1 /* 2 * Copyright (c) 1999, 2016, 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 47 /** 48 * A compiled representation of a regular expression. 49 * 50 * <p> A regular expression, specified as a string, must first be compiled into 51 * an instance of this class. The resulting pattern can then be used to create 52 * a {@link Matcher} object that can match arbitrary {@linkplain 53 * java.lang.CharSequence character sequences} against the regular 54 * expression. All of the state involved in performing a match resides in the 55 * matcher, so many matchers can share the same pattern. 56 * 57 * <p> A typical invocation sequence is thus 58 * 59 * <blockquote><pre> 60 * Pattern p = Pattern.{@link #compile compile}("a*b"); 61 * Matcher m = p.{@link #matcher matcher}("aaaaab"); 62 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote> 63 * 64 * <p> A {@link #matches matches} method is defined by this class as a 65 * convenience for when a regular expression is used just once. This method 66 * compiles an expression and matches an input sequence against it in a single 67 * invocation. The statement 68 * 69 * <blockquote><pre> 70 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote> 71 * 72 * is equivalent to the three statements above, though for repeated matches it 73 * is less efficient since it does not allow the compiled pattern to be reused. 74 * 75 * <p> Instances of this class are immutable and are safe for use by multiple 76 * concurrent threads. Instances of the {@link Matcher} class are not safe for 77 * such use. 78 * 79 * 80 * <h3><a name="sum">Summary of regular-expression constructs</a></h3> 81 * 82 * <table border="0" cellpadding="1" cellspacing="0" 83 * summary="Regular expression constructs, and what they match"> 84 * 85 * <tr align="left"> 86 * <th align="left" id="construct">Construct</th> 87 * <th align="left" id="matches">Matches</th> 88 * </tr> 89 * 90 * <tr><th> </th></tr> 91 * <tr align="left"><th colspan="2" id="characters">Characters</th></tr> 92 * 93 * <tr><td valign="top" headers="construct characters"><i>x</i></td> 94 * <td headers="matches">The character <i>x</i></td></tr> 95 * <tr><td valign="top" headers="construct characters">{@code \\}</td> 96 * <td headers="matches">The backslash character</td></tr> 97 * <tr><td valign="top" headers="construct characters">{@code \0}<i>n</i></td> 98 * <td headers="matches">The character with octal value {@code 0}<i>n</i> 99 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 100 * <tr><td valign="top" headers="construct characters">{@code \0}<i>nn</i></td> 101 * <td headers="matches">The character with octal value {@code 0}<i>nn</i> 102 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 103 * <tr><td valign="top" headers="construct characters">{@code \0}<i>mnn</i></td> 104 * <td headers="matches">The character with octal value {@code 0}<i>mnn</i> 105 * (0 {@code <=} <i>m</i> {@code <=} 3, 106 * 0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 107 * <tr><td valign="top" headers="construct characters">{@code \x}<i>hh</i></td> 108 * <td headers="matches">The character with hexadecimal value {@code 0x}<i>hh</i></td></tr> 109 * <tr><td valign="top" headers="construct characters"><code>\u</code><i>hhhh</i></td> 110 * <td headers="matches">The character with hexadecimal value {@code 0x}<i>hhhh</i></td></tr> 111 * <tr><td valign="top" headers="construct characters"><code>\x</code><i>{h...h}</i></td> 112 * <td headers="matches">The character with hexadecimal value {@code 0x}<i>h...h</i> 113 * ({@link java.lang.Character#MIN_CODE_POINT Character.MIN_CODE_POINT} 114 * <= {@code 0x}<i>h...h</i> <= 115 * {@link java.lang.Character#MAX_CODE_POINT Character.MAX_CODE_POINT})</td></tr> 116 * <tr><td valign="top" headers="construct characters"><code>\N{</code><i>name</i><code>}</code></td> 117 * <td headers="matches">The character with Unicode character name <i>'name'</i></td></tr> 118 * <tr><td valign="top" headers="matches">{@code \t}</td> 119 * <td headers="matches">The tab character (<code>'\u0009'</code>)</td></tr> 120 * <tr><td valign="top" headers="construct characters">{@code \n}</td> 121 * <td headers="matches">The newline (line feed) character (<code>'\u000A'</code>)</td></tr> 122 * <tr><td valign="top" headers="construct characters">{@code \r}</td> 123 * <td headers="matches">The carriage-return character (<code>'\u000D'</code>)</td></tr> 124 * <tr><td valign="top" headers="construct characters">{@code \f}</td> 125 * <td headers="matches">The form-feed character (<code>'\u000C'</code>)</td></tr> 126 * <tr><td valign="top" headers="construct characters">{@code \a}</td> 127 * <td headers="matches">The alert (bell) character (<code>'\u0007'</code>)</td></tr> 128 * <tr><td valign="top" headers="construct characters">{@code \e}</td> 129 * <td headers="matches">The escape character (<code>'\u001B'</code>)</td></tr> 130 * <tr><td valign="top" headers="construct characters">{@code \c}<i>x</i></td> 131 * <td headers="matches">The control character corresponding to <i>x</i></td></tr> 132 * 133 * <tr><th> </th></tr> 134 * <tr align="left"><th colspan="2" id="classes">Character classes</th></tr> 135 * 136 * <tr><td valign="top" headers="construct classes">{@code [abc]}</td> 137 * <td headers="matches">{@code a}, {@code b}, or {@code c} (simple class)</td></tr> 138 * <tr><td valign="top" headers="construct classes">{@code [^abc]}</td> 139 * <td headers="matches">Any character except {@code a}, {@code b}, or {@code c} (negation)</td></tr> 140 * <tr><td valign="top" headers="construct classes">{@code [a-zA-Z]}</td> 141 * <td headers="matches">{@code a} through {@code z} 142 * or {@code A} through {@code Z}, inclusive (range)</td></tr> 143 * <tr><td valign="top" headers="construct classes">{@code [a-d[m-p]]}</td> 144 * <td headers="matches">{@code a} through {@code d}, 145 * or {@code m} through {@code p}: {@code [a-dm-p]} (union)</td></tr> 146 * <tr><td valign="top" headers="construct classes">{@code [a-z&&[def]]}</td> 147 * <td headers="matches">{@code d}, {@code e}, or {@code f} (intersection)</tr> 148 * <tr><td valign="top" headers="construct classes">{@code [a-z&&[^bc]]}</td> 149 * <td headers="matches">{@code a} through {@code z}, 150 * except for {@code b} and {@code c}: {@code [ad-z]} (subtraction)</td></tr> 151 * <tr><td valign="top" headers="construct classes">{@code [a-z&&[^m-p]]}</td> 152 * <td headers="matches">{@code a} through {@code z}, 153 * and not {@code m} through {@code p}: {@code [a-lq-z]}(subtraction)</td></tr> 154 * <tr><th> </th></tr> 155 * 156 * <tr align="left"><th colspan="2" id="predef">Predefined character classes</th></tr> 157 * 158 * <tr><td valign="top" headers="construct predef">{@code .}</td> 159 * <td headers="matches">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr> 160 * <tr><td valign="top" headers="construct predef">{@code \d}</td> 161 * <td headers="matches">A digit: {@code [0-9]}</td></tr> 162 * <tr><td valign="top" headers="construct predef">{@code \D}</td> 163 * <td headers="matches">A non-digit: {@code [^0-9]}</td></tr> 164 * <tr><td valign="top" headers="construct predef">{@code \h}</td> 165 * <td headers="matches">A horizontal whitespace character: 166 * <code>[ \t\xA0\u1680\u180e\u2000-\u200a\u202f\u205f\u3000]</code></td></tr> 167 * <tr><td valign="top" headers="construct predef">{@code \H}</td> 168 * <td headers="matches">A non-horizontal whitespace character: {@code [^\h]}</td></tr> 169 * <tr><td valign="top" headers="construct predef">{@code \s}</td> 170 * <td headers="matches">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 171 * <tr><td valign="top" headers="construct predef">{@code \S}</td> 172 * <td headers="matches">A non-whitespace character: {@code [^\s]}</td></tr> 173 * <tr><td valign="top" headers="construct predef">{@code \v}</td> 174 * <td headers="matches">A vertical whitespace character: <code>[\n\x0B\f\r\x85\u2028\u2029]</code> 175 * </td></tr> 176 * <tr><td valign="top" headers="construct predef">{@code \V}</td> 177 * <td headers="matches">A non-vertical whitespace character: {@code [^\v]}</td></tr> 178 * <tr><td valign="top" headers="construct predef">{@code \w}</td> 179 * <td headers="matches">A word character: {@code [a-zA-Z_0-9]}</td></tr> 180 * <tr><td valign="top" headers="construct predef">{@code \W}</td> 181 * <td headers="matches">A non-word character: {@code [^\w]}</td></tr> 182 * <tr><th> </th></tr> 183 * <tr align="left"><th colspan="2" id="posix"><b>POSIX character classes (US-ASCII only)</b></th></tr> 184 * 185 * <tr><td valign="top" headers="construct posix">{@code \p{Lower}}</td> 186 * <td headers="matches">A lower-case alphabetic character: {@code [a-z]}</td></tr> 187 * <tr><td valign="top" headers="construct posix">{@code \p{Upper}}</td> 188 * <td headers="matches">An upper-case alphabetic character:{@code [A-Z]}</td></tr> 189 * <tr><td valign="top" headers="construct posix">{@code \p{ASCII}}</td> 190 * <td headers="matches">All ASCII:{@code [\x00-\x7F]}</td></tr> 191 * <tr><td valign="top" headers="construct posix">{@code \p{Alpha}}</td> 192 * <td headers="matches">An alphabetic character:{@code [\p{Lower}\p{Upper}]}</td></tr> 193 * <tr><td valign="top" headers="construct posix">{@code \p{Digit}}</td> 194 * <td headers="matches">A decimal digit: {@code [0-9]}</td></tr> 195 * <tr><td valign="top" headers="construct posix">{@code \p{Alnum}}</td> 196 * <td headers="matches">An alphanumeric character:{@code [\p{Alpha}\p{Digit}]}</td></tr> 197 * <tr><td valign="top" headers="construct posix">{@code \p{Punct}}</td> 198 * <td headers="matches">Punctuation: One of {@code !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~}</td></tr> 199 * <!-- {@code [\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]} 200 * {@code [\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]} --> 201 * <tr><td valign="top" headers="construct posix">{@code \p{Graph}}</td> 202 * <td headers="matches">A visible character: {@code [\p{Alnum}\p{Punct}]}</td></tr> 203 * <tr><td valign="top" headers="construct posix">{@code \p{Print}}</td> 204 * <td headers="matches">A printable character: {@code [\p{Graph}\x20]}</td></tr> 205 * <tr><td valign="top" headers="construct posix">{@code \p{Blank}}</td> 206 * <td headers="matches">A space or a tab: {@code [ \t]}</td></tr> 207 * <tr><td valign="top" headers="construct posix">{@code \p{Cntrl}}</td> 208 * <td headers="matches">A control character: {@code [\x00-\x1F\x7F]}</td></tr> 209 * <tr><td valign="top" headers="construct posix">{@code \p{XDigit}}</td> 210 * <td headers="matches">A hexadecimal digit: {@code [0-9a-fA-F]}</td></tr> 211 * <tr><td valign="top" headers="construct posix">{@code \p{Space}}</td> 212 * <td headers="matches">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 213 * 214 * <tr><th> </th></tr> 215 * <tr align="left"><th colspan="2">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr> 216 * 217 * <tr><td valign="top">{@code \p{javaLowerCase}}</td> 218 * <td>Equivalent to java.lang.Character.isLowerCase()</td></tr> 219 * <tr><td valign="top">{@code \p{javaUpperCase}}</td> 220 * <td>Equivalent to java.lang.Character.isUpperCase()</td></tr> 221 * <tr><td valign="top">{@code \p{javaWhitespace}}</td> 222 * <td>Equivalent to java.lang.Character.isWhitespace()</td></tr> 223 * <tr><td valign="top">{@code \p{javaMirrored}}</td> 224 * <td>Equivalent to java.lang.Character.isMirrored()</td></tr> 225 * 226 * <tr><th> </th></tr> 227 * <tr align="left"><th colspan="2" id="unicode">Classes for Unicode scripts, blocks, categories and binary properties</th></tr> 228 * <tr><td valign="top" headers="construct unicode">{@code \p{IsLatin}}</td> 229 * <td headers="matches">A Latin script character (<a href="#usc">script</a>)</td></tr> 230 * <tr><td valign="top" headers="construct unicode">{@code \p{InGreek}}</td> 231 * <td headers="matches">A character in the Greek block (<a href="#ubc">block</a>)</td></tr> 232 * <tr><td valign="top" headers="construct unicode">{@code \p{Lu}}</td> 233 * <td headers="matches">An uppercase letter (<a href="#ucc">category</a>)</td></tr> 234 * <tr><td valign="top" headers="construct unicode">{@code \p{IsAlphabetic}}</td> 235 * <td headers="matches">An alphabetic character (<a href="#ubpc">binary property</a>)</td></tr> 236 * <tr><td valign="top" headers="construct unicode">{@code \p{Sc}}</td> 237 * <td headers="matches">A currency symbol</td></tr> 238 * <tr><td valign="top" headers="construct unicode">{@code \P{InGreek}}</td> 239 * <td headers="matches">Any character except one in the Greek block (negation)</td></tr> 240 * <tr><td valign="top" headers="construct unicode">{@code [\p{L}&&[^\p{Lu}]]}</td> 241 * <td headers="matches">Any letter except an uppercase letter (subtraction)</td></tr> 242 * 243 * <tr><th> </th></tr> 244 * <tr align="left"><th colspan="2" id="bounds">Boundary matchers</th></tr> 245 * 246 * <tr><td valign="top" headers="construct bounds">{@code ^}</td> 247 * <td headers="matches">The beginning of a line</td></tr> 248 * <tr><td valign="top" headers="construct bounds">{@code $}</td> 249 * <td headers="matches">The end of a line</td></tr> 250 * <tr><td valign="top" headers="construct bounds">{@code \b}</td> 251 * <td headers="matches">A word boundary</td></tr> 252 * <tr><td valign="top" headers="construct bounds">{@code \b{g}}</td> 253 * <td headers="matches">A Unicode extended grapheme cluster boundary</td></tr> 254 * <tr><td valign="top" headers="construct bounds">{@code \B}</td> 255 * <td headers="matches">A non-word boundary</td></tr> 256 * <tr><td valign="top" headers="construct bounds">{@code \A}</td> 257 * <td headers="matches">The beginning of the input</td></tr> 258 * <tr><td valign="top" headers="construct bounds">{@code \G}</td> 259 * <td headers="matches">The end of the previous match</td></tr> 260 * <tr><td valign="top" headers="construct bounds">{@code \Z}</td> 261 * <td headers="matches">The end of the input but for the final 262 * <a href="#lt">terminator</a>, if any</td></tr> 263 * <tr><td valign="top" headers="construct bounds">{@code \z}</td> 264 * <td headers="matches">The end of the input</td></tr> 265 * 266 * <tr><th> </th></tr> 267 * <tr align="left"><th colspan="2" id="lineending">Linebreak matcher</th></tr> 268 * <tr><td valign="top" headers="construct lineending">{@code \R}</td> 269 * <td headers="matches">Any Unicode linebreak sequence, is equivalent to 270 * <code>\u000D\u000A|[\u000A\u000B\u000C\u000D\u0085\u2028\u2029] 271 * </code></td></tr> 272 * 273 * <tr><th> </th></tr> 274 * <tr align="left"><th colspan="2" id="grapheme">Unicode Extended Grapheme matcher</th></tr> 275 * <tr><td valign="top" headers="construct grapheme">{@code \X}</td> 276 * <td headers="matches">Any Unicode extended grapheme cluster</td></tr> 277 * 278 * <tr><th> </th></tr> 279 * <tr align="left"><th colspan="2" id="greedy">Greedy quantifiers</th></tr> 280 * 281 * <tr><td valign="top" headers="construct greedy"><i>X</i>{@code ?}</td> 282 * <td headers="matches"><i>X</i>, once or not at all</td></tr> 283 * <tr><td valign="top" headers="construct greedy"><i>X</i>{@code *}</td> 284 * <td headers="matches"><i>X</i>, zero or more times</td></tr> 285 * <tr><td valign="top" headers="construct greedy"><i>X</i>{@code +}</td> 286 * <td headers="matches"><i>X</i>, one or more times</td></tr> 287 * <tr><td valign="top" headers="construct greedy"><i>X</i><code>{</code><i>n</i><code>}</code></td> 288 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr> 289 * <tr><td valign="top" headers="construct greedy"><i>X</i><code>{</code><i>n</i>{@code ,}}</td> 290 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr> 291 * <tr><td valign="top" headers="construct greedy"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}</code></td> 292 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 293 * 294 * <tr><th> </th></tr> 295 * <tr align="left"><th colspan="2" id="reluc">Reluctant quantifiers</th></tr> 296 * 297 * <tr><td valign="top" headers="construct reluc"><i>X</i>{@code ??}</td> 298 * <td headers="matches"><i>X</i>, once or not at all</td></tr> 299 * <tr><td valign="top" headers="construct reluc"><i>X</i>{@code *?}</td> 300 * <td headers="matches"><i>X</i>, zero or more times</td></tr> 301 * <tr><td valign="top" headers="construct reluc"><i>X</i>{@code +?}</td> 302 * <td headers="matches"><i>X</i>, one or more times</td></tr> 303 * <tr><td valign="top" headers="construct reluc"><i>X</i><code>{</code><i>n</i><code>}?</code></td> 304 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr> 305 * <tr><td valign="top" headers="construct reluc"><i>X</i><code>{</code><i>n</i><code>,}?</code></td> 306 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr> 307 * <tr><td valign="top" headers="construct reluc"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}?</code></td> 308 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 309 * 310 * <tr><th> </th></tr> 311 * <tr align="left"><th colspan="2" id="poss">Possessive quantifiers</th></tr> 312 * 313 * <tr><td valign="top" headers="construct poss"><i>X</i>{@code ?+}</td> 314 * <td headers="matches"><i>X</i>, once or not at all</td></tr> 315 * <tr><td valign="top" headers="construct poss"><i>X</i>{@code *+}</td> 316 * <td headers="matches"><i>X</i>, zero or more times</td></tr> 317 * <tr><td valign="top" headers="construct poss"><i>X</i>{@code ++}</td> 318 * <td headers="matches"><i>X</i>, one or more times</td></tr> 319 * <tr><td valign="top" headers="construct poss"><i>X</i><code>{</code><i>n</i><code>}+</code></td> 320 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr> 321 * <tr><td valign="top" headers="construct poss"><i>X</i><code>{</code><i>n</i><code>,}+</code></td> 322 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr> 323 * <tr><td valign="top" headers="construct poss"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}+</code></td> 324 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 325 * 326 * <tr><th> </th></tr> 327 * <tr align="left"><th colspan="2" id="logical">Logical operators</th></tr> 328 * 329 * <tr><td valign="top" headers="construct logical"><i>XY</i></td> 330 * <td headers="matches"><i>X</i> followed by <i>Y</i></td></tr> 331 * <tr><td valign="top" headers="construct logical"><i>X</i>{@code |}<i>Y</i></td> 332 * <td headers="matches">Either <i>X</i> or <i>Y</i></td></tr> 333 * <tr><td valign="top" headers="construct logical">{@code (}<i>X</i>{@code )}</td> 334 * <td headers="matches">X, as a <a href="#cg">capturing group</a></td></tr> 335 * 336 * <tr><th> </th></tr> 337 * <tr align="left"><th colspan="2" id="backref">Back references</th></tr> 338 * 339 * <tr><td valign="bottom" headers="construct backref">{@code \}<i>n</i></td> 340 * <td valign="bottom" headers="matches">Whatever the <i>n</i><sup>th</sup> 341 * <a href="#cg">capturing group</a> matched</td></tr> 342 * 343 * <tr><td valign="bottom" headers="construct backref">{@code \}<i>k</i><<i>name</i>></td> 344 * <td valign="bottom" headers="matches">Whatever the 345 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr> 346 * 347 * <tr><th> </th></tr> 348 * <tr align="left"><th colspan="2" id="quot">Quotation</th></tr> 349 * 350 * <tr><td valign="top" headers="construct quot">{@code \}</td> 351 * <td headers="matches">Nothing, but quotes the following character</td></tr> 352 * <tr><td valign="top" headers="construct quot">{@code \Q}</td> 353 * <td headers="matches">Nothing, but quotes all characters until {@code \E}</td></tr> 354 * <tr><td valign="top" headers="construct quot">{@code \E}</td> 355 * <td headers="matches">Nothing, but ends quoting started by {@code \Q}</td></tr> 356 * <!-- Metachars: !$()*+.<>?[\]^{|} --> 357 * 358 * <tr><th> </th></tr> 359 * <tr align="left"><th colspan="2" id="special">Special constructs (named-capturing and non-capturing)</th></tr> 360 * 361 * <tr><td valign="top" headers="construct special"><code>(?<<a href="#groupname">name</a>></code><i>X</i>{@code )}</td> 362 * <td headers="matches"><i>X</i>, as a named-capturing group</td></tr> 363 * <tr><td valign="top" headers="construct special">{@code (?:}<i>X</i>{@code )}</td> 364 * <td headers="matches"><i>X</i>, as a non-capturing group</td></tr> 365 * <tr><td valign="top" headers="construct special"><code>(?idmsuxU-idmsuxU) </code></td> 366 * <td headers="matches">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a> 367 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> 368 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a> 369 * on - off</td></tr> 370 * <tr><td valign="top" headers="construct special"><code>(?idmsux-idmsux:</code><i>X</i>{@code )} </td> 371 * <td headers="matches"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the 372 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a> 373 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a > 374 * <a href="#COMMENTS">x</a> on - off</td></tr> 375 * <tr><td valign="top" headers="construct special">{@code (?=}<i>X</i>{@code )}</td> 376 * <td headers="matches"><i>X</i>, via zero-width positive lookahead</td></tr> 377 * <tr><td valign="top" headers="construct special">{@code (?!}<i>X</i>{@code )}</td> 378 * <td headers="matches"><i>X</i>, via zero-width negative lookahead</td></tr> 379 * <tr><td valign="top" headers="construct special">{@code (?<=}<i>X</i>{@code )}</td> 380 * <td headers="matches"><i>X</i>, via zero-width positive lookbehind</td></tr> 381 * <tr><td valign="top" headers="construct special">{@code (?<!}<i>X</i>{@code )}</td> 382 * <td headers="matches"><i>X</i>, via zero-width negative lookbehind</td></tr> 383 * <tr><td valign="top" headers="construct special">{@code (?>}<i>X</i>{@code )}</td> 384 * <td headers="matches"><i>X</i>, as an independent, non-capturing group</td></tr> 385 * 386 * </table> 387 * 388 * <hr> 389 * 390 * 391 * <h3><a name="bs">Backslashes, escapes, and quoting</a></h3> 392 * 393 * <p> The backslash character ({@code '\'}) serves to introduce escaped 394 * constructs, as defined in the table above, as well as to quote characters 395 * that otherwise would be interpreted as unescaped constructs. Thus the 396 * expression {@code \\} matches a single backslash and <code>\{</code> matches a 397 * left brace. 398 * 399 * <p> It is an error to use a backslash prior to any alphabetic character that 400 * does not denote an escaped construct; these are reserved for future 401 * extensions to the regular-expression language. A backslash may be used 402 * prior to a non-alphabetic character regardless of whether that character is 403 * part of an unescaped construct. 404 * 405 * <p> Backslashes within string literals in Java source code are interpreted 406 * as required by 407 * <cite>The Java™ Language Specification</cite> 408 * as either Unicode escapes (section 3.3) or other character escapes (section 3.10.6) 409 * It is therefore necessary to double backslashes in string 410 * literals that represent regular expressions to protect them from 411 * interpretation by the Java bytecode compiler. The string literal 412 * <code>"\b"</code>, for example, matches a single backspace character when 413 * interpreted as a regular expression, while {@code "\\b"} matches a 414 * word boundary. The string literal {@code "\(hello\)"} is illegal 415 * and leads to a compile-time error; in order to match the string 416 * {@code (hello)} the string literal {@code "\\(hello\\)"} 417 * must be used. 418 * 419 * <h3><a name="cc">Character Classes</a></h3> 420 * 421 * <p> Character classes may appear within other character classes, and 422 * may be composed by the union operator (implicit) and the intersection 423 * operator ({@code &&}). 424 * The union operator denotes a class that contains every character that is 425 * in at least one of its operand classes. The intersection operator 426 * denotes a class that contains every character that is in both of its 427 * operand classes. 428 * 429 * <p> The precedence of character-class operators is as follows, from 430 * highest to lowest: 431 * 432 * <blockquote><table border="0" cellpadding="1" cellspacing="0" 433 * summary="Precedence of character class operators."> 434 * <tr><th>1 </th> 435 * <td>Literal escape </td> 436 * <td>{@code \x}</td></tr> 437 * <tr><th>2 </th> 438 * <td>Grouping</td> 439 * <td>{@code [...]}</td></tr> 440 * <tr><th>3 </th> 441 * <td>Range</td> 442 * <td>{@code a-z}</td></tr> 443 * <tr><th>4 </th> 444 * <td>Union</td> 445 * <td>{@code [a-e][i-u]}</td></tr> 446 * <tr><th>5 </th> 447 * <td>Intersection</td> 448 * <td>{@code [a-z&&[aeiou]]}</td></tr> 449 * </table></blockquote> 450 * 451 * <p> Note that a different set of metacharacters are in effect inside 452 * a character class than outside a character class. For instance, the 453 * regular expression {@code .} loses its special meaning inside a 454 * character class, while the expression {@code -} becomes a range 455 * forming metacharacter. 456 * 457 * <h3><a name="lt">Line terminators</a></h3> 458 * 459 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks 460 * the end of a line of the input character sequence. The following are 461 * recognized as line terminators: 462 * 463 * <ul> 464 * 465 * <li> A newline (line feed) character ({@code '\n'}), 466 * 467 * <li> A carriage-return character followed immediately by a newline 468 * character ({@code "\r\n"}), 469 * 470 * <li> A standalone carriage-return character ({@code '\r'}), 471 * 472 * <li> A next-line character (<code>'\u0085'</code>), 473 * 474 * <li> A line-separator character (<code>'\u2028'</code>), or 475 * 476 * <li> A paragraph-separator character (<code>'\u2029'</code>). 477 * 478 * </ul> 479 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators 480 * recognized are newline characters. 481 * 482 * <p> The regular expression {@code .} matches any character except a line 483 * terminator unless the {@link #DOTALL} flag is specified. 484 * 485 * <p> By default, the regular expressions {@code ^} and {@code $} ignore 486 * line terminators and only match at the beginning and the end, respectively, 487 * of the entire input sequence. If {@link #MULTILINE} mode is activated then 488 * {@code ^} matches at the beginning of input and after any line terminator 489 * except at the end of input. When in {@link #MULTILINE} mode {@code $} 490 * matches just before a line terminator or the end of the input sequence. 491 * 492 * <h3><a name="cg">Groups and capturing</a></h3> 493 * 494 * <h4><a name="gnumber">Group number</a></h4> 495 * <p> Capturing groups are numbered by counting their opening parentheses from 496 * left to right. In the expression {@code ((A)(B(C)))}, for example, there 497 * are four such groups: </p> 498 * 499 * <blockquote><table cellpadding=1 cellspacing=0 summary="Capturing group numberings"> 500 * <tr><th>1 </th> 501 * <td>{@code ((A)(B(C)))}</td></tr> 502 * <tr><th>2 </th> 503 * <td>{@code (A)}</td></tr> 504 * <tr><th>3 </th> 505 * <td>{@code (B(C))}</td></tr> 506 * <tr><th>4 </th> 507 * <td>{@code (C)}</td></tr> 508 * </table></blockquote> 509 * 510 * <p> Group zero always stands for the entire expression. 511 * 512 * <p> Capturing groups are so named because, during a match, each subsequence 513 * of the input sequence that matches such a group is saved. The captured 514 * subsequence may be used later in the expression, via a back reference, and 515 * may also be retrieved from the matcher once the match operation is complete. 516 * 517 * <h4><a name="groupname">Group name</a></h4> 518 * <p>A capturing group can also be assigned a "name", a {@code named-capturing group}, 519 * and then be back-referenced later by the "name". Group names are composed of 520 * the following characters. The first character must be a {@code letter}. 521 * 522 * <ul> 523 * <li> The uppercase letters {@code 'A'} through {@code 'Z'} 524 * (<code>'\u0041'</code> through <code>'\u005a'</code>), 525 * <li> The lowercase letters {@code 'a'} through {@code 'z'} 526 * (<code>'\u0061'</code> through <code>'\u007a'</code>), 527 * <li> The digits {@code '0'} through {@code '9'} 528 * (<code>'\u0030'</code> through <code>'\u0039'</code>), 529 * </ul> 530 * 531 * <p> A {@code named-capturing group} is still numbered as described in 532 * <a href="#gnumber">Group number</a>. 533 * 534 * <p> The captured input associated with a group is always the subsequence 535 * that the group most recently matched. If a group is evaluated a second time 536 * because of quantification then its previously-captured value, if any, will 537 * be retained if the second evaluation fails. Matching the string 538 * {@code "aba"} against the expression {@code (a(b)?)+}, for example, leaves 539 * group two set to {@code "b"}. All captured input is discarded at the 540 * beginning of each match. 541 * 542 * <p> Groups beginning with {@code (?} are either pure, <i>non-capturing</i> groups 543 * that do not capture text and do not count towards the group total, or 544 * <i>named-capturing</i> group. 545 * 546 * <h3> Unicode support </h3> 547 * 548 * <p> This class is in conformance with Level 1 of <a 549 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 550 * Standard #18: Unicode Regular Expression</i></a>, plus RL2.1 551 * Canonical Equivalents. 552 * <p> 553 * <b>Unicode escape sequences</b> such as <code>\u2014</code> in Java source code 554 * are processed as described in section 3.3 of 555 * <cite>The Java™ Language Specification</cite>. 556 * Such escape sequences are also implemented directly by the regular-expression 557 * parser so that Unicode escapes can be used in expressions that are read from 558 * files or from the keyboard. Thus the strings <code>"\u2014"</code> and 559 * {@code "\\u2014"}, while not equal, compile into the same pattern, which 560 * matches the character with hexadecimal value {@code 0x2014}. 561 * <p> 562 * A Unicode character can also be represented by using its <b>Hex notation</b> 563 * (hexadecimal code point value) directly as described in construct 564 * <code>\x{...}</code>, for example a supplementary character U+2011F can be 565 * specified as <code>\x{2011F}</code>, instead of two consecutive Unicode escape 566 * sequences of the surrogate pair <code>\uD840</code><code>\uDD1F</code>. 567 * <p> 568 * <b>Unicode character names</b> are supported by the named character construct 569 * <code>\N{</code>...<code>}</code>, for example, <code>\N{WHITE SMILING FACE}</code> 570 * specifies character <code>\u263A</code>. The character names supported 571 * by this class are the valid Unicode character names matched by 572 * {@link java.lang.Character#codePointOf(String) Character.codePointOf(name)}. 573 * <p> 574 * <a href="http://www.unicode.org/reports/tr18/#Default_Grapheme_Clusters"> 575 * <b>Unicode extended grapheme clusters</b></a> are supported by the grapheme 576 * cluster matcher {@code \X} and the corresponding boundary matcher {@code \b{g}}. 577 * <p> 578 * Unicode scripts, blocks, categories and binary properties are written with 579 * the {@code \p} and {@code \P} constructs as in Perl. 580 * <code>\p{</code><i>prop</i><code>}</code> matches if 581 * the input has the property <i>prop</i>, while <code>\P{</code><i>prop</i><code>}</code> 582 * does not match if the input has that property. 583 * <p> 584 * Scripts, blocks, categories and binary properties can be used both inside 585 * and outside of a character class. 586 * 587 * <p> 588 * <b><a name="usc">Scripts</a></b> are specified either with the prefix {@code Is}, as in 589 * {@code IsHiragana}, or by using the {@code script} keyword (or its short 590 * form {@code sc}) as in {@code script=Hiragana} or {@code sc=Hiragana}. 591 * <p> 592 * The script names supported by {@code Pattern} are the valid script names 593 * accepted and defined by 594 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}. 595 * 596 * <p> 597 * <b><a name="ubc">Blocks</a></b> are specified with the prefix {@code In}, as in 598 * {@code InMongolian}, or by using the keyword {@code block} (or its short 599 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}. 600 * <p> 601 * The block names supported by {@code Pattern} are the valid block names 602 * accepted and defined by 603 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}. 604 * <p> 605 * 606 * <b><a name="ucc">Categories</a></b> may be specified with the optional prefix {@code Is}: 607 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode 608 * letters. Same as scripts and blocks, categories can also be specified 609 * by using the keyword {@code general_category} (or its short form 610 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}. 611 * <p> 612 * The supported categories are those of 613 * <a href="http://www.unicode.org/unicode/standard/standard.html"> 614 * <i>The Unicode Standard</i></a> in the version specified by the 615 * {@link java.lang.Character Character} class. The category names are those 616 * defined in the Standard, both normative and informative. 617 * <p> 618 * 619 * <b><a name="ubpc">Binary properties</a></b> are specified with the prefix {@code Is}, as in 620 * {@code IsAlphabetic}. The supported binary properties by {@code Pattern} 621 * are 622 * <ul> 623 * <li> Alphabetic 624 * <li> Ideographic 625 * <li> Letter 626 * <li> Lowercase 627 * <li> Uppercase 628 * <li> Titlecase 629 * <li> Punctuation 630 * <Li> Control 631 * <li> White_Space 632 * <li> Digit 633 * <li> Hex_Digit 634 * <li> Join_Control 635 * <li> Noncharacter_Code_Point 636 * <li> Assigned 637 * </ul> 638 * <p> 639 * The following <b>Predefined Character classes</b> and <b>POSIX character classes</b> 640 * are in conformance with the recommendation of <i>Annex C: Compatibility Properties</i> 641 * of <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Regular Expression 642 * </i></a>, when {@link #UNICODE_CHARACTER_CLASS} flag is specified. 643 * 644 * <table border="0" cellpadding="1" cellspacing="0" 645 * summary="predefined and posix character classes in Unicode mode"> 646 * <tr align="left"> 647 * <th align="left" id="predef_classes">Classes</th> 648 * <th align="left" id="predef_matches">Matches</th> 649 *</tr> 650 * <tr><td>{@code \p{Lower}}</td> 651 * <td>A lowercase character:{@code \p{IsLowercase}}</td></tr> 652 * <tr><td>{@code \p{Upper}}</td> 653 * <td>An uppercase character:{@code \p{IsUppercase}}</td></tr> 654 * <tr><td>{@code \p{ASCII}}</td> 655 * <td>All ASCII:{@code [\x00-\x7F]}</td></tr> 656 * <tr><td>{@code \p{Alpha}}</td> 657 * <td>An alphabetic character:{@code \p{IsAlphabetic}}</td></tr> 658 * <tr><td>{@code \p{Digit}}</td> 659 * <td>A decimal digit character:{@code p{IsDigit}}</td></tr> 660 * <tr><td>{@code \p{Alnum}}</td> 661 * <td>An alphanumeric character:{@code [\p{IsAlphabetic}\p{IsDigit}]}</td></tr> 662 * <tr><td>{@code \p{Punct}}</td> 663 * <td>A punctuation character:{@code p{IsPunctuation}}</td></tr> 664 * <tr><td>{@code \p{Graph}}</td> 665 * <td>A visible character: {@code [^\p{IsWhite_Space}\p{gc=Cc}\p{gc=Cs}\p{gc=Cn}]}</td></tr> 666 * <tr><td>{@code \p{Print}}</td> 667 * <td>A printable character: {@code [\p{Graph}\p{Blank}&&[^\p{Cntrl}]]}</td></tr> 668 * <tr><td>{@code \p{Blank}}</td> 669 * <td>A space or a tab: {@code [\p{IsWhite_Space}&&[^\p{gc=Zl}\p{gc=Zp}\x0a\x0b\x0c\x0d\x85]]}</td></tr> 670 * <tr><td>{@code \p{Cntrl}}</td> 671 * <td>A control character: {@code \p{gc=Cc}}</td></tr> 672 * <tr><td>{@code \p{XDigit}}</td> 673 * <td>A hexadecimal digit: {@code [\p{gc=Nd}\p{IsHex_Digit}]}</td></tr> 674 * <tr><td>{@code \p{Space}}</td> 675 * <td>A whitespace character:{@code \p{IsWhite_Space}}</td></tr> 676 * <tr><td>{@code \d}</td> 677 * <td>A digit: {@code \p{IsDigit}}</td></tr> 678 * <tr><td>{@code \D}</td> 679 * <td>A non-digit: {@code [^\d]}</td></tr> 680 * <tr><td>{@code \s}</td> 681 * <td>A whitespace character: {@code \p{IsWhite_Space}}</td></tr> 682 * <tr><td>{@code \S}</td> 683 * <td>A non-whitespace character: {@code [^\s]}</td></tr> 684 * <tr><td>{@code \w}</td> 685 * <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> 686 * <tr><td>{@code \W}</td> 687 * <td>A non-word character: {@code [^\w]}</td></tr> 688 * </table> 689 * <p> 690 * <a name="jcc"> 691 * Categories that behave like the java.lang.Character 692 * boolean is<i>methodname</i> methods (except for the deprecated ones) are 693 * available through the same <code>\p{</code><i>prop</i><code>}</code> syntax where 694 * the specified property has the name <code>java<i>methodname</i></code></a>. 695 * 696 * <h3> Comparison to Perl 5 </h3> 697 * 698 * <p>The {@code Pattern} engine performs traditional NFA-based matching 699 * with ordered alternation as occurs in Perl 5. 700 * 701 * <p> Perl constructs not supported by this class: </p> 702 * 703 * <ul> 704 * <li><p> The backreference constructs, <code>\g{</code><i>n</i><code>}</code> for 705 * the <i>n</i><sup>th</sup><a href="#cg">capturing group</a> and 706 * <code>\g{</code><i>name</i><code>}</code> for 707 * <a href="#groupname">named-capturing group</a>. 708 * </p></li> 709 * 710 * <li><p> The conditional constructs 711 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code )} and 712 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code |}<i>Y</i>{@code )}, 713 * </p></li> 714 * 715 * <li><p> The embedded code constructs <code>(?{</code><i>code</i><code>})</code> 716 * and <code>(??{</code><i>code</i><code>})</code>,</p></li> 717 * 718 * <li><p> The embedded comment syntax {@code (?#comment)}, and </p></li> 719 * 720 * <li><p> The preprocessing operations {@code \l} <code>\u</code>, 721 * {@code \L}, and {@code \U}. </p></li> 722 * 723 * </ul> 724 * 725 * <p> Constructs supported by this class but not by Perl: </p> 726 * 727 * <ul> 728 * 729 * <li><p> Character-class union and intersection as described 730 * <a href="#cc">above</a>.</p></li> 731 * 732 * </ul> 733 * 734 * <p> Notable differences from Perl: </p> 735 * 736 * <ul> 737 * 738 * <li><p> In Perl, {@code \1} through {@code \9} are always interpreted 739 * as back references; a backslash-escaped number greater than {@code 9} is 740 * treated as a back reference if at least that many subexpressions exist, 741 * otherwise it is interpreted, if possible, as an octal escape. In this 742 * class octal escapes must always begin with a zero. In this class, 743 * {@code \1} through {@code \9} are always interpreted as back 744 * references, and a larger number is accepted as a back reference if at 745 * least that many subexpressions exist at that point in the regular 746 * expression, otherwise the parser will drop digits until the number is 747 * smaller or equal to the existing number of groups or it is one digit. 748 * </p></li> 749 * 750 * <li><p> Perl uses the {@code g} flag to request a match that resumes 751 * where the last match left off. This functionality is provided implicitly 752 * by the {@link Matcher} class: Repeated invocations of the {@link 753 * Matcher#find find} method will resume where the last match left off, 754 * unless the matcher is reset. </p></li> 755 * 756 * <li><p> In Perl, embedded flags at the top level of an expression affect 757 * the whole expression. In this class, embedded flags always take effect 758 * at the point at which they appear, whether they are at the top level or 759 * within a group; in the latter case, flags are restored at the end of the 760 * group just as in Perl. </p></li> 761 * 762 * </ul> 763 * 764 * 765 * <p> For a more precise description of the behavior of regular expression 766 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/"> 767 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl, 768 * O'Reilly and Associates, 2006.</a> 769 * </p> 770 * 771 * @see java.lang.String#split(String, int) 772 * @see java.lang.String#split(String) 773 * 774 * @author Mike McCloskey 775 * @author Mark Reinhold 776 * @author JSR-51 Expert Group 777 * @since 1.4 778 * @spec JSR-51 779 */ 780 781 public final class Pattern 782 implements java.io.Serializable 783 { 784 785 /** 786 * Regular expression modifier values. Instead of being passed as 787 * arguments, they can also be passed as inline modifiers. 788 * For example, the following statements have the same effect. 789 * <pre> 790 * RegExp r1 = RegExp.compile("abc", Pattern.I|Pattern.M); 791 * RegExp r2 = RegExp.compile("(?im)abc", 0); 792 * </pre> 793 * 794 * The flags are duplicated so that the familiar Perl match flag 795 * names are available. 796 */ 797 798 /** 799 * Enables Unix lines mode. 800 * 801 * <p> In this mode, only the {@code '\n'} line terminator is recognized 802 * in the behavior of {@code .}, {@code ^}, and {@code $}. 803 * 804 * <p> Unix lines mode can also be enabled via the embedded flag 805 * expression {@code (?d)}. 806 */ 807 public static final int UNIX_LINES = 0x01; 808 809 /** 810 * Enables case-insensitive matching. 811 * 812 * <p> By default, case-insensitive matching assumes that only characters 813 * in the US-ASCII charset are being matched. Unicode-aware 814 * case-insensitive matching can be enabled by specifying the {@link 815 * #UNICODE_CASE} flag in conjunction with this flag. 816 * 817 * <p> Case-insensitive matching can also be enabled via the embedded flag 818 * expression {@code (?i)}. 819 * 820 * <p> Specifying this flag may impose a slight performance penalty. </p> 821 */ 822 public static final int CASE_INSENSITIVE = 0x02; 823 824 /** 825 * Permits whitespace and comments in pattern. 826 * 827 * <p> In this mode, whitespace is ignored, and embedded comments starting 828 * with {@code #} are ignored until the end of a line. 829 * 830 * <p> Comments mode can also be enabled via the embedded flag 831 * expression {@code (?x)}. 832 */ 833 public static final int COMMENTS = 0x04; 834 835 /** 836 * Enables multiline mode. 837 * 838 * <p> In multiline mode the expressions {@code ^} and {@code $} match 839 * just after or just before, respectively, a line terminator or the end of 840 * the input sequence. By default these expressions only match at the 841 * beginning and the end of the entire input sequence. 842 * 843 * <p> Multiline mode can also be enabled via the embedded flag 844 * expression {@code (?m)}. </p> 845 */ 846 public static final int MULTILINE = 0x08; 847 848 /** 849 * Enables literal parsing of the pattern. 850 * 851 * <p> When this flag is specified then the input string that specifies 852 * the pattern is treated as a sequence of literal characters. 853 * Metacharacters or escape sequences in the input sequence will be 854 * given no special meaning. 855 * 856 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on 857 * matching when used in conjunction with this flag. The other flags 858 * become superfluous. 859 * 860 * <p> There is no embedded flag character for enabling literal parsing. 861 * @since 1.5 862 */ 863 public static final int LITERAL = 0x10; 864 865 /** 866 * Enables dotall mode. 867 * 868 * <p> In dotall mode, the expression {@code .} matches any character, 869 * including a line terminator. By default this expression does not match 870 * line terminators. 871 * 872 * <p> Dotall mode can also be enabled via the embedded flag 873 * expression {@code (?s)}. (The {@code s} is a mnemonic for 874 * "single-line" mode, which is what this is called in Perl.) </p> 875 */ 876 public static final int DOTALL = 0x20; 877 878 /** 879 * Enables Unicode-aware case folding. 880 * 881 * <p> When this flag is specified then case-insensitive matching, when 882 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner 883 * consistent with the Unicode Standard. By default, case-insensitive 884 * matching assumes that only characters in the US-ASCII charset are being 885 * matched. 886 * 887 * <p> Unicode-aware case folding can also be enabled via the embedded flag 888 * expression {@code (?u)}. 889 * 890 * <p> Specifying this flag may impose a performance penalty. </p> 891 */ 892 public static final int UNICODE_CASE = 0x40; 893 894 /** 895 * Enables canonical equivalence. 896 * 897 * <p> When this flag is specified then two characters will be considered 898 * to match if, and only if, their full canonical decompositions match. 899 * The expression <code>"a\u030A"</code>, for example, will match the 900 * string <code>"\u00E5"</code> when this flag is specified. By default, 901 * matching does not take canonical equivalence into account. 902 * 903 * <p> There is no embedded flag character for enabling canonical 904 * equivalence. 905 * 906 * <p> Specifying this flag may impose a performance penalty. </p> 907 */ 908 public static final int CANON_EQ = 0x80; 909 910 /** 911 * Enables the Unicode version of <i>Predefined character classes</i> and 912 * <i>POSIX character classes</i>. 913 * 914 * <p> When this flag is specified then the (US-ASCII only) 915 * <i>Predefined character classes</i> and <i>POSIX character classes</i> 916 * are in conformance with 917 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 918 * Standard #18: Unicode Regular Expression</i></a> 919 * <i>Annex C: Compatibility Properties</i>. 920 * <p> 921 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded 922 * flag expression {@code (?U)}. 923 * <p> 924 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case 925 * folding. 926 * <p> 927 * Specifying this flag may impose a performance penalty. </p> 928 * @since 1.7 929 */ 930 public static final int UNICODE_CHARACTER_CLASS = 0x100; 931 932 /** 933 * Contains all possible flags for compile(regex, flags). 934 */ 935 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE | 936 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL | 937 UNICODE_CHARACTER_CLASS | COMMENTS; 938 939 /* Pattern has only two serialized components: The pattern string 940 * and the flags, which are all that is needed to recompile the pattern 941 * when it is deserialized. 942 */ 943 944 /** use serialVersionUID from Merlin b59 for interoperability */ 945 private static final long serialVersionUID = 5073258162644648461L; 946 947 /** 948 * The original regular-expression pattern string. 949 * 950 * @serial 951 */ 952 private String pattern; 953 954 /** 955 * The original pattern flags. 956 * 957 * @serial 958 */ 959 private int flags; 960 961 /** 962 * Boolean indicating this Pattern is compiled; this is necessary in order 963 * to lazily compile deserialized Patterns. 964 */ 965 private transient volatile boolean compiled; 966 967 /** 968 * The normalized pattern string. 969 */ 970 private transient String normalizedPattern; 971 972 /** 973 * The starting point of state machine for the find operation. This allows 974 * a match to start anywhere in the input. 975 */ 976 transient Node root; 977 978 /** 979 * The root of object tree for a match operation. The pattern is matched 980 * at the beginning. This may include a find that uses BnM or a First 981 * node. 982 */ 983 transient Node matchRoot; 984 985 /** 986 * Temporary storage used by parsing pattern slice. 987 */ 988 transient int[] buffer; 989 990 /** 991 * A temporary storage used for predicate for double return. 992 */ 993 transient CharPredicate predicate; 994 995 /** 996 * Map the "name" of the "named capturing group" to its group id 997 * node. 998 */ 999 transient volatile Map<String, Integer> namedGroups; 1000 1001 /** 1002 * Temporary storage used while parsing group references. 1003 */ 1004 transient GroupHead[] groupNodes; 1005 1006 /** 1007 * Temporary storage used to store the top level closure nodes. 1008 */ 1009 transient List<Node> topClosureNodes; 1010 1011 /** 1012 * The number of top greedy closure nodes in this Pattern. Used by 1013 * matchers to allocate storage needed for a IntHashSet to keep the 1014 * beginning pos {@code i} of all failed match. 1015 */ 1016 transient int localTCNCount; 1017 1018 /* 1019 * Turn off the stop-exponential-backtracking optimization if there 1020 * is a group ref in the pattern. 1021 */ 1022 transient boolean hasGroupRef; 1023 1024 /** 1025 * Temporary null terminated code point array used by pattern compiling. 1026 */ 1027 private transient int[] temp; 1028 1029 /** 1030 * The number of capturing groups in this Pattern. Used by matchers to 1031 * allocate storage needed to perform a match. 1032 */ 1033 transient int capturingGroupCount; 1034 1035 /** 1036 * The local variable count used by parsing tree. Used by matchers to 1037 * allocate storage needed to perform a match. 1038 */ 1039 transient int localCount; 1040 1041 /** 1042 * Index into the pattern string that keeps track of how much has been 1043 * parsed. 1044 */ 1045 private transient int cursor; 1046 1047 /** 1048 * Holds the length of the pattern string. 1049 */ 1050 private transient int patternLength; 1051 1052 /** 1053 * If the Start node might possibly match supplementary characters. 1054 * It is set to true during compiling if 1055 * (1) There is supplementary char in pattern, or 1056 * (2) There is complement node of a "family" CharProperty 1057 */ 1058 private transient boolean hasSupplementary; 1059 1060 /** 1061 * Compiles the given regular expression into a pattern. 1062 * 1063 * @param regex 1064 * The expression to be compiled 1065 * @return the given regular expression compiled into a pattern 1066 * @throws PatternSyntaxException 1067 * If the expression's syntax is invalid 1068 */ 1069 public static Pattern compile(String regex) { 1070 return new Pattern(regex, 0); 1071 } 1072 1073 /** 1074 * Compiles the given regular expression into a pattern with the given 1075 * flags. 1076 * 1077 * @param regex 1078 * The expression to be compiled 1079 * 1080 * @param flags 1081 * Match flags, a bit mask that may include 1082 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL}, 1083 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES}, 1084 * {@link #LITERAL}, {@link #UNICODE_CHARACTER_CLASS} 1085 * and {@link #COMMENTS} 1086 * 1087 * @return the given regular expression compiled into a pattern with the given flags 1088 * @throws IllegalArgumentException 1089 * If bit values other than those corresponding to the defined 1090 * match flags are set in {@code flags} 1091 * 1092 * @throws PatternSyntaxException 1093 * If the expression's syntax is invalid 1094 */ 1095 public static Pattern compile(String regex, int flags) { 1096 return new Pattern(regex, flags); 1097 } 1098 1099 /** 1100 * Returns the regular expression from which this pattern was compiled. 1101 * 1102 * @return The source of this pattern 1103 */ 1104 public String pattern() { 1105 return pattern; 1106 } 1107 1108 /** 1109 * <p>Returns the string representation of this pattern. This 1110 * is the regular expression from which this pattern was 1111 * compiled.</p> 1112 * 1113 * @return The string representation of this pattern 1114 * @since 1.5 1115 */ 1116 public String toString() { 1117 return pattern; 1118 } 1119 1120 /** 1121 * Creates a matcher that will match the given input against this pattern. 1122 * 1123 * @param input 1124 * The character sequence to be matched 1125 * 1126 * @return A new matcher for this pattern 1127 */ 1128 public Matcher matcher(CharSequence input) { 1129 if (!compiled) { 1130 synchronized(this) { 1131 if (!compiled) 1132 compile(); 1133 } 1134 } 1135 Matcher m = new Matcher(this, input); 1136 return m; 1137 } 1138 1139 /** 1140 * Returns this pattern's match flags. 1141 * 1142 * @return The match flags specified when this pattern was compiled 1143 */ 1144 public int flags() { 1145 return flags; 1146 } 1147 1148 /** 1149 * Compiles the given regular expression and attempts to match the given 1150 * input against it. 1151 * 1152 * <p> An invocation of this convenience method of the form 1153 * 1154 * <blockquote><pre> 1155 * Pattern.matches(regex, input);</pre></blockquote> 1156 * 1157 * behaves in exactly the same way as the expression 1158 * 1159 * <blockquote><pre> 1160 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote> 1161 * 1162 * <p> If a pattern is to be used multiple times, compiling it once and reusing 1163 * it will be more efficient than invoking this method each time. </p> 1164 * 1165 * @param regex 1166 * The expression to be compiled 1167 * 1168 * @param input 1169 * The character sequence to be matched 1170 * @return whether or not the regular expression matches on the input 1171 * @throws PatternSyntaxException 1172 * If the expression's syntax is invalid 1173 */ 1174 public static boolean matches(String regex, CharSequence input) { 1175 Pattern p = Pattern.compile(regex); 1176 Matcher m = p.matcher(input); 1177 return m.matches(); 1178 } 1179 1180 /** 1181 * Splits the given input sequence around matches of this pattern. 1182 * 1183 * <p> The array returned by this method contains each substring of the 1184 * input sequence that is terminated by another subsequence that matches 1185 * this pattern or is terminated by the end of the input sequence. The 1186 * substrings in the array are in the order in which they occur in the 1187 * input. If this pattern does not match any subsequence of the input then 1188 * the resulting array has just one element, namely the input sequence in 1189 * string form. 1190 * 1191 * <p> When there is a positive-width match at the beginning of the input 1192 * sequence then an empty leading substring is included at the beginning 1193 * of the resulting array. A zero-width match at the beginning however 1194 * never produces such empty leading substring. 1195 * 1196 * <p> The {@code limit} parameter controls the number of times the 1197 * pattern is applied and therefore affects the length of the resulting 1198 * array. If the limit <i>n</i> is greater than zero then the pattern 1199 * will be applied at most <i>n</i> - 1 times, the array's 1200 * length will be no greater than <i>n</i>, and the array's last entry 1201 * will contain all input beyond the last matched delimiter. If <i>n</i> 1202 * is non-positive then the pattern will be applied as many times as 1203 * possible and the array can have any length. If <i>n</i> is zero then 1204 * the pattern will be applied as many times as possible, the array can 1205 * have any length, and trailing empty strings will be discarded. 1206 * 1207 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1208 * results with these parameters: 1209 * 1210 * <blockquote><table cellpadding=1 cellspacing=0 1211 * summary="Split examples showing regex, limit, and result"> 1212 * <tr><th align="left"><i>Regex </i></th> 1213 * <th align="left"><i>Limit </i></th> 1214 * <th align="left"><i>Result </i></th></tr> 1215 * <tr><td align=center>:</td> 1216 * <td align=center>2</td> 1217 * <td>{@code { "boo", "and:foo" }}</td></tr> 1218 * <tr><td align=center>:</td> 1219 * <td align=center>5</td> 1220 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1221 * <tr><td align=center>:</td> 1222 * <td align=center>-2</td> 1223 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1224 * <tr><td align=center>o</td> 1225 * <td align=center>5</td> 1226 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1227 * <tr><td align=center>o</td> 1228 * <td align=center>-2</td> 1229 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1230 * <tr><td align=center>o</td> 1231 * <td align=center>0</td> 1232 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1233 * </table></blockquote> 1234 * 1235 * @param input 1236 * The character sequence to be split 1237 * 1238 * @param limit 1239 * The result threshold, as described above 1240 * 1241 * @return The array of strings computed by splitting the input 1242 * around matches of this pattern 1243 */ 1244 public String[] split(CharSequence input, int limit) { 1245 int index = 0; 1246 boolean matchLimited = limit > 0; 1247 ArrayList<String> matchList = new ArrayList<>(); 1248 Matcher m = matcher(input); 1249 1250 // Add segments before each match found 1251 while(m.find()) { 1252 if (!matchLimited || matchList.size() < limit - 1) { 1253 if (index == 0 && index == m.start() && m.start() == m.end()) { 1254 // no empty leading substring included for zero-width match 1255 // at the beginning of the input char sequence. 1256 continue; 1257 } 1258 String match = input.subSequence(index, m.start()).toString(); 1259 matchList.add(match); 1260 index = m.end(); 1261 } else if (matchList.size() == limit - 1) { // last one 1262 String match = input.subSequence(index, 1263 input.length()).toString(); 1264 matchList.add(match); 1265 index = m.end(); 1266 } 1267 } 1268 1269 // If no match was found, return this 1270 if (index == 0) 1271 return new String[] {input.toString()}; 1272 1273 // Add remaining segment 1274 if (!matchLimited || matchList.size() < limit) 1275 matchList.add(input.subSequence(index, input.length()).toString()); 1276 1277 // Construct result 1278 int resultSize = matchList.size(); 1279 if (limit == 0) 1280 while (resultSize > 0 && matchList.get(resultSize-1).equals("")) 1281 resultSize--; 1282 String[] result = new String[resultSize]; 1283 return matchList.subList(0, resultSize).toArray(result); 1284 } 1285 1286 /** 1287 * Splits the given input sequence around matches of this pattern. 1288 * 1289 * <p> This method works as if by invoking the two-argument {@link 1290 * #split(java.lang.CharSequence, int) split} method with the given input 1291 * sequence and a limit argument of zero. Trailing empty strings are 1292 * therefore not included in the resulting array. </p> 1293 * 1294 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1295 * results with these expressions: 1296 * 1297 * <blockquote><table cellpadding=1 cellspacing=0 1298 * summary="Split examples showing regex and result"> 1299 * <tr><th align="left"><i>Regex </i></th> 1300 * <th align="left"><i>Result</i></th></tr> 1301 * <tr><td align=center>:</td> 1302 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1303 * <tr><td align=center>o</td> 1304 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1305 * </table></blockquote> 1306 * 1307 * 1308 * @param input 1309 * The character sequence to be split 1310 * 1311 * @return The array of strings computed by splitting the input 1312 * around matches of this pattern 1313 */ 1314 public String[] split(CharSequence input) { 1315 return split(input, 0); 1316 } 1317 1318 /** 1319 * Returns a literal pattern {@code String} for the specified 1320 * {@code String}. 1321 * 1322 * <p>This method produces a {@code String} that can be used to 1323 * create a {@code Pattern} that would match the string 1324 * {@code s} as if it were a literal pattern.</p> Metacharacters 1325 * or escape sequences in the input sequence will be given no special 1326 * meaning. 1327 * 1328 * @param s The string to be literalized 1329 * @return A literal string replacement 1330 * @since 1.5 1331 */ 1332 public static String quote(String s) { 1333 int slashEIndex = s.indexOf("\\E"); 1334 if (slashEIndex == -1) 1335 return "\\Q" + s + "\\E"; 1336 1337 int lenHint = s.length(); 1338 lenHint = (lenHint < Integer.MAX_VALUE - 8 - lenHint) ? 1339 (lenHint << 1) : (Integer.MAX_VALUE - 8); 1340 1341 StringBuilder sb = new StringBuilder(lenHint); 1342 sb.append("\\Q"); 1343 int current = 0; 1344 do { 1345 sb.append(s, current, slashEIndex) 1346 .append("\\E\\\\E\\Q"); 1347 current = slashEIndex + 2; 1348 } while ((slashEIndex = s.indexOf("\\E", current)) != -1); 1349 1350 return sb.append(s, current, s.length()) 1351 .append("\\E") 1352 .toString(); 1353 } 1354 1355 /** 1356 * Recompile the Pattern instance from a stream. The original pattern 1357 * string is read in and the object tree is recompiled from it. 1358 */ 1359 private void readObject(java.io.ObjectInputStream s) 1360 throws java.io.IOException, ClassNotFoundException { 1361 1362 // Read in all fields 1363 s.defaultReadObject(); 1364 1365 // Initialize counts 1366 capturingGroupCount = 1; 1367 localCount = 0; 1368 localTCNCount = 0; 1369 1370 // if length > 0, the Pattern is lazily compiled 1371 if (pattern.length() == 0) { 1372 root = new Start(lastAccept); 1373 matchRoot = lastAccept; 1374 compiled = true; 1375 } 1376 } 1377 1378 /** 1379 * This private constructor is used to create all Patterns. The pattern 1380 * string and match flags are all that is needed to completely describe 1381 * a Pattern. An empty pattern string results in an object tree with 1382 * only a Start node and a LastNode node. 1383 */ 1384 private Pattern(String p, int f) { 1385 if ((f & ~ALL_FLAGS) != 0) { 1386 throw new IllegalArgumentException("Unknown flag 0x" 1387 + Integer.toHexString(f)); 1388 } 1389 pattern = p; 1390 flags = f; 1391 1392 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present 1393 if ((flags & UNICODE_CHARACTER_CLASS) != 0) 1394 flags |= UNICODE_CASE; 1395 1396 // Reset group index count 1397 capturingGroupCount = 1; 1398 localCount = 0; 1399 localTCNCount = 0; 1400 1401 if (pattern.length() > 0) { 1402 compile(); 1403 } else { 1404 root = new Start(lastAccept); 1405 matchRoot = lastAccept; 1406 } 1407 } 1408 1409 /** 1410 * The pattern is converted to normalized form ({@link 1411 * java.text.Normalizer.Form.NFC NFC}, canonical decomposition, 1412 * followed by canonical composition for the character class 1413 * part, and {@link java.text.Normalizer.Form.NFD NFD}, 1414 * canonical decomposition) for the rest), and then a pure 1415 * group is constructed to match canonical equivalences of the 1416 * characters. 1417 */ 1418 private static String normalize(String pattern) { 1419 int plen = pattern.length(); 1420 StringBuilder pbuf = new StringBuilder(plen); 1421 char last = 0; 1422 int lastStart = 0; 1423 char cc = 0; 1424 for (int i = 0; i < plen;) { 1425 char c = pattern.charAt(i); 1426 if (cc == 0 && // top level 1427 c == '\\' && i + 1 < plen && pattern.charAt(i + 1) == '\\') { 1428 i += 2; last = 0; 1429 continue; 1430 } 1431 if (c == '[' && last != '\\') { 1432 if (cc == 0) { 1433 if (lastStart < i) 1434 normalizeSlice(pattern, lastStart, i, pbuf); 1435 lastStart = i; 1436 } 1437 cc++; 1438 } else if (c == ']' && last != '\\') { 1439 cc--; 1440 if (cc == 0) { 1441 normalizeClazz(pattern, lastStart, i + 1, pbuf); 1442 lastStart = i + 1; 1443 } 1444 } 1445 last = c; 1446 i++; 1447 } 1448 assert (cc == 0); 1449 if (lastStart < plen) 1450 normalizeSlice(pattern, lastStart, plen, pbuf); 1451 return pbuf.toString(); 1452 } 1453 1454 private static void normalizeSlice(String src, int off, int limit, 1455 StringBuilder dst) 1456 { 1457 int len = src.length(); 1458 int off0 = off; 1459 while (off < limit && ASCII.isAscii(src.charAt(off))) { 1460 off++; 1461 } 1462 if (off == limit) { 1463 dst.append(src, off0, limit); 1464 return; 1465 } 1466 off--; 1467 if (off < off0) 1468 off = off0; 1469 else 1470 dst.append(src, off0, off); 1471 while (off < limit) { 1472 int ch0 = src.codePointAt(off); 1473 if (".$|()[]{}^?*+\\".indexOf(ch0) != -1) { 1474 dst.append((char)ch0); 1475 off++; 1476 continue; 1477 } 1478 int j = off + Character.charCount(ch0); 1479 int ch1; 1480 while (j < limit) { 1481 ch1 = src.codePointAt(j); 1482 if (Grapheme.isBoundary(ch0, ch1)) 1483 break; 1484 ch0 = ch1; 1485 j += Character.charCount(ch1); 1486 } 1487 String seq = src.substring(off, j); 1488 String nfd = Normalizer.normalize(seq, Normalizer.Form.NFD); 1489 off = j; 1490 if (nfd.length() > 1) { 1491 ch0 = nfd.codePointAt(0); 1492 ch1 = nfd.codePointAt(Character.charCount(ch0)); 1493 if (Character.getType(ch1) == Character.NON_SPACING_MARK) { 1494 Set<String> altns = new LinkedHashSet<>(); 1495 altns.add(seq); 1496 produceEquivalentAlternation(nfd, altns); 1497 dst.append("(?:"); 1498 altns.forEach( s -> dst.append(s).append('|')); 1499 dst.delete(dst.length() - 1, dst.length()); 1500 dst.append(")"); 1501 continue; 1502 } 1503 } 1504 String nfc = Normalizer.normalize(seq, Normalizer.Form.NFC); 1505 if (!seq.equals(nfc) && !nfd.equals(nfc)) 1506 dst.append("(?:" + seq + "|" + nfd + "|" + nfc + ")"); 1507 else if (!seq.equals(nfd)) 1508 dst.append("(?:" + seq + "|" + nfd + ")"); 1509 else 1510 dst.append(seq); 1511 } 1512 } 1513 1514 private static void normalizeClazz(String src, int off, int limit, 1515 StringBuilder dst) 1516 { 1517 dst.append(Normalizer.normalize(src.substring(off, limit), Form.NFC)); 1518 } 1519 1520 /** 1521 * Given a specific sequence composed of a regular character and 1522 * combining marks that follow it, produce the alternation that will 1523 * match all canonical equivalences of that sequence. 1524 */ 1525 private static void produceEquivalentAlternation(String src, 1526 Set<String> dst) 1527 { 1528 int len = countChars(src, 0, 1); 1529 if (src.length() == len) { 1530 dst.add(src); // source has one character. 1531 return; 1532 } 1533 String base = src.substring(0,len); 1534 String combiningMarks = src.substring(len); 1535 String[] perms = producePermutations(combiningMarks); 1536 // Add combined permutations 1537 for(int x = 0; x < perms.length; x++) { 1538 String next = base + perms[x]; 1539 dst.add(next); 1540 next = composeOneStep(next); 1541 if (next != null) { 1542 produceEquivalentAlternation(next, dst); 1543 } 1544 } 1545 } 1546 1547 /** 1548 * Returns an array of strings that have all the possible 1549 * permutations of the characters in the input string. 1550 * This is used to get a list of all possible orderings 1551 * of a set of combining marks. Note that some of the permutations 1552 * are invalid because of combining class collisions, and these 1553 * possibilities must be removed because they are not canonically 1554 * equivalent. 1555 */ 1556 private static String[] producePermutations(String input) { 1557 if (input.length() == countChars(input, 0, 1)) 1558 return new String[] {input}; 1559 1560 if (input.length() == countChars(input, 0, 2)) { 1561 int c0 = Character.codePointAt(input, 0); 1562 int c1 = Character.codePointAt(input, Character.charCount(c0)); 1563 if (getClass(c1) == getClass(c0)) { 1564 return new String[] {input}; 1565 } 1566 String[] result = new String[2]; 1567 result[0] = input; 1568 StringBuilder sb = new StringBuilder(2); 1569 sb.appendCodePoint(c1); 1570 sb.appendCodePoint(c0); 1571 result[1] = sb.toString(); 1572 return result; 1573 } 1574 1575 int length = 1; 1576 int nCodePoints = countCodePoints(input); 1577 for(int x=1; x<nCodePoints; x++) 1578 length = length * (x+1); 1579 1580 String[] temp = new String[length]; 1581 1582 int combClass[] = new int[nCodePoints]; 1583 for(int x=0, i=0; x<nCodePoints; x++) { 1584 int c = Character.codePointAt(input, i); 1585 combClass[x] = getClass(c); 1586 i += Character.charCount(c); 1587 } 1588 1589 // For each char, take it out and add the permutations 1590 // of the remaining chars 1591 int index = 0; 1592 int len; 1593 // offset maintains the index in code units. 1594 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) { 1595 len = countChars(input, offset, 1); 1596 for(int y=x-1; y>=0; y--) { 1597 if (combClass[y] == combClass[x]) { 1598 continue loop; 1599 } 1600 } 1601 StringBuilder sb = new StringBuilder(input); 1602 String otherChars = sb.delete(offset, offset+len).toString(); 1603 String[] subResult = producePermutations(otherChars); 1604 1605 String prefix = input.substring(offset, offset+len); 1606 for (String sre : subResult) 1607 temp[index++] = prefix + sre; 1608 } 1609 String[] result = new String[index]; 1610 System.arraycopy(temp, 0, result, 0, index); 1611 return result; 1612 } 1613 1614 private static int getClass(int c) { 1615 return sun.text.Normalizer.getCombiningClass(c); 1616 } 1617 1618 /** 1619 * Attempts to compose input by combining the first character 1620 * with the first combining mark following it. Returns a String 1621 * that is the composition of the leading character with its first 1622 * combining mark followed by the remaining combining marks. Returns 1623 * null if the first two characters cannot be further composed. 1624 */ 1625 private static String composeOneStep(String input) { 1626 int len = countChars(input, 0, 2); 1627 String firstTwoCharacters = input.substring(0, len); 1628 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC); 1629 if (result.equals(firstTwoCharacters)) 1630 return null; 1631 else { 1632 String remainder = input.substring(len); 1633 return result + remainder; 1634 } 1635 } 1636 1637 /** 1638 * Preprocess any \Q...\E sequences in `temp', meta-quoting them. 1639 * See the description of `quotemeta' in perlfunc(1). 1640 */ 1641 private void RemoveQEQuoting() { 1642 final int pLen = patternLength; 1643 int i = 0; 1644 while (i < pLen-1) { 1645 if (temp[i] != '\\') 1646 i += 1; 1647 else if (temp[i + 1] != 'Q') 1648 i += 2; 1649 else 1650 break; 1651 } 1652 if (i >= pLen - 1) // No \Q sequence found 1653 return; 1654 int j = i; 1655 i += 2; 1656 int[] newtemp = new int[j + 3*(pLen-i) + 2]; 1657 System.arraycopy(temp, 0, newtemp, 0, j); 1658 1659 boolean inQuote = true; 1660 boolean beginQuote = true; 1661 while (i < pLen) { 1662 int c = temp[i++]; 1663 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) { 1664 newtemp[j++] = c; 1665 } else if (ASCII.isDigit(c)) { 1666 if (beginQuote) { 1667 /* 1668 * A unicode escape \[0xu] could be before this quote, 1669 * and we don't want this numeric char to processed as 1670 * part of the escape. 1671 */ 1672 newtemp[j++] = '\\'; 1673 newtemp[j++] = 'x'; 1674 newtemp[j++] = '3'; 1675 } 1676 newtemp[j++] = c; 1677 } else if (c != '\\') { 1678 if (inQuote) newtemp[j++] = '\\'; 1679 newtemp[j++] = c; 1680 } else if (inQuote) { 1681 if (temp[i] == 'E') { 1682 i++; 1683 inQuote = false; 1684 } else { 1685 newtemp[j++] = '\\'; 1686 newtemp[j++] = '\\'; 1687 } 1688 } else { 1689 if (temp[i] == 'Q') { 1690 i++; 1691 inQuote = true; 1692 beginQuote = true; 1693 continue; 1694 } else { 1695 newtemp[j++] = c; 1696 if (i != pLen) 1697 newtemp[j++] = temp[i++]; 1698 } 1699 } 1700 1701 beginQuote = false; 1702 } 1703 1704 patternLength = j; 1705 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination 1706 } 1707 1708 /** 1709 * Copies regular expression to an int array and invokes the parsing 1710 * of the expression which will create the object tree. 1711 */ 1712 private void compile() { 1713 // Handle canonical equivalences 1714 if (has(CANON_EQ) && !has(LITERAL)) { 1715 normalizedPattern = normalize(pattern); 1716 } else { 1717 normalizedPattern = pattern; 1718 } 1719 patternLength = normalizedPattern.length(); 1720 1721 // Copy pattern to int array for convenience 1722 // Use double zero to terminate pattern 1723 temp = new int[patternLength + 2]; 1724 1725 hasSupplementary = false; 1726 int c, count = 0; 1727 // Convert all chars into code points 1728 for (int x = 0; x < patternLength; x += Character.charCount(c)) { 1729 c = normalizedPattern.codePointAt(x); 1730 if (isSupplementary(c)) { 1731 hasSupplementary = true; 1732 } 1733 temp[count++] = c; 1734 } 1735 1736 patternLength = count; // patternLength now in code points 1737 1738 if (! has(LITERAL)) 1739 RemoveQEQuoting(); 1740 1741 // Allocate all temporary objects here. 1742 buffer = new int[32]; 1743 groupNodes = new GroupHead[10]; 1744 namedGroups = null; 1745 topClosureNodes = new ArrayList<>(10); 1746 1747 if (has(LITERAL)) { 1748 // Literal pattern handling 1749 matchRoot = newSlice(temp, patternLength, hasSupplementary); 1750 matchRoot.next = lastAccept; 1751 } else { 1752 // Start recursive descent parsing 1753 matchRoot = expr(lastAccept); 1754 // Check extra pattern characters 1755 if (patternLength != cursor) { 1756 if (peek() == ')') { 1757 throw error("Unmatched closing ')'"); 1758 } else { 1759 throw error("Unexpected internal error"); 1760 } 1761 } 1762 } 1763 1764 // Peephole optimization 1765 if (matchRoot instanceof Slice) { 1766 root = BnM.optimize(matchRoot); 1767 if (root == matchRoot) { 1768 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1769 } 1770 } else if (matchRoot instanceof Begin || matchRoot instanceof First) { 1771 root = matchRoot; 1772 } else { 1773 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1774 } 1775 1776 // Optimize the greedy Loop to prevent exponential backtracking, IF there 1777 // is no group ref in this pattern. With a non-negative localTCNCount value, 1778 // the greedy type Loop, Curly will skip the backtracking for any starting 1779 // position "i" that failed in the past. 1780 if (!hasGroupRef) { 1781 for (Node node : topClosureNodes) { 1782 if (node instanceof Loop) { 1783 // non-deterministic-greedy-group 1784 ((Loop)node).posIndex = localTCNCount++; 1785 } 1786 } 1787 } 1788 1789 // Release temporary storage 1790 temp = null; 1791 buffer = null; 1792 groupNodes = null; 1793 patternLength = 0; 1794 compiled = true; 1795 topClosureNodes = null; 1796 } 1797 1798 Map<String, Integer> namedGroups() { 1799 Map<String, Integer> groups = namedGroups; 1800 if (groups == null) { 1801 namedGroups = groups = new HashMap<>(2); 1802 } 1803 return groups; 1804 } 1805 1806 /** 1807 * Used to accumulate information about a subtree of the object graph 1808 * so that optimizations can be applied to the subtree. 1809 */ 1810 static final class TreeInfo { 1811 int minLength; 1812 int maxLength; 1813 boolean maxValid; 1814 boolean deterministic; 1815 1816 TreeInfo() { 1817 reset(); 1818 } 1819 void reset() { 1820 minLength = 0; 1821 maxLength = 0; 1822 maxValid = true; 1823 deterministic = true; 1824 } 1825 } 1826 1827 /* 1828 * The following private methods are mainly used to improve the 1829 * readability of the code. In order to let the Java compiler easily 1830 * inline them, we should not put many assertions or error checks in them. 1831 */ 1832 1833 /** 1834 * Indicates whether a particular flag is set or not. 1835 */ 1836 private boolean has(int f) { 1837 return (flags & f) != 0; 1838 } 1839 1840 /** 1841 * Match next character, signal error if failed. 1842 */ 1843 private void accept(int ch, String s) { 1844 int testChar = temp[cursor++]; 1845 if (has(COMMENTS)) 1846 testChar = parsePastWhitespace(testChar); 1847 if (ch != testChar) { 1848 throw error(s); 1849 } 1850 } 1851 1852 /** 1853 * Mark the end of pattern with a specific character. 1854 */ 1855 private void mark(int c) { 1856 temp[patternLength] = c; 1857 } 1858 1859 /** 1860 * Peek the next character, and do not advance the cursor. 1861 */ 1862 private int peek() { 1863 int ch = temp[cursor]; 1864 if (has(COMMENTS)) 1865 ch = peekPastWhitespace(ch); 1866 return ch; 1867 } 1868 1869 /** 1870 * Read the next character, and advance the cursor by one. 1871 */ 1872 private int read() { 1873 int ch = temp[cursor++]; 1874 if (has(COMMENTS)) 1875 ch = parsePastWhitespace(ch); 1876 return ch; 1877 } 1878 1879 /** 1880 * Read the next character, and advance the cursor by one, 1881 * ignoring the COMMENTS setting 1882 */ 1883 private int readEscaped() { 1884 int ch = temp[cursor++]; 1885 return ch; 1886 } 1887 1888 /** 1889 * Advance the cursor by one, and peek the next character. 1890 */ 1891 private int next() { 1892 int ch = temp[++cursor]; 1893 if (has(COMMENTS)) 1894 ch = peekPastWhitespace(ch); 1895 return ch; 1896 } 1897 1898 /** 1899 * Advance the cursor by one, and peek the next character, 1900 * ignoring the COMMENTS setting 1901 */ 1902 private int nextEscaped() { 1903 int ch = temp[++cursor]; 1904 return ch; 1905 } 1906 1907 /** 1908 * If in xmode peek past whitespace and comments. 1909 */ 1910 private int peekPastWhitespace(int ch) { 1911 while (ASCII.isSpace(ch) || ch == '#') { 1912 while (ASCII.isSpace(ch)) 1913 ch = temp[++cursor]; 1914 if (ch == '#') { 1915 ch = peekPastLine(); 1916 } 1917 } 1918 return ch; 1919 } 1920 1921 /** 1922 * If in xmode parse past whitespace and comments. 1923 */ 1924 private int parsePastWhitespace(int ch) { 1925 while (ASCII.isSpace(ch) || ch == '#') { 1926 while (ASCII.isSpace(ch)) 1927 ch = temp[cursor++]; 1928 if (ch == '#') 1929 ch = parsePastLine(); 1930 } 1931 return ch; 1932 } 1933 1934 /** 1935 * xmode parse past comment to end of line. 1936 */ 1937 private int parsePastLine() { 1938 int ch = temp[cursor++]; 1939 while (ch != 0 && !isLineSeparator(ch)) 1940 ch = temp[cursor++]; 1941 return ch; 1942 } 1943 1944 /** 1945 * xmode peek past comment to end of line. 1946 */ 1947 private int peekPastLine() { 1948 int ch = temp[++cursor]; 1949 while (ch != 0 && !isLineSeparator(ch)) 1950 ch = temp[++cursor]; 1951 return ch; 1952 } 1953 1954 /** 1955 * Determines if character is a line separator in the current mode 1956 */ 1957 private boolean isLineSeparator(int ch) { 1958 if (has(UNIX_LINES)) { 1959 return ch == '\n'; 1960 } else { 1961 return (ch == '\n' || 1962 ch == '\r' || 1963 (ch|1) == '\u2029' || 1964 ch == '\u0085'); 1965 } 1966 } 1967 1968 /** 1969 * Read the character after the next one, and advance the cursor by two. 1970 */ 1971 private int skip() { 1972 int i = cursor; 1973 int ch = temp[i+1]; 1974 cursor = i + 2; 1975 return ch; 1976 } 1977 1978 /** 1979 * Unread one next character, and retreat cursor by one. 1980 */ 1981 private void unread() { 1982 cursor--; 1983 } 1984 1985 /** 1986 * Internal method used for handling all syntax errors. The pattern is 1987 * displayed with a pointer to aid in locating the syntax error. 1988 */ 1989 private PatternSyntaxException error(String s) { 1990 return new PatternSyntaxException(s, normalizedPattern, cursor - 1); 1991 } 1992 1993 /** 1994 * Determines if there is any supplementary character or unpaired 1995 * surrogate in the specified range. 1996 */ 1997 private boolean findSupplementary(int start, int end) { 1998 for (int i = start; i < end; i++) { 1999 if (isSupplementary(temp[i])) 2000 return true; 2001 } 2002 return false; 2003 } 2004 2005 /** 2006 * Determines if the specified code point is a supplementary 2007 * character or unpaired surrogate. 2008 */ 2009 private static final boolean isSupplementary(int ch) { 2010 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || 2011 Character.isSurrogate((char)ch); 2012 } 2013 2014 /** 2015 * The following methods handle the main parsing. They are sorted 2016 * according to their precedence order, the lowest one first. 2017 */ 2018 2019 /** 2020 * The expression is parsed with branch nodes added for alternations. 2021 * This may be called recursively to parse sub expressions that may 2022 * contain alternations. 2023 */ 2024 private Node expr(Node end) { 2025 Node prev = null; 2026 Node firstTail = null; 2027 Branch branch = null; 2028 Node branchConn = null; 2029 2030 for (;;) { 2031 Node node = sequence(end); 2032 Node nodeTail = root; //double return 2033 if (prev == null) { 2034 prev = node; 2035 firstTail = nodeTail; 2036 } else { 2037 // Branch 2038 if (branchConn == null) { 2039 branchConn = new BranchConn(); 2040 branchConn.next = end; 2041 } 2042 if (node == end) { 2043 // if the node returned from sequence() is "end" 2044 // we have an empty expr, set a null atom into 2045 // the branch to indicate to go "next" directly. 2046 node = null; 2047 } else { 2048 // the "tail.next" of each atom goes to branchConn 2049 nodeTail.next = branchConn; 2050 } 2051 if (prev == branch) { 2052 branch.add(node); 2053 } else { 2054 if (prev == end) { 2055 prev = null; 2056 } else { 2057 // replace the "end" with "branchConn" at its tail.next 2058 // when put the "prev" into the branch as the first atom. 2059 firstTail.next = branchConn; 2060 } 2061 prev = branch = new Branch(prev, node, branchConn); 2062 } 2063 } 2064 if (peek() != '|') { 2065 return prev; 2066 } 2067 next(); 2068 } 2069 } 2070 2071 @SuppressWarnings("fallthrough") 2072 /** 2073 * Parsing of sequences between alternations. 2074 */ 2075 private Node sequence(Node end) { 2076 Node head = null; 2077 Node tail = null; 2078 Node node = null; 2079 LOOP: 2080 for (;;) { 2081 int ch = peek(); 2082 switch (ch) { 2083 case '(': 2084 // Because group handles its own closure, 2085 // we need to treat it differently 2086 node = group0(); 2087 // Check for comment or flag group 2088 if (node == null) 2089 continue; 2090 if (head == null) 2091 head = node; 2092 else 2093 tail.next = node; 2094 // Double return: Tail was returned in root 2095 tail = root; 2096 continue; 2097 case '[': 2098 if (has(CANON_EQ) && !has(LITERAL)) 2099 node = new NFCCharProperty(clazz(true)); 2100 else 2101 node = newCharProperty(clazz(true)); 2102 break; 2103 case '\\': 2104 ch = nextEscaped(); 2105 if (ch == 'p' || ch == 'P') { 2106 boolean oneLetter = true; 2107 boolean comp = (ch == 'P'); 2108 ch = next(); // Consume { if present 2109 if (ch != '{') { 2110 unread(); 2111 } else { 2112 oneLetter = false; 2113 } 2114 // node = newCharProperty(family(oneLetter, comp)); 2115 if (has(CANON_EQ) && !has(LITERAL)) 2116 node = new NFCCharProperty(family(oneLetter, comp)); 2117 else 2118 node = newCharProperty(family(oneLetter, comp)); 2119 } else { 2120 unread(); 2121 node = atom(); 2122 } 2123 break; 2124 case '^': 2125 next(); 2126 if (has(MULTILINE)) { 2127 if (has(UNIX_LINES)) 2128 node = new UnixCaret(); 2129 else 2130 node = new Caret(); 2131 } else { 2132 node = new Begin(); 2133 } 2134 break; 2135 case '$': 2136 next(); 2137 if (has(UNIX_LINES)) 2138 node = new UnixDollar(has(MULTILINE)); 2139 else 2140 node = new Dollar(has(MULTILINE)); 2141 break; 2142 case '.': 2143 next(); 2144 if (has(DOTALL)) { 2145 node = new CharProperty(ALL()); 2146 } else { 2147 if (has(UNIX_LINES)) { 2148 node = new CharProperty(UNIXDOT()); 2149 } else { 2150 node = new CharProperty(DOT()); 2151 } 2152 } 2153 break; 2154 case '|': 2155 case ')': 2156 break LOOP; 2157 case ']': // Now interpreting dangling ] and } as literals 2158 case '}': 2159 node = atom(); 2160 break; 2161 case '?': 2162 case '*': 2163 case '+': 2164 next(); 2165 throw error("Dangling meta character '" + ((char)ch) + "'"); 2166 case 0: 2167 if (cursor >= patternLength) { 2168 break LOOP; 2169 } 2170 // Fall through 2171 default: 2172 node = atom(); 2173 break; 2174 } 2175 2176 node = closure(node); 2177 /* save the top dot-greedy nodes (.*, .+) as well 2178 if (node instanceof GreedyCharProperty && 2179 ((GreedyCharProperty)node).cp instanceof Dot) { 2180 topClosureNodes.add(node); 2181 } 2182 */ 2183 if (head == null) { 2184 head = tail = node; 2185 } else { 2186 tail.next = node; 2187 tail = node; 2188 } 2189 } 2190 if (head == null) { 2191 return end; 2192 } 2193 tail.next = end; 2194 root = tail; //double return 2195 return head; 2196 } 2197 2198 @SuppressWarnings("fallthrough") 2199 /** 2200 * Parse and add a new Single or Slice. 2201 */ 2202 private Node atom() { 2203 int first = 0; 2204 int prev = -1; 2205 boolean hasSupplementary = false; 2206 int ch = peek(); 2207 for (;;) { 2208 switch (ch) { 2209 case '*': 2210 case '+': 2211 case '?': 2212 case '{': 2213 if (first > 1) { 2214 cursor = prev; // Unwind one character 2215 first--; 2216 } 2217 break; 2218 case '$': 2219 case '.': 2220 case '^': 2221 case '(': 2222 case '[': 2223 case '|': 2224 case ')': 2225 break; 2226 case '\\': 2227 ch = nextEscaped(); 2228 if (ch == 'p' || ch == 'P') { // Property 2229 if (first > 0) { // Slice is waiting; handle it first 2230 unread(); 2231 break; 2232 } else { // No slice; just return the family node 2233 boolean comp = (ch == 'P'); 2234 boolean oneLetter = true; 2235 ch = next(); // Consume { if present 2236 if (ch != '{') 2237 unread(); 2238 else 2239 oneLetter = false; 2240 if (has(CANON_EQ) && !has(LITERAL)) 2241 return new NFCCharProperty(family(oneLetter, comp)); 2242 else 2243 return newCharProperty(family(oneLetter, comp)); 2244 } 2245 } 2246 unread(); 2247 prev = cursor; 2248 ch = escape(false, first == 0, false); 2249 if (ch >= 0) { 2250 append(ch, first); 2251 first++; 2252 if (isSupplementary(ch)) { 2253 hasSupplementary = true; 2254 } 2255 ch = peek(); 2256 continue; 2257 } else if (first == 0) { 2258 return root; 2259 } 2260 // Unwind meta escape sequence 2261 cursor = prev; 2262 break; 2263 case 0: 2264 if (cursor >= patternLength) { 2265 break; 2266 } 2267 // Fall through 2268 default: 2269 prev = cursor; 2270 append(ch, first); 2271 first++; 2272 if (isSupplementary(ch)) { 2273 hasSupplementary = true; 2274 } 2275 ch = next(); 2276 continue; 2277 } 2278 break; 2279 } 2280 if (first == 1) { 2281 return newCharProperty(single(buffer[0])); 2282 } else { 2283 return newSlice(buffer, first, hasSupplementary); 2284 } 2285 } 2286 2287 private void append(int ch, int len) { 2288 if (len >= buffer.length) { 2289 int[] tmp = new int[len+len]; 2290 System.arraycopy(buffer, 0, tmp, 0, len); 2291 buffer = tmp; 2292 } 2293 buffer[len] = ch; 2294 } 2295 2296 /** 2297 * Parses a backref greedily, taking as many numbers as it 2298 * can. The first digit is always treated as a backref, but 2299 * multi digit numbers are only treated as a backref if at 2300 * least that many backrefs exist at this point in the regex. 2301 */ 2302 private Node ref(int refNum) { 2303 boolean done = false; 2304 while(!done) { 2305 int ch = peek(); 2306 switch(ch) { 2307 case '0': 2308 case '1': 2309 case '2': 2310 case '3': 2311 case '4': 2312 case '5': 2313 case '6': 2314 case '7': 2315 case '8': 2316 case '9': 2317 int newRefNum = (refNum * 10) + (ch - '0'); 2318 // Add another number if it doesn't make a group 2319 // that doesn't exist 2320 if (capturingGroupCount - 1 < newRefNum) { 2321 done = true; 2322 break; 2323 } 2324 refNum = newRefNum; 2325 read(); 2326 break; 2327 default: 2328 done = true; 2329 break; 2330 } 2331 } 2332 hasGroupRef = true; 2333 if (has(CASE_INSENSITIVE)) 2334 return new CIBackRef(refNum, has(UNICODE_CASE)); 2335 else 2336 return new BackRef(refNum); 2337 } 2338 2339 /** 2340 * Parses an escape sequence to determine the actual value that needs 2341 * to be matched. 2342 * If -1 is returned and create was true a new object was added to the tree 2343 * to handle the escape sequence. 2344 * If the returned value is greater than zero, it is the value that 2345 * matches the escape sequence. 2346 */ 2347 private int escape(boolean inclass, boolean create, boolean isrange) { 2348 int ch = skip(); 2349 switch (ch) { 2350 case '0': 2351 return o(); 2352 case '1': 2353 case '2': 2354 case '3': 2355 case '4': 2356 case '5': 2357 case '6': 2358 case '7': 2359 case '8': 2360 case '9': 2361 if (inclass) break; 2362 if (create) { 2363 root = ref((ch - '0')); 2364 } 2365 return -1; 2366 case 'A': 2367 if (inclass) break; 2368 if (create) root = new Begin(); 2369 return -1; 2370 case 'B': 2371 if (inclass) break; 2372 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS)); 2373 return -1; 2374 case 'C': 2375 break; 2376 case 'D': 2377 if (create) { 2378 predicate = has(UNICODE_CHARACTER_CLASS) ? 2379 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2380 predicate = predicate.negate(); 2381 if (!inclass) 2382 root = newCharProperty(predicate); 2383 } 2384 return -1; 2385 case 'E': 2386 case 'F': 2387 break; 2388 case 'G': 2389 if (inclass) break; 2390 if (create) root = new LastMatch(); 2391 return -1; 2392 case 'H': 2393 if (create) { 2394 predicate = HorizWS().negate(); 2395 if (!inclass) 2396 root = newCharProperty(predicate); 2397 } 2398 return -1; 2399 case 'I': 2400 case 'J': 2401 case 'K': 2402 case 'L': 2403 case 'M': 2404 break; 2405 case 'N': 2406 return N(); 2407 case 'O': 2408 case 'P': 2409 case 'Q': 2410 break; 2411 case 'R': 2412 if (inclass) break; 2413 if (create) root = new LineEnding(); 2414 return -1; 2415 case 'S': 2416 if (create) { 2417 predicate = has(UNICODE_CHARACTER_CLASS) ? 2418 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2419 predicate = predicate.negate(); 2420 if (!inclass) 2421 root = newCharProperty(predicate); 2422 } 2423 return -1; 2424 case 'T': 2425 case 'U': 2426 break; 2427 case 'V': 2428 if (create) { 2429 predicate = VertWS().negate(); 2430 if (!inclass) 2431 root = newCharProperty(predicate); 2432 } 2433 return -1; 2434 case 'W': 2435 if (create) { 2436 predicate = has(UNICODE_CHARACTER_CLASS) ? 2437 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2438 predicate = predicate.negate(); 2439 if (!inclass) 2440 root = newCharProperty(predicate); 2441 } 2442 return -1; 2443 case 'X': 2444 if (inclass) break; 2445 if (create) { 2446 root = new XGrapheme(); 2447 } 2448 return -1; 2449 case 'Y': 2450 break; 2451 case 'Z': 2452 if (inclass) break; 2453 if (create) { 2454 if (has(UNIX_LINES)) 2455 root = new UnixDollar(false); 2456 else 2457 root = new Dollar(false); 2458 } 2459 return -1; 2460 case 'a': 2461 return '\007'; 2462 case 'b': 2463 if (inclass) break; 2464 if (create) { 2465 if (peek() == '{') { 2466 if (skip() == 'g') { 2467 if (read() == '}') { 2468 root = new GraphemeBound(); 2469 return -1; 2470 } 2471 break; // error missing trailing } 2472 } 2473 unread(); unread(); 2474 } 2475 root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS)); 2476 } 2477 return -1; 2478 case 'c': 2479 return c(); 2480 case 'd': 2481 if (create) { 2482 predicate = has(UNICODE_CHARACTER_CLASS) ? 2483 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2484 if (!inclass) 2485 root = newCharProperty(predicate); 2486 } 2487 return -1; 2488 case 'e': 2489 return '\033'; 2490 case 'f': 2491 return '\f'; 2492 case 'g': 2493 break; 2494 case 'h': 2495 if (create) { 2496 predicate = HorizWS(); 2497 if (!inclass) 2498 root = newCharProperty(predicate); 2499 } 2500 return -1; 2501 case 'i': 2502 case 'j': 2503 break; 2504 case 'k': 2505 if (inclass) 2506 break; 2507 if (read() != '<') 2508 throw error("\\k is not followed by '<' for named capturing group"); 2509 String name = groupname(read()); 2510 if (!namedGroups().containsKey(name)) 2511 throw error("(named capturing group <"+ name+"> does not exit"); 2512 if (create) { 2513 hasGroupRef = true; 2514 if (has(CASE_INSENSITIVE)) 2515 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE)); 2516 else 2517 root = new BackRef(namedGroups().get(name)); 2518 } 2519 return -1; 2520 case 'l': 2521 case 'm': 2522 break; 2523 case 'n': 2524 return '\n'; 2525 case 'o': 2526 case 'p': 2527 case 'q': 2528 break; 2529 case 'r': 2530 return '\r'; 2531 case 's': 2532 if (create) { 2533 predicate = has(UNICODE_CHARACTER_CLASS) ? 2534 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2535 if (!inclass) 2536 root = newCharProperty(predicate); 2537 } 2538 return -1; 2539 case 't': 2540 return '\t'; 2541 case 'u': 2542 return u(); 2543 case 'v': 2544 // '\v' was implemented as VT/0x0B in releases < 1.8 (though 2545 // undocumented). In JDK8 '\v' is specified as a predefined 2546 // character class for all vertical whitespace characters. 2547 // So [-1, root=VertWS node] pair is returned (instead of a 2548 // single 0x0B). This breaks the range if '\v' is used as 2549 // the start or end value, such as [\v-...] or [...-\v], in 2550 // which a single definite value (0x0B) is expected. For 2551 // compatibility concern '\013'/0x0B is returned if isrange. 2552 if (isrange) 2553 return '\013'; 2554 if (create) { 2555 predicate = VertWS(); 2556 if (!inclass) 2557 root = newCharProperty(predicate); 2558 } 2559 return -1; 2560 case 'w': 2561 if (create) { 2562 predicate = has(UNICODE_CHARACTER_CLASS) ? 2563 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2564 if (!inclass) 2565 root = newCharProperty(predicate); 2566 } 2567 return -1; 2568 case 'x': 2569 return x(); 2570 case 'y': 2571 break; 2572 case 'z': 2573 if (inclass) break; 2574 if (create) root = new End(); 2575 return -1; 2576 default: 2577 return ch; 2578 } 2579 throw error("Illegal/unsupported escape sequence"); 2580 } 2581 2582 /** 2583 * Parse a character class, and return the node that matches it. 2584 * 2585 * Consumes a ] on the way out if consume is true. Usually consume 2586 * is true except for the case of [abc&&def] where def is a separate 2587 * right hand node with "understood" brackets. 2588 */ 2589 private CharPredicate clazz(boolean consume) { 2590 CharPredicate prev = null; 2591 CharPredicate curr = null; 2592 BitClass bits = new BitClass(); 2593 BmpCharPredicate bitsP = ch -> ch < 256 && bits.bits[ch]; 2594 2595 boolean isNeg = false; 2596 boolean hasBits = false; 2597 int ch = next(); 2598 2599 // Negates if first char in a class, otherwise literal 2600 if (ch == '^' && temp[cursor-1] == '[') { 2601 ch = next(); 2602 isNeg = true; 2603 } 2604 for (;;) { 2605 switch (ch) { 2606 case '[': 2607 curr = clazz(true); 2608 if (prev == null) 2609 prev = curr; 2610 else 2611 prev = prev.union(curr); 2612 ch = peek(); 2613 continue; 2614 case '&': 2615 ch = next(); 2616 if (ch == '&') { 2617 ch = next(); 2618 CharPredicate right = null; 2619 while (ch != ']' && ch != '&') { 2620 if (ch == '[') { 2621 if (right == null) 2622 right = clazz(true); 2623 else 2624 right = right.union(clazz(true)); 2625 } else { // abc&&def 2626 unread(); 2627 right = clazz(false); 2628 } 2629 ch = peek(); 2630 } 2631 if (hasBits) { 2632 // bits used, union has high precedence 2633 if (prev == null) { 2634 prev = curr = bitsP; 2635 } else { 2636 prev = prev.union(bitsP); 2637 } 2638 hasBits = false; 2639 } 2640 if (right != null) 2641 curr = right; 2642 if (prev == null) { 2643 if (right == null) 2644 throw error("Bad class syntax"); 2645 else 2646 prev = right; 2647 } else { 2648 prev = prev.and(curr); 2649 } 2650 } else { 2651 // treat as a literal & 2652 unread(); 2653 break; 2654 } 2655 continue; 2656 case 0: 2657 if (cursor >= patternLength) 2658 throw error("Unclosed character class"); 2659 break; 2660 case ']': 2661 if (prev != null || hasBits) { 2662 if (consume) 2663 next(); 2664 if (prev == null) 2665 prev = bitsP; 2666 else if (hasBits) 2667 prev = prev.union(bitsP); 2668 if (isNeg) 2669 return prev.negate(); 2670 return prev; 2671 } 2672 break; 2673 default: 2674 break; 2675 } 2676 curr = range(bits); 2677 if (curr == null) { // the bits used 2678 hasBits = true; 2679 } else { 2680 if (prev == null) 2681 prev = curr; 2682 else if (prev != curr) 2683 prev = prev.union(curr); 2684 } 2685 ch = peek(); 2686 } 2687 } 2688 2689 private CharPredicate bitsOrSingle(BitClass bits, int ch) { 2690 /* Bits can only handle codepoints in [u+0000-u+00ff] range. 2691 Use "single" node instead of bits when dealing with unicode 2692 case folding for codepoints listed below. 2693 (1)Uppercase out of range: u+00ff, u+00b5 2694 toUpperCase(u+00ff) -> u+0178 2695 toUpperCase(u+00b5) -> u+039c 2696 (2)LatinSmallLetterLongS u+17f 2697 toUpperCase(u+017f) -> u+0053 2698 (3)LatinSmallLetterDotlessI u+131 2699 toUpperCase(u+0131) -> u+0049 2700 (4)LatinCapitalLetterIWithDotAbove u+0130 2701 toLowerCase(u+0130) -> u+0069 2702 (5)KelvinSign u+212a 2703 toLowerCase(u+212a) ==> u+006B 2704 (6)AngstromSign u+212b 2705 toLowerCase(u+212b) ==> u+00e5 2706 */ 2707 if (ch < 256 && 2708 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) && 2709 (ch == 0xff || ch == 0xb5 || 2710 ch == 0x49 || ch == 0x69 || //I and i 2711 ch == 0x53 || ch == 0x73 || //S and s 2712 ch == 0x4b || ch == 0x6b || //K and k 2713 ch == 0xc5 || ch == 0xe5))) { //A+ring 2714 bits.add(ch, flags()); 2715 return null; 2716 } 2717 return single(ch); 2718 } 2719 2720 /** 2721 * Returns a suitably optimized, single character predicate 2722 */ 2723 private CharPredicate single(final int ch) { 2724 if (has(CASE_INSENSITIVE)) { 2725 int lower, upper; 2726 if (has(UNICODE_CASE)) { 2727 upper = Character.toUpperCase(ch); 2728 lower = Character.toLowerCase(upper); 2729 // Unicode case insensitive matches 2730 if (upper != lower) 2731 return SingleU(lower); 2732 } else if (ASCII.isAscii(ch)) { 2733 lower = ASCII.toLower(ch); 2734 upper = ASCII.toUpper(ch); 2735 // Case insensitive matches a given BMP character 2736 if (lower != upper) 2737 return SingleI(lower, upper); 2738 } 2739 } 2740 if (isSupplementary(ch)) 2741 return SingleS(ch); 2742 return Single(ch); // Match a given BMP character 2743 } 2744 2745 /** 2746 * Parse a single character or a character range in a character class 2747 * and return its representative node. 2748 */ 2749 private CharPredicate range(BitClass bits) { 2750 int ch = peek(); 2751 if (ch == '\\') { 2752 ch = nextEscaped(); 2753 if (ch == 'p' || ch == 'P') { // A property 2754 boolean comp = (ch == 'P'); 2755 boolean oneLetter = true; 2756 // Consume { if present 2757 ch = next(); 2758 if (ch != '{') 2759 unread(); 2760 else 2761 oneLetter = false; 2762 return family(oneLetter, comp); 2763 } else { // ordinary escape 2764 boolean isrange = temp[cursor+1] == '-'; 2765 unread(); 2766 ch = escape(true, true, isrange); 2767 if (ch == -1) 2768 return predicate; 2769 } 2770 } else { 2771 next(); 2772 } 2773 if (ch >= 0) { 2774 if (peek() == '-') { 2775 int endRange = temp[cursor+1]; 2776 if (endRange == '[') { 2777 return bitsOrSingle(bits, ch); 2778 } 2779 if (endRange != ']') { 2780 next(); 2781 int m = peek(); 2782 if (m == '\\') { 2783 m = escape(true, false, true); 2784 } else { 2785 next(); 2786 } 2787 if (m < ch) { 2788 throw error("Illegal character range"); 2789 } 2790 if (has(CASE_INSENSITIVE)) { 2791 if (has(UNICODE_CASE)) 2792 return CIRangeU(ch, m); 2793 return CIRange(ch, m); 2794 } else { 2795 return Range(ch, m); 2796 } 2797 } 2798 } 2799 return bitsOrSingle(bits, ch); 2800 } 2801 throw error("Unexpected character '"+((char)ch)+"'"); 2802 } 2803 2804 /** 2805 * Parses a Unicode character family and returns its representative node. 2806 */ 2807 private CharPredicate family(boolean singleLetter, boolean isComplement) { 2808 next(); 2809 String name; 2810 CharPredicate p = null; 2811 2812 if (singleLetter) { 2813 int c = temp[cursor]; 2814 if (!Character.isSupplementaryCodePoint(c)) { 2815 name = String.valueOf((char)c); 2816 } else { 2817 name = new String(temp, cursor, 1); 2818 } 2819 read(); 2820 } else { 2821 int i = cursor; 2822 mark('}'); 2823 while(read() != '}') { 2824 } 2825 mark('\000'); 2826 int j = cursor; 2827 if (j > patternLength) 2828 throw error("Unclosed character family"); 2829 if (i + 1 >= j) 2830 throw error("Empty character family"); 2831 name = new String(temp, i, j-i-1); 2832 } 2833 2834 int i = name.indexOf('='); 2835 if (i != -1) { 2836 // property construct \p{name=value} 2837 String value = name.substring(i + 1); 2838 name = name.substring(0, i).toLowerCase(Locale.ENGLISH); 2839 switch (name) { 2840 case "sc": 2841 case "script": 2842 p = CharPredicates.forUnicodeScript(value); 2843 break; 2844 case "blk": 2845 case "block": 2846 p = CharPredicates.forUnicodeBlock(value); 2847 break; 2848 case "gc": 2849 case "general_category": 2850 p = CharPredicates.forProperty(value); 2851 break; 2852 default: 2853 break; 2854 } 2855 if (p == null) 2856 throw error("Unknown Unicode property {name=<" + name + ">, " 2857 + "value=<" + value + ">}"); 2858 2859 } else { 2860 if (name.startsWith("In")) { 2861 // \p{InBlockName} 2862 p = CharPredicates.forUnicodeBlock(name.substring(2)); 2863 } else if (name.startsWith("Is")) { 2864 // \p{IsGeneralCategory} and \p{IsScriptName} 2865 name = name.substring(2); 2866 p = CharPredicates.forUnicodeProperty(name); 2867 if (p == null) 2868 p = CharPredicates.forProperty(name); 2869 if (p == null) 2870 p = CharPredicates.forUnicodeScript(name); 2871 } else { 2872 if (has(UNICODE_CHARACTER_CLASS)) { 2873 p = CharPredicates.forPOSIXName(name); 2874 } 2875 if (p == null) 2876 p = CharPredicates.forProperty(name); 2877 } 2878 if (p == null) 2879 throw error("Unknown character property name {In/Is" + name + "}"); 2880 } 2881 if (isComplement) { 2882 // it might be too expensive to detect if a complement of 2883 // CharProperty can match "certain" supplementary. So just 2884 // go with StartS. 2885 hasSupplementary = true; 2886 p = p.negate(); 2887 } 2888 return p; 2889 } 2890 2891 private CharProperty newCharProperty(CharPredicate p) { 2892 if (p == null) 2893 return null; 2894 if (p instanceof BmpCharPredicate) 2895 return new BmpCharProperty((BmpCharPredicate)p); 2896 else 2897 return new CharProperty(p); 2898 } 2899 2900 /** 2901 * Parses and returns the name of a "named capturing group", the trailing 2902 * ">" is consumed after parsing. 2903 */ 2904 private String groupname(int ch) { 2905 StringBuilder sb = new StringBuilder(); 2906 sb.append(Character.toChars(ch)); 2907 while (ASCII.isLower(ch=read()) || ASCII.isUpper(ch) || 2908 ASCII.isDigit(ch)) { 2909 sb.append(Character.toChars(ch)); 2910 } 2911 if (sb.length() == 0) 2912 throw error("named capturing group has 0 length name"); 2913 if (ch != '>') 2914 throw error("named capturing group is missing trailing '>'"); 2915 return sb.toString(); 2916 } 2917 2918 /** 2919 * Parses a group and returns the head node of a set of nodes that process 2920 * the group. Sometimes a double return system is used where the tail is 2921 * returned in root. 2922 */ 2923 private Node group0() { 2924 boolean capturingGroup = false; 2925 Node head = null; 2926 Node tail = null; 2927 int save = flags; 2928 int saveTCNCount = topClosureNodes.size(); 2929 root = null; 2930 int ch = next(); 2931 if (ch == '?') { 2932 ch = skip(); 2933 switch (ch) { 2934 case ':': // (?:xxx) pure group 2935 head = createGroup(true); 2936 tail = root; 2937 head.next = expr(tail); 2938 break; 2939 case '=': // (?=xxx) and (?!xxx) lookahead 2940 case '!': 2941 head = createGroup(true); 2942 tail = root; 2943 head.next = expr(tail); 2944 if (ch == '=') { 2945 head = tail = new Pos(head); 2946 } else { 2947 head = tail = new Neg(head); 2948 } 2949 break; 2950 case '>': // (?>xxx) independent group 2951 head = createGroup(true); 2952 tail = root; 2953 head.next = expr(tail); 2954 head = tail = new Ques(head, Qtype.INDEPENDENT); 2955 break; 2956 case '<': // (?<xxx) look behind 2957 ch = read(); 2958 if (ASCII.isLower(ch) || ASCII.isUpper(ch)) { 2959 // named captured group 2960 String name = groupname(ch); 2961 if (namedGroups().containsKey(name)) 2962 throw error("Named capturing group <" + name 2963 + "> is already defined"); 2964 capturingGroup = true; 2965 head = createGroup(false); 2966 tail = root; 2967 namedGroups().put(name, capturingGroupCount-1); 2968 head.next = expr(tail); 2969 break; 2970 } 2971 int start = cursor; 2972 head = createGroup(true); 2973 tail = root; 2974 head.next = expr(tail); 2975 tail.next = lookbehindEnd; 2976 TreeInfo info = new TreeInfo(); 2977 head.study(info); 2978 if (info.maxValid == false) { 2979 throw error("Look-behind group does not have " 2980 + "an obvious maximum length"); 2981 } 2982 boolean hasSupplementary = findSupplementary(start, patternLength); 2983 if (ch == '=') { 2984 head = tail = (hasSupplementary ? 2985 new BehindS(head, info.maxLength, 2986 info.minLength) : 2987 new Behind(head, info.maxLength, 2988 info.minLength)); 2989 } else if (ch == '!') { 2990 head = tail = (hasSupplementary ? 2991 new NotBehindS(head, info.maxLength, 2992 info.minLength) : 2993 new NotBehind(head, info.maxLength, 2994 info.minLength)); 2995 } else { 2996 throw error("Unknown look-behind group"); 2997 } 2998 // clear all top-closure-nodes inside lookbehind 2999 if (saveTCNCount < topClosureNodes.size()) 3000 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3001 break; 3002 case '$': 3003 case '@': 3004 throw error("Unknown group type"); 3005 default: // (?xxx:) inlined match flags 3006 unread(); 3007 addFlag(); 3008 ch = read(); 3009 if (ch == ')') { 3010 return null; // Inline modifier only 3011 } 3012 if (ch != ':') { 3013 throw error("Unknown inline modifier"); 3014 } 3015 head = createGroup(true); 3016 tail = root; 3017 head.next = expr(tail); 3018 break; 3019 } 3020 } else { // (xxx) a regular group 3021 capturingGroup = true; 3022 head = createGroup(false); 3023 tail = root; 3024 head.next = expr(tail); 3025 } 3026 3027 accept(')', "Unclosed group"); 3028 flags = save; 3029 3030 // Check for quantifiers 3031 Node node = closure(head); 3032 if (node == head) { // No closure 3033 root = tail; 3034 return node; // Dual return 3035 } 3036 if (head == tail) { // Zero length assertion 3037 root = node; 3038 return node; // Dual return 3039 } 3040 3041 // have group closure, clear all inner closure nodes from the 3042 // top list (no backtracking stopper optimization for inner 3043 if (saveTCNCount < topClosureNodes.size()) 3044 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3045 3046 if (node instanceof Ques) { 3047 Ques ques = (Ques) node; 3048 if (ques.type == Qtype.POSSESSIVE) { 3049 root = node; 3050 return node; 3051 } 3052 tail.next = new BranchConn(); 3053 tail = tail.next; 3054 if (ques.type == Qtype.GREEDY) { 3055 head = new Branch(head, null, tail); 3056 } else { // Reluctant quantifier 3057 head = new Branch(null, head, tail); 3058 } 3059 root = tail; 3060 return head; 3061 } else if (node instanceof Curly) { 3062 Curly curly = (Curly) node; 3063 if (curly.type == Qtype.POSSESSIVE) { 3064 root = node; 3065 return node; 3066 } 3067 // Discover if the group is deterministic 3068 TreeInfo info = new TreeInfo(); 3069 if (head.study(info)) { // Deterministic 3070 GroupTail temp = (GroupTail) tail; 3071 head = root = new GroupCurly(head.next, curly.cmin, 3072 curly.cmax, curly.type, 3073 ((GroupTail)tail).localIndex, 3074 ((GroupTail)tail).groupIndex, 3075 capturingGroup); 3076 return head; 3077 } else { // Non-deterministic 3078 int temp = ((GroupHead) head).localIndex; 3079 Loop loop; 3080 if (curly.type == Qtype.GREEDY) { 3081 loop = new Loop(this.localCount, temp); 3082 // add the max_reps greedy to the top-closure-node list 3083 if (curly.cmax == MAX_REPS) 3084 topClosureNodes.add(loop); 3085 } else { // Reluctant Curly 3086 loop = new LazyLoop(this.localCount, temp); 3087 } 3088 Prolog prolog = new Prolog(loop); 3089 this.localCount += 1; 3090 loop.cmin = curly.cmin; 3091 loop.cmax = curly.cmax; 3092 loop.body = head; 3093 tail.next = loop; 3094 root = loop; 3095 return prolog; // Dual return 3096 } 3097 } 3098 throw error("Internal logic error"); 3099 } 3100 3101 /** 3102 * Create group head and tail nodes using double return. If the group is 3103 * created with anonymous true then it is a pure group and should not 3104 * affect group counting. 3105 */ 3106 private Node createGroup(boolean anonymous) { 3107 int localIndex = localCount++; 3108 int groupIndex = 0; 3109 if (!anonymous) 3110 groupIndex = capturingGroupCount++; 3111 GroupHead head = new GroupHead(localIndex); 3112 root = new GroupTail(localIndex, groupIndex); 3113 3114 // for debug/print only, head.match does NOT need the "tail" info 3115 head.tail = (GroupTail)root; 3116 3117 if (!anonymous && groupIndex < 10) 3118 groupNodes[groupIndex] = head; 3119 return head; 3120 } 3121 3122 @SuppressWarnings("fallthrough") 3123 /** 3124 * Parses inlined match flags and set them appropriately. 3125 */ 3126 private void addFlag() { 3127 int ch = peek(); 3128 for (;;) { 3129 switch (ch) { 3130 case 'i': 3131 flags |= CASE_INSENSITIVE; 3132 break; 3133 case 'm': 3134 flags |= MULTILINE; 3135 break; 3136 case 's': 3137 flags |= DOTALL; 3138 break; 3139 case 'd': 3140 flags |= UNIX_LINES; 3141 break; 3142 case 'u': 3143 flags |= UNICODE_CASE; 3144 break; 3145 case 'c': 3146 flags |= CANON_EQ; 3147 break; 3148 case 'x': 3149 flags |= COMMENTS; 3150 break; 3151 case 'U': 3152 flags |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3153 break; 3154 case '-': // subFlag then fall through 3155 ch = next(); 3156 subFlag(); 3157 default: 3158 return; 3159 } 3160 ch = next(); 3161 } 3162 } 3163 3164 @SuppressWarnings("fallthrough") 3165 /** 3166 * Parses the second part of inlined match flags and turns off 3167 * flags appropriately. 3168 */ 3169 private void subFlag() { 3170 int ch = peek(); 3171 for (;;) { 3172 switch (ch) { 3173 case 'i': 3174 flags &= ~CASE_INSENSITIVE; 3175 break; 3176 case 'm': 3177 flags &= ~MULTILINE; 3178 break; 3179 case 's': 3180 flags &= ~DOTALL; 3181 break; 3182 case 'd': 3183 flags &= ~UNIX_LINES; 3184 break; 3185 case 'u': 3186 flags &= ~UNICODE_CASE; 3187 break; 3188 case 'c': 3189 flags &= ~CANON_EQ; 3190 break; 3191 case 'x': 3192 flags &= ~COMMENTS; 3193 break; 3194 case 'U': 3195 flags &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3196 break; 3197 default: 3198 return; 3199 } 3200 ch = next(); 3201 } 3202 } 3203 3204 static final int MAX_REPS = 0x7FFFFFFF; 3205 3206 static enum Qtype { 3207 GREEDY, LAZY, POSSESSIVE, INDEPENDENT 3208 } 3209 3210 private Node curly(Node prev, int cmin) { 3211 int ch = next(); 3212 if (ch == '?') { 3213 next(); 3214 return new Curly(prev, cmin, MAX_REPS, Qtype.LAZY); 3215 } else if (ch == '+') { 3216 next(); 3217 return new Curly(prev, cmin, MAX_REPS, Qtype.POSSESSIVE); 3218 } 3219 if (prev instanceof BmpCharProperty) { 3220 return new BmpCharPropertyGreedy((BmpCharProperty)prev, cmin); 3221 } else if (prev instanceof CharProperty) { 3222 return new CharPropertyGreedy((CharProperty)prev, cmin); 3223 } 3224 return new Curly(prev, cmin, MAX_REPS, Qtype.GREEDY); 3225 } 3226 3227 /** 3228 * Processes repetition. If the next character peeked is a quantifier 3229 * then new nodes must be appended to handle the repetition. 3230 * Prev could be a single or a group, so it could be a chain of nodes. 3231 */ 3232 private Node closure(Node prev) { 3233 Node atom; 3234 int ch = peek(); 3235 switch (ch) { 3236 case '?': 3237 ch = next(); 3238 if (ch == '?') { 3239 next(); 3240 return new Ques(prev, Qtype.LAZY); 3241 } else if (ch == '+') { 3242 next(); 3243 return new Ques(prev, Qtype.POSSESSIVE); 3244 } 3245 return new Ques(prev, Qtype.GREEDY); 3246 case '*': 3247 return curly(prev, 0); 3248 case '+': 3249 return curly(prev, 1); 3250 case '{': 3251 ch = temp[cursor+1]; 3252 if (ASCII.isDigit(ch)) { 3253 skip(); 3254 int cmin = 0; 3255 do { 3256 cmin = cmin * 10 + (ch - '0'); 3257 } while (ASCII.isDigit(ch = read())); 3258 int cmax = cmin; 3259 if (ch == ',') { 3260 ch = read(); 3261 cmax = MAX_REPS; 3262 if (ch != '}') { 3263 cmax = 0; 3264 while (ASCII.isDigit(ch)) { 3265 cmax = cmax * 10 + (ch - '0'); 3266 ch = read(); 3267 } 3268 } 3269 } 3270 if (ch != '}') 3271 throw error("Unclosed counted closure"); 3272 if (((cmin) | (cmax) | (cmax - cmin)) < 0) 3273 throw error("Illegal repetition range"); 3274 Curly curly; 3275 ch = peek(); 3276 if (ch == '?') { 3277 next(); 3278 curly = new Curly(prev, cmin, cmax, Qtype.LAZY); 3279 } else if (ch == '+') { 3280 next(); 3281 curly = new Curly(prev, cmin, cmax, Qtype.POSSESSIVE); 3282 } else { 3283 curly = new Curly(prev, cmin, cmax, Qtype.GREEDY); 3284 } 3285 return curly; 3286 } else { 3287 throw error("Illegal repetition"); 3288 } 3289 default: 3290 return prev; 3291 } 3292 } 3293 3294 /** 3295 * Utility method for parsing control escape sequences. 3296 */ 3297 private int c() { 3298 if (cursor < patternLength) { 3299 return read() ^ 64; 3300 } 3301 throw error("Illegal control escape sequence"); 3302 } 3303 3304 /** 3305 * Utility method for parsing octal escape sequences. 3306 */ 3307 private int o() { 3308 int n = read(); 3309 if (((n-'0')|('7'-n)) >= 0) { 3310 int m = read(); 3311 if (((m-'0')|('7'-m)) >= 0) { 3312 int o = read(); 3313 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) { 3314 return (n - '0') * 64 + (m - '0') * 8 + (o - '0'); 3315 } 3316 unread(); 3317 return (n - '0') * 8 + (m - '0'); 3318 } 3319 unread(); 3320 return (n - '0'); 3321 } 3322 throw error("Illegal octal escape sequence"); 3323 } 3324 3325 /** 3326 * Utility method for parsing hexadecimal escape sequences. 3327 */ 3328 private int x() { 3329 int n = read(); 3330 if (ASCII.isHexDigit(n)) { 3331 int m = read(); 3332 if (ASCII.isHexDigit(m)) { 3333 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m); 3334 } 3335 } else if (n == '{' && ASCII.isHexDigit(peek())) { 3336 int ch = 0; 3337 while (ASCII.isHexDigit(n = read())) { 3338 ch = (ch << 4) + ASCII.toDigit(n); 3339 if (ch > Character.MAX_CODE_POINT) 3340 throw error("Hexadecimal codepoint is too big"); 3341 } 3342 if (n != '}') 3343 throw error("Unclosed hexadecimal escape sequence"); 3344 return ch; 3345 } 3346 throw error("Illegal hexadecimal escape sequence"); 3347 } 3348 3349 /** 3350 * Utility method for parsing unicode escape sequences. 3351 */ 3352 private int cursor() { 3353 return cursor; 3354 } 3355 3356 private void setcursor(int pos) { 3357 cursor = pos; 3358 } 3359 3360 private int uxxxx() { 3361 int n = 0; 3362 for (int i = 0; i < 4; i++) { 3363 int ch = read(); 3364 if (!ASCII.isHexDigit(ch)) { 3365 throw error("Illegal Unicode escape sequence"); 3366 } 3367 n = n * 16 + ASCII.toDigit(ch); 3368 } 3369 return n; 3370 } 3371 3372 private int u() { 3373 int n = uxxxx(); 3374 if (Character.isHighSurrogate((char)n)) { 3375 int cur = cursor(); 3376 if (read() == '\\' && read() == 'u') { 3377 int n2 = uxxxx(); 3378 if (Character.isLowSurrogate((char)n2)) 3379 return Character.toCodePoint((char)n, (char)n2); 3380 } 3381 setcursor(cur); 3382 } 3383 return n; 3384 } 3385 3386 private int N() { 3387 if (read() == '{') { 3388 int i = cursor; 3389 while (cursor < patternLength && read() != '}') {} 3390 if (cursor > patternLength) 3391 throw error("Unclosed character name escape sequence"); 3392 String name = new String(temp, i, cursor - i - 1); 3393 try { 3394 return Character.codePointOf(name); 3395 } catch (IllegalArgumentException x) { 3396 throw error("Unknown character name [" + name + "]"); 3397 } 3398 } 3399 throw error("Illegal character name escape sequence"); 3400 } 3401 3402 // 3403 // Utility methods for code point support 3404 // 3405 private static final int countChars(CharSequence seq, int index, 3406 int lengthInCodePoints) { 3407 // optimization 3408 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) { 3409 assert (index >= 0 && index < seq.length()); 3410 return 1; 3411 } 3412 int length = seq.length(); 3413 int x = index; 3414 if (lengthInCodePoints >= 0) { 3415 assert (index >= 0 && index < length); 3416 for (int i = 0; x < length && i < lengthInCodePoints; i++) { 3417 if (Character.isHighSurrogate(seq.charAt(x++))) { 3418 if (x < length && Character.isLowSurrogate(seq.charAt(x))) { 3419 x++; 3420 } 3421 } 3422 } 3423 return x - index; 3424 } 3425 3426 assert (index >= 0 && index <= length); 3427 if (index == 0) { 3428 return 0; 3429 } 3430 int len = -lengthInCodePoints; 3431 for (int i = 0; x > 0 && i < len; i++) { 3432 if (Character.isLowSurrogate(seq.charAt(--x))) { 3433 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) { 3434 x--; 3435 } 3436 } 3437 } 3438 return index - x; 3439 } 3440 3441 private static final int countCodePoints(CharSequence seq) { 3442 int length = seq.length(); 3443 int n = 0; 3444 for (int i = 0; i < length; ) { 3445 n++; 3446 if (Character.isHighSurrogate(seq.charAt(i++))) { 3447 if (i < length && Character.isLowSurrogate(seq.charAt(i))) { 3448 i++; 3449 } 3450 } 3451 } 3452 return n; 3453 } 3454 3455 /** 3456 * Creates a bit vector for matching Latin-1 values. A normal BitClass 3457 * never matches values above Latin-1, and a complemented BitClass always 3458 * matches values above Latin-1. 3459 */ 3460 static final class BitClass extends BmpCharProperty { 3461 final boolean[] bits; 3462 BitClass() { 3463 this(new boolean[256]); 3464 } 3465 private BitClass(boolean[] bits) { 3466 super( ch -> ch < 256 && bits[ch]); 3467 this.bits = bits; 3468 } 3469 BitClass add(int c, int flags) { 3470 assert c >= 0 && c <= 255; 3471 if ((flags & CASE_INSENSITIVE) != 0) { 3472 if (ASCII.isAscii(c)) { 3473 bits[ASCII.toUpper(c)] = true; 3474 bits[ASCII.toLower(c)] = true; 3475 } else if ((flags & UNICODE_CASE) != 0) { 3476 bits[Character.toLowerCase(c)] = true; 3477 bits[Character.toUpperCase(c)] = true; 3478 } 3479 } 3480 bits[c] = true; 3481 return this; 3482 } 3483 } 3484 3485 /** 3486 * Utility method for creating a string slice matcher. 3487 */ 3488 private Node newSlice(int[] buf, int count, boolean hasSupplementary) { 3489 int[] tmp = new int[count]; 3490 if (has(CASE_INSENSITIVE)) { 3491 if (has(UNICODE_CASE)) { 3492 for (int i = 0; i < count; i++) { 3493 tmp[i] = Character.toLowerCase( 3494 Character.toUpperCase(buf[i])); 3495 } 3496 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp); 3497 } 3498 for (int i = 0; i < count; i++) { 3499 tmp[i] = ASCII.toLower(buf[i]); 3500 } 3501 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp); 3502 } 3503 for (int i = 0; i < count; i++) { 3504 tmp[i] = buf[i]; 3505 } 3506 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp); 3507 } 3508 3509 /** 3510 * The following classes are the building components of the object 3511 * tree that represents a compiled regular expression. The object tree 3512 * is made of individual elements that handle constructs in the Pattern. 3513 * Each type of object knows how to match its equivalent construct with 3514 * the match() method. 3515 */ 3516 3517 /** 3518 * Base class for all node classes. Subclasses should override the match() 3519 * method as appropriate. This class is an accepting node, so its match() 3520 * always returns true. 3521 */ 3522 static class Node extends Object { 3523 Node next; 3524 Node() { 3525 next = Pattern.accept; 3526 } 3527 /** 3528 * This method implements the classic accept node. 3529 */ 3530 boolean match(Matcher matcher, int i, CharSequence seq) { 3531 matcher.last = i; 3532 matcher.groups[0] = matcher.first; 3533 matcher.groups[1] = matcher.last; 3534 return true; 3535 } 3536 /** 3537 * This method is good for all zero length assertions. 3538 */ 3539 boolean study(TreeInfo info) { 3540 if (next != null) { 3541 return next.study(info); 3542 } else { 3543 return info.deterministic; 3544 } 3545 } 3546 } 3547 3548 static class LastNode extends Node { 3549 /** 3550 * This method implements the classic accept node with 3551 * the addition of a check to see if the match occurred 3552 * using all of the input. 3553 */ 3554 boolean match(Matcher matcher, int i, CharSequence seq) { 3555 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to) 3556 return false; 3557 matcher.last = i; 3558 matcher.groups[0] = matcher.first; 3559 matcher.groups[1] = matcher.last; 3560 return true; 3561 } 3562 } 3563 3564 /** 3565 * Used for REs that can start anywhere within the input string. 3566 * This basically tries to match repeatedly at each spot in the 3567 * input string, moving forward after each try. An anchored search 3568 * or a BnM will bypass this node completely. 3569 */ 3570 static class Start extends Node { 3571 int minLength; 3572 Start(Node node) { 3573 this.next = node; 3574 TreeInfo info = new TreeInfo(); 3575 next.study(info); 3576 minLength = info.minLength; 3577 } 3578 boolean match(Matcher matcher, int i, CharSequence seq) { 3579 if (i > matcher.to - minLength) { 3580 matcher.hitEnd = true; 3581 return false; 3582 } 3583 int guard = matcher.to - minLength; 3584 for (; i <= guard; i++) { 3585 if (next.match(matcher, i, seq)) { 3586 matcher.first = i; 3587 matcher.groups[0] = matcher.first; 3588 matcher.groups[1] = matcher.last; 3589 return true; 3590 } 3591 } 3592 matcher.hitEnd = true; 3593 return false; 3594 } 3595 boolean study(TreeInfo info) { 3596 next.study(info); 3597 info.maxValid = false; 3598 info.deterministic = false; 3599 return false; 3600 } 3601 } 3602 3603 /* 3604 * StartS supports supplementary characters, including unpaired surrogates. 3605 */ 3606 static final class StartS extends Start { 3607 StartS(Node node) { 3608 super(node); 3609 } 3610 boolean match(Matcher matcher, int i, CharSequence seq) { 3611 if (i > matcher.to - minLength) { 3612 matcher.hitEnd = true; 3613 return false; 3614 } 3615 int guard = matcher.to - minLength; 3616 while (i <= guard) { 3617 //if ((ret = next.match(matcher, i, seq)) || i == guard) 3618 if (next.match(matcher, i, seq)) { 3619 matcher.first = i; 3620 matcher.groups[0] = matcher.first; 3621 matcher.groups[1] = matcher.last; 3622 return true; 3623 } 3624 if (i == guard) 3625 break; 3626 // Optimization to move to the next character. This is 3627 // faster than countChars(seq, i, 1). 3628 if (Character.isHighSurrogate(seq.charAt(i++))) { 3629 if (i < seq.length() && 3630 Character.isLowSurrogate(seq.charAt(i))) { 3631 i++; 3632 } 3633 } 3634 } 3635 matcher.hitEnd = true; 3636 return false; 3637 } 3638 } 3639 3640 /** 3641 * Node to anchor at the beginning of input. This object implements the 3642 * match for a \A sequence, and the caret anchor will use this if not in 3643 * multiline mode. 3644 */ 3645 static final class Begin extends Node { 3646 boolean match(Matcher matcher, int i, CharSequence seq) { 3647 int fromIndex = (matcher.anchoringBounds) ? 3648 matcher.from : 0; 3649 if (i == fromIndex && next.match(matcher, i, seq)) { 3650 matcher.first = i; 3651 matcher.groups[0] = i; 3652 matcher.groups[1] = matcher.last; 3653 return true; 3654 } else { 3655 return false; 3656 } 3657 } 3658 } 3659 3660 /** 3661 * Node to anchor at the end of input. This is the absolute end, so this 3662 * should not match at the last newline before the end as $ will. 3663 */ 3664 static final class End extends Node { 3665 boolean match(Matcher matcher, int i, CharSequence seq) { 3666 int endIndex = (matcher.anchoringBounds) ? 3667 matcher.to : matcher.getTextLength(); 3668 if (i == endIndex) { 3669 matcher.hitEnd = true; 3670 return next.match(matcher, i, seq); 3671 } 3672 return false; 3673 } 3674 } 3675 3676 /** 3677 * Node to anchor at the beginning of a line. This is essentially the 3678 * object to match for the multiline ^. 3679 */ 3680 static final class Caret extends Node { 3681 boolean match(Matcher matcher, int i, CharSequence seq) { 3682 int startIndex = matcher.from; 3683 int endIndex = matcher.to; 3684 if (!matcher.anchoringBounds) { 3685 startIndex = 0; 3686 endIndex = matcher.getTextLength(); 3687 } 3688 // Perl does not match ^ at end of input even after newline 3689 if (i == endIndex) { 3690 matcher.hitEnd = true; 3691 return false; 3692 } 3693 if (i > startIndex) { 3694 char ch = seq.charAt(i-1); 3695 if (ch != '\n' && ch != '\r' 3696 && (ch|1) != '\u2029' 3697 && ch != '\u0085' ) { 3698 return false; 3699 } 3700 // Should treat /r/n as one newline 3701 if (ch == '\r' && seq.charAt(i) == '\n') 3702 return false; 3703 } 3704 return next.match(matcher, i, seq); 3705 } 3706 } 3707 3708 /** 3709 * Node to anchor at the beginning of a line when in unixdot mode. 3710 */ 3711 static final class UnixCaret extends Node { 3712 boolean match(Matcher matcher, int i, CharSequence seq) { 3713 int startIndex = matcher.from; 3714 int endIndex = matcher.to; 3715 if (!matcher.anchoringBounds) { 3716 startIndex = 0; 3717 endIndex = matcher.getTextLength(); 3718 } 3719 // Perl does not match ^ at end of input even after newline 3720 if (i == endIndex) { 3721 matcher.hitEnd = true; 3722 return false; 3723 } 3724 if (i > startIndex) { 3725 char ch = seq.charAt(i-1); 3726 if (ch != '\n') { 3727 return false; 3728 } 3729 } 3730 return next.match(matcher, i, seq); 3731 } 3732 } 3733 3734 /** 3735 * Node to match the location where the last match ended. 3736 * This is used for the \G construct. 3737 */ 3738 static final class LastMatch extends Node { 3739 boolean match(Matcher matcher, int i, CharSequence seq) { 3740 if (i != matcher.oldLast) 3741 return false; 3742 return next.match(matcher, i, seq); 3743 } 3744 } 3745 3746 /** 3747 * Node to anchor at the end of a line or the end of input based on the 3748 * multiline mode. 3749 * 3750 * When not in multiline mode, the $ can only match at the very end 3751 * of the input, unless the input ends in a line terminator in which 3752 * it matches right before the last line terminator. 3753 * 3754 * Note that \r\n is considered an atomic line terminator. 3755 * 3756 * Like ^ the $ operator matches at a position, it does not match the 3757 * line terminators themselves. 3758 */ 3759 static final class Dollar extends Node { 3760 boolean multiline; 3761 Dollar(boolean mul) { 3762 multiline = mul; 3763 } 3764 boolean match(Matcher matcher, int i, CharSequence seq) { 3765 int endIndex = (matcher.anchoringBounds) ? 3766 matcher.to : matcher.getTextLength(); 3767 if (!multiline) { 3768 if (i < endIndex - 2) 3769 return false; 3770 if (i == endIndex - 2) { 3771 char ch = seq.charAt(i); 3772 if (ch != '\r') 3773 return false; 3774 ch = seq.charAt(i + 1); 3775 if (ch != '\n') 3776 return false; 3777 } 3778 } 3779 // Matches before any line terminator; also matches at the 3780 // end of input 3781 // Before line terminator: 3782 // If multiline, we match here no matter what 3783 // If not multiline, fall through so that the end 3784 // is marked as hit; this must be a /r/n or a /n 3785 // at the very end so the end was hit; more input 3786 // could make this not match here 3787 if (i < endIndex) { 3788 char ch = seq.charAt(i); 3789 if (ch == '\n') { 3790 // No match between \r\n 3791 if (i > 0 && seq.charAt(i-1) == '\r') 3792 return false; 3793 if (multiline) 3794 return next.match(matcher, i, seq); 3795 } else if (ch == '\r' || ch == '\u0085' || 3796 (ch|1) == '\u2029') { 3797 if (multiline) 3798 return next.match(matcher, i, seq); 3799 } else { // No line terminator, no match 3800 return false; 3801 } 3802 } 3803 // Matched at current end so hit end 3804 matcher.hitEnd = true; 3805 // If a $ matches because of end of input, then more input 3806 // could cause it to fail! 3807 matcher.requireEnd = true; 3808 return next.match(matcher, i, seq); 3809 } 3810 boolean study(TreeInfo info) { 3811 next.study(info); 3812 return info.deterministic; 3813 } 3814 } 3815 3816 /** 3817 * Node to anchor at the end of a line or the end of input based on the 3818 * multiline mode when in unix lines mode. 3819 */ 3820 static final class UnixDollar extends Node { 3821 boolean multiline; 3822 UnixDollar(boolean mul) { 3823 multiline = mul; 3824 } 3825 boolean match(Matcher matcher, int i, CharSequence seq) { 3826 int endIndex = (matcher.anchoringBounds) ? 3827 matcher.to : matcher.getTextLength(); 3828 if (i < endIndex) { 3829 char ch = seq.charAt(i); 3830 if (ch == '\n') { 3831 // If not multiline, then only possible to 3832 // match at very end or one before end 3833 if (multiline == false && i != endIndex - 1) 3834 return false; 3835 // If multiline return next.match without setting 3836 // matcher.hitEnd 3837 if (multiline) 3838 return next.match(matcher, i, seq); 3839 } else { 3840 return false; 3841 } 3842 } 3843 // Matching because at the end or 1 before the end; 3844 // more input could change this so set hitEnd 3845 matcher.hitEnd = true; 3846 // If a $ matches because of end of input, then more input 3847 // could cause it to fail! 3848 matcher.requireEnd = true; 3849 return next.match(matcher, i, seq); 3850 } 3851 boolean study(TreeInfo info) { 3852 next.study(info); 3853 return info.deterministic; 3854 } 3855 } 3856 3857 /** 3858 * Node class that matches a Unicode line ending '\R' 3859 */ 3860 static final class LineEnding extends Node { 3861 boolean match(Matcher matcher, int i, CharSequence seq) { 3862 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029]) 3863 if (i < matcher.to) { 3864 int ch = seq.charAt(i); 3865 if (ch == 0x0A || ch == 0x0B || ch == 0x0C || 3866 ch == 0x85 || ch == 0x2028 || ch == 0x2029) 3867 return next.match(matcher, i + 1, seq); 3868 if (ch == 0x0D) { 3869 i++; 3870 if (i < matcher.to && seq.charAt(i) == 0x0A) 3871 i++; 3872 return next.match(matcher, i, seq); 3873 } 3874 } else { 3875 matcher.hitEnd = true; 3876 } 3877 return false; 3878 } 3879 boolean study(TreeInfo info) { 3880 info.minLength++; 3881 info.maxLength += 2; 3882 return next.study(info); 3883 } 3884 } 3885 3886 /** 3887 * Abstract node class to match one character satisfying some 3888 * boolean property. 3889 */ 3890 static class CharProperty extends Node { 3891 CharPredicate predicate; 3892 3893 CharProperty (CharPredicate predicate) { 3894 this.predicate = predicate; 3895 } 3896 boolean match(Matcher matcher, int i, CharSequence seq) { 3897 if (i < matcher.to) { 3898 int ch = Character.codePointAt(seq, i); 3899 return predicate.is(ch) && 3900 next.match(matcher, i + Character.charCount(ch), seq); 3901 } else { 3902 matcher.hitEnd = true; 3903 return false; 3904 } 3905 } 3906 boolean study(TreeInfo info) { 3907 info.minLength++; 3908 info.maxLength++; 3909 return next.study(info); 3910 } 3911 } 3912 3913 /** 3914 * Optimized version of CharProperty that works only for 3915 * properties never satisfied by Supplementary characters. 3916 */ 3917 private static class BmpCharProperty extends CharProperty { 3918 BmpCharProperty (BmpCharPredicate predicate) { 3919 super(predicate); 3920 } 3921 boolean match(Matcher matcher, int i, CharSequence seq) { 3922 if (i < matcher.to) { 3923 return predicate.is(seq.charAt(i)) && 3924 next.match(matcher, i + 1, seq); 3925 } else { 3926 matcher.hitEnd = true; 3927 return false; 3928 } 3929 } 3930 } 3931 3932 private static class NFCCharProperty extends Node { 3933 CharPredicate predicate; 3934 NFCCharProperty (CharPredicate predicate) { 3935 this.predicate = predicate; 3936 } 3937 3938 boolean match(Matcher matcher, int i, CharSequence seq) { 3939 if (i < matcher.to) { 3940 int ch0 = Character.codePointAt(seq, i); 3941 int n = Character.charCount(ch0); 3942 int j = i + n; 3943 while (j < matcher.to) { 3944 int ch1 = Character.codePointAt(seq, j); 3945 if (Grapheme.isBoundary(ch0, ch1)) 3946 break; 3947 ch0 = ch1; 3948 j += Character.charCount(ch1); 3949 } 3950 if (i + n == j) { // single, assume nfc cp 3951 if (predicate.is(ch0)) 3952 return next.match(matcher, j, seq); 3953 } else { 3954 while (i + n < j) { 3955 String nfc = Normalizer.normalize( 3956 seq.toString().substring(i, j), Normalizer.Form.NFC); 3957 if (nfc.codePointCount(0, nfc.length()) == 1) { 3958 if (predicate.is(nfc.codePointAt(0)) && 3959 next.match(matcher, j, seq)) { 3960 return true; 3961 } 3962 } 3963 3964 ch0 = Character.codePointBefore(seq, j); 3965 j -= Character.charCount(ch0); 3966 } 3967 } 3968 if (j < matcher.to) 3969 return false; 3970 } 3971 matcher.hitEnd = true; 3972 return false; 3973 } 3974 3975 boolean study(TreeInfo info) { 3976 info.minLength++; 3977 info.deterministic = false; 3978 return next.study(info); 3979 } 3980 } 3981 3982 /** 3983 * Node class that matches an unicode extended grapheme cluster 3984 */ 3985 static class XGrapheme extends Node { 3986 boolean match(Matcher matcher, int i, CharSequence seq) { 3987 if (i < matcher.to) { 3988 int ch0 = Character.codePointAt(seq, i); 3989 i += Character.charCount(ch0); 3990 while (i < matcher.to) { 3991 int ch1 = Character.codePointAt(seq, i); 3992 if (Grapheme.isBoundary(ch0, ch1)) 3993 break; 3994 ch0 = ch1; 3995 i += Character.charCount(ch1); 3996 } 3997 return next.match(matcher, i, seq); 3998 } 3999 matcher.hitEnd = true; 4000 return false; 4001 } 4002 4003 boolean study(TreeInfo info) { 4004 info.minLength++; 4005 info.deterministic = false; 4006 return next.study(info); 4007 } 4008 } 4009 4010 /** 4011 * Node class that handles grapheme boundaries 4012 */ 4013 static class GraphemeBound extends Node { 4014 boolean match(Matcher matcher, int i, CharSequence seq) { 4015 int startIndex = matcher.from; 4016 int endIndex = matcher.to; 4017 if (matcher.transparentBounds) { 4018 startIndex = 0; 4019 endIndex = matcher.getTextLength(); 4020 } 4021 if (i == startIndex) { 4022 return next.match(matcher, i, seq); 4023 } 4024 if (i < endIndex) { 4025 if (Character.isSurrogatePair(seq.charAt(i-1), seq.charAt(i)) || 4026 !Grapheme.isBoundary(Character.codePointBefore(seq, i), 4027 Character.codePointAt(seq, i))) { 4028 return false; 4029 } 4030 } else { 4031 matcher.hitEnd = true; 4032 matcher.requireEnd = true; 4033 } 4034 return next.match(matcher, i, seq); 4035 } 4036 } 4037 4038 /** 4039 * Base class for all Slice nodes 4040 */ 4041 static class SliceNode extends Node { 4042 int[] buffer; 4043 SliceNode(int[] buf) { 4044 buffer = buf; 4045 } 4046 boolean study(TreeInfo info) { 4047 info.minLength += buffer.length; 4048 info.maxLength += buffer.length; 4049 return next.study(info); 4050 } 4051 } 4052 4053 /** 4054 * Node class for a case sensitive/BMP-only sequence of literal 4055 * characters. 4056 */ 4057 static class Slice extends SliceNode { 4058 Slice(int[] buf) { 4059 super(buf); 4060 } 4061 boolean match(Matcher matcher, int i, CharSequence seq) { 4062 int[] buf = buffer; 4063 int len = buf.length; 4064 for (int j=0; j<len; j++) { 4065 if ((i+j) >= matcher.to) { 4066 matcher.hitEnd = true; 4067 return false; 4068 } 4069 if (buf[j] != seq.charAt(i+j)) 4070 return false; 4071 } 4072 return next.match(matcher, i+len, seq); 4073 } 4074 } 4075 4076 /** 4077 * Node class for a case_insensitive/BMP-only sequence of literal 4078 * characters. 4079 */ 4080 static class SliceI extends SliceNode { 4081 SliceI(int[] buf) { 4082 super(buf); 4083 } 4084 boolean match(Matcher matcher, int i, CharSequence seq) { 4085 int[] buf = buffer; 4086 int len = buf.length; 4087 for (int j=0; j<len; j++) { 4088 if ((i+j) >= matcher.to) { 4089 matcher.hitEnd = true; 4090 return false; 4091 } 4092 int c = seq.charAt(i+j); 4093 if (buf[j] != c && 4094 buf[j] != ASCII.toLower(c)) 4095 return false; 4096 } 4097 return next.match(matcher, i+len, seq); 4098 } 4099 } 4100 4101 /** 4102 * Node class for a unicode_case_insensitive/BMP-only sequence of 4103 * literal characters. Uses unicode case folding. 4104 */ 4105 static final class SliceU extends SliceNode { 4106 SliceU(int[] buf) { 4107 super(buf); 4108 } 4109 boolean match(Matcher matcher, int i, CharSequence seq) { 4110 int[] buf = buffer; 4111 int len = buf.length; 4112 for (int j=0; j<len; j++) { 4113 if ((i+j) >= matcher.to) { 4114 matcher.hitEnd = true; 4115 return false; 4116 } 4117 int c = seq.charAt(i+j); 4118 if (buf[j] != c && 4119 buf[j] != Character.toLowerCase(Character.toUpperCase(c))) 4120 return false; 4121 } 4122 return next.match(matcher, i+len, seq); 4123 } 4124 } 4125 4126 /** 4127 * Node class for a case sensitive sequence of literal characters 4128 * including supplementary characters. 4129 */ 4130 static final class SliceS extends Slice { 4131 SliceS(int[] buf) { 4132 super(buf); 4133 } 4134 boolean match(Matcher matcher, int i, CharSequence seq) { 4135 int[] buf = buffer; 4136 int x = i; 4137 for (int j = 0; j < buf.length; j++) { 4138 if (x >= matcher.to) { 4139 matcher.hitEnd = true; 4140 return false; 4141 } 4142 int c = Character.codePointAt(seq, x); 4143 if (buf[j] != c) 4144 return false; 4145 x += Character.charCount(c); 4146 if (x > matcher.to) { 4147 matcher.hitEnd = true; 4148 return false; 4149 } 4150 } 4151 return next.match(matcher, x, seq); 4152 } 4153 } 4154 4155 /** 4156 * Node class for a case insensitive sequence of literal characters 4157 * including supplementary characters. 4158 */ 4159 static class SliceIS extends SliceNode { 4160 SliceIS(int[] buf) { 4161 super(buf); 4162 } 4163 int toLower(int c) { 4164 return ASCII.toLower(c); 4165 } 4166 boolean match(Matcher matcher, int i, CharSequence seq) { 4167 int[] buf = buffer; 4168 int x = i; 4169 for (int j = 0; j < buf.length; j++) { 4170 if (x >= matcher.to) { 4171 matcher.hitEnd = true; 4172 return false; 4173 } 4174 int c = Character.codePointAt(seq, x); 4175 if (buf[j] != c && buf[j] != toLower(c)) 4176 return false; 4177 x += Character.charCount(c); 4178 if (x > matcher.to) { 4179 matcher.hitEnd = true; 4180 return false; 4181 } 4182 } 4183 return next.match(matcher, x, seq); 4184 } 4185 } 4186 4187 /** 4188 * Node class for a case insensitive sequence of literal characters. 4189 * Uses unicode case folding. 4190 */ 4191 static final class SliceUS extends SliceIS { 4192 SliceUS(int[] buf) { 4193 super(buf); 4194 } 4195 int toLower(int c) { 4196 return Character.toLowerCase(Character.toUpperCase(c)); 4197 } 4198 } 4199 4200 /** 4201 * The 0 or 1 quantifier. This one class implements all three types. 4202 */ 4203 static final class Ques extends Node { 4204 Node atom; 4205 Qtype type; 4206 Ques(Node node, Qtype type) { 4207 this.atom = node; 4208 this.type = type; 4209 } 4210 boolean match(Matcher matcher, int i, CharSequence seq) { 4211 switch (type) { 4212 case GREEDY: 4213 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)) 4214 || next.match(matcher, i, seq); 4215 case LAZY: 4216 return next.match(matcher, i, seq) 4217 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)); 4218 case POSSESSIVE: 4219 if (atom.match(matcher, i, seq)) i = matcher.last; 4220 return next.match(matcher, i, seq); 4221 default: 4222 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq); 4223 } 4224 } 4225 boolean study(TreeInfo info) { 4226 if (type != Qtype.INDEPENDENT) { 4227 int minL = info.minLength; 4228 atom.study(info); 4229 info.minLength = minL; 4230 info.deterministic = false; 4231 return next.study(info); 4232 } else { 4233 atom.study(info); 4234 return next.study(info); 4235 } 4236 } 4237 } 4238 4239 /** 4240 * Handles the greedy style repetition with the minimum either be 4241 * 0 or 1 and the maximum be MAX_REPS, for * and + quantifier. 4242 */ 4243 static class CharPropertyGreedy extends Node { 4244 final CharPredicate predicate; 4245 final int cmin; 4246 4247 CharPropertyGreedy(CharProperty cp, int cmin) { 4248 this.predicate = cp.predicate; 4249 this.cmin = cmin; 4250 } 4251 boolean match(Matcher matcher, int i, CharSequence seq) { 4252 int n = 0; 4253 int to = matcher.to; 4254 // greedy, all the way down 4255 while (i < to) { 4256 int ch = Character.codePointAt(seq, i); 4257 if (!predicate.is(ch)) 4258 break; 4259 i += Character.charCount(ch); 4260 n++; 4261 } 4262 if (i >= to) { 4263 matcher.hitEnd = true; 4264 } 4265 while (n >= cmin) { 4266 if (next.match(matcher, i, seq)) 4267 return true; 4268 if (n == cmin) 4269 return false; 4270 // backing off if match fails 4271 int ch = Character.codePointBefore(seq, i); 4272 i -= Character.charCount(ch); 4273 n--; 4274 } 4275 return false; 4276 } 4277 4278 boolean study(TreeInfo info) { 4279 info.minLength += cmin; 4280 if (info.maxValid) { 4281 info.maxLength += MAX_REPS; 4282 } 4283 info.deterministic = false; 4284 return next.study(info); 4285 } 4286 } 4287 4288 static final class BmpCharPropertyGreedy extends CharPropertyGreedy { 4289 4290 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) { 4291 super(bcp, cmin); 4292 } 4293 4294 boolean match(Matcher matcher, int i, CharSequence seq) { 4295 int n = 0; 4296 int to = matcher.to; 4297 while (i < to && predicate.is(seq.charAt(i))) { 4298 i++; n++; 4299 } 4300 if (i >= to) { 4301 matcher.hitEnd = true; 4302 } 4303 while (n >= cmin) { 4304 if (next.match(matcher, i, seq)) 4305 return true; 4306 i--; n--; // backing off if match fails 4307 } 4308 return false; 4309 } 4310 } 4311 4312 /** 4313 * Handles the curly-brace style repetition with a specified minimum and 4314 * maximum occurrences. The * quantifier is handled as a special case. 4315 * This class handles the three types. 4316 */ 4317 static final class Curly extends Node { 4318 Node atom; 4319 Qtype type; 4320 int cmin; 4321 int cmax; 4322 4323 Curly(Node node, int cmin, int cmax, Qtype type) { 4324 this.atom = node; 4325 this.type = type; 4326 this.cmin = cmin; 4327 this.cmax = cmax; 4328 } 4329 boolean match(Matcher matcher, int i, CharSequence seq) { 4330 int j; 4331 for (j = 0; j < cmin; j++) { 4332 if (atom.match(matcher, i, seq)) { 4333 i = matcher.last; 4334 continue; 4335 } 4336 return false; 4337 } 4338 if (type == Qtype.GREEDY) 4339 return match0(matcher, i, j, seq); 4340 else if (type == Qtype.LAZY) 4341 return match1(matcher, i, j, seq); 4342 else 4343 return match2(matcher, i, j, seq); 4344 } 4345 // Greedy match. 4346 // i is the index to start matching at 4347 // j is the number of atoms that have matched 4348 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4349 if (j >= cmax) { 4350 // We have matched the maximum... continue with the rest of 4351 // the regular expression 4352 return next.match(matcher, i, seq); 4353 } 4354 int backLimit = j; 4355 while (atom.match(matcher, i, seq)) { 4356 // k is the length of this match 4357 int k = matcher.last - i; 4358 if (k == 0) // Zero length match 4359 break; 4360 // Move up index and number matched 4361 i = matcher.last; 4362 j++; 4363 // We are greedy so match as many as we can 4364 while (j < cmax) { 4365 if (!atom.match(matcher, i, seq)) 4366 break; 4367 if (i + k != matcher.last) { 4368 if (match0(matcher, matcher.last, j+1, seq)) 4369 return true; 4370 break; 4371 } 4372 i += k; 4373 j++; 4374 } 4375 // Handle backing off if match fails 4376 while (j >= backLimit) { 4377 if (next.match(matcher, i, seq)) 4378 return true; 4379 i -= k; 4380 j--; 4381 } 4382 return false; 4383 } 4384 return next.match(matcher, i, seq); 4385 } 4386 // Reluctant match. At this point, the minimum has been satisfied. 4387 // i is the index to start matching at 4388 // j is the number of atoms that have matched 4389 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4390 for (;;) { 4391 // Try finishing match without consuming any more 4392 if (next.match(matcher, i, seq)) 4393 return true; 4394 // At the maximum, no match found 4395 if (j >= cmax) 4396 return false; 4397 // Okay, must try one more atom 4398 if (!atom.match(matcher, i, seq)) 4399 return false; 4400 // If we haven't moved forward then must break out 4401 if (i == matcher.last) 4402 return false; 4403 // Move up index and number matched 4404 i = matcher.last; 4405 j++; 4406 } 4407 } 4408 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4409 for (; j < cmax; j++) { 4410 if (!atom.match(matcher, i, seq)) 4411 break; 4412 if (i == matcher.last) 4413 break; 4414 i = matcher.last; 4415 } 4416 return next.match(matcher, i, seq); 4417 } 4418 boolean study(TreeInfo info) { 4419 // Save original info 4420 int minL = info.minLength; 4421 int maxL = info.maxLength; 4422 boolean maxV = info.maxValid; 4423 boolean detm = info.deterministic; 4424 info.reset(); 4425 4426 atom.study(info); 4427 4428 int temp = info.minLength * cmin + minL; 4429 if (temp < minL) { 4430 temp = 0xFFFFFFF; // arbitrary large number 4431 } 4432 info.minLength = temp; 4433 4434 if (maxV & info.maxValid) { 4435 temp = info.maxLength * cmax + maxL; 4436 info.maxLength = temp; 4437 if (temp < maxL) { 4438 info.maxValid = false; 4439 } 4440 } else { 4441 info.maxValid = false; 4442 } 4443 4444 if (info.deterministic && cmin == cmax) 4445 info.deterministic = detm; 4446 else 4447 info.deterministic = false; 4448 return next.study(info); 4449 } 4450 } 4451 4452 /** 4453 * Handles the curly-brace style repetition with a specified minimum and 4454 * maximum occurrences in deterministic cases. This is an iterative 4455 * optimization over the Prolog and Loop system which would handle this 4456 * in a recursive way. The * quantifier is handled as a special case. 4457 * If capture is true then this class saves group settings and ensures 4458 * that groups are unset when backing off of a group match. 4459 */ 4460 static final class GroupCurly extends Node { 4461 Node atom; 4462 Qtype type; 4463 int cmin; 4464 int cmax; 4465 int localIndex; 4466 int groupIndex; 4467 boolean capture; 4468 4469 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local, 4470 int group, boolean capture) { 4471 this.atom = node; 4472 this.type = type; 4473 this.cmin = cmin; 4474 this.cmax = cmax; 4475 this.localIndex = local; 4476 this.groupIndex = group; 4477 this.capture = capture; 4478 } 4479 boolean match(Matcher matcher, int i, CharSequence seq) { 4480 int[] groups = matcher.groups; 4481 int[] locals = matcher.locals; 4482 int save0 = locals[localIndex]; 4483 int save1 = 0; 4484 int save2 = 0; 4485 4486 if (capture) { 4487 save1 = groups[groupIndex]; 4488 save2 = groups[groupIndex+1]; 4489 } 4490 4491 // Notify GroupTail there is no need to setup group info 4492 // because it will be set here 4493 locals[localIndex] = -1; 4494 4495 boolean ret = true; 4496 for (int j = 0; j < cmin; j++) { 4497 if (atom.match(matcher, i, seq)) { 4498 if (capture) { 4499 groups[groupIndex] = i; 4500 groups[groupIndex+1] = matcher.last; 4501 } 4502 i = matcher.last; 4503 } else { 4504 ret = false; 4505 break; 4506 } 4507 } 4508 if (ret) { 4509 if (type == Qtype.GREEDY) { 4510 ret = match0(matcher, i, cmin, seq); 4511 } else if (type == Qtype.LAZY) { 4512 ret = match1(matcher, i, cmin, seq); 4513 } else { 4514 ret = match2(matcher, i, cmin, seq); 4515 } 4516 } 4517 if (!ret) { 4518 locals[localIndex] = save0; 4519 if (capture) { 4520 groups[groupIndex] = save1; 4521 groups[groupIndex+1] = save2; 4522 } 4523 } 4524 return ret; 4525 } 4526 // Aggressive group match 4527 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4528 // don't back off passing the starting "j" 4529 int min = j; 4530 int[] groups = matcher.groups; 4531 int save0 = 0; 4532 int save1 = 0; 4533 if (capture) { 4534 save0 = groups[groupIndex]; 4535 save1 = groups[groupIndex+1]; 4536 } 4537 for (;;) { 4538 if (j >= cmax) 4539 break; 4540 if (!atom.match(matcher, i, seq)) 4541 break; 4542 int k = matcher.last - i; 4543 if (k <= 0) { 4544 if (capture) { 4545 groups[groupIndex] = i; 4546 groups[groupIndex+1] = i + k; 4547 } 4548 i = i + k; 4549 break; 4550 } 4551 for (;;) { 4552 if (capture) { 4553 groups[groupIndex] = i; 4554 groups[groupIndex+1] = i + k; 4555 } 4556 i = i + k; 4557 if (++j >= cmax) 4558 break; 4559 if (!atom.match(matcher, i, seq)) 4560 break; 4561 if (i + k != matcher.last) { 4562 if (match0(matcher, i, j, seq)) 4563 return true; 4564 break; 4565 } 4566 } 4567 while (j > min) { 4568 if (next.match(matcher, i, seq)) { 4569 if (capture) { 4570 groups[groupIndex+1] = i; 4571 groups[groupIndex] = i - k; 4572 } 4573 return true; 4574 } 4575 // backing off 4576 i = i - k; 4577 if (capture) { 4578 groups[groupIndex+1] = i; 4579 groups[groupIndex] = i - k; 4580 } 4581 j--; 4582 4583 } 4584 break; 4585 } 4586 if (capture) { 4587 groups[groupIndex] = save0; 4588 groups[groupIndex+1] = save1; 4589 } 4590 return next.match(matcher, i, seq); 4591 } 4592 // Reluctant matching 4593 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4594 for (;;) { 4595 if (next.match(matcher, i, seq)) 4596 return true; 4597 if (j >= cmax) 4598 return false; 4599 if (!atom.match(matcher, i, seq)) 4600 return false; 4601 if (i == matcher.last) 4602 return false; 4603 if (capture) { 4604 matcher.groups[groupIndex] = i; 4605 matcher.groups[groupIndex+1] = matcher.last; 4606 } 4607 i = matcher.last; 4608 j++; 4609 } 4610 } 4611 // Possessive matching 4612 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4613 for (; j < cmax; j++) { 4614 if (!atom.match(matcher, i, seq)) { 4615 break; 4616 } 4617 if (capture) { 4618 matcher.groups[groupIndex] = i; 4619 matcher.groups[groupIndex+1] = matcher.last; 4620 } 4621 if (i == matcher.last) { 4622 break; 4623 } 4624 i = matcher.last; 4625 } 4626 return next.match(matcher, i, seq); 4627 } 4628 boolean study(TreeInfo info) { 4629 // Save original info 4630 int minL = info.minLength; 4631 int maxL = info.maxLength; 4632 boolean maxV = info.maxValid; 4633 boolean detm = info.deterministic; 4634 info.reset(); 4635 4636 atom.study(info); 4637 4638 int temp = info.minLength * cmin + minL; 4639 if (temp < minL) { 4640 temp = 0xFFFFFFF; // Arbitrary large number 4641 } 4642 info.minLength = temp; 4643 4644 if (maxV & info.maxValid) { 4645 temp = info.maxLength * cmax + maxL; 4646 info.maxLength = temp; 4647 if (temp < maxL) { 4648 info.maxValid = false; 4649 } 4650 } else { 4651 info.maxValid = false; 4652 } 4653 4654 if (info.deterministic && cmin == cmax) { 4655 info.deterministic = detm; 4656 } else { 4657 info.deterministic = false; 4658 } 4659 return next.study(info); 4660 } 4661 } 4662 4663 /** 4664 * A Guard node at the end of each atom node in a Branch. It 4665 * serves the purpose of chaining the "match" operation to 4666 * "next" but not the "study", so we can collect the TreeInfo 4667 * of each atom node without including the TreeInfo of the 4668 * "next". 4669 */ 4670 static final class BranchConn extends Node { 4671 BranchConn() {}; 4672 boolean match(Matcher matcher, int i, CharSequence seq) { 4673 return next.match(matcher, i, seq); 4674 } 4675 boolean study(TreeInfo info) { 4676 return info.deterministic; 4677 } 4678 } 4679 4680 /** 4681 * Handles the branching of alternations. Note this is also used for 4682 * the ? quantifier to branch between the case where it matches once 4683 * and where it does not occur. 4684 */ 4685 static final class Branch extends Node { 4686 Node[] atoms = new Node[2]; 4687 int size = 2; 4688 Node conn; 4689 Branch(Node first, Node second, Node branchConn) { 4690 conn = branchConn; 4691 atoms[0] = first; 4692 atoms[1] = second; 4693 } 4694 4695 void add(Node node) { 4696 if (size >= atoms.length) { 4697 Node[] tmp = new Node[atoms.length*2]; 4698 System.arraycopy(atoms, 0, tmp, 0, atoms.length); 4699 atoms = tmp; 4700 } 4701 atoms[size++] = node; 4702 } 4703 4704 boolean match(Matcher matcher, int i, CharSequence seq) { 4705 for (int n = 0; n < size; n++) { 4706 if (atoms[n] == null) { 4707 if (conn.next.match(matcher, i, seq)) 4708 return true; 4709 } else if (atoms[n].match(matcher, i, seq)) { 4710 return true; 4711 } 4712 } 4713 return false; 4714 } 4715 4716 boolean study(TreeInfo info) { 4717 int minL = info.minLength; 4718 int maxL = info.maxLength; 4719 boolean maxV = info.maxValid; 4720 4721 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num 4722 int maxL2 = -1; 4723 for (int n = 0; n < size; n++) { 4724 info.reset(); 4725 if (atoms[n] != null) 4726 atoms[n].study(info); 4727 minL2 = Math.min(minL2, info.minLength); 4728 maxL2 = Math.max(maxL2, info.maxLength); 4729 maxV = (maxV & info.maxValid); 4730 } 4731 4732 minL += minL2; 4733 maxL += maxL2; 4734 4735 info.reset(); 4736 conn.next.study(info); 4737 4738 info.minLength += minL; 4739 info.maxLength += maxL; 4740 info.maxValid &= maxV; 4741 info.deterministic = false; 4742 return false; 4743 } 4744 } 4745 4746 /** 4747 * The GroupHead saves the location where the group begins in the locals 4748 * and restores them when the match is done. 4749 * 4750 * The matchRef is used when a reference to this group is accessed later 4751 * in the expression. The locals will have a negative value in them to 4752 * indicate that we do not want to unset the group if the reference 4753 * doesn't match. 4754 */ 4755 static final class GroupHead extends Node { 4756 int localIndex; 4757 GroupTail tail; // for debug/print only, match does not need to know 4758 GroupHead(int localCount) { 4759 localIndex = localCount; 4760 } 4761 boolean match(Matcher matcher, int i, CharSequence seq) { 4762 int save = matcher.locals[localIndex]; 4763 matcher.locals[localIndex] = i; 4764 boolean ret = next.match(matcher, i, seq); 4765 matcher.locals[localIndex] = save; 4766 return ret; 4767 } 4768 boolean matchRef(Matcher matcher, int i, CharSequence seq) { 4769 int save = matcher.locals[localIndex]; 4770 matcher.locals[localIndex] = ~i; // HACK 4771 boolean ret = next.match(matcher, i, seq); 4772 matcher.locals[localIndex] = save; 4773 return ret; 4774 } 4775 } 4776 4777 /** 4778 * Recursive reference to a group in the regular expression. It calls 4779 * matchRef because if the reference fails to match we would not unset 4780 * the group. 4781 */ 4782 static final class GroupRef extends Node { 4783 GroupHead head; 4784 GroupRef(GroupHead head) { 4785 this.head = head; 4786 } 4787 boolean match(Matcher matcher, int i, CharSequence seq) { 4788 return head.matchRef(matcher, i, seq) 4789 && next.match(matcher, matcher.last, seq); 4790 } 4791 boolean study(TreeInfo info) { 4792 info.maxValid = false; 4793 info.deterministic = false; 4794 return next.study(info); 4795 } 4796 } 4797 4798 /** 4799 * The GroupTail handles the setting of group beginning and ending 4800 * locations when groups are successfully matched. It must also be able to 4801 * unset groups that have to be backed off of. 4802 * 4803 * The GroupTail node is also used when a previous group is referenced, 4804 * and in that case no group information needs to be set. 4805 */ 4806 static final class GroupTail extends Node { 4807 int localIndex; 4808 int groupIndex; 4809 GroupTail(int localCount, int groupCount) { 4810 localIndex = localCount; 4811 groupIndex = groupCount + groupCount; 4812 } 4813 boolean match(Matcher matcher, int i, CharSequence seq) { 4814 int tmp = matcher.locals[localIndex]; 4815 if (tmp >= 0) { // This is the normal group case. 4816 // Save the group so we can unset it if it 4817 // backs off of a match. 4818 int groupStart = matcher.groups[groupIndex]; 4819 int groupEnd = matcher.groups[groupIndex+1]; 4820 4821 matcher.groups[groupIndex] = tmp; 4822 matcher.groups[groupIndex+1] = i; 4823 if (next.match(matcher, i, seq)) { 4824 return true; 4825 } 4826 matcher.groups[groupIndex] = groupStart; 4827 matcher.groups[groupIndex+1] = groupEnd; 4828 return false; 4829 } else { 4830 // This is a group reference case. We don't need to save any 4831 // group info because it isn't really a group. 4832 matcher.last = i; 4833 return true; 4834 } 4835 } 4836 } 4837 4838 /** 4839 * This sets up a loop to handle a recursive quantifier structure. 4840 */ 4841 static final class Prolog extends Node { 4842 Loop loop; 4843 Prolog(Loop loop) { 4844 this.loop = loop; 4845 } 4846 boolean match(Matcher matcher, int i, CharSequence seq) { 4847 return loop.matchInit(matcher, i, seq); 4848 } 4849 boolean study(TreeInfo info) { 4850 return loop.study(info); 4851 } 4852 } 4853 4854 /** 4855 * Handles the repetition count for a greedy Curly. The matchInit 4856 * is called from the Prolog to save the index of where the group 4857 * beginning is stored. A zero length group check occurs in the 4858 * normal match but is skipped in the matchInit. 4859 */ 4860 static class Loop extends Node { 4861 Node body; 4862 int countIndex; // local count index in matcher locals 4863 int beginIndex; // group beginning index 4864 int cmin, cmax; 4865 int posIndex; 4866 Loop(int countIndex, int beginIndex) { 4867 this.countIndex = countIndex; 4868 this.beginIndex = beginIndex; 4869 this.posIndex = -1; 4870 } 4871 boolean match(Matcher matcher, int i, CharSequence seq) { 4872 // Avoid infinite loop in zero-length case. 4873 if (i > matcher.locals[beginIndex]) { 4874 int count = matcher.locals[countIndex]; 4875 4876 // This block is for before we reach the minimum 4877 // iterations required for the loop to match 4878 if (count < cmin) { 4879 matcher.locals[countIndex] = count + 1; 4880 boolean b = body.match(matcher, i, seq); 4881 // If match failed we must backtrack, so 4882 // the loop count should NOT be incremented 4883 if (!b) 4884 matcher.locals[countIndex] = count; 4885 // Return success or failure since we are under 4886 // minimum 4887 return b; 4888 } 4889 // This block is for after we have the minimum 4890 // iterations required for the loop to match 4891 if (count < cmax) { 4892 // Let's check if we have already tried and failed 4893 // at this starting position "i" in the past. 4894 // If yes, then just return false wihtout trying 4895 // again, to stop the exponential backtracking. 4896 if (posIndex != -1 && 4897 matcher.localsPos[posIndex].contains(i)) { 4898 return next.match(matcher, i, seq); 4899 } 4900 matcher.locals[countIndex] = count + 1; 4901 boolean b = body.match(matcher, i, seq); 4902 // If match failed we must backtrack, so 4903 // the loop count should NOT be incremented 4904 if (b) 4905 return true; 4906 matcher.locals[countIndex] = count; 4907 // save the failed position 4908 if (posIndex != -1) { 4909 matcher.localsPos[posIndex].add(i); 4910 } 4911 } 4912 } 4913 return next.match(matcher, i, seq); 4914 } 4915 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4916 int save = matcher.locals[countIndex]; 4917 boolean ret = false; 4918 if (posIndex != -1 && matcher.localsPos[posIndex] == null) { 4919 matcher.localsPos[posIndex] = new IntHashSet(); 4920 } 4921 if (0 < cmin) { 4922 matcher.locals[countIndex] = 1; 4923 ret = body.match(matcher, i, seq); 4924 } else if (0 < cmax) { 4925 matcher.locals[countIndex] = 1; 4926 ret = body.match(matcher, i, seq); 4927 if (ret == false) 4928 ret = next.match(matcher, i, seq); 4929 } else { 4930 ret = next.match(matcher, i, seq); 4931 } 4932 matcher.locals[countIndex] = save; 4933 return ret; 4934 } 4935 boolean study(TreeInfo info) { 4936 info.maxValid = false; 4937 info.deterministic = false; 4938 return false; 4939 } 4940 } 4941 4942 /** 4943 * Handles the repetition count for a reluctant Curly. The matchInit 4944 * is called from the Prolog to save the index of where the group 4945 * beginning is stored. A zero length group check occurs in the 4946 * normal match but is skipped in the matchInit. 4947 */ 4948 static final class LazyLoop extends Loop { 4949 LazyLoop(int countIndex, int beginIndex) { 4950 super(countIndex, beginIndex); 4951 } 4952 boolean match(Matcher matcher, int i, CharSequence seq) { 4953 // Check for zero length group 4954 if (i > matcher.locals[beginIndex]) { 4955 int count = matcher.locals[countIndex]; 4956 if (count < cmin) { 4957 matcher.locals[countIndex] = count + 1; 4958 boolean result = body.match(matcher, i, seq); 4959 // If match failed we must backtrack, so 4960 // the loop count should NOT be incremented 4961 if (!result) 4962 matcher.locals[countIndex] = count; 4963 return result; 4964 } 4965 if (next.match(matcher, i, seq)) 4966 return true; 4967 if (count < cmax) { 4968 matcher.locals[countIndex] = count + 1; 4969 boolean result = body.match(matcher, i, seq); 4970 // If match failed we must backtrack, so 4971 // the loop count should NOT be incremented 4972 if (!result) 4973 matcher.locals[countIndex] = count; 4974 return result; 4975 } 4976 return false; 4977 } 4978 return next.match(matcher, i, seq); 4979 } 4980 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4981 int save = matcher.locals[countIndex]; 4982 boolean ret = false; 4983 if (0 < cmin) { 4984 matcher.locals[countIndex] = 1; 4985 ret = body.match(matcher, i, seq); 4986 } else if (next.match(matcher, i, seq)) { 4987 ret = true; 4988 } else if (0 < cmax) { 4989 matcher.locals[countIndex] = 1; 4990 ret = body.match(matcher, i, seq); 4991 } 4992 matcher.locals[countIndex] = save; 4993 return ret; 4994 } 4995 boolean study(TreeInfo info) { 4996 info.maxValid = false; 4997 info.deterministic = false; 4998 return false; 4999 } 5000 } 5001 5002 /** 5003 * Refers to a group in the regular expression. Attempts to match 5004 * whatever the group referred to last matched. 5005 */ 5006 static class BackRef extends Node { 5007 int groupIndex; 5008 BackRef(int groupCount) { 5009 super(); 5010 groupIndex = groupCount + groupCount; 5011 } 5012 boolean match(Matcher matcher, int i, CharSequence seq) { 5013 int j = matcher.groups[groupIndex]; 5014 int k = matcher.groups[groupIndex+1]; 5015 5016 int groupSize = k - j; 5017 // If the referenced group didn't match, neither can this 5018 if (j < 0) 5019 return false; 5020 5021 // If there isn't enough input left no match 5022 if (i + groupSize > matcher.to) { 5023 matcher.hitEnd = true; 5024 return false; 5025 } 5026 // Check each new char to make sure it matches what the group 5027 // referenced matched last time around 5028 for (int index=0; index<groupSize; index++) 5029 if (seq.charAt(i+index) != seq.charAt(j+index)) 5030 return false; 5031 5032 return next.match(matcher, i+groupSize, seq); 5033 } 5034 boolean study(TreeInfo info) { 5035 info.maxValid = false; 5036 return next.study(info); 5037 } 5038 } 5039 5040 static class CIBackRef extends Node { 5041 int groupIndex; 5042 boolean doUnicodeCase; 5043 CIBackRef(int groupCount, boolean doUnicodeCase) { 5044 super(); 5045 groupIndex = groupCount + groupCount; 5046 this.doUnicodeCase = doUnicodeCase; 5047 } 5048 boolean match(Matcher matcher, int i, CharSequence seq) { 5049 int j = matcher.groups[groupIndex]; 5050 int k = matcher.groups[groupIndex+1]; 5051 5052 int groupSize = k - j; 5053 5054 // If the referenced group didn't match, neither can this 5055 if (j < 0) 5056 return false; 5057 5058 // If there isn't enough input left no match 5059 if (i + groupSize > matcher.to) { 5060 matcher.hitEnd = true; 5061 return false; 5062 } 5063 5064 // Check each new char to make sure it matches what the group 5065 // referenced matched last time around 5066 int x = i; 5067 for (int index=0; index<groupSize; index++) { 5068 int c1 = Character.codePointAt(seq, x); 5069 int c2 = Character.codePointAt(seq, j); 5070 if (c1 != c2) { 5071 if (doUnicodeCase) { 5072 int cc1 = Character.toUpperCase(c1); 5073 int cc2 = Character.toUpperCase(c2); 5074 if (cc1 != cc2 && 5075 Character.toLowerCase(cc1) != 5076 Character.toLowerCase(cc2)) 5077 return false; 5078 } else { 5079 if (ASCII.toLower(c1) != ASCII.toLower(c2)) 5080 return false; 5081 } 5082 } 5083 x += Character.charCount(c1); 5084 j += Character.charCount(c2); 5085 } 5086 5087 return next.match(matcher, i+groupSize, seq); 5088 } 5089 boolean study(TreeInfo info) { 5090 info.maxValid = false; 5091 return next.study(info); 5092 } 5093 } 5094 5095 /** 5096 * Searches until the next instance of its atom. This is useful for 5097 * finding the atom efficiently without passing an instance of it 5098 * (greedy problem) and without a lot of wasted search time (reluctant 5099 * problem). 5100 */ 5101 static final class First extends Node { 5102 Node atom; 5103 First(Node node) { 5104 this.atom = BnM.optimize(node); 5105 } 5106 boolean match(Matcher matcher, int i, CharSequence seq) { 5107 if (atom instanceof BnM) { 5108 return atom.match(matcher, i, seq) 5109 && next.match(matcher, matcher.last, seq); 5110 } 5111 for (;;) { 5112 if (i > matcher.to) { 5113 matcher.hitEnd = true; 5114 return false; 5115 } 5116 if (atom.match(matcher, i, seq)) { 5117 return next.match(matcher, matcher.last, seq); 5118 } 5119 i += countChars(seq, i, 1); 5120 matcher.first++; 5121 } 5122 } 5123 boolean study(TreeInfo info) { 5124 atom.study(info); 5125 info.maxValid = false; 5126 info.deterministic = false; 5127 return next.study(info); 5128 } 5129 } 5130 5131 static final class Conditional extends Node { 5132 Node cond, yes, not; 5133 Conditional(Node cond, Node yes, Node not) { 5134 this.cond = cond; 5135 this.yes = yes; 5136 this.not = not; 5137 } 5138 boolean match(Matcher matcher, int i, CharSequence seq) { 5139 if (cond.match(matcher, i, seq)) { 5140 return yes.match(matcher, i, seq); 5141 } else { 5142 return not.match(matcher, i, seq); 5143 } 5144 } 5145 boolean study(TreeInfo info) { 5146 int minL = info.minLength; 5147 int maxL = info.maxLength; 5148 boolean maxV = info.maxValid; 5149 info.reset(); 5150 yes.study(info); 5151 5152 int minL2 = info.minLength; 5153 int maxL2 = info.maxLength; 5154 boolean maxV2 = info.maxValid; 5155 info.reset(); 5156 not.study(info); 5157 5158 info.minLength = minL + Math.min(minL2, info.minLength); 5159 info.maxLength = maxL + Math.max(maxL2, info.maxLength); 5160 info.maxValid = (maxV & maxV2 & info.maxValid); 5161 info.deterministic = false; 5162 return next.study(info); 5163 } 5164 } 5165 5166 /** 5167 * Zero width positive lookahead. 5168 */ 5169 static final class Pos extends Node { 5170 Node cond; 5171 Pos(Node cond) { 5172 this.cond = cond; 5173 } 5174 boolean match(Matcher matcher, int i, CharSequence seq) { 5175 int savedTo = matcher.to; 5176 boolean conditionMatched = false; 5177 5178 // Relax transparent region boundaries for lookahead 5179 if (matcher.transparentBounds) 5180 matcher.to = matcher.getTextLength(); 5181 try { 5182 conditionMatched = cond.match(matcher, i, seq); 5183 } finally { 5184 // Reinstate region boundaries 5185 matcher.to = savedTo; 5186 } 5187 return conditionMatched && next.match(matcher, i, seq); 5188 } 5189 } 5190 5191 /** 5192 * Zero width negative lookahead. 5193 */ 5194 static final class Neg extends Node { 5195 Node cond; 5196 Neg(Node cond) { 5197 this.cond = cond; 5198 } 5199 boolean match(Matcher matcher, int i, CharSequence seq) { 5200 int savedTo = matcher.to; 5201 boolean conditionMatched = false; 5202 5203 // Relax transparent region boundaries for lookahead 5204 if (matcher.transparentBounds) 5205 matcher.to = matcher.getTextLength(); 5206 try { 5207 if (i < matcher.to) { 5208 conditionMatched = !cond.match(matcher, i, seq); 5209 } else { 5210 // If a negative lookahead succeeds then more input 5211 // could cause it to fail! 5212 matcher.requireEnd = true; 5213 conditionMatched = !cond.match(matcher, i, seq); 5214 } 5215 } finally { 5216 // Reinstate region boundaries 5217 matcher.to = savedTo; 5218 } 5219 return conditionMatched && next.match(matcher, i, seq); 5220 } 5221 } 5222 5223 /** 5224 * For use with lookbehinds; matches the position where the lookbehind 5225 * was encountered. 5226 */ 5227 static Node lookbehindEnd = new Node() { 5228 boolean match(Matcher matcher, int i, CharSequence seq) { 5229 return i == matcher.lookbehindTo; 5230 } 5231 }; 5232 5233 /** 5234 * Zero width positive lookbehind. 5235 */ 5236 static class Behind extends Node { 5237 Node cond; 5238 int rmax, rmin; 5239 Behind(Node cond, int rmax, int rmin) { 5240 this.cond = cond; 5241 this.rmax = rmax; 5242 this.rmin = rmin; 5243 } 5244 5245 boolean match(Matcher matcher, int i, CharSequence seq) { 5246 int savedFrom = matcher.from; 5247 boolean conditionMatched = false; 5248 int startIndex = (!matcher.transparentBounds) ? 5249 matcher.from : 0; 5250 int from = Math.max(i - rmax, startIndex); 5251 // Set end boundary 5252 int savedLBT = matcher.lookbehindTo; 5253 matcher.lookbehindTo = i; 5254 // Relax transparent region boundaries for lookbehind 5255 if (matcher.transparentBounds) 5256 matcher.from = 0; 5257 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5258 conditionMatched = cond.match(matcher, j, seq); 5259 } 5260 matcher.from = savedFrom; 5261 matcher.lookbehindTo = savedLBT; 5262 return conditionMatched && next.match(matcher, i, seq); 5263 } 5264 } 5265 5266 /** 5267 * Zero width positive lookbehind, including supplementary 5268 * characters or unpaired surrogates. 5269 */ 5270 static final class BehindS extends Behind { 5271 BehindS(Node cond, int rmax, int rmin) { 5272 super(cond, rmax, rmin); 5273 } 5274 boolean match(Matcher matcher, int i, CharSequence seq) { 5275 int rmaxChars = countChars(seq, i, -rmax); 5276 int rminChars = countChars(seq, i, -rmin); 5277 int savedFrom = matcher.from; 5278 int startIndex = (!matcher.transparentBounds) ? 5279 matcher.from : 0; 5280 boolean conditionMatched = false; 5281 int from = Math.max(i - rmaxChars, startIndex); 5282 // Set end boundary 5283 int savedLBT = matcher.lookbehindTo; 5284 matcher.lookbehindTo = i; 5285 // Relax transparent region boundaries for lookbehind 5286 if (matcher.transparentBounds) 5287 matcher.from = 0; 5288 5289 for (int j = i - rminChars; 5290 !conditionMatched && j >= from; 5291 j -= j>from ? countChars(seq, j, -1) : 1) { 5292 conditionMatched = cond.match(matcher, j, seq); 5293 } 5294 matcher.from = savedFrom; 5295 matcher.lookbehindTo = savedLBT; 5296 return conditionMatched && next.match(matcher, i, seq); 5297 } 5298 } 5299 5300 /** 5301 * Zero width negative lookbehind. 5302 */ 5303 static class NotBehind extends Node { 5304 Node cond; 5305 int rmax, rmin; 5306 NotBehind(Node cond, int rmax, int rmin) { 5307 this.cond = cond; 5308 this.rmax = rmax; 5309 this.rmin = rmin; 5310 } 5311 5312 boolean match(Matcher matcher, int i, CharSequence seq) { 5313 int savedLBT = matcher.lookbehindTo; 5314 int savedFrom = matcher.from; 5315 boolean conditionMatched = false; 5316 int startIndex = (!matcher.transparentBounds) ? 5317 matcher.from : 0; 5318 int from = Math.max(i - rmax, startIndex); 5319 matcher.lookbehindTo = i; 5320 // Relax transparent region boundaries for lookbehind 5321 if (matcher.transparentBounds) 5322 matcher.from = 0; 5323 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5324 conditionMatched = cond.match(matcher, j, seq); 5325 } 5326 // Reinstate region boundaries 5327 matcher.from = savedFrom; 5328 matcher.lookbehindTo = savedLBT; 5329 return !conditionMatched && next.match(matcher, i, seq); 5330 } 5331 } 5332 5333 /** 5334 * Zero width negative lookbehind, including supplementary 5335 * characters or unpaired surrogates. 5336 */ 5337 static final class NotBehindS extends NotBehind { 5338 NotBehindS(Node cond, int rmax, int rmin) { 5339 super(cond, rmax, rmin); 5340 } 5341 boolean match(Matcher matcher, int i, CharSequence seq) { 5342 int rmaxChars = countChars(seq, i, -rmax); 5343 int rminChars = countChars(seq, i, -rmin); 5344 int savedFrom = matcher.from; 5345 int savedLBT = matcher.lookbehindTo; 5346 boolean conditionMatched = false; 5347 int startIndex = (!matcher.transparentBounds) ? 5348 matcher.from : 0; 5349 int from = Math.max(i - rmaxChars, startIndex); 5350 matcher.lookbehindTo = i; 5351 // Relax transparent region boundaries for lookbehind 5352 if (matcher.transparentBounds) 5353 matcher.from = 0; 5354 for (int j = i - rminChars; 5355 !conditionMatched && j >= from; 5356 j -= j>from ? countChars(seq, j, -1) : 1) { 5357 conditionMatched = cond.match(matcher, j, seq); 5358 } 5359 //Reinstate region boundaries 5360 matcher.from = savedFrom; 5361 matcher.lookbehindTo = savedLBT; 5362 return !conditionMatched && next.match(matcher, i, seq); 5363 } 5364 } 5365 5366 /** 5367 * Handles word boundaries. Includes a field to allow this one class to 5368 * deal with the different types of word boundaries we can match. The word 5369 * characters include underscores, letters, and digits. Non spacing marks 5370 * can are also part of a word if they have a base character, otherwise 5371 * they are ignored for purposes of finding word boundaries. 5372 */ 5373 static final class Bound extends Node { 5374 static int LEFT = 0x1; 5375 static int RIGHT= 0x2; 5376 static int BOTH = 0x3; 5377 static int NONE = 0x4; 5378 int type; 5379 boolean useUWORD; 5380 Bound(int n, boolean useUWORD) { 5381 type = n; 5382 this.useUWORD = useUWORD; 5383 } 5384 5385 boolean isWord(int ch) { 5386 return useUWORD ? CharPredicates.WORD().is(ch) 5387 : (ch == '_' || Character.isLetterOrDigit(ch)); 5388 } 5389 5390 int check(Matcher matcher, int i, CharSequence seq) { 5391 int ch; 5392 boolean left = false; 5393 int startIndex = matcher.from; 5394 int endIndex = matcher.to; 5395 if (matcher.transparentBounds) { 5396 startIndex = 0; 5397 endIndex = matcher.getTextLength(); 5398 } 5399 if (i > startIndex) { 5400 ch = Character.codePointBefore(seq, i); 5401 left = (isWord(ch) || 5402 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5403 && hasBaseCharacter(matcher, i-1, seq))); 5404 } 5405 boolean right = false; 5406 if (i < endIndex) { 5407 ch = Character.codePointAt(seq, i); 5408 right = (isWord(ch) || 5409 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5410 && hasBaseCharacter(matcher, i, seq))); 5411 } else { 5412 // Tried to access char past the end 5413 matcher.hitEnd = true; 5414 // The addition of another char could wreck a boundary 5415 matcher.requireEnd = true; 5416 } 5417 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE); 5418 } 5419 boolean match(Matcher matcher, int i, CharSequence seq) { 5420 return (check(matcher, i, seq) & type) > 0 5421 && next.match(matcher, i, seq); 5422 } 5423 } 5424 5425 /** 5426 * Non spacing marks only count as word characters in bounds calculations 5427 * if they have a base character. 5428 */ 5429 private static boolean hasBaseCharacter(Matcher matcher, int i, 5430 CharSequence seq) 5431 { 5432 int start = (!matcher.transparentBounds) ? 5433 matcher.from : 0; 5434 for (int x=i; x >= start; x--) { 5435 int ch = Character.codePointAt(seq, x); 5436 if (Character.isLetterOrDigit(ch)) 5437 return true; 5438 if (Character.getType(ch) == Character.NON_SPACING_MARK) 5439 continue; 5440 return false; 5441 } 5442 return false; 5443 } 5444 5445 /** 5446 * Attempts to match a slice in the input using the Boyer-Moore string 5447 * matching algorithm. The algorithm is based on the idea that the 5448 * pattern can be shifted farther ahead in the search text if it is 5449 * matched right to left. 5450 * <p> 5451 * The pattern is compared to the input one character at a time, from 5452 * the rightmost character in the pattern to the left. If the characters 5453 * all match the pattern has been found. If a character does not match, 5454 * the pattern is shifted right a distance that is the maximum of two 5455 * functions, the bad character shift and the good suffix shift. This 5456 * shift moves the attempted match position through the input more 5457 * quickly than a naive one position at a time check. 5458 * <p> 5459 * The bad character shift is based on the character from the text that 5460 * did not match. If the character does not appear in the pattern, the 5461 * pattern can be shifted completely beyond the bad character. If the 5462 * character does occur in the pattern, the pattern can be shifted to 5463 * line the pattern up with the next occurrence of that character. 5464 * <p> 5465 * The good suffix shift is based on the idea that some subset on the right 5466 * side of the pattern has matched. When a bad character is found, the 5467 * pattern can be shifted right by the pattern length if the subset does 5468 * not occur again in pattern, or by the amount of distance to the 5469 * next occurrence of the subset in the pattern. 5470 * 5471 * Boyer-Moore search methods adapted from code by Amy Yu. 5472 */ 5473 static class BnM extends Node { 5474 int[] buffer; 5475 int[] lastOcc; 5476 int[] optoSft; 5477 5478 /** 5479 * Pre calculates arrays needed to generate the bad character 5480 * shift and the good suffix shift. Only the last seven bits 5481 * are used to see if chars match; This keeps the tables small 5482 * and covers the heavily used ASCII range, but occasionally 5483 * results in an aliased match for the bad character shift. 5484 */ 5485 static Node optimize(Node node) { 5486 if (!(node instanceof Slice)) { 5487 return node; 5488 } 5489 5490 int[] src = ((Slice) node).buffer; 5491 int patternLength = src.length; 5492 // The BM algorithm requires a bit of overhead; 5493 // If the pattern is short don't use it, since 5494 // a shift larger than the pattern length cannot 5495 // be used anyway. 5496 if (patternLength < 4) { 5497 return node; 5498 } 5499 int i, j, k; 5500 int[] lastOcc = new int[128]; 5501 int[] optoSft = new int[patternLength]; 5502 // Precalculate part of the bad character shift 5503 // It is a table for where in the pattern each 5504 // lower 7-bit value occurs 5505 for (i = 0; i < patternLength; i++) { 5506 lastOcc[src[i]&0x7F] = i + 1; 5507 } 5508 // Precalculate the good suffix shift 5509 // i is the shift amount being considered 5510 NEXT: for (i = patternLength; i > 0; i--) { 5511 // j is the beginning index of suffix being considered 5512 for (j = patternLength - 1; j >= i; j--) { 5513 // Testing for good suffix 5514 if (src[j] == src[j-i]) { 5515 // src[j..len] is a good suffix 5516 optoSft[j-1] = i; 5517 } else { 5518 // No match. The array has already been 5519 // filled up with correct values before. 5520 continue NEXT; 5521 } 5522 } 5523 // This fills up the remaining of optoSft 5524 // any suffix can not have larger shift amount 5525 // then its sub-suffix. Why??? 5526 while (j > 0) { 5527 optoSft[--j] = i; 5528 } 5529 } 5530 // Set the guard value because of unicode compression 5531 optoSft[patternLength-1] = 1; 5532 if (node instanceof SliceS) 5533 return new BnMS(src, lastOcc, optoSft, node.next); 5534 return new BnM(src, lastOcc, optoSft, node.next); 5535 } 5536 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5537 this.buffer = src; 5538 this.lastOcc = lastOcc; 5539 this.optoSft = optoSft; 5540 this.next = next; 5541 } 5542 boolean match(Matcher matcher, int i, CharSequence seq) { 5543 int[] src = buffer; 5544 int patternLength = src.length; 5545 int last = matcher.to - patternLength; 5546 5547 // Loop over all possible match positions in text 5548 NEXT: while (i <= last) { 5549 // Loop over pattern from right to left 5550 for (int j = patternLength - 1; j >= 0; j--) { 5551 int ch = seq.charAt(i+j); 5552 if (ch != src[j]) { 5553 // Shift search to the right by the maximum of the 5554 // bad character shift and the good suffix shift 5555 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]); 5556 continue NEXT; 5557 } 5558 } 5559 // Entire pattern matched starting at i 5560 matcher.first = i; 5561 boolean ret = next.match(matcher, i + patternLength, seq); 5562 if (ret) { 5563 matcher.first = i; 5564 matcher.groups[0] = matcher.first; 5565 matcher.groups[1] = matcher.last; 5566 return true; 5567 } 5568 i++; 5569 } 5570 // BnM is only used as the leading node in the unanchored case, 5571 // and it replaced its Start() which always searches to the end 5572 // if it doesn't find what it's looking for, so hitEnd is true. 5573 matcher.hitEnd = true; 5574 return false; 5575 } 5576 boolean study(TreeInfo info) { 5577 info.minLength += buffer.length; 5578 info.maxValid = false; 5579 return next.study(info); 5580 } 5581 } 5582 5583 /** 5584 * Supplementary support version of BnM(). Unpaired surrogates are 5585 * also handled by this class. 5586 */ 5587 static final class BnMS extends BnM { 5588 int lengthInChars; 5589 5590 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5591 super(src, lastOcc, optoSft, next); 5592 for (int cp : buffer) { 5593 lengthInChars += Character.charCount(cp); 5594 } 5595 } 5596 boolean match(Matcher matcher, int i, CharSequence seq) { 5597 int[] src = buffer; 5598 int patternLength = src.length; 5599 int last = matcher.to - lengthInChars; 5600 5601 // Loop over all possible match positions in text 5602 NEXT: while (i <= last) { 5603 // Loop over pattern from right to left 5604 int ch; 5605 for (int j = countChars(seq, i, patternLength), x = patternLength - 1; 5606 j > 0; j -= Character.charCount(ch), x--) { 5607 ch = Character.codePointBefore(seq, i+j); 5608 if (ch != src[x]) { 5609 // Shift search to the right by the maximum of the 5610 // bad character shift and the good suffix shift 5611 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]); 5612 i += countChars(seq, i, n); 5613 continue NEXT; 5614 } 5615 } 5616 // Entire pattern matched starting at i 5617 matcher.first = i; 5618 boolean ret = next.match(matcher, i + lengthInChars, seq); 5619 if (ret) { 5620 matcher.first = i; 5621 matcher.groups[0] = matcher.first; 5622 matcher.groups[1] = matcher.last; 5623 return true; 5624 } 5625 i += countChars(seq, i, 1); 5626 } 5627 matcher.hitEnd = true; 5628 return false; 5629 } 5630 } 5631 5632 @FunctionalInterface 5633 static interface CharPredicate { 5634 boolean is(int ch); 5635 5636 default CharPredicate and(CharPredicate p) { 5637 return ch -> is(ch) && p.is(ch); 5638 } 5639 default CharPredicate union(CharPredicate p) { 5640 return ch -> is(ch) || p.is(ch); 5641 } 5642 default CharPredicate union(CharPredicate p1, 5643 CharPredicate p2 ) { 5644 return ch -> is(ch) || p1.is(ch) || p2.is(ch); 5645 } 5646 default CharPredicate negate() { 5647 return ch -> !is(ch); 5648 } 5649 } 5650 5651 static interface BmpCharPredicate extends CharPredicate { 5652 5653 default CharPredicate and(CharPredicate p) { 5654 if(p instanceof BmpCharPredicate) 5655 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch)); 5656 return ch -> is(ch) && p.is(ch); 5657 } 5658 default CharPredicate union(CharPredicate p) { 5659 if (p instanceof BmpCharPredicate) 5660 return (BmpCharPredicate)(ch -> is(ch) || p.is(ch)); 5661 return ch -> is(ch) || p.is(ch); 5662 } 5663 static CharPredicate union(CharPredicate... predicates) { 5664 CharPredicate cp = ch -> { 5665 for (CharPredicate p : predicates) { 5666 if (!p.is(ch)) 5667 return false; 5668 } 5669 return true; 5670 }; 5671 for (CharPredicate p : predicates) { 5672 if (! (p instanceof BmpCharPredicate)) 5673 return cp; 5674 } 5675 return (BmpCharPredicate)cp; 5676 } 5677 } 5678 5679 /** 5680 * matches a Perl vertical whitespace 5681 */ 5682 static BmpCharPredicate VertWS() { 5683 return cp -> (cp >= 0x0A && cp <= 0x0D) || 5684 cp == 0x85 || cp == 0x2028 || cp == 0x2029; 5685 } 5686 5687 /** 5688 * matches a Perl horizontal whitespace 5689 */ 5690 static BmpCharPredicate HorizWS() { 5691 return cp -> 5692 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 || 5693 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f || 5694 cp == 0x205f || cp == 0x3000; 5695 } 5696 5697 /** 5698 * for the Unicode category ALL and the dot metacharacter when 5699 * in dotall mode. 5700 */ 5701 static CharPredicate ALL() { 5702 return ch -> true; 5703 } 5704 5705 /** 5706 * for the dot metacharacter when dotall is not enabled. 5707 */ 5708 static CharPredicate DOT() { 5709 return ch -> 5710 (ch != '\n' && ch != '\r' 5711 && (ch|1) != '\u2029' 5712 && ch != '\u0085'); 5713 } 5714 5715 /** 5716 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled. 5717 */ 5718 static CharPredicate UNIXDOT() { 5719 return ch -> ch != '\n'; 5720 } 5721 5722 /** 5723 * Indicate that matches a Supplementary Unicode character 5724 */ 5725 static CharPredicate SingleS(int c) { 5726 return ch -> ch == c; 5727 } 5728 5729 /** 5730 * A bmp/optimized predicate of single 5731 */ 5732 static BmpCharPredicate Single(int c) { 5733 return ch -> ch == c; 5734 } 5735 5736 /** 5737 * Case insensitive matches a given BMP character 5738 */ 5739 static BmpCharPredicate SingleI(int lower, int upper) { 5740 return ch -> ch == lower || ch == upper; 5741 } 5742 5743 /** 5744 * Unicode case insensitive matches a given Unicode character 5745 */ 5746 static CharPredicate SingleU(int lower) { 5747 return ch -> lower == ch || 5748 lower == Character.toLowerCase(Character.toUpperCase(ch)); 5749 } 5750 5751 private static boolean inRange(int lower, int ch, int upper) { 5752 return lower <= ch && ch <= upper; 5753 } 5754 5755 /** 5756 * Charactrs within a explicit value range 5757 */ 5758 static CharPredicate Range(int lower, int upper) { 5759 if (upper < Character.MIN_HIGH_SURROGATE || 5760 lower > Character.MAX_HIGH_SURROGATE && 5761 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT) 5762 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper)); 5763 return ch -> inRange(lower, ch, upper); 5764 } 5765 5766 /** 5767 * Charactrs within a explicit value range in a case insensitive manner. 5768 */ 5769 static CharPredicate CIRange(int lower, int upper) { 5770 return ch -> inRange(lower, ch, upper) || 5771 ASCII.isAscii(ch) && 5772 (inRange(lower, ASCII.toUpper(ch), upper) || 5773 inRange(lower, ASCII.toLower(ch), upper)); 5774 } 5775 5776 static CharPredicate CIRangeU(int lower, int upper) { 5777 return ch -> { 5778 if (inRange(lower, ch, upper)) 5779 return true; 5780 int up = Character.toUpperCase(ch); 5781 return inRange(lower, up, upper) || 5782 inRange(lower, Character.toLowerCase(up), upper); 5783 }; 5784 } 5785 5786 /** 5787 * This must be the very first initializer. 5788 */ 5789 static final Node accept = new Node(); 5790 5791 static final Node lastAccept = new LastNode(); 5792 5793 /** 5794 * Creates a predicate which can be used to match a string. 5795 * 5796 * @return The predicate which can be used for matching on a string 5797 * @since 1.8 5798 */ 5799 public Predicate<String> asPredicate() { 5800 return s -> matcher(s).find(); 5801 } 5802 5803 /** 5804 * Creates a stream from the given input sequence around matches of this 5805 * pattern. 5806 * 5807 * <p> The stream returned by this method contains each substring of the 5808 * input sequence that is terminated by another subsequence that matches 5809 * this pattern or is terminated by the end of the input sequence. The 5810 * substrings in the stream are in the order in which they occur in the 5811 * input. Trailing empty strings will be discarded and not encountered in 5812 * the stream. 5813 * 5814 * <p> If this pattern does not match any subsequence of the input then 5815 * the resulting stream has just one element, namely the input sequence in 5816 * string form. 5817 * 5818 * <p> When there is a positive-width match at the beginning of the input 5819 * sequence then an empty leading substring is included at the beginning 5820 * of the stream. A zero-width match at the beginning however never produces 5821 * such empty leading substring. 5822 * 5823 * <p> If the input sequence is mutable, it must remain constant during the 5824 * execution of the terminal stream operation. Otherwise, the result of the 5825 * terminal stream operation is undefined. 5826 * 5827 * @param input 5828 * The character sequence to be split 5829 * 5830 * @return The stream of strings computed by splitting the input 5831 * around matches of this pattern 5832 * @see #split(CharSequence) 5833 * @since 1.8 5834 */ 5835 public Stream<String> splitAsStream(final CharSequence input) { 5836 class MatcherIterator implements Iterator<String> { 5837 private Matcher matcher; 5838 // The start position of the next sub-sequence of input 5839 // when current == input.length there are no more elements 5840 private int current; 5841 // null if the next element, if any, needs to obtained 5842 private String nextElement; 5843 // > 0 if there are N next empty elements 5844 private int emptyElementCount; 5845 5846 public String next() { 5847 if (!hasNext()) 5848 throw new NoSuchElementException(); 5849 5850 if (emptyElementCount == 0) { 5851 String n = nextElement; 5852 nextElement = null; 5853 return n; 5854 } else { 5855 emptyElementCount--; 5856 return ""; 5857 } 5858 } 5859 5860 public boolean hasNext() { 5861 if (matcher == null) { 5862 matcher = matcher(input); 5863 // If the input is an empty string then the result can only be a 5864 // stream of the input. Induce that by setting the empty 5865 // element count to 1 5866 emptyElementCount = input.length() == 0 ? 1 : 0; 5867 } 5868 if (nextElement != null || emptyElementCount > 0) 5869 return true; 5870 5871 if (current == input.length()) 5872 return false; 5873 5874 // Consume the next matching element 5875 // Count sequence of matching empty elements 5876 while (matcher.find()) { 5877 nextElement = input.subSequence(current, matcher.start()).toString(); 5878 current = matcher.end(); 5879 if (!nextElement.isEmpty()) { 5880 return true; 5881 } else if (current > 0) { // no empty leading substring for zero-width 5882 // match at the beginning of the input 5883 emptyElementCount++; 5884 } 5885 } 5886 5887 // Consume last matching element 5888 nextElement = input.subSequence(current, input.length()).toString(); 5889 current = input.length(); 5890 if (!nextElement.isEmpty()) { 5891 return true; 5892 } else { 5893 // Ignore a terminal sequence of matching empty elements 5894 emptyElementCount = 0; 5895 nextElement = null; 5896 return false; 5897 } 5898 } 5899 } 5900 return StreamSupport.stream(Spliterators.spliteratorUnknownSize( 5901 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false); 5902 } 5903 }