1 /* 2 * Copyright 1999-2008 Sun Microsystems, Inc. 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. Sun designates this 8 * particular file as subject to the "Classpath" exception as provided 9 * by Sun 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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, 22 * CA 95054 USA or visit www.sun.com if you need additional information or 23 * have any questions. 24 */ 25 26 package java.util.regex; 27 28 import java.security.AccessController; 29 import java.security.PrivilegedAction; 30 import java.text.CharacterIterator; 31 import java.text.Normalizer; 32 import java.util.Map; 33 import java.util.ArrayList; 34 import java.util.HashMap; 35 import java.util.Arrays; 36 37 38 /** 39 * A compiled representation of a regular expression. 40 * 41 * <p> A regular expression, specified as a string, must first be compiled into 42 * an instance of this class. The resulting pattern can then be used to create 43 * a {@link Matcher} object that can match arbitrary {@link 44 * java.lang.CharSequence </code>character sequences<code>} against the regular 45 * expression. All of the state involved in performing a match resides in the 46 * matcher, so many matchers can share the same pattern. 47 * 48 * <p> A typical invocation sequence is thus 49 * 50 * <blockquote><pre> 51 * Pattern p = Pattern.{@link #compile compile}("a*b"); 52 * Matcher m = p.{@link #matcher matcher}("aaaaab"); 53 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote> 54 * 55 * <p> A {@link #matches matches} method is defined by this class as a 56 * convenience for when a regular expression is used just once. This method 57 * compiles an expression and matches an input sequence against it in a single 58 * invocation. The statement 59 * 60 * <blockquote><pre> 61 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote> 62 * 63 * is equivalent to the three statements above, though for repeated matches it 64 * is less efficient since it does not allow the compiled pattern to be reused. 65 * 66 * <p> Instances of this class are immutable and are safe for use by multiple 67 * concurrent threads. Instances of the {@link Matcher} class are not safe for 68 * such use. 69 * 70 * 71 * <a name="sum"> 72 * <h4> Summary of regular-expression constructs </h4> 73 * 74 * <table border="0" cellpadding="1" cellspacing="0" 75 * summary="Regular expression constructs, and what they match"> 76 * 77 * <tr align="left"> 78 * <th bgcolor="#CCCCFF" align="left" id="construct">Construct</th> 79 * <th bgcolor="#CCCCFF" align="left" id="matches">Matches</th> 80 * </tr> 81 * 82 * <tr><th> </th></tr> 83 * <tr align="left"><th colspan="2" id="characters">Characters</th></tr> 84 * 85 * <tr><td valign="top" headers="construct characters"><i>x</i></td> 86 * <td headers="matches">The character <i>x</i></td></tr> 87 * <tr><td valign="top" headers="construct characters"><tt>\\</tt></td> 88 * <td headers="matches">The backslash character</td></tr> 89 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>n</i></td> 90 * <td headers="matches">The character with octal value <tt>0</tt><i>n</i> 91 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr> 92 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>nn</i></td> 93 * <td headers="matches">The character with octal value <tt>0</tt><i>nn</i> 94 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr> 95 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>mnn</i></td> 96 * <td headers="matches">The character with octal value <tt>0</tt><i>mnn</i> 97 * (0 <tt><=</tt> <i>m</i> <tt><=</tt> 3, 98 * 0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr> 99 * <tr><td valign="top" headers="construct characters"><tt>\x</tt><i>hh</i></td> 100 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hh</i></td></tr> 101 * <tr><td valign="top" headers="construct characters"><tt>\u</tt><i>hhhh</i></td> 102 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hhhh</i></td></tr> 103 * <tr><td valign="top" headers="matches"><tt>\t</tt></td> 104 * <td headers="matches">The tab character (<tt>'\u0009'</tt>)</td></tr> 105 * <tr><td valign="top" headers="construct characters"><tt>\n</tt></td> 106 * <td headers="matches">The newline (line feed) character (<tt>'\u000A'</tt>)</td></tr> 107 * <tr><td valign="top" headers="construct characters"><tt>\r</tt></td> 108 * <td headers="matches">The carriage-return character (<tt>'\u000D'</tt>)</td></tr> 109 * <tr><td valign="top" headers="construct characters"><tt>\f</tt></td> 110 * <td headers="matches">The form-feed character (<tt>'\u000C'</tt>)</td></tr> 111 * <tr><td valign="top" headers="construct characters"><tt>\a</tt></td> 112 * <td headers="matches">The alert (bell) character (<tt>'\u0007'</tt>)</td></tr> 113 * <tr><td valign="top" headers="construct characters"><tt>\e</tt></td> 114 * <td headers="matches">The escape character (<tt>'\u001B'</tt>)</td></tr> 115 * <tr><td valign="top" headers="construct characters"><tt>\c</tt><i>x</i></td> 116 * <td headers="matches">The control character corresponding to <i>x</i></td></tr> 117 * 118 * <tr><th> </th></tr> 119 * <tr align="left"><th colspan="2" id="classes">Character classes</th></tr> 120 * 121 * <tr><td valign="top" headers="construct classes"><tt>[abc]</tt></td> 122 * <td headers="matches"><tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (simple class)</td></tr> 123 * <tr><td valign="top" headers="construct classes"><tt>[^abc]</tt></td> 124 * <td headers="matches">Any character except <tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (negation)</td></tr> 125 * <tr><td valign="top" headers="construct classes"><tt>[a-zA-Z]</tt></td> 126 * <td headers="matches"><tt>a</tt> through <tt>z</tt> 127 * or <tt>A</tt> through <tt>Z</tt>, inclusive (range)</td></tr> 128 * <tr><td valign="top" headers="construct classes"><tt>[a-d[m-p]]</tt></td> 129 * <td headers="matches"><tt>a</tt> through <tt>d</tt>, 130 * or <tt>m</tt> through <tt>p</tt>: <tt>[a-dm-p]</tt> (union)</td></tr> 131 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[def]]</tt></td> 132 * <td headers="matches"><tt>d</tt>, <tt>e</tt>, or <tt>f</tt> (intersection)</tr> 133 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^bc]]</tt></td> 134 * <td headers="matches"><tt>a</tt> through <tt>z</tt>, 135 * except for <tt>b</tt> and <tt>c</tt>: <tt>[ad-z]</tt> (subtraction)</td></tr> 136 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^m-p]]</tt></td> 137 * <td headers="matches"><tt>a</tt> through <tt>z</tt>, 138 * and not <tt>m</tt> through <tt>p</tt>: <tt>[a-lq-z]</tt>(subtraction)</td></tr> 139 * <tr><th> </th></tr> 140 * 141 * <tr align="left"><th colspan="2" id="predef">Predefined character classes</th></tr> 142 * 143 * <tr><td valign="top" headers="construct predef"><tt>.</tt></td> 144 * <td headers="matches">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr> 145 * <tr><td valign="top" headers="construct predef"><tt>\d</tt></td> 146 * <td headers="matches">A digit: <tt>[0-9]</tt></td></tr> 147 * <tr><td valign="top" headers="construct predef"><tt>\D</tt></td> 148 * <td headers="matches">A non-digit: <tt>[^0-9]</tt></td></tr> 149 * <tr><td valign="top" headers="construct predef"><tt>\s</tt></td> 150 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr> 151 * <tr><td valign="top" headers="construct predef"><tt>\S</tt></td> 152 * <td headers="matches">A non-whitespace character: <tt>[^\s]</tt></td></tr> 153 * <tr><td valign="top" headers="construct predef"><tt>\w</tt></td> 154 * <td headers="matches">A word character: <tt>[a-zA-Z_0-9]</tt></td></tr> 155 * <tr><td valign="top" headers="construct predef"><tt>\W</tt></td> 156 * <td headers="matches">A non-word character: <tt>[^\w]</tt></td></tr> 157 * 158 * <tr><th> </th></tr> 159 * <tr align="left"><th colspan="2" id="posix">POSIX character classes</b> (US-ASCII only)<b></th></tr> 160 * 161 * <tr><td valign="top" headers="construct posix"><tt>\p{Lower}</tt></td> 162 * <td headers="matches">A lower-case alphabetic character: <tt>[a-z]</tt></td></tr> 163 * <tr><td valign="top" headers="construct posix"><tt>\p{Upper}</tt></td> 164 * <td headers="matches">An upper-case alphabetic character:<tt>[A-Z]</tt></td></tr> 165 * <tr><td valign="top" headers="construct posix"><tt>\p{ASCII}</tt></td> 166 * <td headers="matches">All ASCII:<tt>[\x00-\x7F]</tt></td></tr> 167 * <tr><td valign="top" headers="construct posix"><tt>\p{Alpha}</tt></td> 168 * <td headers="matches">An alphabetic character:<tt>[\p{Lower}\p{Upper}]</tt></td></tr> 169 * <tr><td valign="top" headers="construct posix"><tt>\p{Digit}</tt></td> 170 * <td headers="matches">A decimal digit: <tt>[0-9]</tt></td></tr> 171 * <tr><td valign="top" headers="construct posix"><tt>\p{Alnum}</tt></td> 172 * <td headers="matches">An alphanumeric character:<tt>[\p{Alpha}\p{Digit}]</tt></td></tr> 173 * <tr><td valign="top" headers="construct posix"><tt>\p{Punct}</tt></td> 174 * <td headers="matches">Punctuation: One of <tt>!"#$%&'()*+,-./:;<=>?@[\]^_`{|}~</tt></td></tr> 175 * <!-- <tt>[\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]</tt> 176 * <tt>[\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]</tt> --> 177 * <tr><td valign="top" headers="construct posix"><tt>\p{Graph}</tt></td> 178 * <td headers="matches">A visible character: <tt>[\p{Alnum}\p{Punct}]</tt></td></tr> 179 * <tr><td valign="top" headers="construct posix"><tt>\p{Print}</tt></td> 180 * <td headers="matches">A printable character: <tt>[\p{Graph}\x20]</tt></td></tr> 181 * <tr><td valign="top" headers="construct posix"><tt>\p{Blank}</tt></td> 182 * <td headers="matches">A space or a tab: <tt>[ \t]</tt></td></tr> 183 * <tr><td valign="top" headers="construct posix"><tt>\p{Cntrl}</tt></td> 184 * <td headers="matches">A control character: <tt>[\x00-\x1F\x7F]</tt></td></tr> 185 * <tr><td valign="top" headers="construct posix"><tt>\p{XDigit}</tt></td> 186 * <td headers="matches">A hexadecimal digit: <tt>[0-9a-fA-F]</tt></td></tr> 187 * <tr><td valign="top" headers="construct posix"><tt>\p{Space}</tt></td> 188 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr> 189 * 190 * <tr><th> </th></tr> 191 * <tr align="left"><th colspan="2">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr> 192 * 193 * <tr><td valign="top"><tt>\p{javaLowerCase}</tt></td> 194 * <td>Equivalent to java.lang.Character.isLowerCase()</td></tr> 195 * <tr><td valign="top"><tt>\p{javaUpperCase}</tt></td> 196 * <td>Equivalent to java.lang.Character.isUpperCase()</td></tr> 197 * <tr><td valign="top"><tt>\p{javaWhitespace}</tt></td> 198 * <td>Equivalent to java.lang.Character.isWhitespace()</td></tr> 199 * <tr><td valign="top"><tt>\p{javaMirrored}</tt></td> 200 * <td>Equivalent to java.lang.Character.isMirrored()</td></tr> 201 * 202 * <tr><th> </th></tr> 203 * <tr align="left"><th colspan="2" id="unicode">Classes for Unicode blocks and categories</th></tr> 204 * 205 * <tr><td valign="top" headers="construct unicode"><tt>\p{InGreek}</tt></td> 206 * <td headers="matches">A character in the Greek block (simple <a href="#ubc">block</a>)</td></tr> 207 * <tr><td valign="top" headers="construct unicode"><tt>\p{Lu}</tt></td> 208 * <td headers="matches">An uppercase letter (simple <a href="#ubc">category</a>)</td></tr> 209 * <tr><td valign="top" headers="construct unicode"><tt>\p{Sc}</tt></td> 210 * <td headers="matches">A currency symbol</td></tr> 211 * <tr><td valign="top" headers="construct unicode"><tt>\P{InGreek}</tt></td> 212 * <td headers="matches">Any character except one in the Greek block (negation)</td></tr> 213 * <tr><td valign="top" headers="construct unicode"><tt>[\p{L}&&[^\p{Lu}]] </tt></td> 214 * <td headers="matches">Any letter except an uppercase letter (subtraction)</td></tr> 215 * 216 * <tr><th> </th></tr> 217 * <tr align="left"><th colspan="2" id="bounds">Boundary matchers</th></tr> 218 * 219 * <tr><td valign="top" headers="construct bounds"><tt>^</tt></td> 220 * <td headers="matches">The beginning of a line</td></tr> 221 * <tr><td valign="top" headers="construct bounds"><tt>$</tt></td> 222 * <td headers="matches">The end of a line</td></tr> 223 * <tr><td valign="top" headers="construct bounds"><tt>\b</tt></td> 224 * <td headers="matches">A word boundary</td></tr> 225 * <tr><td valign="top" headers="construct bounds"><tt>\B</tt></td> 226 * <td headers="matches">A non-word boundary</td></tr> 227 * <tr><td valign="top" headers="construct bounds"><tt>\A</tt></td> 228 * <td headers="matches">The beginning of the input</td></tr> 229 * <tr><td valign="top" headers="construct bounds"><tt>\G</tt></td> 230 * <td headers="matches">The end of the previous match</td></tr> 231 * <tr><td valign="top" headers="construct bounds"><tt>\Z</tt></td> 232 * <td headers="matches">The end of the input but for the final 233 * <a href="#lt">terminator</a>, if any</td></tr> 234 * <tr><td valign="top" headers="construct bounds"><tt>\z</tt></td> 235 * <td headers="matches">The end of the input</td></tr> 236 * 237 * <tr><th> </th></tr> 238 * <tr align="left"><th colspan="2" id="greedy">Greedy quantifiers</th></tr> 239 * 240 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>?</tt></td> 241 * <td headers="matches"><i>X</i>, once or not at all</td></tr> 242 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>*</tt></td> 243 * <td headers="matches"><i>X</i>, zero or more times</td></tr> 244 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>+</tt></td> 245 * <td headers="matches"><i>X</i>, one or more times</td></tr> 246 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>}</tt></td> 247 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr> 248 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>,}</tt></td> 249 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr> 250 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}</tt></td> 251 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 252 * 253 * <tr><th> </th></tr> 254 * <tr align="left"><th colspan="2" id="reluc">Reluctant quantifiers</th></tr> 255 * 256 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>??</tt></td> 257 * <td headers="matches"><i>X</i>, once or not at all</td></tr> 258 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>*?</tt></td> 259 * <td headers="matches"><i>X</i>, zero or more times</td></tr> 260 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>+?</tt></td> 261 * <td headers="matches"><i>X</i>, one or more times</td></tr> 262 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>}?</tt></td> 263 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr> 264 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>,}?</tt></td> 265 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr> 266 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}?</tt></td> 267 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 268 * 269 * <tr><th> </th></tr> 270 * <tr align="left"><th colspan="2" id="poss">Possessive quantifiers</th></tr> 271 * 272 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>?+</tt></td> 273 * <td headers="matches"><i>X</i>, once or not at all</td></tr> 274 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>*+</tt></td> 275 * <td headers="matches"><i>X</i>, zero or more times</td></tr> 276 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>++</tt></td> 277 * <td headers="matches"><i>X</i>, one or more times</td></tr> 278 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>}+</tt></td> 279 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr> 280 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>,}+</tt></td> 281 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr> 282 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}+</tt></td> 283 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 284 * 285 * <tr><th> </th></tr> 286 * <tr align="left"><th colspan="2" id="logical">Logical operators</th></tr> 287 * 288 * <tr><td valign="top" headers="construct logical"><i>XY</i></td> 289 * <td headers="matches"><i>X</i> followed by <i>Y</i></td></tr> 290 * <tr><td valign="top" headers="construct logical"><i>X</i><tt>|</tt><i>Y</i></td> 291 * <td headers="matches">Either <i>X</i> or <i>Y</i></td></tr> 292 * <tr><td valign="top" headers="construct logical"><tt>(</tt><i>X</i><tt>)</tt></td> 293 * <td headers="matches">X, as a <a href="#cg">capturing group</a></td></tr> 294 * 295 * <tr><th> </th></tr> 296 * <tr align="left"><th colspan="2" id="backref">Back references</th></tr> 297 * 298 * <tr><td valign="bottom" headers="construct backref"><tt>\</tt><i>n</i></td> 299 * <td valign="bottom" headers="matches">Whatever the <i>n</i><sup>th</sup> 300 * <a href="#cg">capturing group</a> matched</td></tr> 301 * 302 * <tr><td valign="bottom" headers="construct backref"><tt>\</tt><i>k</i><<i>name</i>></td> 303 * <td valign="bottom" headers="matches">Whatever the 304 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr> 305 * 306 * <tr><th> </th></tr> 307 * <tr align="left"><th colspan="2" id="quot">Quotation</th></tr> 308 * 309 * <tr><td valign="top" headers="construct quot"><tt>\</tt></td> 310 * <td headers="matches">Nothing, but quotes the following character</td></tr> 311 * <tr><td valign="top" headers="construct quot"><tt>\Q</tt></td> 312 * <td headers="matches">Nothing, but quotes all characters until <tt>\E</tt></td></tr> 313 * <tr><td valign="top" headers="construct quot"><tt>\E</tt></td> 314 * <td headers="matches">Nothing, but ends quoting started by <tt>\Q</tt></td></tr> 315 * <!-- Metachars: !$()*+.<>?[\]^{|} --> 316 * 317 * <tr><th> </th></tr> 318 * <tr align="left"><th colspan="2" id="special">Special constructs (named-capturing and non-capturing)</th></tr> 319 * 320 * <tr><td valign="top" headers="construct special"><tt>(?<<a href="#groupname">name</a>></tt><i>X</i><tt>)</tt></td> 321 * <td headers="matches"><i>X</i>, as a named-capturing group</td></tr> 322 * <tr><td valign="top" headers="construct special"><tt>(?:</tt><i>X</i><tt>)</tt></td> 323 * <td headers="matches"><i>X</i>, as a non-capturing group</td></tr> 324 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux) </tt></td> 325 * <td headers="matches">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a> 326 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> 327 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> on - off</td></tr> 328 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux:</tt><i>X</i><tt>)</tt> </td> 329 * <td headers="matches"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the 330 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a> 331 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a > 332 * <a href="#COMMENTS">x</a> on - off</td></tr> 333 * <tr><td valign="top" headers="construct special"><tt>(?=</tt><i>X</i><tt>)</tt></td> 334 * <td headers="matches"><i>X</i>, via zero-width positive lookahead</td></tr> 335 * <tr><td valign="top" headers="construct special"><tt>(?!</tt><i>X</i><tt>)</tt></td> 336 * <td headers="matches"><i>X</i>, via zero-width negative lookahead</td></tr> 337 * <tr><td valign="top" headers="construct special"><tt>(?<=</tt><i>X</i><tt>)</tt></td> 338 * <td headers="matches"><i>X</i>, via zero-width positive lookbehind</td></tr> 339 * <tr><td valign="top" headers="construct special"><tt>(?<!</tt><i>X</i><tt>)</tt></td> 340 * <td headers="matches"><i>X</i>, via zero-width negative lookbehind</td></tr> 341 * <tr><td valign="top" headers="construct special"><tt>(?></tt><i>X</i><tt>)</tt></td> 342 * <td headers="matches"><i>X</i>, as an independent, non-capturing group</td></tr> 343 * 344 * </table> 345 * 346 * <hr> 347 * 348 * 349 * <a name="bs"> 350 * <h4> Backslashes, escapes, and quoting </h4> 351 * 352 * <p> The backslash character (<tt>'\'</tt>) serves to introduce escaped 353 * constructs, as defined in the table above, as well as to quote characters 354 * that otherwise would be interpreted as unescaped constructs. Thus the 355 * expression <tt>\\</tt> matches a single backslash and <tt>\{</tt> matches a 356 * left brace. 357 * 358 * <p> It is an error to use a backslash prior to any alphabetic character that 359 * does not denote an escaped construct; these are reserved for future 360 * extensions to the regular-expression language. A backslash may be used 361 * prior to a non-alphabetic character regardless of whether that character is 362 * part of an unescaped construct. 363 * 364 * <p> Backslashes within string literals in Java source code are interpreted 365 * as required by the <a 366 * href="http://java.sun.com/docs/books/jls">Java Language 367 * Specification</a> as either <a 368 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">Unicode 369 * escapes</a> or other <a 370 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#101089">character 371 * escapes</a>. It is therefore necessary to double backslashes in string 372 * literals that represent regular expressions to protect them from 373 * interpretation by the Java bytecode compiler. The string literal 374 * <tt>"\b"</tt>, for example, matches a single backspace character when 375 * interpreted as a regular expression, while <tt>"\\b"</tt> matches a 376 * word boundary. The string literal <tt>"\(hello\)"</tt> is illegal 377 * and leads to a compile-time error; in order to match the string 378 * <tt>(hello)</tt> the string literal <tt>"\\(hello\\)"</tt> 379 * must be used. 380 * 381 * <a name="cc"> 382 * <h4> Character Classes </h4> 383 * 384 * <p> Character classes may appear within other character classes, and 385 * may be composed by the union operator (implicit) and the intersection 386 * operator (<tt>&&</tt>). 387 * The union operator denotes a class that contains every character that is 388 * in at least one of its operand classes. The intersection operator 389 * denotes a class that contains every character that is in both of its 390 * operand classes. 391 * 392 * <p> The precedence of character-class operators is as follows, from 393 * highest to lowest: 394 * 395 * <blockquote><table border="0" cellpadding="1" cellspacing="0" 396 * summary="Precedence of character class operators."> 397 * <tr><th>1 </th> 398 * <td>Literal escape </td> 399 * <td><tt>\x</tt></td></tr> 400 * <tr><th>2 </th> 401 * <td>Grouping</td> 402 * <td><tt>[...]</tt></td></tr> 403 * <tr><th>3 </th> 404 * <td>Range</td> 405 * <td><tt>a-z</tt></td></tr> 406 * <tr><th>4 </th> 407 * <td>Union</td> 408 * <td><tt>[a-e][i-u]</tt></td></tr> 409 * <tr><th>5 </th> 410 * <td>Intersection</td> 411 * <td><tt>[a-z&&[aeiou]]</tt></td></tr> 412 * </table></blockquote> 413 * 414 * <p> Note that a different set of metacharacters are in effect inside 415 * a character class than outside a character class. For instance, the 416 * regular expression <tt>.</tt> loses its special meaning inside a 417 * character class, while the expression <tt>-</tt> becomes a range 418 * forming metacharacter. 419 * 420 * <a name="lt"> 421 * <h4> Line terminators </h4> 422 * 423 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks 424 * the end of a line of the input character sequence. The following are 425 * recognized as line terminators: 426 * 427 * <ul> 428 * 429 * <li> A newline (line feed) character (<tt>'\n'</tt>), 430 * 431 * <li> A carriage-return character followed immediately by a newline 432 * character (<tt>"\r\n"</tt>), 433 * 434 * <li> A standalone carriage-return character (<tt>'\r'</tt>), 435 * 436 * <li> A next-line character (<tt>'\u0085'</tt>), 437 * 438 * <li> A line-separator character (<tt>'\u2028'</tt>), or 439 * 440 * <li> A paragraph-separator character (<tt>'\u2029</tt>). 441 * 442 * </ul> 443 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators 444 * recognized are newline characters. 445 * 446 * <p> The regular expression <tt>.</tt> matches any character except a line 447 * terminator unless the {@link #DOTALL} flag is specified. 448 * 449 * <p> By default, the regular expressions <tt>^</tt> and <tt>$</tt> ignore 450 * line terminators and only match at the beginning and the end, respectively, 451 * of the entire input sequence. If {@link #MULTILINE} mode is activated then 452 * <tt>^</tt> matches at the beginning of input and after any line terminator 453 * except at the end of input. When in {@link #MULTILINE} mode <tt>$</tt> 454 * matches just before a line terminator or the end of the input sequence. 455 * 456 * <a name="cg"> 457 * <h4> Groups and capturing </h4> 458 * 459 * <a name="gnumber"> 460 * <h5> Group number </h5> 461 * <p> Capturing groups are numbered by counting their opening parentheses from 462 * left to right. In the expression <tt>((A)(B(C)))</tt>, for example, there 463 * are four such groups: </p> 464 * 465 * <blockquote><table cellpadding=1 cellspacing=0 summary="Capturing group numberings"> 466 * <tr><th>1 </th> 467 * <td><tt>((A)(B(C)))</tt></td></tr> 468 * <tr><th>2 </th> 469 * <td><tt>(A)</tt></td></tr> 470 * <tr><th>3 </th> 471 * <td><tt>(B(C))</tt></td></tr> 472 * <tr><th>4 </th> 473 * <td><tt>(C)</tt></td></tr> 474 * </table></blockquote> 475 * 476 * <p> Group zero always stands for the entire expression. 477 * 478 * <p> Capturing groups are so named because, during a match, each subsequence 479 * of the input sequence that matches such a group is saved. The captured 480 * subsequence may be used later in the expression, via a back reference, and 481 * may also be retrieved from the matcher once the match operation is complete. 482 * 483 * <a name="groupname"> 484 * <h5> Group name </h5> 485 * <p>A capturing group can also be assigned a "name", a <tt>named-capturing group</tt>, 486 * and then be back-referenced later by the "name". Group names are composed of 487 * the following characters: 488 * 489 * <ul> 490 * <li> The uppercase letters <tt>'A'</tt> through <tt>'Z'</tt> 491 * (<tt>'\u0041'</tt> through <tt>'\u005a'</tt>), 492 * <li> The lowercase letters <tt>'a'</tt> through <tt>'z'</tt> 493 * (<tt>'\u0061'</tt> through <tt>'\u007a'</tt>), 494 * <li> The digits <tt>'0'</tt> through <tt>'9'</tt> 495 * (<tt>'\u0030'</tt> through <tt>'\u0039'</tt>), 496 * </ul> 497 * 498 * <p> A <tt>named-capturing group</tt> is still numbered as described in 499 * <a href="#gnumber">Group number</a>. 500 * 501 * <p> The captured input associated with a group is always the subsequence 502 * that the group most recently matched. If a group is evaluated a second time 503 * because of quantification then its previously-captured value, if any, will 504 * be retained if the second evaluation fails. Matching the string 505 * <tt>"aba"</tt> against the expression <tt>(a(b)?)+</tt>, for example, leaves 506 * group two set to <tt>"b"</tt>. All captured input is discarded at the 507 * beginning of each match. 508 * 509 * <p> Groups beginning with <tt>(?</tt> are either pure, <i>non-capturing</i> groups 510 * that do not capture text and do not count towards the group total, or 511 * <i>named-capturing</i> group. 512 * 513 * <h4> Unicode support </h4> 514 * 515 * <p> This class is in conformance with Level 1 of <a 516 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 517 * Standard #18: Unicode Regular Expression Guidelines</i></a>, plus RL2.1 518 * Canonical Equivalents. 519 * 520 * <p> Unicode escape sequences such as <tt>\u2014</tt> in Java source code 521 * are processed as described in <a 522 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">\u00A73.3</a> 523 * of the Java Language Specification. Such escape sequences are also 524 * implemented directly by the regular-expression parser so that Unicode 525 * escapes can be used in expressions that are read from files or from the 526 * keyboard. Thus the strings <tt>"\u2014"</tt> and <tt>"\\u2014"</tt>, 527 * while not equal, compile into the same pattern, which matches the character 528 * with hexadecimal value <tt>0x2014</tt>. 529 * 530 * <a name="ubc"> <p>Unicode blocks and categories are written with the 531 * <tt>\p</tt> and <tt>\P</tt> constructs as in 532 * Perl. <tt>\p{</tt><i>prop</i><tt>}</tt> matches if the input has the 533 * property <i>prop</i>, while <tt>\P{</tt><i>prop</i><tt>}</tt> does not match if 534 * the input has that property. Blocks are specified with the prefix 535 * <tt>In</tt>, as in <tt>InMongolian</tt>. Categories may be specified with 536 * the optional prefix <tt>Is</tt>: Both <tt>\p{L}</tt> and <tt>\p{IsL}</tt> 537 * denote the category of Unicode letters. Blocks and categories can be used 538 * both inside and outside of a character class. 539 * 540 * <p> The supported categories are those of 541 * <a href="http://www.unicode.org/unicode/standard/standard.html"> 542 * <i>The Unicode Standard</i></a> in the version specified by the 543 * {@link java.lang.Character Character} class. The category names are those 544 * defined in the Standard, both normative and informative. 545 * The block names supported by <code>Pattern</code> are the valid block names 546 * accepted and defined by 547 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}. 548 * 549 * <a name="jcc"> <p>Categories that behave like the java.lang.Character 550 * boolean is<i>methodname</i> methods (except for the deprecated ones) are 551 * available through the same <tt>\p{</tt><i>prop</i><tt>}</tt> syntax where 552 * the specified property has the name <tt>java<i>methodname</i></tt>. 553 * 554 * <h4> Comparison to Perl 5 </h4> 555 * 556 * <p>The <code>Pattern</code> engine performs traditional NFA-based matching 557 * with ordered alternation as occurs in Perl 5. 558 * 559 * <p> Perl constructs not supported by this class: </p> 560 * 561 * <ul> 562 * 563 * <li><p> The conditional constructs <tt>(?{</tt><i>X</i><tt>})</tt> and 564 * <tt>(?(</tt><i>condition</i><tt>)</tt><i>X</i><tt>|</tt><i>Y</i><tt>)</tt>, 565 * </p></li> 566 * 567 * <li><p> The embedded code constructs <tt>(?{</tt><i>code</i><tt>})</tt> 568 * and <tt>(??{</tt><i>code</i><tt>})</tt>,</p></li> 569 * 570 * <li><p> The embedded comment syntax <tt>(?#comment)</tt>, and </p></li> 571 * 572 * <li><p> The preprocessing operations <tt>\l</tt> <tt>\u</tt>, 573 * <tt>\L</tt>, and <tt>\U</tt>. </p></li> 574 * 575 * </ul> 576 * 577 * <p> Constructs supported by this class but not by Perl: </p> 578 * 579 * <ul> 580 * 581 * <li><p> Possessive quantifiers, which greedily match as much as they can 582 * and do not back off, even when doing so would allow the overall match to 583 * succeed. </p></li> 584 * 585 * <li><p> Character-class union and intersection as described 586 * <a href="#cc">above</a>.</p></li> 587 * 588 * </ul> 589 * 590 * <p> Notable differences from Perl: </p> 591 * 592 * <ul> 593 * 594 * <li><p> In Perl, <tt>\1</tt> through <tt>\9</tt> are always interpreted 595 * as back references; a backslash-escaped number greater than <tt>9</tt> is 596 * treated as a back reference if at least that many subexpressions exist, 597 * otherwise it is interpreted, if possible, as an octal escape. In this 598 * class octal escapes must always begin with a zero. In this class, 599 * <tt>\1</tt> through <tt>\9</tt> are always interpreted as back 600 * references, and a larger number is accepted as a back reference if at 601 * least that many subexpressions exist at that point in the regular 602 * expression, otherwise the parser will drop digits until the number is 603 * smaller or equal to the existing number of groups or it is one digit. 604 * </p></li> 605 * 606 * <li><p> Perl uses the <tt>g</tt> flag to request a match that resumes 607 * where the last match left off. This functionality is provided implicitly 608 * by the {@link Matcher} class: Repeated invocations of the {@link 609 * Matcher#find find} method will resume where the last match left off, 610 * unless the matcher is reset. </p></li> 611 * 612 * <li><p> In Perl, embedded flags at the top level of an expression affect 613 * the whole expression. In this class, embedded flags always take effect 614 * at the point at which they appear, whether they are at the top level or 615 * within a group; in the latter case, flags are restored at the end of the 616 * group just as in Perl. </p></li> 617 * 618 * <li><p> Perl is forgiving about malformed matching constructs, as in the 619 * expression <tt>*a</tt>, as well as dangling brackets, as in the 620 * expression <tt>abc]</tt>, and treats them as literals. This 621 * class also accepts dangling brackets but is strict about dangling 622 * metacharacters like +, ? and *, and will throw a 623 * {@link PatternSyntaxException} if it encounters them. </p></li> 624 * 625 * </ul> 626 * 627 * 628 * <p> For a more precise description of the behavior of regular expression 629 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/"> 630 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl, 631 * O'Reilly and Associates, 2006.</a> 632 * </p> 633 * 634 * @see java.lang.String#split(String, int) 635 * @see java.lang.String#split(String) 636 * 637 * @author Mike McCloskey 638 * @author Mark Reinhold 639 * @author JSR-51 Expert Group 640 * @since 1.4 641 * @spec JSR-51 642 */ 643 644 public final class Pattern 645 implements java.io.Serializable 646 { 647 648 /** 649 * Regular expression modifier values. Instead of being passed as 650 * arguments, they can also be passed as inline modifiers. 651 * For example, the following statements have the same effect. 652 * <pre> 653 * RegExp r1 = RegExp.compile("abc", Pattern.I|Pattern.M); 654 * RegExp r2 = RegExp.compile("(?im)abc", 0); 655 * </pre> 656 * 657 * The flags are duplicated so that the familiar Perl match flag 658 * names are available. 659 */ 660 661 /** 662 * Enables Unix lines mode. 663 * 664 * <p> In this mode, only the <tt>'\n'</tt> line terminator is recognized 665 * in the behavior of <tt>.</tt>, <tt>^</tt>, and <tt>$</tt>. 666 * 667 * <p> Unix lines mode can also be enabled via the embedded flag 668 * expression <tt>(?d)</tt>. 669 */ 670 public static final int UNIX_LINES = 0x01; 671 672 /** 673 * Enables case-insensitive matching. 674 * 675 * <p> By default, case-insensitive matching assumes that only characters 676 * in the US-ASCII charset are being matched. Unicode-aware 677 * case-insensitive matching can be enabled by specifying the {@link 678 * #UNICODE_CASE} flag in conjunction with this flag. 679 * 680 * <p> Case-insensitive matching can also be enabled via the embedded flag 681 * expression <tt>(?i)</tt>. 682 * 683 * <p> Specifying this flag may impose a slight performance penalty. </p> 684 */ 685 public static final int CASE_INSENSITIVE = 0x02; 686 687 /** 688 * Permits whitespace and comments in pattern. 689 * 690 * <p> In this mode, whitespace is ignored, and embedded comments starting 691 * with <tt>#</tt> are ignored until the end of a line. 692 * 693 * <p> Comments mode can also be enabled via the embedded flag 694 * expression <tt>(?x)</tt>. 695 */ 696 public static final int COMMENTS = 0x04; 697 698 /** 699 * Enables multiline mode. 700 * 701 * <p> In multiline mode the expressions <tt>^</tt> and <tt>$</tt> match 702 * just after or just before, respectively, a line terminator or the end of 703 * the input sequence. By default these expressions only match at the 704 * beginning and the end of the entire input sequence. 705 * 706 * <p> Multiline mode can also be enabled via the embedded flag 707 * expression <tt>(?m)</tt>. </p> 708 */ 709 public static final int MULTILINE = 0x08; 710 711 /** 712 * Enables literal parsing of the pattern. 713 * 714 * <p> When this flag is specified then the input string that specifies 715 * the pattern is treated as a sequence of literal characters. 716 * Metacharacters or escape sequences in the input sequence will be 717 * given no special meaning. 718 * 719 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on 720 * matching when used in conjunction with this flag. The other flags 721 * become superfluous. 722 * 723 * <p> There is no embedded flag character for enabling literal parsing. 724 * @since 1.5 725 */ 726 public static final int LITERAL = 0x10; 727 728 /** 729 * Enables dotall mode. 730 * 731 * <p> In dotall mode, the expression <tt>.</tt> matches any character, 732 * including a line terminator. By default this expression does not match 733 * line terminators. 734 * 735 * <p> Dotall mode can also be enabled via the embedded flag 736 * expression <tt>(?s)</tt>. (The <tt>s</tt> is a mnemonic for 737 * "single-line" mode, which is what this is called in Perl.) </p> 738 */ 739 public static final int DOTALL = 0x20; 740 741 /** 742 * Enables Unicode-aware case folding. 743 * 744 * <p> When this flag is specified then case-insensitive matching, when 745 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner 746 * consistent with the Unicode Standard. By default, case-insensitive 747 * matching assumes that only characters in the US-ASCII charset are being 748 * matched. 749 * 750 * <p> Unicode-aware case folding can also be enabled via the embedded flag 751 * expression <tt>(?u)</tt>. 752 * 753 * <p> Specifying this flag may impose a performance penalty. </p> 754 */ 755 public static final int UNICODE_CASE = 0x40; 756 757 /** 758 * Enables canonical equivalence. 759 * 760 * <p> When this flag is specified then two characters will be considered 761 * to match if, and only if, their full canonical decompositions match. 762 * The expression <tt>"a\u030A"</tt>, for example, will match the 763 * string <tt>"\u00E5"</tt> when this flag is specified. By default, 764 * matching does not take canonical equivalence into account. 765 * 766 * <p> There is no embedded flag character for enabling canonical 767 * equivalence. 768 * 769 * <p> Specifying this flag may impose a performance penalty. </p> 770 */ 771 public static final int CANON_EQ = 0x80; 772 773 /* Pattern has only two serialized components: The pattern string 774 * and the flags, which are all that is needed to recompile the pattern 775 * when it is deserialized. 776 */ 777 778 /** use serialVersionUID from Merlin b59 for interoperability */ 779 private static final long serialVersionUID = 5073258162644648461L; 780 781 /** 782 * The original regular-expression pattern string. 783 * 784 * @serial 785 */ 786 private String pattern; 787 788 /** 789 * The original pattern flags. 790 * 791 * @serial 792 */ 793 private int flags; 794 795 /** 796 * Boolean indicating this Pattern is compiled; this is necessary in order 797 * to lazily compile deserialized Patterns. 798 */ 799 private transient volatile boolean compiled = false; 800 801 /** 802 * The normalized pattern string. 803 */ 804 private transient String normalizedPattern; 805 806 /** 807 * The starting point of state machine for the find operation. This allows 808 * a match to start anywhere in the input. 809 */ 810 transient Node root; 811 812 /** 813 * The root of object tree for a match operation. The pattern is matched 814 * at the beginning. This may include a find that uses BnM or a First 815 * node. 816 */ 817 transient Node matchRoot; 818 819 /** 820 * Temporary storage used by parsing pattern slice. 821 */ 822 transient int[] buffer; 823 824 /** 825 * Map the "name" of the "named capturing group" to its group id 826 * node. 827 */ 828 transient volatile Map<String, Integer> namedGroups; 829 830 /** 831 * Temporary storage used while parsing group references. 832 */ 833 transient GroupHead[] groupNodes; 834 835 /** 836 * Temporary null terminated code point array used by pattern compiling. 837 */ 838 private transient int[] temp; 839 840 /** 841 * The number of capturing groups in this Pattern. Used by matchers to 842 * allocate storage needed to perform a match. 843 */ 844 transient int capturingGroupCount; 845 846 /** 847 * The local variable count used by parsing tree. Used by matchers to 848 * allocate storage needed to perform a match. 849 */ 850 transient int localCount; 851 852 /** 853 * Index into the pattern string that keeps track of how much has been 854 * parsed. 855 */ 856 private transient int cursor; 857 858 /** 859 * Holds the length of the pattern string. 860 */ 861 private transient int patternLength; 862 863 /** 864 * Compiles the given regular expression into a pattern. </p> 865 * 866 * @param regex 867 * The expression to be compiled 868 * 869 * @throws PatternSyntaxException 870 * If the expression's syntax is invalid 871 */ 872 public static Pattern compile(String regex) { 873 return new Pattern(regex, 0); 874 } 875 876 /** 877 * Compiles the given regular expression into a pattern with the given 878 * flags. </p> 879 * 880 * @param regex 881 * The expression to be compiled 882 * 883 * @param flags 884 * Match flags, a bit mask that may include 885 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL}, 886 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES}, 887 * {@link #LITERAL} and {@link #COMMENTS} 888 * 889 * @throws IllegalArgumentException 890 * If bit values other than those corresponding to the defined 891 * match flags are set in <tt>flags</tt> 892 * 893 * @throws PatternSyntaxException 894 * If the expression's syntax is invalid 895 */ 896 public static Pattern compile(String regex, int flags) { 897 return new Pattern(regex, flags); 898 } 899 900 /** 901 * Returns the regular expression from which this pattern was compiled. 902 * </p> 903 * 904 * @return The source of this pattern 905 */ 906 public String pattern() { 907 return pattern; 908 } 909 910 /** 911 * <p>Returns the string representation of this pattern. This 912 * is the regular expression from which this pattern was 913 * compiled.</p> 914 * 915 * @return The string representation of this pattern 916 * @since 1.5 917 */ 918 public String toString() { 919 return pattern; 920 } 921 922 /** 923 * Creates a matcher that will match the given input against this pattern. 924 * </p> 925 * 926 * @param input 927 * The character sequence to be matched 928 * 929 * @return A new matcher for this pattern 930 */ 931 public Matcher matcher(CharSequence input) { 932 if (!compiled) { 933 synchronized(this) { 934 if (!compiled) 935 compile(); 936 } 937 } 938 Matcher m = new Matcher(this, input); 939 return m; 940 } 941 942 /** 943 * Returns this pattern's match flags. </p> 944 * 945 * @return The match flags specified when this pattern was compiled 946 */ 947 public int flags() { 948 return flags; 949 } 950 951 /** 952 * Compiles the given regular expression and attempts to match the given 953 * input against it. 954 * 955 * <p> An invocation of this convenience method of the form 956 * 957 * <blockquote><pre> 958 * Pattern.matches(regex, input);</pre></blockquote> 959 * 960 * behaves in exactly the same way as the expression 961 * 962 * <blockquote><pre> 963 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote> 964 * 965 * <p> If a pattern is to be used multiple times, compiling it once and reusing 966 * it will be more efficient than invoking this method each time. </p> 967 * 968 * @param regex 969 * The expression to be compiled 970 * 971 * @param input 972 * The character sequence to be matched 973 * 974 * @throws PatternSyntaxException 975 * If the expression's syntax is invalid 976 */ 977 public static boolean matches(String regex, CharSequence input) { 978 Pattern p = Pattern.compile(regex); 979 Matcher m = p.matcher(input); 980 return m.matches(); 981 } 982 983 /** 984 * Splits the given input sequence around matches of this pattern. 985 * 986 * <p> The array returned by this method contains each substring of the 987 * input sequence that is terminated by another subsequence that matches 988 * this pattern or is terminated by the end of the input sequence. The 989 * substrings in the array are in the order in which they occur in the 990 * input. If this pattern does not match any subsequence of the input then 991 * the resulting array has just one element, namely the input sequence in 992 * string form. 993 * 994 * <p> The <tt>limit</tt> parameter controls the number of times the 995 * pattern is applied and therefore affects the length of the resulting 996 * array. If the limit <i>n</i> is greater than zero then the pattern 997 * will be applied at most <i>n</i> - 1 times, the array's 998 * length will be no greater than <i>n</i>, and the array's last entry 999 * will contain all input beyond the last matched delimiter. If <i>n</i> 1000 * is non-positive then the pattern will be applied as many times as 1001 * possible and the array can have any length. If <i>n</i> is zero then 1002 * the pattern will be applied as many times as possible, the array can 1003 * have any length, and trailing empty strings will be discarded. 1004 * 1005 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following 1006 * results with these parameters: 1007 * 1008 * <blockquote><table cellpadding=1 cellspacing=0 1009 * summary="Split examples showing regex, limit, and result"> 1010 * <tr><th><P align="left"><i>Regex </i></th> 1011 * <th><P align="left"><i>Limit </i></th> 1012 * <th><P align="left"><i>Result </i></th></tr> 1013 * <tr><td align=center>:</td> 1014 * <td align=center>2</td> 1015 * <td><tt>{ "boo", "and:foo" }</tt></td></tr> 1016 * <tr><td align=center>:</td> 1017 * <td align=center>5</td> 1018 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr> 1019 * <tr><td align=center>:</td> 1020 * <td align=center>-2</td> 1021 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr> 1022 * <tr><td align=center>o</td> 1023 * <td align=center>5</td> 1024 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr> 1025 * <tr><td align=center>o</td> 1026 * <td align=center>-2</td> 1027 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr> 1028 * <tr><td align=center>o</td> 1029 * <td align=center>0</td> 1030 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr> 1031 * </table></blockquote> 1032 * 1033 * 1034 * @param input 1035 * The character sequence to be split 1036 * 1037 * @param limit 1038 * The result threshold, as described above 1039 * 1040 * @return The array of strings computed by splitting the input 1041 * around matches of this pattern 1042 */ 1043 public String[] split(CharSequence input, int limit) { 1044 int index = 0; 1045 boolean matchLimited = limit > 0; 1046 ArrayList<String> matchList = new ArrayList<String>(); 1047 Matcher m = matcher(input); 1048 1049 // Add segments before each match found 1050 while(m.find()) { 1051 if (!matchLimited || matchList.size() < limit - 1) { 1052 String match = input.subSequence(index, m.start()).toString(); 1053 matchList.add(match); 1054 index = m.end(); 1055 } else if (matchList.size() == limit - 1) { // last one 1056 String match = input.subSequence(index, 1057 input.length()).toString(); 1058 matchList.add(match); 1059 index = m.end(); 1060 } 1061 } 1062 1063 // If no match was found, return this 1064 if (index == 0) 1065 return new String[] {input.toString()}; 1066 1067 // Add remaining segment 1068 if (!matchLimited || matchList.size() < limit) 1069 matchList.add(input.subSequence(index, input.length()).toString()); 1070 1071 // Construct result 1072 int resultSize = matchList.size(); 1073 if (limit == 0) 1074 while (resultSize > 0 && matchList.get(resultSize-1).equals("")) 1075 resultSize--; 1076 String[] result = new String[resultSize]; 1077 return matchList.subList(0, resultSize).toArray(result); 1078 } 1079 1080 /** 1081 * Splits the given input sequence around matches of this pattern. 1082 * 1083 * <p> This method works as if by invoking the two-argument {@link 1084 * #split(java.lang.CharSequence, int) split} method with the given input 1085 * sequence and a limit argument of zero. Trailing empty strings are 1086 * therefore not included in the resulting array. </p> 1087 * 1088 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following 1089 * results with these expressions: 1090 * 1091 * <blockquote><table cellpadding=1 cellspacing=0 1092 * summary="Split examples showing regex and result"> 1093 * <tr><th><P align="left"><i>Regex </i></th> 1094 * <th><P align="left"><i>Result</i></th></tr> 1095 * <tr><td align=center>:</td> 1096 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr> 1097 * <tr><td align=center>o</td> 1098 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr> 1099 * </table></blockquote> 1100 * 1101 * 1102 * @param input 1103 * The character sequence to be split 1104 * 1105 * @return The array of strings computed by splitting the input 1106 * around matches of this pattern 1107 */ 1108 public String[] split(CharSequence input) { 1109 return split(input, 0); 1110 } 1111 1112 /** 1113 * Returns a literal pattern <code>String</code> for the specified 1114 * <code>String</code>. 1115 * 1116 * <p>This method produces a <code>String</code> that can be used to 1117 * create a <code>Pattern</code> that would match the string 1118 * <code>s</code> as if it were a literal pattern.</p> Metacharacters 1119 * or escape sequences in the input sequence will be given no special 1120 * meaning. 1121 * 1122 * @param s The string to be literalized 1123 * @return A literal string replacement 1124 * @since 1.5 1125 */ 1126 public static String quote(String s) { 1127 int slashEIndex = s.indexOf("\\E"); 1128 if (slashEIndex == -1) 1129 return "\\Q" + s + "\\E"; 1130 1131 StringBuilder sb = new StringBuilder(s.length() * 2); 1132 sb.append("\\Q"); 1133 slashEIndex = 0; 1134 int current = 0; 1135 while ((slashEIndex = s.indexOf("\\E", current)) != -1) { 1136 sb.append(s.substring(current, slashEIndex)); 1137 current = slashEIndex + 2; 1138 sb.append("\\E\\\\E\\Q"); 1139 } 1140 sb.append(s.substring(current, s.length())); 1141 sb.append("\\E"); 1142 return sb.toString(); 1143 } 1144 1145 /** 1146 * Recompile the Pattern instance from a stream. The original pattern 1147 * string is read in and the object tree is recompiled from it. 1148 */ 1149 private void readObject(java.io.ObjectInputStream s) 1150 throws java.io.IOException, ClassNotFoundException { 1151 1152 // Read in all fields 1153 s.defaultReadObject(); 1154 1155 // Initialize counts 1156 capturingGroupCount = 1; 1157 localCount = 0; 1158 1159 // if length > 0, the Pattern is lazily compiled 1160 compiled = false; 1161 if (pattern.length() == 0) { 1162 root = new Start(lastAccept); 1163 matchRoot = lastAccept; 1164 compiled = true; 1165 } 1166 } 1167 1168 /** 1169 * This private constructor is used to create all Patterns. The pattern 1170 * string and match flags are all that is needed to completely describe 1171 * a Pattern. An empty pattern string results in an object tree with 1172 * only a Start node and a LastNode node. 1173 */ 1174 private Pattern(String p, int f) { 1175 pattern = p; 1176 flags = f; 1177 1178 // Reset group index count 1179 capturingGroupCount = 1; 1180 localCount = 0; 1181 1182 if (pattern.length() > 0) { 1183 compile(); 1184 } else { 1185 root = new Start(lastAccept); 1186 matchRoot = lastAccept; 1187 } 1188 } 1189 1190 /** 1191 * The pattern is converted to normalizedD form and then a pure group 1192 * is constructed to match canonical equivalences of the characters. 1193 */ 1194 private void normalize() { 1195 boolean inCharClass = false; 1196 int lastCodePoint = -1; 1197 1198 // Convert pattern into normalizedD form 1199 normalizedPattern = Normalizer.normalize(pattern, Normalizer.Form.NFD); 1200 patternLength = normalizedPattern.length(); 1201 1202 // Modify pattern to match canonical equivalences 1203 StringBuilder newPattern = new StringBuilder(patternLength); 1204 for(int i=0; i<patternLength; ) { 1205 int c = normalizedPattern.codePointAt(i); 1206 StringBuilder sequenceBuffer; 1207 if ((Character.getType(c) == Character.NON_SPACING_MARK) 1208 && (lastCodePoint != -1)) { 1209 sequenceBuffer = new StringBuilder(); 1210 sequenceBuffer.appendCodePoint(lastCodePoint); 1211 sequenceBuffer.appendCodePoint(c); 1212 while(Character.getType(c) == Character.NON_SPACING_MARK) { 1213 i += Character.charCount(c); 1214 if (i >= patternLength) 1215 break; 1216 c = normalizedPattern.codePointAt(i); 1217 sequenceBuffer.appendCodePoint(c); 1218 } 1219 String ea = produceEquivalentAlternation( 1220 sequenceBuffer.toString()); 1221 newPattern.setLength(newPattern.length()-Character.charCount(lastCodePoint)); 1222 newPattern.append("(?:").append(ea).append(")"); 1223 } else if (c == '[' && lastCodePoint != '\\') { 1224 i = normalizeCharClass(newPattern, i); 1225 } else { 1226 newPattern.appendCodePoint(c); 1227 } 1228 lastCodePoint = c; 1229 i += Character.charCount(c); 1230 } 1231 normalizedPattern = newPattern.toString(); 1232 } 1233 1234 /** 1235 * Complete the character class being parsed and add a set 1236 * of alternations to it that will match the canonical equivalences 1237 * of the characters within the class. 1238 */ 1239 private int normalizeCharClass(StringBuilder newPattern, int i) { 1240 StringBuilder charClass = new StringBuilder(); 1241 StringBuilder eq = null; 1242 int lastCodePoint = -1; 1243 String result; 1244 1245 i++; 1246 charClass.append("["); 1247 while(true) { 1248 int c = normalizedPattern.codePointAt(i); 1249 StringBuilder sequenceBuffer; 1250 1251 if (c == ']' && lastCodePoint != '\\') { 1252 charClass.append((char)c); 1253 break; 1254 } else if (Character.getType(c) == Character.NON_SPACING_MARK) { 1255 sequenceBuffer = new StringBuilder(); 1256 sequenceBuffer.appendCodePoint(lastCodePoint); 1257 while(Character.getType(c) == Character.NON_SPACING_MARK) { 1258 sequenceBuffer.appendCodePoint(c); 1259 i += Character.charCount(c); 1260 if (i >= normalizedPattern.length()) 1261 break; 1262 c = normalizedPattern.codePointAt(i); 1263 } 1264 String ea = produceEquivalentAlternation( 1265 sequenceBuffer.toString()); 1266 1267 charClass.setLength(charClass.length()-Character.charCount(lastCodePoint)); 1268 if (eq == null) 1269 eq = new StringBuilder(); 1270 eq.append('|'); 1271 eq.append(ea); 1272 } else { 1273 charClass.appendCodePoint(c); 1274 i++; 1275 } 1276 if (i == normalizedPattern.length()) 1277 throw error("Unclosed character class"); 1278 lastCodePoint = c; 1279 } 1280 1281 if (eq != null) { 1282 result = "(?:"+charClass.toString()+eq.toString()+")"; 1283 } else { 1284 result = charClass.toString(); 1285 } 1286 1287 newPattern.append(result); 1288 return i; 1289 } 1290 1291 /** 1292 * Given a specific sequence composed of a regular character and 1293 * combining marks that follow it, produce the alternation that will 1294 * match all canonical equivalences of that sequence. 1295 */ 1296 private String produceEquivalentAlternation(String source) { 1297 int len = countChars(source, 0, 1); 1298 if (source.length() == len) 1299 // source has one character. 1300 return source; 1301 1302 String base = source.substring(0,len); 1303 String combiningMarks = source.substring(len); 1304 1305 String[] perms = producePermutations(combiningMarks); 1306 StringBuilder result = new StringBuilder(source); 1307 1308 // Add combined permutations 1309 for(int x=0; x<perms.length; x++) { 1310 String next = base + perms[x]; 1311 if (x>0) 1312 result.append("|"+next); 1313 next = composeOneStep(next); 1314 if (next != null) 1315 result.append("|"+produceEquivalentAlternation(next)); 1316 } 1317 return result.toString(); 1318 } 1319 1320 /** 1321 * Returns an array of strings that have all the possible 1322 * permutations of the characters in the input string. 1323 * This is used to get a list of all possible orderings 1324 * of a set of combining marks. Note that some of the permutations 1325 * are invalid because of combining class collisions, and these 1326 * possibilities must be removed because they are not canonically 1327 * equivalent. 1328 */ 1329 private String[] producePermutations(String input) { 1330 if (input.length() == countChars(input, 0, 1)) 1331 return new String[] {input}; 1332 1333 if (input.length() == countChars(input, 0, 2)) { 1334 int c0 = Character.codePointAt(input, 0); 1335 int c1 = Character.codePointAt(input, Character.charCount(c0)); 1336 if (getClass(c1) == getClass(c0)) { 1337 return new String[] {input}; 1338 } 1339 String[] result = new String[2]; 1340 result[0] = input; 1341 StringBuilder sb = new StringBuilder(2); 1342 sb.appendCodePoint(c1); 1343 sb.appendCodePoint(c0); 1344 result[1] = sb.toString(); 1345 return result; 1346 } 1347 1348 int length = 1; 1349 int nCodePoints = countCodePoints(input); 1350 for(int x=1; x<nCodePoints; x++) 1351 length = length * (x+1); 1352 1353 String[] temp = new String[length]; 1354 1355 int combClass[] = new int[nCodePoints]; 1356 for(int x=0, i=0; x<nCodePoints; x++) { 1357 int c = Character.codePointAt(input, i); 1358 combClass[x] = getClass(c); 1359 i += Character.charCount(c); 1360 } 1361 1362 // For each char, take it out and add the permutations 1363 // of the remaining chars 1364 int index = 0; 1365 int len; 1366 // offset maintains the index in code units. 1367 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) { 1368 len = countChars(input, offset, 1); 1369 boolean skip = false; 1370 for(int y=x-1; y>=0; y--) { 1371 if (combClass[y] == combClass[x]) { 1372 continue loop; 1373 } 1374 } 1375 StringBuilder sb = new StringBuilder(input); 1376 String otherChars = sb.delete(offset, offset+len).toString(); 1377 String[] subResult = producePermutations(otherChars); 1378 1379 String prefix = input.substring(offset, offset+len); 1380 for(int y=0; y<subResult.length; y++) 1381 temp[index++] = prefix + subResult[y]; 1382 } 1383 String[] result = new String[index]; 1384 for (int x=0; x<index; x++) 1385 result[x] = temp[x]; 1386 return result; 1387 } 1388 1389 private int getClass(int c) { 1390 return sun.text.Normalizer.getCombiningClass(c); 1391 } 1392 1393 /** 1394 * Attempts to compose input by combining the first character 1395 * with the first combining mark following it. Returns a String 1396 * that is the composition of the leading character with its first 1397 * combining mark followed by the remaining combining marks. Returns 1398 * null if the first two characters cannot be further composed. 1399 */ 1400 private String composeOneStep(String input) { 1401 int len = countChars(input, 0, 2); 1402 String firstTwoCharacters = input.substring(0, len); 1403 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC); 1404 1405 if (result.equals(firstTwoCharacters)) 1406 return null; 1407 else { 1408 String remainder = input.substring(len); 1409 return result + remainder; 1410 } 1411 } 1412 1413 /** 1414 * Preprocess any \Q...\E sequences in `temp', meta-quoting them. 1415 * See the description of `quotemeta' in perlfunc(1). 1416 */ 1417 private void RemoveQEQuoting() { 1418 final int pLen = patternLength; 1419 int i = 0; 1420 while (i < pLen-1) { 1421 if (temp[i] != '\\') 1422 i += 1; 1423 else if (temp[i + 1] != 'Q') 1424 i += 2; 1425 else 1426 break; 1427 } 1428 if (i >= pLen - 1) // No \Q sequence found 1429 return; 1430 int j = i; 1431 i += 2; 1432 int[] newtemp = new int[j + 2*(pLen-i) + 2]; 1433 System.arraycopy(temp, 0, newtemp, 0, j); 1434 1435 boolean inQuote = true; 1436 while (i < pLen) { 1437 int c = temp[i++]; 1438 if (! ASCII.isAscii(c) || ASCII.isAlnum(c)) { 1439 newtemp[j++] = c; 1440 } else if (c != '\\') { 1441 if (inQuote) newtemp[j++] = '\\'; 1442 newtemp[j++] = c; 1443 } else if (inQuote) { 1444 if (temp[i] == 'E') { 1445 i++; 1446 inQuote = false; 1447 } else { 1448 newtemp[j++] = '\\'; 1449 newtemp[j++] = '\\'; 1450 } 1451 } else { 1452 if (temp[i] == 'Q') { 1453 i++; 1454 inQuote = true; 1455 } else { 1456 newtemp[j++] = c; 1457 if (i != pLen) 1458 newtemp[j++] = temp[i++]; 1459 } 1460 } 1461 } 1462 1463 patternLength = j; 1464 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination 1465 } 1466 1467 /** 1468 * Copies regular expression to an int array and invokes the parsing 1469 * of the expression which will create the object tree. 1470 */ 1471 private void compile() { 1472 // Handle canonical equivalences 1473 if (has(CANON_EQ) && !has(LITERAL)) { 1474 normalize(); 1475 } else { 1476 normalizedPattern = pattern; 1477 } 1478 patternLength = normalizedPattern.length(); 1479 1480 // Copy pattern to int array for convenience 1481 // Use double zero to terminate pattern 1482 temp = new int[patternLength + 2]; 1483 1484 boolean hasSupplementary = false; 1485 int c, count = 0; 1486 // Convert all chars into code points 1487 for (int x = 0; x < patternLength; x += Character.charCount(c)) { 1488 c = normalizedPattern.codePointAt(x); 1489 if (isSupplementary(c)) { 1490 hasSupplementary = true; 1491 } 1492 temp[count++] = c; 1493 } 1494 1495 patternLength = count; // patternLength now in code points 1496 1497 if (! has(LITERAL)) 1498 RemoveQEQuoting(); 1499 1500 // Allocate all temporary objects here. 1501 buffer = new int[32]; 1502 groupNodes = new GroupHead[10]; 1503 namedGroups = null; 1504 1505 if (has(LITERAL)) { 1506 // Literal pattern handling 1507 matchRoot = newSlice(temp, patternLength, hasSupplementary); 1508 matchRoot.next = lastAccept; 1509 } else { 1510 // Start recursive descent parsing 1511 matchRoot = expr(lastAccept); 1512 // Check extra pattern characters 1513 if (patternLength != cursor) { 1514 if (peek() == ')') { 1515 throw error("Unmatched closing ')'"); 1516 } else { 1517 throw error("Unexpected internal error"); 1518 } 1519 } 1520 } 1521 1522 // Peephole optimization 1523 if (matchRoot instanceof Slice) { 1524 root = BnM.optimize(matchRoot); 1525 if (root == matchRoot) { 1526 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1527 } 1528 } else if (matchRoot instanceof Begin || matchRoot instanceof First) { 1529 root = matchRoot; 1530 } else { 1531 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1532 } 1533 1534 // Release temporary storage 1535 temp = null; 1536 buffer = null; 1537 groupNodes = null; 1538 patternLength = 0; 1539 compiled = true; 1540 } 1541 1542 Map<String, Integer> namedGroups() { 1543 if (namedGroups == null) 1544 namedGroups = new HashMap<String, Integer>(2); 1545 return namedGroups; 1546 } 1547 1548 /** 1549 * Used to print out a subtree of the Pattern to help with debugging. 1550 */ 1551 private static void printObjectTree(Node node) { 1552 while(node != null) { 1553 if (node instanceof Prolog) { 1554 System.out.println(node); 1555 printObjectTree(((Prolog)node).loop); 1556 System.out.println("**** end contents prolog loop"); 1557 } else if (node instanceof Loop) { 1558 System.out.println(node); 1559 printObjectTree(((Loop)node).body); 1560 System.out.println("**** end contents Loop body"); 1561 } else if (node instanceof Curly) { 1562 System.out.println(node); 1563 printObjectTree(((Curly)node).atom); 1564 System.out.println("**** end contents Curly body"); 1565 } else if (node instanceof GroupCurly) { 1566 System.out.println(node); 1567 printObjectTree(((GroupCurly)node).atom); 1568 System.out.println("**** end contents GroupCurly body"); 1569 } else if (node instanceof GroupTail) { 1570 System.out.println(node); 1571 System.out.println("Tail next is "+node.next); 1572 return; 1573 } else { 1574 System.out.println(node); 1575 } 1576 node = node.next; 1577 if (node != null) 1578 System.out.println("->next:"); 1579 if (node == Pattern.accept) { 1580 System.out.println("Accept Node"); 1581 node = null; 1582 } 1583 } 1584 } 1585 1586 /** 1587 * Used to accumulate information about a subtree of the object graph 1588 * so that optimizations can be applied to the subtree. 1589 */ 1590 static final class TreeInfo { 1591 int minLength; 1592 int maxLength; 1593 boolean maxValid; 1594 boolean deterministic; 1595 1596 TreeInfo() { 1597 reset(); 1598 } 1599 void reset() { 1600 minLength = 0; 1601 maxLength = 0; 1602 maxValid = true; 1603 deterministic = true; 1604 } 1605 } 1606 1607 /* 1608 * The following private methods are mainly used to improve the 1609 * readability of the code. In order to let the Java compiler easily 1610 * inline them, we should not put many assertions or error checks in them. 1611 */ 1612 1613 /** 1614 * Indicates whether a particular flag is set or not. 1615 */ 1616 private boolean has(int f) { 1617 return (flags & f) != 0; 1618 } 1619 1620 /** 1621 * Match next character, signal error if failed. 1622 */ 1623 private void accept(int ch, String s) { 1624 int testChar = temp[cursor++]; 1625 if (has(COMMENTS)) 1626 testChar = parsePastWhitespace(testChar); 1627 if (ch != testChar) { 1628 throw error(s); 1629 } 1630 } 1631 1632 /** 1633 * Mark the end of pattern with a specific character. 1634 */ 1635 private void mark(int c) { 1636 temp[patternLength] = c; 1637 } 1638 1639 /** 1640 * Peek the next character, and do not advance the cursor. 1641 */ 1642 private int peek() { 1643 int ch = temp[cursor]; 1644 if (has(COMMENTS)) 1645 ch = peekPastWhitespace(ch); 1646 return ch; 1647 } 1648 1649 /** 1650 * Read the next character, and advance the cursor by one. 1651 */ 1652 private int read() { 1653 int ch = temp[cursor++]; 1654 if (has(COMMENTS)) 1655 ch = parsePastWhitespace(ch); 1656 return ch; 1657 } 1658 1659 /** 1660 * Read the next character, and advance the cursor by one, 1661 * ignoring the COMMENTS setting 1662 */ 1663 private int readEscaped() { 1664 int ch = temp[cursor++]; 1665 return ch; 1666 } 1667 1668 /** 1669 * Advance the cursor by one, and peek the next character. 1670 */ 1671 private int next() { 1672 int ch = temp[++cursor]; 1673 if (has(COMMENTS)) 1674 ch = peekPastWhitespace(ch); 1675 return ch; 1676 } 1677 1678 /** 1679 * Advance the cursor by one, and peek the next character, 1680 * ignoring the COMMENTS setting 1681 */ 1682 private int nextEscaped() { 1683 int ch = temp[++cursor]; 1684 return ch; 1685 } 1686 1687 /** 1688 * If in xmode peek past whitespace and comments. 1689 */ 1690 private int peekPastWhitespace(int ch) { 1691 while (ASCII.isSpace(ch) || ch == '#') { 1692 while (ASCII.isSpace(ch)) 1693 ch = temp[++cursor]; 1694 if (ch == '#') { 1695 ch = peekPastLine(); 1696 } 1697 } 1698 return ch; 1699 } 1700 1701 /** 1702 * If in xmode parse past whitespace and comments. 1703 */ 1704 private int parsePastWhitespace(int ch) { 1705 while (ASCII.isSpace(ch) || ch == '#') { 1706 while (ASCII.isSpace(ch)) 1707 ch = temp[cursor++]; 1708 if (ch == '#') 1709 ch = parsePastLine(); 1710 } 1711 return ch; 1712 } 1713 1714 /** 1715 * xmode parse past comment to end of line. 1716 */ 1717 private int parsePastLine() { 1718 int ch = temp[cursor++]; 1719 while (ch != 0 && !isLineSeparator(ch)) 1720 ch = temp[cursor++]; 1721 return ch; 1722 } 1723 1724 /** 1725 * xmode peek past comment to end of line. 1726 */ 1727 private int peekPastLine() { 1728 int ch = temp[++cursor]; 1729 while (ch != 0 && !isLineSeparator(ch)) 1730 ch = temp[++cursor]; 1731 return ch; 1732 } 1733 1734 /** 1735 * Determines if character is a line separator in the current mode 1736 */ 1737 private boolean isLineSeparator(int ch) { 1738 if (has(UNIX_LINES)) { 1739 return ch == '\n'; 1740 } else { 1741 return (ch == '\n' || 1742 ch == '\r' || 1743 (ch|1) == '\u2029' || 1744 ch == '\u0085'); 1745 } 1746 } 1747 1748 /** 1749 * Read the character after the next one, and advance the cursor by two. 1750 */ 1751 private int skip() { 1752 int i = cursor; 1753 int ch = temp[i+1]; 1754 cursor = i + 2; 1755 return ch; 1756 } 1757 1758 /** 1759 * Unread one next character, and retreat cursor by one. 1760 */ 1761 private void unread() { 1762 cursor--; 1763 } 1764 1765 /** 1766 * Internal method used for handling all syntax errors. The pattern is 1767 * displayed with a pointer to aid in locating the syntax error. 1768 */ 1769 private PatternSyntaxException error(String s) { 1770 return new PatternSyntaxException(s, normalizedPattern, cursor - 1); 1771 } 1772 1773 /** 1774 * Determines if there is any supplementary character or unpaired 1775 * surrogate in the specified range. 1776 */ 1777 private boolean findSupplementary(int start, int end) { 1778 for (int i = start; i < end; i++) { 1779 if (isSupplementary(temp[i])) 1780 return true; 1781 } 1782 return false; 1783 } 1784 1785 /** 1786 * Determines if the specified code point is a supplementary 1787 * character or unpaired surrogate. 1788 */ 1789 private static final boolean isSupplementary(int ch) { 1790 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || isSurrogate(ch); 1791 } 1792 1793 /** 1794 * The following methods handle the main parsing. They are sorted 1795 * according to their precedence order, the lowest one first. 1796 */ 1797 1798 /** 1799 * The expression is parsed with branch nodes added for alternations. 1800 * This may be called recursively to parse sub expressions that may 1801 * contain alternations. 1802 */ 1803 private Node expr(Node end) { 1804 Node prev = null; 1805 Node firstTail = null; 1806 Node branchConn = null; 1807 1808 for (;;) { 1809 Node node = sequence(end); 1810 Node nodeTail = root; //double return 1811 if (prev == null) { 1812 prev = node; 1813 firstTail = nodeTail; 1814 } else { 1815 // Branch 1816 if (branchConn == null) { 1817 branchConn = new BranchConn(); 1818 branchConn.next = end; 1819 } 1820 if (node == end) { 1821 // if the node returned from sequence() is "end" 1822 // we have an empty expr, set a null atom into 1823 // the branch to indicate to go "next" directly. 1824 node = null; 1825 } else { 1826 // the "tail.next" of each atom goes to branchConn 1827 nodeTail.next = branchConn; 1828 } 1829 if (prev instanceof Branch) { 1830 ((Branch)prev).add(node); 1831 } else { 1832 if (prev == end) { 1833 prev = null; 1834 } else { 1835 // replace the "end" with "branchConn" at its tail.next 1836 // when put the "prev" into the branch as the first atom. 1837 firstTail.next = branchConn; 1838 } 1839 prev = new Branch(prev, node, branchConn); 1840 } 1841 } 1842 if (peek() != '|') { 1843 return prev; 1844 } 1845 next(); 1846 } 1847 } 1848 1849 /** 1850 * Parsing of sequences between alternations. 1851 */ 1852 private Node sequence(Node end) { 1853 Node head = null; 1854 Node tail = null; 1855 Node node = null; 1856 LOOP: 1857 for (;;) { 1858 int ch = peek(); 1859 switch (ch) { 1860 case '(': 1861 // Because group handles its own closure, 1862 // we need to treat it differently 1863 node = group0(); 1864 // Check for comment or flag group 1865 if (node == null) 1866 continue; 1867 if (head == null) 1868 head = node; 1869 else 1870 tail.next = node; 1871 // Double return: Tail was returned in root 1872 tail = root; 1873 continue; 1874 case '[': 1875 node = clazz(true); 1876 break; 1877 case '\\': 1878 ch = nextEscaped(); 1879 if (ch == 'p' || ch == 'P') { 1880 boolean oneLetter = true; 1881 boolean comp = (ch == 'P'); 1882 ch = next(); // Consume { if present 1883 if (ch != '{') { 1884 unread(); 1885 } else { 1886 oneLetter = false; 1887 } 1888 node = family(oneLetter).maybeComplement(comp); 1889 } else { 1890 unread(); 1891 node = atom(); 1892 } 1893 break; 1894 case '^': 1895 next(); 1896 if (has(MULTILINE)) { 1897 if (has(UNIX_LINES)) 1898 node = new UnixCaret(); 1899 else 1900 node = new Caret(); 1901 } else { 1902 node = new Begin(); 1903 } 1904 break; 1905 case '$': 1906 next(); 1907 if (has(UNIX_LINES)) 1908 node = new UnixDollar(has(MULTILINE)); 1909 else 1910 node = new Dollar(has(MULTILINE)); 1911 break; 1912 case '.': 1913 next(); 1914 if (has(DOTALL)) { 1915 node = new All(); 1916 } else { 1917 if (has(UNIX_LINES)) 1918 node = new UnixDot(); 1919 else { 1920 node = new Dot(); 1921 } 1922 } 1923 break; 1924 case '|': 1925 case ')': 1926 break LOOP; 1927 case ']': // Now interpreting dangling ] and } as literals 1928 case '}': 1929 node = atom(); 1930 break; 1931 case '?': 1932 case '*': 1933 case '+': 1934 next(); 1935 throw error("Dangling meta character '" + ((char)ch) + "'"); 1936 case 0: 1937 if (cursor >= patternLength) { 1938 break LOOP; 1939 } 1940 // Fall through 1941 default: 1942 node = atom(); 1943 break; 1944 } 1945 1946 node = closure(node); 1947 1948 if (head == null) { 1949 head = tail = node; 1950 } else { 1951 tail.next = node; 1952 tail = node; 1953 } 1954 } 1955 if (head == null) { 1956 return end; 1957 } 1958 tail.next = end; 1959 root = tail; //double return 1960 return head; 1961 } 1962 1963 /** 1964 * Parse and add a new Single or Slice. 1965 */ 1966 private Node atom() { 1967 int first = 0; 1968 int prev = -1; 1969 boolean hasSupplementary = false; 1970 int ch = peek(); 1971 for (;;) { 1972 switch (ch) { 1973 case '*': 1974 case '+': 1975 case '?': 1976 case '{': 1977 if (first > 1) { 1978 cursor = prev; // Unwind one character 1979 first--; 1980 } 1981 break; 1982 case '$': 1983 case '.': 1984 case '^': 1985 case '(': 1986 case '[': 1987 case '|': 1988 case ')': 1989 break; 1990 case '\\': 1991 ch = nextEscaped(); 1992 if (ch == 'p' || ch == 'P') { // Property 1993 if (first > 0) { // Slice is waiting; handle it first 1994 unread(); 1995 break; 1996 } else { // No slice; just return the family node 1997 boolean comp = (ch == 'P'); 1998 boolean oneLetter = true; 1999 ch = next(); // Consume { if present 2000 if (ch != '{') 2001 unread(); 2002 else 2003 oneLetter = false; 2004 return family(oneLetter).maybeComplement(comp); 2005 } 2006 } 2007 unread(); 2008 prev = cursor; 2009 ch = escape(false, first == 0); 2010 if (ch >= 0) { 2011 append(ch, first); 2012 first++; 2013 if (isSupplementary(ch)) { 2014 hasSupplementary = true; 2015 } 2016 ch = peek(); 2017 continue; 2018 } else if (first == 0) { 2019 return root; 2020 } 2021 // Unwind meta escape sequence 2022 cursor = prev; 2023 break; 2024 case 0: 2025 if (cursor >= patternLength) { 2026 break; 2027 } 2028 // Fall through 2029 default: 2030 prev = cursor; 2031 append(ch, first); 2032 first++; 2033 if (isSupplementary(ch)) { 2034 hasSupplementary = true; 2035 } 2036 ch = next(); 2037 continue; 2038 } 2039 break; 2040 } 2041 if (first == 1) { 2042 return newSingle(buffer[0]); 2043 } else { 2044 return newSlice(buffer, first, hasSupplementary); 2045 } 2046 } 2047 2048 private void append(int ch, int len) { 2049 if (len >= buffer.length) { 2050 int[] tmp = new int[len+len]; 2051 System.arraycopy(buffer, 0, tmp, 0, len); 2052 buffer = tmp; 2053 } 2054 buffer[len] = ch; 2055 } 2056 2057 /** 2058 * Parses a backref greedily, taking as many numbers as it 2059 * can. The first digit is always treated as a backref, but 2060 * multi digit numbers are only treated as a backref if at 2061 * least that many backrefs exist at this point in the regex. 2062 */ 2063 private Node ref(int refNum) { 2064 boolean done = false; 2065 while(!done) { 2066 int ch = peek(); 2067 switch(ch) { 2068 case '0': 2069 case '1': 2070 case '2': 2071 case '3': 2072 case '4': 2073 case '5': 2074 case '6': 2075 case '7': 2076 case '8': 2077 case '9': 2078 int newRefNum = (refNum * 10) + (ch - '0'); 2079 // Add another number if it doesn't make a group 2080 // that doesn't exist 2081 if (capturingGroupCount - 1 < newRefNum) { 2082 done = true; 2083 break; 2084 } 2085 refNum = newRefNum; 2086 read(); 2087 break; 2088 default: 2089 done = true; 2090 break; 2091 } 2092 } 2093 if (has(CASE_INSENSITIVE)) 2094 return new CIBackRef(refNum, has(UNICODE_CASE)); 2095 else 2096 return new BackRef(refNum); 2097 } 2098 2099 /** 2100 * Parses an escape sequence to determine the actual value that needs 2101 * to be matched. 2102 * If -1 is returned and create was true a new object was added to the tree 2103 * to handle the escape sequence. 2104 * If the returned value is greater than zero, it is the value that 2105 * matches the escape sequence. 2106 */ 2107 private int escape(boolean inclass, boolean create) { 2108 int ch = skip(); 2109 switch (ch) { 2110 case '0': 2111 return o(); 2112 case '1': 2113 case '2': 2114 case '3': 2115 case '4': 2116 case '5': 2117 case '6': 2118 case '7': 2119 case '8': 2120 case '9': 2121 if (inclass) break; 2122 if (create) { 2123 root = ref((ch - '0')); 2124 } 2125 return -1; 2126 case 'A': 2127 if (inclass) break; 2128 if (create) root = new Begin(); 2129 return -1; 2130 case 'B': 2131 if (inclass) break; 2132 if (create) root = new Bound(Bound.NONE); 2133 return -1; 2134 case 'C': 2135 break; 2136 case 'D': 2137 if (create) root = new Ctype(ASCII.DIGIT).complement(); 2138 return -1; 2139 case 'E': 2140 case 'F': 2141 break; 2142 case 'G': 2143 if (inclass) break; 2144 if (create) root = new LastMatch(); 2145 return -1; 2146 case 'H': 2147 case 'I': 2148 case 'J': 2149 case 'K': 2150 case 'L': 2151 case 'M': 2152 case 'N': 2153 case 'O': 2154 case 'P': 2155 case 'Q': 2156 case 'R': 2157 break; 2158 case 'S': 2159 if (create) root = new Ctype(ASCII.SPACE).complement(); 2160 return -1; 2161 case 'T': 2162 case 'U': 2163 case 'V': 2164 break; 2165 case 'W': 2166 if (create) root = new Ctype(ASCII.WORD).complement(); 2167 return -1; 2168 case 'X': 2169 case 'Y': 2170 break; 2171 case 'Z': 2172 if (inclass) break; 2173 if (create) { 2174 if (has(UNIX_LINES)) 2175 root = new UnixDollar(false); 2176 else 2177 root = new Dollar(false); 2178 } 2179 return -1; 2180 case 'a': 2181 return '\007'; 2182 case 'b': 2183 if (inclass) break; 2184 if (create) root = new Bound(Bound.BOTH); 2185 return -1; 2186 case 'c': 2187 return c(); 2188 case 'd': 2189 if (create) root = new Ctype(ASCII.DIGIT); 2190 return -1; 2191 case 'e': 2192 return '\033'; 2193 case 'f': 2194 return '\f'; 2195 case 'g': 2196 case 'h': 2197 case 'i': 2198 case 'j': 2199 break; 2200 case 'k': 2201 if (inclass) 2202 break; 2203 if (read() != '<') 2204 throw error("\\k is not followed by '<' for named capturing group"); 2205 String name = groupname(read()); 2206 if (!namedGroups().containsKey(name)) 2207 throw error("(named capturing group <"+ name+"> does not exit"); 2208 if (create) { 2209 if (has(CASE_INSENSITIVE)) 2210 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE)); 2211 else 2212 root = new BackRef(namedGroups().get(name)); 2213 } 2214 return -1; 2215 case 'l': 2216 case 'm': 2217 break; 2218 case 'n': 2219 return '\n'; 2220 case 'o': 2221 case 'p': 2222 case 'q': 2223 break; 2224 case 'r': 2225 return '\r'; 2226 case 's': 2227 if (create) root = new Ctype(ASCII.SPACE); 2228 return -1; 2229 case 't': 2230 return '\t'; 2231 case 'u': 2232 return u(); 2233 case 'v': 2234 return '\013'; 2235 case 'w': 2236 if (create) root = new Ctype(ASCII.WORD); 2237 return -1; 2238 case 'x': 2239 return x(); 2240 case 'y': 2241 break; 2242 case 'z': 2243 if (inclass) break; 2244 if (create) root = new End(); 2245 return -1; 2246 default: 2247 return ch; 2248 } 2249 throw error("Illegal/unsupported escape sequence"); 2250 } 2251 2252 /** 2253 * Parse a character class, and return the node that matches it. 2254 * 2255 * Consumes a ] on the way out if consume is true. Usually consume 2256 * is true except for the case of [abc&&def] where def is a separate 2257 * right hand node with "understood" brackets. 2258 */ 2259 private CharProperty clazz(boolean consume) { 2260 CharProperty prev = null; 2261 CharProperty node = null; 2262 BitClass bits = new BitClass(); 2263 boolean include = true; 2264 boolean firstInClass = true; 2265 int ch = next(); 2266 for (;;) { 2267 switch (ch) { 2268 case '^': 2269 // Negates if first char in a class, otherwise literal 2270 if (firstInClass) { 2271 if (temp[cursor-1] != '[') 2272 break; 2273 ch = next(); 2274 include = !include; 2275 continue; 2276 } else { 2277 // ^ not first in class, treat as literal 2278 break; 2279 } 2280 case '[': 2281 firstInClass = false; 2282 node = clazz(true); 2283 if (prev == null) 2284 prev = node; 2285 else 2286 prev = union(prev, node); 2287 ch = peek(); 2288 continue; 2289 case '&': 2290 firstInClass = false; 2291 ch = next(); 2292 if (ch == '&') { 2293 ch = next(); 2294 CharProperty rightNode = null; 2295 while (ch != ']' && ch != '&') { 2296 if (ch == '[') { 2297 if (rightNode == null) 2298 rightNode = clazz(true); 2299 else 2300 rightNode = union(rightNode, clazz(true)); 2301 } else { // abc&&def 2302 unread(); 2303 rightNode = clazz(false); 2304 } 2305 ch = peek(); 2306 } 2307 if (rightNode != null) 2308 node = rightNode; 2309 if (prev == null) { 2310 if (rightNode == null) 2311 throw error("Bad class syntax"); 2312 else 2313 prev = rightNode; 2314 } else { 2315 prev = intersection(prev, node); 2316 } 2317 } else { 2318 // treat as a literal & 2319 unread(); 2320 break; 2321 } 2322 continue; 2323 case 0: 2324 firstInClass = false; 2325 if (cursor >= patternLength) 2326 throw error("Unclosed character class"); 2327 break; 2328 case ']': 2329 firstInClass = false; 2330 if (prev != null) { 2331 if (consume) 2332 next(); 2333 return prev; 2334 } 2335 break; 2336 default: 2337 firstInClass = false; 2338 break; 2339 } 2340 node = range(bits); 2341 if (include) { 2342 if (prev == null) { 2343 prev = node; 2344 } else { 2345 if (prev != node) 2346 prev = union(prev, node); 2347 } 2348 } else { 2349 if (prev == null) { 2350 prev = node.complement(); 2351 } else { 2352 if (prev != node) 2353 prev = setDifference(prev, node); 2354 } 2355 } 2356 ch = peek(); 2357 } 2358 } 2359 2360 private CharProperty bitsOrSingle(BitClass bits, int ch) { 2361 /* Bits can only handle codepoints in [u+0000-u+00ff] range. 2362 Use "single" node instead of bits when dealing with unicode 2363 case folding for codepoints listed below. 2364 (1)Uppercase out of range: u+00ff, u+00b5 2365 toUpperCase(u+00ff) -> u+0178 2366 toUpperCase(u+00b5) -> u+039c 2367 (2)LatinSmallLetterLongS u+17f 2368 toUpperCase(u+017f) -> u+0053 2369 (3)LatinSmallLetterDotlessI u+131 2370 toUpperCase(u+0131) -> u+0049 2371 (4)LatinCapitalLetterIWithDotAbove u+0130 2372 toLowerCase(u+0130) -> u+0069 2373 (5)KelvinSign u+212a 2374 toLowerCase(u+212a) ==> u+006B 2375 (6)AngstromSign u+212b 2376 toLowerCase(u+212b) ==> u+00e5 2377 */ 2378 int d; 2379 if (ch < 256 && 2380 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) && 2381 (ch == 0xff || ch == 0xb5 || 2382 ch == 0x49 || ch == 0x69 || //I and i 2383 ch == 0x53 || ch == 0x73 || //S and s 2384 ch == 0x4b || ch == 0x6b || //K and k 2385 ch == 0xc5 || ch == 0xe5))) //A+ring 2386 return bits.add(ch, flags()); 2387 return newSingle(ch); 2388 } 2389 2390 /** 2391 * Parse a single character or a character range in a character class 2392 * and return its representative node. 2393 */ 2394 private CharProperty range(BitClass bits) { 2395 int ch = peek(); 2396 if (ch == '\\') { 2397 ch = nextEscaped(); 2398 if (ch == 'p' || ch == 'P') { // A property 2399 boolean comp = (ch == 'P'); 2400 boolean oneLetter = true; 2401 // Consume { if present 2402 ch = next(); 2403 if (ch != '{') 2404 unread(); 2405 else 2406 oneLetter = false; 2407 return family(oneLetter).maybeComplement(comp); 2408 } else { // ordinary escape 2409 unread(); 2410 ch = escape(true, true); 2411 if (ch == -1) 2412 return (CharProperty) root; 2413 } 2414 } else { 2415 ch = single(); 2416 } 2417 if (ch >= 0) { 2418 if (peek() == '-') { 2419 int endRange = temp[cursor+1]; 2420 if (endRange == '[') { 2421 return bitsOrSingle(bits, ch); 2422 } 2423 if (endRange != ']') { 2424 next(); 2425 int m = single(); 2426 if (m < ch) 2427 throw error("Illegal character range"); 2428 if (has(CASE_INSENSITIVE)) 2429 return caseInsensitiveRangeFor(ch, m); 2430 else 2431 return rangeFor(ch, m); 2432 } 2433 } 2434 return bitsOrSingle(bits, ch); 2435 } 2436 throw error("Unexpected character '"+((char)ch)+"'"); 2437 } 2438 2439 private int single() { 2440 int ch = peek(); 2441 switch (ch) { 2442 case '\\': 2443 return escape(true, false); 2444 default: 2445 next(); 2446 return ch; 2447 } 2448 } 2449 2450 /** 2451 * Parses a Unicode character family and returns its representative node. 2452 */ 2453 private CharProperty family(boolean singleLetter) { 2454 next(); 2455 String name; 2456 2457 if (singleLetter) { 2458 int c = temp[cursor]; 2459 if (!Character.isSupplementaryCodePoint(c)) { 2460 name = String.valueOf((char)c); 2461 } else { 2462 name = new String(temp, cursor, 1); 2463 } 2464 read(); 2465 } else { 2466 int i = cursor; 2467 mark('}'); 2468 while(read() != '}') { 2469 } 2470 mark('\000'); 2471 int j = cursor; 2472 if (j > patternLength) 2473 throw error("Unclosed character family"); 2474 if (i + 1 >= j) 2475 throw error("Empty character family"); 2476 name = new String(temp, i, j-i-1); 2477 } 2478 2479 if (name.startsWith("In")) { 2480 return unicodeBlockPropertyFor(name.substring(2)); 2481 } else { 2482 if (name.startsWith("Is")) 2483 name = name.substring(2); 2484 return charPropertyNodeFor(name); 2485 } 2486 } 2487 2488 /** 2489 * Returns a CharProperty matching all characters in a UnicodeBlock. 2490 */ 2491 private CharProperty unicodeBlockPropertyFor(String name) { 2492 final Character.UnicodeBlock block; 2493 try { 2494 block = Character.UnicodeBlock.forName(name); 2495 } catch (IllegalArgumentException iae) { 2496 throw error("Unknown character block name {" + name + "}"); 2497 } 2498 return new CharProperty() { 2499 boolean isSatisfiedBy(int ch) { 2500 return block == Character.UnicodeBlock.of(ch);}}; 2501 } 2502 2503 /** 2504 * Returns a CharProperty matching all characters in a named property. 2505 */ 2506 private CharProperty charPropertyNodeFor(String name) { 2507 CharProperty p = CharPropertyNames.charPropertyFor(name); 2508 if (p == null) 2509 throw error("Unknown character property name {" + name + "}"); 2510 return p; 2511 } 2512 2513 /** 2514 * Parses and returns the name of a "named capturing group", the trailing 2515 * ">" is consumed after parsing. 2516 */ 2517 private String groupname(int ch) { 2518 StringBuilder sb = new StringBuilder(); 2519 sb.append(Character.toChars(ch)); 2520 while (ASCII.isLower(ch=read()) || ASCII.isUpper(ch) || 2521 ASCII.isDigit(ch)) { 2522 sb.append(Character.toChars(ch)); 2523 } 2524 if (sb.length() == 0) 2525 throw error("named capturing group has 0 length name"); 2526 if (ch != '>') 2527 throw error("named capturing group is missing trailing '>'"); 2528 return sb.toString(); 2529 } 2530 2531 /** 2532 * Parses a group and returns the head node of a set of nodes that process 2533 * the group. Sometimes a double return system is used where the tail is 2534 * returned in root. 2535 */ 2536 private Node group0() { 2537 boolean capturingGroup = false; 2538 Node head = null; 2539 Node tail = null; 2540 int save = flags; 2541 root = null; 2542 int ch = next(); 2543 if (ch == '?') { 2544 ch = skip(); 2545 switch (ch) { 2546 case ':': // (?:xxx) pure group 2547 head = createGroup(true); 2548 tail = root; 2549 head.next = expr(tail); 2550 break; 2551 case '=': // (?=xxx) and (?!xxx) lookahead 2552 case '!': 2553 head = createGroup(true); 2554 tail = root; 2555 head.next = expr(tail); 2556 if (ch == '=') { 2557 head = tail = new Pos(head); 2558 } else { 2559 head = tail = new Neg(head); 2560 } 2561 break; 2562 case '>': // (?>xxx) independent group 2563 head = createGroup(true); 2564 tail = root; 2565 head.next = expr(tail); 2566 head = tail = new Ques(head, INDEPENDENT); 2567 break; 2568 case '<': // (?<xxx) look behind 2569 ch = read(); 2570 if (Character.isLetter(ch)) { // named captured group 2571 String name = groupname(ch); 2572 if (namedGroups().containsKey(name)) 2573 throw error("Named capturing group <" + name 2574 + "> is already defined"); 2575 capturingGroup = true; 2576 head = createGroup(false); 2577 tail = root; 2578 namedGroups().put(name, capturingGroupCount-1); 2579 head.next = expr(tail); 2580 break; 2581 } 2582 int start = cursor; 2583 head = createGroup(true); 2584 tail = root; 2585 head.next = expr(tail); 2586 tail.next = lookbehindEnd; 2587 TreeInfo info = new TreeInfo(); 2588 head.study(info); 2589 if (info.maxValid == false) { 2590 throw error("Look-behind group does not have " 2591 + "an obvious maximum length"); 2592 } 2593 boolean hasSupplementary = findSupplementary(start, patternLength); 2594 if (ch == '=') { 2595 head = tail = (hasSupplementary ? 2596 new BehindS(head, info.maxLength, 2597 info.minLength) : 2598 new Behind(head, info.maxLength, 2599 info.minLength)); 2600 } else if (ch == '!') { 2601 head = tail = (hasSupplementary ? 2602 new NotBehindS(head, info.maxLength, 2603 info.minLength) : 2604 new NotBehind(head, info.maxLength, 2605 info.minLength)); 2606 } else { 2607 throw error("Unknown look-behind group"); 2608 } 2609 break; 2610 case '$': 2611 case '@': 2612 throw error("Unknown group type"); 2613 default: // (?xxx:) inlined match flags 2614 unread(); 2615 addFlag(); 2616 ch = read(); 2617 if (ch == ')') { 2618 return null; // Inline modifier only 2619 } 2620 if (ch != ':') { 2621 throw error("Unknown inline modifier"); 2622 } 2623 head = createGroup(true); 2624 tail = root; 2625 head.next = expr(tail); 2626 break; 2627 } 2628 } else { // (xxx) a regular group 2629 capturingGroup = true; 2630 head = createGroup(false); 2631 tail = root; 2632 head.next = expr(tail); 2633 } 2634 2635 accept(')', "Unclosed group"); 2636 flags = save; 2637 2638 // Check for quantifiers 2639 Node node = closure(head); 2640 if (node == head) { // No closure 2641 root = tail; 2642 return node; // Dual return 2643 } 2644 if (head == tail) { // Zero length assertion 2645 root = node; 2646 return node; // Dual return 2647 } 2648 2649 if (node instanceof Ques) { 2650 Ques ques = (Ques) node; 2651 if (ques.type == POSSESSIVE) { 2652 root = node; 2653 return node; 2654 } 2655 tail.next = new BranchConn(); 2656 tail = tail.next; 2657 if (ques.type == GREEDY) { 2658 head = new Branch(head, null, tail); 2659 } else { // Reluctant quantifier 2660 head = new Branch(null, head, tail); 2661 } 2662 root = tail; 2663 return head; 2664 } else if (node instanceof Curly) { 2665 Curly curly = (Curly) node; 2666 if (curly.type == POSSESSIVE) { 2667 root = node; 2668 return node; 2669 } 2670 // Discover if the group is deterministic 2671 TreeInfo info = new TreeInfo(); 2672 if (head.study(info)) { // Deterministic 2673 GroupTail temp = (GroupTail) tail; 2674 head = root = new GroupCurly(head.next, curly.cmin, 2675 curly.cmax, curly.type, 2676 ((GroupTail)tail).localIndex, 2677 ((GroupTail)tail).groupIndex, 2678 capturingGroup); 2679 return head; 2680 } else { // Non-deterministic 2681 int temp = ((GroupHead) head).localIndex; 2682 Loop loop; 2683 if (curly.type == GREEDY) 2684 loop = new Loop(this.localCount, temp); 2685 else // Reluctant Curly 2686 loop = new LazyLoop(this.localCount, temp); 2687 Prolog prolog = new Prolog(loop); 2688 this.localCount += 1; 2689 loop.cmin = curly.cmin; 2690 loop.cmax = curly.cmax; 2691 loop.body = head; 2692 tail.next = loop; 2693 root = loop; 2694 return prolog; // Dual return 2695 } 2696 } 2697 throw error("Internal logic error"); 2698 } 2699 2700 /** 2701 * Create group head and tail nodes using double return. If the group is 2702 * created with anonymous true then it is a pure group and should not 2703 * affect group counting. 2704 */ 2705 private Node createGroup(boolean anonymous) { 2706 int localIndex = localCount++; 2707 int groupIndex = 0; 2708 if (!anonymous) 2709 groupIndex = capturingGroupCount++; 2710 GroupHead head = new GroupHead(localIndex); 2711 root = new GroupTail(localIndex, groupIndex); 2712 if (!anonymous && groupIndex < 10) 2713 groupNodes[groupIndex] = head; 2714 return head; 2715 } 2716 2717 /** 2718 * Parses inlined match flags and set them appropriately. 2719 */ 2720 private void addFlag() { 2721 int ch = peek(); 2722 for (;;) { 2723 switch (ch) { 2724 case 'i': 2725 flags |= CASE_INSENSITIVE; 2726 break; 2727 case 'm': 2728 flags |= MULTILINE; 2729 break; 2730 case 's': 2731 flags |= DOTALL; 2732 break; 2733 case 'd': 2734 flags |= UNIX_LINES; 2735 break; 2736 case 'u': 2737 flags |= UNICODE_CASE; 2738 break; 2739 case 'c': 2740 flags |= CANON_EQ; 2741 break; 2742 case 'x': 2743 flags |= COMMENTS; 2744 break; 2745 case '-': // subFlag then fall through 2746 ch = next(); 2747 subFlag(); 2748 default: 2749 return; 2750 } 2751 ch = next(); 2752 } 2753 } 2754 2755 /** 2756 * Parses the second part of inlined match flags and turns off 2757 * flags appropriately. 2758 */ 2759 private void subFlag() { 2760 int ch = peek(); 2761 for (;;) { 2762 switch (ch) { 2763 case 'i': 2764 flags &= ~CASE_INSENSITIVE; 2765 break; 2766 case 'm': 2767 flags &= ~MULTILINE; 2768 break; 2769 case 's': 2770 flags &= ~DOTALL; 2771 break; 2772 case 'd': 2773 flags &= ~UNIX_LINES; 2774 break; 2775 case 'u': 2776 flags &= ~UNICODE_CASE; 2777 break; 2778 case 'c': 2779 flags &= ~CANON_EQ; 2780 break; 2781 case 'x': 2782 flags &= ~COMMENTS; 2783 break; 2784 default: 2785 return; 2786 } 2787 ch = next(); 2788 } 2789 } 2790 2791 static final int MAX_REPS = 0x7FFFFFFF; 2792 2793 static final int GREEDY = 0; 2794 2795 static final int LAZY = 1; 2796 2797 static final int POSSESSIVE = 2; 2798 2799 static final int INDEPENDENT = 3; 2800 2801 /** 2802 * Processes repetition. If the next character peeked is a quantifier 2803 * then new nodes must be appended to handle the repetition. 2804 * Prev could be a single or a group, so it could be a chain of nodes. 2805 */ 2806 private Node closure(Node prev) { 2807 Node atom; 2808 int ch = peek(); 2809 switch (ch) { 2810 case '?': 2811 ch = next(); 2812 if (ch == '?') { 2813 next(); 2814 return new Ques(prev, LAZY); 2815 } else if (ch == '+') { 2816 next(); 2817 return new Ques(prev, POSSESSIVE); 2818 } 2819 return new Ques(prev, GREEDY); 2820 case '*': 2821 ch = next(); 2822 if (ch == '?') { 2823 next(); 2824 return new Curly(prev, 0, MAX_REPS, LAZY); 2825 } else if (ch == '+') { 2826 next(); 2827 return new Curly(prev, 0, MAX_REPS, POSSESSIVE); 2828 } 2829 return new Curly(prev, 0, MAX_REPS, GREEDY); 2830 case '+': 2831 ch = next(); 2832 if (ch == '?') { 2833 next(); 2834 return new Curly(prev, 1, MAX_REPS, LAZY); 2835 } else if (ch == '+') { 2836 next(); 2837 return new Curly(prev, 1, MAX_REPS, POSSESSIVE); 2838 } 2839 return new Curly(prev, 1, MAX_REPS, GREEDY); 2840 case '{': 2841 ch = temp[cursor+1]; 2842 if (ASCII.isDigit(ch)) { 2843 skip(); 2844 int cmin = 0; 2845 do { 2846 cmin = cmin * 10 + (ch - '0'); 2847 } while (ASCII.isDigit(ch = read())); 2848 int cmax = cmin; 2849 if (ch == ',') { 2850 ch = read(); 2851 cmax = MAX_REPS; 2852 if (ch != '}') { 2853 cmax = 0; 2854 while (ASCII.isDigit(ch)) { 2855 cmax = cmax * 10 + (ch - '0'); 2856 ch = read(); 2857 } 2858 } 2859 } 2860 if (ch != '}') 2861 throw error("Unclosed counted closure"); 2862 if (((cmin) | (cmax) | (cmax - cmin)) < 0) 2863 throw error("Illegal repetition range"); 2864 Curly curly; 2865 ch = peek(); 2866 if (ch == '?') { 2867 next(); 2868 curly = new Curly(prev, cmin, cmax, LAZY); 2869 } else if (ch == '+') { 2870 next(); 2871 curly = new Curly(prev, cmin, cmax, POSSESSIVE); 2872 } else { 2873 curly = new Curly(prev, cmin, cmax, GREEDY); 2874 } 2875 return curly; 2876 } else { 2877 throw error("Illegal repetition"); 2878 } 2879 default: 2880 return prev; 2881 } 2882 } 2883 2884 /** 2885 * Utility method for parsing control escape sequences. 2886 */ 2887 private int c() { 2888 if (cursor < patternLength) { 2889 return read() ^ 64; 2890 } 2891 throw error("Illegal control escape sequence"); 2892 } 2893 2894 /** 2895 * Utility method for parsing octal escape sequences. 2896 */ 2897 private int o() { 2898 int n = read(); 2899 if (((n-'0')|('7'-n)) >= 0) { 2900 int m = read(); 2901 if (((m-'0')|('7'-m)) >= 0) { 2902 int o = read(); 2903 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) { 2904 return (n - '0') * 64 + (m - '0') * 8 + (o - '0'); 2905 } 2906 unread(); 2907 return (n - '0') * 8 + (m - '0'); 2908 } 2909 unread(); 2910 return (n - '0'); 2911 } 2912 throw error("Illegal octal escape sequence"); 2913 } 2914 2915 /** 2916 * Utility method for parsing hexadecimal escape sequences. 2917 */ 2918 private int x() { 2919 int n = read(); 2920 if (ASCII.isHexDigit(n)) { 2921 int m = read(); 2922 if (ASCII.isHexDigit(m)) { 2923 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m); 2924 } 2925 } 2926 throw error("Illegal hexadecimal escape sequence"); 2927 } 2928 2929 /** 2930 * Utility method for parsing unicode escape sequences. 2931 */ 2932 private int cursor() { 2933 return cursor; 2934 } 2935 2936 private void setcursor(int pos) { 2937 cursor = pos; 2938 } 2939 2940 private int uxxxx() { 2941 int n = 0; 2942 for (int i = 0; i < 4; i++) { 2943 int ch = read(); 2944 if (!ASCII.isHexDigit(ch)) { 2945 throw error("Illegal Unicode escape sequence"); 2946 } 2947 n = n * 16 + ASCII.toDigit(ch); 2948 } 2949 return n; 2950 } 2951 2952 private int u() { 2953 int n = uxxxx(); 2954 if (Character.isHighSurrogate((char)n)) { 2955 int cur = cursor(); 2956 if (read() == '\\' && read() == 'u') { 2957 int n2 = uxxxx(); 2958 if (Character.isLowSurrogate((char)n2)) 2959 return Character.toCodePoint((char)n, (char)n2); 2960 } 2961 setcursor(cur); 2962 } 2963 return n; 2964 } 2965 2966 // 2967 // Utility methods for code point support 2968 // 2969 2970 /** 2971 * Tests a surrogate value. 2972 */ 2973 private static final boolean isSurrogate(int c) { 2974 return c >= Character.MIN_HIGH_SURROGATE && c <= Character.MAX_LOW_SURROGATE; 2975 } 2976 2977 private static final int countChars(CharSequence seq, int index, 2978 int lengthInCodePoints) { 2979 // optimization 2980 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) { 2981 assert (index >= 0 && index < seq.length()); 2982 return 1; 2983 } 2984 int length = seq.length(); 2985 int x = index; 2986 if (lengthInCodePoints >= 0) { 2987 assert (index >= 0 && index < length); 2988 for (int i = 0; x < length && i < lengthInCodePoints; i++) { 2989 if (Character.isHighSurrogate(seq.charAt(x++))) { 2990 if (x < length && Character.isLowSurrogate(seq.charAt(x))) { 2991 x++; 2992 } 2993 } 2994 } 2995 return x - index; 2996 } 2997 2998 assert (index >= 0 && index <= length); 2999 if (index == 0) { 3000 return 0; 3001 } 3002 int len = -lengthInCodePoints; 3003 for (int i = 0; x > 0 && i < len; i++) { 3004 if (Character.isLowSurrogate(seq.charAt(--x))) { 3005 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) { 3006 x--; 3007 } 3008 } 3009 } 3010 return index - x; 3011 } 3012 3013 private static final int countCodePoints(CharSequence seq) { 3014 int length = seq.length(); 3015 int n = 0; 3016 for (int i = 0; i < length; ) { 3017 n++; 3018 if (Character.isHighSurrogate(seq.charAt(i++))) { 3019 if (i < length && Character.isLowSurrogate(seq.charAt(i))) { 3020 i++; 3021 } 3022 } 3023 } 3024 return n; 3025 } 3026 3027 /** 3028 * Creates a bit vector for matching Latin-1 values. A normal BitClass 3029 * never matches values above Latin-1, and a complemented BitClass always 3030 * matches values above Latin-1. 3031 */ 3032 private static final class BitClass extends BmpCharProperty { 3033 final boolean[] bits; 3034 BitClass() { bits = new boolean[256]; } 3035 private BitClass(boolean[] bits) { this.bits = bits; } 3036 BitClass add(int c, int flags) { 3037 assert c >= 0 && c <= 255; 3038 if ((flags & CASE_INSENSITIVE) != 0) { 3039 if (ASCII.isAscii(c)) { 3040 bits[ASCII.toUpper(c)] = true; 3041 bits[ASCII.toLower(c)] = true; 3042 } else if ((flags & UNICODE_CASE) != 0) { 3043 bits[Character.toLowerCase(c)] = true; 3044 bits[Character.toUpperCase(c)] = true; 3045 } 3046 } 3047 bits[c] = true; 3048 return this; 3049 } 3050 boolean isSatisfiedBy(int ch) { 3051 return ch < 256 && bits[ch]; 3052 } 3053 } 3054 3055 /** 3056 * Returns a suitably optimized, single character matcher. 3057 */ 3058 private CharProperty newSingle(final int ch) { 3059 if (has(CASE_INSENSITIVE)) { 3060 int lower, upper; 3061 if (has(UNICODE_CASE)) { 3062 upper = Character.toUpperCase(ch); 3063 lower = Character.toLowerCase(upper); 3064 if (upper != lower) 3065 return new SingleU(lower); 3066 } else if (ASCII.isAscii(ch)) { 3067 lower = ASCII.toLower(ch); 3068 upper = ASCII.toUpper(ch); 3069 if (lower != upper) 3070 return new SingleI(lower, upper); 3071 } 3072 } 3073 if (isSupplementary(ch)) 3074 return new SingleS(ch); // Match a given Unicode character 3075 return new Single(ch); // Match a given BMP character 3076 } 3077 3078 /** 3079 * Utility method for creating a string slice matcher. 3080 */ 3081 private Node newSlice(int[] buf, int count, boolean hasSupplementary) { 3082 int[] tmp = new int[count]; 3083 if (has(CASE_INSENSITIVE)) { 3084 if (has(UNICODE_CASE)) { 3085 for (int i = 0; i < count; i++) { 3086 tmp[i] = Character.toLowerCase( 3087 Character.toUpperCase(buf[i])); 3088 } 3089 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp); 3090 } 3091 for (int i = 0; i < count; i++) { 3092 tmp[i] = ASCII.toLower(buf[i]); 3093 } 3094 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp); 3095 } 3096 for (int i = 0; i < count; i++) { 3097 tmp[i] = buf[i]; 3098 } 3099 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp); 3100 } 3101 3102 /** 3103 * The following classes are the building components of the object 3104 * tree that represents a compiled regular expression. The object tree 3105 * is made of individual elements that handle constructs in the Pattern. 3106 * Each type of object knows how to match its equivalent construct with 3107 * the match() method. 3108 */ 3109 3110 /** 3111 * Base class for all node classes. Subclasses should override the match() 3112 * method as appropriate. This class is an accepting node, so its match() 3113 * always returns true. 3114 */ 3115 static class Node extends Object { 3116 Node next; 3117 Node() { 3118 next = Pattern.accept; 3119 } 3120 /** 3121 * This method implements the classic accept node. 3122 */ 3123 boolean match(Matcher matcher, int i, CharSequence seq) { 3124 matcher.last = i; 3125 matcher.groups[0] = matcher.first; 3126 matcher.groups[1] = matcher.last; 3127 return true; 3128 } 3129 /** 3130 * This method is good for all zero length assertions. 3131 */ 3132 boolean study(TreeInfo info) { 3133 if (next != null) { 3134 return next.study(info); 3135 } else { 3136 return info.deterministic; 3137 } 3138 } 3139 } 3140 3141 static class LastNode extends Node { 3142 /** 3143 * This method implements the classic accept node with 3144 * the addition of a check to see if the match occurred 3145 * using all of the input. 3146 */ 3147 boolean match(Matcher matcher, int i, CharSequence seq) { 3148 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to) 3149 return false; 3150 matcher.last = i; 3151 matcher.groups[0] = matcher.first; 3152 matcher.groups[1] = matcher.last; 3153 return true; 3154 } 3155 } 3156 3157 /** 3158 * Used for REs that can start anywhere within the input string. 3159 * This basically tries to match repeatedly at each spot in the 3160 * input string, moving forward after each try. An anchored search 3161 * or a BnM will bypass this node completely. 3162 */ 3163 static class Start extends Node { 3164 int minLength; 3165 Start(Node node) { 3166 this.next = node; 3167 TreeInfo info = new TreeInfo(); 3168 next.study(info); 3169 minLength = info.minLength; 3170 } 3171 boolean match(Matcher matcher, int i, CharSequence seq) { 3172 if (i > matcher.to - minLength) { 3173 matcher.hitEnd = true; 3174 return false; 3175 } 3176 boolean ret = false; 3177 int guard = matcher.to - minLength; 3178 for (; i <= guard; i++) { 3179 if (ret = next.match(matcher, i, seq)) 3180 break; 3181 if (i == guard) 3182 matcher.hitEnd = true; 3183 } 3184 if (ret) { 3185 matcher.first = i; 3186 matcher.groups[0] = matcher.first; 3187 matcher.groups[1] = matcher.last; 3188 } 3189 return ret; 3190 } 3191 boolean study(TreeInfo info) { 3192 next.study(info); 3193 info.maxValid = false; 3194 info.deterministic = false; 3195 return false; 3196 } 3197 } 3198 3199 /* 3200 * StartS supports supplementary characters, including unpaired surrogates. 3201 */ 3202 static final class StartS extends Start { 3203 StartS(Node node) { 3204 super(node); 3205 } 3206 boolean match(Matcher matcher, int i, CharSequence seq) { 3207 if (i > matcher.to - minLength) { 3208 matcher.hitEnd = true; 3209 return false; 3210 } 3211 boolean ret = false; 3212 int guard = matcher.to - minLength; 3213 while (i <= guard) { 3214 if ((ret = next.match(matcher, i, seq)) || i == guard) 3215 break; 3216 // Optimization to move to the next character. This is 3217 // faster than countChars(seq, i, 1). 3218 if (Character.isHighSurrogate(seq.charAt(i++))) { 3219 if (i < seq.length() && Character.isLowSurrogate(seq.charAt(i))) { 3220 i++; 3221 } 3222 } 3223 if (i == guard) 3224 matcher.hitEnd = true; 3225 } 3226 if (ret) { 3227 matcher.first = i; 3228 matcher.groups[0] = matcher.first; 3229 matcher.groups[1] = matcher.last; 3230 } 3231 return ret; 3232 } 3233 } 3234 3235 /** 3236 * Node to anchor at the beginning of input. This object implements the 3237 * match for a \A sequence, and the caret anchor will use this if not in 3238 * multiline mode. 3239 */ 3240 static final class Begin extends Node { 3241 boolean match(Matcher matcher, int i, CharSequence seq) { 3242 int fromIndex = (matcher.anchoringBounds) ? 3243 matcher.from : 0; 3244 if (i == fromIndex && next.match(matcher, i, seq)) { 3245 matcher.first = i; 3246 matcher.groups[0] = i; 3247 matcher.groups[1] = matcher.last; 3248 return true; 3249 } else { 3250 return false; 3251 } 3252 } 3253 } 3254 3255 /** 3256 * Node to anchor at the end of input. This is the absolute end, so this 3257 * should not match at the last newline before the end as $ will. 3258 */ 3259 static final class End extends Node { 3260 boolean match(Matcher matcher, int i, CharSequence seq) { 3261 int endIndex = (matcher.anchoringBounds) ? 3262 matcher.to : matcher.getTextLength(); 3263 if (i == endIndex) { 3264 matcher.hitEnd = true; 3265 return next.match(matcher, i, seq); 3266 } 3267 return false; 3268 } 3269 } 3270 3271 /** 3272 * Node to anchor at the beginning of a line. This is essentially the 3273 * object to match for the multiline ^. 3274 */ 3275 static final class Caret extends Node { 3276 boolean match(Matcher matcher, int i, CharSequence seq) { 3277 int startIndex = matcher.from; 3278 int endIndex = matcher.to; 3279 if (!matcher.anchoringBounds) { 3280 startIndex = 0; 3281 endIndex = matcher.getTextLength(); 3282 } 3283 // Perl does not match ^ at end of input even after newline 3284 if (i == endIndex) { 3285 matcher.hitEnd = true; 3286 return false; 3287 } 3288 if (i > startIndex) { 3289 char ch = seq.charAt(i-1); 3290 if (ch != '\n' && ch != '\r' 3291 && (ch|1) != '\u2029' 3292 && ch != '\u0085' ) { 3293 return false; 3294 } 3295 // Should treat /r/n as one newline 3296 if (ch == '\r' && seq.charAt(i) == '\n') 3297 return false; 3298 } 3299 return next.match(matcher, i, seq); 3300 } 3301 } 3302 3303 /** 3304 * Node to anchor at the beginning of a line when in unixdot mode. 3305 */ 3306 static final class UnixCaret extends Node { 3307 boolean match(Matcher matcher, int i, CharSequence seq) { 3308 int startIndex = matcher.from; 3309 int endIndex = matcher.to; 3310 if (!matcher.anchoringBounds) { 3311 startIndex = 0; 3312 endIndex = matcher.getTextLength(); 3313 } 3314 // Perl does not match ^ at end of input even after newline 3315 if (i == endIndex) { 3316 matcher.hitEnd = true; 3317 return false; 3318 } 3319 if (i > startIndex) { 3320 char ch = seq.charAt(i-1); 3321 if (ch != '\n') { 3322 return false; 3323 } 3324 } 3325 return next.match(matcher, i, seq); 3326 } 3327 } 3328 3329 /** 3330 * Node to match the location where the last match ended. 3331 * This is used for the \G construct. 3332 */ 3333 static final class LastMatch extends Node { 3334 boolean match(Matcher matcher, int i, CharSequence seq) { 3335 if (i != matcher.oldLast) 3336 return false; 3337 return next.match(matcher, i, seq); 3338 } 3339 } 3340 3341 /** 3342 * Node to anchor at the end of a line or the end of input based on the 3343 * multiline mode. 3344 * 3345 * When not in multiline mode, the $ can only match at the very end 3346 * of the input, unless the input ends in a line terminator in which 3347 * it matches right before the last line terminator. 3348 * 3349 * Note that \r\n is considered an atomic line terminator. 3350 * 3351 * Like ^ the $ operator matches at a position, it does not match the 3352 * line terminators themselves. 3353 */ 3354 static final class Dollar extends Node { 3355 boolean multiline; 3356 Dollar(boolean mul) { 3357 multiline = mul; 3358 } 3359 boolean match(Matcher matcher, int i, CharSequence seq) { 3360 int endIndex = (matcher.anchoringBounds) ? 3361 matcher.to : matcher.getTextLength(); 3362 if (!multiline) { 3363 if (i < endIndex - 2) 3364 return false; 3365 if (i == endIndex - 2) { 3366 char ch = seq.charAt(i); 3367 if (ch != '\r') 3368 return false; 3369 ch = seq.charAt(i + 1); 3370 if (ch != '\n') 3371 return false; 3372 } 3373 } 3374 // Matches before any line terminator; also matches at the 3375 // end of input 3376 // Before line terminator: 3377 // If multiline, we match here no matter what 3378 // If not multiline, fall through so that the end 3379 // is marked as hit; this must be a /r/n or a /n 3380 // at the very end so the end was hit; more input 3381 // could make this not match here 3382 if (i < endIndex) { 3383 char ch = seq.charAt(i); 3384 if (ch == '\n') { 3385 // No match between \r\n 3386 if (i > 0 && seq.charAt(i-1) == '\r') 3387 return false; 3388 if (multiline) 3389 return next.match(matcher, i, seq); 3390 } else if (ch == '\r' || ch == '\u0085' || 3391 (ch|1) == '\u2029') { 3392 if (multiline) 3393 return next.match(matcher, i, seq); 3394 } else { // No line terminator, no match 3395 return false; 3396 } 3397 } 3398 // Matched at current end so hit end 3399 matcher.hitEnd = true; 3400 // If a $ matches because of end of input, then more input 3401 // could cause it to fail! 3402 matcher.requireEnd = true; 3403 return next.match(matcher, i, seq); 3404 } 3405 boolean study(TreeInfo info) { 3406 next.study(info); 3407 return info.deterministic; 3408 } 3409 } 3410 3411 /** 3412 * Node to anchor at the end of a line or the end of input based on the 3413 * multiline mode when in unix lines mode. 3414 */ 3415 static final class UnixDollar extends Node { 3416 boolean multiline; 3417 UnixDollar(boolean mul) { 3418 multiline = mul; 3419 } 3420 boolean match(Matcher matcher, int i, CharSequence seq) { 3421 int endIndex = (matcher.anchoringBounds) ? 3422 matcher.to : matcher.getTextLength(); 3423 if (i < endIndex) { 3424 char ch = seq.charAt(i); 3425 if (ch == '\n') { 3426 // If not multiline, then only possible to 3427 // match at very end or one before end 3428 if (multiline == false && i != endIndex - 1) 3429 return false; 3430 // If multiline return next.match without setting 3431 // matcher.hitEnd 3432 if (multiline) 3433 return next.match(matcher, i, seq); 3434 } else { 3435 return false; 3436 } 3437 } 3438 // Matching because at the end or 1 before the end; 3439 // more input could change this so set hitEnd 3440 matcher.hitEnd = true; 3441 // If a $ matches because of end of input, then more input 3442 // could cause it to fail! 3443 matcher.requireEnd = true; 3444 return next.match(matcher, i, seq); 3445 } 3446 boolean study(TreeInfo info) { 3447 next.study(info); 3448 return info.deterministic; 3449 } 3450 } 3451 3452 /** 3453 * Abstract node class to match one character satisfying some 3454 * boolean property. 3455 */ 3456 private static abstract class CharProperty extends Node { 3457 abstract boolean isSatisfiedBy(int ch); 3458 CharProperty complement() { 3459 return new CharProperty() { 3460 boolean isSatisfiedBy(int ch) { 3461 return ! CharProperty.this.isSatisfiedBy(ch);}}; 3462 } 3463 CharProperty maybeComplement(boolean complement) { 3464 return complement ? complement() : this; 3465 } 3466 boolean match(Matcher matcher, int i, CharSequence seq) { 3467 if (i < matcher.to) { 3468 int ch = Character.codePointAt(seq, i); 3469 return isSatisfiedBy(ch) 3470 && next.match(matcher, i+Character.charCount(ch), seq); 3471 } else { 3472 matcher.hitEnd = true; 3473 return false; 3474 } 3475 } 3476 boolean study(TreeInfo info) { 3477 info.minLength++; 3478 info.maxLength++; 3479 return next.study(info); 3480 } 3481 } 3482 3483 /** 3484 * Optimized version of CharProperty that works only for 3485 * properties never satisfied by Supplementary characters. 3486 */ 3487 private static abstract class BmpCharProperty extends CharProperty { 3488 boolean match(Matcher matcher, int i, CharSequence seq) { 3489 if (i < matcher.to) { 3490 return isSatisfiedBy(seq.charAt(i)) 3491 && next.match(matcher, i+1, seq); 3492 } else { 3493 matcher.hitEnd = true; 3494 return false; 3495 } 3496 } 3497 } 3498 3499 /** 3500 * Node class that matches a Supplementary Unicode character 3501 */ 3502 static final class SingleS extends CharProperty { 3503 final int c; 3504 SingleS(int c) { this.c = c; } 3505 boolean isSatisfiedBy(int ch) { 3506 return ch == c; 3507 } 3508 } 3509 3510 /** 3511 * Optimization -- matches a given BMP character 3512 */ 3513 static final class Single extends BmpCharProperty { 3514 final int c; 3515 Single(int c) { this.c = c; } 3516 boolean isSatisfiedBy(int ch) { 3517 return ch == c; 3518 } 3519 } 3520 3521 /** 3522 * Case insensitive matches a given BMP character 3523 */ 3524 static final class SingleI extends BmpCharProperty { 3525 final int lower; 3526 final int upper; 3527 SingleI(int lower, int upper) { 3528 this.lower = lower; 3529 this.upper = upper; 3530 } 3531 boolean isSatisfiedBy(int ch) { 3532 return ch == lower || ch == upper; 3533 } 3534 } 3535 3536 /** 3537 * Unicode case insensitive matches a given Unicode character 3538 */ 3539 static final class SingleU extends CharProperty { 3540 final int lower; 3541 SingleU(int lower) { 3542 this.lower = lower; 3543 } 3544 boolean isSatisfiedBy(int ch) { 3545 return lower == ch || 3546 lower == Character.toLowerCase(Character.toUpperCase(ch)); 3547 } 3548 } 3549 3550 /** 3551 * Node class that matches a Unicode category. 3552 */ 3553 static final class Category extends CharProperty { 3554 final int typeMask; 3555 Category(int typeMask) { this.typeMask = typeMask; } 3556 boolean isSatisfiedBy(int ch) { 3557 return (typeMask & (1 << Character.getType(ch))) != 0; 3558 } 3559 } 3560 3561 /** 3562 * Node class that matches a POSIX type. 3563 */ 3564 static final class Ctype extends BmpCharProperty { 3565 final int ctype; 3566 Ctype(int ctype) { this.ctype = ctype; } 3567 boolean isSatisfiedBy(int ch) { 3568 return ch < 128 && ASCII.isType(ch, ctype); 3569 } 3570 } 3571 3572 /** 3573 * Base class for all Slice nodes 3574 */ 3575 static class SliceNode extends Node { 3576 int[] buffer; 3577 SliceNode(int[] buf) { 3578 buffer = buf; 3579 } 3580 boolean study(TreeInfo info) { 3581 info.minLength += buffer.length; 3582 info.maxLength += buffer.length; 3583 return next.study(info); 3584 } 3585 } 3586 3587 /** 3588 * Node class for a case sensitive/BMP-only sequence of literal 3589 * characters. 3590 */ 3591 static final class Slice extends SliceNode { 3592 Slice(int[] buf) { 3593 super(buf); 3594 } 3595 boolean match(Matcher matcher, int i, CharSequence seq) { 3596 int[] buf = buffer; 3597 int len = buf.length; 3598 for (int j=0; j<len; j++) { 3599 if ((i+j) >= matcher.to) { 3600 matcher.hitEnd = true; 3601 return false; 3602 } 3603 if (buf[j] != seq.charAt(i+j)) 3604 return false; 3605 } 3606 return next.match(matcher, i+len, seq); 3607 } 3608 } 3609 3610 /** 3611 * Node class for a case_insensitive/BMP-only sequence of literal 3612 * characters. 3613 */ 3614 static class SliceI extends SliceNode { 3615 SliceI(int[] buf) { 3616 super(buf); 3617 } 3618 boolean match(Matcher matcher, int i, CharSequence seq) { 3619 int[] buf = buffer; 3620 int len = buf.length; 3621 for (int j=0; j<len; j++) { 3622 if ((i+j) >= matcher.to) { 3623 matcher.hitEnd = true; 3624 return false; 3625 } 3626 int c = seq.charAt(i+j); 3627 if (buf[j] != c && 3628 buf[j] != ASCII.toLower(c)) 3629 return false; 3630 } 3631 return next.match(matcher, i+len, seq); 3632 } 3633 } 3634 3635 /** 3636 * Node class for a unicode_case_insensitive/BMP-only sequence of 3637 * literal characters. Uses unicode case folding. 3638 */ 3639 static final class SliceU extends SliceNode { 3640 SliceU(int[] buf) { 3641 super(buf); 3642 } 3643 boolean match(Matcher matcher, int i, CharSequence seq) { 3644 int[] buf = buffer; 3645 int len = buf.length; 3646 for (int j=0; j<len; j++) { 3647 if ((i+j) >= matcher.to) { 3648 matcher.hitEnd = true; 3649 return false; 3650 } 3651 int c = seq.charAt(i+j); 3652 if (buf[j] != c && 3653 buf[j] != Character.toLowerCase(Character.toUpperCase(c))) 3654 return false; 3655 } 3656 return next.match(matcher, i+len, seq); 3657 } 3658 } 3659 3660 /** 3661 * Node class for a case sensitive sequence of literal characters 3662 * including supplementary characters. 3663 */ 3664 static final class SliceS extends SliceNode { 3665 SliceS(int[] buf) { 3666 super(buf); 3667 } 3668 boolean match(Matcher matcher, int i, CharSequence seq) { 3669 int[] buf = buffer; 3670 int x = i; 3671 for (int j = 0; j < buf.length; j++) { 3672 if (x >= matcher.to) { 3673 matcher.hitEnd = true; 3674 return false; 3675 } 3676 int c = Character.codePointAt(seq, x); 3677 if (buf[j] != c) 3678 return false; 3679 x += Character.charCount(c); 3680 if (x > matcher.to) { 3681 matcher.hitEnd = true; 3682 return false; 3683 } 3684 } 3685 return next.match(matcher, x, seq); 3686 } 3687 } 3688 3689 /** 3690 * Node class for a case insensitive sequence of literal characters 3691 * including supplementary characters. 3692 */ 3693 static class SliceIS extends SliceNode { 3694 SliceIS(int[] buf) { 3695 super(buf); 3696 } 3697 int toLower(int c) { 3698 return ASCII.toLower(c); 3699 } 3700 boolean match(Matcher matcher, int i, CharSequence seq) { 3701 int[] buf = buffer; 3702 int x = i; 3703 for (int j = 0; j < buf.length; j++) { 3704 if (x >= matcher.to) { 3705 matcher.hitEnd = true; 3706 return false; 3707 } 3708 int c = Character.codePointAt(seq, x); 3709 if (buf[j] != c && buf[j] != toLower(c)) 3710 return false; 3711 x += Character.charCount(c); 3712 if (x > matcher.to) { 3713 matcher.hitEnd = true; 3714 return false; 3715 } 3716 } 3717 return next.match(matcher, x, seq); 3718 } 3719 } 3720 3721 /** 3722 * Node class for a case insensitive sequence of literal characters. 3723 * Uses unicode case folding. 3724 */ 3725 static final class SliceUS extends SliceIS { 3726 SliceUS(int[] buf) { 3727 super(buf); 3728 } 3729 int toLower(int c) { 3730 return Character.toLowerCase(Character.toUpperCase(c)); 3731 } 3732 } 3733 3734 private static boolean inRange(int lower, int ch, int upper) { 3735 return lower <= ch && ch <= upper; 3736 } 3737 3738 /** 3739 * Returns node for matching characters within an explicit value range. 3740 */ 3741 private static CharProperty rangeFor(final int lower, 3742 final int upper) { 3743 return new CharProperty() { 3744 boolean isSatisfiedBy(int ch) { 3745 return inRange(lower, ch, upper);}}; 3746 } 3747 3748 /** 3749 * Returns node for matching characters within an explicit value 3750 * range in a case insensitive manner. 3751 */ 3752 private CharProperty caseInsensitiveRangeFor(final int lower, 3753 final int upper) { 3754 if (has(UNICODE_CASE)) 3755 return new CharProperty() { 3756 boolean isSatisfiedBy(int ch) { 3757 if (inRange(lower, ch, upper)) 3758 return true; 3759 int up = Character.toUpperCase(ch); 3760 return inRange(lower, up, upper) || 3761 inRange(lower, Character.toLowerCase(up), upper);}}; 3762 return new CharProperty() { 3763 boolean isSatisfiedBy(int ch) { 3764 return inRange(lower, ch, upper) || 3765 ASCII.isAscii(ch) && 3766 (inRange(lower, ASCII.toUpper(ch), upper) || 3767 inRange(lower, ASCII.toLower(ch), upper)); 3768 }}; 3769 } 3770 3771 /** 3772 * Implements the Unicode category ALL and the dot metacharacter when 3773 * in dotall mode. 3774 */ 3775 static final class All extends CharProperty { 3776 boolean isSatisfiedBy(int ch) { 3777 return true; 3778 } 3779 } 3780 3781 /** 3782 * Node class for the dot metacharacter when dotall is not enabled. 3783 */ 3784 static final class Dot extends CharProperty { 3785 boolean isSatisfiedBy(int ch) { 3786 return (ch != '\n' && ch != '\r' 3787 && (ch|1) != '\u2029' 3788 && ch != '\u0085'); 3789 } 3790 } 3791 3792 /** 3793 * Node class for the dot metacharacter when dotall is not enabled 3794 * but UNIX_LINES is enabled. 3795 */ 3796 static final class UnixDot extends CharProperty { 3797 boolean isSatisfiedBy(int ch) { 3798 return ch != '\n'; 3799 } 3800 } 3801 3802 /** 3803 * The 0 or 1 quantifier. This one class implements all three types. 3804 */ 3805 static final class Ques extends Node { 3806 Node atom; 3807 int type; 3808 Ques(Node node, int type) { 3809 this.atom = node; 3810 this.type = type; 3811 } 3812 boolean match(Matcher matcher, int i, CharSequence seq) { 3813 switch (type) { 3814 case GREEDY: 3815 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)) 3816 || next.match(matcher, i, seq); 3817 case LAZY: 3818 return next.match(matcher, i, seq) 3819 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)); 3820 case POSSESSIVE: 3821 if (atom.match(matcher, i, seq)) i = matcher.last; 3822 return next.match(matcher, i, seq); 3823 default: 3824 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq); 3825 } 3826 } 3827 boolean study(TreeInfo info) { 3828 if (type != INDEPENDENT) { 3829 int minL = info.minLength; 3830 atom.study(info); 3831 info.minLength = minL; 3832 info.deterministic = false; 3833 return next.study(info); 3834 } else { 3835 atom.study(info); 3836 return next.study(info); 3837 } 3838 } 3839 } 3840 3841 /** 3842 * Handles the curly-brace style repetition with a specified minimum and 3843 * maximum occurrences. The * quantifier is handled as a special case. 3844 * This class handles the three types. 3845 */ 3846 static final class Curly extends Node { 3847 Node atom; 3848 int type; 3849 int cmin; 3850 int cmax; 3851 3852 Curly(Node node, int cmin, int cmax, int type) { 3853 this.atom = node; 3854 this.type = type; 3855 this.cmin = cmin; 3856 this.cmax = cmax; 3857 } 3858 boolean match(Matcher matcher, int i, CharSequence seq) { 3859 int j; 3860 for (j = 0; j < cmin; j++) { 3861 if (atom.match(matcher, i, seq)) { 3862 i = matcher.last; 3863 continue; 3864 } 3865 return false; 3866 } 3867 if (type == GREEDY) 3868 return match0(matcher, i, j, seq); 3869 else if (type == LAZY) 3870 return match1(matcher, i, j, seq); 3871 else 3872 return match2(matcher, i, j, seq); 3873 } 3874 // Greedy match. 3875 // i is the index to start matching at 3876 // j is the number of atoms that have matched 3877 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 3878 if (j >= cmax) { 3879 // We have matched the maximum... continue with the rest of 3880 // the regular expression 3881 return next.match(matcher, i, seq); 3882 } 3883 int backLimit = j; 3884 while (atom.match(matcher, i, seq)) { 3885 // k is the length of this match 3886 int k = matcher.last - i; 3887 if (k == 0) // Zero length match 3888 break; 3889 // Move up index and number matched 3890 i = matcher.last; 3891 j++; 3892 // We are greedy so match as many as we can 3893 while (j < cmax) { 3894 if (!atom.match(matcher, i, seq)) 3895 break; 3896 if (i + k != matcher.last) { 3897 if (match0(matcher, matcher.last, j+1, seq)) 3898 return true; 3899 break; 3900 } 3901 i += k; 3902 j++; 3903 } 3904 // Handle backing off if match fails 3905 while (j >= backLimit) { 3906 if (next.match(matcher, i, seq)) 3907 return true; 3908 i -= k; 3909 j--; 3910 } 3911 return false; 3912 } 3913 return next.match(matcher, i, seq); 3914 } 3915 // Reluctant match. At this point, the minimum has been satisfied. 3916 // i is the index to start matching at 3917 // j is the number of atoms that have matched 3918 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 3919 for (;;) { 3920 // Try finishing match without consuming any more 3921 if (next.match(matcher, i, seq)) 3922 return true; 3923 // At the maximum, no match found 3924 if (j >= cmax) 3925 return false; 3926 // Okay, must try one more atom 3927 if (!atom.match(matcher, i, seq)) 3928 return false; 3929 // If we haven't moved forward then must break out 3930 if (i == matcher.last) 3931 return false; 3932 // Move up index and number matched 3933 i = matcher.last; 3934 j++; 3935 } 3936 } 3937 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 3938 for (; j < cmax; j++) { 3939 if (!atom.match(matcher, i, seq)) 3940 break; 3941 if (i == matcher.last) 3942 break; 3943 i = matcher.last; 3944 } 3945 return next.match(matcher, i, seq); 3946 } 3947 boolean study(TreeInfo info) { 3948 // Save original info 3949 int minL = info.minLength; 3950 int maxL = info.maxLength; 3951 boolean maxV = info.maxValid; 3952 boolean detm = info.deterministic; 3953 info.reset(); 3954 3955 atom.study(info); 3956 3957 int temp = info.minLength * cmin + minL; 3958 if (temp < minL) { 3959 temp = 0xFFFFFFF; // arbitrary large number 3960 } 3961 info.minLength = temp; 3962 3963 if (maxV & info.maxValid) { 3964 temp = info.maxLength * cmax + maxL; 3965 info.maxLength = temp; 3966 if (temp < maxL) { 3967 info.maxValid = false; 3968 } 3969 } else { 3970 info.maxValid = false; 3971 } 3972 3973 if (info.deterministic && cmin == cmax) 3974 info.deterministic = detm; 3975 else 3976 info.deterministic = false; 3977 3978 return next.study(info); 3979 } 3980 } 3981 3982 /** 3983 * Handles the curly-brace style repetition with a specified minimum and 3984 * maximum occurrences in deterministic cases. This is an iterative 3985 * optimization over the Prolog and Loop system which would handle this 3986 * in a recursive way. The * quantifier is handled as a special case. 3987 * If capture is true then this class saves group settings and ensures 3988 * that groups are unset when backing off of a group match. 3989 */ 3990 static final class GroupCurly extends Node { 3991 Node atom; 3992 int type; 3993 int cmin; 3994 int cmax; 3995 int localIndex; 3996 int groupIndex; 3997 boolean capture; 3998 3999 GroupCurly(Node node, int cmin, int cmax, int type, int local, 4000 int group, boolean capture) { 4001 this.atom = node; 4002 this.type = type; 4003 this.cmin = cmin; 4004 this.cmax = cmax; 4005 this.localIndex = local; 4006 this.groupIndex = group; 4007 this.capture = capture; 4008 } 4009 boolean match(Matcher matcher, int i, CharSequence seq) { 4010 int[] groups = matcher.groups; 4011 int[] locals = matcher.locals; 4012 int save0 = locals[localIndex]; 4013 int save1 = 0; 4014 int save2 = 0; 4015 4016 if (capture) { 4017 save1 = groups[groupIndex]; 4018 save2 = groups[groupIndex+1]; 4019 } 4020 4021 // Notify GroupTail there is no need to setup group info 4022 // because it will be set here 4023 locals[localIndex] = -1; 4024 4025 boolean ret = true; 4026 for (int j = 0; j < cmin; j++) { 4027 if (atom.match(matcher, i, seq)) { 4028 if (capture) { 4029 groups[groupIndex] = i; 4030 groups[groupIndex+1] = matcher.last; 4031 } 4032 i = matcher.last; 4033 } else { 4034 ret = false; 4035 break; 4036 } 4037 } 4038 if (ret) { 4039 if (type == GREEDY) { 4040 ret = match0(matcher, i, cmin, seq); 4041 } else if (type == LAZY) { 4042 ret = match1(matcher, i, cmin, seq); 4043 } else { 4044 ret = match2(matcher, i, cmin, seq); 4045 } 4046 } 4047 if (!ret) { 4048 locals[localIndex] = save0; 4049 if (capture) { 4050 groups[groupIndex] = save1; 4051 groups[groupIndex+1] = save2; 4052 } 4053 } 4054 return ret; 4055 } 4056 // Aggressive group match 4057 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4058 int[] groups = matcher.groups; 4059 int save0 = 0; 4060 int save1 = 0; 4061 if (capture) { 4062 save0 = groups[groupIndex]; 4063 save1 = groups[groupIndex+1]; 4064 } 4065 for (;;) { 4066 if (j >= cmax) 4067 break; 4068 if (!atom.match(matcher, i, seq)) 4069 break; 4070 int k = matcher.last - i; 4071 if (k <= 0) { 4072 if (capture) { 4073 groups[groupIndex] = i; 4074 groups[groupIndex+1] = i + k; 4075 } 4076 i = i + k; 4077 break; 4078 } 4079 for (;;) { 4080 if (capture) { 4081 groups[groupIndex] = i; 4082 groups[groupIndex+1] = i + k; 4083 } 4084 i = i + k; 4085 if (++j >= cmax) 4086 break; 4087 if (!atom.match(matcher, i, seq)) 4088 break; 4089 if (i + k != matcher.last) { 4090 if (match0(matcher, i, j, seq)) 4091 return true; 4092 break; 4093 } 4094 } 4095 while (j > cmin) { 4096 if (next.match(matcher, i, seq)) { 4097 if (capture) { 4098 groups[groupIndex+1] = i; 4099 groups[groupIndex] = i - k; 4100 } 4101 i = i - k; 4102 return true; 4103 } 4104 // backing off 4105 if (capture) { 4106 groups[groupIndex+1] = i; 4107 groups[groupIndex] = i - k; 4108 } 4109 i = i - k; 4110 j--; 4111 } 4112 break; 4113 } 4114 if (capture) { 4115 groups[groupIndex] = save0; 4116 groups[groupIndex+1] = save1; 4117 } 4118 return next.match(matcher, i, seq); 4119 } 4120 // Reluctant matching 4121 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4122 for (;;) { 4123 if (next.match(matcher, i, seq)) 4124 return true; 4125 if (j >= cmax) 4126 return false; 4127 if (!atom.match(matcher, i, seq)) 4128 return false; 4129 if (i == matcher.last) 4130 return false; 4131 if (capture) { 4132 matcher.groups[groupIndex] = i; 4133 matcher.groups[groupIndex+1] = matcher.last; 4134 } 4135 i = matcher.last; 4136 j++; 4137 } 4138 } 4139 // Possessive matching 4140 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4141 for (; j < cmax; j++) { 4142 if (!atom.match(matcher, i, seq)) { 4143 break; 4144 } 4145 if (capture) { 4146 matcher.groups[groupIndex] = i; 4147 matcher.groups[groupIndex+1] = matcher.last; 4148 } 4149 if (i == matcher.last) { 4150 break; 4151 } 4152 i = matcher.last; 4153 } 4154 return next.match(matcher, i, seq); 4155 } 4156 boolean study(TreeInfo info) { 4157 // Save original info 4158 int minL = info.minLength; 4159 int maxL = info.maxLength; 4160 boolean maxV = info.maxValid; 4161 boolean detm = info.deterministic; 4162 info.reset(); 4163 4164 atom.study(info); 4165 4166 int temp = info.minLength * cmin + minL; 4167 if (temp < minL) { 4168 temp = 0xFFFFFFF; // Arbitrary large number 4169 } 4170 info.minLength = temp; 4171 4172 if (maxV & info.maxValid) { 4173 temp = info.maxLength * cmax + maxL; 4174 info.maxLength = temp; 4175 if (temp < maxL) { 4176 info.maxValid = false; 4177 } 4178 } else { 4179 info.maxValid = false; 4180 } 4181 4182 if (info.deterministic && cmin == cmax) { 4183 info.deterministic = detm; 4184 } else { 4185 info.deterministic = false; 4186 } 4187 4188 return next.study(info); 4189 } 4190 } 4191 4192 /** 4193 * A Guard node at the end of each atom node in a Branch. It 4194 * serves the purpose of chaining the "match" operation to 4195 * "next" but not the "study", so we can collect the TreeInfo 4196 * of each atom node without including the TreeInfo of the 4197 * "next". 4198 */ 4199 static final class BranchConn extends Node { 4200 BranchConn() {}; 4201 boolean match(Matcher matcher, int i, CharSequence seq) { 4202 return next.match(matcher, i, seq); 4203 } 4204 boolean study(TreeInfo info) { 4205 return info.deterministic; 4206 } 4207 } 4208 4209 /** 4210 * Handles the branching of alternations. Note this is also used for 4211 * the ? quantifier to branch between the case where it matches once 4212 * and where it does not occur. 4213 */ 4214 static final class Branch extends Node { 4215 Node[] atoms = new Node[2]; 4216 int size = 2; 4217 Node conn; 4218 Branch(Node first, Node second, Node branchConn) { 4219 conn = branchConn; 4220 atoms[0] = first; 4221 atoms[1] = second; 4222 } 4223 4224 void add(Node node) { 4225 if (size >= atoms.length) { 4226 Node[] tmp = new Node[atoms.length*2]; 4227 System.arraycopy(atoms, 0, tmp, 0, atoms.length); 4228 atoms = tmp; 4229 } 4230 atoms[size++] = node; 4231 } 4232 4233 boolean match(Matcher matcher, int i, CharSequence seq) { 4234 for (int n = 0; n < size; n++) { 4235 if (atoms[n] == null) { 4236 if (conn.next.match(matcher, i, seq)) 4237 return true; 4238 } else if (atoms[n].match(matcher, i, seq)) { 4239 return true; 4240 } 4241 } 4242 return false; 4243 } 4244 4245 boolean study(TreeInfo info) { 4246 int minL = info.minLength; 4247 int maxL = info.maxLength; 4248 boolean maxV = info.maxValid; 4249 4250 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num 4251 int maxL2 = -1; 4252 for (int n = 0; n < size; n++) { 4253 info.reset(); 4254 if (atoms[n] != null) 4255 atoms[n].study(info); 4256 minL2 = Math.min(minL2, info.minLength); 4257 maxL2 = Math.max(maxL2, info.maxLength); 4258 maxV = (maxV & info.maxValid); 4259 } 4260 4261 minL += minL2; 4262 maxL += maxL2; 4263 4264 info.reset(); 4265 conn.next.study(info); 4266 4267 info.minLength += minL; 4268 info.maxLength += maxL; 4269 info.maxValid &= maxV; 4270 info.deterministic = false; 4271 return false; 4272 } 4273 } 4274 4275 /** 4276 * The GroupHead saves the location where the group begins in the locals 4277 * and restores them when the match is done. 4278 * 4279 * The matchRef is used when a reference to this group is accessed later 4280 * in the expression. The locals will have a negative value in them to 4281 * indicate that we do not want to unset the group if the reference 4282 * doesn't match. 4283 */ 4284 static final class GroupHead extends Node { 4285 int localIndex; 4286 GroupHead(int localCount) { 4287 localIndex = localCount; 4288 } 4289 boolean match(Matcher matcher, int i, CharSequence seq) { 4290 int save = matcher.locals[localIndex]; 4291 matcher.locals[localIndex] = i; 4292 boolean ret = next.match(matcher, i, seq); 4293 matcher.locals[localIndex] = save; 4294 return ret; 4295 } 4296 boolean matchRef(Matcher matcher, int i, CharSequence seq) { 4297 int save = matcher.locals[localIndex]; 4298 matcher.locals[localIndex] = ~i; // HACK 4299 boolean ret = next.match(matcher, i, seq); 4300 matcher.locals[localIndex] = save; 4301 return ret; 4302 } 4303 } 4304 4305 /** 4306 * Recursive reference to a group in the regular expression. It calls 4307 * matchRef because if the reference fails to match we would not unset 4308 * the group. 4309 */ 4310 static final class GroupRef extends Node { 4311 GroupHead head; 4312 GroupRef(GroupHead head) { 4313 this.head = head; 4314 } 4315 boolean match(Matcher matcher, int i, CharSequence seq) { 4316 return head.matchRef(matcher, i, seq) 4317 && next.match(matcher, matcher.last, seq); 4318 } 4319 boolean study(TreeInfo info) { 4320 info.maxValid = false; 4321 info.deterministic = false; 4322 return next.study(info); 4323 } 4324 } 4325 4326 /** 4327 * The GroupTail handles the setting of group beginning and ending 4328 * locations when groups are successfully matched. It must also be able to 4329 * unset groups that have to be backed off of. 4330 * 4331 * The GroupTail node is also used when a previous group is referenced, 4332 * and in that case no group information needs to be set. 4333 */ 4334 static final class GroupTail extends Node { 4335 int localIndex; 4336 int groupIndex; 4337 GroupTail(int localCount, int groupCount) { 4338 localIndex = localCount; 4339 groupIndex = groupCount + groupCount; 4340 } 4341 boolean match(Matcher matcher, int i, CharSequence seq) { 4342 int tmp = matcher.locals[localIndex]; 4343 if (tmp >= 0) { // This is the normal group case. 4344 // Save the group so we can unset it if it 4345 // backs off of a match. 4346 int groupStart = matcher.groups[groupIndex]; 4347 int groupEnd = matcher.groups[groupIndex+1]; 4348 4349 matcher.groups[groupIndex] = tmp; 4350 matcher.groups[groupIndex+1] = i; 4351 if (next.match(matcher, i, seq)) { 4352 return true; 4353 } 4354 matcher.groups[groupIndex] = groupStart; 4355 matcher.groups[groupIndex+1] = groupEnd; 4356 return false; 4357 } else { 4358 // This is a group reference case. We don't need to save any 4359 // group info because it isn't really a group. 4360 matcher.last = i; 4361 return true; 4362 } 4363 } 4364 } 4365 4366 /** 4367 * This sets up a loop to handle a recursive quantifier structure. 4368 */ 4369 static final class Prolog extends Node { 4370 Loop loop; 4371 Prolog(Loop loop) { 4372 this.loop = loop; 4373 } 4374 boolean match(Matcher matcher, int i, CharSequence seq) { 4375 return loop.matchInit(matcher, i, seq); 4376 } 4377 boolean study(TreeInfo info) { 4378 return loop.study(info); 4379 } 4380 } 4381 4382 /** 4383 * Handles the repetition count for a greedy Curly. The matchInit 4384 * is called from the Prolog to save the index of where the group 4385 * beginning is stored. A zero length group check occurs in the 4386 * normal match but is skipped in the matchInit. 4387 */ 4388 static class Loop extends Node { 4389 Node body; 4390 int countIndex; // local count index in matcher locals 4391 int beginIndex; // group beginning index 4392 int cmin, cmax; 4393 Loop(int countIndex, int beginIndex) { 4394 this.countIndex = countIndex; 4395 this.beginIndex = beginIndex; 4396 } 4397 boolean match(Matcher matcher, int i, CharSequence seq) { 4398 // Avoid infinite loop in zero-length case. 4399 if (i > matcher.locals[beginIndex]) { 4400 int count = matcher.locals[countIndex]; 4401 4402 // This block is for before we reach the minimum 4403 // iterations required for the loop to match 4404 if (count < cmin) { 4405 matcher.locals[countIndex] = count + 1; 4406 boolean b = body.match(matcher, i, seq); 4407 // If match failed we must backtrack, so 4408 // the loop count should NOT be incremented 4409 if (!b) 4410 matcher.locals[countIndex] = count; 4411 // Return success or failure since we are under 4412 // minimum 4413 return b; 4414 } 4415 // This block is for after we have the minimum 4416 // iterations required for the loop to match 4417 if (count < cmax) { 4418 matcher.locals[countIndex] = count + 1; 4419 boolean b = body.match(matcher, i, seq); 4420 // If match failed we must backtrack, so 4421 // the loop count should NOT be incremented 4422 if (!b) 4423 matcher.locals[countIndex] = count; 4424 else 4425 return true; 4426 } 4427 } 4428 return next.match(matcher, i, seq); 4429 } 4430 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4431 int save = matcher.locals[countIndex]; 4432 boolean ret = false; 4433 if (0 < cmin) { 4434 matcher.locals[countIndex] = 1; 4435 ret = body.match(matcher, i, seq); 4436 } else if (0 < cmax) { 4437 matcher.locals[countIndex] = 1; 4438 ret = body.match(matcher, i, seq); 4439 if (ret == false) 4440 ret = next.match(matcher, i, seq); 4441 } else { 4442 ret = next.match(matcher, i, seq); 4443 } 4444 matcher.locals[countIndex] = save; 4445 return ret; 4446 } 4447 boolean study(TreeInfo info) { 4448 info.maxValid = false; 4449 info.deterministic = false; 4450 return false; 4451 } 4452 } 4453 4454 /** 4455 * Handles the repetition count for a reluctant Curly. The matchInit 4456 * is called from the Prolog to save the index of where the group 4457 * beginning is stored. A zero length group check occurs in the 4458 * normal match but is skipped in the matchInit. 4459 */ 4460 static final class LazyLoop extends Loop { 4461 LazyLoop(int countIndex, int beginIndex) { 4462 super(countIndex, beginIndex); 4463 } 4464 boolean match(Matcher matcher, int i, CharSequence seq) { 4465 // Check for zero length group 4466 if (i > matcher.locals[beginIndex]) { 4467 int count = matcher.locals[countIndex]; 4468 if (count < cmin) { 4469 matcher.locals[countIndex] = count + 1; 4470 boolean result = body.match(matcher, i, seq); 4471 // If match failed we must backtrack, so 4472 // the loop count should NOT be incremented 4473 if (!result) 4474 matcher.locals[countIndex] = count; 4475 return result; 4476 } 4477 if (next.match(matcher, i, seq)) 4478 return true; 4479 if (count < cmax) { 4480 matcher.locals[countIndex] = count + 1; 4481 boolean result = body.match(matcher, i, seq); 4482 // If match failed we must backtrack, so 4483 // the loop count should NOT be incremented 4484 if (!result) 4485 matcher.locals[countIndex] = count; 4486 return result; 4487 } 4488 return false; 4489 } 4490 return next.match(matcher, i, seq); 4491 } 4492 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4493 int save = matcher.locals[countIndex]; 4494 boolean ret = false; 4495 if (0 < cmin) { 4496 matcher.locals[countIndex] = 1; 4497 ret = body.match(matcher, i, seq); 4498 } else if (next.match(matcher, i, seq)) { 4499 ret = true; 4500 } else if (0 < cmax) { 4501 matcher.locals[countIndex] = 1; 4502 ret = body.match(matcher, i, seq); 4503 } 4504 matcher.locals[countIndex] = save; 4505 return ret; 4506 } 4507 boolean study(TreeInfo info) { 4508 info.maxValid = false; 4509 info.deterministic = false; 4510 return false; 4511 } 4512 } 4513 4514 /** 4515 * Refers to a group in the regular expression. Attempts to match 4516 * whatever the group referred to last matched. 4517 */ 4518 static class BackRef extends Node { 4519 int groupIndex; 4520 BackRef(int groupCount) { 4521 super(); 4522 groupIndex = groupCount + groupCount; 4523 } 4524 boolean match(Matcher matcher, int i, CharSequence seq) { 4525 int j = matcher.groups[groupIndex]; 4526 int k = matcher.groups[groupIndex+1]; 4527 4528 int groupSize = k - j; 4529 4530 // If the referenced group didn't match, neither can this 4531 if (j < 0) 4532 return false; 4533 4534 // If there isn't enough input left no match 4535 if (i + groupSize > matcher.to) { 4536 matcher.hitEnd = true; 4537 return false; 4538 } 4539 4540 // Check each new char to make sure it matches what the group 4541 // referenced matched last time around 4542 for (int index=0; index<groupSize; index++) 4543 if (seq.charAt(i+index) != seq.charAt(j+index)) 4544 return false; 4545 4546 return next.match(matcher, i+groupSize, seq); 4547 } 4548 boolean study(TreeInfo info) { 4549 info.maxValid = false; 4550 return next.study(info); 4551 } 4552 } 4553 4554 static class CIBackRef extends Node { 4555 int groupIndex; 4556 boolean doUnicodeCase; 4557 CIBackRef(int groupCount, boolean doUnicodeCase) { 4558 super(); 4559 groupIndex = groupCount + groupCount; 4560 this.doUnicodeCase = doUnicodeCase; 4561 } 4562 boolean match(Matcher matcher, int i, CharSequence seq) { 4563 int j = matcher.groups[groupIndex]; 4564 int k = matcher.groups[groupIndex+1]; 4565 4566 int groupSize = k - j; 4567 4568 // If the referenced group didn't match, neither can this 4569 if (j < 0) 4570 return false; 4571 4572 // If there isn't enough input left no match 4573 if (i + groupSize > matcher.to) { 4574 matcher.hitEnd = true; 4575 return false; 4576 } 4577 4578 // Check each new char to make sure it matches what the group 4579 // referenced matched last time around 4580 int x = i; 4581 for (int index=0; index<groupSize; index++) { 4582 int c1 = Character.codePointAt(seq, x); 4583 int c2 = Character.codePointAt(seq, j); 4584 if (c1 != c2) { 4585 if (doUnicodeCase) { 4586 int cc1 = Character.toUpperCase(c1); 4587 int cc2 = Character.toUpperCase(c2); 4588 if (cc1 != cc2 && 4589 Character.toLowerCase(cc1) != 4590 Character.toLowerCase(cc2)) 4591 return false; 4592 } else { 4593 if (ASCII.toLower(c1) != ASCII.toLower(c2)) 4594 return false; 4595 } 4596 } 4597 x += Character.charCount(c1); 4598 j += Character.charCount(c2); 4599 } 4600 4601 return next.match(matcher, i+groupSize, seq); 4602 } 4603 boolean study(TreeInfo info) { 4604 info.maxValid = false; 4605 return next.study(info); 4606 } 4607 } 4608 4609 /** 4610 * Searches until the next instance of its atom. This is useful for 4611 * finding the atom efficiently without passing an instance of it 4612 * (greedy problem) and without a lot of wasted search time (reluctant 4613 * problem). 4614 */ 4615 static final class First extends Node { 4616 Node atom; 4617 First(Node node) { 4618 this.atom = BnM.optimize(node); 4619 } 4620 boolean match(Matcher matcher, int i, CharSequence seq) { 4621 if (atom instanceof BnM) { 4622 return atom.match(matcher, i, seq) 4623 && next.match(matcher, matcher.last, seq); 4624 } 4625 for (;;) { 4626 if (i > matcher.to) { 4627 matcher.hitEnd = true; 4628 return false; 4629 } 4630 if (atom.match(matcher, i, seq)) { 4631 return next.match(matcher, matcher.last, seq); 4632 } 4633 i += countChars(seq, i, 1); 4634 matcher.first++; 4635 } 4636 } 4637 boolean study(TreeInfo info) { 4638 atom.study(info); 4639 info.maxValid = false; 4640 info.deterministic = false; 4641 return next.study(info); 4642 } 4643 } 4644 4645 static final class Conditional extends Node { 4646 Node cond, yes, not; 4647 Conditional(Node cond, Node yes, Node not) { 4648 this.cond = cond; 4649 this.yes = yes; 4650 this.not = not; 4651 } 4652 boolean match(Matcher matcher, int i, CharSequence seq) { 4653 if (cond.match(matcher, i, seq)) { 4654 return yes.match(matcher, i, seq); 4655 } else { 4656 return not.match(matcher, i, seq); 4657 } 4658 } 4659 boolean study(TreeInfo info) { 4660 int minL = info.minLength; 4661 int maxL = info.maxLength; 4662 boolean maxV = info.maxValid; 4663 info.reset(); 4664 yes.study(info); 4665 4666 int minL2 = info.minLength; 4667 int maxL2 = info.maxLength; 4668 boolean maxV2 = info.maxValid; 4669 info.reset(); 4670 not.study(info); 4671 4672 info.minLength = minL + Math.min(minL2, info.minLength); 4673 info.maxLength = maxL + Math.max(maxL2, info.maxLength); 4674 info.maxValid = (maxV & maxV2 & info.maxValid); 4675 info.deterministic = false; 4676 return next.study(info); 4677 } 4678 } 4679 4680 /** 4681 * Zero width positive lookahead. 4682 */ 4683 static final class Pos extends Node { 4684 Node cond; 4685 Pos(Node cond) { 4686 this.cond = cond; 4687 } 4688 boolean match(Matcher matcher, int i, CharSequence seq) { 4689 int savedTo = matcher.to; 4690 boolean conditionMatched = false; 4691 4692 // Relax transparent region boundaries for lookahead 4693 if (matcher.transparentBounds) 4694 matcher.to = matcher.getTextLength(); 4695 try { 4696 conditionMatched = cond.match(matcher, i, seq); 4697 } finally { 4698 // Reinstate region boundaries 4699 matcher.to = savedTo; 4700 } 4701 return conditionMatched && next.match(matcher, i, seq); 4702 } 4703 } 4704 4705 /** 4706 * Zero width negative lookahead. 4707 */ 4708 static final class Neg extends Node { 4709 Node cond; 4710 Neg(Node cond) { 4711 this.cond = cond; 4712 } 4713 boolean match(Matcher matcher, int i, CharSequence seq) { 4714 int savedTo = matcher.to; 4715 boolean conditionMatched = false; 4716 4717 // Relax transparent region boundaries for lookahead 4718 if (matcher.transparentBounds) 4719 matcher.to = matcher.getTextLength(); 4720 try { 4721 if (i < matcher.to) { 4722 conditionMatched = !cond.match(matcher, i, seq); 4723 } else { 4724 // If a negative lookahead succeeds then more input 4725 // could cause it to fail! 4726 matcher.requireEnd = true; 4727 conditionMatched = !cond.match(matcher, i, seq); 4728 } 4729 } finally { 4730 // Reinstate region boundaries 4731 matcher.to = savedTo; 4732 } 4733 return conditionMatched && next.match(matcher, i, seq); 4734 } 4735 } 4736 4737 /** 4738 * For use with lookbehinds; matches the position where the lookbehind 4739 * was encountered. 4740 */ 4741 static Node lookbehindEnd = new Node() { 4742 boolean match(Matcher matcher, int i, CharSequence seq) { 4743 return i == matcher.lookbehindTo; 4744 } 4745 }; 4746 4747 /** 4748 * Zero width positive lookbehind. 4749 */ 4750 static class Behind extends Node { 4751 Node cond; 4752 int rmax, rmin; 4753 Behind(Node cond, int rmax, int rmin) { 4754 this.cond = cond; 4755 this.rmax = rmax; 4756 this.rmin = rmin; 4757 } 4758 4759 boolean match(Matcher matcher, int i, CharSequence seq) { 4760 int savedFrom = matcher.from; 4761 boolean conditionMatched = false; 4762 int startIndex = (!matcher.transparentBounds) ? 4763 matcher.from : 0; 4764 int from = Math.max(i - rmax, startIndex); 4765 // Set end boundary 4766 int savedLBT = matcher.lookbehindTo; 4767 matcher.lookbehindTo = i; 4768 // Relax transparent region boundaries for lookbehind 4769 if (matcher.transparentBounds) 4770 matcher.from = 0; 4771 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 4772 conditionMatched = cond.match(matcher, j, seq); 4773 } 4774 matcher.from = savedFrom; 4775 matcher.lookbehindTo = savedLBT; 4776 return conditionMatched && next.match(matcher, i, seq); 4777 } 4778 } 4779 4780 /** 4781 * Zero width positive lookbehind, including supplementary 4782 * characters or unpaired surrogates. 4783 */ 4784 static final class BehindS extends Behind { 4785 BehindS(Node cond, int rmax, int rmin) { 4786 super(cond, rmax, rmin); 4787 } 4788 boolean match(Matcher matcher, int i, CharSequence seq) { 4789 int rmaxChars = countChars(seq, i, -rmax); 4790 int rminChars = countChars(seq, i, -rmin); 4791 int savedFrom = matcher.from; 4792 int startIndex = (!matcher.transparentBounds) ? 4793 matcher.from : 0; 4794 boolean conditionMatched = false; 4795 int from = Math.max(i - rmaxChars, startIndex); 4796 // Set end boundary 4797 int savedLBT = matcher.lookbehindTo; 4798 matcher.lookbehindTo = i; 4799 // Relax transparent region boundaries for lookbehind 4800 if (matcher.transparentBounds) 4801 matcher.from = 0; 4802 4803 for (int j = i - rminChars; 4804 !conditionMatched && j >= from; 4805 j -= j>from ? countChars(seq, j, -1) : 1) { 4806 conditionMatched = cond.match(matcher, j, seq); 4807 } 4808 matcher.from = savedFrom; 4809 matcher.lookbehindTo = savedLBT; 4810 return conditionMatched && next.match(matcher, i, seq); 4811 } 4812 } 4813 4814 /** 4815 * Zero width negative lookbehind. 4816 */ 4817 static class NotBehind extends Node { 4818 Node cond; 4819 int rmax, rmin; 4820 NotBehind(Node cond, int rmax, int rmin) { 4821 this.cond = cond; 4822 this.rmax = rmax; 4823 this.rmin = rmin; 4824 } 4825 4826 boolean match(Matcher matcher, int i, CharSequence seq) { 4827 int savedLBT = matcher.lookbehindTo; 4828 int savedFrom = matcher.from; 4829 boolean conditionMatched = false; 4830 int startIndex = (!matcher.transparentBounds) ? 4831 matcher.from : 0; 4832 int from = Math.max(i - rmax, startIndex); 4833 matcher.lookbehindTo = i; 4834 // Relax transparent region boundaries for lookbehind 4835 if (matcher.transparentBounds) 4836 matcher.from = 0; 4837 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 4838 conditionMatched = cond.match(matcher, j, seq); 4839 } 4840 // Reinstate region boundaries 4841 matcher.from = savedFrom; 4842 matcher.lookbehindTo = savedLBT; 4843 return !conditionMatched && next.match(matcher, i, seq); 4844 } 4845 } 4846 4847 /** 4848 * Zero width negative lookbehind, including supplementary 4849 * characters or unpaired surrogates. 4850 */ 4851 static final class NotBehindS extends NotBehind { 4852 NotBehindS(Node cond, int rmax, int rmin) { 4853 super(cond, rmax, rmin); 4854 } 4855 boolean match(Matcher matcher, int i, CharSequence seq) { 4856 int rmaxChars = countChars(seq, i, -rmax); 4857 int rminChars = countChars(seq, i, -rmin); 4858 int savedFrom = matcher.from; 4859 int savedLBT = matcher.lookbehindTo; 4860 boolean conditionMatched = false; 4861 int startIndex = (!matcher.transparentBounds) ? 4862 matcher.from : 0; 4863 int from = Math.max(i - rmaxChars, startIndex); 4864 matcher.lookbehindTo = i; 4865 // Relax transparent region boundaries for lookbehind 4866 if (matcher.transparentBounds) 4867 matcher.from = 0; 4868 for (int j = i - rminChars; 4869 !conditionMatched && j >= from; 4870 j -= j>from ? countChars(seq, j, -1) : 1) { 4871 conditionMatched = cond.match(matcher, j, seq); 4872 } 4873 //Reinstate region boundaries 4874 matcher.from = savedFrom; 4875 matcher.lookbehindTo = savedLBT; 4876 return !conditionMatched && next.match(matcher, i, seq); 4877 } 4878 } 4879 4880 /** 4881 * Returns the set union of two CharProperty nodes. 4882 */ 4883 private static CharProperty union(final CharProperty lhs, 4884 final CharProperty rhs) { 4885 return new CharProperty() { 4886 boolean isSatisfiedBy(int ch) { 4887 return lhs.isSatisfiedBy(ch) || rhs.isSatisfiedBy(ch);}}; 4888 } 4889 4890 /** 4891 * Returns the set intersection of two CharProperty nodes. 4892 */ 4893 private static CharProperty intersection(final CharProperty lhs, 4894 final CharProperty rhs) { 4895 return new CharProperty() { 4896 boolean isSatisfiedBy(int ch) { 4897 return lhs.isSatisfiedBy(ch) && rhs.isSatisfiedBy(ch);}}; 4898 } 4899 4900 /** 4901 * Returns the set difference of two CharProperty nodes. 4902 */ 4903 private static CharProperty setDifference(final CharProperty lhs, 4904 final CharProperty rhs) { 4905 return new CharProperty() { 4906 boolean isSatisfiedBy(int ch) { 4907 return ! rhs.isSatisfiedBy(ch) && lhs.isSatisfiedBy(ch);}}; 4908 } 4909 4910 /** 4911 * Handles word boundaries. Includes a field to allow this one class to 4912 * deal with the different types of word boundaries we can match. The word 4913 * characters include underscores, letters, and digits. Non spacing marks 4914 * can are also part of a word if they have a base character, otherwise 4915 * they are ignored for purposes of finding word boundaries. 4916 */ 4917 static final class Bound extends Node { 4918 static int LEFT = 0x1; 4919 static int RIGHT= 0x2; 4920 static int BOTH = 0x3; 4921 static int NONE = 0x4; 4922 int type; 4923 Bound(int n) { 4924 type = n; 4925 } 4926 int check(Matcher matcher, int i, CharSequence seq) { 4927 int ch; 4928 boolean left = false; 4929 int startIndex = matcher.from; 4930 int endIndex = matcher.to; 4931 if (matcher.transparentBounds) { 4932 startIndex = 0; 4933 endIndex = matcher.getTextLength(); 4934 } 4935 if (i > startIndex) { 4936 ch = Character.codePointBefore(seq, i); 4937 left = (ch == '_' || Character.isLetterOrDigit(ch) || 4938 ((Character.getType(ch) == Character.NON_SPACING_MARK) 4939 && hasBaseCharacter(matcher, i-1, seq))); 4940 } 4941 boolean right = false; 4942 if (i < endIndex) { 4943 ch = Character.codePointAt(seq, i); 4944 right = (ch == '_' || Character.isLetterOrDigit(ch) || 4945 ((Character.getType(ch) == Character.NON_SPACING_MARK) 4946 && hasBaseCharacter(matcher, i, seq))); 4947 } else { 4948 // Tried to access char past the end 4949 matcher.hitEnd = true; 4950 // The addition of another char could wreck a boundary 4951 matcher.requireEnd = true; 4952 } 4953 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE); 4954 } 4955 boolean match(Matcher matcher, int i, CharSequence seq) { 4956 return (check(matcher, i, seq) & type) > 0 4957 && next.match(matcher, i, seq); 4958 } 4959 } 4960 4961 /** 4962 * Non spacing marks only count as word characters in bounds calculations 4963 * if they have a base character. 4964 */ 4965 private static boolean hasBaseCharacter(Matcher matcher, int i, 4966 CharSequence seq) 4967 { 4968 int start = (!matcher.transparentBounds) ? 4969 matcher.from : 0; 4970 for (int x=i; x >= start; x--) { 4971 int ch = Character.codePointAt(seq, x); 4972 if (Character.isLetterOrDigit(ch)) 4973 return true; 4974 if (Character.getType(ch) == Character.NON_SPACING_MARK) 4975 continue; 4976 return false; 4977 } 4978 return false; 4979 } 4980 4981 /** 4982 * Attempts to match a slice in the input using the Boyer-Moore string 4983 * matching algorithm. The algorithm is based on the idea that the 4984 * pattern can be shifted farther ahead in the search text if it is 4985 * matched right to left. 4986 * <p> 4987 * The pattern is compared to the input one character at a time, from 4988 * the rightmost character in the pattern to the left. If the characters 4989 * all match the pattern has been found. If a character does not match, 4990 * the pattern is shifted right a distance that is the maximum of two 4991 * functions, the bad character shift and the good suffix shift. This 4992 * shift moves the attempted match position through the input more 4993 * quickly than a naive one position at a time check. 4994 * <p> 4995 * The bad character shift is based on the character from the text that 4996 * did not match. If the character does not appear in the pattern, the 4997 * pattern can be shifted completely beyond the bad character. If the 4998 * character does occur in the pattern, the pattern can be shifted to 4999 * line the pattern up with the next occurrence of that character. 5000 * <p> 5001 * The good suffix shift is based on the idea that some subset on the right 5002 * side of the pattern has matched. When a bad character is found, the 5003 * pattern can be shifted right by the pattern length if the subset does 5004 * not occur again in pattern, or by the amount of distance to the 5005 * next occurrence of the subset in the pattern. 5006 * 5007 * Boyer-Moore search methods adapted from code by Amy Yu. 5008 */ 5009 static class BnM extends Node { 5010 int[] buffer; 5011 int[] lastOcc; 5012 int[] optoSft; 5013 5014 /** 5015 * Pre calculates arrays needed to generate the bad character 5016 * shift and the good suffix shift. Only the last seven bits 5017 * are used to see if chars match; This keeps the tables small 5018 * and covers the heavily used ASCII range, but occasionally 5019 * results in an aliased match for the bad character shift. 5020 */ 5021 static Node optimize(Node node) { 5022 if (!(node instanceof Slice)) { 5023 return node; 5024 } 5025 5026 int[] src = ((Slice) node).buffer; 5027 int patternLength = src.length; 5028 // The BM algorithm requires a bit of overhead; 5029 // If the pattern is short don't use it, since 5030 // a shift larger than the pattern length cannot 5031 // be used anyway. 5032 if (patternLength < 4) { 5033 return node; 5034 } 5035 int i, j, k; 5036 int[] lastOcc = new int[128]; 5037 int[] optoSft = new int[patternLength]; 5038 // Precalculate part of the bad character shift 5039 // It is a table for where in the pattern each 5040 // lower 7-bit value occurs 5041 for (i = 0; i < patternLength; i++) { 5042 lastOcc[src[i]&0x7F] = i + 1; 5043 } 5044 // Precalculate the good suffix shift 5045 // i is the shift amount being considered 5046 NEXT: for (i = patternLength; i > 0; i--) { 5047 // j is the beginning index of suffix being considered 5048 for (j = patternLength - 1; j >= i; j--) { 5049 // Testing for good suffix 5050 if (src[j] == src[j-i]) { 5051 // src[j..len] is a good suffix 5052 optoSft[j-1] = i; 5053 } else { 5054 // No match. The array has already been 5055 // filled up with correct values before. 5056 continue NEXT; 5057 } 5058 } 5059 // This fills up the remaining of optoSft 5060 // any suffix can not have larger shift amount 5061 // then its sub-suffix. Why??? 5062 while (j > 0) { 5063 optoSft[--j] = i; 5064 } 5065 } 5066 // Set the guard value because of unicode compression 5067 optoSft[patternLength-1] = 1; 5068 if (node instanceof SliceS) 5069 return new BnMS(src, lastOcc, optoSft, node.next); 5070 return new BnM(src, lastOcc, optoSft, node.next); 5071 } 5072 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5073 this.buffer = src; 5074 this.lastOcc = lastOcc; 5075 this.optoSft = optoSft; 5076 this.next = next; 5077 } 5078 boolean match(Matcher matcher, int i, CharSequence seq) { 5079 int[] src = buffer; 5080 int patternLength = src.length; 5081 int last = matcher.to - patternLength; 5082 5083 // Loop over all possible match positions in text 5084 NEXT: while (i <= last) { 5085 // Loop over pattern from right to left 5086 for (int j = patternLength - 1; j >= 0; j--) { 5087 int ch = seq.charAt(i+j); 5088 if (ch != src[j]) { 5089 // Shift search to the right by the maximum of the 5090 // bad character shift and the good suffix shift 5091 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]); 5092 continue NEXT; 5093 } 5094 } 5095 // Entire pattern matched starting at i 5096 matcher.first = i; 5097 boolean ret = next.match(matcher, i + patternLength, seq); 5098 if (ret) { 5099 matcher.first = i; 5100 matcher.groups[0] = matcher.first; 5101 matcher.groups[1] = matcher.last; 5102 return true; 5103 } 5104 i++; 5105 } 5106 // BnM is only used as the leading node in the unanchored case, 5107 // and it replaced its Start() which always searches to the end 5108 // if it doesn't find what it's looking for, so hitEnd is true. 5109 matcher.hitEnd = true; 5110 return false; 5111 } 5112 boolean study(TreeInfo info) { 5113 info.minLength += buffer.length; 5114 info.maxValid = false; 5115 return next.study(info); 5116 } 5117 } 5118 5119 /** 5120 * Supplementary support version of BnM(). Unpaired surrogates are 5121 * also handled by this class. 5122 */ 5123 static final class BnMS extends BnM { 5124 int lengthInChars; 5125 5126 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5127 super(src, lastOcc, optoSft, next); 5128 for (int x = 0; x < buffer.length; x++) { 5129 lengthInChars += Character.charCount(buffer[x]); 5130 } 5131 } 5132 boolean match(Matcher matcher, int i, CharSequence seq) { 5133 int[] src = buffer; 5134 int patternLength = src.length; 5135 int last = matcher.to - lengthInChars; 5136 5137 // Loop over all possible match positions in text 5138 NEXT: while (i <= last) { 5139 // Loop over pattern from right to left 5140 int ch; 5141 for (int j = countChars(seq, i, patternLength), x = patternLength - 1; 5142 j > 0; j -= Character.charCount(ch), x--) { 5143 ch = Character.codePointBefore(seq, i+j); 5144 if (ch != src[x]) { 5145 // Shift search to the right by the maximum of the 5146 // bad character shift and the good suffix shift 5147 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]); 5148 i += countChars(seq, i, n); 5149 continue NEXT; 5150 } 5151 } 5152 // Entire pattern matched starting at i 5153 matcher.first = i; 5154 boolean ret = next.match(matcher, i + lengthInChars, seq); 5155 if (ret) { 5156 matcher.first = i; 5157 matcher.groups[0] = matcher.first; 5158 matcher.groups[1] = matcher.last; 5159 return true; 5160 } 5161 i += countChars(seq, i, 1); 5162 } 5163 matcher.hitEnd = true; 5164 return false; 5165 } 5166 } 5167 5168 /////////////////////////////////////////////////////////////////////////////// 5169 /////////////////////////////////////////////////////////////////////////////// 5170 5171 /** 5172 * This must be the very first initializer. 5173 */ 5174 static Node accept = new Node(); 5175 5176 static Node lastAccept = new LastNode(); 5177 5178 private static class CharPropertyNames { 5179 5180 static CharProperty charPropertyFor(String name) { 5181 CharPropertyFactory m = map.get(name); 5182 return m == null ? null : m.make(); 5183 } 5184 5185 private static abstract class CharPropertyFactory { 5186 abstract CharProperty make(); 5187 } 5188 5189 private static void defCategory(String name, 5190 final int typeMask) { 5191 map.put(name, new CharPropertyFactory() { 5192 CharProperty make() { return new Category(typeMask);}}); 5193 } 5194 5195 private static void defRange(String name, 5196 final int lower, final int upper) { 5197 map.put(name, new CharPropertyFactory() { 5198 CharProperty make() { return rangeFor(lower, upper);}}); 5199 } 5200 5201 private static void defCtype(String name, 5202 final int ctype) { 5203 map.put(name, new CharPropertyFactory() { 5204 CharProperty make() { return new Ctype(ctype);}}); 5205 } 5206 5207 private static abstract class CloneableProperty 5208 extends CharProperty implements Cloneable 5209 { 5210 public CloneableProperty clone() { 5211 try { 5212 return (CloneableProperty) super.clone(); 5213 } catch (CloneNotSupportedException e) { 5214 throw new AssertionError(e); 5215 } 5216 } 5217 } 5218 5219 private static void defClone(String name, 5220 final CloneableProperty p) { 5221 map.put(name, new CharPropertyFactory() { 5222 CharProperty make() { return p.clone();}}); 5223 } 5224 5225 private static final HashMap<String, CharPropertyFactory> map 5226 = new HashMap<String, CharPropertyFactory>(); 5227 5228 static { 5229 // Unicode character property aliases, defined in 5230 // http://www.unicode.org/Public/UNIDATA/PropertyValueAliases.txt 5231 defCategory("Cn", 1<<Character.UNASSIGNED); 5232 defCategory("Lu", 1<<Character.UPPERCASE_LETTER); 5233 defCategory("Ll", 1<<Character.LOWERCASE_LETTER); 5234 defCategory("Lt", 1<<Character.TITLECASE_LETTER); 5235 defCategory("Lm", 1<<Character.MODIFIER_LETTER); 5236 defCategory("Lo", 1<<Character.OTHER_LETTER); 5237 defCategory("Mn", 1<<Character.NON_SPACING_MARK); 5238 defCategory("Me", 1<<Character.ENCLOSING_MARK); 5239 defCategory("Mc", 1<<Character.COMBINING_SPACING_MARK); 5240 defCategory("Nd", 1<<Character.DECIMAL_DIGIT_NUMBER); 5241 defCategory("Nl", 1<<Character.LETTER_NUMBER); 5242 defCategory("No", 1<<Character.OTHER_NUMBER); 5243 defCategory("Zs", 1<<Character.SPACE_SEPARATOR); 5244 defCategory("Zl", 1<<Character.LINE_SEPARATOR); 5245 defCategory("Zp", 1<<Character.PARAGRAPH_SEPARATOR); 5246 defCategory("Cc", 1<<Character.CONTROL); 5247 defCategory("Cf", 1<<Character.FORMAT); 5248 defCategory("Co", 1<<Character.PRIVATE_USE); 5249 defCategory("Cs", 1<<Character.SURROGATE); 5250 defCategory("Pd", 1<<Character.DASH_PUNCTUATION); 5251 defCategory("Ps", 1<<Character.START_PUNCTUATION); 5252 defCategory("Pe", 1<<Character.END_PUNCTUATION); 5253 defCategory("Pc", 1<<Character.CONNECTOR_PUNCTUATION); 5254 defCategory("Po", 1<<Character.OTHER_PUNCTUATION); 5255 defCategory("Sm", 1<<Character.MATH_SYMBOL); 5256 defCategory("Sc", 1<<Character.CURRENCY_SYMBOL); 5257 defCategory("Sk", 1<<Character.MODIFIER_SYMBOL); 5258 defCategory("So", 1<<Character.OTHER_SYMBOL); 5259 defCategory("Pi", 1<<Character.INITIAL_QUOTE_PUNCTUATION); 5260 defCategory("Pf", 1<<Character.FINAL_QUOTE_PUNCTUATION); 5261 defCategory("L", ((1<<Character.UPPERCASE_LETTER) | 5262 (1<<Character.LOWERCASE_LETTER) | 5263 (1<<Character.TITLECASE_LETTER) | 5264 (1<<Character.MODIFIER_LETTER) | 5265 (1<<Character.OTHER_LETTER))); 5266 defCategory("M", ((1<<Character.NON_SPACING_MARK) | 5267 (1<<Character.ENCLOSING_MARK) | 5268 (1<<Character.COMBINING_SPACING_MARK))); 5269 defCategory("N", ((1<<Character.DECIMAL_DIGIT_NUMBER) | 5270 (1<<Character.LETTER_NUMBER) | 5271 (1<<Character.OTHER_NUMBER))); 5272 defCategory("Z", ((1<<Character.SPACE_SEPARATOR) | 5273 (1<<Character.LINE_SEPARATOR) | 5274 (1<<Character.PARAGRAPH_SEPARATOR))); 5275 defCategory("C", ((1<<Character.CONTROL) | 5276 (1<<Character.FORMAT) | 5277 (1<<Character.PRIVATE_USE) | 5278 (1<<Character.SURROGATE))); // Other 5279 defCategory("P", ((1<<Character.DASH_PUNCTUATION) | 5280 (1<<Character.START_PUNCTUATION) | 5281 (1<<Character.END_PUNCTUATION) | 5282 (1<<Character.CONNECTOR_PUNCTUATION) | 5283 (1<<Character.OTHER_PUNCTUATION) | 5284 (1<<Character.INITIAL_QUOTE_PUNCTUATION) | 5285 (1<<Character.FINAL_QUOTE_PUNCTUATION))); 5286 defCategory("S", ((1<<Character.MATH_SYMBOL) | 5287 (1<<Character.CURRENCY_SYMBOL) | 5288 (1<<Character.MODIFIER_SYMBOL) | 5289 (1<<Character.OTHER_SYMBOL))); 5290 defCategory("LC", ((1<<Character.UPPERCASE_LETTER) | 5291 (1<<Character.LOWERCASE_LETTER) | 5292 (1<<Character.TITLECASE_LETTER))); 5293 defCategory("LD", ((1<<Character.UPPERCASE_LETTER) | 5294 (1<<Character.LOWERCASE_LETTER) | 5295 (1<<Character.TITLECASE_LETTER) | 5296 (1<<Character.MODIFIER_LETTER) | 5297 (1<<Character.OTHER_LETTER) | 5298 (1<<Character.DECIMAL_DIGIT_NUMBER))); 5299 defRange("L1", 0x00, 0xFF); // Latin-1 5300 map.put("all", new CharPropertyFactory() { 5301 CharProperty make() { return new All(); }}); 5302 5303 // Posix regular expression character classes, defined in 5304 // http://www.unix.org/onlinepubs/009695399/basedefs/xbd_chap09.html 5305 defRange("ASCII", 0x00, 0x7F); // ASCII 5306 defCtype("Alnum", ASCII.ALNUM); // Alphanumeric characters 5307 defCtype("Alpha", ASCII.ALPHA); // Alphabetic characters 5308 defCtype("Blank", ASCII.BLANK); // Space and tab characters 5309 defCtype("Cntrl", ASCII.CNTRL); // Control characters 5310 defRange("Digit", '0', '9'); // Numeric characters 5311 defCtype("Graph", ASCII.GRAPH); // printable and visible 5312 defRange("Lower", 'a', 'z'); // Lower-case alphabetic 5313 defRange("Print", 0x20, 0x7E); // Printable characters 5314 defCtype("Punct", ASCII.PUNCT); // Punctuation characters 5315 defCtype("Space", ASCII.SPACE); // Space characters 5316 defRange("Upper", 'A', 'Z'); // Upper-case alphabetic 5317 defCtype("XDigit",ASCII.XDIGIT); // hexadecimal digits 5318 5319 // Java character properties, defined by methods in Character.java 5320 defClone("javaLowerCase", new CloneableProperty() { 5321 boolean isSatisfiedBy(int ch) { 5322 return Character.isLowerCase(ch);}}); 5323 defClone("javaUpperCase", new CloneableProperty() { 5324 boolean isSatisfiedBy(int ch) { 5325 return Character.isUpperCase(ch);}}); 5326 defClone("javaTitleCase", new CloneableProperty() { 5327 boolean isSatisfiedBy(int ch) { 5328 return Character.isTitleCase(ch);}}); 5329 defClone("javaDigit", new CloneableProperty() { 5330 boolean isSatisfiedBy(int ch) { 5331 return Character.isDigit(ch);}}); 5332 defClone("javaDefined", new CloneableProperty() { 5333 boolean isSatisfiedBy(int ch) { 5334 return Character.isDefined(ch);}}); 5335 defClone("javaLetter", new CloneableProperty() { 5336 boolean isSatisfiedBy(int ch) { 5337 return Character.isLetter(ch);}}); 5338 defClone("javaLetterOrDigit", new CloneableProperty() { 5339 boolean isSatisfiedBy(int ch) { 5340 return Character.isLetterOrDigit(ch);}}); 5341 defClone("javaJavaIdentifierStart", new CloneableProperty() { 5342 boolean isSatisfiedBy(int ch) { 5343 return Character.isJavaIdentifierStart(ch);}}); 5344 defClone("javaJavaIdentifierPart", new CloneableProperty() { 5345 boolean isSatisfiedBy(int ch) { 5346 return Character.isJavaIdentifierPart(ch);}}); 5347 defClone("javaUnicodeIdentifierStart", new CloneableProperty() { 5348 boolean isSatisfiedBy(int ch) { 5349 return Character.isUnicodeIdentifierStart(ch);}}); 5350 defClone("javaUnicodeIdentifierPart", new CloneableProperty() { 5351 boolean isSatisfiedBy(int ch) { 5352 return Character.isUnicodeIdentifierPart(ch);}}); 5353 defClone("javaIdentifierIgnorable", new CloneableProperty() { 5354 boolean isSatisfiedBy(int ch) { 5355 return Character.isIdentifierIgnorable(ch);}}); 5356 defClone("javaSpaceChar", new CloneableProperty() { 5357 boolean isSatisfiedBy(int ch) { 5358 return Character.isSpaceChar(ch);}}); 5359 defClone("javaWhitespace", new CloneableProperty() { 5360 boolean isSatisfiedBy(int ch) { 5361 return Character.isWhitespace(ch);}}); 5362 defClone("javaISOControl", new CloneableProperty() { 5363 boolean isSatisfiedBy(int ch) { 5364 return Character.isISOControl(ch);}}); 5365 defClone("javaMirrored", new CloneableProperty() { 5366 boolean isSatisfiedBy(int ch) { 5367 return Character.isMirrored(ch);}}); 5368 } 5369 } 5370 }