1 /* 2 * Copyright (c) 1999, 2019, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. Oracle designates this 8 * particular file as subject to the "Classpath" exception as provided 9 * by Oracle in the LICENSE file that accompanied this code. 10 * 11 * This code is distributed in the hope that it will be useful, but WITHOUT 12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 14 * version 2 for more details (a copy is included in the LICENSE file that 15 * accompanied this code). 16 * 17 * You should have received a copy of the GNU General Public License version 18 * 2 along with this work; if not, write to the Free Software Foundation, 19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 20 * 21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 22 * or visit www.oracle.com if you need additional information or have any 23 * questions. 24 */ 25 26 package java.util.regex; 27 28 import java.text.Normalizer; 29 import java.text.Normalizer.Form; 30 import java.util.Locale; 31 import java.util.Iterator; 32 import java.util.Map; 33 import java.util.ArrayList; 34 import java.util.HashMap; 35 import java.util.LinkedHashSet; 36 import java.util.List; 37 import java.util.Set; 38 import java.util.Arrays; 39 import java.util.NoSuchElementException; 40 import java.util.Spliterator; 41 import java.util.Spliterators; 42 import java.util.function.Predicate; 43 import java.util.stream.Stream; 44 import java.util.stream.StreamSupport; 45 46 import jdk.internal.util.ArraysSupport; 47 48 /** 49 * A compiled representation of a regular expression. 50 * 51 * <p> A regular expression, specified as a string, must first be compiled into 52 * an instance of this class. The resulting pattern can then be used to create 53 * a {@link Matcher} object that can match arbitrary {@linkplain 54 * java.lang.CharSequence character sequences} against the regular 55 * expression. All of the state involved in performing a match resides in the 56 * matcher, so many matchers can share the same pattern. 57 * 58 * <p> A typical invocation sequence is thus 59 * 60 * <blockquote><pre> 61 * Pattern p = Pattern.{@link #compile compile}("a*b"); 62 * Matcher m = p.{@link #matcher matcher}("aaaaab"); 63 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote> 64 * 65 * <p> A {@link #matches matches} method is defined by this class as a 66 * convenience for when a regular expression is used just once. This method 67 * compiles an expression and matches an input sequence against it in a single 68 * invocation. The statement 69 * 70 * <blockquote><pre> 71 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote> 72 * 73 * is equivalent to the three statements above, though for repeated matches it 74 * is less efficient since it does not allow the compiled pattern to be reused. 75 * 76 * <p> Instances of this class are immutable and are safe for use by multiple 77 * concurrent threads. Instances of the {@link Matcher} class are not safe for 78 * such use. 79 * 80 * 81 * <h2><a id="sum">Summary of regular-expression constructs</a></h2> 82 * 83 * <table class="borderless"> 84 * <caption style="display:none">Regular expression constructs, and what they match</caption> 85 * <thead style="text-align:left"> 86 * <tr> 87 * <th id="construct">Construct</th> 88 * <th id="matches">Matches</th> 89 * </tr> 90 * </thead> 91 * <tbody style="text-align:left"> 92 * 93 * <tr><th colspan="2" style="padding-top:20px" id="characters">Characters</th></tr> 94 * 95 * <tr><th style="vertical-align:top; font-weight: normal" id="x"><i>x</i></th> 96 * <td headers="matches characters x">The character <i>x</i></td></tr> 97 * <tr><th style="vertical-align:top; font-weight: normal" id="backslash">{@code \\}</th> 98 * <td headers="matches characters backslash">The backslash character</td></tr> 99 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_n">{@code \0}<i>n</i></th> 100 * <td headers="matches characters octal_n">The character with octal value {@code 0}<i>n</i> 101 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 102 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nn">{@code \0}<i>nn</i></th> 103 * <td headers="matches characters octal_nn">The character with octal value {@code 0}<i>nn</i> 104 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 105 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nnn">{@code \0}<i>mnn</i></th> 106 * <td headers="matches characters octal_nnn">The character with octal value {@code 0}<i>mnn</i> 107 * (0 {@code <=} <i>m</i> {@code <=} 3, 108 * 0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 109 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hh">{@code \x}<i>hh</i></th> 110 * <td headers="matches characters hex_hh">The character with hexadecimal value {@code 0x}<i>hh</i></td></tr> 111 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hhhh"><code>\u</code><i>hhhh</i></th> 112 * <td headers="matches characters hex_hhhh">The character with hexadecimal value {@code 0x}<i>hhhh</i></td></tr> 113 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_h_h"><code>\x</code><i>{h...h}</i></th> 114 * <td headers="matches characters hex_h_h">The character with hexadecimal value {@code 0x}<i>h...h</i> 115 * ({@link java.lang.Character#MIN_CODE_POINT Character.MIN_CODE_POINT} 116 * <= {@code 0x}<i>h...h</i> <= 117 * {@link java.lang.Character#MAX_CODE_POINT Character.MAX_CODE_POINT})</td></tr> 118 * <tr><th style="vertical-align:top; font-weight: normal" id="unicode_name"><code>\N{</code><i>name</i><code>}</code></th> 119 * <td headers="matches characters unicode_name">The character with Unicode character name <i>'name'</i></td></tr> 120 * <tr><th style="vertical-align:top; font-weight:normal" id="tab">{@code \t}</th> 121 * <td headers="matches characters tab">The tab character (<code>'\u0009'</code>)</td></tr> 122 * <tr><th style="vertical-align:top; font-weight:normal" id="newline">{@code \n}</th> 123 * <td headers="matches characters newline">The newline (line feed) character (<code>'\u000A'</code>)</td></tr> 124 * <tr><th style="vertical-align:top; font-weight:normal" id="return">{@code \r}</th> 125 * <td headers="matches characters return">The carriage-return character (<code>'\u000D'</code>)</td></tr> 126 * <tr><th style="vertical-align:top; font-weight:normal" id="form_feed">{@code \f}</th> 127 * <td headers="matches characters form_feed">The form-feed character (<code>'\u000C'</code>)</td></tr> 128 * <tr><th style="vertical-align:top; font-weight:normal" id="bell">{@code \a}</th> 129 * <td headers="matches characters bell">The alert (bell) character (<code>'\u0007'</code>)</td></tr> 130 * <tr><th style="vertical-align:top; font-weight:normal" id="escape">{@code \e}</th> 131 * <td headers="matches characters escape">The escape character (<code>'\u001B'</code>)</td></tr> 132 * <tr><th style="vertical-align:top; font-weight:normal" id="ctrl_x">{@code \c}<i>x</i></th> 133 * <td headers="matches characters ctrl_x">The control character corresponding to <i>x</i></td></tr> 134 * 135 * <tr><th colspan="2" style="padding-top:20px" id="classes">Character classes</th></tr> 136 * 137 * <tr><th style="vertical-align:top; font-weight:normal" id="simple">{@code [abc]}</th> 138 * <td headers="matches classes simple">{@code a}, {@code b}, or {@code c} (simple class)</td></tr> 139 * <tr><th style="vertical-align:top; font-weight:normal" id="negation">{@code [^abc]}</th> 140 * <td headers="matches classes negation">Any character except {@code a}, {@code b}, or {@code c} (negation)</td></tr> 141 * <tr><th style="vertical-align:top; font-weight:normal" id="range">{@code [a-zA-Z]}</th> 142 * <td headers="matches classes range">{@code a} through {@code z} 143 * or {@code A} through {@code Z}, inclusive (range)</td></tr> 144 * <tr><th style="vertical-align:top; font-weight:normal" id="union">{@code [a-d[m-p]]}</th> 145 * <td headers="matches classes union">{@code a} through {@code d}, 146 * or {@code m} through {@code p}: {@code [a-dm-p]} (union)</td></tr> 147 * <tr><th style="vertical-align:top; font-weight:normal" id="intersection">{@code [a-z&&[def]]}</th> 148 * <td headers="matches classes intersection">{@code d}, {@code e}, or {@code f} (intersection)</tr> 149 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction1">{@code [a-z&&[^bc]]}</th> 150 * <td headers="matches classes subtraction1">{@code a} through {@code z}, 151 * except for {@code b} and {@code c}: {@code [ad-z]} (subtraction)</td></tr> 152 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction2">{@code [a-z&&[^m-p]]}</th> 153 * <td headers="matches classes subtraction2">{@code a} through {@code z}, 154 * and not {@code m} through {@code p}: {@code [a-lq-z]}(subtraction)</td></tr> 155 * 156 * <tr><th colspan="2" style="padding-top:20px" id="predef">Predefined character classes</th></tr> 157 * 158 * <tr><th style="vertical-align:top; font-weight:normal" id="any">{@code .}</th> 159 * <td headers="matches predef any">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr> 160 * <tr><th style="vertical-align:top; font-weight:normal" id="digit">{@code \d}</th> 161 * <td headers="matches predef digit">A digit: {@code [0-9]}</td></tr> 162 * <tr><th style="vertical-align:top; font-weight:normal" id="non_digit">{@code \D}</th> 163 * <td headers="matches predef non_digit">A non-digit: {@code [^0-9]}</td></tr> 164 * <tr><th style="vertical-align:top; font-weight:normal" id="horiz_white">{@code \h}</th> 165 * <td headers="matches predef horiz_white">A horizontal whitespace character: 166 * <code>[ \t\xA0\u1680\u180e\u2000-\u200a\u202f\u205f\u3000]</code></td></tr> 167 * <tr><th style="vertical-align:top; font-weight:normal" id="non_horiz_white">{@code \H}</th> 168 * <td headers="matches predef non_horiz_white">A non-horizontal whitespace character: {@code [^\h]}</td></tr> 169 * <tr><th style="vertical-align:top; font-weight:normal" id="white">{@code \s}</th> 170 * <td headers="matches predef white">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 171 * <tr><th style="vertical-align:top; font-weight:normal" id="non_white">{@code \S}</th> 172 * <td headers="matches predef non_white">A non-whitespace character: {@code [^\s]}</td></tr> 173 * <tr><th style="vertical-align:top; font-weight:normal" id="vert_white">{@code \v}</th> 174 * <td headers="matches predef vert_white">A vertical whitespace character: <code>[\n\x0B\f\r\x85\u2028\u2029]</code> 175 * </td></tr> 176 * <tr><th style="vertical-align:top; font-weight:normal" id="non_vert_white">{@code \V}</th> 177 * <td headers="matches predef non_vert_white">A non-vertical whitespace character: {@code [^\v]}</td></tr> 178 * <tr><th style="vertical-align:top; font-weight:normal" id="word">{@code \w}</th> 179 * <td headers="matches predef word">A word character: {@code [a-zA-Z_0-9]}</td></tr> 180 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word">{@code \W}</th> 181 * <td headers="matches predef non_word">A non-word character: {@code [^\w]}</td></tr> 182 * 183 * <tr><th colspan="2" style="padding-top:20px" id="posix"><b>POSIX character classes (US-ASCII only)</b></th></tr> 184 * 185 * <tr><th style="vertical-align:top; font-weight:normal" id="Lower">{@code \p{Lower}}</th> 186 * <td headers="matches posix Lower">A lower-case alphabetic character: {@code [a-z]}</td></tr> 187 * <tr><th style="vertical-align:top; font-weight:normal" id="Upper">{@code \p{Upper}}</th> 188 * <td headers="matches posix Upper">An upper-case alphabetic character:{@code [A-Z]}</td></tr> 189 * <tr><th style="vertical-align:top; font-weight:normal" id="ASCII">{@code \p{ASCII}}</th> 190 * <td headers="matches posix ASCII">All ASCII:{@code [\x00-\x7F]}</td></tr> 191 * <tr><th style="vertical-align:top; font-weight:normal" id="Alpha">{@code \p{Alpha}}</th> 192 * <td headers="matches posix Alpha">An alphabetic character:{@code [\p{Lower}\p{Upper}]}</td></tr> 193 * <tr><th style="vertical-align:top; font-weight:normal" id="Digit">{@code \p{Digit}}</th> 194 * <td headers="matches posix Digit">A decimal digit: {@code [0-9]}</td></tr> 195 * <tr><th style="vertical-align:top; font-weight:normal" id="Alnum">{@code \p{Alnum}}</th> 196 * <td headers="matches posix Alnum">An alphanumeric character:{@code [\p{Alpha}\p{Digit}]}</td></tr> 197 * <tr><th style="vertical-align:top; font-weight:normal" id="Punct">{@code \p{Punct}}</th> 198 * <td headers="matches posix Punct">Punctuation: One of {@code !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~}</td></tr> 199 * <!-- {@code [\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]} 200 * {@code [\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]} --> 201 * <tr><th style="vertical-align:top; font-weight:normal" id="Graph">{@code \p{Graph}}</th> 202 * <td headers="matches posix Graph">A visible character: {@code [\p{Alnum}\p{Punct}]}</td></tr> 203 * <tr><th style="vertical-align:top; font-weight:normal" id="Print">{@code \p{Print}}</th> 204 * <td headers="matches posix Print">A printable character: {@code [\p{Graph}\x20]}</td></tr> 205 * <tr><th style="vertical-align:top; font-weight:normal" id="Blank">{@code \p{Blank}}</th> 206 * <td headers="matches posix Blank">A space or a tab: {@code [ \t]}</td></tr> 207 * <tr><th style="vertical-align:top; font-weight:normal" id="Cntrl">{@code \p{Cntrl}}</th> 208 * <td headers="matches posix Cntrl">A control character: {@code [\x00-\x1F\x7F]}</td></tr> 209 * <tr><th style="vertical-align:top; font-weight:normal" id="XDigit">{@code \p{XDigit}}</th> 210 * <td headers="matches posix XDigit">A hexadecimal digit: {@code [0-9a-fA-F]}</td></tr> 211 * <tr><th style="vertical-align:top; font-weight:normal" id="Space">{@code \p{Space}}</th> 212 * <td headers="matches posix Space">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 213 * 214 * <tr><th colspan="2" style="padding-top:20px" id="java">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr> 215 * 216 * <tr><th style="vertical-align:top; font-weight:normal" id="javaLowerCase">{@code \p{javaLowerCase}}</th> 217 * <td headers="matches java javaLowerCase">Equivalent to java.lang.Character.isLowerCase()</td></tr> 218 * <tr><th style="vertical-align:top; font-weight:normal" id="javaUpperCase">{@code \p{javaUpperCase}}</th> 219 * <td headers="matches java javaUpperCase">Equivalent to java.lang.Character.isUpperCase()</td></tr> 220 * <tr><th style="vertical-align:top; font-weight:normal" id="javaWhitespace">{@code \p{javaWhitespace}}</th> 221 * <td headers="matches java javaWhitespace">Equivalent to java.lang.Character.isWhitespace()</td></tr> 222 * <tr><th style="vertical-align:top; font-weight:normal" id="javaMirrored">{@code \p{javaMirrored}}</th> 223 * <td headers="matches java javaMirrored">Equivalent to java.lang.Character.isMirrored()</td></tr> 224 * 225 * <tr><th colspan="2" style="padding-top:20px" id="unicode">Classes for Unicode scripts, blocks, categories and binary properties</th></tr> 226 * 227 * <tr><th style="vertical-align:top; font-weight:normal" id="IsLatin">{@code \p{IsLatin}}</th> 228 * <td headers="matches unicode IsLatin">A Latin script character (<a href="#usc">script</a>)</td></tr> 229 * <tr><th style="vertical-align:top; font-weight:normal" id="InGreek">{@code \p{InGreek}}</th> 230 * <td headers="matches unicode InGreek">A character in the Greek block (<a href="#ubc">block</a>)</td></tr> 231 * <tr><th style="vertical-align:top; font-weight:normal" id="Lu">{@code \p{Lu}}</th> 232 * <td headers="matches unicode Lu">An uppercase letter (<a href="#ucc">category</a>)</td></tr> 233 * <tr><th style="vertical-align:top; font-weight:normal" id="IsAlphabetic">{@code \p{IsAlphabetic}}</th> 234 * <td headers="matches unicode IsAlphabetic">An alphabetic character (<a href="#ubpc">binary property</a>)</td></tr> 235 * <tr><th style="vertical-align:top; font-weight:normal" id="Sc">{@code \p{Sc}}</th> 236 * <td headers="matches unicode Sc">A currency symbol</td></tr> 237 * <tr><th style="vertical-align:top; font-weight:normal" id="not_InGreek">{@code \P{InGreek}}</th> 238 * <td headers="matches unicode not_InGreek">Any character except one in the Greek block (negation)</td></tr> 239 * <tr><th style="vertical-align:top; font-weight:normal" id="not_uppercase">{@code [\p{L}&&[^\p{Lu}]]}</th> 240 * <td headers="matches unicode not_uppercase">Any letter except an uppercase letter (subtraction)</td></tr> 241 * 242 * <tr><th colspan="2" style="padding-top:20px" id="bounds">Boundary matchers</th></tr> 243 * 244 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_line">{@code ^}</th> 245 * <td headers="matches bounds begin_line">The beginning of a line</td></tr> 246 * <tr><th style="vertical-align:top; font-weight:normal" id="end_line">{@code $}</th> 247 * <td headers="matches bounds end_line">The end of a line</td></tr> 248 * <tr><th style="vertical-align:top; font-weight:normal" id="word_boundary">{@code \b}</th> 249 * <td headers="matches bounds word_boundary">A word boundary</td></tr> 250 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_cluster_boundary">{@code \b{g}}</th> 251 * <td headers="matches bounds grapheme_cluster_boundary">A Unicode extended grapheme cluster boundary</td></tr> 252 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word_boundary">{@code \B}</th> 253 * <td headers="matches bounds non_word_boundary">A non-word boundary</td></tr> 254 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_input">{@code \A}</th> 255 * <td headers="matches bounds begin_input">The beginning of the input</td></tr> 256 * <tr><th style="vertical-align:top; font-weight:normal" id="end_prev_match">{@code \G}</th> 257 * <td headers="matches bounds end_prev_match">The end of the previous match</td></tr> 258 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input_except_term">{@code \Z}</th> 259 * <td headers="matches bounds end_input_except_term">The end of the input but for the final 260 * <a href="#lt">terminator</a>, if any</td></tr> 261 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input">{@code \z}</th> 262 * <td headers="matches bounds end_input">The end of the input</td></tr> 263 * 264 * <tr><th colspan="2" style="padding-top:20px" id="linebreak">Linebreak matcher</th></tr> 265 * 266 * <tr><th style="vertical-align:top; font-weight:normal" id="any_unicode_linebreak">{@code \R}</th> 267 * <td headers="matches linebreak any_unicode_linebreak">Any Unicode linebreak sequence, is equivalent to 268 * <code>\u000D\u000A|[\u000A\u000B\u000C\u000D\u0085\u2028\u2029] 269 * </code></td></tr> 270 * 271 * <tr><th colspan="2" style="padding-top:20px" id="grapheme">Unicode Extended Grapheme matcher</th></tr> 272 * 273 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_any">{@code \X}</th> 274 * <td headers="matches grapheme grapheme_any">Any Unicode extended grapheme cluster</td></tr> 275 * 276 * <tr><th colspan="2" style="padding-top:20px" id="greedy">Greedy quantifiers</th></tr> 277 * 278 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_once_or_not"><i>X</i>{@code ?}</th> 279 * <td headers="matches greedy greedy_once_or_not"><i>X</i>, once or not at all</td></tr> 280 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_zero_or_more"><i>X</i>{@code *}</th> 281 * <td headers="matches greedy greedy_zero_or_more"><i>X</i>, zero or more times</td></tr> 282 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_one_or_more"><i>X</i>{@code +}</th> 283 * <td headers="matches greedy greedy_one_or_more"><i>X</i>, one or more times</td></tr> 284 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_exactly"><i>X</i><code>{</code><i>n</i><code>}</code></th> 285 * <td headers="matches greedy greedy_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 286 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least"><i>X</i><code>{</code><i>n</i>{@code ,}}</th> 287 * <td headers="matches greedy greedy_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 288 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}</code></th> 289 * <td headers="matches greedy greedy_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 290 * 291 * <tr><th colspan="2" style="padding-top:20px" id="reluc">Reluctant quantifiers</th></tr> 292 * 293 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_once_or_not"><i>X</i>{@code ??}</th> 294 * <td headers="matches reluc reluc_once_or_not"><i>X</i>, once or not at all</td></tr> 295 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_zero_or_more"><i>X</i>{@code *?}</th> 296 * <td headers="matches reluc reluc_zero_or_more"><i>X</i>, zero or more times</td></tr> 297 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_one_or_more"><i>X</i>{@code +?}</th> 298 * <td headers="matches reluc reluc_one_or_more"><i>X</i>, one or more times</td></tr> 299 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_exactly"><i>X</i><code>{</code><i>n</i><code>}?</code></th> 300 * <td headers="matches reluc reluc_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 301 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least"><i>X</i><code>{</code><i>n</i><code>,}?</code></th> 302 * <td headers="matches reluc reluc_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 303 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}?</code></th> 304 * <td headers="matches reluc reluc_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 305 * 306 * <tr><th colspan="2" style="padding-top:20px" id="poss">Possessive quantifiers</th></tr> 307 * 308 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_once_or_not"><i>X</i>{@code ?+}</th> 309 * <td headers="matches poss poss_once_or_not"><i>X</i>, once or not at all</td></tr> 310 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_zero_or_more"><i>X</i>{@code *+}</th> 311 * <td headers="matches poss poss_zero_or_more"><i>X</i>, zero or more times</td></tr> 312 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_one_or_more"><i>X</i>{@code ++}</th> 313 * <td headers="matches poss poss_one_or_more"><i>X</i>, one or more times</td></tr> 314 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_exactly"><i>X</i><code>{</code><i>n</i><code>}+</code></th> 315 * <td headers="matches poss poss_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 316 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least"><i>X</i><code>{</code><i>n</i><code>,}+</code></th> 317 * <td headers="matches poss poss_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 318 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}+</code></th> 319 * <td headers="matches poss poss_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 320 * 321 * <tr><th colspan="2" style="padding-top:20px" id="logical">Logical operators</th></tr> 322 * 323 * <tr><th style="vertical-align:top; font-weight:normal" id="concat"><i>XY</i></th> 324 * <td headers="matches logical concat"><i>X</i> followed by <i>Y</i></td></tr> 325 * <tr><th style="vertical-align:top; font-weight:normal" id="alternate"><i>X</i>{@code |}<i>Y</i></th> 326 * <td headers="matches logical alternate">Either <i>X</i> or <i>Y</i></td></tr> 327 * <tr><th style="vertical-align:top; font-weight:normal" id="group">{@code (}<i>X</i>{@code )}</th> 328 * <td headers="matches logical group">X, as a <a href="#cg">capturing group</a></td></tr> 329 * 330 * <tr><th colspan="2" style="padding-top:20px" id="backref">Back references</th></tr> 331 * 332 * <tr><th style="vertical-align:top; font-weight:normal" id="back_nth">{@code \}<i>n</i></th> 333 * <td headers="matches backref back_nth">Whatever the <i>n</i><sup>th</sup> 334 * <a href="#cg">capturing group</a> matched</td></tr> 335 * <tr><th style="vertical-align:top; font-weight:normal" id="back_named">{@code \}<i>k</i><<i>name</i>></th> 336 * <td headers="matches backref back_named">Whatever the 337 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr> 338 * 339 * <tr><th colspan="2" style="padding-top:20px" id="quote">Quotation</th></tr> 340 * 341 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_follow">{@code \}</th> 342 * <td headers="matches quote quote_follow">Nothing, but quotes the following character</td></tr> 343 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_begin">{@code \Q}</th> 344 * <td headers="matches quote quote_begin">Nothing, but quotes all characters until {@code \E}</td></tr> 345 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_end">{@code \E}</th> 346 * <td headers="matches quote quote_end">Nothing, but ends quoting started by {@code \Q}</td></tr> 347 * <!-- Metachars: !$()*+.<>?[\]^{|} --> 348 * 349 * <tr><th colspan="2" style="padding-top:20px" id="special">Special constructs (named-capturing and non-capturing)</th></tr> 350 * 351 * <tr><th style="vertical-align:top; font-weight:normal" id="named_group"><code>(?<<a href="#groupname">name</a>></code><i>X</i>{@code )}</th> 352 * <td headers="matches special named_group"><i>X</i>, as a named-capturing group</td></tr> 353 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group">{@code (?:}<i>X</i>{@code )}</th> 354 * <td headers="matches special non_capture_group"><i>X</i>, as a non-capturing group</td></tr> 355 * <tr><th style="vertical-align:top; font-weight:normal" id="flags"><code>(?idmsuxU-idmsuxU) </code></th> 356 * <td headers="matches special flags">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a> 357 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> 358 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a> 359 * on - off</td></tr> 360 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group_flags"><code>(?idmsux-idmsux:</code><i>X</i>{@code )} </th> 361 * <td headers="matches special non_capture_group_flags"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the 362 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a> 363 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a > 364 * <a href="#COMMENTS">x</a> 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. 544 * <p> 545 * <b>Unicode escape sequences</b> such as <code>\u2014</code> in Java source code 546 * are processed as described in section 3.3 of 547 * <cite>The Java™ Language Specification</cite>. 548 * Such escape sequences are also implemented directly by the regular-expression 549 * parser so that Unicode escapes can be used in expressions that are read from 550 * files or from the keyboard. Thus the strings <code>"\u2014"</code> and 551 * {@code "\\u2014"}, while not equal, compile into the same pattern, which 552 * matches the character with hexadecimal value {@code 0x2014}. 553 * <p> 554 * A Unicode character can also be represented by using its <b>Hex notation</b> 555 * (hexadecimal code point value) directly as described in construct 556 * <code>\x{...}</code>, for example a supplementary character U+2011F can be 557 * specified as <code>\x{2011F}</code>, instead of two consecutive Unicode escape 558 * sequences of the surrogate pair <code>\uD840</code><code>\uDD1F</code>. 559 * <p> 560 * <b>Unicode character names</b> are supported by the named character construct 561 * <code>\N{</code>...<code>}</code>, for example, <code>\N{WHITE SMILING FACE}</code> 562 * specifies character <code>\u263A</code>. The character names supported 563 * by this class are the valid Unicode character names matched by 564 * {@link java.lang.Character#codePointOf(String) Character.codePointOf(name)}. 565 * <p> 566 * <a href="http://www.unicode.org/reports/tr18/#Default_Grapheme_Clusters"> 567 * <b>Unicode extended grapheme clusters</b></a> are supported by the grapheme 568 * cluster matcher {@code \X} and the corresponding boundary matcher {@code \b{g}}. 569 * <p> 570 * Unicode scripts, blocks, categories and binary properties are written with 571 * the {@code \p} and {@code \P} constructs as in Perl. 572 * <code>\p{</code><i>prop</i><code>}</code> matches if 573 * the input has the property <i>prop</i>, while <code>\P{</code><i>prop</i><code>}</code> 574 * does not match if the input has that property. 575 * <p> 576 * Scripts, blocks, categories and binary properties can be used both inside 577 * and outside of a character class. 578 * 579 * <p> 580 * <b><a id="usc">Scripts</a></b> are specified either with the prefix {@code Is}, as in 581 * {@code IsHiragana}, or by using the {@code script} keyword (or its short 582 * form {@code sc}) as in {@code script=Hiragana} or {@code sc=Hiragana}. 583 * <p> 584 * The script names supported by {@code Pattern} are the valid script names 585 * accepted and defined by 586 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}. 587 * 588 * <p> 589 * <b><a id="ubc">Blocks</a></b> are specified with the prefix {@code In}, as in 590 * {@code InMongolian}, or by using the keyword {@code block} (or its short 591 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}. 592 * <p> 593 * The block names supported by {@code Pattern} are the valid block names 594 * accepted and defined by 595 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}. 596 * <p> 597 * 598 * <b><a id="ucc">Categories</a></b> may be specified with the optional prefix {@code Is}: 599 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode 600 * letters. Same as scripts and blocks, categories can also be specified 601 * by using the keyword {@code general_category} (or its short form 602 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}. 603 * <p> 604 * The supported categories are those of 605 * <a href="http://www.unicode.org/unicode/standard/standard.html"> 606 * <i>The Unicode Standard</i></a> in the version specified by the 607 * {@link java.lang.Character Character} class. The category names are those 608 * defined in the Standard, both normative and informative. 609 * <p> 610 * 611 * <b><a id="ubpc">Binary properties</a></b> are specified with the prefix {@code Is}, as in 612 * {@code IsAlphabetic}. The supported binary properties by {@code Pattern} 613 * are 614 * <ul> 615 * <li> Alphabetic 616 * <li> Ideographic 617 * <li> Letter 618 * <li> Lowercase 619 * <li> Uppercase 620 * <li> Titlecase 621 * <li> Punctuation 622 * <Li> Control 623 * <li> White_Space 624 * <li> Digit 625 * <li> Hex_Digit 626 * <li> Join_Control 627 * <li> Noncharacter_Code_Point 628 * <li> Assigned 629 * </ul> 630 * <p> 631 * The following <b>Predefined Character classes</b> and <b>POSIX character classes</b> 632 * are in conformance with the recommendation of <i>Annex C: Compatibility Properties</i> 633 * of <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Regular Expression 634 * </i></a>, when {@link #UNICODE_CHARACTER_CLASS} flag is specified. 635 * 636 * <table class="striped"> 637 * <caption style="display:none">predefined and posix character classes in Unicode mode</caption> 638 * <thead> 639 * <tr> 640 * <th scope="col" id="predef_classes">Classes</th> 641 * <th scope="col" id="predef_matches">Matches</th> 642 * </tr> 643 * </thead> 644 * <tbody> 645 * <tr><th scope="row">{@code \p{Lower}}</th> 646 * <td>A lowercase character:{@code \p{IsLowercase}}</td></tr> 647 * <tr><th scope="row">{@code \p{Upper}}</th> 648 * <td>An uppercase character:{@code \p{IsUppercase}}</td></tr> 649 * <tr><th scope="row">{@code \p{ASCII}}</th> 650 * <td>All ASCII:{@code [\x00-\x7F]}</td></tr> 651 * <tr><th scope="row">{@code \p{Alpha}}</th> 652 * <td>An alphabetic character:{@code \p{IsAlphabetic}}</td></tr> 653 * <tr><th scope="row">{@code \p{Digit}}</th> 654 * <td>A decimal digit character:{@code \p{IsDigit}}</td></tr> 655 * <tr><th scope="row">{@code \p{Alnum}}</th> 656 * <td>An alphanumeric character:{@code [\p{IsAlphabetic}\p{IsDigit}]}</td></tr> 657 * <tr><th scope="row">{@code \p{Punct}}</th> 658 * <td>A punctuation character:{@code \p{IsPunctuation}}</td></tr> 659 * <tr><th scope="row">{@code \p{Graph}}</th> 660 * <td>A visible character: {@code [^\p{IsWhite_Space}\p{gc=Cc}\p{gc=Cs}\p{gc=Cn}]}</td></tr> 661 * <tr><th scope="row">{@code \p{Print}}</th> 662 * <td>A printable character: {@code [\p{Graph}\p{Blank}&&[^\p{Cntrl}]]}</td></tr> 663 * <tr><th scope="row">{@code \p{Blank}}</th> 664 * <td>A space or a tab: {@code [\p{IsWhite_Space}&&[^\p{gc=Zl}\p{gc=Zp}\x0a\x0b\x0c\x0d\x85]]}</td></tr> 665 * <tr><th scope="row">{@code \p{Cntrl}}</th> 666 * <td>A control character: {@code \p{gc=Cc}}</td></tr> 667 * <tr><th scope="row">{@code \p{XDigit}}</th> 668 * <td>A hexadecimal digit: {@code [\p{gc=Nd}\p{IsHex_Digit}]}</td></tr> 669 * <tr><th scope="row">{@code \p{Space}}</th> 670 * <td>A whitespace character:{@code \p{IsWhite_Space}}</td></tr> 671 * <tr><th scope="row">{@code \d}</th> 672 * <td>A digit: {@code \p{IsDigit}}</td></tr> 673 * <tr><th scope="row">{@code \D}</th> 674 * <td>A non-digit: {@code [^\d]}</td></tr> 675 * <tr><th scope="row">{@code \s}</th> 676 * <td>A whitespace character: {@code \p{IsWhite_Space}}</td></tr> 677 * <tr><th scope="row">{@code \S}</th> 678 * <td>A non-whitespace character: {@code [^\s]}</td></tr> 679 * <tr><th scope="row">{@code \w}</th> 680 * <td>A word character: {@code [\p{Alpha}\p{gc=Mn}\p{gc=Me}\p{gc=Mc}\p{Digit}\p{gc=Pc}\p{IsJoin_Control}]}</td></tr> 681 * <tr><th scope="row">{@code \W}</th> 682 * <td>A non-word character: {@code [^\w]}</td></tr> 683 * </tbody> 684 * </table> 685 * <p> 686 * <a id="jcc"> 687 * Categories that behave like the java.lang.Character 688 * boolean is<i>methodname</i> methods (except for the deprecated ones) are 689 * available through the same <code>\p{</code><i>prop</i><code>}</code> syntax where 690 * the specified property has the name <code>java<i>methodname</i></code></a>. 691 * 692 * <h2> Comparison to Perl 5 </h2> 693 * 694 * <p>The {@code Pattern} engine performs traditional NFA-based matching 695 * with ordered alternation as occurs in Perl 5. 696 * 697 * <p> Perl constructs not supported by this class: </p> 698 * 699 * <ul> 700 * <li><p> The backreference constructs, <code>\g{</code><i>n</i><code>}</code> for 701 * the <i>n</i><sup>th</sup><a href="#cg">capturing group</a> and 702 * <code>\g{</code><i>name</i><code>}</code> for 703 * <a href="#groupname">named-capturing group</a>. 704 * </p></li> 705 * 706 * <li><p> The conditional constructs 707 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code )} and 708 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code |}<i>Y</i>{@code )}, 709 * </p></li> 710 * 711 * <li><p> The embedded code constructs <code>(?{</code><i>code</i><code>})</code> 712 * and <code>(??{</code><i>code</i><code>})</code>,</p></li> 713 * 714 * <li><p> The embedded comment syntax {@code (?#comment)}, and </p></li> 715 * 716 * <li><p> The preprocessing operations {@code \l} <code>\u</code>, 717 * {@code \L}, and {@code \U}. </p></li> 718 * 719 * </ul> 720 * 721 * <p> Constructs supported by this class but not by Perl: </p> 722 * 723 * <ul> 724 * 725 * <li><p> Character-class union and intersection as described 726 * <a href="#cc">above</a>.</p></li> 727 * 728 * </ul> 729 * 730 * <p> Notable differences from Perl: </p> 731 * 732 * <ul> 733 * 734 * <li><p> In Perl, {@code \1} through {@code \9} are always interpreted 735 * as back references; a backslash-escaped number greater than {@code 9} is 736 * treated as a back reference if at least that many subexpressions exist, 737 * otherwise it is interpreted, if possible, as an octal escape. In this 738 * class octal escapes must always begin with a zero. In this class, 739 * {@code \1} through {@code \9} are always interpreted as back 740 * references, and a larger number is accepted as a back reference if at 741 * least that many subexpressions exist at that point in the regular 742 * expression, otherwise the parser will drop digits until the number is 743 * smaller or equal to the existing number of groups or it is one digit. 744 * </p></li> 745 * 746 * <li><p> Perl uses the {@code g} flag to request a match that resumes 747 * where the last match left off. This functionality is provided implicitly 748 * by the {@link Matcher} class: Repeated invocations of the {@link 749 * Matcher#find find} method will resume where the last match left off, 750 * unless the matcher is reset. </p></li> 751 * 752 * <li><p> In Perl, embedded flags at the top level of an expression affect 753 * the whole expression. In this class, embedded flags always take effect 754 * at the point at which they appear, whether they are at the top level or 755 * within a group; in the latter case, flags are restored at the end of the 756 * group just as in Perl. </p></li> 757 * 758 * </ul> 759 * 760 * 761 * <p> For a more precise description of the behavior of regular expression 762 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/"> 763 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl, 764 * O'Reilly and Associates, 2006.</a> 765 * </p> 766 * 767 * @see java.lang.String#split(String, int) 768 * @see java.lang.String#split(String) 769 * 770 * @author Mike McCloskey 771 * @author Mark Reinhold 772 * @author JSR-51 Expert Group 773 * @since 1.4 774 * @spec JSR-51 775 */ 776 777 public final class Pattern 778 implements java.io.Serializable 779 { 780 781 /** 782 * Regular expression modifier values. Instead of being passed as 783 * arguments, they can also be passed as inline modifiers. 784 * For example, the following statements have the same effect. 785 * <pre> 786 * Pattern p1 = Pattern.compile("abc", Pattern.CASE_INSENSITIVE|Pattern.MULTILINE); 787 * Pattern p2 = Pattern.compile("(?im)abc", 0); 788 * </pre> 789 */ 790 791 /** 792 * Enables Unix lines mode. 793 * 794 * <p> In this mode, only the {@code '\n'} line terminator is recognized 795 * in the behavior of {@code .}, {@code ^}, and {@code $}. 796 * 797 * <p> Unix lines mode can also be enabled via the embedded flag 798 * expression {@code (?d)}. 799 */ 800 public static final int UNIX_LINES = 0x01; 801 802 /** 803 * Enables case-insensitive matching. 804 * 805 * <p> By default, case-insensitive matching assumes that only characters 806 * in the US-ASCII charset are being matched. Unicode-aware 807 * case-insensitive matching can be enabled by specifying the {@link 808 * #UNICODE_CASE} flag in conjunction with this flag. 809 * 810 * <p> Case-insensitive matching can also be enabled via the embedded flag 811 * expression {@code (?i)}. 812 * 813 * <p> Specifying this flag may impose a slight performance penalty. </p> 814 */ 815 public static final int CASE_INSENSITIVE = 0x02; 816 817 /** 818 * Permits whitespace and comments in pattern. 819 * 820 * <p> In this mode, whitespace is ignored, and embedded comments starting 821 * with {@code #} are ignored until the end of a line. 822 * 823 * <p> Comments mode can also be enabled via the embedded flag 824 * expression {@code (?x)}. 825 */ 826 public static final int COMMENTS = 0x04; 827 828 /** 829 * Enables multiline mode. 830 * 831 * <p> In multiline mode the expressions {@code ^} and {@code $} match 832 * just after or just before, respectively, a line terminator or the end of 833 * the input sequence. By default these expressions only match at the 834 * beginning and the end of the entire input sequence. 835 * 836 * <p> Multiline mode can also be enabled via the embedded flag 837 * expression {@code (?m)}. </p> 838 */ 839 public static final int MULTILINE = 0x08; 840 841 /** 842 * Enables literal parsing of the pattern. 843 * 844 * <p> When this flag is specified then the input string that specifies 845 * the pattern is treated as a sequence of literal characters. 846 * Metacharacters or escape sequences in the input sequence will be 847 * given no special meaning. 848 * 849 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on 850 * matching when used in conjunction with this flag. The other flags 851 * become superfluous. 852 * 853 * <p> There is no embedded flag character for enabling literal parsing. 854 * @since 1.5 855 */ 856 public static final int LITERAL = 0x10; 857 858 /** 859 * Enables dotall mode. 860 * 861 * <p> In dotall mode, the expression {@code .} matches any character, 862 * including a line terminator. By default this expression does not match 863 * line terminators. 864 * 865 * <p> Dotall mode can also be enabled via the embedded flag 866 * expression {@code (?s)}. (The {@code s} is a mnemonic for 867 * "single-line" mode, which is what this is called in Perl.) </p> 868 */ 869 public static final int DOTALL = 0x20; 870 871 /** 872 * Enables Unicode-aware case folding. 873 * 874 * <p> When this flag is specified then case-insensitive matching, when 875 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner 876 * consistent with the Unicode Standard. By default, case-insensitive 877 * matching assumes that only characters in the US-ASCII charset are being 878 * matched. 879 * 880 * <p> Unicode-aware case folding can also be enabled via the embedded flag 881 * expression {@code (?u)}. 882 * 883 * <p> Specifying this flag may impose a performance penalty. </p> 884 */ 885 public static final int UNICODE_CASE = 0x40; 886 887 /** 888 * Enables canonical equivalence. 889 * 890 * <p> When this flag is specified then two characters will be considered 891 * to match if, and only if, their full canonical decompositions match. 892 * The expression <code>"a\u030A"</code>, for example, will match the 893 * string <code>"\u00E5"</code> when this flag is specified. By default, 894 * matching does not take canonical equivalence into account. 895 * 896 * <p> There is no embedded flag character for enabling canonical 897 * equivalence. 898 * 899 * <p> Specifying this flag may impose a performance penalty. </p> 900 */ 901 public static final int CANON_EQ = 0x80; 902 903 /** 904 * Enables the Unicode version of <i>Predefined character classes</i> and 905 * <i>POSIX character classes</i>. 906 * 907 * <p> When this flag is specified then the (US-ASCII only) 908 * <i>Predefined character classes</i> and <i>POSIX character classes</i> 909 * are in conformance with 910 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 911 * Standard #18: Unicode Regular Expression</i></a> 912 * <i>Annex C: Compatibility Properties</i>. 913 * <p> 914 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded 915 * flag expression {@code (?U)}. 916 * <p> 917 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case 918 * folding. 919 * <p> 920 * Specifying this flag may impose a performance penalty. </p> 921 * @since 1.7 922 */ 923 public static final int UNICODE_CHARACTER_CLASS = 0x100; 924 925 /** 926 * Contains all possible flags for compile(regex, flags). 927 */ 928 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE | 929 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL | 930 UNICODE_CHARACTER_CLASS | COMMENTS; 931 932 /* Pattern has only two serialized components: The pattern string 933 * and the flags, which are all that is needed to recompile the pattern 934 * when it is deserialized. 935 */ 936 937 /** use serialVersionUID from Merlin b59 for interoperability */ 938 private static final long serialVersionUID = 5073258162644648461L; 939 940 /** 941 * The original regular-expression pattern string. 942 * 943 * @serial 944 */ 945 private String pattern; 946 947 /** 948 * The original pattern flags. 949 * 950 * @serial 951 */ 952 private int flags; 953 954 /** 955 * The temporary pattern flags used during compiling. The flags might be turn 956 * on and off by embedded flag. 957 */ 958 private transient int flags0; 959 960 /** 961 * Boolean indicating this Pattern is compiled; this is necessary in order 962 * to lazily compile deserialized Patterns. 963 */ 964 private transient volatile boolean compiled; 965 966 /** 967 * The normalized pattern string. 968 */ 969 private transient String normalizedPattern; 970 971 /** 972 * The starting point of state machine for the find operation. This allows 973 * a match to start anywhere in the input. 974 */ 975 transient Node root; 976 977 /** 978 * The root of object tree for a match operation. The pattern is matched 979 * at the beginning. This may include a find that uses BnM or a First 980 * node. 981 */ 982 transient Node matchRoot; 983 984 /** 985 * Temporary storage used by parsing pattern slice. 986 */ 987 transient int[] buffer; 988 989 /** 990 * A temporary storage used for predicate for double return. 991 */ 992 transient CharPredicate predicate; 993 994 /** 995 * Map the "name" of the "named capturing group" to its group id 996 * node. 997 */ 998 transient volatile Map<String, Integer> namedGroups; 999 1000 /** 1001 * Temporary storage used while parsing group references. 1002 */ 1003 transient GroupHead[] groupNodes; 1004 1005 /** 1006 * Temporary storage used to store the top level closure nodes. 1007 */ 1008 transient List<Node> topClosureNodes; 1009 1010 /** 1011 * The number of top greedy closure nodes in this Pattern. Used by 1012 * matchers to allocate storage needed for a IntHashSet to keep the 1013 * beginning pos {@code i} of all failed match. 1014 */ 1015 transient int localTCNCount; 1016 1017 /* 1018 * Turn off the stop-exponential-backtracking optimization if there 1019 * is a group ref in the pattern. 1020 */ 1021 transient boolean hasGroupRef; 1022 1023 /** 1024 * Temporary null terminated code point array used by pattern compiling. 1025 */ 1026 private transient int[] temp; 1027 1028 /** 1029 * The number of capturing groups in this Pattern. Used by matchers to 1030 * allocate storage needed to perform a match. 1031 */ 1032 transient int capturingGroupCount; 1033 1034 /** 1035 * The local variable count used by parsing tree. Used by matchers to 1036 * allocate storage needed to perform a match. 1037 */ 1038 transient int localCount; 1039 1040 /** 1041 * Index into the pattern string that keeps track of how much has been 1042 * parsed. 1043 */ 1044 private transient int cursor; 1045 1046 /** 1047 * Holds the length of the pattern string. 1048 */ 1049 private transient int patternLength; 1050 1051 /** 1052 * If the Start node might possibly match supplementary characters. 1053 * It is set to true during compiling if 1054 * (1) There is supplementary char in pattern, or 1055 * (2) There is complement node of a "family" CharProperty 1056 */ 1057 private transient boolean hasSupplementary; 1058 1059 /** 1060 * Compiles the given regular expression into a pattern. 1061 * 1062 * @param regex 1063 * The expression to be compiled 1064 * @return the given regular expression compiled into a pattern 1065 * @throws PatternSyntaxException 1066 * If the expression's syntax is invalid 1067 */ 1068 public static Pattern compile(String regex) { 1069 return new Pattern(regex, 0); 1070 } 1071 1072 /** 1073 * Compiles the given regular expression into a pattern with the given 1074 * flags. 1075 * 1076 * @param regex 1077 * The expression to be compiled 1078 * 1079 * @param flags 1080 * Match flags, a bit mask that may include 1081 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL}, 1082 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES}, 1083 * {@link #LITERAL}, {@link #UNICODE_CHARACTER_CLASS} 1084 * and {@link #COMMENTS} 1085 * 1086 * @return the given regular expression compiled into a pattern with the given flags 1087 * @throws IllegalArgumentException 1088 * If bit values other than those corresponding to the defined 1089 * match flags are set in {@code flags} 1090 * 1091 * @throws PatternSyntaxException 1092 * If the expression's syntax is invalid 1093 */ 1094 public static Pattern compile(String regex, int flags) { 1095 return new Pattern(regex, flags); 1096 } 1097 1098 /** 1099 * Returns the regular expression from which this pattern was compiled. 1100 * 1101 * @return The source of this pattern 1102 */ 1103 public String pattern() { 1104 return pattern; 1105 } 1106 1107 /** 1108 * <p>Returns the string representation of this pattern. This 1109 * is the regular expression from which this pattern was 1110 * compiled.</p> 1111 * 1112 * @return The string representation of this pattern 1113 * @since 1.5 1114 */ 1115 public String toString() { 1116 return pattern; 1117 } 1118 1119 /** 1120 * Creates a matcher that will match the given input against this pattern. 1121 * 1122 * @param input 1123 * The character sequence to be matched 1124 * 1125 * @return A new matcher for this pattern 1126 */ 1127 public Matcher matcher(CharSequence input) { 1128 if (!compiled) { 1129 synchronized(this) { 1130 if (!compiled) 1131 compile(); 1132 } 1133 } 1134 Matcher m = new Matcher(this, input); 1135 return m; 1136 } 1137 1138 /** 1139 * Returns this pattern's match flags. 1140 * 1141 * @return The match flags specified when this pattern was compiled 1142 */ 1143 public int flags() { 1144 return flags0; 1145 } 1146 1147 /** 1148 * Compiles the given regular expression and attempts to match the given 1149 * input against it. 1150 * 1151 * <p> An invocation of this convenience method of the form 1152 * 1153 * <blockquote><pre> 1154 * Pattern.matches(regex, input);</pre></blockquote> 1155 * 1156 * behaves in exactly the same way as the expression 1157 * 1158 * <blockquote><pre> 1159 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote> 1160 * 1161 * <p> If a pattern is to be used multiple times, compiling it once and reusing 1162 * it will be more efficient than invoking this method each time. </p> 1163 * 1164 * @param regex 1165 * The expression to be compiled 1166 * 1167 * @param input 1168 * The character sequence to be matched 1169 * @return whether or not the regular expression matches on the input 1170 * @throws PatternSyntaxException 1171 * If the expression's syntax is invalid 1172 */ 1173 public static boolean matches(String regex, CharSequence input) { 1174 Pattern p = Pattern.compile(regex); 1175 Matcher m = p.matcher(input); 1176 return m.matches(); 1177 } 1178 1179 /** 1180 * Splits the given input sequence around matches of this pattern. 1181 * 1182 * <p> The array returned by this method contains each substring of the 1183 * input sequence that is terminated by another subsequence that matches 1184 * this pattern or is terminated by the end of the input sequence. The 1185 * substrings in the array are in the order in which they occur in the 1186 * input. If this pattern does not match any subsequence of the input then 1187 * the resulting array has just one element, namely the input sequence in 1188 * string form. 1189 * 1190 * <p> When there is a positive-width match at the beginning of the input 1191 * sequence then an empty leading substring is included at the beginning 1192 * of the resulting array. A zero-width match at the beginning however 1193 * never produces such empty leading substring. 1194 * 1195 * <p> The {@code limit} parameter controls the number of times the 1196 * pattern is applied and therefore affects the length of the resulting 1197 * array. 1198 * <ul> 1199 * <li><p> 1200 * If the <i>limit</i> is positive then the pattern will be applied 1201 * at most <i>limit</i> - 1 times, the array's length will be 1202 * no greater than <i>limit</i>, and the array's last entry will contain 1203 * all input beyond the last matched delimiter.</p></li> 1204 * 1205 * <li><p> 1206 * If the <i>limit</i> is zero then the pattern will be applied as 1207 * many times as possible, the array can have any length, and trailing 1208 * empty strings will be discarded.</p></li> 1209 * 1210 * <li><p> 1211 * If the <i>limit</i> is negative then the pattern will be applied 1212 * as many times as possible and the array can have any length.</p></li> 1213 * </ul> 1214 * 1215 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1216 * results with these parameters: 1217 * 1218 * <table class="plain" style="margin-left:2em;"> 1219 * <caption style="display:none">Split example showing regex, limit, and result</caption> 1220 * <thead> 1221 * <tr> 1222 * <th scope="col">Regex</th> 1223 * <th scope="col">Limit</th> 1224 * <th scope="col">Result</th> 1225 * </tr> 1226 * </thead> 1227 * <tbody> 1228 * <tr><th scope="row" rowspan="3" style="font-weight:normal">:</th> 1229 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">2</th> 1230 * <td>{@code { "boo", "and:foo" }}</td></tr> 1231 * <tr><!-- : --> 1232 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1233 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1234 * <tr><!-- : --> 1235 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1236 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1237 * <tr><th scope="row" rowspan="3" style="font-weight:normal">o</th> 1238 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1239 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1240 * <tr><!-- o --> 1241 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1242 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1243 * <tr><!-- o --> 1244 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">0</th> 1245 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1246 * </tbody> 1247 * </table> 1248 * 1249 * @param input 1250 * The character sequence to be split 1251 * 1252 * @param limit 1253 * The result threshold, as described above 1254 * 1255 * @return The array of strings computed by splitting the input 1256 * around matches of this pattern 1257 */ 1258 public String[] split(CharSequence input, int limit) { 1259 int index = 0; 1260 boolean matchLimited = limit > 0; 1261 ArrayList<String> matchList = new ArrayList<>(); 1262 Matcher m = matcher(input); 1263 1264 // Add segments before each match found 1265 while(m.find()) { 1266 if (!matchLimited || matchList.size() < limit - 1) { 1267 if (index == 0 && index == m.start() && m.start() == m.end()) { 1268 // no empty leading substring included for zero-width match 1269 // at the beginning of the input char sequence. 1270 continue; 1271 } 1272 String match = input.subSequence(index, m.start()).toString(); 1273 matchList.add(match); 1274 index = m.end(); 1275 } else if (matchList.size() == limit - 1) { // last one 1276 String match = input.subSequence(index, 1277 input.length()).toString(); 1278 matchList.add(match); 1279 index = m.end(); 1280 } 1281 } 1282 1283 // If no match was found, return this 1284 if (index == 0) 1285 return new String[] {input.toString()}; 1286 1287 // Add remaining segment 1288 if (!matchLimited || matchList.size() < limit) 1289 matchList.add(input.subSequence(index, input.length()).toString()); 1290 1291 // Construct result 1292 int resultSize = matchList.size(); 1293 if (limit == 0) 1294 while (resultSize > 0 && matchList.get(resultSize-1).isEmpty()) 1295 resultSize--; 1296 String[] result = new String[resultSize]; 1297 return matchList.subList(0, resultSize).toArray(result); 1298 } 1299 1300 /** 1301 * Splits the given input sequence around matches of this pattern. 1302 * 1303 * <p> This method works as if by invoking the two-argument {@link 1304 * #split(java.lang.CharSequence, int) split} method with the given input 1305 * sequence and a limit argument of zero. Trailing empty strings are 1306 * therefore not included in the resulting array. </p> 1307 * 1308 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1309 * results with these expressions: 1310 * 1311 * <table class="plain" style="margin-left:2em"> 1312 * <caption style="display:none">Split examples showing regex and result</caption> 1313 * <thead> 1314 * <tr> 1315 * <th scope="col">Regex</th> 1316 * <th scope="col">Result</th> 1317 * </tr> 1318 * </thead> 1319 * <tbody> 1320 * <tr><th scope="row" style="text-weight:normal">:</th> 1321 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1322 * <tr><th scope="row" style="text-weight:normal">o</th> 1323 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1324 * </tbody> 1325 * </table> 1326 * 1327 * 1328 * @param input 1329 * The character sequence to be split 1330 * 1331 * @return The array of strings computed by splitting the input 1332 * around matches of this pattern 1333 */ 1334 public String[] split(CharSequence input) { 1335 return split(input, 0); 1336 } 1337 1338 /** 1339 * Returns a literal pattern {@code String} for the specified 1340 * {@code String}. 1341 * 1342 * <p>This method produces a {@code String} that can be used to 1343 * create a {@code Pattern} that would match the string 1344 * {@code s} as if it were a literal pattern.</p> Metacharacters 1345 * or escape sequences in the input sequence will be given no special 1346 * meaning. 1347 * 1348 * @param s The string to be literalized 1349 * @return A literal string replacement 1350 * @since 1.5 1351 */ 1352 public static String quote(String s) { 1353 int slashEIndex = s.indexOf("\\E"); 1354 if (slashEIndex == -1) 1355 return "\\Q" + s + "\\E"; 1356 1357 int lenHint = s.length(); 1358 lenHint = (lenHint < Integer.MAX_VALUE - 8 - lenHint) ? 1359 (lenHint << 1) : (Integer.MAX_VALUE - 8); 1360 1361 StringBuilder sb = new StringBuilder(lenHint); 1362 sb.append("\\Q"); 1363 int current = 0; 1364 do { 1365 sb.append(s, current, slashEIndex) 1366 .append("\\E\\\\E\\Q"); 1367 current = slashEIndex + 2; 1368 } while ((slashEIndex = s.indexOf("\\E", current)) != -1); 1369 1370 return sb.append(s, current, s.length()) 1371 .append("\\E") 1372 .toString(); 1373 } 1374 1375 /** 1376 * Recompile the Pattern instance from a stream. The original pattern 1377 * string is read in and the object tree is recompiled from it. 1378 */ 1379 private void readObject(java.io.ObjectInputStream s) 1380 throws java.io.IOException, ClassNotFoundException { 1381 1382 // Read in all fields 1383 s.defaultReadObject(); 1384 1385 // reset the flags 1386 flags0 = flags; 1387 1388 // Initialize counts 1389 capturingGroupCount = 1; 1390 localCount = 0; 1391 localTCNCount = 0; 1392 1393 // if length > 0, the Pattern is lazily compiled 1394 if (pattern.isEmpty()) { 1395 root = new Start(lastAccept); 1396 matchRoot = lastAccept; 1397 compiled = true; 1398 } 1399 } 1400 1401 /** 1402 * This private constructor is used to create all Patterns. The pattern 1403 * string and match flags are all that is needed to completely describe 1404 * a Pattern. An empty pattern string results in an object tree with 1405 * only a Start node and a LastNode node. 1406 */ 1407 private Pattern(String p, int f) { 1408 if ((f & ~ALL_FLAGS) != 0) { 1409 throw new IllegalArgumentException("Unknown flag 0x" 1410 + Integer.toHexString(f)); 1411 } 1412 pattern = p; 1413 flags = f; 1414 1415 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present 1416 if ((flags & UNICODE_CHARACTER_CLASS) != 0) 1417 flags |= UNICODE_CASE; 1418 1419 // 'flags' for compiling 1420 flags0 = flags; 1421 1422 // Reset group index count 1423 capturingGroupCount = 1; 1424 localCount = 0; 1425 localTCNCount = 0; 1426 1427 if (!pattern.isEmpty()) { 1428 compile(); 1429 } else { 1430 root = new Start(lastAccept); 1431 matchRoot = lastAccept; 1432 } 1433 } 1434 1435 /** 1436 * The pattern is converted to normalized form ({@link 1437 * java.text.Normalizer.Form#NFC NFC}, canonical decomposition, 1438 * followed by canonical composition for the character class 1439 * part, and {@link java.text.Normalizer.Form#NFD NFD}, 1440 * canonical decomposition for the rest), and then a pure 1441 * group is constructed to match canonical equivalences of the 1442 * characters. 1443 */ 1444 private static String normalize(String pattern) { 1445 int plen = pattern.length(); 1446 StringBuilder pbuf = new StringBuilder(plen); 1447 char last = 0; 1448 int lastStart = 0; 1449 char cc = 0; 1450 for (int i = 0; i < plen;) { 1451 char c = pattern.charAt(i); 1452 if (cc == 0 && // top level 1453 c == '\\' && i + 1 < plen && pattern.charAt(i + 1) == '\\') { 1454 i += 2; last = 0; 1455 continue; 1456 } 1457 if (c == '[' && last != '\\') { 1458 if (cc == 0) { 1459 if (lastStart < i) 1460 normalizeSlice(pattern, lastStart, i, pbuf); 1461 lastStart = i; 1462 } 1463 cc++; 1464 } else if (c == ']' && last != '\\') { 1465 cc--; 1466 if (cc == 0) { 1467 normalizeClazz(pattern, lastStart, i + 1, pbuf); 1468 lastStart = i + 1; 1469 } 1470 } 1471 last = c; 1472 i++; 1473 } 1474 assert (cc == 0); 1475 if (lastStart < plen) 1476 normalizeSlice(pattern, lastStart, plen, pbuf); 1477 return pbuf.toString(); 1478 } 1479 1480 private static void normalizeSlice(String src, int off, int limit, 1481 StringBuilder dst) 1482 { 1483 int len = src.length(); 1484 int off0 = off; 1485 while (off < limit && ASCII.isAscii(src.charAt(off))) { 1486 off++; 1487 } 1488 if (off == limit) { 1489 dst.append(src, off0, limit); 1490 return; 1491 } 1492 off--; 1493 if (off < off0) 1494 off = off0; 1495 else 1496 dst.append(src, off0, off); 1497 while (off < limit) { 1498 int ch0 = src.codePointAt(off); 1499 if (".$|()[]{}^?*+\\".indexOf(ch0) != -1) { 1500 dst.append((char)ch0); 1501 off++; 1502 continue; 1503 } 1504 int j = off + Character.charCount(ch0); 1505 int ch1; 1506 while (j < limit) { 1507 ch1 = src.codePointAt(j); 1508 if (Grapheme.isBoundary(ch0, ch1)) 1509 break; 1510 ch0 = ch1; 1511 j += Character.charCount(ch1); 1512 } 1513 String seq = src.substring(off, j); 1514 String nfd = Normalizer.normalize(seq, Normalizer.Form.NFD); 1515 off = j; 1516 if (nfd.length() > 1) { 1517 ch0 = nfd.codePointAt(0); 1518 ch1 = nfd.codePointAt(Character.charCount(ch0)); 1519 if (Character.getType(ch1) == Character.NON_SPACING_MARK) { 1520 Set<String> altns = new LinkedHashSet<>(); 1521 altns.add(seq); 1522 produceEquivalentAlternation(nfd, altns); 1523 dst.append("(?:"); 1524 altns.forEach( s -> dst.append(s).append('|')); 1525 dst.delete(dst.length() - 1, dst.length()); 1526 dst.append(")"); 1527 continue; 1528 } 1529 } 1530 String nfc = Normalizer.normalize(seq, Normalizer.Form.NFC); 1531 if (!seq.equals(nfc) && !nfd.equals(nfc)) 1532 dst.append("(?:" + seq + "|" + nfd + "|" + nfc + ")"); 1533 else if (!seq.equals(nfd)) 1534 dst.append("(?:" + seq + "|" + nfd + ")"); 1535 else 1536 dst.append(seq); 1537 } 1538 } 1539 1540 private static void normalizeClazz(String src, int off, int limit, 1541 StringBuilder dst) 1542 { 1543 dst.append(Normalizer.normalize(src.substring(off, limit), Form.NFC)); 1544 } 1545 1546 /** 1547 * Given a specific sequence composed of a regular character and 1548 * combining marks that follow it, produce the alternation that will 1549 * match all canonical equivalences of that sequence. 1550 */ 1551 private static void produceEquivalentAlternation(String src, 1552 Set<String> dst) 1553 { 1554 int len = countChars(src, 0, 1); 1555 if (src.length() == len) { 1556 dst.add(src); // source has one character. 1557 return; 1558 } 1559 String base = src.substring(0,len); 1560 String combiningMarks = src.substring(len); 1561 String[] perms = producePermutations(combiningMarks); 1562 // Add combined permutations 1563 for(int x = 0; x < perms.length; x++) { 1564 String next = base + perms[x]; 1565 dst.add(next); 1566 next = composeOneStep(next); 1567 if (next != null) { 1568 produceEquivalentAlternation(next, dst); 1569 } 1570 } 1571 } 1572 1573 /** 1574 * Returns an array of strings that have all the possible 1575 * permutations of the characters in the input string. 1576 * This is used to get a list of all possible orderings 1577 * of a set of combining marks. Note that some of the permutations 1578 * are invalid because of combining class collisions, and these 1579 * possibilities must be removed because they are not canonically 1580 * equivalent. 1581 */ 1582 private static String[] producePermutations(String input) { 1583 if (input.length() == countChars(input, 0, 1)) 1584 return new String[] {input}; 1585 1586 if (input.length() == countChars(input, 0, 2)) { 1587 int c0 = Character.codePointAt(input, 0); 1588 int c1 = Character.codePointAt(input, Character.charCount(c0)); 1589 if (getClass(c1) == getClass(c0)) { 1590 return new String[] {input}; 1591 } 1592 String[] result = new String[2]; 1593 result[0] = input; 1594 StringBuilder sb = new StringBuilder(2); 1595 sb.appendCodePoint(c1); 1596 sb.appendCodePoint(c0); 1597 result[1] = sb.toString(); 1598 return result; 1599 } 1600 1601 int length = 1; 1602 int nCodePoints = countCodePoints(input); 1603 for(int x=1; x<nCodePoints; x++) 1604 length = length * (x+1); 1605 1606 String[] temp = new String[length]; 1607 1608 int combClass[] = new int[nCodePoints]; 1609 for(int x=0, i=0; x<nCodePoints; x++) { 1610 int c = Character.codePointAt(input, i); 1611 combClass[x] = getClass(c); 1612 i += Character.charCount(c); 1613 } 1614 1615 // For each char, take it out and add the permutations 1616 // of the remaining chars 1617 int index = 0; 1618 int len; 1619 // offset maintains the index in code units. 1620 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) { 1621 len = countChars(input, offset, 1); 1622 for(int y=x-1; y>=0; y--) { 1623 if (combClass[y] == combClass[x]) { 1624 continue loop; 1625 } 1626 } 1627 StringBuilder sb = new StringBuilder(input); 1628 String otherChars = sb.delete(offset, offset+len).toString(); 1629 String[] subResult = producePermutations(otherChars); 1630 1631 String prefix = input.substring(offset, offset+len); 1632 for (String sre : subResult) 1633 temp[index++] = prefix + sre; 1634 } 1635 String[] result = new String[index]; 1636 System.arraycopy(temp, 0, result, 0, index); 1637 return result; 1638 } 1639 1640 private static int getClass(int c) { 1641 return sun.text.Normalizer.getCombiningClass(c); 1642 } 1643 1644 /** 1645 * Attempts to compose input by combining the first character 1646 * with the first combining mark following it. Returns a String 1647 * that is the composition of the leading character with its first 1648 * combining mark followed by the remaining combining marks. Returns 1649 * null if the first two characters cannot be further composed. 1650 */ 1651 private static String composeOneStep(String input) { 1652 int len = countChars(input, 0, 2); 1653 String firstTwoCharacters = input.substring(0, len); 1654 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC); 1655 if (result.equals(firstTwoCharacters)) 1656 return null; 1657 else { 1658 String remainder = input.substring(len); 1659 return result + remainder; 1660 } 1661 } 1662 1663 /** 1664 * Preprocess any \Q...\E sequences in `temp', meta-quoting them. 1665 * See the description of `quotemeta' in perlfunc(1). 1666 */ 1667 private void RemoveQEQuoting() { 1668 final int pLen = patternLength; 1669 int i = 0; 1670 while (i < pLen-1) { 1671 if (temp[i] != '\\') 1672 i += 1; 1673 else if (temp[i + 1] != 'Q') 1674 i += 2; 1675 else 1676 break; 1677 } 1678 if (i >= pLen - 1) // No \Q sequence found 1679 return; 1680 int j = i; 1681 i += 2; 1682 int newTempLen; 1683 try { 1684 newTempLen = Math.addExact(j + 2, Math.multiplyExact(3, pLen - i)); 1685 } catch (ArithmeticException ae) { 1686 throw new OutOfMemoryError(); 1687 } 1688 int[] newtemp = new int[newTempLen]; 1689 System.arraycopy(temp, 0, newtemp, 0, j); 1690 1691 boolean inQuote = true; 1692 boolean beginQuote = true; 1693 while (i < pLen) { 1694 int c = temp[i++]; 1695 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) { 1696 newtemp[j++] = c; 1697 } else if (ASCII.isDigit(c)) { 1698 if (beginQuote) { 1699 /* 1700 * A unicode escape \[0xu] could be before this quote, 1701 * and we don't want this numeric char to processed as 1702 * part of the escape. 1703 */ 1704 newtemp[j++] = '\\'; 1705 newtemp[j++] = 'x'; 1706 newtemp[j++] = '3'; 1707 } 1708 newtemp[j++] = c; 1709 } else if (c != '\\') { 1710 if (inQuote) newtemp[j++] = '\\'; 1711 newtemp[j++] = c; 1712 } else if (inQuote) { 1713 if (temp[i] == 'E') { 1714 i++; 1715 inQuote = false; 1716 } else { 1717 newtemp[j++] = '\\'; 1718 newtemp[j++] = '\\'; 1719 } 1720 } else { 1721 if (temp[i] == 'Q') { 1722 i++; 1723 inQuote = true; 1724 beginQuote = true; 1725 continue; 1726 } else { 1727 newtemp[j++] = c; 1728 if (i != pLen) 1729 newtemp[j++] = temp[i++]; 1730 } 1731 } 1732 1733 beginQuote = false; 1734 } 1735 1736 patternLength = j; 1737 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination 1738 } 1739 1740 /** 1741 * Copies regular expression to an int array and invokes the parsing 1742 * of the expression which will create the object tree. 1743 */ 1744 private void compile() { 1745 // Handle canonical equivalences 1746 if (has(CANON_EQ) && !has(LITERAL)) { 1747 normalizedPattern = normalize(pattern); 1748 } else { 1749 normalizedPattern = pattern; 1750 } 1751 patternLength = normalizedPattern.length(); 1752 1753 // Copy pattern to int array for convenience 1754 // Use double zero to terminate pattern 1755 temp = new int[patternLength + 2]; 1756 1757 hasSupplementary = false; 1758 int c, count = 0; 1759 // Convert all chars into code points 1760 for (int x = 0; x < patternLength; x += Character.charCount(c)) { 1761 c = normalizedPattern.codePointAt(x); 1762 if (isSupplementary(c)) { 1763 hasSupplementary = true; 1764 } 1765 temp[count++] = c; 1766 } 1767 1768 patternLength = count; // patternLength now in code points 1769 1770 if (! has(LITERAL)) 1771 RemoveQEQuoting(); 1772 1773 // Allocate all temporary objects here. 1774 buffer = new int[32]; 1775 groupNodes = new GroupHead[10]; 1776 namedGroups = null; 1777 topClosureNodes = new ArrayList<>(10); 1778 1779 if (has(LITERAL)) { 1780 // Literal pattern handling 1781 matchRoot = newSlice(temp, patternLength, hasSupplementary); 1782 matchRoot.next = lastAccept; 1783 } else { 1784 // Start recursive descent parsing 1785 matchRoot = expr(lastAccept); 1786 // Check extra pattern characters 1787 if (patternLength != cursor) { 1788 if (peek() == ')') { 1789 throw error("Unmatched closing ')'"); 1790 } else { 1791 throw error("Unexpected internal error"); 1792 } 1793 } 1794 } 1795 1796 // Peephole optimization 1797 if (matchRoot instanceof Slice) { 1798 root = BnM.optimize(matchRoot); 1799 if (root == matchRoot) { 1800 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1801 } 1802 } else if (matchRoot instanceof Begin || matchRoot instanceof First) { 1803 root = matchRoot; 1804 } else { 1805 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1806 } 1807 1808 // Optimize the greedy Loop to prevent exponential backtracking, IF there 1809 // is no group ref in this pattern. With a non-negative localTCNCount value, 1810 // the greedy type Loop, Curly will skip the backtracking for any starting 1811 // position "i" that failed in the past. 1812 if (!hasGroupRef) { 1813 for (Node node : topClosureNodes) { 1814 if (node instanceof Loop) { 1815 // non-deterministic-greedy-group 1816 ((Loop)node).posIndex = localTCNCount++; 1817 } 1818 } 1819 } 1820 1821 // Release temporary storage 1822 temp = null; 1823 buffer = null; 1824 groupNodes = null; 1825 patternLength = 0; 1826 compiled = true; 1827 topClosureNodes = null; 1828 } 1829 1830 Map<String, Integer> namedGroups() { 1831 Map<String, Integer> groups = namedGroups; 1832 if (groups == null) { 1833 namedGroups = groups = new HashMap<>(2); 1834 } 1835 return groups; 1836 } 1837 1838 /** 1839 * Used to accumulate information about a subtree of the object graph 1840 * so that optimizations can be applied to the subtree. 1841 */ 1842 static final class TreeInfo { 1843 int minLength; 1844 int maxLength; 1845 boolean maxValid; 1846 boolean deterministic; 1847 1848 TreeInfo() { 1849 reset(); 1850 } 1851 void reset() { 1852 minLength = 0; 1853 maxLength = 0; 1854 maxValid = true; 1855 deterministic = true; 1856 } 1857 } 1858 1859 /* 1860 * The following private methods are mainly used to improve the 1861 * readability of the code. In order to let the Java compiler easily 1862 * inline them, we should not put many assertions or error checks in them. 1863 */ 1864 1865 /** 1866 * Indicates whether a particular flag is set or not. 1867 */ 1868 private boolean has(int f) { 1869 return (flags0 & f) != 0; 1870 } 1871 1872 /** 1873 * Match next character, signal error if failed. 1874 */ 1875 private void accept(int ch, String s) { 1876 int testChar = temp[cursor++]; 1877 if (has(COMMENTS)) 1878 testChar = parsePastWhitespace(testChar); 1879 if (ch != testChar) { 1880 throw error(s); 1881 } 1882 } 1883 1884 /** 1885 * Mark the end of pattern with a specific character. 1886 */ 1887 private void mark(int c) { 1888 temp[patternLength] = c; 1889 } 1890 1891 /** 1892 * Peek the next character, and do not advance the cursor. 1893 */ 1894 private int peek() { 1895 int ch = temp[cursor]; 1896 if (has(COMMENTS)) 1897 ch = peekPastWhitespace(ch); 1898 return ch; 1899 } 1900 1901 /** 1902 * Read the next character, and advance the cursor by one. 1903 */ 1904 private int read() { 1905 int ch = temp[cursor++]; 1906 if (has(COMMENTS)) 1907 ch = parsePastWhitespace(ch); 1908 return ch; 1909 } 1910 1911 /** 1912 * Read the next character, and advance the cursor by one, 1913 * ignoring the COMMENTS setting 1914 */ 1915 private int readEscaped() { 1916 int ch = temp[cursor++]; 1917 return ch; 1918 } 1919 1920 /** 1921 * Advance the cursor by one, and peek the next character. 1922 */ 1923 private int next() { 1924 int ch = temp[++cursor]; 1925 if (has(COMMENTS)) 1926 ch = peekPastWhitespace(ch); 1927 return ch; 1928 } 1929 1930 /** 1931 * Advance the cursor by one, and peek the next character, 1932 * ignoring the COMMENTS setting 1933 */ 1934 private int nextEscaped() { 1935 int ch = temp[++cursor]; 1936 return ch; 1937 } 1938 1939 /** 1940 * If in xmode peek past whitespace and comments. 1941 */ 1942 private int peekPastWhitespace(int ch) { 1943 while (ASCII.isSpace(ch) || ch == '#') { 1944 while (ASCII.isSpace(ch)) 1945 ch = temp[++cursor]; 1946 if (ch == '#') { 1947 ch = peekPastLine(); 1948 } 1949 } 1950 return ch; 1951 } 1952 1953 /** 1954 * If in xmode parse past whitespace and comments. 1955 */ 1956 private int parsePastWhitespace(int ch) { 1957 while (ASCII.isSpace(ch) || ch == '#') { 1958 while (ASCII.isSpace(ch)) 1959 ch = temp[cursor++]; 1960 if (ch == '#') 1961 ch = parsePastLine(); 1962 } 1963 return ch; 1964 } 1965 1966 /** 1967 * xmode parse past comment to end of line. 1968 */ 1969 private int parsePastLine() { 1970 int ch = temp[cursor++]; 1971 while (ch != 0 && !isLineSeparator(ch)) 1972 ch = temp[cursor++]; 1973 return ch; 1974 } 1975 1976 /** 1977 * xmode peek past comment to end of line. 1978 */ 1979 private int peekPastLine() { 1980 int ch = temp[++cursor]; 1981 while (ch != 0 && !isLineSeparator(ch)) 1982 ch = temp[++cursor]; 1983 return ch; 1984 } 1985 1986 /** 1987 * Determines if character is a line separator in the current mode 1988 */ 1989 private boolean isLineSeparator(int ch) { 1990 if (has(UNIX_LINES)) { 1991 return ch == '\n'; 1992 } else { 1993 return (ch == '\n' || 1994 ch == '\r' || 1995 (ch|1) == '\u2029' || 1996 ch == '\u0085'); 1997 } 1998 } 1999 2000 /** 2001 * Read the character after the next one, and advance the cursor by two. 2002 */ 2003 private int skip() { 2004 int i = cursor; 2005 int ch = temp[i+1]; 2006 cursor = i + 2; 2007 return ch; 2008 } 2009 2010 /** 2011 * Unread one next character, and retreat cursor by one. 2012 */ 2013 private void unread() { 2014 cursor--; 2015 } 2016 2017 /** 2018 * Internal method used for handling all syntax errors. The pattern is 2019 * displayed with a pointer to aid in locating the syntax error. 2020 */ 2021 private PatternSyntaxException error(String s) { 2022 return new PatternSyntaxException(s, normalizedPattern, cursor - 1); 2023 } 2024 2025 /** 2026 * Determines if there is any supplementary character or unpaired 2027 * surrogate in the specified range. 2028 */ 2029 private boolean findSupplementary(int start, int end) { 2030 for (int i = start; i < end; i++) { 2031 if (isSupplementary(temp[i])) 2032 return true; 2033 } 2034 return false; 2035 } 2036 2037 /** 2038 * Determines if the specified code point is a supplementary 2039 * character or unpaired surrogate. 2040 */ 2041 private static final boolean isSupplementary(int ch) { 2042 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || 2043 Character.isSurrogate((char)ch); 2044 } 2045 2046 /** 2047 * The following methods handle the main parsing. They are sorted 2048 * according to their precedence order, the lowest one first. 2049 */ 2050 2051 /** 2052 * The expression is parsed with branch nodes added for alternations. 2053 * This may be called recursively to parse sub expressions that may 2054 * contain alternations. 2055 */ 2056 private Node expr(Node end) { 2057 Node prev = null; 2058 Node firstTail = null; 2059 Branch branch = null; 2060 Node branchConn = null; 2061 2062 for (;;) { 2063 Node node = sequence(end); 2064 Node nodeTail = root; //double return 2065 if (prev == null) { 2066 prev = node; 2067 firstTail = nodeTail; 2068 } else { 2069 // Branch 2070 if (branchConn == null) { 2071 branchConn = new BranchConn(); 2072 branchConn.next = end; 2073 } 2074 if (node == end) { 2075 // if the node returned from sequence() is "end" 2076 // we have an empty expr, set a null atom into 2077 // the branch to indicate to go "next" directly. 2078 node = null; 2079 } else { 2080 // the "tail.next" of each atom goes to branchConn 2081 nodeTail.next = branchConn; 2082 } 2083 if (prev == branch) { 2084 branch.add(node); 2085 } else { 2086 if (prev == end) { 2087 prev = null; 2088 } else { 2089 // replace the "end" with "branchConn" at its tail.next 2090 // when put the "prev" into the branch as the first atom. 2091 firstTail.next = branchConn; 2092 } 2093 prev = branch = new Branch(prev, node, branchConn); 2094 } 2095 } 2096 if (peek() != '|') { 2097 return prev; 2098 } 2099 next(); 2100 } 2101 } 2102 2103 @SuppressWarnings("fallthrough") 2104 /** 2105 * Parsing of sequences between alternations. 2106 */ 2107 private Node sequence(Node end) { 2108 Node head = null; 2109 Node tail = null; 2110 Node node = null; 2111 LOOP: 2112 for (;;) { 2113 int ch = peek(); 2114 switch (ch) { 2115 case '(': 2116 // Because group handles its own closure, 2117 // we need to treat it differently 2118 node = group0(); 2119 // Check for comment or flag group 2120 if (node == null) 2121 continue; 2122 if (head == null) 2123 head = node; 2124 else 2125 tail.next = node; 2126 // Double return: Tail was returned in root 2127 tail = root; 2128 continue; 2129 case '[': 2130 if (has(CANON_EQ) && !has(LITERAL)) 2131 node = new NFCCharProperty(clazz(true)); 2132 else 2133 node = newCharProperty(clazz(true)); 2134 break; 2135 case '\\': 2136 ch = nextEscaped(); 2137 if (ch == 'p' || ch == 'P') { 2138 boolean oneLetter = true; 2139 boolean comp = (ch == 'P'); 2140 ch = next(); // Consume { if present 2141 if (ch != '{') { 2142 unread(); 2143 } else { 2144 oneLetter = false; 2145 } 2146 // node = newCharProperty(family(oneLetter, comp)); 2147 if (has(CANON_EQ) && !has(LITERAL)) 2148 node = new NFCCharProperty(family(oneLetter, comp)); 2149 else 2150 node = newCharProperty(family(oneLetter, comp)); 2151 } else { 2152 unread(); 2153 node = atom(); 2154 } 2155 break; 2156 case '^': 2157 next(); 2158 if (has(MULTILINE)) { 2159 if (has(UNIX_LINES)) 2160 node = new UnixCaret(); 2161 else 2162 node = new Caret(); 2163 } else { 2164 node = new Begin(); 2165 } 2166 break; 2167 case '$': 2168 next(); 2169 if (has(UNIX_LINES)) 2170 node = new UnixDollar(has(MULTILINE)); 2171 else 2172 node = new Dollar(has(MULTILINE)); 2173 break; 2174 case '.': 2175 next(); 2176 if (has(DOTALL)) { 2177 node = new CharProperty(ALL()); 2178 } else { 2179 if (has(UNIX_LINES)) { 2180 node = new CharProperty(UNIXDOT()); 2181 } else { 2182 node = new CharProperty(DOT()); 2183 } 2184 } 2185 break; 2186 case '|': 2187 case ')': 2188 break LOOP; 2189 case ']': // Now interpreting dangling ] and } as literals 2190 case '}': 2191 node = atom(); 2192 break; 2193 case '?': 2194 case '*': 2195 case '+': 2196 next(); 2197 throw error("Dangling meta character '" + ((char)ch) + "'"); 2198 case 0: 2199 if (cursor >= patternLength) { 2200 break LOOP; 2201 } 2202 // Fall through 2203 default: 2204 node = atom(); 2205 break; 2206 } 2207 2208 node = closure(node); 2209 /* save the top dot-greedy nodes (.*, .+) as well 2210 if (node instanceof GreedyCharProperty && 2211 ((GreedyCharProperty)node).cp instanceof Dot) { 2212 topClosureNodes.add(node); 2213 } 2214 */ 2215 if (head == null) { 2216 head = tail = node; 2217 } else { 2218 tail.next = node; 2219 tail = node; 2220 } 2221 } 2222 if (head == null) { 2223 return end; 2224 } 2225 tail.next = end; 2226 root = tail; //double return 2227 return head; 2228 } 2229 2230 @SuppressWarnings("fallthrough") 2231 /** 2232 * Parse and add a new Single or Slice. 2233 */ 2234 private Node atom() { 2235 int first = 0; 2236 int prev = -1; 2237 boolean hasSupplementary = false; 2238 int ch = peek(); 2239 for (;;) { 2240 switch (ch) { 2241 case '*': 2242 case '+': 2243 case '?': 2244 case '{': 2245 if (first > 1) { 2246 cursor = prev; // Unwind one character 2247 first--; 2248 } 2249 break; 2250 case '$': 2251 case '.': 2252 case '^': 2253 case '(': 2254 case '[': 2255 case '|': 2256 case ')': 2257 break; 2258 case '\\': 2259 ch = nextEscaped(); 2260 if (ch == 'p' || ch == 'P') { // Property 2261 if (first > 0) { // Slice is waiting; handle it first 2262 unread(); 2263 break; 2264 } else { // No slice; just return the family node 2265 boolean comp = (ch == 'P'); 2266 boolean oneLetter = true; 2267 ch = next(); // Consume { if present 2268 if (ch != '{') 2269 unread(); 2270 else 2271 oneLetter = false; 2272 if (has(CANON_EQ) && !has(LITERAL)) 2273 return new NFCCharProperty(family(oneLetter, comp)); 2274 else 2275 return newCharProperty(family(oneLetter, comp)); 2276 } 2277 } 2278 unread(); 2279 prev = cursor; 2280 ch = escape(false, first == 0, false); 2281 if (ch >= 0) { 2282 append(ch, first); 2283 first++; 2284 if (isSupplementary(ch)) { 2285 hasSupplementary = true; 2286 } 2287 ch = peek(); 2288 continue; 2289 } else if (first == 0) { 2290 return root; 2291 } 2292 // Unwind meta escape sequence 2293 cursor = prev; 2294 break; 2295 case 0: 2296 if (cursor >= patternLength) { 2297 break; 2298 } 2299 // Fall through 2300 default: 2301 prev = cursor; 2302 append(ch, first); 2303 first++; 2304 if (isSupplementary(ch)) { 2305 hasSupplementary = true; 2306 } 2307 ch = next(); 2308 continue; 2309 } 2310 break; 2311 } 2312 if (first == 1) { 2313 return newCharProperty(single(buffer[0])); 2314 } else { 2315 return newSlice(buffer, first, hasSupplementary); 2316 } 2317 } 2318 2319 private void append(int ch, int index) { 2320 int len = buffer.length; 2321 if (index - len >= 0) { 2322 len = ArraysSupport.newLength(len, 2323 1 + index - len, /* minimum growth */ 2324 len /* preferred growth */); 2325 buffer = Arrays.copyOf(buffer, len); 2326 } 2327 buffer[index] = ch; 2328 } 2329 2330 /** 2331 * Parses a backref greedily, taking as many numbers as it 2332 * can. The first digit is always treated as a backref, but 2333 * multi digit numbers are only treated as a backref if at 2334 * least that many backrefs exist at this point in the regex. 2335 */ 2336 private Node ref(int refNum) { 2337 boolean done = false; 2338 while(!done) { 2339 int ch = peek(); 2340 switch(ch) { 2341 case '0': 2342 case '1': 2343 case '2': 2344 case '3': 2345 case '4': 2346 case '5': 2347 case '6': 2348 case '7': 2349 case '8': 2350 case '9': 2351 int newRefNum = (refNum * 10) + (ch - '0'); 2352 // Add another number if it doesn't make a group 2353 // that doesn't exist 2354 if (capturingGroupCount - 1 < newRefNum) { 2355 done = true; 2356 break; 2357 } 2358 refNum = newRefNum; 2359 read(); 2360 break; 2361 default: 2362 done = true; 2363 break; 2364 } 2365 } 2366 hasGroupRef = true; 2367 if (has(CASE_INSENSITIVE)) 2368 return new CIBackRef(refNum, has(UNICODE_CASE)); 2369 else 2370 return new BackRef(refNum); 2371 } 2372 2373 /** 2374 * Parses an escape sequence to determine the actual value that needs 2375 * to be matched. 2376 * If -1 is returned and create was true a new object was added to the tree 2377 * to handle the escape sequence. 2378 * If the returned value is greater than zero, it is the value that 2379 * matches the escape sequence. 2380 */ 2381 private int escape(boolean inclass, boolean create, boolean isrange) { 2382 int ch = skip(); 2383 switch (ch) { 2384 case '0': 2385 return o(); 2386 case '1': 2387 case '2': 2388 case '3': 2389 case '4': 2390 case '5': 2391 case '6': 2392 case '7': 2393 case '8': 2394 case '9': 2395 if (inclass) break; 2396 if (create) { 2397 root = ref((ch - '0')); 2398 } 2399 return -1; 2400 case 'A': 2401 if (inclass) break; 2402 if (create) root = new Begin(); 2403 return -1; 2404 case 'B': 2405 if (inclass) break; 2406 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS)); 2407 return -1; 2408 case 'C': 2409 break; 2410 case 'D': 2411 if (create) { 2412 predicate = has(UNICODE_CHARACTER_CLASS) ? 2413 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2414 predicate = predicate.negate(); 2415 if (!inclass) 2416 root = newCharProperty(predicate); 2417 } 2418 return -1; 2419 case 'E': 2420 case 'F': 2421 break; 2422 case 'G': 2423 if (inclass) break; 2424 if (create) root = new LastMatch(); 2425 return -1; 2426 case 'H': 2427 if (create) { 2428 predicate = HorizWS().negate(); 2429 if (!inclass) 2430 root = newCharProperty(predicate); 2431 } 2432 return -1; 2433 case 'I': 2434 case 'J': 2435 case 'K': 2436 case 'L': 2437 case 'M': 2438 break; 2439 case 'N': 2440 return N(); 2441 case 'O': 2442 case 'P': 2443 case 'Q': 2444 break; 2445 case 'R': 2446 if (inclass) break; 2447 if (create) root = new LineEnding(); 2448 return -1; 2449 case 'S': 2450 if (create) { 2451 predicate = has(UNICODE_CHARACTER_CLASS) ? 2452 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2453 predicate = predicate.negate(); 2454 if (!inclass) 2455 root = newCharProperty(predicate); 2456 } 2457 return -1; 2458 case 'T': 2459 case 'U': 2460 break; 2461 case 'V': 2462 if (create) { 2463 predicate = VertWS().negate(); 2464 if (!inclass) 2465 root = newCharProperty(predicate); 2466 } 2467 return -1; 2468 case 'W': 2469 if (create) { 2470 predicate = has(UNICODE_CHARACTER_CLASS) ? 2471 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2472 predicate = predicate.negate(); 2473 if (!inclass) 2474 root = newCharProperty(predicate); 2475 } 2476 return -1; 2477 case 'X': 2478 if (inclass) break; 2479 if (create) { 2480 root = new XGrapheme(); 2481 } 2482 return -1; 2483 case 'Y': 2484 break; 2485 case 'Z': 2486 if (inclass) break; 2487 if (create) { 2488 if (has(UNIX_LINES)) 2489 root = new UnixDollar(false); 2490 else 2491 root = new Dollar(false); 2492 } 2493 return -1; 2494 case 'a': 2495 return '\007'; 2496 case 'b': 2497 if (inclass) break; 2498 if (create) { 2499 if (peek() == '{') { 2500 if (skip() == 'g') { 2501 if (read() == '}') { 2502 root = new GraphemeBound(); 2503 return -1; 2504 } 2505 break; // error missing trailing } 2506 } 2507 unread(); unread(); 2508 } 2509 root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS)); 2510 } 2511 return -1; 2512 case 'c': 2513 return c(); 2514 case 'd': 2515 if (create) { 2516 predicate = has(UNICODE_CHARACTER_CLASS) ? 2517 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2518 if (!inclass) 2519 root = newCharProperty(predicate); 2520 } 2521 return -1; 2522 case 'e': 2523 return '\033'; 2524 case 'f': 2525 return '\f'; 2526 case 'g': 2527 break; 2528 case 'h': 2529 if (create) { 2530 predicate = HorizWS(); 2531 if (!inclass) 2532 root = newCharProperty(predicate); 2533 } 2534 return -1; 2535 case 'i': 2536 case 'j': 2537 break; 2538 case 'k': 2539 if (inclass) 2540 break; 2541 if (read() != '<') 2542 throw error("\\k is not followed by '<' for named capturing group"); 2543 String name = groupname(read()); 2544 if (!namedGroups().containsKey(name)) 2545 throw error("named capturing group <" + name + "> does not exist"); 2546 if (create) { 2547 hasGroupRef = true; 2548 if (has(CASE_INSENSITIVE)) 2549 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE)); 2550 else 2551 root = new BackRef(namedGroups().get(name)); 2552 } 2553 return -1; 2554 case 'l': 2555 case 'm': 2556 break; 2557 case 'n': 2558 return '\n'; 2559 case 'o': 2560 case 'p': 2561 case 'q': 2562 break; 2563 case 'r': 2564 return '\r'; 2565 case 's': 2566 if (create) { 2567 predicate = has(UNICODE_CHARACTER_CLASS) ? 2568 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2569 if (!inclass) 2570 root = newCharProperty(predicate); 2571 } 2572 return -1; 2573 case 't': 2574 return '\t'; 2575 case 'u': 2576 return u(); 2577 case 'v': 2578 // '\v' was implemented as VT/0x0B in releases < 1.8 (though 2579 // undocumented). In JDK8 '\v' is specified as a predefined 2580 // character class for all vertical whitespace characters. 2581 // So [-1, root=VertWS node] pair is returned (instead of a 2582 // single 0x0B). This breaks the range if '\v' is used as 2583 // the start or end value, such as [\v-...] or [...-\v], in 2584 // which a single definite value (0x0B) is expected. For 2585 // compatibility concern '\013'/0x0B is returned if isrange. 2586 if (isrange) 2587 return '\013'; 2588 if (create) { 2589 predicate = VertWS(); 2590 if (!inclass) 2591 root = newCharProperty(predicate); 2592 } 2593 return -1; 2594 case 'w': 2595 if (create) { 2596 predicate = has(UNICODE_CHARACTER_CLASS) ? 2597 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2598 if (!inclass) 2599 root = newCharProperty(predicate); 2600 } 2601 return -1; 2602 case 'x': 2603 return x(); 2604 case 'y': 2605 break; 2606 case 'z': 2607 if (inclass) break; 2608 if (create) root = new End(); 2609 return -1; 2610 default: 2611 return ch; 2612 } 2613 throw error("Illegal/unsupported escape sequence"); 2614 } 2615 2616 /** 2617 * Parse a character class, and return the node that matches it. 2618 * 2619 * Consumes a ] on the way out if consume is true. Usually consume 2620 * is true except for the case of [abc&&def] where def is a separate 2621 * right hand node with "understood" brackets. 2622 */ 2623 private CharPredicate clazz(boolean consume) { 2624 CharPredicate prev = null; 2625 CharPredicate curr = null; 2626 BitClass bits = new BitClass(); 2627 BmpCharPredicate bitsP = ch -> ch < 256 && bits.bits[ch]; 2628 2629 boolean isNeg = false; 2630 boolean hasBits = false; 2631 int ch = next(); 2632 2633 // Negates if first char in a class, otherwise literal 2634 if (ch == '^' && temp[cursor-1] == '[') { 2635 ch = next(); 2636 isNeg = true; 2637 } 2638 for (;;) { 2639 switch (ch) { 2640 case '[': 2641 curr = clazz(true); 2642 if (prev == null) 2643 prev = curr; 2644 else 2645 prev = prev.union(curr); 2646 ch = peek(); 2647 continue; 2648 case '&': 2649 ch = next(); 2650 if (ch == '&') { 2651 ch = next(); 2652 CharPredicate right = null; 2653 while (ch != ']' && ch != '&') { 2654 if (ch == '[') { 2655 if (right == null) 2656 right = clazz(true); 2657 else 2658 right = right.union(clazz(true)); 2659 } else { // abc&&def 2660 unread(); 2661 right = clazz(false); 2662 } 2663 ch = peek(); 2664 } 2665 if (hasBits) { 2666 // bits used, union has high precedence 2667 if (prev == null) { 2668 prev = curr = bitsP; 2669 } else { 2670 prev = prev.union(bitsP); 2671 } 2672 hasBits = false; 2673 } 2674 if (right != null) 2675 curr = right; 2676 if (prev == null) { 2677 if (right == null) 2678 throw error("Bad class syntax"); 2679 else 2680 prev = right; 2681 } else { 2682 prev = prev.and(curr); 2683 } 2684 } else { 2685 // treat as a literal & 2686 unread(); 2687 break; 2688 } 2689 continue; 2690 case 0: 2691 if (cursor >= patternLength) 2692 throw error("Unclosed character class"); 2693 break; 2694 case ']': 2695 if (prev != null || hasBits) { 2696 if (consume) 2697 next(); 2698 if (prev == null) 2699 prev = bitsP; 2700 else if (hasBits) 2701 prev = prev.union(bitsP); 2702 if (isNeg) 2703 return prev.negate(); 2704 return prev; 2705 } 2706 break; 2707 default: 2708 break; 2709 } 2710 curr = range(bits); 2711 if (curr == null) { // the bits used 2712 hasBits = true; 2713 } else { 2714 if (prev == null) 2715 prev = curr; 2716 else if (prev != curr) 2717 prev = prev.union(curr); 2718 } 2719 ch = peek(); 2720 } 2721 } 2722 2723 private CharPredicate bitsOrSingle(BitClass bits, int ch) { 2724 /* Bits can only handle codepoints in [u+0000-u+00ff] range. 2725 Use "single" node instead of bits when dealing with unicode 2726 case folding for codepoints listed below. 2727 (1)Uppercase out of range: u+00ff, u+00b5 2728 toUpperCase(u+00ff) -> u+0178 2729 toUpperCase(u+00b5) -> u+039c 2730 (2)LatinSmallLetterLongS u+17f 2731 toUpperCase(u+017f) -> u+0053 2732 (3)LatinSmallLetterDotlessI u+131 2733 toUpperCase(u+0131) -> u+0049 2734 (4)LatinCapitalLetterIWithDotAbove u+0130 2735 toLowerCase(u+0130) -> u+0069 2736 (5)KelvinSign u+212a 2737 toLowerCase(u+212a) ==> u+006B 2738 (6)AngstromSign u+212b 2739 toLowerCase(u+212b) ==> u+00e5 2740 */ 2741 if (ch < 256 && 2742 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) && 2743 (ch == 0xff || ch == 0xb5 || 2744 ch == 0x49 || ch == 0x69 || //I and i 2745 ch == 0x53 || ch == 0x73 || //S and s 2746 ch == 0x4b || ch == 0x6b || //K and k 2747 ch == 0xc5 || ch == 0xe5))) { //A+ring 2748 bits.add(ch, flags0); 2749 return null; 2750 } 2751 return single(ch); 2752 } 2753 2754 /** 2755 * Returns a suitably optimized, single character predicate 2756 */ 2757 private CharPredicate single(final int ch) { 2758 if (has(CASE_INSENSITIVE)) { 2759 int lower, upper; 2760 if (has(UNICODE_CASE)) { 2761 upper = Character.toUpperCase(ch); 2762 lower = Character.toLowerCase(upper); 2763 // Unicode case insensitive matches 2764 if (upper != lower) 2765 return SingleU(lower); 2766 } else if (ASCII.isAscii(ch)) { 2767 lower = ASCII.toLower(ch); 2768 upper = ASCII.toUpper(ch); 2769 // Case insensitive matches a given BMP character 2770 if (lower != upper) 2771 return SingleI(lower, upper); 2772 } 2773 } 2774 if (isSupplementary(ch)) 2775 return SingleS(ch); 2776 return Single(ch); // Match a given BMP character 2777 } 2778 2779 /** 2780 * Parse a single character or a character range in a character class 2781 * and return its representative node. 2782 */ 2783 private CharPredicate range(BitClass bits) { 2784 int ch = peek(); 2785 if (ch == '\\') { 2786 ch = nextEscaped(); 2787 if (ch == 'p' || ch == 'P') { // A property 2788 boolean comp = (ch == 'P'); 2789 boolean oneLetter = true; 2790 // Consume { if present 2791 ch = next(); 2792 if (ch != '{') 2793 unread(); 2794 else 2795 oneLetter = false; 2796 return family(oneLetter, comp); 2797 } else { // ordinary escape 2798 boolean isrange = temp[cursor+1] == '-'; 2799 unread(); 2800 ch = escape(true, true, isrange); 2801 if (ch == -1) 2802 return predicate; 2803 } 2804 } else { 2805 next(); 2806 } 2807 if (ch >= 0) { 2808 if (peek() == '-') { 2809 int endRange = temp[cursor+1]; 2810 if (endRange == '[') { 2811 return bitsOrSingle(bits, ch); 2812 } 2813 if (endRange != ']') { 2814 next(); 2815 int m = peek(); 2816 if (m == '\\') { 2817 m = escape(true, false, true); 2818 } else { 2819 next(); 2820 } 2821 if (m < ch) { 2822 throw error("Illegal character range"); 2823 } 2824 if (has(CASE_INSENSITIVE)) { 2825 if (has(UNICODE_CASE)) 2826 return CIRangeU(ch, m); 2827 return CIRange(ch, m); 2828 } else { 2829 return Range(ch, m); 2830 } 2831 } 2832 } 2833 return bitsOrSingle(bits, ch); 2834 } 2835 throw error("Unexpected character '"+((char)ch)+"'"); 2836 } 2837 2838 /** 2839 * Parses a Unicode character family and returns its representative node. 2840 */ 2841 private CharPredicate family(boolean singleLetter, boolean isComplement) { 2842 next(); 2843 String name; 2844 CharPredicate p = null; 2845 2846 if (singleLetter) { 2847 int c = temp[cursor]; 2848 if (!Character.isSupplementaryCodePoint(c)) { 2849 name = String.valueOf((char)c); 2850 } else { 2851 name = new String(temp, cursor, 1); 2852 } 2853 read(); 2854 } else { 2855 int i = cursor; 2856 mark('}'); 2857 while(read() != '}') { 2858 } 2859 mark('\000'); 2860 int j = cursor; 2861 if (j > patternLength) 2862 throw error("Unclosed character family"); 2863 if (i + 1 >= j) 2864 throw error("Empty character family"); 2865 name = new String(temp, i, j-i-1); 2866 } 2867 2868 int i = name.indexOf('='); 2869 if (i != -1) { 2870 // property construct \p{name=value} 2871 String value = name.substring(i + 1); 2872 name = name.substring(0, i).toLowerCase(Locale.ENGLISH); 2873 switch (name) { 2874 case "sc": 2875 case "script": 2876 p = CharPredicates.forUnicodeScript(value); 2877 break; 2878 case "blk": 2879 case "block": 2880 p = CharPredicates.forUnicodeBlock(value); 2881 break; 2882 case "gc": 2883 case "general_category": 2884 p = CharPredicates.forProperty(value); 2885 break; 2886 default: 2887 break; 2888 } 2889 if (p == null) 2890 throw error("Unknown Unicode property {name=<" + name + ">, " 2891 + "value=<" + value + ">}"); 2892 2893 } else { 2894 if (name.startsWith("In")) { 2895 // \p{InBlockName} 2896 p = CharPredicates.forUnicodeBlock(name.substring(2)); 2897 } else if (name.startsWith("Is")) { 2898 // \p{IsGeneralCategory} and \p{IsScriptName} 2899 name = name.substring(2); 2900 p = CharPredicates.forUnicodeProperty(name); 2901 if (p == null) 2902 p = CharPredicates.forProperty(name); 2903 if (p == null) 2904 p = CharPredicates.forUnicodeScript(name); 2905 } else { 2906 if (has(UNICODE_CHARACTER_CLASS)) { 2907 p = CharPredicates.forPOSIXName(name); 2908 } 2909 if (p == null) 2910 p = CharPredicates.forProperty(name); 2911 } 2912 if (p == null) 2913 throw error("Unknown character property name {In/Is" + name + "}"); 2914 } 2915 if (isComplement) { 2916 // it might be too expensive to detect if a complement of 2917 // CharProperty can match "certain" supplementary. So just 2918 // go with StartS. 2919 hasSupplementary = true; 2920 p = p.negate(); 2921 } 2922 return p; 2923 } 2924 2925 private CharProperty newCharProperty(CharPredicate p) { 2926 if (p == null) 2927 return null; 2928 if (p instanceof BmpCharPredicate) 2929 return new BmpCharProperty((BmpCharPredicate)p); 2930 else 2931 return new CharProperty(p); 2932 } 2933 2934 /** 2935 * Parses and returns the name of a "named capturing group", the trailing 2936 * ">" is consumed after parsing. 2937 */ 2938 private String groupname(int ch) { 2939 StringBuilder sb = new StringBuilder(); 2940 if (!ASCII.isAlpha(ch)) 2941 throw error("capturing group name does not start with a Latin letter"); 2942 do { 2943 sb.append((char) ch); 2944 } while (ASCII.isAlnum(ch=read())); 2945 if (ch != '>') 2946 throw error("named capturing group is missing trailing '>'"); 2947 return sb.toString(); 2948 } 2949 2950 /** 2951 * Parses a group and returns the head node of a set of nodes that process 2952 * the group. Sometimes a double return system is used where the tail is 2953 * returned in root. 2954 */ 2955 private Node group0() { 2956 boolean capturingGroup = false; 2957 Node head = null; 2958 Node tail = null; 2959 int save = flags0; 2960 int saveTCNCount = topClosureNodes.size(); 2961 root = null; 2962 int ch = next(); 2963 if (ch == '?') { 2964 ch = skip(); 2965 switch (ch) { 2966 case ':': // (?:xxx) pure group 2967 head = createGroup(true); 2968 tail = root; 2969 head.next = expr(tail); 2970 break; 2971 case '=': // (?=xxx) and (?!xxx) lookahead 2972 case '!': 2973 head = createGroup(true); 2974 tail = root; 2975 head.next = expr(tail); 2976 if (ch == '=') { 2977 head = tail = new Pos(head); 2978 } else { 2979 head = tail = new Neg(head); 2980 } 2981 break; 2982 case '>': // (?>xxx) independent group 2983 head = createGroup(true); 2984 tail = root; 2985 head.next = expr(tail); 2986 head = tail = new Ques(head, Qtype.INDEPENDENT); 2987 break; 2988 case '<': // (?<xxx) look behind 2989 ch = read(); 2990 if (ch != '=' && ch != '!') { 2991 // named captured group 2992 String name = groupname(ch); 2993 if (namedGroups().containsKey(name)) 2994 throw error("Named capturing group <" + name 2995 + "> is already defined"); 2996 capturingGroup = true; 2997 head = createGroup(false); 2998 tail = root; 2999 namedGroups().put(name, capturingGroupCount-1); 3000 head.next = expr(tail); 3001 break; 3002 } 3003 int start = cursor; 3004 head = createGroup(true); 3005 tail = root; 3006 head.next = expr(tail); 3007 tail.next = lookbehindEnd; 3008 TreeInfo info = new TreeInfo(); 3009 head.study(info); 3010 if (info.maxValid == false) { 3011 throw error("Look-behind group does not have " 3012 + "an obvious maximum length"); 3013 } 3014 boolean hasSupplementary = findSupplementary(start, patternLength); 3015 if (ch == '=') { 3016 head = tail = (hasSupplementary ? 3017 new BehindS(head, info.maxLength, 3018 info.minLength) : 3019 new Behind(head, info.maxLength, 3020 info.minLength)); 3021 } else { // if (ch == '!') 3022 head = tail = (hasSupplementary ? 3023 new NotBehindS(head, info.maxLength, 3024 info.minLength) : 3025 new NotBehind(head, info.maxLength, 3026 info.minLength)); 3027 } 3028 // clear all top-closure-nodes inside lookbehind 3029 if (saveTCNCount < topClosureNodes.size()) 3030 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3031 break; 3032 case '$': 3033 case '@': 3034 throw error("Unknown group type"); 3035 default: // (?xxx:) inlined match flags 3036 unread(); 3037 addFlag(); 3038 ch = read(); 3039 if (ch == ')') { 3040 return null; // Inline modifier only 3041 } 3042 if (ch != ':') { 3043 throw error("Unknown inline modifier"); 3044 } 3045 head = createGroup(true); 3046 tail = root; 3047 head.next = expr(tail); 3048 break; 3049 } 3050 } else { // (xxx) a regular group 3051 capturingGroup = true; 3052 head = createGroup(false); 3053 tail = root; 3054 head.next = expr(tail); 3055 } 3056 3057 accept(')', "Unclosed group"); 3058 flags0 = save; 3059 3060 // Check for quantifiers 3061 Node node = closure(head); 3062 if (node == head) { // No closure 3063 root = tail; 3064 return node; // Dual return 3065 } 3066 if (head == tail) { // Zero length assertion 3067 root = node; 3068 return node; // Dual return 3069 } 3070 3071 // have group closure, clear all inner closure nodes from the 3072 // top list (no backtracking stopper optimization for inner 3073 if (saveTCNCount < topClosureNodes.size()) 3074 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3075 3076 if (node instanceof Ques) { 3077 Ques ques = (Ques) node; 3078 if (ques.type == Qtype.POSSESSIVE) { 3079 root = node; 3080 return node; 3081 } 3082 tail.next = new BranchConn(); 3083 tail = tail.next; 3084 if (ques.type == Qtype.GREEDY) { 3085 head = new Branch(head, null, tail); 3086 } else { // Reluctant quantifier 3087 head = new Branch(null, head, tail); 3088 } 3089 root = tail; 3090 return head; 3091 } else if (node instanceof Curly) { 3092 Curly curly = (Curly) node; 3093 if (curly.type == Qtype.POSSESSIVE) { 3094 root = node; 3095 return node; 3096 } 3097 // Discover if the group is deterministic 3098 TreeInfo info = new TreeInfo(); 3099 if (head.study(info)) { // Deterministic 3100 GroupTail temp = (GroupTail) tail; 3101 head = root = new GroupCurly(head.next, curly.cmin, 3102 curly.cmax, curly.type, 3103 ((GroupTail)tail).localIndex, 3104 ((GroupTail)tail).groupIndex, 3105 capturingGroup); 3106 return head; 3107 } else { // Non-deterministic 3108 int temp = ((GroupHead) head).localIndex; 3109 Loop loop; 3110 if (curly.type == Qtype.GREEDY) { 3111 loop = new Loop(this.localCount, temp); 3112 // add the max_reps greedy to the top-closure-node list 3113 if (curly.cmax == MAX_REPS) 3114 topClosureNodes.add(loop); 3115 } else { // Reluctant Curly 3116 loop = new LazyLoop(this.localCount, temp); 3117 } 3118 Prolog prolog = new Prolog(loop); 3119 this.localCount += 1; 3120 loop.cmin = curly.cmin; 3121 loop.cmax = curly.cmax; 3122 loop.body = head; 3123 tail.next = loop; 3124 root = loop; 3125 return prolog; // Dual return 3126 } 3127 } 3128 throw error("Internal logic error"); 3129 } 3130 3131 /** 3132 * Create group head and tail nodes using double return. If the group is 3133 * created with anonymous true then it is a pure group and should not 3134 * affect group counting. 3135 */ 3136 private Node createGroup(boolean anonymous) { 3137 int localIndex = localCount++; 3138 int groupIndex = 0; 3139 if (!anonymous) 3140 groupIndex = capturingGroupCount++; 3141 GroupHead head = new GroupHead(localIndex); 3142 root = new GroupTail(localIndex, groupIndex); 3143 3144 // for debug/print only, head.match does NOT need the "tail" info 3145 head.tail = (GroupTail)root; 3146 3147 if (!anonymous && groupIndex < 10) 3148 groupNodes[groupIndex] = head; 3149 return head; 3150 } 3151 3152 @SuppressWarnings("fallthrough") 3153 /** 3154 * Parses inlined match flags and set them appropriately. 3155 */ 3156 private void addFlag() { 3157 int ch = peek(); 3158 for (;;) { 3159 switch (ch) { 3160 case 'i': 3161 flags0 |= CASE_INSENSITIVE; 3162 break; 3163 case 'm': 3164 flags0 |= MULTILINE; 3165 break; 3166 case 's': 3167 flags0 |= DOTALL; 3168 break; 3169 case 'd': 3170 flags0 |= UNIX_LINES; 3171 break; 3172 case 'u': 3173 flags0 |= UNICODE_CASE; 3174 break; 3175 case 'c': 3176 flags0 |= CANON_EQ; 3177 break; 3178 case 'x': 3179 flags0 |= COMMENTS; 3180 break; 3181 case 'U': 3182 flags0 |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3183 break; 3184 case '-': // subFlag then fall through 3185 ch = next(); 3186 subFlag(); 3187 default: 3188 return; 3189 } 3190 ch = next(); 3191 } 3192 } 3193 3194 @SuppressWarnings("fallthrough") 3195 /** 3196 * Parses the second part of inlined match flags and turns off 3197 * flags appropriately. 3198 */ 3199 private void subFlag() { 3200 int ch = peek(); 3201 for (;;) { 3202 switch (ch) { 3203 case 'i': 3204 flags0 &= ~CASE_INSENSITIVE; 3205 break; 3206 case 'm': 3207 flags0 &= ~MULTILINE; 3208 break; 3209 case 's': 3210 flags0 &= ~DOTALL; 3211 break; 3212 case 'd': 3213 flags0 &= ~UNIX_LINES; 3214 break; 3215 case 'u': 3216 flags0 &= ~UNICODE_CASE; 3217 break; 3218 case 'c': 3219 flags0 &= ~CANON_EQ; 3220 break; 3221 case 'x': 3222 flags0 &= ~COMMENTS; 3223 break; 3224 case 'U': 3225 flags0 &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3226 break; 3227 default: 3228 return; 3229 } 3230 ch = next(); 3231 } 3232 } 3233 3234 static final int MAX_REPS = 0x7FFFFFFF; 3235 3236 static enum Qtype { 3237 GREEDY, LAZY, POSSESSIVE, INDEPENDENT 3238 } 3239 3240 private Node curly(Node prev, int cmin) { 3241 int ch = next(); 3242 if (ch == '?') { 3243 next(); 3244 return new Curly(prev, cmin, MAX_REPS, Qtype.LAZY); 3245 } else if (ch == '+') { 3246 next(); 3247 return new Curly(prev, cmin, MAX_REPS, Qtype.POSSESSIVE); 3248 } 3249 if (prev instanceof BmpCharProperty) { 3250 return new BmpCharPropertyGreedy((BmpCharProperty)prev, cmin); 3251 } else if (prev instanceof CharProperty) { 3252 return new CharPropertyGreedy((CharProperty)prev, cmin); 3253 } 3254 return new Curly(prev, cmin, MAX_REPS, Qtype.GREEDY); 3255 } 3256 3257 /** 3258 * Processes repetition. If the next character peeked is a quantifier 3259 * then new nodes must be appended to handle the repetition. 3260 * Prev could be a single or a group, so it could be a chain of nodes. 3261 */ 3262 private Node closure(Node prev) { 3263 Node atom; 3264 int ch = peek(); 3265 switch (ch) { 3266 case '?': 3267 ch = next(); 3268 if (ch == '?') { 3269 next(); 3270 return new Ques(prev, Qtype.LAZY); 3271 } else if (ch == '+') { 3272 next(); 3273 return new Ques(prev, Qtype.POSSESSIVE); 3274 } 3275 return new Ques(prev, Qtype.GREEDY); 3276 case '*': 3277 return curly(prev, 0); 3278 case '+': 3279 return curly(prev, 1); 3280 case '{': 3281 ch = temp[cursor+1]; 3282 if (ASCII.isDigit(ch)) { 3283 skip(); 3284 int cmin = 0; 3285 do { 3286 cmin = cmin * 10 + (ch - '0'); 3287 } while (ASCII.isDigit(ch = read())); 3288 int cmax = cmin; 3289 if (ch == ',') { 3290 ch = read(); 3291 cmax = MAX_REPS; 3292 if (ch != '}') { 3293 cmax = 0; 3294 while (ASCII.isDigit(ch)) { 3295 cmax = cmax * 10 + (ch - '0'); 3296 ch = read(); 3297 } 3298 } 3299 } 3300 if (ch != '}') 3301 throw error("Unclosed counted closure"); 3302 if (((cmin) | (cmax) | (cmax - cmin)) < 0) 3303 throw error("Illegal repetition range"); 3304 Curly curly; 3305 ch = peek(); 3306 if (ch == '?') { 3307 next(); 3308 curly = new Curly(prev, cmin, cmax, Qtype.LAZY); 3309 } else if (ch == '+') { 3310 next(); 3311 curly = new Curly(prev, cmin, cmax, Qtype.POSSESSIVE); 3312 } else { 3313 curly = new Curly(prev, cmin, cmax, Qtype.GREEDY); 3314 } 3315 return curly; 3316 } else { 3317 throw error("Illegal repetition"); 3318 } 3319 default: 3320 return prev; 3321 } 3322 } 3323 3324 /** 3325 * Utility method for parsing control escape sequences. 3326 */ 3327 private int c() { 3328 if (cursor < patternLength) { 3329 return read() ^ 64; 3330 } 3331 throw error("Illegal control escape sequence"); 3332 } 3333 3334 /** 3335 * Utility method for parsing octal escape sequences. 3336 */ 3337 private int o() { 3338 int n = read(); 3339 if (((n-'0')|('7'-n)) >= 0) { 3340 int m = read(); 3341 if (((m-'0')|('7'-m)) >= 0) { 3342 int o = read(); 3343 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) { 3344 return (n - '0') * 64 + (m - '0') * 8 + (o - '0'); 3345 } 3346 unread(); 3347 return (n - '0') * 8 + (m - '0'); 3348 } 3349 unread(); 3350 return (n - '0'); 3351 } 3352 throw error("Illegal octal escape sequence"); 3353 } 3354 3355 /** 3356 * Utility method for parsing hexadecimal escape sequences. 3357 */ 3358 private int x() { 3359 int n = read(); 3360 if (ASCII.isHexDigit(n)) { 3361 int m = read(); 3362 if (ASCII.isHexDigit(m)) { 3363 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m); 3364 } 3365 } else if (n == '{' && ASCII.isHexDigit(peek())) { 3366 int ch = 0; 3367 while (ASCII.isHexDigit(n = read())) { 3368 ch = (ch << 4) + ASCII.toDigit(n); 3369 if (ch > Character.MAX_CODE_POINT) 3370 throw error("Hexadecimal codepoint is too big"); 3371 } 3372 if (n != '}') 3373 throw error("Unclosed hexadecimal escape sequence"); 3374 return ch; 3375 } 3376 throw error("Illegal hexadecimal escape sequence"); 3377 } 3378 3379 /** 3380 * Utility method for parsing unicode escape sequences. 3381 */ 3382 private int cursor() { 3383 return cursor; 3384 } 3385 3386 private void setcursor(int pos) { 3387 cursor = pos; 3388 } 3389 3390 private int uxxxx() { 3391 int n = 0; 3392 for (int i = 0; i < 4; i++) { 3393 int ch = read(); 3394 if (!ASCII.isHexDigit(ch)) { 3395 throw error("Illegal Unicode escape sequence"); 3396 } 3397 n = n * 16 + ASCII.toDigit(ch); 3398 } 3399 return n; 3400 } 3401 3402 private int u() { 3403 int n = uxxxx(); 3404 if (Character.isHighSurrogate((char)n)) { 3405 int cur = cursor(); 3406 if (read() == '\\' && read() == 'u') { 3407 int n2 = uxxxx(); 3408 if (Character.isLowSurrogate((char)n2)) 3409 return Character.toCodePoint((char)n, (char)n2); 3410 } 3411 setcursor(cur); 3412 } 3413 return n; 3414 } 3415 3416 private int N() { 3417 if (read() == '{') { 3418 int i = cursor; 3419 while (cursor < patternLength && read() != '}') {} 3420 if (cursor > patternLength) 3421 throw error("Unclosed character name escape sequence"); 3422 String name = new String(temp, i, cursor - i - 1); 3423 try { 3424 return Character.codePointOf(name); 3425 } catch (IllegalArgumentException x) { 3426 throw error("Unknown character name [" + name + "]"); 3427 } 3428 } 3429 throw error("Illegal character name escape sequence"); 3430 } 3431 3432 // 3433 // Utility methods for code point support 3434 // 3435 private static final int countChars(CharSequence seq, int index, 3436 int lengthInCodePoints) { 3437 // optimization 3438 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) { 3439 assert (index >= 0 && index < seq.length()); 3440 return 1; 3441 } 3442 int length = seq.length(); 3443 int x = index; 3444 if (lengthInCodePoints >= 0) { 3445 assert (index >= 0 && index < length); 3446 for (int i = 0; x < length && i < lengthInCodePoints; i++) { 3447 if (Character.isHighSurrogate(seq.charAt(x++))) { 3448 if (x < length && Character.isLowSurrogate(seq.charAt(x))) { 3449 x++; 3450 } 3451 } 3452 } 3453 return x - index; 3454 } 3455 3456 assert (index >= 0 && index <= length); 3457 if (index == 0) { 3458 return 0; 3459 } 3460 int len = -lengthInCodePoints; 3461 for (int i = 0; x > 0 && i < len; i++) { 3462 if (Character.isLowSurrogate(seq.charAt(--x))) { 3463 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) { 3464 x--; 3465 } 3466 } 3467 } 3468 return index - x; 3469 } 3470 3471 private static final int countCodePoints(CharSequence seq) { 3472 int length = seq.length(); 3473 int n = 0; 3474 for (int i = 0; i < length; ) { 3475 n++; 3476 if (Character.isHighSurrogate(seq.charAt(i++))) { 3477 if (i < length && Character.isLowSurrogate(seq.charAt(i))) { 3478 i++; 3479 } 3480 } 3481 } 3482 return n; 3483 } 3484 3485 /** 3486 * Creates a bit vector for matching Latin-1 values. A normal BitClass 3487 * never matches values above Latin-1, and a complemented BitClass always 3488 * matches values above Latin-1. 3489 */ 3490 static final class BitClass extends BmpCharProperty { 3491 final boolean[] bits; 3492 BitClass() { 3493 this(new boolean[256]); 3494 } 3495 private BitClass(boolean[] bits) { 3496 super( ch -> ch < 256 && bits[ch]); 3497 this.bits = bits; 3498 } 3499 BitClass add(int c, int flags) { 3500 assert c >= 0 && c <= 255; 3501 if ((flags & CASE_INSENSITIVE) != 0) { 3502 if (ASCII.isAscii(c)) { 3503 bits[ASCII.toUpper(c)] = true; 3504 bits[ASCII.toLower(c)] = true; 3505 } else if ((flags & UNICODE_CASE) != 0) { 3506 bits[Character.toLowerCase(c)] = true; 3507 bits[Character.toUpperCase(c)] = true; 3508 } 3509 } 3510 bits[c] = true; 3511 return this; 3512 } 3513 } 3514 3515 /** 3516 * Utility method for creating a string slice matcher. 3517 */ 3518 private Node newSlice(int[] buf, int count, boolean hasSupplementary) { 3519 int[] tmp = new int[count]; 3520 if (has(CASE_INSENSITIVE)) { 3521 if (has(UNICODE_CASE)) { 3522 for (int i = 0; i < count; i++) { 3523 tmp[i] = Character.toLowerCase( 3524 Character.toUpperCase(buf[i])); 3525 } 3526 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp); 3527 } 3528 for (int i = 0; i < count; i++) { 3529 tmp[i] = ASCII.toLower(buf[i]); 3530 } 3531 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp); 3532 } 3533 for (int i = 0; i < count; i++) { 3534 tmp[i] = buf[i]; 3535 } 3536 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp); 3537 } 3538 3539 /** 3540 * The following classes are the building components of the object 3541 * tree that represents a compiled regular expression. The object tree 3542 * is made of individual elements that handle constructs in the Pattern. 3543 * Each type of object knows how to match its equivalent construct with 3544 * the match() method. 3545 */ 3546 3547 /** 3548 * Base class for all node classes. Subclasses should override the match() 3549 * method as appropriate. This class is an accepting node, so its match() 3550 * always returns true. 3551 */ 3552 static class Node extends Object { 3553 Node next; 3554 Node() { 3555 next = Pattern.accept; 3556 } 3557 /** 3558 * This method implements the classic accept node. 3559 */ 3560 boolean match(Matcher matcher, int i, CharSequence seq) { 3561 matcher.last = i; 3562 matcher.groups[0] = matcher.first; 3563 matcher.groups[1] = matcher.last; 3564 return true; 3565 } 3566 /** 3567 * This method is good for all zero length assertions. 3568 */ 3569 boolean study(TreeInfo info) { 3570 if (next != null) { 3571 return next.study(info); 3572 } else { 3573 return info.deterministic; 3574 } 3575 } 3576 } 3577 3578 static class LastNode extends Node { 3579 /** 3580 * This method implements the classic accept node with 3581 * the addition of a check to see if the match occurred 3582 * using all of the input. 3583 */ 3584 boolean match(Matcher matcher, int i, CharSequence seq) { 3585 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to) 3586 return false; 3587 matcher.last = i; 3588 matcher.groups[0] = matcher.first; 3589 matcher.groups[1] = matcher.last; 3590 return true; 3591 } 3592 } 3593 3594 /** 3595 * Used for REs that can start anywhere within the input string. 3596 * This basically tries to match repeatedly at each spot in the 3597 * input string, moving forward after each try. An anchored search 3598 * or a BnM will bypass this node completely. 3599 */ 3600 static class Start extends Node { 3601 int minLength; 3602 Start(Node node) { 3603 this.next = node; 3604 TreeInfo info = new TreeInfo(); 3605 next.study(info); 3606 minLength = info.minLength; 3607 } 3608 boolean match(Matcher matcher, int i, CharSequence seq) { 3609 if (i > matcher.to - minLength) { 3610 matcher.hitEnd = true; 3611 return false; 3612 } 3613 int guard = matcher.to - minLength; 3614 for (; i <= guard; i++) { 3615 if (next.match(matcher, i, seq)) { 3616 matcher.first = i; 3617 matcher.groups[0] = matcher.first; 3618 matcher.groups[1] = matcher.last; 3619 return true; 3620 } 3621 } 3622 matcher.hitEnd = true; 3623 return false; 3624 } 3625 boolean study(TreeInfo info) { 3626 next.study(info); 3627 info.maxValid = false; 3628 info.deterministic = false; 3629 return false; 3630 } 3631 } 3632 3633 /* 3634 * StartS supports supplementary characters, including unpaired surrogates. 3635 */ 3636 static final class StartS extends Start { 3637 StartS(Node node) { 3638 super(node); 3639 } 3640 boolean match(Matcher matcher, int i, CharSequence seq) { 3641 if (i > matcher.to - minLength) { 3642 matcher.hitEnd = true; 3643 return false; 3644 } 3645 int guard = matcher.to - minLength; 3646 while (i <= guard) { 3647 //if ((ret = next.match(matcher, i, seq)) || i == guard) 3648 if (next.match(matcher, i, seq)) { 3649 matcher.first = i; 3650 matcher.groups[0] = matcher.first; 3651 matcher.groups[1] = matcher.last; 3652 return true; 3653 } 3654 if (i == guard) 3655 break; 3656 // Optimization to move to the next character. This is 3657 // faster than countChars(seq, i, 1). 3658 if (Character.isHighSurrogate(seq.charAt(i++))) { 3659 if (i < seq.length() && 3660 Character.isLowSurrogate(seq.charAt(i))) { 3661 i++; 3662 } 3663 } 3664 } 3665 matcher.hitEnd = true; 3666 return false; 3667 } 3668 } 3669 3670 /** 3671 * Node to anchor at the beginning of input. This object implements the 3672 * match for a \A sequence, and the caret anchor will use this if not in 3673 * multiline mode. 3674 */ 3675 static final class Begin extends Node { 3676 boolean match(Matcher matcher, int i, CharSequence seq) { 3677 int fromIndex = (matcher.anchoringBounds) ? 3678 matcher.from : 0; 3679 if (i == fromIndex && next.match(matcher, i, seq)) { 3680 matcher.first = i; 3681 matcher.groups[0] = i; 3682 matcher.groups[1] = matcher.last; 3683 return true; 3684 } else { 3685 return false; 3686 } 3687 } 3688 } 3689 3690 /** 3691 * Node to anchor at the end of input. This is the absolute end, so this 3692 * should not match at the last newline before the end as $ will. 3693 */ 3694 static final class End extends Node { 3695 boolean match(Matcher matcher, int i, CharSequence seq) { 3696 int endIndex = (matcher.anchoringBounds) ? 3697 matcher.to : matcher.getTextLength(); 3698 if (i == endIndex) { 3699 matcher.hitEnd = true; 3700 return next.match(matcher, i, seq); 3701 } 3702 return false; 3703 } 3704 } 3705 3706 /** 3707 * Node to anchor at the beginning of a line. This is essentially the 3708 * object to match for the multiline ^. 3709 */ 3710 static final class Caret extends Node { 3711 boolean match(Matcher matcher, int i, CharSequence seq) { 3712 int startIndex = matcher.from; 3713 int endIndex = matcher.to; 3714 if (!matcher.anchoringBounds) { 3715 startIndex = 0; 3716 endIndex = matcher.getTextLength(); 3717 } 3718 // Perl does not match ^ at end of input even after newline 3719 if (i == endIndex) { 3720 matcher.hitEnd = true; 3721 return false; 3722 } 3723 if (i > startIndex) { 3724 char ch = seq.charAt(i-1); 3725 if (ch != '\n' && ch != '\r' 3726 && (ch|1) != '\u2029' 3727 && ch != '\u0085' ) { 3728 return false; 3729 } 3730 // Should treat /r/n as one newline 3731 if (ch == '\r' && seq.charAt(i) == '\n') 3732 return false; 3733 } 3734 return next.match(matcher, i, seq); 3735 } 3736 } 3737 3738 /** 3739 * Node to anchor at the beginning of a line when in unixdot mode. 3740 */ 3741 static final class UnixCaret extends Node { 3742 boolean match(Matcher matcher, int i, CharSequence seq) { 3743 int startIndex = matcher.from; 3744 int endIndex = matcher.to; 3745 if (!matcher.anchoringBounds) { 3746 startIndex = 0; 3747 endIndex = matcher.getTextLength(); 3748 } 3749 // Perl does not match ^ at end of input even after newline 3750 if (i == endIndex) { 3751 matcher.hitEnd = true; 3752 return false; 3753 } 3754 if (i > startIndex) { 3755 char ch = seq.charAt(i-1); 3756 if (ch != '\n') { 3757 return false; 3758 } 3759 } 3760 return next.match(matcher, i, seq); 3761 } 3762 } 3763 3764 /** 3765 * Node to match the location where the last match ended. 3766 * This is used for the \G construct. 3767 */ 3768 static final class LastMatch extends Node { 3769 boolean match(Matcher matcher, int i, CharSequence seq) { 3770 if (i != matcher.oldLast) 3771 return false; 3772 return next.match(matcher, i, seq); 3773 } 3774 } 3775 3776 /** 3777 * Node to anchor at the end of a line or the end of input based on the 3778 * multiline mode. 3779 * 3780 * When not in multiline mode, the $ can only match at the very end 3781 * of the input, unless the input ends in a line terminator in which 3782 * it matches right before the last line terminator. 3783 * 3784 * Note that \r\n is considered an atomic line terminator. 3785 * 3786 * Like ^ the $ operator matches at a position, it does not match the 3787 * line terminators themselves. 3788 */ 3789 static final class Dollar extends Node { 3790 boolean multiline; 3791 Dollar(boolean mul) { 3792 multiline = mul; 3793 } 3794 boolean match(Matcher matcher, int i, CharSequence seq) { 3795 int endIndex = (matcher.anchoringBounds) ? 3796 matcher.to : matcher.getTextLength(); 3797 if (!multiline) { 3798 if (i < endIndex - 2) 3799 return false; 3800 if (i == endIndex - 2) { 3801 char ch = seq.charAt(i); 3802 if (ch != '\r') 3803 return false; 3804 ch = seq.charAt(i + 1); 3805 if (ch != '\n') 3806 return false; 3807 } 3808 } 3809 // Matches before any line terminator; also matches at the 3810 // end of input 3811 // Before line terminator: 3812 // If multiline, we match here no matter what 3813 // If not multiline, fall through so that the end 3814 // is marked as hit; this must be a /r/n or a /n 3815 // at the very end so the end was hit; more input 3816 // could make this not match here 3817 if (i < endIndex) { 3818 char ch = seq.charAt(i); 3819 if (ch == '\n') { 3820 // No match between \r\n 3821 if (i > 0 && seq.charAt(i-1) == '\r') 3822 return false; 3823 if (multiline) 3824 return next.match(matcher, i, seq); 3825 } else if (ch == '\r' || ch == '\u0085' || 3826 (ch|1) == '\u2029') { 3827 if (multiline) 3828 return next.match(matcher, i, seq); 3829 } else { // No line terminator, no match 3830 return false; 3831 } 3832 } 3833 // Matched at current end so hit end 3834 matcher.hitEnd = true; 3835 // If a $ matches because of end of input, then more input 3836 // could cause it to fail! 3837 matcher.requireEnd = true; 3838 return next.match(matcher, i, seq); 3839 } 3840 boolean study(TreeInfo info) { 3841 next.study(info); 3842 return info.deterministic; 3843 } 3844 } 3845 3846 /** 3847 * Node to anchor at the end of a line or the end of input based on the 3848 * multiline mode when in unix lines mode. 3849 */ 3850 static final class UnixDollar extends Node { 3851 boolean multiline; 3852 UnixDollar(boolean mul) { 3853 multiline = mul; 3854 } 3855 boolean match(Matcher matcher, int i, CharSequence seq) { 3856 int endIndex = (matcher.anchoringBounds) ? 3857 matcher.to : matcher.getTextLength(); 3858 if (i < endIndex) { 3859 char ch = seq.charAt(i); 3860 if (ch == '\n') { 3861 // If not multiline, then only possible to 3862 // match at very end or one before end 3863 if (multiline == false && i != endIndex - 1) 3864 return false; 3865 // If multiline return next.match without setting 3866 // matcher.hitEnd 3867 if (multiline) 3868 return next.match(matcher, i, seq); 3869 } else { 3870 return false; 3871 } 3872 } 3873 // Matching because at the end or 1 before the end; 3874 // more input could change this so set hitEnd 3875 matcher.hitEnd = true; 3876 // If a $ matches because of end of input, then more input 3877 // could cause it to fail! 3878 matcher.requireEnd = true; 3879 return next.match(matcher, i, seq); 3880 } 3881 boolean study(TreeInfo info) { 3882 next.study(info); 3883 return info.deterministic; 3884 } 3885 } 3886 3887 /** 3888 * Node class that matches a Unicode line ending '\R' 3889 */ 3890 static final class LineEnding extends Node { 3891 boolean match(Matcher matcher, int i, CharSequence seq) { 3892 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029]) 3893 if (i < matcher.to) { 3894 int ch = seq.charAt(i); 3895 if (ch == 0x0A || ch == 0x0B || ch == 0x0C || 3896 ch == 0x85 || ch == 0x2028 || ch == 0x2029) 3897 return next.match(matcher, i + 1, seq); 3898 if (ch == 0x0D) { 3899 i++; 3900 if (i < matcher.to) { 3901 if (seq.charAt(i) == 0x0A && 3902 next.match(matcher, i + 1, seq)) { 3903 return true; 3904 } 3905 } else { 3906 matcher.hitEnd = true; 3907 } 3908 return next.match(matcher, i, seq); 3909 } 3910 } else { 3911 matcher.hitEnd = true; 3912 } 3913 return false; 3914 } 3915 boolean study(TreeInfo info) { 3916 info.minLength++; 3917 info.maxLength += 2; 3918 return next.study(info); 3919 } 3920 } 3921 3922 /** 3923 * Abstract node class to match one character satisfying some 3924 * boolean property. 3925 */ 3926 static class CharProperty extends Node { 3927 CharPredicate predicate; 3928 3929 CharProperty (CharPredicate predicate) { 3930 this.predicate = predicate; 3931 } 3932 boolean match(Matcher matcher, int i, CharSequence seq) { 3933 if (i < matcher.to) { 3934 int ch = Character.codePointAt(seq, i); 3935 return predicate.is(ch) && 3936 next.match(matcher, i + Character.charCount(ch), seq); 3937 } else { 3938 matcher.hitEnd = true; 3939 return false; 3940 } 3941 } 3942 boolean study(TreeInfo info) { 3943 info.minLength++; 3944 info.maxLength++; 3945 return next.study(info); 3946 } 3947 } 3948 3949 /** 3950 * Optimized version of CharProperty that works only for 3951 * properties never satisfied by Supplementary characters. 3952 */ 3953 private static class BmpCharProperty extends CharProperty { 3954 BmpCharProperty (BmpCharPredicate predicate) { 3955 super(predicate); 3956 } 3957 boolean match(Matcher matcher, int i, CharSequence seq) { 3958 if (i < matcher.to) { 3959 return predicate.is(seq.charAt(i)) && 3960 next.match(matcher, i + 1, seq); 3961 } else { 3962 matcher.hitEnd = true; 3963 return false; 3964 } 3965 } 3966 } 3967 3968 private static class NFCCharProperty extends Node { 3969 CharPredicate predicate; 3970 NFCCharProperty (CharPredicate predicate) { 3971 this.predicate = predicate; 3972 } 3973 3974 boolean match(Matcher matcher, int i, CharSequence seq) { 3975 if (i < matcher.to) { 3976 int ch0 = Character.codePointAt(seq, i); 3977 int n = Character.charCount(ch0); 3978 int j = i + n; 3979 while (j < matcher.to) { 3980 int ch1 = Character.codePointAt(seq, j); 3981 if (Grapheme.isBoundary(ch0, ch1)) 3982 break; 3983 ch0 = ch1; 3984 j += Character.charCount(ch1); 3985 } 3986 if (i + n == j) { // single, assume nfc cp 3987 if (predicate.is(ch0)) 3988 return next.match(matcher, j, seq); 3989 } else { 3990 while (i + n < j) { 3991 String nfc = Normalizer.normalize( 3992 seq.toString().substring(i, j), Normalizer.Form.NFC); 3993 if (nfc.codePointCount(0, nfc.length()) == 1) { 3994 if (predicate.is(nfc.codePointAt(0)) && 3995 next.match(matcher, j, seq)) { 3996 return true; 3997 } 3998 } 3999 4000 ch0 = Character.codePointBefore(seq, j); 4001 j -= Character.charCount(ch0); 4002 } 4003 } 4004 if (j < matcher.to) 4005 return false; 4006 } 4007 matcher.hitEnd = true; 4008 return false; 4009 } 4010 4011 boolean study(TreeInfo info) { 4012 info.minLength++; 4013 info.deterministic = false; 4014 return next.study(info); 4015 } 4016 } 4017 4018 /** 4019 * Node class that matches an unicode extended grapheme cluster 4020 */ 4021 static class XGrapheme extends Node { 4022 boolean match(Matcher matcher, int i, CharSequence seq) { 4023 if (i < matcher.to) { 4024 int ch0 = Character.codePointAt(seq, i); 4025 i += Character.charCount(ch0); 4026 while (i < matcher.to) { 4027 int ch1 = Character.codePointAt(seq, i); 4028 if (Grapheme.isBoundary(ch0, ch1)) 4029 break; 4030 ch0 = ch1; 4031 i += Character.charCount(ch1); 4032 } 4033 return next.match(matcher, i, seq); 4034 } 4035 matcher.hitEnd = true; 4036 return false; 4037 } 4038 4039 boolean study(TreeInfo info) { 4040 info.minLength++; 4041 info.deterministic = false; 4042 return next.study(info); 4043 } 4044 } 4045 4046 /** 4047 * Node class that handles grapheme boundaries 4048 */ 4049 static class GraphemeBound extends Node { 4050 boolean match(Matcher matcher, int i, CharSequence seq) { 4051 int startIndex = matcher.from; 4052 int endIndex = matcher.to; 4053 if (matcher.transparentBounds) { 4054 startIndex = 0; 4055 endIndex = matcher.getTextLength(); 4056 } 4057 if (i == startIndex) { 4058 return next.match(matcher, i, seq); 4059 } 4060 if (i < endIndex) { 4061 if (Character.isSurrogatePair(seq.charAt(i-1), seq.charAt(i)) || 4062 !Grapheme.isBoundary(Character.codePointBefore(seq, i), 4063 Character.codePointAt(seq, i))) { 4064 return false; 4065 } 4066 } else { 4067 matcher.hitEnd = true; 4068 matcher.requireEnd = true; 4069 } 4070 return next.match(matcher, i, seq); 4071 } 4072 } 4073 4074 /** 4075 * Base class for all Slice nodes 4076 */ 4077 static class SliceNode extends Node { 4078 int[] buffer; 4079 SliceNode(int[] buf) { 4080 buffer = buf; 4081 } 4082 boolean study(TreeInfo info) { 4083 info.minLength += buffer.length; 4084 info.maxLength += buffer.length; 4085 return next.study(info); 4086 } 4087 } 4088 4089 /** 4090 * Node class for a case sensitive/BMP-only sequence of literal 4091 * characters. 4092 */ 4093 static class Slice extends SliceNode { 4094 Slice(int[] buf) { 4095 super(buf); 4096 } 4097 boolean match(Matcher matcher, int i, CharSequence seq) { 4098 int[] buf = buffer; 4099 int len = buf.length; 4100 for (int j=0; j<len; j++) { 4101 if ((i+j) >= matcher.to) { 4102 matcher.hitEnd = true; 4103 return false; 4104 } 4105 if (buf[j] != seq.charAt(i+j)) 4106 return false; 4107 } 4108 return next.match(matcher, i+len, seq); 4109 } 4110 } 4111 4112 /** 4113 * Node class for a case_insensitive/BMP-only sequence of literal 4114 * characters. 4115 */ 4116 static class SliceI extends SliceNode { 4117 SliceI(int[] buf) { 4118 super(buf); 4119 } 4120 boolean match(Matcher matcher, int i, CharSequence seq) { 4121 int[] buf = buffer; 4122 int len = buf.length; 4123 for (int j=0; j<len; j++) { 4124 if ((i+j) >= matcher.to) { 4125 matcher.hitEnd = true; 4126 return false; 4127 } 4128 int c = seq.charAt(i+j); 4129 if (buf[j] != c && 4130 buf[j] != ASCII.toLower(c)) 4131 return false; 4132 } 4133 return next.match(matcher, i+len, seq); 4134 } 4135 } 4136 4137 /** 4138 * Node class for a unicode_case_insensitive/BMP-only sequence of 4139 * literal characters. Uses unicode case folding. 4140 */ 4141 static final class SliceU extends SliceNode { 4142 SliceU(int[] buf) { 4143 super(buf); 4144 } 4145 boolean match(Matcher matcher, int i, CharSequence seq) { 4146 int[] buf = buffer; 4147 int len = buf.length; 4148 for (int j=0; j<len; j++) { 4149 if ((i+j) >= matcher.to) { 4150 matcher.hitEnd = true; 4151 return false; 4152 } 4153 int c = seq.charAt(i+j); 4154 if (buf[j] != c && 4155 buf[j] != Character.toLowerCase(Character.toUpperCase(c))) 4156 return false; 4157 } 4158 return next.match(matcher, i+len, seq); 4159 } 4160 } 4161 4162 /** 4163 * Node class for a case sensitive sequence of literal characters 4164 * including supplementary characters. 4165 */ 4166 static final class SliceS extends Slice { 4167 SliceS(int[] buf) { 4168 super(buf); 4169 } 4170 boolean match(Matcher matcher, int i, CharSequence seq) { 4171 int[] buf = buffer; 4172 int x = i; 4173 for (int j = 0; j < buf.length; j++) { 4174 if (x >= matcher.to) { 4175 matcher.hitEnd = true; 4176 return false; 4177 } 4178 int c = Character.codePointAt(seq, x); 4179 if (buf[j] != c) 4180 return false; 4181 x += Character.charCount(c); 4182 if (x > matcher.to) { 4183 matcher.hitEnd = true; 4184 return false; 4185 } 4186 } 4187 return next.match(matcher, x, seq); 4188 } 4189 } 4190 4191 /** 4192 * Node class for a case insensitive sequence of literal characters 4193 * including supplementary characters. 4194 */ 4195 static class SliceIS extends SliceNode { 4196 SliceIS(int[] buf) { 4197 super(buf); 4198 } 4199 int toLower(int c) { 4200 return ASCII.toLower(c); 4201 } 4202 boolean match(Matcher matcher, int i, CharSequence seq) { 4203 int[] buf = buffer; 4204 int x = i; 4205 for (int j = 0; j < buf.length; j++) { 4206 if (x >= matcher.to) { 4207 matcher.hitEnd = true; 4208 return false; 4209 } 4210 int c = Character.codePointAt(seq, x); 4211 if (buf[j] != c && buf[j] != toLower(c)) 4212 return false; 4213 x += Character.charCount(c); 4214 if (x > matcher.to) { 4215 matcher.hitEnd = true; 4216 return false; 4217 } 4218 } 4219 return next.match(matcher, x, seq); 4220 } 4221 } 4222 4223 /** 4224 * Node class for a case insensitive sequence of literal characters. 4225 * Uses unicode case folding. 4226 */ 4227 static final class SliceUS extends SliceIS { 4228 SliceUS(int[] buf) { 4229 super(buf); 4230 } 4231 int toLower(int c) { 4232 return Character.toLowerCase(Character.toUpperCase(c)); 4233 } 4234 } 4235 4236 /** 4237 * The 0 or 1 quantifier. This one class implements all three types. 4238 */ 4239 static final class Ques extends Node { 4240 Node atom; 4241 Qtype type; 4242 Ques(Node node, Qtype type) { 4243 this.atom = node; 4244 this.type = type; 4245 } 4246 boolean match(Matcher matcher, int i, CharSequence seq) { 4247 switch (type) { 4248 case GREEDY: 4249 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)) 4250 || next.match(matcher, i, seq); 4251 case LAZY: 4252 return next.match(matcher, i, seq) 4253 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)); 4254 case POSSESSIVE: 4255 if (atom.match(matcher, i, seq)) i = matcher.last; 4256 return next.match(matcher, i, seq); 4257 default: 4258 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq); 4259 } 4260 } 4261 boolean study(TreeInfo info) { 4262 if (type != Qtype.INDEPENDENT) { 4263 int minL = info.minLength; 4264 atom.study(info); 4265 info.minLength = minL; 4266 info.deterministic = false; 4267 return next.study(info); 4268 } else { 4269 atom.study(info); 4270 return next.study(info); 4271 } 4272 } 4273 } 4274 4275 /** 4276 * Handles the greedy style repetition with the minimum either be 4277 * 0 or 1 and the maximum be MAX_REPS, for * and + quantifier. 4278 */ 4279 static class CharPropertyGreedy extends Node { 4280 final CharPredicate predicate; 4281 final int cmin; 4282 4283 CharPropertyGreedy(CharProperty cp, int cmin) { 4284 this.predicate = cp.predicate; 4285 this.cmin = cmin; 4286 } 4287 boolean match(Matcher matcher, int i, CharSequence seq) { 4288 int n = 0; 4289 int to = matcher.to; 4290 // greedy, all the way down 4291 while (i < to) { 4292 int ch = Character.codePointAt(seq, i); 4293 if (!predicate.is(ch)) 4294 break; 4295 i += Character.charCount(ch); 4296 n++; 4297 } 4298 if (i >= to) { 4299 matcher.hitEnd = true; 4300 } 4301 while (n >= cmin) { 4302 if (next.match(matcher, i, seq)) 4303 return true; 4304 if (n == cmin) 4305 return false; 4306 // backing off if match fails 4307 int ch = Character.codePointBefore(seq, i); 4308 i -= Character.charCount(ch); 4309 n--; 4310 } 4311 return false; 4312 } 4313 4314 boolean study(TreeInfo info) { 4315 info.minLength += cmin; 4316 if (info.maxValid) { 4317 info.maxLength += MAX_REPS; 4318 } 4319 info.deterministic = false; 4320 return next.study(info); 4321 } 4322 } 4323 4324 static final class BmpCharPropertyGreedy extends CharPropertyGreedy { 4325 4326 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) { 4327 super(bcp, cmin); 4328 } 4329 4330 boolean match(Matcher matcher, int i, CharSequence seq) { 4331 int n = 0; 4332 int to = matcher.to; 4333 while (i < to && predicate.is(seq.charAt(i))) { 4334 i++; n++; 4335 } 4336 if (i >= to) { 4337 matcher.hitEnd = true; 4338 } 4339 while (n >= cmin) { 4340 if (next.match(matcher, i, seq)) 4341 return true; 4342 i--; n--; // backing off if match fails 4343 } 4344 return false; 4345 } 4346 } 4347 4348 /** 4349 * Handles the curly-brace style repetition with a specified minimum and 4350 * maximum occurrences. The * quantifier is handled as a special case. 4351 * This class handles the three types. 4352 */ 4353 static final class Curly extends Node { 4354 Node atom; 4355 Qtype type; 4356 int cmin; 4357 int cmax; 4358 4359 Curly(Node node, int cmin, int cmax, Qtype type) { 4360 this.atom = node; 4361 this.type = type; 4362 this.cmin = cmin; 4363 this.cmax = cmax; 4364 } 4365 boolean match(Matcher matcher, int i, CharSequence seq) { 4366 int j; 4367 for (j = 0; j < cmin; j++) { 4368 if (atom.match(matcher, i, seq)) { 4369 i = matcher.last; 4370 continue; 4371 } 4372 return false; 4373 } 4374 if (type == Qtype.GREEDY) 4375 return match0(matcher, i, j, seq); 4376 else if (type == Qtype.LAZY) 4377 return match1(matcher, i, j, seq); 4378 else 4379 return match2(matcher, i, j, seq); 4380 } 4381 // Greedy match. 4382 // i is the index to start matching at 4383 // j is the number of atoms that have matched 4384 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4385 if (j >= cmax) { 4386 // We have matched the maximum... continue with the rest of 4387 // the regular expression 4388 return next.match(matcher, i, seq); 4389 } 4390 int backLimit = j; 4391 while (atom.match(matcher, i, seq)) { 4392 // k is the length of this match 4393 int k = matcher.last - i; 4394 if (k == 0) // Zero length match 4395 break; 4396 // Move up index and number matched 4397 i = matcher.last; 4398 j++; 4399 // We are greedy so match as many as we can 4400 while (j < cmax) { 4401 if (!atom.match(matcher, i, seq)) 4402 break; 4403 if (i + k != matcher.last) { 4404 if (match0(matcher, matcher.last, j+1, seq)) 4405 return true; 4406 break; 4407 } 4408 i += k; 4409 j++; 4410 } 4411 // Handle backing off if match fails 4412 while (j >= backLimit) { 4413 if (next.match(matcher, i, seq)) 4414 return true; 4415 i -= k; 4416 j--; 4417 } 4418 return false; 4419 } 4420 return next.match(matcher, i, seq); 4421 } 4422 // Reluctant match. At this point, the minimum has been satisfied. 4423 // i is the index to start matching at 4424 // j is the number of atoms that have matched 4425 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4426 for (;;) { 4427 // Try finishing match without consuming any more 4428 if (next.match(matcher, i, seq)) 4429 return true; 4430 // At the maximum, no match found 4431 if (j >= cmax) 4432 return false; 4433 // Okay, must try one more atom 4434 if (!atom.match(matcher, i, seq)) 4435 return false; 4436 // If we haven't moved forward then must break out 4437 if (i == matcher.last) 4438 return false; 4439 // Move up index and number matched 4440 i = matcher.last; 4441 j++; 4442 } 4443 } 4444 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4445 for (; j < cmax; j++) { 4446 if (!atom.match(matcher, i, seq)) 4447 break; 4448 if (i == matcher.last) 4449 break; 4450 i = matcher.last; 4451 } 4452 return next.match(matcher, i, seq); 4453 } 4454 boolean study(TreeInfo info) { 4455 // Save original info 4456 int minL = info.minLength; 4457 int maxL = info.maxLength; 4458 boolean maxV = info.maxValid; 4459 boolean detm = info.deterministic; 4460 info.reset(); 4461 4462 atom.study(info); 4463 4464 int temp = info.minLength * cmin + minL; 4465 if (temp < minL) { 4466 temp = 0xFFFFFFF; // arbitrary large number 4467 } 4468 info.minLength = temp; 4469 4470 if (maxV & info.maxValid) { 4471 temp = info.maxLength * cmax + maxL; 4472 info.maxLength = temp; 4473 if (temp < maxL) { 4474 info.maxValid = false; 4475 } 4476 } else { 4477 info.maxValid = false; 4478 } 4479 4480 if (info.deterministic && cmin == cmax) 4481 info.deterministic = detm; 4482 else 4483 info.deterministic = false; 4484 return next.study(info); 4485 } 4486 } 4487 4488 /** 4489 * Handles the curly-brace style repetition with a specified minimum and 4490 * maximum occurrences in deterministic cases. This is an iterative 4491 * optimization over the Prolog and Loop system which would handle this 4492 * in a recursive way. The * quantifier is handled as a special case. 4493 * If capture is true then this class saves group settings and ensures 4494 * that groups are unset when backing off of a group match. 4495 */ 4496 static final class GroupCurly extends Node { 4497 Node atom; 4498 Qtype type; 4499 int cmin; 4500 int cmax; 4501 int localIndex; 4502 int groupIndex; 4503 boolean capture; 4504 4505 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local, 4506 int group, boolean capture) { 4507 this.atom = node; 4508 this.type = type; 4509 this.cmin = cmin; 4510 this.cmax = cmax; 4511 this.localIndex = local; 4512 this.groupIndex = group; 4513 this.capture = capture; 4514 } 4515 boolean match(Matcher matcher, int i, CharSequence seq) { 4516 int[] groups = matcher.groups; 4517 int[] locals = matcher.locals; 4518 int save0 = locals[localIndex]; 4519 int save1 = 0; 4520 int save2 = 0; 4521 4522 if (capture) { 4523 save1 = groups[groupIndex]; 4524 save2 = groups[groupIndex+1]; 4525 } 4526 4527 // Notify GroupTail there is no need to setup group info 4528 // because it will be set here 4529 locals[localIndex] = -1; 4530 4531 boolean ret = true; 4532 for (int j = 0; j < cmin; j++) { 4533 if (atom.match(matcher, i, seq)) { 4534 if (capture) { 4535 groups[groupIndex] = i; 4536 groups[groupIndex+1] = matcher.last; 4537 } 4538 i = matcher.last; 4539 } else { 4540 ret = false; 4541 break; 4542 } 4543 } 4544 if (ret) { 4545 if (type == Qtype.GREEDY) { 4546 ret = match0(matcher, i, cmin, seq); 4547 } else if (type == Qtype.LAZY) { 4548 ret = match1(matcher, i, cmin, seq); 4549 } else { 4550 ret = match2(matcher, i, cmin, seq); 4551 } 4552 } 4553 if (!ret) { 4554 locals[localIndex] = save0; 4555 if (capture) { 4556 groups[groupIndex] = save1; 4557 groups[groupIndex+1] = save2; 4558 } 4559 } 4560 return ret; 4561 } 4562 // Aggressive group match 4563 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4564 // don't back off passing the starting "j" 4565 int min = j; 4566 int[] groups = matcher.groups; 4567 int save0 = 0; 4568 int save1 = 0; 4569 if (capture) { 4570 save0 = groups[groupIndex]; 4571 save1 = groups[groupIndex+1]; 4572 } 4573 for (;;) { 4574 if (j >= cmax) 4575 break; 4576 if (!atom.match(matcher, i, seq)) 4577 break; 4578 int k = matcher.last - i; 4579 if (k <= 0) { 4580 if (capture) { 4581 groups[groupIndex] = i; 4582 groups[groupIndex+1] = i + k; 4583 } 4584 i = i + k; 4585 break; 4586 } 4587 for (;;) { 4588 if (capture) { 4589 groups[groupIndex] = i; 4590 groups[groupIndex+1] = i + k; 4591 } 4592 i = i + k; 4593 if (++j >= cmax) 4594 break; 4595 if (!atom.match(matcher, i, seq)) 4596 break; 4597 if (i + k != matcher.last) { 4598 if (match0(matcher, i, j, seq)) 4599 return true; 4600 break; 4601 } 4602 } 4603 while (j > min) { 4604 if (next.match(matcher, i, seq)) { 4605 if (capture) { 4606 groups[groupIndex+1] = i; 4607 groups[groupIndex] = i - k; 4608 } 4609 return true; 4610 } 4611 // backing off 4612 i = i - k; 4613 if (capture) { 4614 groups[groupIndex+1] = i; 4615 groups[groupIndex] = i - k; 4616 } 4617 j--; 4618 4619 } 4620 break; 4621 } 4622 if (capture) { 4623 groups[groupIndex] = save0; 4624 groups[groupIndex+1] = save1; 4625 } 4626 return next.match(matcher, i, seq); 4627 } 4628 // Reluctant matching 4629 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4630 for (;;) { 4631 if (next.match(matcher, i, seq)) 4632 return true; 4633 if (j >= cmax) 4634 return false; 4635 if (!atom.match(matcher, i, seq)) 4636 return false; 4637 if (i == matcher.last) 4638 return false; 4639 if (capture) { 4640 matcher.groups[groupIndex] = i; 4641 matcher.groups[groupIndex+1] = matcher.last; 4642 } 4643 i = matcher.last; 4644 j++; 4645 } 4646 } 4647 // Possessive matching 4648 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4649 for (; j < cmax; j++) { 4650 if (!atom.match(matcher, i, seq)) { 4651 break; 4652 } 4653 if (capture) { 4654 matcher.groups[groupIndex] = i; 4655 matcher.groups[groupIndex+1] = matcher.last; 4656 } 4657 if (i == matcher.last) { 4658 break; 4659 } 4660 i = matcher.last; 4661 } 4662 return next.match(matcher, i, seq); 4663 } 4664 boolean study(TreeInfo info) { 4665 // Save original info 4666 int minL = info.minLength; 4667 int maxL = info.maxLength; 4668 boolean maxV = info.maxValid; 4669 boolean detm = info.deterministic; 4670 info.reset(); 4671 4672 atom.study(info); 4673 4674 int temp = info.minLength * cmin + minL; 4675 if (temp < minL) { 4676 temp = 0xFFFFFFF; // Arbitrary large number 4677 } 4678 info.minLength = temp; 4679 4680 if (maxV & info.maxValid) { 4681 temp = info.maxLength * cmax + maxL; 4682 info.maxLength = temp; 4683 if (temp < maxL) { 4684 info.maxValid = false; 4685 } 4686 } else { 4687 info.maxValid = false; 4688 } 4689 4690 if (info.deterministic && cmin == cmax) { 4691 info.deterministic = detm; 4692 } else { 4693 info.deterministic = false; 4694 } 4695 return next.study(info); 4696 } 4697 } 4698 4699 /** 4700 * A Guard node at the end of each atom node in a Branch. It 4701 * serves the purpose of chaining the "match" operation to 4702 * "next" but not the "study", so we can collect the TreeInfo 4703 * of each atom node without including the TreeInfo of the 4704 * "next". 4705 */ 4706 static final class BranchConn extends Node { 4707 BranchConn() {}; 4708 boolean match(Matcher matcher, int i, CharSequence seq) { 4709 return next.match(matcher, i, seq); 4710 } 4711 boolean study(TreeInfo info) { 4712 return info.deterministic; 4713 } 4714 } 4715 4716 /** 4717 * Handles the branching of alternations. Note this is also used for 4718 * the ? quantifier to branch between the case where it matches once 4719 * and where it does not occur. 4720 */ 4721 static final class Branch extends Node { 4722 Node[] atoms = new Node[2]; 4723 int size = 2; 4724 Node conn; 4725 Branch(Node first, Node second, Node branchConn) { 4726 conn = branchConn; 4727 atoms[0] = first; 4728 atoms[1] = second; 4729 } 4730 4731 void add(Node node) { 4732 if (size >= atoms.length) { 4733 Node[] tmp = new Node[atoms.length*2]; 4734 System.arraycopy(atoms, 0, tmp, 0, atoms.length); 4735 atoms = tmp; 4736 } 4737 atoms[size++] = node; 4738 } 4739 4740 boolean match(Matcher matcher, int i, CharSequence seq) { 4741 for (int n = 0; n < size; n++) { 4742 if (atoms[n] == null) { 4743 if (conn.next.match(matcher, i, seq)) 4744 return true; 4745 } else if (atoms[n].match(matcher, i, seq)) { 4746 return true; 4747 } 4748 } 4749 return false; 4750 } 4751 4752 boolean study(TreeInfo info) { 4753 int minL = info.minLength; 4754 int maxL = info.maxLength; 4755 boolean maxV = info.maxValid; 4756 4757 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num 4758 int maxL2 = -1; 4759 for (int n = 0; n < size; n++) { 4760 info.reset(); 4761 if (atoms[n] != null) 4762 atoms[n].study(info); 4763 minL2 = Math.min(minL2, info.minLength); 4764 maxL2 = Math.max(maxL2, info.maxLength); 4765 maxV = (maxV & info.maxValid); 4766 } 4767 4768 minL += minL2; 4769 maxL += maxL2; 4770 4771 info.reset(); 4772 conn.next.study(info); 4773 4774 info.minLength += minL; 4775 info.maxLength += maxL; 4776 info.maxValid &= maxV; 4777 info.deterministic = false; 4778 return false; 4779 } 4780 } 4781 4782 /** 4783 * The GroupHead saves the location where the group begins in the locals 4784 * and restores them when the match is done. 4785 * 4786 * The matchRef is used when a reference to this group is accessed later 4787 * in the expression. The locals will have a negative value in them to 4788 * indicate that we do not want to unset the group if the reference 4789 * doesn't match. 4790 */ 4791 static final class GroupHead extends Node { 4792 int localIndex; 4793 GroupTail tail; // for debug/print only, match does not need to know 4794 GroupHead(int localCount) { 4795 localIndex = localCount; 4796 } 4797 boolean match(Matcher matcher, int i, CharSequence seq) { 4798 int save = matcher.locals[localIndex]; 4799 matcher.locals[localIndex] = i; 4800 boolean ret = next.match(matcher, i, seq); 4801 matcher.locals[localIndex] = save; 4802 return ret; 4803 } 4804 boolean matchRef(Matcher matcher, int i, CharSequence seq) { 4805 int save = matcher.locals[localIndex]; 4806 matcher.locals[localIndex] = ~i; // HACK 4807 boolean ret = next.match(matcher, i, seq); 4808 matcher.locals[localIndex] = save; 4809 return ret; 4810 } 4811 } 4812 4813 /** 4814 * Recursive reference to a group in the regular expression. It calls 4815 * matchRef because if the reference fails to match we would not unset 4816 * the group. 4817 */ 4818 static final class GroupRef extends Node { 4819 GroupHead head; 4820 GroupRef(GroupHead head) { 4821 this.head = head; 4822 } 4823 boolean match(Matcher matcher, int i, CharSequence seq) { 4824 return head.matchRef(matcher, i, seq) 4825 && next.match(matcher, matcher.last, seq); 4826 } 4827 boolean study(TreeInfo info) { 4828 info.maxValid = false; 4829 info.deterministic = false; 4830 return next.study(info); 4831 } 4832 } 4833 4834 /** 4835 * The GroupTail handles the setting of group beginning and ending 4836 * locations when groups are successfully matched. It must also be able to 4837 * unset groups that have to be backed off of. 4838 * 4839 * The GroupTail node is also used when a previous group is referenced, 4840 * and in that case no group information needs to be set. 4841 */ 4842 static final class GroupTail extends Node { 4843 int localIndex; 4844 int groupIndex; 4845 GroupTail(int localCount, int groupCount) { 4846 localIndex = localCount; 4847 groupIndex = groupCount + groupCount; 4848 } 4849 boolean match(Matcher matcher, int i, CharSequence seq) { 4850 int tmp = matcher.locals[localIndex]; 4851 if (tmp >= 0) { // This is the normal group case. 4852 // Save the group so we can unset it if it 4853 // backs off of a match. 4854 int groupStart = matcher.groups[groupIndex]; 4855 int groupEnd = matcher.groups[groupIndex+1]; 4856 4857 matcher.groups[groupIndex] = tmp; 4858 matcher.groups[groupIndex+1] = i; 4859 if (next.match(matcher, i, seq)) { 4860 return true; 4861 } 4862 matcher.groups[groupIndex] = groupStart; 4863 matcher.groups[groupIndex+1] = groupEnd; 4864 return false; 4865 } else { 4866 // This is a group reference case. We don't need to save any 4867 // group info because it isn't really a group. 4868 matcher.last = i; 4869 return true; 4870 } 4871 } 4872 } 4873 4874 /** 4875 * This sets up a loop to handle a recursive quantifier structure. 4876 */ 4877 static final class Prolog extends Node { 4878 Loop loop; 4879 Prolog(Loop loop) { 4880 this.loop = loop; 4881 } 4882 boolean match(Matcher matcher, int i, CharSequence seq) { 4883 return loop.matchInit(matcher, i, seq); 4884 } 4885 boolean study(TreeInfo info) { 4886 return loop.study(info); 4887 } 4888 } 4889 4890 /** 4891 * Handles the repetition count for a greedy Curly. The matchInit 4892 * is called from the Prolog to save the index of where the group 4893 * beginning is stored. A zero length group check occurs in the 4894 * normal match but is skipped in the matchInit. 4895 */ 4896 static class Loop extends Node { 4897 Node body; 4898 int countIndex; // local count index in matcher locals 4899 int beginIndex; // group beginning index 4900 int cmin, cmax; 4901 int posIndex; 4902 Loop(int countIndex, int beginIndex) { 4903 this.countIndex = countIndex; 4904 this.beginIndex = beginIndex; 4905 this.posIndex = -1; 4906 } 4907 boolean match(Matcher matcher, int i, CharSequence seq) { 4908 // Avoid infinite loop in zero-length case. 4909 if (i > matcher.locals[beginIndex]) { 4910 int count = matcher.locals[countIndex]; 4911 4912 // This block is for before we reach the minimum 4913 // iterations required for the loop to match 4914 if (count < cmin) { 4915 matcher.locals[countIndex] = count + 1; 4916 boolean b = body.match(matcher, i, seq); 4917 // If match failed we must backtrack, so 4918 // the loop count should NOT be incremented 4919 if (!b) 4920 matcher.locals[countIndex] = count; 4921 // Return success or failure since we are under 4922 // minimum 4923 return b; 4924 } 4925 // This block is for after we have the minimum 4926 // iterations required for the loop to match 4927 if (count < cmax) { 4928 // Let's check if we have already tried and failed 4929 // at this starting position "i" in the past. 4930 // If yes, then just return false wihtout trying 4931 // again, to stop the exponential backtracking. 4932 if (posIndex != -1 && 4933 matcher.localsPos[posIndex].contains(i)) { 4934 return next.match(matcher, i, seq); 4935 } 4936 matcher.locals[countIndex] = count + 1; 4937 boolean b = body.match(matcher, i, seq); 4938 // If match failed we must backtrack, so 4939 // the loop count should NOT be incremented 4940 if (b) 4941 return true; 4942 matcher.locals[countIndex] = count; 4943 // save the failed position 4944 if (posIndex != -1) { 4945 matcher.localsPos[posIndex].add(i); 4946 } 4947 } 4948 } 4949 return next.match(matcher, i, seq); 4950 } 4951 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4952 int save = matcher.locals[countIndex]; 4953 boolean ret = false; 4954 if (posIndex != -1 && matcher.localsPos[posIndex] == null) { 4955 matcher.localsPos[posIndex] = new IntHashSet(); 4956 } 4957 if (0 < cmin) { 4958 matcher.locals[countIndex] = 1; 4959 ret = body.match(matcher, i, seq); 4960 } else if (0 < cmax) { 4961 matcher.locals[countIndex] = 1; 4962 ret = body.match(matcher, i, seq); 4963 if (ret == false) 4964 ret = next.match(matcher, i, seq); 4965 } else { 4966 ret = next.match(matcher, i, seq); 4967 } 4968 matcher.locals[countIndex] = save; 4969 return ret; 4970 } 4971 boolean study(TreeInfo info) { 4972 info.maxValid = false; 4973 info.deterministic = false; 4974 return false; 4975 } 4976 } 4977 4978 /** 4979 * Handles the repetition count for a reluctant Curly. The matchInit 4980 * is called from the Prolog to save the index of where the group 4981 * beginning is stored. A zero length group check occurs in the 4982 * normal match but is skipped in the matchInit. 4983 */ 4984 static final class LazyLoop extends Loop { 4985 LazyLoop(int countIndex, int beginIndex) { 4986 super(countIndex, beginIndex); 4987 } 4988 boolean match(Matcher matcher, int i, CharSequence seq) { 4989 // Check for zero length group 4990 if (i > matcher.locals[beginIndex]) { 4991 int count = matcher.locals[countIndex]; 4992 if (count < cmin) { 4993 matcher.locals[countIndex] = count + 1; 4994 boolean result = body.match(matcher, i, seq); 4995 // If match failed we must backtrack, so 4996 // the loop count should NOT be incremented 4997 if (!result) 4998 matcher.locals[countIndex] = count; 4999 return result; 5000 } 5001 if (next.match(matcher, i, seq)) 5002 return true; 5003 if (count < cmax) { 5004 matcher.locals[countIndex] = count + 1; 5005 boolean result = body.match(matcher, i, seq); 5006 // If match failed we must backtrack, so 5007 // the loop count should NOT be incremented 5008 if (!result) 5009 matcher.locals[countIndex] = count; 5010 return result; 5011 } 5012 return false; 5013 } 5014 return next.match(matcher, i, seq); 5015 } 5016 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 5017 int save = matcher.locals[countIndex]; 5018 boolean ret = false; 5019 if (0 < cmin) { 5020 matcher.locals[countIndex] = 1; 5021 ret = body.match(matcher, i, seq); 5022 } else if (next.match(matcher, i, seq)) { 5023 ret = true; 5024 } else if (0 < cmax) { 5025 matcher.locals[countIndex] = 1; 5026 ret = body.match(matcher, i, seq); 5027 } 5028 matcher.locals[countIndex] = save; 5029 return ret; 5030 } 5031 boolean study(TreeInfo info) { 5032 info.maxValid = false; 5033 info.deterministic = false; 5034 return false; 5035 } 5036 } 5037 5038 /** 5039 * Refers to a group in the regular expression. Attempts to match 5040 * whatever the group referred to last matched. 5041 */ 5042 static class BackRef extends Node { 5043 int groupIndex; 5044 BackRef(int groupCount) { 5045 super(); 5046 groupIndex = groupCount + groupCount; 5047 } 5048 boolean match(Matcher matcher, int i, CharSequence seq) { 5049 int j = matcher.groups[groupIndex]; 5050 int k = matcher.groups[groupIndex+1]; 5051 5052 int groupSize = k - j; 5053 // If the referenced group didn't match, neither can this 5054 if (j < 0) 5055 return false; 5056 5057 // If there isn't enough input left no match 5058 if (i + groupSize > matcher.to) { 5059 matcher.hitEnd = true; 5060 return false; 5061 } 5062 // Check each new char to make sure it matches what the group 5063 // referenced matched last time around 5064 for (int index=0; index<groupSize; index++) 5065 if (seq.charAt(i+index) != seq.charAt(j+index)) 5066 return false; 5067 5068 return next.match(matcher, i+groupSize, seq); 5069 } 5070 boolean study(TreeInfo info) { 5071 info.maxValid = false; 5072 return next.study(info); 5073 } 5074 } 5075 5076 static class CIBackRef extends Node { 5077 int groupIndex; 5078 boolean doUnicodeCase; 5079 CIBackRef(int groupCount, boolean doUnicodeCase) { 5080 super(); 5081 groupIndex = groupCount + groupCount; 5082 this.doUnicodeCase = doUnicodeCase; 5083 } 5084 boolean match(Matcher matcher, int i, CharSequence seq) { 5085 int j = matcher.groups[groupIndex]; 5086 int k = matcher.groups[groupIndex+1]; 5087 5088 int groupSize = k - j; 5089 5090 // If the referenced group didn't match, neither can this 5091 if (j < 0) 5092 return false; 5093 5094 // If there isn't enough input left no match 5095 if (i + groupSize > matcher.to) { 5096 matcher.hitEnd = true; 5097 return false; 5098 } 5099 5100 // Check each new char to make sure it matches what the group 5101 // referenced matched last time around 5102 int x = i; 5103 for (int index=0; index<groupSize; index++) { 5104 int c1 = Character.codePointAt(seq, x); 5105 int c2 = Character.codePointAt(seq, j); 5106 if (c1 != c2) { 5107 if (doUnicodeCase) { 5108 int cc1 = Character.toUpperCase(c1); 5109 int cc2 = Character.toUpperCase(c2); 5110 if (cc1 != cc2 && 5111 Character.toLowerCase(cc1) != 5112 Character.toLowerCase(cc2)) 5113 return false; 5114 } else { 5115 if (ASCII.toLower(c1) != ASCII.toLower(c2)) 5116 return false; 5117 } 5118 } 5119 x += Character.charCount(c1); 5120 j += Character.charCount(c2); 5121 } 5122 5123 return next.match(matcher, i+groupSize, seq); 5124 } 5125 boolean study(TreeInfo info) { 5126 info.maxValid = false; 5127 return next.study(info); 5128 } 5129 } 5130 5131 /** 5132 * Searches until the next instance of its atom. This is useful for 5133 * finding the atom efficiently without passing an instance of it 5134 * (greedy problem) and without a lot of wasted search time (reluctant 5135 * problem). 5136 */ 5137 static final class First extends Node { 5138 Node atom; 5139 First(Node node) { 5140 this.atom = BnM.optimize(node); 5141 } 5142 boolean match(Matcher matcher, int i, CharSequence seq) { 5143 if (atom instanceof BnM) { 5144 return atom.match(matcher, i, seq) 5145 && next.match(matcher, matcher.last, seq); 5146 } 5147 for (;;) { 5148 if (i > matcher.to) { 5149 matcher.hitEnd = true; 5150 return false; 5151 } 5152 if (atom.match(matcher, i, seq)) { 5153 return next.match(matcher, matcher.last, seq); 5154 } 5155 i += countChars(seq, i, 1); 5156 matcher.first++; 5157 } 5158 } 5159 boolean study(TreeInfo info) { 5160 atom.study(info); 5161 info.maxValid = false; 5162 info.deterministic = false; 5163 return next.study(info); 5164 } 5165 } 5166 5167 static final class Conditional extends Node { 5168 Node cond, yes, not; 5169 Conditional(Node cond, Node yes, Node not) { 5170 this.cond = cond; 5171 this.yes = yes; 5172 this.not = not; 5173 } 5174 boolean match(Matcher matcher, int i, CharSequence seq) { 5175 if (cond.match(matcher, i, seq)) { 5176 return yes.match(matcher, i, seq); 5177 } else { 5178 return not.match(matcher, i, seq); 5179 } 5180 } 5181 boolean study(TreeInfo info) { 5182 int minL = info.minLength; 5183 int maxL = info.maxLength; 5184 boolean maxV = info.maxValid; 5185 info.reset(); 5186 yes.study(info); 5187 5188 int minL2 = info.minLength; 5189 int maxL2 = info.maxLength; 5190 boolean maxV2 = info.maxValid; 5191 info.reset(); 5192 not.study(info); 5193 5194 info.minLength = minL + Math.min(minL2, info.minLength); 5195 info.maxLength = maxL + Math.max(maxL2, info.maxLength); 5196 info.maxValid = (maxV & maxV2 & info.maxValid); 5197 info.deterministic = false; 5198 return next.study(info); 5199 } 5200 } 5201 5202 /** 5203 * Zero width positive lookahead. 5204 */ 5205 static final class Pos extends Node { 5206 Node cond; 5207 Pos(Node cond) { 5208 this.cond = cond; 5209 } 5210 boolean match(Matcher matcher, int i, CharSequence seq) { 5211 int savedTo = matcher.to; 5212 boolean conditionMatched = false; 5213 5214 // Relax transparent region boundaries for lookahead 5215 if (matcher.transparentBounds) 5216 matcher.to = matcher.getTextLength(); 5217 try { 5218 conditionMatched = cond.match(matcher, i, seq); 5219 } finally { 5220 // Reinstate region boundaries 5221 matcher.to = savedTo; 5222 } 5223 return conditionMatched && next.match(matcher, i, seq); 5224 } 5225 } 5226 5227 /** 5228 * Zero width negative lookahead. 5229 */ 5230 static final class Neg extends Node { 5231 Node cond; 5232 Neg(Node cond) { 5233 this.cond = cond; 5234 } 5235 boolean match(Matcher matcher, int i, CharSequence seq) { 5236 int savedTo = matcher.to; 5237 boolean conditionMatched = false; 5238 5239 // Relax transparent region boundaries for lookahead 5240 if (matcher.transparentBounds) 5241 matcher.to = matcher.getTextLength(); 5242 try { 5243 if (i < matcher.to) { 5244 conditionMatched = !cond.match(matcher, i, seq); 5245 } else { 5246 // If a negative lookahead succeeds then more input 5247 // could cause it to fail! 5248 matcher.requireEnd = true; 5249 conditionMatched = !cond.match(matcher, i, seq); 5250 } 5251 } finally { 5252 // Reinstate region boundaries 5253 matcher.to = savedTo; 5254 } 5255 return conditionMatched && next.match(matcher, i, seq); 5256 } 5257 } 5258 5259 /** 5260 * For use with lookbehinds; matches the position where the lookbehind 5261 * was encountered. 5262 */ 5263 static Node lookbehindEnd = new Node() { 5264 boolean match(Matcher matcher, int i, CharSequence seq) { 5265 return i == matcher.lookbehindTo; 5266 } 5267 }; 5268 5269 /** 5270 * Zero width positive lookbehind. 5271 */ 5272 static class Behind extends Node { 5273 Node cond; 5274 int rmax, rmin; 5275 Behind(Node cond, int rmax, int rmin) { 5276 this.cond = cond; 5277 this.rmax = rmax; 5278 this.rmin = rmin; 5279 } 5280 5281 boolean match(Matcher matcher, int i, CharSequence seq) { 5282 int savedFrom = matcher.from; 5283 boolean conditionMatched = false; 5284 int startIndex = (!matcher.transparentBounds) ? 5285 matcher.from : 0; 5286 int from = Math.max(i - rmax, startIndex); 5287 // Set end boundary 5288 int savedLBT = matcher.lookbehindTo; 5289 matcher.lookbehindTo = i; 5290 // Relax transparent region boundaries for lookbehind 5291 if (matcher.transparentBounds) 5292 matcher.from = 0; 5293 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5294 conditionMatched = cond.match(matcher, j, seq); 5295 } 5296 matcher.from = savedFrom; 5297 matcher.lookbehindTo = savedLBT; 5298 return conditionMatched && next.match(matcher, i, seq); 5299 } 5300 } 5301 5302 /** 5303 * Zero width positive lookbehind, including supplementary 5304 * characters or unpaired surrogates. 5305 */ 5306 static final class BehindS extends Behind { 5307 BehindS(Node cond, int rmax, int rmin) { 5308 super(cond, rmax, rmin); 5309 } 5310 boolean match(Matcher matcher, int i, CharSequence seq) { 5311 int rmaxChars = countChars(seq, i, -rmax); 5312 int rminChars = countChars(seq, i, -rmin); 5313 int savedFrom = matcher.from; 5314 int startIndex = (!matcher.transparentBounds) ? 5315 matcher.from : 0; 5316 boolean conditionMatched = false; 5317 int from = Math.max(i - rmaxChars, startIndex); 5318 // Set end boundary 5319 int savedLBT = matcher.lookbehindTo; 5320 matcher.lookbehindTo = i; 5321 // Relax transparent region boundaries for lookbehind 5322 if (matcher.transparentBounds) 5323 matcher.from = 0; 5324 5325 for (int j = i - rminChars; 5326 !conditionMatched && j >= from; 5327 j -= j>from ? countChars(seq, j, -1) : 1) { 5328 conditionMatched = cond.match(matcher, j, seq); 5329 } 5330 matcher.from = savedFrom; 5331 matcher.lookbehindTo = savedLBT; 5332 return conditionMatched && next.match(matcher, i, seq); 5333 } 5334 } 5335 5336 /** 5337 * Zero width negative lookbehind. 5338 */ 5339 static class NotBehind extends Node { 5340 Node cond; 5341 int rmax, rmin; 5342 NotBehind(Node cond, int rmax, int rmin) { 5343 this.cond = cond; 5344 this.rmax = rmax; 5345 this.rmin = rmin; 5346 } 5347 5348 boolean match(Matcher matcher, int i, CharSequence seq) { 5349 int savedLBT = matcher.lookbehindTo; 5350 int savedFrom = matcher.from; 5351 boolean conditionMatched = false; 5352 int startIndex = (!matcher.transparentBounds) ? 5353 matcher.from : 0; 5354 int from = Math.max(i - rmax, startIndex); 5355 matcher.lookbehindTo = i; 5356 // Relax transparent region boundaries for lookbehind 5357 if (matcher.transparentBounds) 5358 matcher.from = 0; 5359 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5360 conditionMatched = cond.match(matcher, j, seq); 5361 } 5362 // Reinstate region boundaries 5363 matcher.from = savedFrom; 5364 matcher.lookbehindTo = savedLBT; 5365 return !conditionMatched && next.match(matcher, i, seq); 5366 } 5367 } 5368 5369 /** 5370 * Zero width negative lookbehind, including supplementary 5371 * characters or unpaired surrogates. 5372 */ 5373 static final class NotBehindS extends NotBehind { 5374 NotBehindS(Node cond, int rmax, int rmin) { 5375 super(cond, rmax, rmin); 5376 } 5377 boolean match(Matcher matcher, int i, CharSequence seq) { 5378 int rmaxChars = countChars(seq, i, -rmax); 5379 int rminChars = countChars(seq, i, -rmin); 5380 int savedFrom = matcher.from; 5381 int savedLBT = matcher.lookbehindTo; 5382 boolean conditionMatched = false; 5383 int startIndex = (!matcher.transparentBounds) ? 5384 matcher.from : 0; 5385 int from = Math.max(i - rmaxChars, startIndex); 5386 matcher.lookbehindTo = i; 5387 // Relax transparent region boundaries for lookbehind 5388 if (matcher.transparentBounds) 5389 matcher.from = 0; 5390 for (int j = i - rminChars; 5391 !conditionMatched && j >= from; 5392 j -= j>from ? countChars(seq, j, -1) : 1) { 5393 conditionMatched = cond.match(matcher, j, seq); 5394 } 5395 //Reinstate region boundaries 5396 matcher.from = savedFrom; 5397 matcher.lookbehindTo = savedLBT; 5398 return !conditionMatched && next.match(matcher, i, seq); 5399 } 5400 } 5401 5402 /** 5403 * Handles word boundaries. Includes a field to allow this one class to 5404 * deal with the different types of word boundaries we can match. The word 5405 * characters include underscores, letters, and digits. Non spacing marks 5406 * can are also part of a word if they have a base character, otherwise 5407 * they are ignored for purposes of finding word boundaries. 5408 */ 5409 static final class Bound extends Node { 5410 static int LEFT = 0x1; 5411 static int RIGHT= 0x2; 5412 static int BOTH = 0x3; 5413 static int NONE = 0x4; 5414 int type; 5415 boolean useUWORD; 5416 Bound(int n, boolean useUWORD) { 5417 type = n; 5418 this.useUWORD = useUWORD; 5419 } 5420 5421 boolean isWord(int ch) { 5422 return useUWORD ? CharPredicates.WORD().is(ch) 5423 : (ch == '_' || Character.isLetterOrDigit(ch)); 5424 } 5425 5426 int check(Matcher matcher, int i, CharSequence seq) { 5427 int ch; 5428 boolean left = false; 5429 int startIndex = matcher.from; 5430 int endIndex = matcher.to; 5431 if (matcher.transparentBounds) { 5432 startIndex = 0; 5433 endIndex = matcher.getTextLength(); 5434 } 5435 if (i > startIndex) { 5436 ch = Character.codePointBefore(seq, i); 5437 left = (isWord(ch) || 5438 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5439 && hasBaseCharacter(matcher, i-1, seq))); 5440 } 5441 boolean right = false; 5442 if (i < endIndex) { 5443 ch = Character.codePointAt(seq, i); 5444 right = (isWord(ch) || 5445 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5446 && hasBaseCharacter(matcher, i, seq))); 5447 } else { 5448 // Tried to access char past the end 5449 matcher.hitEnd = true; 5450 // The addition of another char could wreck a boundary 5451 matcher.requireEnd = true; 5452 } 5453 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE); 5454 } 5455 boolean match(Matcher matcher, int i, CharSequence seq) { 5456 return (check(matcher, i, seq) & type) > 0 5457 && next.match(matcher, i, seq); 5458 } 5459 } 5460 5461 /** 5462 * Non spacing marks only count as word characters in bounds calculations 5463 * if they have a base character. 5464 */ 5465 private static boolean hasBaseCharacter(Matcher matcher, int i, 5466 CharSequence seq) 5467 { 5468 int start = (!matcher.transparentBounds) ? 5469 matcher.from : 0; 5470 for (int x=i; x >= start; x--) { 5471 int ch = Character.codePointAt(seq, x); 5472 if (Character.isLetterOrDigit(ch)) 5473 return true; 5474 if (Character.getType(ch) == Character.NON_SPACING_MARK) 5475 continue; 5476 return false; 5477 } 5478 return false; 5479 } 5480 5481 /** 5482 * Attempts to match a slice in the input using the Boyer-Moore string 5483 * matching algorithm. The algorithm is based on the idea that the 5484 * pattern can be shifted farther ahead in the search text if it is 5485 * matched right to left. 5486 * <p> 5487 * The pattern is compared to the input one character at a time, from 5488 * the rightmost character in the pattern to the left. If the characters 5489 * all match the pattern has been found. If a character does not match, 5490 * the pattern is shifted right a distance that is the maximum of two 5491 * functions, the bad character shift and the good suffix shift. This 5492 * shift moves the attempted match position through the input more 5493 * quickly than a naive one position at a time check. 5494 * <p> 5495 * The bad character shift is based on the character from the text that 5496 * did not match. If the character does not appear in the pattern, the 5497 * pattern can be shifted completely beyond the bad character. If the 5498 * character does occur in the pattern, the pattern can be shifted to 5499 * line the pattern up with the next occurrence of that character. 5500 * <p> 5501 * The good suffix shift is based on the idea that some subset on the right 5502 * side of the pattern has matched. When a bad character is found, the 5503 * pattern can be shifted right by the pattern length if the subset does 5504 * not occur again in pattern, or by the amount of distance to the 5505 * next occurrence of the subset in the pattern. 5506 * 5507 * Boyer-Moore search methods adapted from code by Amy Yu. 5508 */ 5509 static class BnM extends Node { 5510 int[] buffer; 5511 int[] lastOcc; 5512 int[] optoSft; 5513 5514 /** 5515 * Pre calculates arrays needed to generate the bad character 5516 * shift and the good suffix shift. Only the last seven bits 5517 * are used to see if chars match; This keeps the tables small 5518 * and covers the heavily used ASCII range, but occasionally 5519 * results in an aliased match for the bad character shift. 5520 */ 5521 static Node optimize(Node node) { 5522 if (!(node instanceof Slice)) { 5523 return node; 5524 } 5525 5526 int[] src = ((Slice) node).buffer; 5527 int patternLength = src.length; 5528 // The BM algorithm requires a bit of overhead; 5529 // If the pattern is short don't use it, since 5530 // a shift larger than the pattern length cannot 5531 // be used anyway. 5532 if (patternLength < 4) { 5533 return node; 5534 } 5535 int i, j, k; 5536 int[] lastOcc = new int[128]; 5537 int[] optoSft = new int[patternLength]; 5538 // Precalculate part of the bad character shift 5539 // It is a table for where in the pattern each 5540 // lower 7-bit value occurs 5541 for (i = 0; i < patternLength; i++) { 5542 lastOcc[src[i]&0x7F] = i + 1; 5543 } 5544 // Precalculate the good suffix shift 5545 // i is the shift amount being considered 5546 NEXT: for (i = patternLength; i > 0; i--) { 5547 // j is the beginning index of suffix being considered 5548 for (j = patternLength - 1; j >= i; j--) { 5549 // Testing for good suffix 5550 if (src[j] == src[j-i]) { 5551 // src[j..len] is a good suffix 5552 optoSft[j-1] = i; 5553 } else { 5554 // No match. The array has already been 5555 // filled up with correct values before. 5556 continue NEXT; 5557 } 5558 } 5559 // This fills up the remaining of optoSft 5560 // any suffix can not have larger shift amount 5561 // then its sub-suffix. Why??? 5562 while (j > 0) { 5563 optoSft[--j] = i; 5564 } 5565 } 5566 // Set the guard value because of unicode compression 5567 optoSft[patternLength-1] = 1; 5568 if (node instanceof SliceS) 5569 return new BnMS(src, lastOcc, optoSft, node.next); 5570 return new BnM(src, lastOcc, optoSft, node.next); 5571 } 5572 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5573 this.buffer = src; 5574 this.lastOcc = lastOcc; 5575 this.optoSft = optoSft; 5576 this.next = next; 5577 } 5578 boolean match(Matcher matcher, int i, CharSequence seq) { 5579 int[] src = buffer; 5580 int patternLength = src.length; 5581 int last = matcher.to - patternLength; 5582 5583 // Loop over all possible match positions in text 5584 NEXT: while (i <= last) { 5585 // Loop over pattern from right to left 5586 for (int j = patternLength - 1; j >= 0; j--) { 5587 int ch = seq.charAt(i+j); 5588 if (ch != src[j]) { 5589 // Shift search to the right by the maximum of the 5590 // bad character shift and the good suffix shift 5591 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]); 5592 continue NEXT; 5593 } 5594 } 5595 // Entire pattern matched starting at i 5596 matcher.first = i; 5597 boolean ret = next.match(matcher, i + patternLength, seq); 5598 if (ret) { 5599 matcher.first = i; 5600 matcher.groups[0] = matcher.first; 5601 matcher.groups[1] = matcher.last; 5602 return true; 5603 } 5604 i++; 5605 } 5606 // BnM is only used as the leading node in the unanchored case, 5607 // and it replaced its Start() which always searches to the end 5608 // if it doesn't find what it's looking for, so hitEnd is true. 5609 matcher.hitEnd = true; 5610 return false; 5611 } 5612 boolean study(TreeInfo info) { 5613 info.minLength += buffer.length; 5614 info.maxValid = false; 5615 return next.study(info); 5616 } 5617 } 5618 5619 /** 5620 * Supplementary support version of BnM(). Unpaired surrogates are 5621 * also handled by this class. 5622 */ 5623 static final class BnMS extends BnM { 5624 int lengthInChars; 5625 5626 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5627 super(src, lastOcc, optoSft, next); 5628 for (int cp : buffer) { 5629 lengthInChars += Character.charCount(cp); 5630 } 5631 } 5632 boolean match(Matcher matcher, int i, CharSequence seq) { 5633 int[] src = buffer; 5634 int patternLength = src.length; 5635 int last = matcher.to - lengthInChars; 5636 5637 // Loop over all possible match positions in text 5638 NEXT: while (i <= last) { 5639 // Loop over pattern from right to left 5640 int ch; 5641 for (int j = countChars(seq, i, patternLength), x = patternLength - 1; 5642 j > 0; j -= Character.charCount(ch), x--) { 5643 ch = Character.codePointBefore(seq, i+j); 5644 if (ch != src[x]) { 5645 // Shift search to the right by the maximum of the 5646 // bad character shift and the good suffix shift 5647 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]); 5648 i += countChars(seq, i, n); 5649 continue NEXT; 5650 } 5651 } 5652 // Entire pattern matched starting at i 5653 matcher.first = i; 5654 boolean ret = next.match(matcher, i + lengthInChars, seq); 5655 if (ret) { 5656 matcher.first = i; 5657 matcher.groups[0] = matcher.first; 5658 matcher.groups[1] = matcher.last; 5659 return true; 5660 } 5661 i += countChars(seq, i, 1); 5662 } 5663 matcher.hitEnd = true; 5664 return false; 5665 } 5666 } 5667 5668 @FunctionalInterface 5669 static interface CharPredicate { 5670 boolean is(int ch); 5671 5672 default CharPredicate and(CharPredicate p) { 5673 return ch -> is(ch) && p.is(ch); 5674 } 5675 default CharPredicate union(CharPredicate p) { 5676 return ch -> is(ch) || p.is(ch); 5677 } 5678 default CharPredicate union(CharPredicate p1, 5679 CharPredicate p2 ) { 5680 return ch -> is(ch) || p1.is(ch) || p2.is(ch); 5681 } 5682 default CharPredicate negate() { 5683 return ch -> !is(ch); 5684 } 5685 } 5686 5687 static interface BmpCharPredicate extends CharPredicate { 5688 5689 default CharPredicate and(CharPredicate p) { 5690 if(p instanceof BmpCharPredicate) 5691 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch)); 5692 return ch -> is(ch) && p.is(ch); 5693 } 5694 default CharPredicate union(CharPredicate p) { 5695 if (p instanceof BmpCharPredicate) 5696 return (BmpCharPredicate)(ch -> is(ch) || p.is(ch)); 5697 return ch -> is(ch) || p.is(ch); 5698 } 5699 static CharPredicate union(CharPredicate... predicates) { 5700 CharPredicate cp = ch -> { 5701 for (CharPredicate p : predicates) { 5702 if (!p.is(ch)) 5703 return false; 5704 } 5705 return true; 5706 }; 5707 for (CharPredicate p : predicates) { 5708 if (! (p instanceof BmpCharPredicate)) 5709 return cp; 5710 } 5711 return (BmpCharPredicate)cp; 5712 } 5713 } 5714 5715 /** 5716 * matches a Perl vertical whitespace 5717 */ 5718 static BmpCharPredicate VertWS() { 5719 return cp -> (cp >= 0x0A && cp <= 0x0D) || 5720 cp == 0x85 || cp == 0x2028 || cp == 0x2029; 5721 } 5722 5723 /** 5724 * matches a Perl horizontal whitespace 5725 */ 5726 static BmpCharPredicate HorizWS() { 5727 return cp -> 5728 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 || 5729 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f || 5730 cp == 0x205f || cp == 0x3000; 5731 } 5732 5733 /** 5734 * for the Unicode category ALL and the dot metacharacter when 5735 * in dotall mode. 5736 */ 5737 static CharPredicate ALL() { 5738 return ch -> true; 5739 } 5740 5741 /** 5742 * for the dot metacharacter when dotall is not enabled. 5743 */ 5744 static CharPredicate DOT() { 5745 return ch -> 5746 (ch != '\n' && ch != '\r' 5747 && (ch|1) != '\u2029' 5748 && ch != '\u0085'); 5749 } 5750 5751 /** 5752 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled. 5753 */ 5754 static CharPredicate UNIXDOT() { 5755 return ch -> ch != '\n'; 5756 } 5757 5758 /** 5759 * Indicate that matches a Supplementary Unicode character 5760 */ 5761 static CharPredicate SingleS(int c) { 5762 return ch -> ch == c; 5763 } 5764 5765 /** 5766 * A bmp/optimized predicate of single 5767 */ 5768 static BmpCharPredicate Single(int c) { 5769 return ch -> ch == c; 5770 } 5771 5772 /** 5773 * Case insensitive matches a given BMP character 5774 */ 5775 static BmpCharPredicate SingleI(int lower, int upper) { 5776 return ch -> ch == lower || ch == upper; 5777 } 5778 5779 /** 5780 * Unicode case insensitive matches a given Unicode character 5781 */ 5782 static CharPredicate SingleU(int lower) { 5783 return ch -> lower == ch || 5784 lower == Character.toLowerCase(Character.toUpperCase(ch)); 5785 } 5786 5787 private static boolean inRange(int lower, int ch, int upper) { 5788 return lower <= ch && ch <= upper; 5789 } 5790 5791 /** 5792 * Charactrs within a explicit value range 5793 */ 5794 static CharPredicate Range(int lower, int upper) { 5795 if (upper < Character.MIN_HIGH_SURROGATE || 5796 lower > Character.MAX_HIGH_SURROGATE && 5797 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT) 5798 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper)); 5799 return ch -> inRange(lower, ch, upper); 5800 } 5801 5802 /** 5803 * Charactrs within a explicit value range in a case insensitive manner. 5804 */ 5805 static CharPredicate CIRange(int lower, int upper) { 5806 return ch -> inRange(lower, ch, upper) || 5807 ASCII.isAscii(ch) && 5808 (inRange(lower, ASCII.toUpper(ch), upper) || 5809 inRange(lower, ASCII.toLower(ch), upper)); 5810 } 5811 5812 static CharPredicate CIRangeU(int lower, int upper) { 5813 return ch -> { 5814 if (inRange(lower, ch, upper)) 5815 return true; 5816 int up = Character.toUpperCase(ch); 5817 return inRange(lower, up, upper) || 5818 inRange(lower, Character.toLowerCase(up), upper); 5819 }; 5820 } 5821 5822 /** 5823 * This must be the very first initializer. 5824 */ 5825 static final Node accept = new Node(); 5826 5827 static final Node lastAccept = new LastNode(); 5828 5829 /** 5830 * Creates a predicate that tests if this pattern is found in a given input 5831 * string. 5832 * 5833 * @apiNote 5834 * This method creates a predicate that behaves as if it creates a matcher 5835 * from the input sequence and then calls {@code find}, for example a 5836 * predicate of the form: 5837 * <pre>{@code 5838 * s -> matcher(s).find(); 5839 * }</pre> 5840 * 5841 * @return The predicate which can be used for finding a match on a 5842 * subsequence of a string 5843 * @since 1.8 5844 * @see Matcher#find 5845 */ 5846 public Predicate<String> asPredicate() { 5847 return s -> matcher(s).find(); 5848 } 5849 5850 /** 5851 * Creates a predicate that tests if this pattern matches a given input string. 5852 * 5853 * @apiNote 5854 * This method creates a predicate that behaves as if it creates a matcher 5855 * from the input sequence and then calls {@code matches}, for example a 5856 * predicate of the form: 5857 * <pre>{@code 5858 * s -> matcher(s).matches(); 5859 * }</pre> 5860 * 5861 * @return The predicate which can be used for matching an input string 5862 * against this pattern. 5863 * @since 11 5864 * @see Matcher#matches 5865 */ 5866 public Predicate<String> asMatchPredicate() { 5867 return s -> matcher(s).matches(); 5868 } 5869 5870 /** 5871 * Creates a stream from the given input sequence around matches of this 5872 * pattern. 5873 * 5874 * <p> The stream returned by this method contains each substring of the 5875 * input sequence that is terminated by another subsequence that matches 5876 * this pattern or is terminated by the end of the input sequence. The 5877 * substrings in the stream are in the order in which they occur in the 5878 * input. Trailing empty strings will be discarded and not encountered in 5879 * the stream. 5880 * 5881 * <p> If this pattern does not match any subsequence of the input then 5882 * the resulting stream has just one element, namely the input sequence in 5883 * string form. 5884 * 5885 * <p> When there is a positive-width match at the beginning of the input 5886 * sequence then an empty leading substring is included at the beginning 5887 * of the stream. A zero-width match at the beginning however never produces 5888 * such empty leading substring. 5889 * 5890 * <p> If the input sequence is mutable, it must remain constant during the 5891 * execution of the terminal stream operation. Otherwise, the result of the 5892 * terminal stream operation is undefined. 5893 * 5894 * @param input 5895 * The character sequence to be split 5896 * 5897 * @return The stream of strings computed by splitting the input 5898 * around matches of this pattern 5899 * @see #split(CharSequence) 5900 * @since 1.8 5901 */ 5902 public Stream<String> splitAsStream(final CharSequence input) { 5903 class MatcherIterator implements Iterator<String> { 5904 private Matcher matcher; 5905 // The start position of the next sub-sequence of input 5906 // when current == input.length there are no more elements 5907 private int current; 5908 // null if the next element, if any, needs to obtained 5909 private String nextElement; 5910 // > 0 if there are N next empty elements 5911 private int emptyElementCount; 5912 5913 public String next() { 5914 if (!hasNext()) 5915 throw new NoSuchElementException(); 5916 5917 if (emptyElementCount == 0) { 5918 String n = nextElement; 5919 nextElement = null; 5920 return n; 5921 } else { 5922 emptyElementCount--; 5923 return ""; 5924 } 5925 } 5926 5927 public boolean hasNext() { 5928 if (matcher == null) { 5929 matcher = matcher(input); 5930 // If the input is an empty string then the result can only be a 5931 // stream of the input. Induce that by setting the empty 5932 // element count to 1 5933 emptyElementCount = input.length() == 0 ? 1 : 0; 5934 } 5935 if (nextElement != null || emptyElementCount > 0) 5936 return true; 5937 5938 if (current == input.length()) 5939 return false; 5940 5941 // Consume the next matching element 5942 // Count sequence of matching empty elements 5943 while (matcher.find()) { 5944 nextElement = input.subSequence(current, matcher.start()).toString(); 5945 current = matcher.end(); 5946 if (!nextElement.isEmpty()) { 5947 return true; 5948 } else if (current > 0) { // no empty leading substring for zero-width 5949 // match at the beginning of the input 5950 emptyElementCount++; 5951 } 5952 } 5953 5954 // Consume last matching element 5955 nextElement = input.subSequence(current, input.length()).toString(); 5956 current = input.length(); 5957 if (!nextElement.isEmpty()) { 5958 return true; 5959 } else { 5960 // Ignore a terminal sequence of matching empty elements 5961 emptyElementCount = 0; 5962 nextElement = null; 5963 return false; 5964 } 5965 } 5966 } 5967 return StreamSupport.stream(Spliterators.spliteratorUnknownSize( 5968 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false); 5969 } 5970 }