(true); 3026 tail = root; 3027 head.next = expr(tail); 3028 tail.next = LookBehindEndNode.INSTANCE; 3029 TreeInfo info = new TreeInfo(); 3030 head.study(info); 3031 if (info.maxValid == false) { 3032 throw error("Look-behind group does not have " 3033 + "an obvious maximum length"); 3034 } 3035 boolean hasSupplementary = findSupplementary(start, patternLength); 3036 if (ch == '=') { 3037 head = tail = (hasSupplementary ? 3038 new BehindS(head, info.maxLength, 3039 info.minLength) : 3040 new Behind(head, info.maxLength, 3041 info.minLength)); 3042 } else { // if (ch == '!') 3043 head = tail = (hasSupplementary ? 3044 new NotBehindS(head, info.maxLength, 3045 info.minLength) : 3046 new NotBehind(head, info.maxLength, 3047 info.minLength)); 3048 } 3049 // clear all top-closure-nodes inside lookbehind 3050 if (saveTCNCount < topClosureNodes.size()) 3051 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3052 break; 3053 case '$': 3054 case '@': 3055 throw error("Unknown group type"); 3056 default: // (?xxx:) inlined match flags 3057 unread(); 3058 addFlag(); 3059 ch = read(); 3060 if (ch == ')') { 3061 return null; // Inline modifier only 3062 } 3063 if (ch != ':') { 3064 throw error("Unknown inline modifier"); 3065 } 3066 head = createGroup(true); 3067 tail = root; 3068 head.next = expr(tail); 3069 break; 3070 } 3071 } else { // (xxx) a regular group 3072 capturingGroup = true; 3073 head = createGroup(false); 3074 tail = root; 3075 head.next = expr(tail); 3076 } 3077 3078 accept(')', "Unclosed group"); 3079 flags0 = save; 3080 3081 // Check for quantifiers 3082 Node node = closure(head); 3083 if (node == head) { // No closure 3084 root = tail; 3085 return node; // Dual return 3086 } 3087 if (head == tail) { // Zero length assertion 3088 root = node; 3089 return node; // Dual return 3090 } 3091 3092 // have group closure, clear all inner closure nodes from the 3093 // top list (no backtracking stopper optimization for inner 3094 if (saveTCNCount < topClosureNodes.size()) 3095 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3096 3097 return groupWithClosure(node, head, tail, capturingGroup); 3098 } 3099 3100 /** 3101 * Transforms a Group with quantifiers into some special constructs 3102 * (such as Branch or Loop/GroupCurly), if necessary. 3103 * 3104 * This method is applied either to actual groups or to the Unicode 3105 * linebreak (aka \\R) represented as an anonymous group. 3106 */ 3107 private Node groupWithClosure(Node node, Node head, Node tail, 3108 boolean capturingGroup) 3109 { 3110 if (node instanceof Ques) { 3111 Ques ques = (Ques) node; 3112 if (ques.type == Qtype.POSSESSIVE) { 3113 root = node; 3114 return node; 3115 } 3116 BranchConn branchConn = new BranchConn(); 3117 tail = tail.next = branchConn; 3118 if (ques.type == Qtype.GREEDY) { 3119 head = new Branch(head, null, branchConn); 3120 } else { // Reluctant quantifier 3121 head = new Branch(null, head, branchConn); 3122 } 3123 root = tail; 3124 return head; 3125 } else if (node instanceof Curly) { 3126 Curly curly = (Curly) node; 3127 if (curly.type == Qtype.POSSESSIVE) { 3128 root = node; 3129 return node; 3130 } 3131 // Discover if the group is deterministic 3132 TreeInfo info = new TreeInfo(); 3133 if (head.study(info)) { // Deterministic 3134 GroupTail temp = (GroupTail) tail; 3135 head = root = new GroupCurly(head.next, curly.cmin, 3136 curly.cmax, curly.type, 3137 ((GroupTail)tail).localIndex, 3138 ((GroupTail)tail).groupIndex, 3139 capturingGroup); 3140 return head; 3141 } else { // Non-deterministic 3142 int temp = ((GroupHead) head).localIndex; 3143 Loop loop; 3144 if (curly.type == Qtype.GREEDY) { 3145 loop = new Loop(this.localCount, temp); 3146 // add the max_reps greedy to the top-closure-node list 3147 if (curly.cmax == MAX_REPS) 3148 topClosureNodes.add(loop); 3149 } else { // Reluctant Curly 3150 loop = new LazyLoop(this.localCount, temp); 3151 } 3152 Prolog prolog = new Prolog(loop); 3153 this.localCount += 1; 3154 loop.cmin = curly.cmin; 3155 loop.cmax = curly.cmax; 3156 loop.body = head; 3157 tail.next = loop; 3158 root = loop; 3159 return prolog; // Dual return 3160 } 3161 } 3162 throw error("Internal logic error"); 3163 } 3164 3165 /** 3166 * Create group head and tail nodes using double return. If the group is 3167 * created with anonymous true then it is a pure group and should not 3168 * affect group counting. 3169 */ 3170 private Node createGroup(boolean anonymous) { 3171 int localIndex = localCount++; 3172 int groupIndex = 0; 3173 if (!anonymous) 3174 groupIndex = capturingGroupCount++; 3175 GroupHead head = new GroupHead(localIndex); 3176 root = new GroupTail(localIndex, groupIndex); 3177 3178 // for debug/print only, head.match does NOT need the "tail" info 3179 head.tail = (GroupTail)root; 3180 3181 if (!anonymous && groupIndex < 10) 3182 groupNodes[groupIndex] = head; 3183 return head; 3184 } 3185 3186 @SuppressWarnings("fallthrough") 3187 /** 3188 * Parses inlined match flags and set them appropriately. 3189 */ 3190 private void addFlag() { 3191 int ch = peek(); 3192 for (;;) { 3193 switch (ch) { 3194 case 'i': 3195 flags0 |= CASE_INSENSITIVE; 3196 break; 3197 case 'm': 3198 flags0 |= MULTILINE; 3199 break; 3200 case 's': 3201 flags0 |= DOTALL; 3202 break; 3203 case 'd': 3204 flags0 |= UNIX_LINES; 3205 break; 3206 case 'u': 3207 flags0 |= UNICODE_CASE; 3208 break; 3209 case 'c': 3210 flags0 |= CANON_EQ; 3211 break; 3212 case 'x': 3213 flags0 |= COMMENTS; 3214 break; 3215 case 'U': 3216 flags0 |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3217 break; 3218 case '-': // subFlag then fall through 3219 ch = next(); 3220 subFlag(); 3221 default: 3222 return; 3223 } 3224 ch = next(); 3225 } 3226 } 3227 3228 @SuppressWarnings("fallthrough") 3229 /** 3230 * Parses the second part of inlined match flags and turns off 3231 * flags appropriately. 3232 */ 3233 private void subFlag() { 3234 int ch = peek(); 3235 for (;;) { 3236 switch (ch) { 3237 case 'i': 3238 flags0 &= ~CASE_INSENSITIVE; 3239 break; 3240 case 'm': 3241 flags0 &= ~MULTILINE; 3242 break; 3243 case 's': 3244 flags0 &= ~DOTALL; 3245 break; 3246 case 'd': 3247 flags0 &= ~UNIX_LINES; 3248 break; 3249 case 'u': 3250 flags0 &= ~UNICODE_CASE; 3251 break; 3252 case 'c': 3253 flags0 &= ~CANON_EQ; 3254 break; 3255 case 'x': 3256 flags0 &= ~COMMENTS; 3257 break; 3258 case 'U': 3259 flags0 &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3260 break; 3261 default: 3262 return; 3263 } 3264 ch = next(); 3265 } 3266 } 3267 3268 static final int MAX_REPS = 0x7FFFFFFF; 3269 3270 static enum Qtype { 3271 GREEDY, LAZY, POSSESSIVE, INDEPENDENT 3272 } 3273 3274 private Qtype qtype() { 3275 int ch = next(); 3276 if (ch == '?') { 3277 next(); 3278 return Qtype.LAZY; 3279 } else if (ch == '+') { 3280 next(); 3281 return Qtype.POSSESSIVE; 3282 } 3283 return Qtype.GREEDY; 3284 } 3285 3286 private Node curly(Node prev, int cmin) { 3287 Qtype qtype = qtype(); 3288 if (qtype == Qtype.GREEDY) { 3289 if (prev instanceof BmpCharProperty) { 3290 return new BmpCharPropertyGreedy((BmpCharProperty)prev, cmin); 3291 } else if (prev instanceof CharProperty) { 3292 return new CharPropertyGreedy((CharProperty)prev, cmin); 3293 } 3294 } 3295 return new Curly(prev, cmin, MAX_REPS, qtype); 3296 } 3297 3298 /** 3299 * Processing repetition of a Unicode linebreak \\R. 3300 */ 3301 private Node closureOfLineEnding(LineEnding le) { 3302 int ch = peek(); 3303 if (ch != '?' && ch != '*' && ch != '+' && ch != '{') { 3304 return le; 3305 } 3306 3307 // Replace the LineEnding with an anonymous group 3308 // (?:\\u000D\\u000A|[\\u000A\\u000B\\u000C\\u000D\\u0085\\u2028\\u2029]) 3309 Node grHead = createGroup(true); 3310 Node grTail = root; 3311 BranchConn branchConn = new BranchConn(); 3312 branchConn.next = grTail; 3313 Node slice = new Slice(new int[] {0x0D, 0x0A}); 3314 slice.next = branchConn; 3315 Node chClass = newCharProperty(x -> x == 0x0A || x == 0x0B || 3316 x == 0x0C || x == 0x0D || x == 0x85 || x == 0x2028 || 3317 x == 0x2029); 3318 chClass.next = branchConn; 3319 grHead.next = new Branch(slice, chClass, branchConn); 3320 return groupWithClosure(closure(grHead), grHead, grTail, false); 3321 } 3322 3323 /** 3324 * Processes repetition. If the next character peeked is a quantifier 3325 * then new nodes must be appended to handle the repetition. 3326 * Prev could be a single or a group, so it could be a chain of nodes. 3327 */ 3328 private Node closure(Node prev) { 3329 int ch = peek(); 3330 switch (ch) { 3331 case '?': 3332 return new Ques(prev, qtype()); 3333 case '*': 3334 return curly(prev, 0); 3335 case '+': 3336 return curly(prev, 1); 3337 case '{': 3338 ch = skip(); 3339 if (ASCII.isDigit(ch)) { 3340 int cmin = 0, cmax; 3341 try { 3342 do { 3343 cmin = Math.addExact(Math.multiplyExact(cmin, 10), 3344 ch - '0'); 3345 } while (ASCII.isDigit(ch = read())); 3346 if (ch == ',') { 3347 ch = read(); 3348 if (ch == '}') { 3349 unread(); 3350 return curly(prev, cmin); 3351 } else { 3352 cmax = 0; 3353 while (ASCII.isDigit(ch)) { 3354 cmax = Math.addExact(Math.multiplyExact(cmax, 10), 3355 ch - '0'); 3356 ch = read(); 3357 } 3358 } 3359 } else { 3360 cmax = cmin; 3361 } 3362 } catch (ArithmeticException ae) { 3363 throw error("Illegal repetition range"); 3364 } 3365 if (ch != '}') 3366 throw error("Unclosed counted closure"); 3367 if (cmax < cmin) 3368 throw error("Illegal repetition range"); 3369 unread(); 3370 return (cmin == 0 && cmax == 1) 3371 ? new Ques(prev, qtype()) 3372 : new Curly(prev, cmin, cmax, qtype()); 3373 } else { 3374 throw error("Illegal repetition"); 3375 } 3376 default: 3377 return prev; 3378 } 3379 } 3380 3381 /** 3382 * Utility method for parsing control escape sequences. 3383 */ 3384 private int c() { 3385 if (cursor < patternLength) { 3386 return read() ^ 64; 3387 } 3388 throw error("Illegal control escape sequence"); 3389 } 3390 3391 /** 3392 * Utility method for parsing octal escape sequences. 3393 */ 3394 private int o() { 3395 int n = read(); 3396 if (((n-'0')|('7'-n)) >= 0) { 3397 int m = read(); 3398 if (((m-'0')|('7'-m)) >= 0) { 3399 int o = read(); 3400 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) { 3401 return (n - '0') * 64 + (m - '0') * 8 + (o - '0'); 3402 } 3403 unread(); 3404 return (n - '0') * 8 + (m - '0'); 3405 } 3406 unread(); 3407 return (n - '0'); 3408 } 3409 throw error("Illegal octal escape sequence"); 3410 } 3411 3412 /** 3413 * Utility method for parsing hexadecimal escape sequences. 3414 */ 3415 private int x() { 3416 int n = read(); 3417 if (ASCII.isHexDigit(n)) { 3418 int m = read(); 3419 if (ASCII.isHexDigit(m)) { 3420 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m); 3421 } 3422 } else if (n == '{' && ASCII.isHexDigit(peek())) { 3423 int ch = 0; 3424 while (ASCII.isHexDigit(n = read())) { 3425 ch = (ch << 4) + ASCII.toDigit(n); 3426 if (ch > Character.MAX_CODE_POINT) 3427 throw error("Hexadecimal codepoint is too big"); 3428 } 3429 if (n != '}') 3430 throw error("Unclosed hexadecimal escape sequence"); 3431 return ch; 3432 } 3433 throw error("Illegal hexadecimal escape sequence"); 3434 } 3435 3436 /** 3437 * Utility method for parsing unicode escape sequences. 3438 */ 3439 private int cursor() { 3440 return cursor; 3441 } 3442 3443 private void setcursor(int pos) { 3444 cursor = pos; 3445 } 3446 3447 private int uxxxx() { 3448 int n = 0; 3449 for (int i = 0; i < 4; i++) { 3450 int ch = read(); 3451 if (!ASCII.isHexDigit(ch)) { 3452 throw error("Illegal Unicode escape sequence"); 3453 } 3454 n = n * 16 + ASCII.toDigit(ch); 3455 } 3456 return n; 3457 } 3458 3459 private int u() { 3460 int n = uxxxx(); 3461 if (Character.isHighSurrogate((char)n)) { 3462 int cur = cursor(); 3463 if (read() == '\\' && read() == 'u') { 3464 int n2 = uxxxx(); 3465 if (Character.isLowSurrogate((char)n2)) 3466 return Character.toCodePoint((char)n, (char)n2); 3467 } 3468 setcursor(cur); 3469 } 3470 return n; 3471 } 3472 3473 private int N() { 3474 if (read() == '{') { 3475 int i = cursor; 3476 while (read() != '}') { 3477 if (cursor >= patternLength) 3478 throw error("Unclosed character name escape sequence"); 3479 } 3480 String name = new String(temp, i, cursor - i - 1); 3481 try { 3482 return Character.codePointOf(name); 3483 } catch (IllegalArgumentException x) { 3484 throw error("Unknown character name [" + name + "]"); 3485 } 3486 } 3487 throw error("Illegal character name escape sequence"); 3488 } 3489 3490 // 3491 // Utility methods for code point support 3492 // 3493 private static final int countChars(CharSequence seq, int index, 3494 int lengthInCodePoints) { 3495 // optimization 3496 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) { 3497 assert (index >= 0 && index < seq.length()); 3498 return 1; 3499 } 3500 int length = seq.length(); 3501 int x = index; 3502 if (lengthInCodePoints >= 0) { 3503 assert (index >= 0 && index < length); 3504 for (int i = 0; x < length && i < lengthInCodePoints; i++) { 3505 if (Character.isHighSurrogate(seq.charAt(x++))) { 3506 if (x < length && Character.isLowSurrogate(seq.charAt(x))) { 3507 x++; 3508 } 3509 } 3510 } 3511 return x - index; 3512 } 3513 3514 assert (index >= 0 && index <= length); 3515 if (index == 0) { 3516 return 0; 3517 } 3518 int len = -lengthInCodePoints; 3519 for (int i = 0; x > 0 && i < len; i++) { 3520 if (Character.isLowSurrogate(seq.charAt(--x))) { 3521 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) { 3522 x--; 3523 } 3524 } 3525 } 3526 return index - x; 3527 } 3528 3529 private static final int countCodePoints(CharSequence seq) { 3530 int length = seq.length(); 3531 int n = 0; 3532 for (int i = 0; i < length; ) { 3533 n++; 3534 if (Character.isHighSurrogate(seq.charAt(i++))) { 3535 if (i < length && Character.isLowSurrogate(seq.charAt(i))) { 3536 i++; 3537 } 3538 } 3539 } 3540 return n; 3541 } 3542 3543 /** 3544 * Creates a bit vector for matching Latin-1 values. A normal BitClass 3545 * never matches values above Latin-1, and a complemented BitClass always 3546 * matches values above Latin-1. 3547 */ 3548 static final class BitClass implements BmpCharPredicate { 3549 final boolean[] bits; 3550 BitClass() { 3551 bits = new boolean[256]; 3552 } 3553 BitClass add(int c, int flags) { 3554 assert c >= 0 && c <= 255; 3555 if ((flags & CASE_INSENSITIVE) != 0) { 3556 if (ASCII.isAscii(c)) { 3557 bits[ASCII.toUpper(c)] = true; 3558 bits[ASCII.toLower(c)] = true; 3559 } else if ((flags & UNICODE_CASE) != 0) { 3560 bits[Character.toLowerCase(c)] = true; 3561 bits[Character.toUpperCase(c)] = true; 3562 } 3563 } 3564 bits[c] = true; 3565 return this; 3566 } 3567 public boolean is(int ch) { 3568 return ch < 256 && bits[ch]; 3569 } 3570 } 3571 3572 3573 /** 3574 * Utility method for creating a string slice matcher. 3575 */ 3576 private Node newSlice(int[] buf, int count, boolean hasSupplementary) { 3577 int[] tmp = new int[count]; 3578 if (has(CASE_INSENSITIVE)) { 3579 if (has(UNICODE_CASE)) { 3580 for (int i = 0; i < count; i++) { 3581 tmp[i] = Character.toLowerCase( 3582 Character.toUpperCase(buf[i])); 3583 } 3584 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp); 3585 } 3586 for (int i = 0; i < count; i++) { 3587 tmp[i] = ASCII.toLower(buf[i]); 3588 } 3589 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp); 3590 } 3591 for (int i = 0; i < count; i++) { 3592 tmp[i] = buf[i]; 3593 } 3594 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp); 3595 } 3596 3597 /** 3598 * The following classes are the building components of the object 3599 * tree that represents a compiled regular expression. The object tree 3600 * is made of individual elements that handle constructs in the Pattern. 3601 * Each type of object knows how to match its equivalent construct with 3602 * the match() method. 3603 */ 3604 3605 /** 3606 * Base class for all node classes. Subclasses should override the match() 3607 * method as appropriate. This class is an accepting node, so its match() 3608 * always returns true. 3609 */ 3610 static class Node extends Object { 3611 Node next; 3612 Node() { 3613 next = Pattern.accept; 3614 } 3615 /** 3616 * This method implements the classic accept node. 3617 */ 3618 boolean match(Matcher matcher, int i, CharSequence seq) { 3619 matcher.last = i; 3620 matcher.groups[0] = matcher.first; 3621 matcher.groups[1] = matcher.last; 3622 return true; 3623 } 3624 /** 3625 * This method is good for all zero length assertions. 3626 */ 3627 boolean study(TreeInfo info) { 3628 if (next != null) { 3629 return next.study(info); 3630 } else { 3631 return info.deterministic; 3632 } 3633 } 3634 } 3635 3636 static class LastNode extends Node { 3637 /** 3638 * This method implements the classic accept node with 3639 * the addition of a check to see if the match occurred 3640 * using all of the input. 3641 */ 3642 boolean match(Matcher matcher, int i, CharSequence seq) { 3643 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to) 3644 return false; 3645 matcher.last = i; 3646 matcher.groups[0] = matcher.first; 3647 matcher.groups[1] = matcher.last; 3648 return true; 3649 } 3650 } 3651 3652 /** 3653 * Used for REs that can start anywhere within the input string. 3654 * This basically tries to match repeatedly at each spot in the 3655 * input string, moving forward after each try. An anchored search 3656 * or a BnM will bypass this node completely. 3657 */ 3658 static class Start extends Node { 3659 int minLength; 3660 Start(Node node) { 3661 this.next = node; 3662 TreeInfo info = new TreeInfo(); 3663 next.study(info); 3664 minLength = info.minLength; 3665 } 3666 boolean match(Matcher matcher, int i, CharSequence seq) { 3667 if (i > matcher.to - minLength) { 3668 matcher.hitEnd = true; 3669 return false; 3670 } 3671 int guard = matcher.to - minLength; 3672 for (; i <= guard; i++) { 3673 if (next.match(matcher, i, seq)) { 3674 matcher.first = i; 3675 matcher.groups[0] = matcher.first; 3676 matcher.groups[1] = matcher.last; 3677 return true; 3678 } 3679 } 3680 matcher.hitEnd = true; 3681 return false; 3682 } 3683 boolean study(TreeInfo info) { 3684 next.study(info); 3685 info.maxValid = false; 3686 info.deterministic = false; 3687 return false; 3688 } 3689 } 3690 3691 /* 3692 * StartS supports supplementary characters, including unpaired surrogates. 3693 */ 3694 static final class StartS extends Start { 3695 StartS(Node node) { 3696 super(node); 3697 } 3698 boolean match(Matcher matcher, int i, CharSequence seq) { 3699 if (i > matcher.to - minLength) { 3700 matcher.hitEnd = true; 3701 return false; 3702 } 3703 int guard = matcher.to - minLength; 3704 while (i <= guard) { 3705 //if ((ret = next.match(matcher, i, seq)) || i == guard) 3706 if (next.match(matcher, i, seq)) { 3707 matcher.first = i; 3708 matcher.groups[0] = matcher.first; 3709 matcher.groups[1] = matcher.last; 3710 return true; 3711 } 3712 if (i == guard) 3713 break; 3714 // Optimization to move to the next character. This is 3715 // faster than countChars(seq, i, 1). 3716 if (Character.isHighSurrogate(seq.charAt(i++))) { 3717 if (i < seq.length() && 3718 Character.isLowSurrogate(seq.charAt(i))) { 3719 i++; 3720 } 3721 } 3722 } 3723 matcher.hitEnd = true; 3724 return false; 3725 } 3726 } 3727 3728 /** 3729 * Node to anchor at the beginning of input. This object implements the 3730 * match for a \A sequence, and the caret anchor will use this if not in 3731 * multiline mode. 3732 */ 3733 static final class Begin extends Node { 3734 boolean match(Matcher matcher, int i, CharSequence seq) { 3735 int fromIndex = (matcher.anchoringBounds) ? 3736 matcher.from : 0; 3737 if (i == fromIndex && next.match(matcher, i, seq)) { 3738 matcher.first = i; 3739 matcher.groups[0] = i; 3740 matcher.groups[1] = matcher.last; 3741 return true; 3742 } else { 3743 return false; 3744 } 3745 } 3746 } 3747 3748 /** 3749 * Node to anchor at the end of input. This is the absolute end, so this 3750 * should not match at the last newline before the end as $ will. 3751 */ 3752 static final class End extends Node { 3753 boolean match(Matcher matcher, int i, CharSequence seq) { 3754 int endIndex = (matcher.anchoringBounds) ? 3755 matcher.to : matcher.getTextLength(); 3756 if (i == endIndex) { 3757 matcher.hitEnd = true; 3758 return next.match(matcher, i, seq); 3759 } 3760 return false; 3761 } 3762 } 3763 3764 /** 3765 * Node to anchor at the beginning of a line. This is essentially the 3766 * object to match for the multiline ^. 3767 */ 3768 static final class Caret extends Node { 3769 boolean match(Matcher matcher, int i, CharSequence seq) { 3770 int startIndex = matcher.from; 3771 int endIndex = matcher.to; 3772 if (!matcher.anchoringBounds) { 3773 startIndex = 0; 3774 endIndex = matcher.getTextLength(); 3775 } 3776 // Perl does not match ^ at end of input even after newline 3777 if (i == endIndex) { 3778 matcher.hitEnd = true; 3779 return false; 3780 } 3781 if (i > startIndex) { 3782 char ch = seq.charAt(i-1); 3783 if (ch != '\n' && ch != '\r' 3784 && (ch|1) != '\u2029' 3785 && ch != '\u0085' ) { 3786 return false; 3787 } 3788 // Should treat /r/n as one newline 3789 if (ch == '\r' && seq.charAt(i) == '\n') 3790 return false; 3791 } 3792 return next.match(matcher, i, seq); 3793 } 3794 } 3795 3796 /** 3797 * Node to anchor at the beginning of a line when in unixdot mode. 3798 */ 3799 static final class UnixCaret extends Node { 3800 boolean match(Matcher matcher, int i, CharSequence seq) { 3801 int startIndex = matcher.from; 3802 int endIndex = matcher.to; 3803 if (!matcher.anchoringBounds) { 3804 startIndex = 0; 3805 endIndex = matcher.getTextLength(); 3806 } 3807 // Perl does not match ^ at end of input even after newline 3808 if (i == endIndex) { 3809 matcher.hitEnd = true; 3810 return false; 3811 } 3812 if (i > startIndex) { 3813 char ch = seq.charAt(i-1); 3814 if (ch != '\n') { 3815 return false; 3816 } 3817 } 3818 return next.match(matcher, i, seq); 3819 } 3820 } 3821 3822 /** 3823 * Node to match the location where the last match ended. 3824 * This is used for the \G construct. 3825 */ 3826 static final class LastMatch extends Node { 3827 boolean match(Matcher matcher, int i, CharSequence seq) { 3828 if (i != matcher.oldLast) 3829 return false; 3830 return next.match(matcher, i, seq); 3831 } 3832 } 3833 3834 /** 3835 * Node to anchor at the end of a line or the end of input based on the 3836 * multiline mode. 3837 * 3838 * When not in multiline mode, the $ can only match at the very end 3839 * of the input, unless the input ends in a line terminator in which 3840 * it matches right before the last line terminator. 3841 * 3842 * Note that \r\n is considered an atomic line terminator. 3843 * 3844 * Like ^ the $ operator matches at a position, it does not match the 3845 * line terminators themselves. 3846 */ 3847 static final class Dollar extends Node { 3848 boolean multiline; 3849 Dollar(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 (!multiline) { 3856 if (i < endIndex - 2) 3857 return false; 3858 if (i == endIndex - 2) { 3859 char ch = seq.charAt(i); 3860 if (ch != '\r') 3861 return false; 3862 ch = seq.charAt(i + 1); 3863 if (ch != '\n') 3864 return false; 3865 } 3866 } 3867 // Matches before any line terminator; also matches at the 3868 // end of input 3869 // Before line terminator: 3870 // If multiline, we match here no matter what 3871 // If not multiline, fall through so that the end 3872 // is marked as hit; this must be a /r/n or a /n 3873 // at the very end so the end was hit; more input 3874 // could make this not match here 3875 if (i < endIndex) { 3876 char ch = seq.charAt(i); 3877 if (ch == '\n') { 3878 // No match between \r\n 3879 if (i > 0 && seq.charAt(i-1) == '\r') 3880 return false; 3881 if (multiline) 3882 return next.match(matcher, i, seq); 3883 } else if (ch == '\r' || ch == '\u0085' || 3884 (ch|1) == '\u2029') { 3885 if (multiline) 3886 return next.match(matcher, i, seq); 3887 } else { // No line terminator, no match 3888 return false; 3889 } 3890 } 3891 // Matched at current end so hit end 3892 matcher.hitEnd = true; 3893 // If a $ matches because of end of input, then more input 3894 // could cause it to fail! 3895 matcher.requireEnd = true; 3896 return next.match(matcher, i, seq); 3897 } 3898 boolean study(TreeInfo info) { 3899 next.study(info); 3900 return info.deterministic; 3901 } 3902 } 3903 3904 /** 3905 * Node to anchor at the end of a line or the end of input based on the 3906 * multiline mode when in unix lines mode. 3907 */ 3908 static final class UnixDollar extends Node { 3909 boolean multiline; 3910 UnixDollar(boolean mul) { 3911 multiline = mul; 3912 } 3913 boolean match(Matcher matcher, int i, CharSequence seq) { 3914 int endIndex = (matcher.anchoringBounds) ? 3915 matcher.to : matcher.getTextLength(); 3916 if (i < endIndex) { 3917 char ch = seq.charAt(i); 3918 if (ch == '\n') { 3919 // If not multiline, then only possible to 3920 // match at very end or one before end 3921 if (multiline == false && i != endIndex - 1) 3922 return false; 3923 // If multiline return next.match without setting 3924 // matcher.hitEnd 3925 if (multiline) 3926 return next.match(matcher, i, seq); 3927 } else { 3928 return false; 3929 } 3930 } 3931 // Matching because at the end or 1 before the end; 3932 // more input could change this so set hitEnd 3933 matcher.hitEnd = true; 3934 // If a $ matches because of end of input, then more input 3935 // could cause it to fail! 3936 matcher.requireEnd = true; 3937 return next.match(matcher, i, seq); 3938 } 3939 boolean study(TreeInfo info) { 3940 next.study(info); 3941 return info.deterministic; 3942 } 3943 } 3944 3945 /** 3946 * Node class that matches a Unicode line ending '\R' 3947 */ 3948 static final class LineEnding extends Node { 3949 boolean match(Matcher matcher, int i, CharSequence seq) { 3950 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029]) 3951 if (i < matcher.to) { 3952 int ch = seq.charAt(i); 3953 if (ch == 0x0A || ch == 0x0B || ch == 0x0C || 3954 ch == 0x85 || ch == 0x2028 || ch == 0x2029) 3955 return next.match(matcher, i + 1, seq); 3956 if (ch == 0x0D) { 3957 i++; 3958 if (i < matcher.to) { 3959 if (seq.charAt(i) == 0x0A && 3960 next.match(matcher, i + 1, seq)) { 3961 return true; 3962 } 3963 } else { 3964 matcher.hitEnd = true; 3965 } 3966 return next.match(matcher, i, seq); 3967 } 3968 } else { 3969 matcher.hitEnd = true; 3970 } 3971 return false; 3972 } 3973 boolean study(TreeInfo info) { 3974 info.minLength++; 3975 info.maxLength += 2; 3976 return next.study(info); 3977 } 3978 } 3979 3980 /** 3981 * Abstract node class to match one character satisfying some 3982 * boolean property. 3983 */ 3984 static class CharProperty extends Node { 3985 final CharPredicate predicate; 3986 3987 CharProperty (CharPredicate predicate) { 3988 this.predicate = predicate; 3989 } 3990 boolean match(Matcher matcher, int i, CharSequence seq) { 3991 if (i < matcher.to) { 3992 int ch = Character.codePointAt(seq, i); 3993 i += Character.charCount(ch); 3994 if (i <= matcher.to) { 3995 return predicate.is(ch) && 3996 next.match(matcher, i, seq); 3997 } 3998 } 3999 matcher.hitEnd = true; 4000 return false; 4001 } 4002 boolean study(TreeInfo info) { 4003 info.minLength++; 4004 info.maxLength++; 4005 return next.study(info); 4006 } 4007 } 4008 4009 /** 4010 * Optimized version of CharProperty that works only for 4011 * properties never satisfied by Supplementary characters. 4012 */ 4013 private static class BmpCharProperty extends CharProperty { 4014 BmpCharProperty (BmpCharPredicate predicate) { 4015 super(predicate); 4016 } 4017 boolean match(Matcher matcher, int i, CharSequence seq) { 4018 if (i < matcher.to) { 4019 return predicate.is(seq.charAt(i)) && 4020 next.match(matcher, i + 1, seq); 4021 } else { 4022 matcher.hitEnd = true; 4023 return false; 4024 } 4025 } 4026 } 4027 4028 private static class NFCCharProperty extends Node { 4029 CharPredicate predicate; 4030 NFCCharProperty (CharPredicate predicate) { 4031 this.predicate = predicate; 4032 } 4033 4034 boolean match(Matcher matcher, int i, CharSequence seq) { 4035 if (i < matcher.to) { 4036 int ch0 = Character.codePointAt(seq, i); 4037 int n = Character.charCount(ch0); 4038 int j = Grapheme.nextBoundary(seq, i, matcher.to); 4039 if (i + n == j) { // single cp grapheme, assume nfc 4040 if (predicate.is(ch0)) 4041 return next.match(matcher, j, seq); 4042 } else { 4043 while (i + n < j) { 4044 String nfc = Normalizer.normalize( 4045 seq.toString().substring(i, j), Normalizer.Form.NFC); 4046 if (nfc.codePointCount(0, nfc.length()) == 1) { 4047 if (predicate.is(nfc.codePointAt(0)) && 4048 next.match(matcher, j, seq)) { 4049 return true; 4050 } 4051 } 4052 4053 ch0 = Character.codePointBefore(seq, j); 4054 j -= Character.charCount(ch0); 4055 } 4056 } 4057 if (j < matcher.to) 4058 return false; 4059 } 4060 matcher.hitEnd = true; 4061 return false; 4062 } 4063 4064 boolean study(TreeInfo info) { 4065 info.minLength++; 4066 info.deterministic = false; 4067 return next.study(info); 4068 } 4069 } 4070 4071 /** 4072 * Node class that matches an unicode extended grapheme cluster 4073 */ 4074 static class XGrapheme extends Node { 4075 boolean match(Matcher matcher, int i, CharSequence seq) { 4076 if (i < matcher.to) { 4077 i = Grapheme.nextBoundary(seq, i, matcher.to); 4078 return next.match(matcher, i, seq); 4079 } 4080 matcher.hitEnd = true; 4081 return false; 4082 } 4083 4084 boolean study(TreeInfo info) { 4085 info.minLength++; 4086 info.deterministic = false; 4087 return next.study(info); 4088 } 4089 } 4090 4091 /** 4092 * Node class that handles grapheme boundaries 4093 */ 4094 static class GraphemeBound extends Node { 4095 boolean match(Matcher matcher, int i, CharSequence seq) { 4096 int startIndex = matcher.from; 4097 int endIndex = matcher.to; 4098 if (matcher.transparentBounds) { 4099 startIndex = 0; 4100 endIndex = matcher.getTextLength(); 4101 } 4102 if (i == startIndex) { 4103 // continue with return below 4104 } else if (i < endIndex) { 4105 if (Character.isSurrogatePair(seq.charAt(i - 1), seq.charAt(i))) { 4106 return false; 4107 } 4108 if (Grapheme.nextBoundary(seq, matcher.last, endIndex) > i) { 4109 return false; 4110 } 4111 } else { 4112 matcher.hitEnd = true; 4113 matcher.requireEnd = true; 4114 } 4115 return next.match(matcher, i, seq); 4116 } 4117 } 4118 4119 /** 4120 * Base class for all Slice nodes 4121 */ 4122 static class SliceNode extends Node { 4123 int[] buffer; 4124 SliceNode(int[] buf) { 4125 buffer = buf; 4126 } 4127 boolean study(TreeInfo info) { 4128 info.minLength += buffer.length; 4129 info.maxLength += buffer.length; 4130 return next.study(info); 4131 } 4132 } 4133 4134 /** 4135 * Node class for a case sensitive/BMP-only sequence of literal 4136 * characters. 4137 */ 4138 static class Slice extends SliceNode { 4139 Slice(int[] buf) { 4140 super(buf); 4141 } 4142 boolean match(Matcher matcher, int i, CharSequence seq) { 4143 int[] buf = buffer; 4144 int len = buf.length; 4145 for (int j=0; j<len; j++) { 4146 if ((i+j) >= matcher.to) { 4147 matcher.hitEnd = true; 4148 return false; 4149 } 4150 if (buf[j] != seq.charAt(i+j)) 4151 return false; 4152 } 4153 return next.match(matcher, i+len, seq); 4154 } 4155 } 4156 4157 /** 4158 * Node class for a case_insensitive/BMP-only sequence of literal 4159 * characters. 4160 */ 4161 static class SliceI extends SliceNode { 4162 SliceI(int[] buf) { 4163 super(buf); 4164 } 4165 boolean match(Matcher matcher, int i, CharSequence seq) { 4166 int[] buf = buffer; 4167 int len = buf.length; 4168 for (int j=0; j<len; j++) { 4169 if ((i+j) >= matcher.to) { 4170 matcher.hitEnd = true; 4171 return false; 4172 } 4173 int c = seq.charAt(i+j); 4174 if (buf[j] != c && 4175 buf[j] != ASCII.toLower(c)) 4176 return false; 4177 } 4178 return next.match(matcher, i+len, seq); 4179 } 4180 } 4181 4182 /** 4183 * Node class for a unicode_case_insensitive/BMP-only sequence of 4184 * literal characters. Uses unicode case folding. 4185 */ 4186 static final class SliceU extends SliceNode { 4187 SliceU(int[] buf) { 4188 super(buf); 4189 } 4190 boolean match(Matcher matcher, int i, CharSequence seq) { 4191 int[] buf = buffer; 4192 int len = buf.length; 4193 for (int j=0; j<len; j++) { 4194 if ((i+j) >= matcher.to) { 4195 matcher.hitEnd = true; 4196 return false; 4197 } 4198 int c = seq.charAt(i+j); 4199 if (buf[j] != c && 4200 buf[j] != Character.toLowerCase(Character.toUpperCase(c))) 4201 return false; 4202 } 4203 return next.match(matcher, i+len, seq); 4204 } 4205 } 4206 4207 /** 4208 * Node class for a case sensitive sequence of literal characters 4209 * including supplementary characters. 4210 */ 4211 static final class SliceS extends Slice { 4212 SliceS(int[] buf) { 4213 super(buf); 4214 } 4215 boolean match(Matcher matcher, int i, CharSequence seq) { 4216 int[] buf = buffer; 4217 int x = i; 4218 for (int j = 0; j < buf.length; j++) { 4219 if (x >= matcher.to) { 4220 matcher.hitEnd = true; 4221 return false; 4222 } 4223 int c = Character.codePointAt(seq, x); 4224 if (buf[j] != c) 4225 return false; 4226 x += Character.charCount(c); 4227 if (x > matcher.to) { 4228 matcher.hitEnd = true; 4229 return false; 4230 } 4231 } 4232 return next.match(matcher, x, seq); 4233 } 4234 } 4235 4236 /** 4237 * Node class for a case insensitive sequence of literal characters 4238 * including supplementary characters. 4239 */ 4240 static class SliceIS extends SliceNode { 4241 SliceIS(int[] buf) { 4242 super(buf); 4243 } 4244 int toLower(int c) { 4245 return ASCII.toLower(c); 4246 } 4247 boolean match(Matcher matcher, int i, CharSequence seq) { 4248 int[] buf = buffer; 4249 int x = i; 4250 for (int j = 0; j < buf.length; j++) { 4251 if (x >= matcher.to) { 4252 matcher.hitEnd = true; 4253 return false; 4254 } 4255 int c = Character.codePointAt(seq, x); 4256 if (buf[j] != c && buf[j] != toLower(c)) 4257 return false; 4258 x += Character.charCount(c); 4259 if (x > matcher.to) { 4260 matcher.hitEnd = true; 4261 return false; 4262 } 4263 } 4264 return next.match(matcher, x, seq); 4265 } 4266 } 4267 4268 /** 4269 * Node class for a case insensitive sequence of literal characters. 4270 * Uses unicode case folding. 4271 */ 4272 static final class SliceUS extends SliceIS { 4273 SliceUS(int[] buf) { 4274 super(buf); 4275 } 4276 int toLower(int c) { 4277 return Character.toLowerCase(Character.toUpperCase(c)); 4278 } 4279 } 4280 4281 /** 4282 * The 0 or 1 quantifier. This one class implements all three types. 4283 */ 4284 static final class Ques extends Node { 4285 Node atom; 4286 Qtype type; 4287 Ques(Node node, Qtype type) { 4288 this.atom = node; 4289 this.type = type; 4290 } 4291 boolean match(Matcher matcher, int i, CharSequence seq) { 4292 switch (type) { 4293 case GREEDY: 4294 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)) 4295 || next.match(matcher, i, seq); 4296 case LAZY: 4297 return next.match(matcher, i, seq) 4298 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)); 4299 case POSSESSIVE: 4300 if (atom.match(matcher, i, seq)) i = matcher.last; 4301 return next.match(matcher, i, seq); 4302 default: 4303 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq); 4304 } 4305 } 4306 boolean study(TreeInfo info) { 4307 if (type != Qtype.INDEPENDENT) { 4308 int minL = info.minLength; 4309 atom.study(info); 4310 info.minLength = minL; 4311 info.deterministic = false; 4312 return next.study(info); 4313 } else { 4314 atom.study(info); 4315 return next.study(info); 4316 } 4317 } 4318 } 4319 4320 /** 4321 * Handles the greedy style repetition with the specified minimum 4322 * and the maximum equal to MAX_REPS, for *, + and {N,} quantifiers. 4323 */ 4324 static class CharPropertyGreedy extends Node { 4325 final CharPredicate predicate; 4326 final int cmin; 4327 4328 CharPropertyGreedy(CharProperty cp, int cmin) { 4329 this.predicate = cp.predicate; 4330 this.cmin = cmin; 4331 } 4332 boolean match(Matcher matcher, int i, CharSequence seq) { 4333 int starti = i; 4334 int n = 0; 4335 int to = matcher.to; 4336 // greedy, all the way down 4337 while (i < to) { 4338 int ch = Character.codePointAt(seq, i); 4339 int len = Character.charCount(ch); 4340 if (i + len > to) { 4341 // the region cut off the high half of a surrogate pair 4342 matcher.hitEnd = true; 4343 ch = seq.charAt(i); 4344 len = 1; 4345 } 4346 if (!predicate.is(ch)) 4347 break; 4348 i += len; 4349 n++; 4350 } 4351 if (i >= to) { 4352 matcher.hitEnd = true; 4353 } 4354 while (n >= cmin) { 4355 if (next.match(matcher, i, seq)) 4356 return true; 4357 if (n == cmin) 4358 return false; 4359 // backing off if match fails 4360 int ch = Character.codePointBefore(seq, i); 4361 // check if the region cut off the low half of a surrogate pair 4362 i = Math.max(starti, i - Character.charCount(ch)); 4363 n--; 4364 } 4365 return false; 4366 } 4367 4368 boolean study(TreeInfo info) { 4369 info.minLength += cmin; 4370 if (info.maxValid) { 4371 info.maxLength += MAX_REPS; 4372 } 4373 info.deterministic = false; 4374 return next.study(info); 4375 } 4376 } 4377 4378 static final class BmpCharPropertyGreedy extends CharPropertyGreedy { 4379 4380 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) { 4381 super(bcp, cmin); 4382 } 4383 4384 boolean match(Matcher matcher, int i, CharSequence seq) { 4385 int n = 0; 4386 int to = matcher.to; 4387 while (i < to && predicate.is(seq.charAt(i))) { 4388 i++; n++; 4389 } 4390 if (i >= to) { 4391 matcher.hitEnd = true; 4392 } 4393 while (n >= cmin) { 4394 if (next.match(matcher, i, seq)) 4395 return true; 4396 i--; n--; // backing off if match fails 4397 } 4398 return false; 4399 } 4400 } 4401 4402 /** 4403 * Handles the curly-brace style repetition with a specified minimum and 4404 * maximum occurrences. The * quantifier is handled as a special case. 4405 * This class handles the three types. 4406 */ 4407 static final class Curly extends Node { 4408 Node atom; 4409 Qtype type; 4410 int cmin; 4411 int cmax; 4412 4413 Curly(Node node, int cmin, int cmax, Qtype type) { 4414 this.atom = node; 4415 this.type = type; 4416 this.cmin = cmin; 4417 this.cmax = cmax; 4418 } 4419 boolean match(Matcher matcher, int i, CharSequence seq) { 4420 int j; 4421 for (j = 0; j < cmin; j++) { 4422 if (atom.match(matcher, i, seq)) { 4423 i = matcher.last; 4424 continue; 4425 } 4426 return false; 4427 } 4428 if (type == Qtype.GREEDY) 4429 return match0(matcher, i, j, seq); 4430 else if (type == Qtype.LAZY) 4431 return match1(matcher, i, j, seq); 4432 else 4433 return match2(matcher, i, j, seq); 4434 } 4435 // Greedy match. 4436 // i is the index to start matching at 4437 // j is the number of atoms that have matched 4438 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4439 if (j >= cmax) { 4440 // We have matched the maximum... continue with the rest of 4441 // the regular expression 4442 return next.match(matcher, i, seq); 4443 } 4444 int backLimit = j; 4445 while (atom.match(matcher, i, seq)) { 4446 // k is the length of this match 4447 int k = matcher.last - i; 4448 if (k == 0) // Zero length match 4449 break; 4450 // Move up index and number matched 4451 i = matcher.last; 4452 j++; 4453 // We are greedy so match as many as we can 4454 while (j < cmax) { 4455 if (!atom.match(matcher, i, seq)) 4456 break; 4457 if (i + k != matcher.last) { 4458 if (match0(matcher, matcher.last, j+1, seq)) 4459 return true; 4460 break; 4461 } 4462 i += k; 4463 j++; 4464 } 4465 // Handle backing off if match fails 4466 while (j >= backLimit) { 4467 if (next.match(matcher, i, seq)) 4468 return true; 4469 i -= k; 4470 j--; 4471 } 4472 return false; 4473 } 4474 return next.match(matcher, i, seq); 4475 } 4476 // Reluctant match. At this point, the minimum has been satisfied. 4477 // i is the index to start matching at 4478 // j is the number of atoms that have matched 4479 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4480 for (;;) { 4481 // Try finishing match without consuming any more 4482 if (next.match(matcher, i, seq)) 4483 return true; 4484 // At the maximum, no match found 4485 if (j >= cmax) 4486 return false; 4487 // Okay, must try one more atom 4488 if (!atom.match(matcher, i, seq)) 4489 return false; 4490 // If we haven't moved forward then must break out 4491 if (i == matcher.last) 4492 return false; 4493 // Move up index and number matched 4494 i = matcher.last; 4495 j++; 4496 } 4497 } 4498 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4499 for (; j < cmax; j++) { 4500 if (!atom.match(matcher, i, seq)) 4501 break; 4502 if (i == matcher.last) 4503 break; 4504 i = matcher.last; 4505 } 4506 return next.match(matcher, i, seq); 4507 } 4508 boolean study(TreeInfo info) { 4509 // Save original info 4510 int minL = info.minLength; 4511 int maxL = info.maxLength; 4512 boolean maxV = info.maxValid; 4513 boolean detm = info.deterministic; 4514 info.reset(); 4515 4516 atom.study(info); 4517 4518 int temp = info.minLength * cmin + minL; 4519 if (temp < minL) { 4520 temp = 0xFFFFFFF; // arbitrary large number 4521 } 4522 info.minLength = temp; 4523 4524 if (maxV & info.maxValid) { 4525 temp = info.maxLength * cmax + maxL; 4526 info.maxLength = temp; 4527 if (temp < maxL) { 4528 info.maxValid = false; 4529 } 4530 } else { 4531 info.maxValid = false; 4532 } 4533 4534 if (info.deterministic && cmin == cmax) 4535 info.deterministic = detm; 4536 else 4537 info.deterministic = false; 4538 return next.study(info); 4539 } 4540 } 4541 4542 /** 4543 * Handles the curly-brace style repetition with a specified minimum and 4544 * maximum occurrences in deterministic cases. This is an iterative 4545 * optimization over the Prolog and Loop system which would handle this 4546 * in a recursive way. The * quantifier is handled as a special case. 4547 * If capture is true then this class saves group settings and ensures 4548 * that groups are unset when backing off of a group match. 4549 */ 4550 static final class GroupCurly extends Node { 4551 Node atom; 4552 Qtype type; 4553 int cmin; 4554 int cmax; 4555 int localIndex; 4556 int groupIndex; 4557 boolean capture; 4558 4559 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local, 4560 int group, boolean capture) { 4561 this.atom = node; 4562 this.type = type; 4563 this.cmin = cmin; 4564 this.cmax = cmax; 4565 this.localIndex = local; 4566 this.groupIndex = group; 4567 this.capture = capture; 4568 } 4569 boolean match(Matcher matcher, int i, CharSequence seq) { 4570 int[] groups = matcher.groups; 4571 int[] locals = matcher.locals; 4572 int save0 = locals[localIndex]; 4573 int save1 = 0; 4574 int save2 = 0; 4575 4576 if (capture) { 4577 save1 = groups[groupIndex]; 4578 save2 = groups[groupIndex+1]; 4579 } 4580 4581 // Notify GroupTail there is no need to setup group info 4582 // because it will be set here 4583 locals[localIndex] = -1; 4584 4585 boolean ret = true; 4586 for (int j = 0; j < cmin; j++) { 4587 if (atom.match(matcher, i, seq)) { 4588 if (capture) { 4589 groups[groupIndex] = i; 4590 groups[groupIndex+1] = matcher.last; 4591 } 4592 i = matcher.last; 4593 } else { 4594 ret = false; 4595 break; 4596 } 4597 } 4598 if (ret) { 4599 if (type == Qtype.GREEDY) { 4600 ret = match0(matcher, i, cmin, seq); 4601 } else if (type == Qtype.LAZY) { 4602 ret = match1(matcher, i, cmin, seq); 4603 } else { 4604 ret = match2(matcher, i, cmin, seq); 4605 } 4606 } 4607 if (!ret) { 4608 locals[localIndex] = save0; 4609 if (capture) { 4610 groups[groupIndex] = save1; 4611 groups[groupIndex+1] = save2; 4612 } 4613 } 4614 return ret; 4615 } 4616 // Aggressive group match 4617 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4618 // don't back off passing the starting "j" 4619 int min = j; 4620 int[] groups = matcher.groups; 4621 int save0 = 0; 4622 int save1 = 0; 4623 if (capture) { 4624 save0 = groups[groupIndex]; 4625 save1 = groups[groupIndex+1]; 4626 } 4627 for (;;) { 4628 if (j >= cmax) 4629 break; 4630 if (!atom.match(matcher, i, seq)) 4631 break; 4632 int k = matcher.last - i; 4633 if (k <= 0) { 4634 if (capture) { 4635 groups[groupIndex] = i; 4636 groups[groupIndex+1] = i + k; 4637 } 4638 i = i + k; 4639 break; 4640 } 4641 for (;;) { 4642 if (capture) { 4643 groups[groupIndex] = i; 4644 groups[groupIndex+1] = i + k; 4645 } 4646 i = i + k; 4647 if (++j >= cmax) 4648 break; 4649 if (!atom.match(matcher, i, seq)) 4650 break; 4651 if (i + k != matcher.last) { 4652 if (match0(matcher, i, j, seq)) 4653 return true; 4654 break; 4655 } 4656 } 4657 while (j > min) { 4658 if (next.match(matcher, i, seq)) { 4659 if (capture) { 4660 groups[groupIndex+1] = i; 4661 groups[groupIndex] = i - k; 4662 } 4663 return true; 4664 } 4665 // backing off 4666 i = i - k; 4667 if (capture) { 4668 groups[groupIndex+1] = i; 4669 groups[groupIndex] = i - k; 4670 } 4671 j--; 4672 4673 } 4674 break; 4675 } 4676 if (capture) { 4677 groups[groupIndex] = save0; 4678 groups[groupIndex+1] = save1; 4679 } 4680 return next.match(matcher, i, seq); 4681 } 4682 // Reluctant matching 4683 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4684 for (;;) { 4685 if (next.match(matcher, i, seq)) 4686 return true; 4687 if (j >= cmax) 4688 return false; 4689 if (!atom.match(matcher, i, seq)) 4690 return false; 4691 if (i == matcher.last) 4692 return false; 4693 if (capture) { 4694 matcher.groups[groupIndex] = i; 4695 matcher.groups[groupIndex+1] = matcher.last; 4696 } 4697 i = matcher.last; 4698 j++; 4699 } 4700 } 4701 // Possessive matching 4702 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4703 for (; j < cmax; j++) { 4704 if (!atom.match(matcher, i, seq)) { 4705 break; 4706 } 4707 if (capture) { 4708 matcher.groups[groupIndex] = i; 4709 matcher.groups[groupIndex+1] = matcher.last; 4710 } 4711 if (i == matcher.last) { 4712 break; 4713 } 4714 i = matcher.last; 4715 } 4716 return next.match(matcher, i, seq); 4717 } 4718 boolean study(TreeInfo info) { 4719 // Save original info 4720 int minL = info.minLength; 4721 int maxL = info.maxLength; 4722 boolean maxV = info.maxValid; 4723 boolean detm = info.deterministic; 4724 info.reset(); 4725 4726 atom.study(info); 4727 4728 int temp = info.minLength * cmin + minL; 4729 if (temp < minL) { 4730 temp = 0xFFFFFFF; // Arbitrary large number 4731 } 4732 info.minLength = temp; 4733 4734 if (maxV & info.maxValid) { 4735 temp = info.maxLength * cmax + maxL; 4736 info.maxLength = temp; 4737 if (temp < maxL) { 4738 info.maxValid = false; 4739 } 4740 } else { 4741 info.maxValid = false; 4742 } 4743 4744 if (info.deterministic && cmin == cmax) { 4745 info.deterministic = detm; 4746 } else { 4747 info.deterministic = false; 4748 } 4749 return next.study(info); 4750 } 4751 } 4752 4753 /** 4754 * A Guard node at the end of each atom node in a Branch. It 4755 * serves the purpose of chaining the "match" operation to 4756 * "next" but not the "study", so we can collect the TreeInfo 4757 * of each atom node without including the TreeInfo of the 4758 * "next". 4759 */ 4760 static final class BranchConn extends Node { 4761 BranchConn() {} 4762 boolean match(Matcher matcher, int i, CharSequence seq) { 4763 return next.match(matcher, i, seq); 4764 } 4765 boolean study(TreeInfo info) { 4766 return info.deterministic; 4767 } 4768 } 4769 4770 /** 4771 * Handles the branching of alternations. Note this is also used for 4772 * the ? quantifier to branch between the case where it matches once 4773 * and where it does not occur. 4774 */ 4775 static final class Branch extends Node { 4776 Node[] atoms = new Node[2]; 4777 int size = 2; 4778 BranchConn conn; 4779 Branch(Node first, Node second, BranchConn branchConn) { 4780 conn = branchConn; 4781 atoms[0] = first; 4782 atoms[1] = second; 4783 } 4784 4785 void add(Node node) { 4786 if (size >= atoms.length) { 4787 int len = ArraysSupport.newLength(size, 4788 1, /* minimum growth */ 4789 size /* preferred growth */); 4790 atoms = Arrays.copyOf(atoms, len); 4791 } 4792 atoms[size++] = node; 4793 } 4794 4795 boolean match(Matcher matcher, int i, CharSequence seq) { 4796 for (int n = 0; n < size; n++) { 4797 if (atoms[n] == null) { 4798 if (conn.next.match(matcher, i, seq)) 4799 return true; 4800 } else if (atoms[n].match(matcher, i, seq)) { 4801 return true; 4802 } 4803 } 4804 return false; 4805 } 4806 4807 boolean study(TreeInfo info) { 4808 int minL = info.minLength; 4809 int maxL = info.maxLength; 4810 boolean maxV = info.maxValid; 4811 4812 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num 4813 int maxL2 = -1; 4814 for (int n = 0; n < size; n++) { 4815 info.reset(); 4816 if (atoms[n] != null) 4817 atoms[n].study(info); 4818 minL2 = Math.min(minL2, info.minLength); 4819 maxL2 = Math.max(maxL2, info.maxLength); 4820 maxV = (maxV & info.maxValid); 4821 } 4822 4823 minL += minL2; 4824 maxL += maxL2; 4825 4826 info.reset(); 4827 conn.next.study(info); 4828 4829 info.minLength += minL; 4830 info.maxLength += maxL; 4831 info.maxValid &= maxV; 4832 info.deterministic = false; 4833 return false; 4834 } 4835 } 4836 4837 /** 4838 * The GroupHead saves the location where the group begins in the locals 4839 * and restores them when the match is done. 4840 * 4841 * The matchRef is used when a reference to this group is accessed later 4842 * in the expression. The locals will have a negative value in them to 4843 * indicate that we do not want to unset the group if the reference 4844 * doesn't match. 4845 */ 4846 static final class GroupHead extends Node { 4847 int localIndex; 4848 GroupTail tail; // for debug/print only, match does not need to know 4849 GroupHead(int localCount) { 4850 localIndex = localCount; 4851 } 4852 boolean match(Matcher matcher, int i, CharSequence seq) { 4853 int save = matcher.locals[localIndex]; 4854 matcher.locals[localIndex] = i; 4855 boolean ret = next.match(matcher, i, seq); 4856 matcher.locals[localIndex] = save; 4857 return ret; 4858 } 4859 } 4860 4861 /** 4862 * The GroupTail handles the setting of group beginning and ending 4863 * locations when groups are successfully matched. It must also be able to 4864 * unset groups that have to be backed off of. 4865 * 4866 * The GroupTail node is also used when a previous group is referenced, 4867 * and in that case no group information needs to be set. 4868 */ 4869 static final class GroupTail extends Node { 4870 int localIndex; 4871 int groupIndex; 4872 GroupTail(int localCount, int groupCount) { 4873 localIndex = localCount; 4874 groupIndex = groupCount + groupCount; 4875 } 4876 boolean match(Matcher matcher, int i, CharSequence seq) { 4877 int tmp = matcher.locals[localIndex]; 4878 if (tmp >= 0) { // This is the normal group case. 4879 // Save the group so we can unset it if it 4880 // backs off of a match. 4881 int groupStart = matcher.groups[groupIndex]; 4882 int groupEnd = matcher.groups[groupIndex+1]; 4883 4884 matcher.groups[groupIndex] = tmp; 4885 matcher.groups[groupIndex+1] = i; 4886 if (next.match(matcher, i, seq)) { 4887 return true; 4888 } 4889 matcher.groups[groupIndex] = groupStart; 4890 matcher.groups[groupIndex+1] = groupEnd; 4891 return false; 4892 } else { 4893 // This is a group reference case. We don't need to save any 4894 // group info because it isn't really a group. 4895 matcher.last = i; 4896 return true; 4897 } 4898 } 4899 } 4900 4901 /** 4902 * This sets up a loop to handle a recursive quantifier structure. 4903 */ 4904 static final class Prolog extends Node { 4905 Loop loop; 4906 Prolog(Loop loop) { 4907 this.loop = loop; 4908 } 4909 boolean match(Matcher matcher, int i, CharSequence seq) { 4910 return loop.matchInit(matcher, i, seq); 4911 } 4912 boolean study(TreeInfo info) { 4913 return loop.study(info); 4914 } 4915 } 4916 4917 /** 4918 * Handles the repetition count for a greedy Curly. The matchInit 4919 * is called from the Prolog to save the index of where the group 4920 * beginning is stored. A zero length group check occurs in the 4921 * normal match but is skipped in the matchInit. 4922 */ 4923 static class Loop extends Node { 4924 Node body; 4925 int countIndex; // local count index in matcher locals 4926 int beginIndex; // group beginning index 4927 int cmin, cmax; 4928 int posIndex; 4929 Loop(int countIndex, int beginIndex) { 4930 this.countIndex = countIndex; 4931 this.beginIndex = beginIndex; 4932 this.posIndex = -1; 4933 } 4934 boolean match(Matcher matcher, int i, CharSequence seq) { 4935 // Avoid infinite loop in zero-length case. 4936 if (i > matcher.locals[beginIndex]) { 4937 int count = matcher.locals[countIndex]; 4938 4939 // This block is for before we reach the minimum 4940 // iterations required for the loop to match 4941 if (count < cmin) { 4942 matcher.locals[countIndex] = count + 1; 4943 boolean b = body.match(matcher, i, seq); 4944 // If match failed we must backtrack, so 4945 // the loop count should NOT be incremented 4946 if (!b) 4947 matcher.locals[countIndex] = count; 4948 // Return success or failure since we are under 4949 // minimum 4950 return b; 4951 } 4952 // This block is for after we have the minimum 4953 // iterations required for the loop to match 4954 if (count < cmax) { 4955 // Let's check if we have already tried and failed 4956 // at this starting position "i" in the past. 4957 // If yes, then just return false wihtout trying 4958 // again, to stop the exponential backtracking. 4959 if (posIndex != -1 && 4960 matcher.localsPos[posIndex].contains(i)) { 4961 return next.match(matcher, i, seq); 4962 } 4963 matcher.locals[countIndex] = count + 1; 4964 boolean b = body.match(matcher, i, seq); 4965 // If match failed we must backtrack, so 4966 // the loop count should NOT be incremented 4967 if (b) 4968 return true; 4969 matcher.locals[countIndex] = count; 4970 // save the failed position 4971 if (posIndex != -1) { 4972 matcher.localsPos[posIndex].add(i); 4973 } 4974 } 4975 } 4976 return next.match(matcher, i, seq); 4977 } 4978 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4979 int save = matcher.locals[countIndex]; 4980 boolean ret; 4981 if (posIndex != -1 && matcher.localsPos[posIndex] == null) { 4982 matcher.localsPos[posIndex] = new IntHashSet(); 4983 } 4984 if (0 < cmin) { 4985 matcher.locals[countIndex] = 1; 4986 ret = body.match(matcher, i, seq); 4987 } else if (0 < cmax) { 4988 matcher.locals[countIndex] = 1; 4989 ret = body.match(matcher, i, seq); 4990 if (ret == false) 4991 ret = next.match(matcher, i, seq); 4992 } else { 4993 ret = next.match(matcher, i, seq); 4994 } 4995 matcher.locals[countIndex] = save; 4996 return ret; 4997 } 4998 boolean study(TreeInfo info) { 4999 info.maxValid = false; 5000 info.deterministic = false; 5001 return false; 5002 } 5003 } 5004 5005 /** 5006 * Handles the repetition count for a reluctant Curly. The matchInit 5007 * is called from the Prolog to save the index of where the group 5008 * beginning is stored. A zero length group check occurs in the 5009 * normal match but is skipped in the matchInit. 5010 */ 5011 static final class LazyLoop extends Loop { 5012 LazyLoop(int countIndex, int beginIndex) { 5013 super(countIndex, beginIndex); 5014 } 5015 boolean match(Matcher matcher, int i, CharSequence seq) { 5016 // Check for zero length group 5017 if (i > matcher.locals[beginIndex]) { 5018 int count = matcher.locals[countIndex]; 5019 if (count < cmin) { 5020 matcher.locals[countIndex] = count + 1; 5021 boolean result = body.match(matcher, i, seq); 5022 // If match failed we must backtrack, so 5023 // the loop count should NOT be incremented 5024 if (!result) 5025 matcher.locals[countIndex] = count; 5026 return result; 5027 } 5028 if (next.match(matcher, i, seq)) 5029 return true; 5030 if (count < cmax) { 5031 matcher.locals[countIndex] = count + 1; 5032 boolean result = body.match(matcher, i, seq); 5033 // If match failed we must backtrack, so 5034 // the loop count should NOT be incremented 5035 if (!result) 5036 matcher.locals[countIndex] = count; 5037 return result; 5038 } 5039 return false; 5040 } 5041 return next.match(matcher, i, seq); 5042 } 5043 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 5044 int save = matcher.locals[countIndex]; 5045 boolean ret = false; 5046 if (0 < cmin) { 5047 matcher.locals[countIndex] = 1; 5048 ret = body.match(matcher, i, seq); 5049 } else if (next.match(matcher, i, seq)) { 5050 ret = true; 5051 } else if (0 < cmax) { 5052 matcher.locals[countIndex] = 1; 5053 ret = body.match(matcher, i, seq); 5054 } 5055 matcher.locals[countIndex] = save; 5056 return ret; 5057 } 5058 boolean study(TreeInfo info) { 5059 info.maxValid = false; 5060 info.deterministic = false; 5061 return false; 5062 } 5063 } 5064 5065 /** 5066 * Refers to a group in the regular expression. Attempts to match 5067 * whatever the group referred to last matched. 5068 */ 5069 static class BackRef extends Node { 5070 int groupIndex; 5071 BackRef(int groupCount) { 5072 super(); 5073 groupIndex = groupCount + groupCount; 5074 } 5075 boolean match(Matcher matcher, int i, CharSequence seq) { 5076 int j = matcher.groups[groupIndex]; 5077 int k = matcher.groups[groupIndex+1]; 5078 5079 int groupSize = k - j; 5080 // If the referenced group didn't match, neither can this 5081 if (j < 0) 5082 return false; 5083 5084 // If there isn't enough input left no match 5085 if (i + groupSize > matcher.to) { 5086 matcher.hitEnd = true; 5087 return false; 5088 } 5089 // Check each new char to make sure it matches what the group 5090 // referenced matched last time around 5091 for (int index=0; index<groupSize; index++) 5092 if (seq.charAt(i+index) != seq.charAt(j+index)) 5093 return false; 5094 5095 return next.match(matcher, i+groupSize, seq); 5096 } 5097 boolean study(TreeInfo info) { 5098 info.maxValid = false; 5099 return next.study(info); 5100 } 5101 } 5102 5103 static class CIBackRef extends Node { 5104 int groupIndex; 5105 boolean doUnicodeCase; 5106 CIBackRef(int groupCount, boolean doUnicodeCase) { 5107 super(); 5108 groupIndex = groupCount + groupCount; 5109 this.doUnicodeCase = doUnicodeCase; 5110 } 5111 boolean match(Matcher matcher, int i, CharSequence seq) { 5112 int j = matcher.groups[groupIndex]; 5113 int k = matcher.groups[groupIndex+1]; 5114 5115 int groupSize = k - j; 5116 5117 // If the referenced group didn't match, neither can this 5118 if (j < 0) 5119 return false; 5120 5121 // If there isn't enough input left no match 5122 if (i + groupSize > matcher.to) { 5123 matcher.hitEnd = true; 5124 return false; 5125 } 5126 5127 // Check each new char to make sure it matches what the group 5128 // referenced matched last time around 5129 int x = i; 5130 for (int index=0; index<groupSize; index++) { 5131 int c1 = Character.codePointAt(seq, x); 5132 int c2 = Character.codePointAt(seq, j); 5133 if (c1 != c2) { 5134 if (doUnicodeCase) { 5135 int cc1 = Character.toUpperCase(c1); 5136 int cc2 = Character.toUpperCase(c2); 5137 if (cc1 != cc2 && 5138 Character.toLowerCase(cc1) != 5139 Character.toLowerCase(cc2)) 5140 return false; 5141 } else { 5142 if (ASCII.toLower(c1) != ASCII.toLower(c2)) 5143 return false; 5144 } 5145 } 5146 x += Character.charCount(c1); 5147 j += Character.charCount(c2); 5148 } 5149 5150 return next.match(matcher, i+groupSize, seq); 5151 } 5152 boolean study(TreeInfo info) { 5153 info.maxValid = false; 5154 return next.study(info); 5155 } 5156 } 5157 5158 /** 5159 * Searches until the next instance of its atom. This is useful for 5160 * finding the atom efficiently without passing an instance of it 5161 * (greedy problem) and without a lot of wasted search time (reluctant 5162 * problem). 5163 */ 5164 static final class First extends Node { 5165 Node atom; 5166 First(Node node) { 5167 this.atom = BnM.optimize(node); 5168 } 5169 boolean match(Matcher matcher, int i, CharSequence seq) { 5170 if (atom instanceof BnM) { 5171 return atom.match(matcher, i, seq) 5172 && next.match(matcher, matcher.last, seq); 5173 } 5174 for (;;) { 5175 if (i > matcher.to) { 5176 matcher.hitEnd = true; 5177 return false; 5178 } 5179 if (atom.match(matcher, i, seq)) { 5180 return next.match(matcher, matcher.last, seq); 5181 } 5182 i += countChars(seq, i, 1); 5183 matcher.first++; 5184 } 5185 } 5186 boolean study(TreeInfo info) { 5187 atom.study(info); 5188 info.maxValid = false; 5189 info.deterministic = false; 5190 return next.study(info); 5191 } 5192 } 5193 5194 /** 5195 * Zero width positive lookahead. 5196 */ 5197 static final class Pos extends Node { 5198 Node cond; 5199 Pos(Node cond) { 5200 this.cond = cond; 5201 } 5202 boolean match(Matcher matcher, int i, CharSequence seq) { 5203 int savedTo = matcher.to; 5204 boolean conditionMatched; 5205 5206 // Relax transparent region boundaries for lookahead 5207 if (matcher.transparentBounds) 5208 matcher.to = matcher.getTextLength(); 5209 try { 5210 conditionMatched = cond.match(matcher, i, seq); 5211 } finally { 5212 // Reinstate region boundaries 5213 matcher.to = savedTo; 5214 } 5215 return conditionMatched && next.match(matcher, i, seq); 5216 } 5217 } 5218 5219 /** 5220 * Zero width negative lookahead. 5221 */ 5222 static final class Neg extends Node { 5223 Node cond; 5224 Neg(Node cond) { 5225 this.cond = cond; 5226 } 5227 boolean match(Matcher matcher, int i, CharSequence seq) { 5228 int savedTo = matcher.to; 5229 boolean conditionMatched; 5230 5231 // Relax transparent region boundaries for lookahead 5232 if (matcher.transparentBounds) 5233 matcher.to = matcher.getTextLength(); 5234 try { 5235 if (i < matcher.to) { 5236 conditionMatched = !cond.match(matcher, i, seq); 5237 } else { 5238 // If a negative lookahead succeeds then more input 5239 // could cause it to fail! 5240 matcher.requireEnd = true; 5241 conditionMatched = !cond.match(matcher, i, seq); 5242 } 5243 } finally { 5244 // Reinstate region boundaries 5245 matcher.to = savedTo; 5246 } 5247 return conditionMatched && next.match(matcher, i, seq); 5248 } 5249 } 5250 5251 /** 5252 * For use with lookbehinds; matches the position where the lookbehind 5253 * was encountered. 5254 */ 5255 static class LookBehindEndNode extends Node { 5256 private LookBehindEndNode() {} // Singleton 5257 5258 static LookBehindEndNode INSTANCE = new LookBehindEndNode(); 5259 5260 boolean match(Matcher matcher, int i, CharSequence seq) { 5261 return i == matcher.lookbehindTo; 5262 } 5263 } 5264 5265 /** 5266 * Zero width positive lookbehind. 5267 */ 5268 static class Behind extends Node { 5269 Node cond; 5270 int rmax, rmin; 5271 Behind(Node cond, int rmax, int rmin) { 5272 this.cond = cond; 5273 this.rmax = rmax; 5274 this.rmin = rmin; 5275 } 5276 5277 boolean match(Matcher matcher, int i, CharSequence seq) { 5278 int savedFrom = matcher.from; 5279 boolean conditionMatched = false; 5280 int startIndex = (!matcher.transparentBounds) ? 5281 matcher.from : 0; 5282 int from = Math.max(i - rmax, startIndex); 5283 // Set end boundary 5284 int savedLBT = matcher.lookbehindTo; 5285 matcher.lookbehindTo = i; 5286 // Relax transparent region boundaries for lookbehind 5287 if (matcher.transparentBounds) 5288 matcher.from = 0; 5289 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5290 conditionMatched = cond.match(matcher, j, seq); 5291 } 5292 matcher.from = savedFrom; 5293 matcher.lookbehindTo = savedLBT; 5294 return conditionMatched && next.match(matcher, i, seq); 5295 } 5296 } 5297 5298 /** 5299 * Zero width positive lookbehind, including supplementary 5300 * characters or unpaired surrogates. 5301 */ 5302 static final class BehindS extends Behind { 5303 BehindS(Node cond, int rmax, int rmin) { 5304 super(cond, rmax, rmin); 5305 } 5306 boolean match(Matcher matcher, int i, CharSequence seq) { 5307 int rmaxChars = countChars(seq, i, -rmax); 5308 int rminChars = countChars(seq, i, -rmin); 5309 int savedFrom = matcher.from; 5310 int startIndex = (!matcher.transparentBounds) ? 5311 matcher.from : 0; 5312 boolean conditionMatched = false; 5313 int from = Math.max(i - rmaxChars, startIndex); 5314 // Set end boundary 5315 int savedLBT = matcher.lookbehindTo; 5316 matcher.lookbehindTo = i; 5317 // Relax transparent region boundaries for lookbehind 5318 if (matcher.transparentBounds) 5319 matcher.from = 0; 5320 5321 for (int j = i - rminChars; 5322 !conditionMatched && j >= from; 5323 j -= j>from ? countChars(seq, j, -1) : 1) { 5324 conditionMatched = cond.match(matcher, j, seq); 5325 } 5326 matcher.from = savedFrom; 5327 matcher.lookbehindTo = savedLBT; 5328 return conditionMatched && next.match(matcher, i, seq); 5329 } 5330 } 5331 5332 /** 5333 * Zero width negative lookbehind. 5334 */ 5335 static class NotBehind extends Node { 5336 Node cond; 5337 int rmax, rmin; 5338 NotBehind(Node cond, int rmax, int rmin) { 5339 this.cond = cond; 5340 this.rmax = rmax; 5341 this.rmin = rmin; 5342 } 5343 5344 boolean match(Matcher matcher, int i, CharSequence seq) { 5345 int savedLBT = matcher.lookbehindTo; 5346 int savedFrom = matcher.from; 5347 boolean conditionMatched = false; 5348 int startIndex = (!matcher.transparentBounds) ? 5349 matcher.from : 0; 5350 int from = Math.max(i - rmax, startIndex); 5351 matcher.lookbehindTo = i; 5352 // Relax transparent region boundaries for lookbehind 5353 if (matcher.transparentBounds) 5354 matcher.from = 0; 5355 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5356 conditionMatched = cond.match(matcher, j, seq); 5357 } 5358 // Reinstate region boundaries 5359 matcher.from = savedFrom; 5360 matcher.lookbehindTo = savedLBT; 5361 return !conditionMatched && next.match(matcher, i, seq); 5362 } 5363 } 5364 5365 /** 5366 * Zero width negative lookbehind, including supplementary 5367 * characters or unpaired surrogates. 5368 */ 5369 static final class NotBehindS extends NotBehind { 5370 NotBehindS(Node cond, int rmax, int rmin) { 5371 super(cond, rmax, rmin); 5372 } 5373 boolean match(Matcher matcher, int i, CharSequence seq) { 5374 int rmaxChars = countChars(seq, i, -rmax); 5375 int rminChars = countChars(seq, i, -rmin); 5376 int savedFrom = matcher.from; 5377 int savedLBT = matcher.lookbehindTo; 5378 boolean conditionMatched = false; 5379 int startIndex = (!matcher.transparentBounds) ? 5380 matcher.from : 0; 5381 int from = Math.max(i - rmaxChars, startIndex); 5382 matcher.lookbehindTo = i; 5383 // Relax transparent region boundaries for lookbehind 5384 if (matcher.transparentBounds) 5385 matcher.from = 0; 5386 for (int j = i - rminChars; 5387 !conditionMatched && j >= from; 5388 j -= j>from ? countChars(seq, j, -1) : 1) { 5389 conditionMatched = cond.match(matcher, j, seq); 5390 } 5391 //Reinstate region boundaries 5392 matcher.from = savedFrom; 5393 matcher.lookbehindTo = savedLBT; 5394 return !conditionMatched && next.match(matcher, i, seq); 5395 } 5396 } 5397 5398 /** 5399 * Handles word boundaries. Includes a field to allow this one class to 5400 * deal with the different types of word boundaries we can match. The word 5401 * characters include underscores, letters, and digits. Non spacing marks 5402 * can are also part of a word if they have a base character, otherwise 5403 * they are ignored for purposes of finding word boundaries. 5404 */ 5405 static final class Bound extends Node { 5406 static int LEFT = 0x1; 5407 static int RIGHT= 0x2; 5408 static int BOTH = 0x3; 5409 static int NONE = 0x4; 5410 int type; 5411 boolean useUWORD; 5412 Bound(int n, boolean useUWORD) { 5413 type = n; 5414 this.useUWORD = useUWORD; 5415 } 5416 5417 boolean isWord(int ch) { 5418 return useUWORD ? CharPredicates.WORD().is(ch) 5419 : (ch == '_' || Character.isLetterOrDigit(ch)); 5420 } 5421 5422 int check(Matcher matcher, int i, CharSequence seq) { 5423 int ch; 5424 boolean left = false; 5425 int startIndex = matcher.from; 5426 int endIndex = matcher.to; 5427 if (matcher.transparentBounds) { 5428 startIndex = 0; 5429 endIndex = matcher.getTextLength(); 5430 } 5431 if (i > startIndex) { 5432 ch = Character.codePointBefore(seq, i); 5433 left = (isWord(ch) || 5434 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5435 && hasBaseCharacter(matcher, i-1, seq))); 5436 } 5437 boolean right = false; 5438 if (i < endIndex) { 5439 ch = Character.codePointAt(seq, i); 5440 right = (isWord(ch) || 5441 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5442 && hasBaseCharacter(matcher, i, seq))); 5443 } else { 5444 // Tried to access char past the end 5445 matcher.hitEnd = true; 5446 // The addition of another char could wreck a boundary 5447 matcher.requireEnd = true; 5448 } 5449 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE); 5450 } 5451 boolean match(Matcher matcher, int i, CharSequence seq) { 5452 return (check(matcher, i, seq) & type) > 0 5453 && next.match(matcher, i, seq); 5454 } 5455 } 5456 5457 /** 5458 * Non spacing marks only count as word characters in bounds calculations 5459 * if they have a base character. 5460 */ 5461 private static boolean hasBaseCharacter(Matcher matcher, int i, 5462 CharSequence seq) 5463 { 5464 int start = (!matcher.transparentBounds) ? 5465 matcher.from : 0; 5466 for (int x=i; x >= start; x--) { 5467 int ch = Character.codePointAt(seq, x); 5468 if (Character.isLetterOrDigit(ch)) 5469 return true; 5470 if (Character.getType(ch) == Character.NON_SPACING_MARK) 5471 continue; 5472 return false; 5473 } 5474 return false; 5475 } 5476 5477 /** 5478 * Attempts to match a slice in the input using the Boyer-Moore string 5479 * matching algorithm. The algorithm is based on the idea that the 5480 * pattern can be shifted farther ahead in the search text if it is 5481 * matched right to left. 5482 * <p> 5483 * The pattern is compared to the input one character at a time, from 5484 * the rightmost character in the pattern to the left. If the characters 5485 * all match the pattern has been found. If a character does not match, 5486 * the pattern is shifted right a distance that is the maximum of two 5487 * functions, the bad character shift and the good suffix shift. This 5488 * shift moves the attempted match position through the input more 5489 * quickly than a naive one position at a time check. 5490 * <p> 5491 * The bad character shift is based on the character from the text that 5492 * did not match. If the character does not appear in the pattern, the 5493 * pattern can be shifted completely beyond the bad character. If the 5494 * character does occur in the pattern, the pattern can be shifted to 5495 * line the pattern up with the next occurrence of that character. 5496 * <p> 5497 * The good suffix shift is based on the idea that some subset on the right 5498 * side of the pattern has matched. When a bad character is found, the 5499 * pattern can be shifted right by the pattern length if the subset does 5500 * not occur again in pattern, or by the amount of distance to the 5501 * next occurrence of the subset in the pattern. 5502 * 5503 * Boyer-Moore search methods adapted from code by Amy Yu. 5504 */ 5505 static class BnM extends Node { 5506 int[] buffer; 5507 int[] lastOcc; 5508 int[] optoSft; 5509 5510 /** 5511 * Pre calculates arrays needed to generate the bad character 5512 * shift and the good suffix shift. Only the last seven bits 5513 * are used to see if chars match; This keeps the tables small 5514 * and covers the heavily used ASCII range, but occasionally 5515 * results in an aliased match for the bad character shift. 5516 */ 5517 static Node optimize(Node node) { 5518 if (!(node instanceof Slice)) { 5519 return node; 5520 } 5521 5522 int[] src = ((Slice) node).buffer; 5523 int patternLength = src.length; 5524 // The BM algorithm requires a bit of overhead; 5525 // If the pattern is short don't use it, since 5526 // a shift larger than the pattern length cannot 5527 // be used anyway. 5528 if (patternLength < 4) { 5529 return node; 5530 } 5531 int i, j; 5532 int[] lastOcc = new int[128]; 5533 int[] optoSft = new int[patternLength]; 5534 // Precalculate part of the bad character shift 5535 // It is a table for where in the pattern each 5536 // lower 7-bit value occurs 5537 for (i = 0; i < patternLength; i++) { 5538 lastOcc[src[i]&0x7F] = i + 1; 5539 } 5540 // Precalculate the good suffix shift 5541 // i is the shift amount being considered 5542 NEXT: for (i = patternLength; i > 0; i--) { 5543 // j is the beginning index of suffix being considered 5544 for (j = patternLength - 1; j >= i; j--) { 5545 // Testing for good suffix 5546 if (src[j] == src[j-i]) { 5547 // src[j..len] is a good suffix 5548 optoSft[j-1] = i; 5549 } else { 5550 // No match. The array has already been 5551 // filled up with correct values before. 5552 continue NEXT; 5553 } 5554 } 5555 // This fills up the remaining of optoSft 5556 // any suffix can not have larger shift amount 5557 // then its sub-suffix. Why??? 5558 while (j > 0) { 5559 optoSft[--j] = i; 5560 } 5561 } 5562 // Set the guard value because of unicode compression 5563 optoSft[patternLength-1] = 1; 5564 if (node instanceof SliceS) 5565 return new BnMS(src, lastOcc, optoSft, node.next); 5566 return new BnM(src, lastOcc, optoSft, node.next); 5567 } 5568 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5569 this.buffer = src; 5570 this.lastOcc = lastOcc; 5571 this.optoSft = optoSft; 5572 this.next = next; 5573 } 5574 boolean match(Matcher matcher, int i, CharSequence seq) { 5575 int[] src = buffer; 5576 int patternLength = src.length; 5577 int last = matcher.to - patternLength; 5578 5579 // Loop over all possible match positions in text 5580 NEXT: while (i <= last) { 5581 // Loop over pattern from right to left 5582 for (int j = patternLength - 1; j >= 0; j--) { 5583 int ch = seq.charAt(i+j); 5584 if (ch != src[j]) { 5585 // Shift search to the right by the maximum of the 5586 // bad character shift and the good suffix shift 5587 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]); 5588 continue NEXT; 5589 } 5590 } 5591 // Entire pattern matched starting at i 5592 matcher.first = i; 5593 boolean ret = next.match(matcher, i + patternLength, seq); 5594 if (ret) { 5595 matcher.first = i; 5596 matcher.groups[0] = matcher.first; 5597 matcher.groups[1] = matcher.last; 5598 return true; 5599 } 5600 i++; 5601 } 5602 // BnM is only used as the leading node in the unanchored case, 5603 // and it replaced its Start() which always searches to the end 5604 // if it doesn't find what it's looking for, so hitEnd is true. 5605 matcher.hitEnd = true; 5606 return false; 5607 } 5608 boolean study(TreeInfo info) { 5609 info.minLength += buffer.length; 5610 info.maxValid = false; 5611 return next.study(info); 5612 } 5613 } 5614 5615 /** 5616 * Supplementary support version of BnM(). Unpaired surrogates are 5617 * also handled by this class. 5618 */ 5619 static final class BnMS extends BnM { 5620 int lengthInChars; 5621 5622 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5623 super(src, lastOcc, optoSft, next); 5624 for (int cp : buffer) { 5625 lengthInChars += Character.charCount(cp); 5626 } 5627 } 5628 boolean match(Matcher matcher, int i, CharSequence seq) { 5629 int[] src = buffer; 5630 int patternLength = src.length; 5631 int last = matcher.to - lengthInChars; 5632 5633 // Loop over all possible match positions in text 5634 NEXT: while (i <= last) { 5635 // Loop over pattern from right to left 5636 int ch; 5637 for (int j = countChars(seq, i, patternLength), x = patternLength - 1; 5638 j > 0; j -= Character.charCount(ch), x--) { 5639 ch = Character.codePointBefore(seq, i+j); 5640 if (ch != src[x]) { 5641 // Shift search to the right by the maximum of the 5642 // bad character shift and the good suffix shift 5643 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]); 5644 i += countChars(seq, i, n); 5645 continue NEXT; 5646 } 5647 } 5648 // Entire pattern matched starting at i 5649 matcher.first = i; 5650 boolean ret = next.match(matcher, i + lengthInChars, seq); 5651 if (ret) { 5652 matcher.first = i; 5653 matcher.groups[0] = matcher.first; 5654 matcher.groups[1] = matcher.last; 5655 return true; 5656 } 5657 i += countChars(seq, i, 1); 5658 } 5659 matcher.hitEnd = true; 5660 return false; 5661 } 5662 } 5663 5664 @FunctionalInterface 5665 static interface CharPredicate { 5666 boolean is(int ch); 5667 5668 default CharPredicate and(CharPredicate p) { 5669 return ch -> is(ch) && p.is(ch); 5670 } 5671 default CharPredicate union(CharPredicate p) { 5672 return ch -> is(ch) || p.is(ch); 5673 } 5674 default CharPredicate union(CharPredicate p1, 5675 CharPredicate p2) { 5676 return ch -> is(ch) || p1.is(ch) || p2.is(ch); 5677 } 5678 default CharPredicate negate() { 5679 return ch -> !is(ch); 5680 } 5681 } 5682 5683 static interface BmpCharPredicate extends CharPredicate { 5684 5685 default CharPredicate and(CharPredicate p) { 5686 if (p instanceof BmpCharPredicate) 5687 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch)); 5688 return ch -> is(ch) && p.is(ch); 5689 } 5690 default CharPredicate union(CharPredicate p) { 5691 if (p instanceof BmpCharPredicate) 5692 return (BmpCharPredicate)(ch -> is(ch) || p.is(ch)); 5693 return ch -> is(ch) || p.is(ch); 5694 } 5695 static CharPredicate union(CharPredicate... predicates) { 5696 CharPredicate cp = ch -> { 5697 for (CharPredicate p : predicates) { 5698 if (!p.is(ch)) 5699 return false; 5700 } 5701 return true; 5702 }; 5703 for (CharPredicate p : predicates) { 5704 if (! (p instanceof BmpCharPredicate)) 5705 return cp; 5706 } 5707 return (BmpCharPredicate)cp; 5708 } 5709 } 5710 5711 /** 5712 * matches a Perl vertical whitespace 5713 */ 5714 static BmpCharPredicate VertWS() { 5715 return cp -> (cp >= 0x0A && cp <= 0x0D) || 5716 cp == 0x85 || cp == 0x2028 || cp == 0x2029; 5717 } 5718 5719 /** 5720 * matches a Perl horizontal whitespace 5721 */ 5722 static BmpCharPredicate HorizWS() { 5723 return cp -> 5724 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 || 5725 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f || 5726 cp == 0x205f || cp == 0x3000; 5727 } 5728 5729 /** 5730 * for the Unicode category ALL and the dot metacharacter when 5731 * in dotall mode. 5732 */ 5733 static CharPredicate ALL() { 5734 return ch -> true; 5735 } 5736 5737 /** 5738 * for the dot metacharacter when dotall is not enabled. 5739 */ 5740 static CharPredicate DOT() { 5741 return ch -> 5742 (ch != '\n' && ch != '\r' 5743 && (ch|1) != '\u2029' 5744 && ch != '\u0085'); 5745 } 5746 5747 /** 5748 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled. 5749 */ 5750 static CharPredicate UNIXDOT() { 5751 return ch -> ch != '\n'; 5752 } 5753 5754 /** 5755 * Indicate that matches a Supplementary Unicode character 5756 */ 5757 static CharPredicate SingleS(int c) { 5758 return ch -> ch == c; 5759 } 5760 5761 /** 5762 * A bmp/optimized predicate of single 5763 */ 5764 static BmpCharPredicate Single(int c) { 5765 return ch -> ch == c; 5766 } 5767 5768 /** 5769 * Case insensitive matches a given BMP character 5770 */ 5771 static BmpCharPredicate SingleI(int lower, int upper) { 5772 return ch -> ch == lower || ch == upper; 5773 } 5774 5775 /** 5776 * Unicode case insensitive matches a given Unicode character 5777 */ 5778 static CharPredicate SingleU(int lower) { 5779 return ch -> lower == ch || 5780 lower == Character.toLowerCase(Character.toUpperCase(ch)); 5781 } 5782 5783 private static boolean inRange(int lower, int ch, int upper) { 5784 return lower <= ch && ch <= upper; 5785 } 5786 5787 /** 5788 * Charactrs within a explicit value range 5789 */ 5790 static CharPredicate Range(int lower, int upper) { 5791 if (upper < Character.MIN_HIGH_SURROGATE || 5792 lower > Character.MAX_HIGH_SURROGATE && 5793 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT) 5794 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper)); 5795 return ch -> inRange(lower, ch, upper); 5796 } 5797 5798 /** 5799 * Charactrs within a explicit value range in a case insensitive manner. 5800 */ 5801 static CharPredicate CIRange(int lower, int upper) { 5802 return ch -> inRange(lower, ch, upper) || 5803 ASCII.isAscii(ch) && 5804 (inRange(lower, ASCII.toUpper(ch), upper) || 5805 inRange(lower, ASCII.toLower(ch), upper)); 5806 } 5807 5808 static CharPredicate CIRangeU(int lower, int upper) { 5809 return ch -> { 5810 if (inRange(lower, ch, upper)) 5811 return true; 5812 int up = Character.toUpperCase(ch); 5813 return inRange(lower, up, upper) || 5814 inRange(lower, Character.toLowerCase(up), upper); 5815 }; 5816 } 5817 5818 /** 5819 * This must be the very first initializer. 5820 */ 5821 static final Node accept = new Node(); 5822 5823 static final Node lastAccept = new LastNode(); 5824 5825 /** 5826 * Creates a predicate that tests if this pattern is found in a given input 5827 * string. 5828 * 5829 * @apiNote 5830 * This method creates a predicate that behaves as if it creates a matcher 5831 * from the input sequence and then calls {@code find}, for example a 5832 * predicate of the form: 5833 * <pre>{@code 5834 * s -> matcher(s).find(); 5835 * }</pre> 5836 * 5837 * @return The predicate which can be used for finding a match on a 5838 * subsequence of a string 5839 * @since 1.8 5840 * @see Matcher#find 5841 */ 5842 public Predicate<String> asPredicate() { 5843 return s -> matcher(s).find(); 5844 } 5845 5846 /** 5847 * Creates a predicate that tests if this pattern matches a given input string. 5848 * 5849 * @apiNote 5850 * This method creates a predicate that behaves as if it creates a matcher 5851 * from the input sequence and then calls {@code matches}, for example a 5852 * predicate of the form: 5853 * <pre>{@code 5854 * s -> matcher(s).matches(); 5855 * }</pre> 5856 * 5857 * @return The predicate which can be used for matching an input string 5858 * against this pattern. 5859 * @since 11 5860 * @see Matcher#matches 5861 */ 5862 public Predicate<String> asMatchPredicate() { 5863 return s -> matcher(s).matches(); 5864 } 5865 5866 /** 5867 * Creates a stream from the given input sequence around matches of this 5868 * pattern. 5869 * 5870 * <p> The stream returned by this method contains each substring of the 5871 * input sequence that is terminated by another subsequence that matches 5872 * this pattern or is terminated by the end of the input sequence. The 5873 * substrings in the stream are in the order in which they occur in the 5874 * input. Trailing empty strings will be discarded and not encountered in 5875 * the stream. 5876 * 5877 * <p> If this pattern does not match any subsequence of the input then 5878 * the resulting stream has just one element, namely the input sequence in 5879 * string form. 5880 * 5881 * <p> When there is a positive-width match at the beginning of the input 5882 * sequence then an empty leading substring is included at the beginning 5883 * of the stream. A zero-width match at the beginning however never produces 5884 * such empty leading substring. 5885 * 5886 * <p> If the input sequence is mutable, it must remain constant during the 5887 * execution of the terminal stream operation. Otherwise, the result of the 5888 * terminal stream operation is undefined. 5889 * 5890 * @param input 5891 * The character sequence to be split 5892 * 5893 * @return The stream of strings computed by splitting the input 5894 * around matches of this pattern 5895 * @see #split(CharSequence) 5896 * @since 1.8 5897 */ 5898 public Stream<String> splitAsStream(final CharSequence input) { 5899 class MatcherIterator implements Iterator<String> { 5900 private Matcher matcher; 5901 // The start position of the next sub-sequence of input 5902 // when current == input.length there are no more elements 5903 private int current; 5904 // null if the next element, if any, needs to obtained 5905 private String nextElement; 5906 // > 0 if there are N next empty elements 5907 private int emptyElementCount; 5908 5909 public String next() { 5910 if (!hasNext()) 5911 throw new NoSuchElementException(); 5912 5913 if (emptyElementCount == 0) { 5914 String n = nextElement; 5915 nextElement = null; 5916 return n; 5917 } else { 5918 emptyElementCount--; 5919 return ""; 5920 } 5921 } 5922 5923 public boolean hasNext() { 5924 if (matcher == null) { 5925 matcher = matcher(input); 5926 // If the input is an empty string then the result can only be a 5927 // stream of the input. Induce that by setting the empty 5928 // element count to 1 5929 emptyElementCount = input.length() == 0 ? 1 : 0; 5930 } 5931 if (nextElement != null || emptyElementCount > 0) 5932 return true; 5933 5934 if (current == input.length()) 5935 return false; 5936 5937 // Consume the next matching element 5938 // Count sequence of matching empty elements 5939 while (matcher.find()) { 5940 nextElement = input.subSequence(current, matcher.start()).toString(); 5941 current = matcher.end(); 5942 if (!nextElement.isEmpty()) { 5943 return true; 5944 } else if (current > 0) { // no empty leading substring for zero-width 5945 // match at the beginning of the input 5946 emptyElementCount++; 5947 } 5948 } 5949 5950 // Consume last matching element 5951 nextElement = input.subSequence(current, input.length()).toString(); 5952 current = input.length(); 5953 if (!nextElement.isEmpty()) { 5954 return true; 5955 } else { 5956 // Ignore a terminal sequence of matching empty elements 5957 emptyElementCount = 0; 5958 nextElement = null; 5959 return false; 5960 } 5961 } 5962 } 5963 return StreamSupport.stream(Spliterators.spliteratorUnknownSize( 5964 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false); 5965 } 5966 }