1 /* 2 * Copyright (c) 1999, 2018, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. Oracle designates this 8 * particular file as subject to the "Classpath" exception as provided 9 * by Oracle in the LICENSE file that accompanied this code. 10 * 11 * This code is distributed in the hope that it will be useful, but WITHOUT 12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 14 * version 2 for more details (a copy is included in the LICENSE file that 15 * accompanied this code). 16 * 17 * You should have received a copy of the GNU General Public License version 18 * 2 along with this work; if not, write to the Free Software Foundation, 19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 20 * 21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 22 * or visit www.oracle.com if you need additional information or have any 23 * questions. 24 */ 25 26 package java.util.regex; 27 28 import java.text.Normalizer; 29 import java.text.Normalizer.Form; 30 import java.util.Locale; 31 import java.util.Iterator; 32 import java.util.Map; 33 import java.util.ArrayList; 34 import java.util.HashMap; 35 import java.util.LinkedHashSet; 36 import java.util.List; 37 import java.util.Set; 38 import java.util.Arrays; 39 import java.util.NoSuchElementException; 40 import java.util.Spliterator; 41 import java.util.Spliterators; 42 import java.util.function.Predicate; 43 import java.util.stream.Stream; 44 import java.util.stream.StreamSupport; 45 46 47 /** 48 * A compiled representation of a regular expression. 49 * 50 * <p> A regular expression, specified as a string, must first be compiled into 51 * an instance of this class. The resulting pattern can then be used to create 52 * a {@link Matcher} object that can match arbitrary {@linkplain 53 * java.lang.CharSequence character sequences} against the regular 54 * expression. All of the state involved in performing a match resides in the 55 * matcher, so many matchers can share the same pattern. 56 * 57 * <p> A typical invocation sequence is thus 58 * 59 * <blockquote><pre> 60 * Pattern p = Pattern.{@link #compile compile}("a*b"); 61 * Matcher m = p.{@link #matcher matcher}("aaaaab"); 62 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote> 63 * 64 * <p> A {@link #matches matches} method is defined by this class as a 65 * convenience for when a regular expression is used just once. This method 66 * compiles an expression and matches an input sequence against it in a single 67 * invocation. The statement 68 * 69 * <blockquote><pre> 70 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote> 71 * 72 * is equivalent to the three statements above, though for repeated matches it 73 * is less efficient since it does not allow the compiled pattern to be reused. 74 * 75 * <p> Instances of this class are immutable and are safe for use by multiple 76 * concurrent threads. Instances of the {@link Matcher} class are not safe for 77 * such use. 78 * 79 * 80 * <h3><a id="sum">Summary of regular-expression constructs</a></h3> 81 * 82 * <table class="borderless"> 83 * <caption style="display:none">Regular expression constructs, and what they match</caption> 84 * <thead style="text-align:left"> 85 * <tr> 86 * <th id="construct">Construct</th> 87 * <th id="matches">Matches</th> 88 * </tr> 89 * </thead> 90 * <tbody style="text-align:left"> 91 * 92 * <tr><th colspan="2" style="padding-top:20px" id="characters">Characters</th></tr> 93 * 94 * <tr><th style="vertical-align:top; font-weight: normal" id="x"><i>x</i></th> 95 * <td headers="matches characters x">The character <i>x</i></td></tr> 96 * <tr><th style="vertical-align:top; font-weight: normal" id="backslash">{@code \\}</th> 97 * <td headers="matches characters backslash">The backslash character</td></tr> 98 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_n">{@code \0}<i>n</i></th> 99 * <td headers="matches characters octal_n">The character with octal value {@code 0}<i>n</i> 100 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 101 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nn">{@code \0}<i>nn</i></th> 102 * <td headers="matches characters octal_nn">The character with octal value {@code 0}<i>nn</i> 103 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 104 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nnn">{@code \0}<i>mnn</i></th> 105 * <td headers="matches characters octal_nnn">The character with octal value {@code 0}<i>mnn</i> 106 * (0 {@code <=} <i>m</i> {@code <=} 3, 107 * 0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 108 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hh">{@code \x}<i>hh</i></th> 109 * <td headers="matches characters hex_hh">The character with hexadecimal value {@code 0x}<i>hh</i></td></tr> 110 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hhhh"><code>\u</code><i>hhhh</i></th> 111 * <td headers="matches characters hex_hhhh">The character with hexadecimal value {@code 0x}<i>hhhh</i></td></tr> 112 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_h_h"><code>\x</code><i>{h...h}</i></th> 113 * <td headers="matches characters hex_h_h">The character with hexadecimal value {@code 0x}<i>h...h</i> 114 * ({@link java.lang.Character#MIN_CODE_POINT Character.MIN_CODE_POINT} 115 * <= {@code 0x}<i>h...h</i> <= 116 * {@link java.lang.Character#MAX_CODE_POINT Character.MAX_CODE_POINT})</td></tr> 117 * <tr><th style="vertical-align:top; font-weight: normal" id="unicode_name"><code>\N{</code><i>name</i><code>}</code></th> 118 * <td headers="matches characters unicode_name">The character with Unicode character name <i>'name'</i></td></tr> 119 * <tr><th style="vertical-align:top; font-weight:normal" id="tab">{@code \t}</th> 120 * <td headers="matches characters tab">The tab character (<code>'\u0009'</code>)</td></tr> 121 * <tr><th style="vertical-align:top; font-weight:normal" id="newline">{@code \n}</th> 122 * <td headers="matches characters newline">The newline (line feed) character (<code>'\u000A'</code>)</td></tr> 123 * <tr><th style="vertical-align:top; font-weight:normal" id="return">{@code \r}</th> 124 * <td headers="matches characters return">The carriage-return character (<code>'\u000D'</code>)</td></tr> 125 * <tr><th style="vertical-align:top; font-weight:normal" id="form_feed">{@code \f}</th> 126 * <td headers="matches characters form_feed">The form-feed character (<code>'\u000C'</code>)</td></tr> 127 * <tr><th style="vertical-align:top; font-weight:normal" id="bell">{@code \a}</th> 128 * <td headers="matches characters bell">The alert (bell) character (<code>'\u0007'</code>)</td></tr> 129 * <tr><th style="vertical-align:top; font-weight:normal" id="escape">{@code \e}</th> 130 * <td headers="matches characters escape">The escape character (<code>'\u001B'</code>)</td></tr> 131 * <tr><th style="vertical-align:top; font-weight:normal" id="ctrl_x">{@code \c}<i>x</i></th> 132 * <td headers="matches characters ctrl_x">The control character corresponding to <i>x</i></td></tr> 133 * 134 * <tr><th colspan="2" style="padding-top:20px" id="classes">Character classes</th></tr> 135 * 136 * <tr><th style="vertical-align:top; font-weight:normal" id="simple">{@code [abc]}</th> 137 * <td headers="matches classes simple">{@code a}, {@code b}, or {@code c} (simple class)</td></tr> 138 * <tr><th style="vertical-align:top; font-weight:normal" id="negation">{@code [^abc]}</th> 139 * <td headers="matches classes negation">Any character except {@code a}, {@code b}, or {@code c} (negation)</td></tr> 140 * <tr><th style="vertical-align:top; font-weight:normal" id="range">{@code [a-zA-Z]}</th> 141 * <td headers="matches classes range">{@code a} through {@code z} 142 * or {@code A} through {@code Z}, inclusive (range)</td></tr> 143 * <tr><th style="vertical-align:top; font-weight:normal" id="union">{@code [a-d[m-p]]}</th> 144 * <td headers="matches classes union">{@code a} through {@code d}, 145 * or {@code m} through {@code p}: {@code [a-dm-p]} (union)</td></tr> 146 * <tr><th style="vertical-align:top; font-weight:normal" id="intersection">{@code [a-z&&[def]]}</th> 147 * <td headers="matches classes intersection">{@code d}, {@code e}, or {@code f} (intersection)</tr> 148 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction1">{@code [a-z&&[^bc]]}</th> 149 * <td headers="matches classes subtraction1">{@code a} through {@code z}, 150 * except for {@code b} and {@code c}: {@code [ad-z]} (subtraction)</td></tr> 151 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction2">{@code [a-z&&[^m-p]]}</th> 152 * <td headers="matches classes subtraction2">{@code a} through {@code z}, 153 * and not {@code m} through {@code p}: {@code [a-lq-z]}(subtraction)</td></tr> 154 * 155 * <tr><th colspan="2" style="padding-top:20px" id="predef">Predefined character classes</th></tr> 156 * 157 * <tr><th style="vertical-align:top; font-weight:normal" id="any">{@code .}</th> 158 * <td headers="matches predef any">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr> 159 * <tr><th style="vertical-align:top; font-weight:normal" id="digit">{@code \d}</th> 160 * <td headers="matches predef digit">A digit: {@code [0-9]}</td></tr> 161 * <tr><th style="vertical-align:top; font-weight:normal" id="non_digit">{@code \D}</th> 162 * <td headers="matches predef non_digit">A non-digit: {@code [^0-9]}</td></tr> 163 * <tr><th style="vertical-align:top; font-weight:normal" id="horiz_white">{@code \h}</th> 164 * <td headers="matches predef horiz_white">A horizontal whitespace character: 165 * <code>[ \t\xA0\u1680\u180e\u2000-\u200a\u202f\u205f\u3000]</code></td></tr> 166 * <tr><th style="vertical-align:top; font-weight:normal" id="non_horiz_white">{@code \H}</th> 167 * <td headers="matches predef non_horiz_white">A non-horizontal whitespace character: {@code [^\h]}</td></tr> 168 * <tr><th style="vertical-align:top; font-weight:normal" id="white">{@code \s}</th> 169 * <td headers="matches predef white">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 170 * <tr><th style="vertical-align:top; font-weight:normal" id="non_white">{@code \S}</th> 171 * <td headers="matches predef non_white">A non-whitespace character: {@code [^\s]}</td></tr> 172 * <tr><th style="vertical-align:top; font-weight:normal" id="vert_white">{@code \v}</th> 173 * <td headers="matches predef vert_white">A vertical whitespace character: <code>[\n\x0B\f\r\x85\u2028\u2029]</code> 174 * </td></tr> 175 * <tr><th style="vertical-align:top; font-weight:normal" id="non_vert_white">{@code \V}</th> 176 * <td headers="matches predef non_vert_white">A non-vertical whitespace character: {@code [^\v]}</td></tr> 177 * <tr><th style="vertical-align:top; font-weight:normal" id="word">{@code \w}</th> 178 * <td headers="matches predef word">A word character: {@code [a-zA-Z_0-9]}</td></tr> 179 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word">{@code \W}</th> 180 * <td headers="matches predef non_word">A non-word character: {@code [^\w]}</td></tr> 181 * 182 * <tr><th colspan="2" style="padding-top:20px" id="posix"><b>POSIX character classes (US-ASCII only)</b></th></tr> 183 * 184 * <tr><th style="vertical-align:top; font-weight:normal" id="Lower">{@code \p{Lower}}</th> 185 * <td headers="matches posix Lower">A lower-case alphabetic character: {@code [a-z]}</td></tr> 186 * <tr><th style="vertical-align:top; font-weight:normal" id="Upper">{@code \p{Upper}}</th> 187 * <td headers="matches posix Upper">An upper-case alphabetic character:{@code [A-Z]}</td></tr> 188 * <tr><th style="vertical-align:top; font-weight:normal" id="ASCII">{@code \p{ASCII}}</th> 189 * <td headers="matches posix ASCII">All ASCII:{@code [\x00-\x7F]}</td></tr> 190 * <tr><th style="vertical-align:top; font-weight:normal" id="Alpha">{@code \p{Alpha}}</th> 191 * <td headers="matches posix Alpha">An alphabetic character:{@code [\p{Lower}\p{Upper}]}</td></tr> 192 * <tr><th style="vertical-align:top; font-weight:normal" id="Digit">{@code \p{Digit}}</th> 193 * <td headers="matches posix Digit">A decimal digit: {@code [0-9]}</td></tr> 194 * <tr><th style="vertical-align:top; font-weight:normal" id="Alnum">{@code \p{Alnum}}</th> 195 * <td headers="matches posix Alnum">An alphanumeric character:{@code [\p{Alpha}\p{Digit}]}</td></tr> 196 * <tr><th style="vertical-align:top; font-weight:normal" id="Punct">{@code \p{Punct}}</th> 197 * <td headers="matches posix Punct">Punctuation: One of {@code !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~}</td></tr> 198 * <!-- {@code [\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]} 199 * {@code [\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]} --> 200 * <tr><th style="vertical-align:top; font-weight:normal" id="Graph">{@code \p{Graph}}</th> 201 * <td headers="matches posix Graph">A visible character: {@code [\p{Alnum}\p{Punct}]}</td></tr> 202 * <tr><th style="vertical-align:top; font-weight:normal" id="Print">{@code \p{Print}}</th> 203 * <td headers="matches posix Print">A printable character: {@code [\p{Graph}\x20]}</td></tr> 204 * <tr><th style="vertical-align:top; font-weight:normal" id="Blank">{@code \p{Blank}}</th> 205 * <td headers="matches posix Blank">A space or a tab: {@code [ \t]}</td></tr> 206 * <tr><th style="vertical-align:top; font-weight:normal" id="Cntrl">{@code \p{Cntrl}}</th> 207 * <td headers="matches posix Cntrl">A control character: {@code [\x00-\x1F\x7F]}</td></tr> 208 * <tr><th style="vertical-align:top; font-weight:normal" id="XDigit">{@code \p{XDigit}}</th> 209 * <td headers="matches posix XDigit">A hexadecimal digit: {@code [0-9a-fA-F]}</td></tr> 210 * <tr><th style="vertical-align:top; font-weight:normal" id="Space">{@code \p{Space}}</th> 211 * <td headers="matches posix Space">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 212 * 213 * <tr><th colspan="2" style="padding-top:20px" id="java">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr> 214 * 215 * <tr><th style="vertical-align:top; font-weight:normal" id="javaLowerCase">{@code \p{javaLowerCase}}</th> 216 * <td headers="matches java javaLowerCase">Equivalent to java.lang.Character.isLowerCase()</td></tr> 217 * <tr><th style="vertical-align:top; font-weight:normal" id="javaUpperCase">{@code \p{javaUpperCase}}</th> 218 * <td headers="matches java javaUpperCase">Equivalent to java.lang.Character.isUpperCase()</td></tr> 219 * <tr><th style="vertical-align:top; font-weight:normal" id="javaWhitespace">{@code \p{javaWhitespace}}</th> 220 * <td headers="matches java javaWhitespace">Equivalent to java.lang.Character.isWhitespace()</td></tr> 221 * <tr><th style="vertical-align:top; font-weight:normal" id="javaMirrored">{@code \p{javaMirrored}}</th> 222 * <td headers="matches java javaMirrored">Equivalent to java.lang.Character.isMirrored()</td></tr> 223 * 224 * <tr><th colspan="2" style="padding-top:20px" id="unicode">Classes for Unicode scripts, blocks, categories and binary properties</th></tr> 225 * 226 * <tr><th style="vertical-align:top; font-weight:normal" id="IsLatin">{@code \p{IsLatin}}</th> 227 * <td headers="matches unicode IsLatin">A Latin script character (<a href="#usc">script</a>)</td></tr> 228 * <tr><th style="vertical-align:top; font-weight:normal" id="InGreek">{@code \p{InGreek}}</th> 229 * <td headers="matches unicode InGreek">A character in the Greek block (<a href="#ubc">block</a>)</td></tr> 230 * <tr><th style="vertical-align:top; font-weight:normal" id="Lu">{@code \p{Lu}}</th> 231 * <td headers="matches unicode Lu">An uppercase letter (<a href="#ucc">category</a>)</td></tr> 232 * <tr><th style="vertical-align:top; font-weight:normal" id="IsAlphabetic">{@code \p{IsAlphabetic}}</th> 233 * <td headers="matches unicode IsAlphabetic">An alphabetic character (<a href="#ubpc">binary property</a>)</td></tr> 234 * <tr><th style="vertical-align:top; font-weight:normal" id="Sc">{@code \p{Sc}}</th> 235 * <td headers="matches unicode Sc">A currency symbol</td></tr> 236 * <tr><th style="vertical-align:top; font-weight:normal" id="not_InGreek">{@code \P{InGreek}}</th> 237 * <td headers="matches unicode not_InGreek">Any character except one in the Greek block (negation)</td></tr> 238 * <tr><th style="vertical-align:top; font-weight:normal" id="not_uppercase">{@code [\p{L}&&[^\p{Lu}]]}</th> 239 * <td headers="matches unicode not_uppercase">Any letter except an uppercase letter (subtraction)</td></tr> 240 * 241 * <tr><th colspan="2" style="padding-top:20px" id="bounds">Boundary matchers</th></tr> 242 * 243 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_line">{@code ^}</th> 244 * <td headers="matches bounds begin_line">The beginning of a line</td></tr> 245 * <tr><th style="vertical-align:top; font-weight:normal" id="end_line">{@code $}</th> 246 * <td headers="matches bounds end_line">The end of a line</td></tr> 247 * <tr><th style="vertical-align:top; font-weight:normal" id="word_boundary">{@code \b}</th> 248 * <td headers="matches bounds word_boundary">A word boundary</td></tr> 249 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_cluster_boundary">{@code \b{g}}</th> 250 * <td headers="matches bounds grapheme_cluster_boundary">A Unicode extended grapheme cluster boundary</td></tr> 251 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word_boundary">{@code \B}</th> 252 * <td headers="matches bounds non_word_boundary">A non-word boundary</td></tr> 253 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_input">{@code \A}</th> 254 * <td headers="matches bounds begin_input">The beginning of the input</td></tr> 255 * <tr><th style="vertical-align:top; font-weight:normal" id="end_prev_match">{@code \G}</th> 256 * <td headers="matches bounds end_prev_match">The end of the previous match</td></tr> 257 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input_except_term">{@code \Z}</th> 258 * <td headers="matches bounds end_input_except_term">The end of the input but for the final 259 * <a href="#lt">terminator</a>, if any</td></tr> 260 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input">{@code \z}</th> 261 * <td headers="matches bounds end_input">The end of the input</td></tr> 262 * 263 * <tr><th colspan="2" style="padding-top:20px" id="linebreak">Linebreak matcher</th></tr> 264 * 265 * <tr><th style="vertical-align:top; font-weight:normal" id="any_unicode_linebreak">{@code \R}</th> 266 * <td headers="matches linebreak any_unicode_linebreak">Any Unicode linebreak sequence, is equivalent to 267 * <code>\u000D\u000A|[\u000A\u000B\u000C\u000D\u0085\u2028\u2029] 268 * </code></td></tr> 269 * 270 * <tr><th colspan="2" style="padding-top:20px" id="grapheme">Unicode Extended Grapheme matcher</th></tr> 271 * 272 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_any">{@code \X}</th> 273 * <td headers="matches grapheme grapheme_any">Any Unicode extended grapheme cluster</td></tr> 274 * 275 * <tr><th colspan="2" style="padding-top:20px" id="greedy">Greedy quantifiers</th></tr> 276 * 277 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_once_or_not"><i>X</i>{@code ?}</th> 278 * <td headers="matches greedy greedy_once_or_not"><i>X</i>, once or not at all</td></tr> 279 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_zero_or_more"><i>X</i>{@code *}</th> 280 * <td headers="matches greedy greedy_zero_or_more"><i>X</i>, zero or more times</td></tr> 281 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_one_or_more"><i>X</i>{@code +}</th> 282 * <td headers="matches greedy greedy_one_or_more"><i>X</i>, one or more times</td></tr> 283 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_exactly"><i>X</i><code>{</code><i>n</i><code>}</code></th> 284 * <td headers="matches greedy greedy_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 285 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least"><i>X</i><code>{</code><i>n</i>{@code ,}}</th> 286 * <td headers="matches greedy greedy_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 287 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}</code></th> 288 * <td headers="matches greedy greedy_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 289 * 290 * <tr><th colspan="2" style="padding-top:20px" id="reluc">Reluctant quantifiers</th></tr> 291 * 292 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_once_or_not"><i>X</i>{@code ??}</th> 293 * <td headers="matches reluc reluc_once_or_not"><i>X</i>, once or not at all</td></tr> 294 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_zero_or_more"><i>X</i>{@code *?}</th> 295 * <td headers="matches reluc reluc_zero_or_more"><i>X</i>, zero or more times</td></tr> 296 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_one_or_more"><i>X</i>{@code +?}</th> 297 * <td headers="matches reluc reluc_one_or_more"><i>X</i>, one or more times</td></tr> 298 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_exactly"><i>X</i><code>{</code><i>n</i><code>}?</code></th> 299 * <td headers="matches reluc reluc_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 300 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least"><i>X</i><code>{</code><i>n</i><code>,}?</code></th> 301 * <td headers="matches reluc reluc_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 302 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}?</code></th> 303 * <td headers="matches reluc reluc_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 304 * 305 * <tr><th colspan="2" style="padding-top:20px" id="poss">Possessive quantifiers</th></tr> 306 * 307 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_once_or_not"><i>X</i>{@code ?+}</th> 308 * <td headers="matches poss poss_once_or_not"><i>X</i>, once or not at all</td></tr> 309 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_zero_or_more"><i>X</i>{@code *+}</th> 310 * <td headers="matches poss poss_zero_or_more"><i>X</i>, zero or more times</td></tr> 311 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_one_or_more"><i>X</i>{@code ++}</th> 312 * <td headers="matches poss poss_one_or_more"><i>X</i>, one or more times</td></tr> 313 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_exactly"><i>X</i><code>{</code><i>n</i><code>}+</code></th> 314 * <td headers="matches poss poss_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 315 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least"><i>X</i><code>{</code><i>n</i><code>,}+</code></th> 316 * <td headers="matches poss poss_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 317 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}+</code></th> 318 * <td headers="matches poss poss_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 319 * 320 * <tr><th colspan="2" style="padding-top:20px" id="logical">Logical operators</th></tr> 321 * 322 * <tr><th style="vertical-align:top; font-weight:normal" id="concat"><i>XY</i></th> 323 * <td headers="matches logical concat"><i>X</i> followed by <i>Y</i></td></tr> 324 * <tr><th style="vertical-align:top; font-weight:normal" id="alternate"><i>X</i>{@code |}<i>Y</i></th> 325 * <td headers="matches logical alternate">Either <i>X</i> or <i>Y</i></td></tr> 326 * <tr><th style="vertical-align:top; font-weight:normal" id="group">{@code (}<i>X</i>{@code )}</th> 327 * <td headers="matches logical group">X, as a <a href="#cg">capturing group</a></td></tr> 328 * 329 * <tr><th colspan="2" style="padding-top:20px" id="backref">Back references</th></tr> 330 * 331 * <tr><th style="vertical-align:top; font-weight:normal" id="back_nth">{@code \}<i>n</i></th> 332 * <td headers="matches backref back_nth">Whatever the <i>n</i><sup>th</sup> 333 * <a href="#cg">capturing group</a> matched</td></tr> 334 * <tr><th style="vertical-align:top; font-weight:normal" id="back_named">{@code \}<i>k</i><<i>name</i>></th> 335 * <td headers="matches backref back_named">Whatever the 336 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr> 337 * 338 * <tr><th colspan="2" style="padding-top:20px" id="quote">Quotation</th></tr> 339 * 340 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_follow">{@code \}</th> 341 * <td headers="matches quote quote_follow">Nothing, but quotes the following character</td></tr> 342 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_begin">{@code \Q}</th> 343 * <td headers="matches quote quote_begin">Nothing, but quotes all characters until {@code \E}</td></tr> 344 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_end">{@code \E}</th> 345 * <td headers="matches quote quote_end">Nothing, but ends quoting started by {@code \Q}</td></tr> 346 * <!-- Metachars: !$()*+.<>?[\]^{|} --> 347 * 348 * <tr><th colspan="2" style="padding-top:20px" id="special">Special constructs (named-capturing and non-capturing)</th></tr> 349 * 350 * <tr><th style="vertical-align:top; font-weight:normal" id="named_group"><code>(?<<a href="#groupname">name</a>></code><i>X</i>{@code )}</th> 351 * <td headers="matches special named_group"><i>X</i>, as a named-capturing group</td></tr> 352 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group">{@code (?:}<i>X</i>{@code )}</th> 353 * <td headers="matches special non_capture_group"><i>X</i>, as a non-capturing group</td></tr> 354 * <tr><th style="vertical-align:top; font-weight:normal" id="flags"><code>(?idmsuxU-idmsuxU) </code></th> 355 * <td headers="matches special flags">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a> 356 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> 357 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a> 358 * on - off</td></tr> 359 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group_flags"><code>(?idmsux-idmsux:</code><i>X</i>{@code )} </th> 360 * <td headers="matches special non_capture_group_flags"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the 361 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a> 362 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a > 363 * <a href="#COMMENTS">x</a> on - off</td></tr> 364 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookahead">{@code (?=}<i>X</i>{@code )}</th> 365 * <td headers="matches special pos_lookahead"><i>X</i>, via zero-width positive lookahead</td></tr> 366 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookahead">{@code (?!}<i>X</i>{@code )}</th> 367 * <td headers="matches special neg_lookahead"><i>X</i>, via zero-width negative lookahead</td></tr> 368 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookbehind">{@code (?<=}<i>X</i>{@code )}</th> 369 * <td headers="matches special pos_lookbehind"><i>X</i>, via zero-width positive lookbehind</td></tr> 370 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookbehind">{@code (?<!}<i>X</i>{@code )}</th> 371 * <td headers="matches special neg_lookbehind"><i>X</i>, via zero-width negative lookbehind</td></tr> 372 * <tr><th style="vertical-align:top; font-weight:normal" id="indep_non_capture_group">{@code (?>}<i>X</i>{@code )}</th> 373 * <td headers="matches special indep_non_capture_group"><i>X</i>, as an independent, non-capturing group</td></tr> 374 * 375 * </tbody> 376 * </table> 377 * 378 * <hr> 379 * 380 * 381 * <h3><a id="bs">Backslashes, escapes, and quoting</a></h3> 382 * 383 * <p> The backslash character ({@code '\'}) serves to introduce escaped 384 * constructs, as defined in the table above, as well as to quote characters 385 * that otherwise would be interpreted as unescaped constructs. Thus the 386 * expression {@code \\} matches a single backslash and <code>\{</code> matches a 387 * left brace. 388 * 389 * <p> It is an error to use a backslash prior to any alphabetic character that 390 * does not denote an escaped construct; these are reserved for future 391 * extensions to the regular-expression language. A backslash may be used 392 * prior to a non-alphabetic character regardless of whether that character is 393 * part of an unescaped construct. 394 * 395 * <p> Backslashes within string literals in Java source code are interpreted 396 * as required by 397 * <cite>The Java™ Language Specification</cite> 398 * as either Unicode escapes (section 3.3) or other character escapes (section 3.10.6) 399 * It is therefore necessary to double backslashes in string 400 * literals that represent regular expressions to protect them from 401 * interpretation by the Java bytecode compiler. The string literal 402 * <code>"\b"</code>, for example, matches a single backspace character when 403 * interpreted as a regular expression, while {@code "\\b"} matches a 404 * word boundary. The string literal {@code "\(hello\)"} is illegal 405 * and leads to a compile-time error; in order to match the string 406 * {@code (hello)} the string literal {@code "\\(hello\\)"} 407 * must be used. 408 * 409 * <h3><a id="cc">Character Classes</a></h3> 410 * 411 * <p> Character classes may appear within other character classes, and 412 * may be composed by the union operator (implicit) and the intersection 413 * operator ({@code &&}). 414 * The union operator denotes a class that contains every character that is 415 * in at least one of its operand classes. The intersection operator 416 * denotes a class that contains every character that is in both of its 417 * operand classes. 418 * 419 * <p> The precedence of character-class operators is as follows, from 420 * highest to lowest: 421 * 422 * <table class="striped" style="margin-left: 2em;"> 423 * <caption style="display:none">Precedence of character class operators.</caption> 424 * <thead> 425 * <tr><th scope="col">Precedence<th scope="col">Name<th scope="col">Example 426 * </thead> 427 * <tbody> 428 * <tr><th scope="row">1</th> 429 * <td>Literal escape </td> 430 * <td>{@code \x}</td></tr> 431 * <tr><th scope="row">2</th> 432 * <td>Grouping</td> 433 * <td>{@code [...]}</td></tr> 434 * <tr><th scope="row">3</th> 435 * <td>Range</td> 436 * <td>{@code a-z}</td></tr> 437 * <tr><th scope="row">4</th> 438 * <td>Union</td> 439 * <td>{@code [a-e][i-u]}</td></tr> 440 * <tr><th scope="row">5</th> 441 * <td>Intersection</td> 442 * <td>{@code [a-z&&[aeiou]]}</td></tr> 443 * </tbody> 444 * </table> 445 * 446 * <p> Note that a different set of metacharacters are in effect inside 447 * a character class than outside a character class. For instance, the 448 * regular expression {@code .} loses its special meaning inside a 449 * character class, while the expression {@code -} becomes a range 450 * forming metacharacter. 451 * 452 * <h3><a id="lt">Line terminators</a></h3> 453 * 454 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks 455 * the end of a line of the input character sequence. The following are 456 * recognized as line terminators: 457 * 458 * <ul> 459 * 460 * <li> A newline (line feed) character ({@code '\n'}), 461 * 462 * <li> A carriage-return character followed immediately by a newline 463 * character ({@code "\r\n"}), 464 * 465 * <li> A standalone carriage-return character ({@code '\r'}), 466 * 467 * <li> A next-line character (<code>'\u0085'</code>), 468 * 469 * <li> A line-separator character (<code>'\u2028'</code>), or 470 * 471 * <li> A paragraph-separator character (<code>'\u2029'</code>). 472 * 473 * </ul> 474 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators 475 * recognized are newline characters. 476 * 477 * <p> The regular expression {@code .} matches any character except a line 478 * terminator unless the {@link #DOTALL} flag is specified. 479 * 480 * <p> By default, the regular expressions {@code ^} and {@code $} ignore 481 * line terminators and only match at the beginning and the end, respectively, 482 * of the entire input sequence. If {@link #MULTILINE} mode is activated then 483 * {@code ^} matches at the beginning of input and after any line terminator 484 * except at the end of input. When in {@link #MULTILINE} mode {@code $} 485 * matches just before a line terminator or the end of the input sequence. 486 * 487 * <h3><a id="cg">Groups and capturing</a></h3> 488 * 489 * <h4><a id="gnumber">Group number</a></h4> 490 * <p> Capturing groups are numbered by counting their opening parentheses from 491 * left to right. In the expression {@code ((A)(B(C)))}, for example, there 492 * are four such groups: </p> 493 * 494 * <ol style="margin-left:2em;"> 495 * <li> {@code ((A)(B(C)))} 496 * <li> {@code (A)} 497 * <li> {@code (B(C))} 498 * <li> {@code (C)} 499 * </ol> 500 * 501 * <p> Group zero always stands for the entire expression. 502 * 503 * <p> Capturing groups are so named because, during a match, each subsequence 504 * of the input sequence that matches such a group is saved. The captured 505 * subsequence may be used later in the expression, via a back reference, and 506 * may also be retrieved from the matcher once the match operation is complete. 507 * 508 * <h4><a id="groupname">Group name</a></h4> 509 * <p>A capturing group can also be assigned a "name", a {@code named-capturing group}, 510 * and then be back-referenced later by the "name". Group names are composed of 511 * the following characters. The first character must be a {@code letter}. 512 * 513 * <ul> 514 * <li> The uppercase letters {@code 'A'} through {@code 'Z'} 515 * (<code>'\u0041'</code> through <code>'\u005a'</code>), 516 * <li> The lowercase letters {@code 'a'} through {@code 'z'} 517 * (<code>'\u0061'</code> through <code>'\u007a'</code>), 518 * <li> The digits {@code '0'} through {@code '9'} 519 * (<code>'\u0030'</code> through <code>'\u0039'</code>), 520 * </ul> 521 * 522 * <p> A {@code named-capturing group} is still numbered as described in 523 * <a href="#gnumber">Group number</a>. 524 * 525 * <p> The captured input associated with a group is always the subsequence 526 * that the group most recently matched. If a group is evaluated a second time 527 * because of quantification then its previously-captured value, if any, will 528 * be retained if the second evaluation fails. Matching the string 529 * {@code "aba"} against the expression {@code (a(b)?)+}, for example, leaves 530 * group two set to {@code "b"}. All captured input is discarded at the 531 * beginning of each match. 532 * 533 * <p> Groups beginning with {@code (?} are either pure, <i>non-capturing</i> groups 534 * that do not capture text and do not count towards the group total, or 535 * <i>named-capturing</i> group. 536 * 537 * <h3> Unicode support </h3> 538 * 539 * <p> This class is in conformance with Level 1 of <a 540 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 541 * Standard #18: Unicode Regular Expression</i></a>, plus RL2.1 542 * Canonical Equivalents. 543 * <p> 544 * <b>Unicode escape sequences</b> such as <code>\u2014</code> in Java source code 545 * are processed as described in section 3.3 of 546 * <cite>The Java™ Language Specification</cite>. 547 * Such escape sequences are also implemented directly by the regular-expression 548 * parser so that Unicode escapes can be used in expressions that are read from 549 * files or from the keyboard. Thus the strings <code>"\u2014"</code> and 550 * {@code "\\u2014"}, while not equal, compile into the same pattern, which 551 * matches the character with hexadecimal value {@code 0x2014}. 552 * <p> 553 * A Unicode character can also be represented by using its <b>Hex notation</b> 554 * (hexadecimal code point value) directly as described in construct 555 * <code>\x{...}</code>, for example a supplementary character U+2011F can be 556 * specified as <code>\x{2011F}</code>, instead of two consecutive Unicode escape 557 * sequences of the surrogate pair <code>\uD840</code><code>\uDD1F</code>. 558 * <p> 559 * <b>Unicode character names</b> are supported by the named character construct 560 * <code>\N{</code>...<code>}</code>, for example, <code>\N{WHITE SMILING FACE}</code> 561 * specifies character <code>\u263A</code>. The character names supported 562 * by this class are the valid Unicode character names matched by 563 * {@link java.lang.Character#codePointOf(String) Character.codePointOf(name)}. 564 * <p> 565 * <a href="http://www.unicode.org/reports/tr18/#Default_Grapheme_Clusters"> 566 * <b>Unicode extended grapheme clusters</b></a> are supported by the grapheme 567 * cluster matcher {@code \X} and the corresponding boundary matcher {@code \b{g}}. 568 * <p> 569 * Unicode scripts, blocks, categories and binary properties are written with 570 * the {@code \p} and {@code \P} constructs as in Perl. 571 * <code>\p{</code><i>prop</i><code>}</code> matches if 572 * the input has the property <i>prop</i>, while <code>\P{</code><i>prop</i><code>}</code> 573 * does not match if the input has that property. 574 * <p> 575 * Scripts, blocks, categories and binary properties can be used both inside 576 * and outside of a character class. 577 * 578 * <p> 579 * <b><a id="usc">Scripts</a></b> are specified either with the prefix {@code Is}, as in 580 * {@code IsHiragana}, or by using the {@code script} keyword (or its short 581 * form {@code sc}) as in {@code script=Hiragana} or {@code sc=Hiragana}. 582 * <p> 583 * The script names supported by {@code Pattern} are the valid script names 584 * accepted and defined by 585 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}. 586 * 587 * <p> 588 * <b><a id="ubc">Blocks</a></b> are specified with the prefix {@code In}, as in 589 * {@code InMongolian}, or by using the keyword {@code block} (or its short 590 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}. 591 * <p> 592 * The block names supported by {@code Pattern} are the valid block names 593 * accepted and defined by 594 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}. 595 * <p> 596 * 597 * <b><a id="ucc">Categories</a></b> may be specified with the optional prefix {@code Is}: 598 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode 599 * letters. Same as scripts and blocks, categories can also be specified 600 * by using the keyword {@code general_category} (or its short form 601 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}. 602 * <p> 603 * The supported categories are those of 604 * <a href="http://www.unicode.org/unicode/standard/standard.html"> 605 * <i>The Unicode Standard</i></a> in the version specified by the 606 * {@link java.lang.Character Character} class. The category names are those 607 * defined in the Standard, both normative and informative. 608 * <p> 609 * 610 * <b><a id="ubpc">Binary properties</a></b> are specified with the prefix {@code Is}, as in 611 * {@code IsAlphabetic}. The supported binary properties by {@code Pattern} 612 * are 613 * <ul> 614 * <li> Alphabetic 615 * <li> Ideographic 616 * <li> Letter 617 * <li> Lowercase 618 * <li> Uppercase 619 * <li> Titlecase 620 * <li> Punctuation 621 * <Li> Control 622 * <li> White_Space 623 * <li> Digit 624 * <li> Hex_Digit 625 * <li> Join_Control 626 * <li> Noncharacter_Code_Point 627 * <li> Assigned 628 * </ul> 629 * <p> 630 * The following <b>Predefined Character classes</b> and <b>POSIX character classes</b> 631 * are in conformance with the recommendation of <i>Annex C: Compatibility Properties</i> 632 * of <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Regular Expression 633 * </i></a>, when {@link #UNICODE_CHARACTER_CLASS} flag is specified. 634 * 635 * <table class="striped"> 636 * <caption style="display:none">predefined and posix character classes in Unicode mode</caption> 637 * <thead> 638 * <tr> 639 * <th scope="col" id="predef_classes">Classes</th> 640 * <th scope="col" id="predef_matches">Matches</th> 641 * </tr> 642 * </thead> 643 * <tbody> 644 * <tr><th scope="row">{@code \p{Lower}}</th> 645 * <td>A lowercase character:{@code \p{IsLowercase}}</td></tr> 646 * <tr><th scope="row">{@code \p{Upper}}</th> 647 * <td>An uppercase character:{@code \p{IsUppercase}}</td></tr> 648 * <tr><th scope="row">{@code \p{ASCII}}</th> 649 * <td>All ASCII:{@code [\x00-\x7F]}</td></tr> 650 * <tr><th scope="row">{@code \p{Alpha}}</th> 651 * <td>An alphabetic character:{@code \p{IsAlphabetic}}</td></tr> 652 * <tr><th scope="row">{@code \p{Digit}}</th> 653 * <td>A decimal digit character:{@code \p{IsDigit}}</td></tr> 654 * <tr><th scope="row">{@code \p{Alnum}}</th> 655 * <td>An alphanumeric character:{@code [\p{IsAlphabetic}\p{IsDigit}]}</td></tr> 656 * <tr><th scope="row">{@code \p{Punct}}</th> 657 * <td>A punctuation character:{@code \p{IsPunctuation}}</td></tr> 658 * <tr><th scope="row">{@code \p{Graph}}</th> 659 * <td>A visible character: {@code [^\p{IsWhite_Space}\p{gc=Cc}\p{gc=Cs}\p{gc=Cn}]}</td></tr> 660 * <tr><th scope="row">{@code \p{Print}}</th> 661 * <td>A printable character: {@code [\p{Graph}\p{Blank}&&[^\p{Cntrl}]]}</td></tr> 662 * <tr><th scope="row">{@code \p{Blank}}</th> 663 * <td>A space or a tab: {@code [\p{IsWhite_Space}&&[^\p{gc=Zl}\p{gc=Zp}\x0a\x0b\x0c\x0d\x85]]}</td></tr> 664 * <tr><th scope="row">{@code \p{Cntrl}}</th> 665 * <td>A control character: {@code \p{gc=Cc}}</td></tr> 666 * <tr><th scope="row">{@code \p{XDigit}}</th> 667 * <td>A hexadecimal digit: {@code [\p{gc=Nd}\p{IsHex_Digit}]}</td></tr> 668 * <tr><th scope="row">{@code \p{Space}}</th> 669 * <td>A whitespace character:{@code \p{IsWhite_Space}}</td></tr> 670 * <tr><th scope="row">{@code \d}</th> 671 * <td>A digit: {@code \p{IsDigit}}</td></tr> 672 * <tr><th scope="row">{@code \D}</th> 673 * <td>A non-digit: {@code [^\d]}</td></tr> 674 * <tr><th scope="row">{@code \s}</th> 675 * <td>A whitespace character: {@code \p{IsWhite_Space}}</td></tr> 676 * <tr><th scope="row">{@code \S}</th> 677 * <td>A non-whitespace character: {@code [^\s]}</td></tr> 678 * <tr><th scope="row">{@code \w}</th> 679 * <td>A word character: {@code [\p{Alpha}\p{gc=Mn}\p{gc=Me}\p{gc=Mc}\p{Digit}\p{gc=Pc}\p{IsJoin_Control}]}</td></tr> 680 * <tr><th scope="row">{@code \W}</th> 681 * <td>A non-word character: {@code [^\w]}</td></tr> 682 * </tbody> 683 * </table> 684 * <p> 685 * <a id="jcc"> 686 * Categories that behave like the java.lang.Character 687 * boolean is<i>methodname</i> methods (except for the deprecated ones) are 688 * available through the same <code>\p{</code><i>prop</i><code>}</code> syntax where 689 * the specified property has the name <code>java<i>methodname</i></code></a>. 690 * 691 * <h3> Comparison to Perl 5 </h3> 692 * 693 * <p>The {@code Pattern} engine performs traditional NFA-based matching 694 * with ordered alternation as occurs in Perl 5. 695 * 696 * <p> Perl constructs not supported by this class: </p> 697 * 698 * <ul> 699 * <li><p> The backreference constructs, <code>\g{</code><i>n</i><code>}</code> for 700 * the <i>n</i><sup>th</sup><a href="#cg">capturing group</a> and 701 * <code>\g{</code><i>name</i><code>}</code> for 702 * <a href="#groupname">named-capturing group</a>. 703 * </p></li> 704 * 705 * <li><p> The conditional constructs 706 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code )} and 707 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code |}<i>Y</i>{@code )}, 708 * </p></li> 709 * 710 * <li><p> The embedded code constructs <code>(?{</code><i>code</i><code>})</code> 711 * and <code>(??{</code><i>code</i><code>})</code>,</p></li> 712 * 713 * <li><p> The embedded comment syntax {@code (?#comment)}, and </p></li> 714 * 715 * <li><p> The preprocessing operations {@code \l} <code>\u</code>, 716 * {@code \L}, and {@code \U}. </p></li> 717 * 718 * </ul> 719 * 720 * <p> Constructs supported by this class but not by Perl: </p> 721 * 722 * <ul> 723 * 724 * <li><p> Character-class union and intersection as described 725 * <a href="#cc">above</a>.</p></li> 726 * 727 * </ul> 728 * 729 * <p> Notable differences from Perl: </p> 730 * 731 * <ul> 732 * 733 * <li><p> In Perl, {@code \1} through {@code \9} are always interpreted 734 * as back references; a backslash-escaped number greater than {@code 9} is 735 * treated as a back reference if at least that many subexpressions exist, 736 * otherwise it is interpreted, if possible, as an octal escape. In this 737 * class octal escapes must always begin with a zero. In this class, 738 * {@code \1} through {@code \9} are always interpreted as back 739 * references, and a larger number is accepted as a back reference if at 740 * least that many subexpressions exist at that point in the regular 741 * expression, otherwise the parser will drop digits until the number is 742 * smaller or equal to the existing number of groups or it is one digit. 743 * </p></li> 744 * 745 * <li><p> Perl uses the {@code g} flag to request a match that resumes 746 * where the last match left off. This functionality is provided implicitly 747 * by the {@link Matcher} class: Repeated invocations of the {@link 748 * Matcher#find find} method will resume where the last match left off, 749 * unless the matcher is reset. </p></li> 750 * 751 * <li><p> In Perl, embedded flags at the top level of an expression affect 752 * the whole expression. In this class, embedded flags always take effect 753 * at the point at which they appear, whether they are at the top level or 754 * within a group; in the latter case, flags are restored at the end of the 755 * group just as in Perl. </p></li> 756 * 757 * </ul> 758 * 759 * 760 * <p> For a more precise description of the behavior of regular expression 761 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/"> 762 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl, 763 * O'Reilly and Associates, 2006.</a> 764 * </p> 765 * 766 * @see java.lang.String#split(String, int) 767 * @see java.lang.String#split(String) 768 * 769 * @author Mike McCloskey 770 * @author Mark Reinhold 771 * @author JSR-51 Expert Group 772 * @since 1.4 773 * @spec JSR-51 774 */ 775 776 public final class Pattern 777 implements java.io.Serializable 778 { 779 780 /** 781 * Regular expression modifier values. Instead of being passed as 782 * arguments, they can also be passed as inline modifiers. 783 * For example, the following statements have the same effect. 784 * <pre> 785 * RegExp r1 = RegExp.compile("abc", Pattern.I|Pattern.M); 786 * RegExp r2 = RegExp.compile("(?im)abc", 0); 787 * </pre> 788 * 789 * The flags are duplicated so that the familiar Perl match flag 790 * names are available. 791 */ 792 793 /** 794 * Enables Unix lines mode. 795 * 796 * <p> In this mode, only the {@code '\n'} line terminator is recognized 797 * in the behavior of {@code .}, {@code ^}, and {@code $}. 798 * 799 * <p> Unix lines mode can also be enabled via the embedded flag 800 * expression {@code (?d)}. 801 */ 802 public static final int UNIX_LINES = 0x01; 803 804 /** 805 * Enables case-insensitive matching. 806 * 807 * <p> By default, case-insensitive matching assumes that only characters 808 * in the US-ASCII charset are being matched. Unicode-aware 809 * case-insensitive matching can be enabled by specifying the {@link 810 * #UNICODE_CASE} flag in conjunction with this flag. 811 * 812 * <p> Case-insensitive matching can also be enabled via the embedded flag 813 * expression {@code (?i)}. 814 * 815 * <p> Specifying this flag may impose a slight performance penalty. </p> 816 */ 817 public static final int CASE_INSENSITIVE = 0x02; 818 819 /** 820 * Permits whitespace and comments in pattern. 821 * 822 * <p> In this mode, whitespace is ignored, and embedded comments starting 823 * with {@code #} are ignored until the end of a line. 824 * 825 * <p> Comments mode can also be enabled via the embedded flag 826 * expression {@code (?x)}. 827 */ 828 public static final int COMMENTS = 0x04; 829 830 /** 831 * Enables multiline mode. 832 * 833 * <p> In multiline mode the expressions {@code ^} and {@code $} match 834 * just after or just before, respectively, a line terminator or the end of 835 * the input sequence. By default these expressions only match at the 836 * beginning and the end of the entire input sequence. 837 * 838 * <p> Multiline mode can also be enabled via the embedded flag 839 * expression {@code (?m)}. </p> 840 */ 841 public static final int MULTILINE = 0x08; 842 843 /** 844 * Enables literal parsing of the pattern. 845 * 846 * <p> When this flag is specified then the input string that specifies 847 * the pattern is treated as a sequence of literal characters. 848 * Metacharacters or escape sequences in the input sequence will be 849 * given no special meaning. 850 * 851 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on 852 * matching when used in conjunction with this flag. The other flags 853 * become superfluous. 854 * 855 * <p> There is no embedded flag character for enabling literal parsing. 856 * @since 1.5 857 */ 858 public static final int LITERAL = 0x10; 859 860 /** 861 * Enables dotall mode. 862 * 863 * <p> In dotall mode, the expression {@code .} matches any character, 864 * including a line terminator. By default this expression does not match 865 * line terminators. 866 * 867 * <p> Dotall mode can also be enabled via the embedded flag 868 * expression {@code (?s)}. (The {@code s} is a mnemonic for 869 * "single-line" mode, which is what this is called in Perl.) </p> 870 */ 871 public static final int DOTALL = 0x20; 872 873 /** 874 * Enables Unicode-aware case folding. 875 * 876 * <p> When this flag is specified then case-insensitive matching, when 877 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner 878 * consistent with the Unicode Standard. By default, case-insensitive 879 * matching assumes that only characters in the US-ASCII charset are being 880 * matched. 881 * 882 * <p> Unicode-aware case folding can also be enabled via the embedded flag 883 * expression {@code (?u)}. 884 * 885 * <p> Specifying this flag may impose a performance penalty. </p> 886 */ 887 public static final int UNICODE_CASE = 0x40; 888 889 /** 890 * Enables canonical equivalence. 891 * 892 * <p> When this flag is specified then two characters will be considered 893 * to match if, and only if, their full canonical decompositions match. 894 * The expression <code>"a\u030A"</code>, for example, will match the 895 * string <code>"\u00E5"</code> when this flag is specified. By default, 896 * matching does not take canonical equivalence into account. 897 * 898 * <p> There is no embedded flag character for enabling canonical 899 * equivalence. 900 * 901 * <p> Specifying this flag may impose a performance penalty. </p> 902 */ 903 public static final int CANON_EQ = 0x80; 904 905 /** 906 * Enables the Unicode version of <i>Predefined character classes</i> and 907 * <i>POSIX character classes</i>. 908 * 909 * <p> When this flag is specified then the (US-ASCII only) 910 * <i>Predefined character classes</i> and <i>POSIX character classes</i> 911 * are in conformance with 912 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 913 * Standard #18: Unicode Regular Expression</i></a> 914 * <i>Annex C: Compatibility Properties</i>. 915 * <p> 916 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded 917 * flag expression {@code (?U)}. 918 * <p> 919 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case 920 * folding. 921 * <p> 922 * Specifying this flag may impose a performance penalty. </p> 923 * @since 1.7 924 */ 925 public static final int UNICODE_CHARACTER_CLASS = 0x100; 926 927 /** 928 * Contains all possible flags for compile(regex, flags). 929 */ 930 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE | 931 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL | 932 UNICODE_CHARACTER_CLASS | COMMENTS; 933 934 /* Pattern has only two serialized components: The pattern string 935 * and the flags, which are all that is needed to recompile the pattern 936 * when it is deserialized. 937 */ 938 939 /** use serialVersionUID from Merlin b59 for interoperability */ 940 private static final long serialVersionUID = 5073258162644648461L; 941 942 /** 943 * The original regular-expression pattern string. 944 * 945 * @serial 946 */ 947 private String pattern; 948 949 /** 950 * The original pattern flags. 951 * 952 * @serial 953 */ 954 private int flags; 955 956 /** 957 * The temporary pattern flags used during compiling. The flags might be turn 958 * on and off by embedded flag. 959 */ 960 private transient int flags0; 961 962 /** 963 * Boolean indicating this Pattern is compiled; this is necessary in order 964 * to lazily compile deserialized Patterns. 965 */ 966 private transient volatile boolean compiled; 967 968 /** 969 * The normalized pattern string. 970 */ 971 private transient String normalizedPattern; 972 973 /** 974 * The starting point of state machine for the find operation. This allows 975 * a match to start anywhere in the input. 976 */ 977 transient Node root; 978 979 /** 980 * The root of object tree for a match operation. The pattern is matched 981 * at the beginning. This may include a find that uses BnM or a First 982 * node. 983 */ 984 transient Node matchRoot; 985 986 /** 987 * Temporary storage used by parsing pattern slice. 988 */ 989 transient int[] buffer; 990 991 /** 992 * A temporary storage used for predicate for double return. 993 */ 994 transient CharPredicate predicate; 995 996 /** 997 * Map the "name" of the "named capturing group" to its group id 998 * node. 999 */ 1000 transient volatile Map<String, Integer> namedGroups; 1001 1002 /** 1003 * Temporary storage used while parsing group references. 1004 */ 1005 transient GroupHead[] groupNodes; 1006 1007 /** 1008 * Temporary storage used to store the top level closure nodes. 1009 */ 1010 transient List<Node> topClosureNodes; 1011 1012 /** 1013 * The number of top greedy closure nodes in this Pattern. Used by 1014 * matchers to allocate storage needed for a IntHashSet to keep the 1015 * beginning pos {@code i} of all failed match. 1016 */ 1017 transient int localTCNCount; 1018 1019 /* 1020 * Turn off the stop-exponential-backtracking optimization if there 1021 * is a group ref in the pattern. 1022 */ 1023 transient boolean hasGroupRef; 1024 1025 /** 1026 * Temporary null terminated code point array used by pattern compiling. 1027 */ 1028 private transient int[] temp; 1029 1030 /** 1031 * The number of capturing groups in this Pattern. Used by matchers to 1032 * allocate storage needed to perform a match. 1033 */ 1034 transient int capturingGroupCount; 1035 1036 /** 1037 * The local variable count used by parsing tree. Used by matchers to 1038 * allocate storage needed to perform a match. 1039 */ 1040 transient int localCount; 1041 1042 /** 1043 * Index into the pattern string that keeps track of how much has been 1044 * parsed. 1045 */ 1046 private transient int cursor; 1047 1048 /** 1049 * Holds the length of the pattern string. 1050 */ 1051 private transient int patternLength; 1052 1053 /** 1054 * If the Start node might possibly match supplementary characters. 1055 * It is set to true during compiling if 1056 * (1) There is supplementary char in pattern, or 1057 * (2) There is complement node of a "family" CharProperty 1058 */ 1059 private transient boolean hasSupplementary; 1060 1061 /** 1062 * Compiles the given regular expression into a pattern. 1063 * 1064 * @param regex 1065 * The expression to be compiled 1066 * @return the given regular expression compiled into a pattern 1067 * @throws PatternSyntaxException 1068 * If the expression's syntax is invalid 1069 */ 1070 public static Pattern compile(String regex) { 1071 return new Pattern(regex, 0); 1072 } 1073 1074 /** 1075 * Compiles the given regular expression into a pattern with the given 1076 * flags. 1077 * 1078 * @param regex 1079 * The expression to be compiled 1080 * 1081 * @param flags 1082 * Match flags, a bit mask that may include 1083 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL}, 1084 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES}, 1085 * {@link #LITERAL}, {@link #UNICODE_CHARACTER_CLASS} 1086 * and {@link #COMMENTS} 1087 * 1088 * @return the given regular expression compiled into a pattern with the given flags 1089 * @throws IllegalArgumentException 1090 * If bit values other than those corresponding to the defined 1091 * match flags are set in {@code flags} 1092 * 1093 * @throws PatternSyntaxException 1094 * If the expression's syntax is invalid 1095 */ 1096 public static Pattern compile(String regex, int flags) { 1097 return new Pattern(regex, flags); 1098 } 1099 1100 /** 1101 * Returns the regular expression from which this pattern was compiled. 1102 * 1103 * @return The source of this pattern 1104 */ 1105 public String pattern() { 1106 return pattern; 1107 } 1108 1109 /** 1110 * <p>Returns the string representation of this pattern. This 1111 * is the regular expression from which this pattern was 1112 * compiled.</p> 1113 * 1114 * @return The string representation of this pattern 1115 * @since 1.5 1116 */ 1117 public String toString() { 1118 return pattern; 1119 } 1120 1121 /** 1122 * Creates a matcher that will match the given input against this pattern. 1123 * 1124 * @param input 1125 * The character sequence to be matched 1126 * 1127 * @return A new matcher for this pattern 1128 */ 1129 public Matcher matcher(CharSequence input) { 1130 if (!compiled) { 1131 synchronized(this) { 1132 if (!compiled) 1133 compile(); 1134 } 1135 } 1136 Matcher m = new Matcher(this, input); 1137 return m; 1138 } 1139 1140 /** 1141 * Returns this pattern's match flags. 1142 * 1143 * @return The match flags specified when this pattern was compiled 1144 */ 1145 public int flags() { 1146 return flags0; 1147 } 1148 1149 /** 1150 * Compiles the given regular expression and attempts to match the given 1151 * input against it. 1152 * 1153 * <p> An invocation of this convenience method of the form 1154 * 1155 * <blockquote><pre> 1156 * Pattern.matches(regex, input);</pre></blockquote> 1157 * 1158 * behaves in exactly the same way as the expression 1159 * 1160 * <blockquote><pre> 1161 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote> 1162 * 1163 * <p> If a pattern is to be used multiple times, compiling it once and reusing 1164 * it will be more efficient than invoking this method each time. </p> 1165 * 1166 * @param regex 1167 * The expression to be compiled 1168 * 1169 * @param input 1170 * The character sequence to be matched 1171 * @return whether or not the regular expression matches on the input 1172 * @throws PatternSyntaxException 1173 * If the expression's syntax is invalid 1174 */ 1175 public static boolean matches(String regex, CharSequence input) { 1176 Pattern p = Pattern.compile(regex); 1177 Matcher m = p.matcher(input); 1178 return m.matches(); 1179 } 1180 1181 /** 1182 * Splits the given input sequence around matches of this pattern. 1183 * 1184 * <p> The array returned by this method contains each substring of the 1185 * input sequence that is terminated by another subsequence that matches 1186 * this pattern or is terminated by the end of the input sequence. The 1187 * substrings in the array are in the order in which they occur in the 1188 * input. If this pattern does not match any subsequence of the input then 1189 * the resulting array has just one element, namely the input sequence in 1190 * string form. 1191 * 1192 * <p> When there is a positive-width match at the beginning of the input 1193 * sequence then an empty leading substring is included at the beginning 1194 * of the resulting array. A zero-width match at the beginning however 1195 * never produces such empty leading substring. 1196 * 1197 * <p> The {@code limit} parameter controls the number of times the 1198 * pattern is applied and therefore affects the length of the resulting 1199 * array. If the limit <i>n</i> is greater than zero then the pattern 1200 * will be applied at most <i>n</i> - 1 times, the array's 1201 * length will be no greater than <i>n</i>, and the array's last entry 1202 * will contain all input beyond the last matched delimiter. If <i>n</i> 1203 * is non-positive then the pattern will be applied as many times as 1204 * possible and the array can have any length. If <i>n</i> is zero then 1205 * the pattern will be applied as many times as possible, the array can 1206 * have any length, and trailing empty strings will be discarded. 1207 * 1208 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1209 * results with these parameters: 1210 * 1211 * <table class="plain" style="margin-left:2em;"> 1212 * <caption style="display:none">Split example showing regex, limit, and result</caption> 1213 * <thead> 1214 * <tr> 1215 * <th scope="col">Regex</th> 1216 * <th scope="col">Limit</th> 1217 * <th scope="col">Result</th> 1218 * </tr> 1219 * </thead> 1220 * <tbody> 1221 * <tr><th scope="row" rowspan="3" style="font-weight:normal">:</th> 1222 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">2</th> 1223 * <td>{@code { "boo", "and:foo" }}</td></tr> 1224 * <tr><!-- : --> 1225 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1226 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1227 * <tr><!-- : --> 1228 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1229 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1230 * <tr><th scope="row" rowspan="3" style="font-weight:normal">o</th> 1231 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1232 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1233 * <tr><!-- o --> 1234 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1235 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1236 * <tr><!-- o --> 1237 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">0</th> 1238 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1239 * </tbody> 1240 * </table> 1241 * 1242 * @param input 1243 * The character sequence to be split 1244 * 1245 * @param limit 1246 * The result threshold, as described above 1247 * 1248 * @return The array of strings computed by splitting the input 1249 * around matches of this pattern 1250 */ 1251 public String[] split(CharSequence input, int limit) { 1252 int index = 0; 1253 boolean matchLimited = limit > 0; 1254 ArrayList<String> matchList = new ArrayList<>(); 1255 Matcher m = matcher(input); 1256 1257 // Add segments before each match found 1258 while(m.find()) { 1259 if (!matchLimited || matchList.size() < limit - 1) { 1260 if (index == 0 && index == m.start() && m.start() == m.end()) { 1261 // no empty leading substring included for zero-width match 1262 // at the beginning of the input char sequence. 1263 continue; 1264 } 1265 String match = input.subSequence(index, m.start()).toString(); 1266 matchList.add(match); 1267 index = m.end(); 1268 } else if (matchList.size() == limit - 1) { // last one 1269 String match = input.subSequence(index, 1270 input.length()).toString(); 1271 matchList.add(match); 1272 index = m.end(); 1273 } 1274 } 1275 1276 // If no match was found, return this 1277 if (index == 0) 1278 return new String[] {input.toString()}; 1279 1280 // Add remaining segment 1281 if (!matchLimited || matchList.size() < limit) 1282 matchList.add(input.subSequence(index, input.length()).toString()); 1283 1284 // Construct result 1285 int resultSize = matchList.size(); 1286 if (limit == 0) 1287 while (resultSize > 0 && matchList.get(resultSize-1).equals("")) 1288 resultSize--; 1289 String[] result = new String[resultSize]; 1290 return matchList.subList(0, resultSize).toArray(result); 1291 } 1292 1293 /** 1294 * Splits the given input sequence around matches of this pattern. 1295 * 1296 * <p> This method works as if by invoking the two-argument {@link 1297 * #split(java.lang.CharSequence, int) split} method with the given input 1298 * sequence and a limit argument of zero. Trailing empty strings are 1299 * therefore not included in the resulting array. </p> 1300 * 1301 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1302 * results with these expressions: 1303 * 1304 * <table class="plain" style="margin-left:2em"> 1305 * <caption style="display:none">Split examples showing regex and result</caption> 1306 * <thead> 1307 * <tr> 1308 * <th scope="col">Regex</th> 1309 * <th scope="col">Result</th> 1310 * </tr> 1311 * </thead> 1312 * <tbody> 1313 * <tr><th scope="row" style="text-weight:normal">:</th> 1314 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1315 * <tr><th scope="row" style="text-weight:normal">o</th> 1316 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1317 * </tbody> 1318 * </table> 1319 * 1320 * 1321 * @param input 1322 * The character sequence to be split 1323 * 1324 * @return The array of strings computed by splitting the input 1325 * around matches of this pattern 1326 */ 1327 public String[] split(CharSequence input) { 1328 return split(input, 0); 1329 } 1330 1331 /** 1332 * Returns a literal pattern {@code String} for the specified 1333 * {@code String}. 1334 * 1335 * <p>This method produces a {@code String} that can be used to 1336 * create a {@code Pattern} that would match the string 1337 * {@code s} as if it were a literal pattern.</p> Metacharacters 1338 * or escape sequences in the input sequence will be given no special 1339 * meaning. 1340 * 1341 * @param s The string to be literalized 1342 * @return A literal string replacement 1343 * @since 1.5 1344 */ 1345 public static String quote(String s) { 1346 int slashEIndex = s.indexOf("\\E"); 1347 if (slashEIndex == -1) 1348 return "\\Q" + s + "\\E"; 1349 1350 int lenHint = s.length(); 1351 lenHint = (lenHint < Integer.MAX_VALUE - 8 - lenHint) ? 1352 (lenHint << 1) : (Integer.MAX_VALUE - 8); 1353 1354 StringBuilder sb = new StringBuilder(lenHint); 1355 sb.append("\\Q"); 1356 int current = 0; 1357 do { 1358 sb.append(s, current, slashEIndex) 1359 .append("\\E\\\\E\\Q"); 1360 current = slashEIndex + 2; 1361 } while ((slashEIndex = s.indexOf("\\E", current)) != -1); 1362 1363 return sb.append(s, current, s.length()) 1364 .append("\\E") 1365 .toString(); 1366 } 1367 1368 /** 1369 * Recompile the Pattern instance from a stream. The original pattern 1370 * string is read in and the object tree is recompiled from it. 1371 */ 1372 private void readObject(java.io.ObjectInputStream s) 1373 throws java.io.IOException, ClassNotFoundException { 1374 1375 // Read in all fields 1376 s.defaultReadObject(); 1377 1378 // reset the flags 1379 flags0 = flags; 1380 1381 // Initialize counts 1382 capturingGroupCount = 1; 1383 localCount = 0; 1384 localTCNCount = 0; 1385 1386 // if length > 0, the Pattern is lazily compiled 1387 if (pattern.length() == 0) { 1388 root = new Start(lastAccept); 1389 matchRoot = lastAccept; 1390 compiled = true; 1391 } 1392 } 1393 1394 /** 1395 * This private constructor is used to create all Patterns. The pattern 1396 * string and match flags are all that is needed to completely describe 1397 * a Pattern. An empty pattern string results in an object tree with 1398 * only a Start node and a LastNode node. 1399 */ 1400 private Pattern(String p, int f) { 1401 if ((f & ~ALL_FLAGS) != 0) { 1402 throw new IllegalArgumentException("Unknown flag 0x" 1403 + Integer.toHexString(f)); 1404 } 1405 pattern = p; 1406 flags = f; 1407 1408 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present 1409 if ((flags & UNICODE_CHARACTER_CLASS) != 0) 1410 flags |= UNICODE_CASE; 1411 1412 // 'flags' for compiling 1413 flags0 = flags; 1414 1415 // Reset group index count 1416 capturingGroupCount = 1; 1417 localCount = 0; 1418 localTCNCount = 0; 1419 1420 if (pattern.length() > 0) { 1421 compile(); 1422 } else { 1423 root = new Start(lastAccept); 1424 matchRoot = lastAccept; 1425 } 1426 } 1427 1428 /** 1429 * The pattern is converted to normalized form ({@link 1430 * java.text.Normalizer.Form.NFC NFC}, canonical decomposition, 1431 * followed by canonical composition for the character class 1432 * part, and {@link java.text.Normalizer.Form.NFD NFD}, 1433 * canonical decomposition) for the rest), and then a pure 1434 * group is constructed to match canonical equivalences of the 1435 * characters. 1436 */ 1437 private static String normalize(String pattern) { 1438 int plen = pattern.length(); 1439 StringBuilder pbuf = new StringBuilder(plen); 1440 char last = 0; 1441 int lastStart = 0; 1442 char cc = 0; 1443 for (int i = 0; i < plen;) { 1444 char c = pattern.charAt(i); 1445 if (cc == 0 && // top level 1446 c == '\\' && i + 1 < plen && pattern.charAt(i + 1) == '\\') { 1447 i += 2; last = 0; 1448 continue; 1449 } 1450 if (c == '[' && last != '\\') { 1451 if (cc == 0) { 1452 if (lastStart < i) 1453 normalizeSlice(pattern, lastStart, i, pbuf); 1454 lastStart = i; 1455 } 1456 cc++; 1457 } else if (c == ']' && last != '\\') { 1458 cc--; 1459 if (cc == 0) { 1460 normalizeClazz(pattern, lastStart, i + 1, pbuf); 1461 lastStart = i + 1; 1462 } 1463 } 1464 last = c; 1465 i++; 1466 } 1467 assert (cc == 0); 1468 if (lastStart < plen) 1469 normalizeSlice(pattern, lastStart, plen, pbuf); 1470 return pbuf.toString(); 1471 } 1472 1473 private static void normalizeSlice(String src, int off, int limit, 1474 StringBuilder dst) 1475 { 1476 int len = src.length(); 1477 int off0 = off; 1478 while (off < limit && ASCII.isAscii(src.charAt(off))) { 1479 off++; 1480 } 1481 if (off == limit) { 1482 dst.append(src, off0, limit); 1483 return; 1484 } 1485 off--; 1486 if (off < off0) 1487 off = off0; 1488 else 1489 dst.append(src, off0, off); 1490 while (off < limit) { 1491 int ch0 = src.codePointAt(off); 1492 if (".$|()[]{}^?*+\\".indexOf(ch0) != -1) { 1493 dst.append((char)ch0); 1494 off++; 1495 continue; 1496 } 1497 int j = off + Character.charCount(ch0); 1498 int ch1; 1499 while (j < limit) { 1500 ch1 = src.codePointAt(j); 1501 if (Grapheme.isBoundary(ch0, ch1)) 1502 break; 1503 ch0 = ch1; 1504 j += Character.charCount(ch1); 1505 } 1506 String seq = src.substring(off, j); 1507 String nfd = Normalizer.normalize(seq, Normalizer.Form.NFD); 1508 off = j; 1509 if (nfd.length() > 1) { 1510 ch0 = nfd.codePointAt(0); 1511 ch1 = nfd.codePointAt(Character.charCount(ch0)); 1512 if (Character.getType(ch1) == Character.NON_SPACING_MARK) { 1513 Set<String> altns = new LinkedHashSet<>(); 1514 altns.add(seq); 1515 produceEquivalentAlternation(nfd, altns); 1516 dst.append("(?:"); 1517 altns.forEach( s -> dst.append(s).append('|')); 1518 dst.delete(dst.length() - 1, dst.length()); 1519 dst.append(")"); 1520 continue; 1521 } 1522 } 1523 String nfc = Normalizer.normalize(seq, Normalizer.Form.NFC); 1524 if (!seq.equals(nfc) && !nfd.equals(nfc)) 1525 dst.append("(?:" + seq + "|" + nfd + "|" + nfc + ")"); 1526 else if (!seq.equals(nfd)) 1527 dst.append("(?:" + seq + "|" + nfd + ")"); 1528 else 1529 dst.append(seq); 1530 } 1531 } 1532 1533 private static void normalizeClazz(String src, int off, int limit, 1534 StringBuilder dst) 1535 { 1536 dst.append(Normalizer.normalize(src.substring(off, limit), Form.NFC)); 1537 } 1538 1539 /** 1540 * Given a specific sequence composed of a regular character and 1541 * combining marks that follow it, produce the alternation that will 1542 * match all canonical equivalences of that sequence. 1543 */ 1544 private static void produceEquivalentAlternation(String src, 1545 Set<String> dst) 1546 { 1547 int len = countChars(src, 0, 1); 1548 if (src.length() == len) { 1549 dst.add(src); // source has one character. 1550 return; 1551 } 1552 String base = src.substring(0,len); 1553 String combiningMarks = src.substring(len); 1554 String[] perms = producePermutations(combiningMarks); 1555 // Add combined permutations 1556 for(int x = 0; x < perms.length; x++) { 1557 String next = base + perms[x]; 1558 dst.add(next); 1559 next = composeOneStep(next); 1560 if (next != null) { 1561 produceEquivalentAlternation(next, dst); 1562 } 1563 } 1564 } 1565 1566 /** 1567 * Returns an array of strings that have all the possible 1568 * permutations of the characters in the input string. 1569 * This is used to get a list of all possible orderings 1570 * of a set of combining marks. Note that some of the permutations 1571 * are invalid because of combining class collisions, and these 1572 * possibilities must be removed because they are not canonically 1573 * equivalent. 1574 */ 1575 private static String[] producePermutations(String input) { 1576 if (input.length() == countChars(input, 0, 1)) 1577 return new String[] {input}; 1578 1579 if (input.length() == countChars(input, 0, 2)) { 1580 int c0 = Character.codePointAt(input, 0); 1581 int c1 = Character.codePointAt(input, Character.charCount(c0)); 1582 if (getClass(c1) == getClass(c0)) { 1583 return new String[] {input}; 1584 } 1585 String[] result = new String[2]; 1586 result[0] = input; 1587 StringBuilder sb = new StringBuilder(2); 1588 sb.appendCodePoint(c1); 1589 sb.appendCodePoint(c0); 1590 result[1] = sb.toString(); 1591 return result; 1592 } 1593 1594 int length = 1; 1595 int nCodePoints = countCodePoints(input); 1596 for(int x=1; x<nCodePoints; x++) 1597 length = length * (x+1); 1598 1599 String[] temp = new String[length]; 1600 1601 int combClass[] = new int[nCodePoints]; 1602 for(int x=0, i=0; x<nCodePoints; x++) { 1603 int c = Character.codePointAt(input, i); 1604 combClass[x] = getClass(c); 1605 i += Character.charCount(c); 1606 } 1607 1608 // For each char, take it out and add the permutations 1609 // of the remaining chars 1610 int index = 0; 1611 int len; 1612 // offset maintains the index in code units. 1613 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) { 1614 len = countChars(input, offset, 1); 1615 for(int y=x-1; y>=0; y--) { 1616 if (combClass[y] == combClass[x]) { 1617 continue loop; 1618 } 1619 } 1620 StringBuilder sb = new StringBuilder(input); 1621 String otherChars = sb.delete(offset, offset+len).toString(); 1622 String[] subResult = producePermutations(otherChars); 1623 1624 String prefix = input.substring(offset, offset+len); 1625 for (String sre : subResult) 1626 temp[index++] = prefix + sre; 1627 } 1628 String[] result = new String[index]; 1629 System.arraycopy(temp, 0, result, 0, index); 1630 return result; 1631 } 1632 1633 private static int getClass(int c) { 1634 return sun.text.Normalizer.getCombiningClass(c); 1635 } 1636 1637 /** 1638 * Attempts to compose input by combining the first character 1639 * with the first combining mark following it. Returns a String 1640 * that is the composition of the leading character with its first 1641 * combining mark followed by the remaining combining marks. Returns 1642 * null if the first two characters cannot be further composed. 1643 */ 1644 private static String composeOneStep(String input) { 1645 int len = countChars(input, 0, 2); 1646 String firstTwoCharacters = input.substring(0, len); 1647 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC); 1648 if (result.equals(firstTwoCharacters)) 1649 return null; 1650 else { 1651 String remainder = input.substring(len); 1652 return result + remainder; 1653 } 1654 } 1655 1656 /** 1657 * Preprocess any \Q...\E sequences in `temp', meta-quoting them. 1658 * See the description of `quotemeta' in perlfunc(1). 1659 */ 1660 private void RemoveQEQuoting() { 1661 final int pLen = patternLength; 1662 int i = 0; 1663 while (i < pLen-1) { 1664 if (temp[i] != '\\') 1665 i += 1; 1666 else if (temp[i + 1] != 'Q') 1667 i += 2; 1668 else 1669 break; 1670 } 1671 if (i >= pLen - 1) // No \Q sequence found 1672 return; 1673 int j = i; 1674 i += 2; 1675 int[] newtemp = new int[j + 3*(pLen-i) + 2]; 1676 System.arraycopy(temp, 0, newtemp, 0, j); 1677 1678 boolean inQuote = true; 1679 boolean beginQuote = true; 1680 while (i < pLen) { 1681 int c = temp[i++]; 1682 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) { 1683 newtemp[j++] = c; 1684 } else if (ASCII.isDigit(c)) { 1685 if (beginQuote) { 1686 /* 1687 * A unicode escape \[0xu] could be before this quote, 1688 * and we don't want this numeric char to processed as 1689 * part of the escape. 1690 */ 1691 newtemp[j++] = '\\'; 1692 newtemp[j++] = 'x'; 1693 newtemp[j++] = '3'; 1694 } 1695 newtemp[j++] = c; 1696 } else if (c != '\\') { 1697 if (inQuote) newtemp[j++] = '\\'; 1698 newtemp[j++] = c; 1699 } else if (inQuote) { 1700 if (temp[i] == 'E') { 1701 i++; 1702 inQuote = false; 1703 } else { 1704 newtemp[j++] = '\\'; 1705 newtemp[j++] = '\\'; 1706 } 1707 } else { 1708 if (temp[i] == 'Q') { 1709 i++; 1710 inQuote = true; 1711 beginQuote = true; 1712 continue; 1713 } else { 1714 newtemp[j++] = c; 1715 if (i != pLen) 1716 newtemp[j++] = temp[i++]; 1717 } 1718 } 1719 1720 beginQuote = false; 1721 } 1722 1723 patternLength = j; 1724 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination 1725 } 1726 1727 /** 1728 * Copies regular expression to an int array and invokes the parsing 1729 * of the expression which will create the object tree. 1730 */ 1731 private void compile() { 1732 // Handle canonical equivalences 1733 if (has(CANON_EQ) && !has(LITERAL)) { 1734 normalizedPattern = normalize(pattern); 1735 } else { 1736 normalizedPattern = pattern; 1737 } 1738 patternLength = normalizedPattern.length(); 1739 1740 // Copy pattern to int array for convenience 1741 // Use double zero to terminate pattern 1742 temp = new int[patternLength + 2]; 1743 1744 hasSupplementary = false; 1745 int c, count = 0; 1746 // Convert all chars into code points 1747 for (int x = 0; x < patternLength; x += Character.charCount(c)) { 1748 c = normalizedPattern.codePointAt(x); 1749 if (isSupplementary(c)) { 1750 hasSupplementary = true; 1751 } 1752 temp[count++] = c; 1753 } 1754 1755 patternLength = count; // patternLength now in code points 1756 1757 if (! has(LITERAL)) 1758 RemoveQEQuoting(); 1759 1760 // Allocate all temporary objects here. 1761 buffer = new int[32]; 1762 groupNodes = new GroupHead[10]; 1763 namedGroups = null; 1764 topClosureNodes = new ArrayList<>(10); 1765 1766 if (has(LITERAL)) { 1767 // Literal pattern handling 1768 matchRoot = newSlice(temp, patternLength, hasSupplementary); 1769 matchRoot.next = lastAccept; 1770 } else { 1771 // Start recursive descent parsing 1772 matchRoot = expr(lastAccept); 1773 // Check extra pattern characters 1774 if (patternLength != cursor) { 1775 if (peek() == ')') { 1776 throw error("Unmatched closing ')'"); 1777 } else { 1778 throw error("Unexpected internal error"); 1779 } 1780 } 1781 } 1782 1783 // Peephole optimization 1784 if (matchRoot instanceof Slice) { 1785 root = BnM.optimize(matchRoot); 1786 if (root == matchRoot) { 1787 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1788 } 1789 } else if (matchRoot instanceof Begin || matchRoot instanceof First) { 1790 root = matchRoot; 1791 } else { 1792 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1793 } 1794 1795 // Optimize the greedy Loop to prevent exponential backtracking, IF there 1796 // is no group ref in this pattern. With a non-negative localTCNCount value, 1797 // the greedy type Loop, Curly will skip the backtracking for any starting 1798 // position "i" that failed in the past. 1799 if (!hasGroupRef) { 1800 for (Node node : topClosureNodes) { 1801 if (node instanceof Loop) { 1802 // non-deterministic-greedy-group 1803 ((Loop)node).posIndex = localTCNCount++; 1804 } 1805 } 1806 } 1807 1808 // Release temporary storage 1809 temp = null; 1810 buffer = null; 1811 groupNodes = null; 1812 patternLength = 0; 1813 compiled = true; 1814 topClosureNodes = null; 1815 } 1816 1817 Map<String, Integer> namedGroups() { 1818 Map<String, Integer> groups = namedGroups; 1819 if (groups == null) { 1820 namedGroups = groups = new HashMap<>(2); 1821 } 1822 return groups; 1823 } 1824 1825 /** 1826 * Used to accumulate information about a subtree of the object graph 1827 * so that optimizations can be applied to the subtree. 1828 */ 1829 static final class TreeInfo { 1830 int minLength; 1831 int maxLength; 1832 boolean maxValid; 1833 boolean deterministic; 1834 1835 TreeInfo() { 1836 reset(); 1837 } 1838 void reset() { 1839 minLength = 0; 1840 maxLength = 0; 1841 maxValid = true; 1842 deterministic = true; 1843 } 1844 } 1845 1846 /* 1847 * The following private methods are mainly used to improve the 1848 * readability of the code. In order to let the Java compiler easily 1849 * inline them, we should not put many assertions or error checks in them. 1850 */ 1851 1852 /** 1853 * Indicates whether a particular flag is set or not. 1854 */ 1855 private boolean has(int f) { 1856 return (flags0 & f) != 0; 1857 } 1858 1859 /** 1860 * Match next character, signal error if failed. 1861 */ 1862 private void accept(int ch, String s) { 1863 int testChar = temp[cursor++]; 1864 if (has(COMMENTS)) 1865 testChar = parsePastWhitespace(testChar); 1866 if (ch != testChar) { 1867 throw error(s); 1868 } 1869 } 1870 1871 /** 1872 * Mark the end of pattern with a specific character. 1873 */ 1874 private void mark(int c) { 1875 temp[patternLength] = c; 1876 } 1877 1878 /** 1879 * Peek the next character, and do not advance the cursor. 1880 */ 1881 private int peek() { 1882 int ch = temp[cursor]; 1883 if (has(COMMENTS)) 1884 ch = peekPastWhitespace(ch); 1885 return ch; 1886 } 1887 1888 /** 1889 * Read the next character, and advance the cursor by one. 1890 */ 1891 private int read() { 1892 int ch = temp[cursor++]; 1893 if (has(COMMENTS)) 1894 ch = parsePastWhitespace(ch); 1895 return ch; 1896 } 1897 1898 /** 1899 * Read the next character, and advance the cursor by one, 1900 * ignoring the COMMENTS setting 1901 */ 1902 private int readEscaped() { 1903 int ch = temp[cursor++]; 1904 return ch; 1905 } 1906 1907 /** 1908 * Advance the cursor by one, and peek the next character. 1909 */ 1910 private int next() { 1911 int ch = temp[++cursor]; 1912 if (has(COMMENTS)) 1913 ch = peekPastWhitespace(ch); 1914 return ch; 1915 } 1916 1917 /** 1918 * Advance the cursor by one, and peek the next character, 1919 * ignoring the COMMENTS setting 1920 */ 1921 private int nextEscaped() { 1922 int ch = temp[++cursor]; 1923 return ch; 1924 } 1925 1926 /** 1927 * If in xmode peek past whitespace and comments. 1928 */ 1929 private int peekPastWhitespace(int ch) { 1930 while (ASCII.isSpace(ch) || ch == '#') { 1931 while (ASCII.isSpace(ch)) 1932 ch = temp[++cursor]; 1933 if (ch == '#') { 1934 ch = peekPastLine(); 1935 } 1936 } 1937 return ch; 1938 } 1939 1940 /** 1941 * If in xmode parse past whitespace and comments. 1942 */ 1943 private int parsePastWhitespace(int ch) { 1944 while (ASCII.isSpace(ch) || ch == '#') { 1945 while (ASCII.isSpace(ch)) 1946 ch = temp[cursor++]; 1947 if (ch == '#') 1948 ch = parsePastLine(); 1949 } 1950 return ch; 1951 } 1952 1953 /** 1954 * xmode parse past comment to end of line. 1955 */ 1956 private int parsePastLine() { 1957 int ch = temp[cursor++]; 1958 while (ch != 0 && !isLineSeparator(ch)) 1959 ch = temp[cursor++]; 1960 return ch; 1961 } 1962 1963 /** 1964 * xmode peek past comment to end of line. 1965 */ 1966 private int peekPastLine() { 1967 int ch = temp[++cursor]; 1968 while (ch != 0 && !isLineSeparator(ch)) 1969 ch = temp[++cursor]; 1970 return ch; 1971 } 1972 1973 /** 1974 * Determines if character is a line separator in the current mode 1975 */ 1976 private boolean isLineSeparator(int ch) { 1977 if (has(UNIX_LINES)) { 1978 return ch == '\n'; 1979 } else { 1980 return (ch == '\n' || 1981 ch == '\r' || 1982 (ch|1) == '\u2029' || 1983 ch == '\u0085'); 1984 } 1985 } 1986 1987 /** 1988 * Read the character after the next one, and advance the cursor by two. 1989 */ 1990 private int skip() { 1991 int i = cursor; 1992 int ch = temp[i+1]; 1993 cursor = i + 2; 1994 return ch; 1995 } 1996 1997 /** 1998 * Unread one next character, and retreat cursor by one. 1999 */ 2000 private void unread() { 2001 cursor--; 2002 } 2003 2004 /** 2005 * Internal method used for handling all syntax errors. The pattern is 2006 * displayed with a pointer to aid in locating the syntax error. 2007 */ 2008 private PatternSyntaxException error(String s) { 2009 return new PatternSyntaxException(s, normalizedPattern, cursor - 1); 2010 } 2011 2012 /** 2013 * Determines if there is any supplementary character or unpaired 2014 * surrogate in the specified range. 2015 */ 2016 private boolean findSupplementary(int start, int end) { 2017 for (int i = start; i < end; i++) { 2018 if (isSupplementary(temp[i])) 2019 return true; 2020 } 2021 return false; 2022 } 2023 2024 /** 2025 * Determines if the specified code point is a supplementary 2026 * character or unpaired surrogate. 2027 */ 2028 private static final boolean isSupplementary(int ch) { 2029 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || 2030 Character.isSurrogate((char)ch); 2031 } 2032 2033 /** 2034 * The following methods handle the main parsing. They are sorted 2035 * according to their precedence order, the lowest one first. 2036 */ 2037 2038 /** 2039 * The expression is parsed with branch nodes added for alternations. 2040 * This may be called recursively to parse sub expressions that may 2041 * contain alternations. 2042 */ 2043 private Node expr(Node end) { 2044 Node prev = null; 2045 Node firstTail = null; 2046 Branch branch = null; 2047 Node branchConn = null; 2048 2049 for (;;) { 2050 Node node = sequence(end); 2051 Node nodeTail = root; //double return 2052 if (prev == null) { 2053 prev = node; 2054 firstTail = nodeTail; 2055 } else { 2056 // Branch 2057 if (branchConn == null) { 2058 branchConn = new BranchConn(); 2059 branchConn.next = end; 2060 } 2061 if (node == end) { 2062 // if the node returned from sequence() is "end" 2063 // we have an empty expr, set a null atom into 2064 // the branch to indicate to go "next" directly. 2065 node = null; 2066 } else { 2067 // the "tail.next" of each atom goes to branchConn 2068 nodeTail.next = branchConn; 2069 } 2070 if (prev == branch) { 2071 branch.add(node); 2072 } else { 2073 if (prev == end) { 2074 prev = null; 2075 } else { 2076 // replace the "end" with "branchConn" at its tail.next 2077 // when put the "prev" into the branch as the first atom. 2078 firstTail.next = branchConn; 2079 } 2080 prev = branch = new Branch(prev, node, branchConn); 2081 } 2082 } 2083 if (peek() != '|') { 2084 return prev; 2085 } 2086 next(); 2087 } 2088 } 2089 2090 @SuppressWarnings("fallthrough") 2091 /** 2092 * Parsing of sequences between alternations. 2093 */ 2094 private Node sequence(Node end) { 2095 Node head = null; 2096 Node tail = null; 2097 Node node = null; 2098 LOOP: 2099 for (;;) { 2100 int ch = peek(); 2101 switch (ch) { 2102 case '(': 2103 // Because group handles its own closure, 2104 // we need to treat it differently 2105 node = group0(); 2106 // Check for comment or flag group 2107 if (node == null) 2108 continue; 2109 if (head == null) 2110 head = node; 2111 else 2112 tail.next = node; 2113 // Double return: Tail was returned in root 2114 tail = root; 2115 continue; 2116 case '[': 2117 if (has(CANON_EQ) && !has(LITERAL)) 2118 node = new NFCCharProperty(clazz(true)); 2119 else 2120 node = newCharProperty(clazz(true)); 2121 break; 2122 case '\\': 2123 ch = nextEscaped(); 2124 if (ch == 'p' || ch == 'P') { 2125 boolean oneLetter = true; 2126 boolean comp = (ch == 'P'); 2127 ch = next(); // Consume { if present 2128 if (ch != '{') { 2129 unread(); 2130 } else { 2131 oneLetter = false; 2132 } 2133 // node = newCharProperty(family(oneLetter, comp)); 2134 if (has(CANON_EQ) && !has(LITERAL)) 2135 node = new NFCCharProperty(family(oneLetter, comp)); 2136 else 2137 node = newCharProperty(family(oneLetter, comp)); 2138 } else { 2139 unread(); 2140 node = atom(); 2141 } 2142 break; 2143 case '^': 2144 next(); 2145 if (has(MULTILINE)) { 2146 if (has(UNIX_LINES)) 2147 node = new UnixCaret(); 2148 else 2149 node = new Caret(); 2150 } else { 2151 node = new Begin(); 2152 } 2153 break; 2154 case '$': 2155 next(); 2156 if (has(UNIX_LINES)) 2157 node = new UnixDollar(has(MULTILINE)); 2158 else 2159 node = new Dollar(has(MULTILINE)); 2160 break; 2161 case '.': 2162 next(); 2163 if (has(DOTALL)) { 2164 node = new CharProperty(ALL()); 2165 } else { 2166 if (has(UNIX_LINES)) { 2167 node = new CharProperty(UNIXDOT()); 2168 } else { 2169 node = new CharProperty(DOT()); 2170 } 2171 } 2172 break; 2173 case '|': 2174 case ')': 2175 break LOOP; 2176 case ']': // Now interpreting dangling ] and } as literals 2177 case '}': 2178 node = atom(); 2179 break; 2180 case '?': 2181 case '*': 2182 case '+': 2183 next(); 2184 throw error("Dangling meta character '" + ((char)ch) + "'"); 2185 case 0: 2186 if (cursor >= patternLength) { 2187 break LOOP; 2188 } 2189 // Fall through 2190 default: 2191 node = atom(); 2192 break; 2193 } 2194 2195 node = closure(node); 2196 /* save the top dot-greedy nodes (.*, .+) as well 2197 if (node instanceof GreedyCharProperty && 2198 ((GreedyCharProperty)node).cp instanceof Dot) { 2199 topClosureNodes.add(node); 2200 } 2201 */ 2202 if (head == null) { 2203 head = tail = node; 2204 } else { 2205 tail.next = node; 2206 tail = node; 2207 } 2208 } 2209 if (head == null) { 2210 return end; 2211 } 2212 tail.next = end; 2213 root = tail; //double return 2214 return head; 2215 } 2216 2217 @SuppressWarnings("fallthrough") 2218 /** 2219 * Parse and add a new Single or Slice. 2220 */ 2221 private Node atom() { 2222 int first = 0; 2223 int prev = -1; 2224 boolean hasSupplementary = false; 2225 int ch = peek(); 2226 for (;;) { 2227 switch (ch) { 2228 case '*': 2229 case '+': 2230 case '?': 2231 case '{': 2232 if (first > 1) { 2233 cursor = prev; // Unwind one character 2234 first--; 2235 } 2236 break; 2237 case '$': 2238 case '.': 2239 case '^': 2240 case '(': 2241 case '[': 2242 case '|': 2243 case ')': 2244 break; 2245 case '\\': 2246 ch = nextEscaped(); 2247 if (ch == 'p' || ch == 'P') { // Property 2248 if (first > 0) { // Slice is waiting; handle it first 2249 unread(); 2250 break; 2251 } else { // No slice; just return the family node 2252 boolean comp = (ch == 'P'); 2253 boolean oneLetter = true; 2254 ch = next(); // Consume { if present 2255 if (ch != '{') 2256 unread(); 2257 else 2258 oneLetter = false; 2259 if (has(CANON_EQ) && !has(LITERAL)) 2260 return new NFCCharProperty(family(oneLetter, comp)); 2261 else 2262 return newCharProperty(family(oneLetter, comp)); 2263 } 2264 } 2265 unread(); 2266 prev = cursor; 2267 ch = escape(false, first == 0, false); 2268 if (ch >= 0) { 2269 append(ch, first); 2270 first++; 2271 if (isSupplementary(ch)) { 2272 hasSupplementary = true; 2273 } 2274 ch = peek(); 2275 continue; 2276 } else if (first == 0) { 2277 return root; 2278 } 2279 // Unwind meta escape sequence 2280 cursor = prev; 2281 break; 2282 case 0: 2283 if (cursor >= patternLength) { 2284 break; 2285 } 2286 // Fall through 2287 default: 2288 prev = cursor; 2289 append(ch, first); 2290 first++; 2291 if (isSupplementary(ch)) { 2292 hasSupplementary = true; 2293 } 2294 ch = next(); 2295 continue; 2296 } 2297 break; 2298 } 2299 if (first == 1) { 2300 return newCharProperty(single(buffer[0])); 2301 } else { 2302 return newSlice(buffer, first, hasSupplementary); 2303 } 2304 } 2305 2306 private void append(int ch, int len) { 2307 if (len >= buffer.length) { 2308 int[] tmp = new int[len+len]; 2309 System.arraycopy(buffer, 0, tmp, 0, len); 2310 buffer = tmp; 2311 } 2312 buffer[len] = ch; 2313 } 2314 2315 /** 2316 * Parses a backref greedily, taking as many numbers as it 2317 * can. The first digit is always treated as a backref, but 2318 * multi digit numbers are only treated as a backref if at 2319 * least that many backrefs exist at this point in the regex. 2320 */ 2321 private Node ref(int refNum) { 2322 boolean done = false; 2323 while(!done) { 2324 int ch = peek(); 2325 switch(ch) { 2326 case '0': 2327 case '1': 2328 case '2': 2329 case '3': 2330 case '4': 2331 case '5': 2332 case '6': 2333 case '7': 2334 case '8': 2335 case '9': 2336 int newRefNum = (refNum * 10) + (ch - '0'); 2337 // Add another number if it doesn't make a group 2338 // that doesn't exist 2339 if (capturingGroupCount - 1 < newRefNum) { 2340 done = true; 2341 break; 2342 } 2343 refNum = newRefNum; 2344 read(); 2345 break; 2346 default: 2347 done = true; 2348 break; 2349 } 2350 } 2351 hasGroupRef = true; 2352 if (has(CASE_INSENSITIVE)) 2353 return new CIBackRef(refNum, has(UNICODE_CASE)); 2354 else 2355 return new BackRef(refNum); 2356 } 2357 2358 /** 2359 * Parses an escape sequence to determine the actual value that needs 2360 * to be matched. 2361 * If -1 is returned and create was true a new object was added to the tree 2362 * to handle the escape sequence. 2363 * If the returned value is greater than zero, it is the value that 2364 * matches the escape sequence. 2365 */ 2366 private int escape(boolean inclass, boolean create, boolean isrange) { 2367 int ch = skip(); 2368 switch (ch) { 2369 case '0': 2370 return o(); 2371 case '1': 2372 case '2': 2373 case '3': 2374 case '4': 2375 case '5': 2376 case '6': 2377 case '7': 2378 case '8': 2379 case '9': 2380 if (inclass) break; 2381 if (create) { 2382 root = ref((ch - '0')); 2383 } 2384 return -1; 2385 case 'A': 2386 if (inclass) break; 2387 if (create) root = new Begin(); 2388 return -1; 2389 case 'B': 2390 if (inclass) break; 2391 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS)); 2392 return -1; 2393 case 'C': 2394 break; 2395 case 'D': 2396 if (create) { 2397 predicate = has(UNICODE_CHARACTER_CLASS) ? 2398 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2399 predicate = predicate.negate(); 2400 if (!inclass) 2401 root = newCharProperty(predicate); 2402 } 2403 return -1; 2404 case 'E': 2405 case 'F': 2406 break; 2407 case 'G': 2408 if (inclass) break; 2409 if (create) root = new LastMatch(); 2410 return -1; 2411 case 'H': 2412 if (create) { 2413 predicate = HorizWS().negate(); 2414 if (!inclass) 2415 root = newCharProperty(predicate); 2416 } 2417 return -1; 2418 case 'I': 2419 case 'J': 2420 case 'K': 2421 case 'L': 2422 case 'M': 2423 break; 2424 case 'N': 2425 return N(); 2426 case 'O': 2427 case 'P': 2428 case 'Q': 2429 break; 2430 case 'R': 2431 if (inclass) break; 2432 if (create) root = new LineEnding(); 2433 return -1; 2434 case 'S': 2435 if (create) { 2436 predicate = has(UNICODE_CHARACTER_CLASS) ? 2437 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2438 predicate = predicate.negate(); 2439 if (!inclass) 2440 root = newCharProperty(predicate); 2441 } 2442 return -1; 2443 case 'T': 2444 case 'U': 2445 break; 2446 case 'V': 2447 if (create) { 2448 predicate = VertWS().negate(); 2449 if (!inclass) 2450 root = newCharProperty(predicate); 2451 } 2452 return -1; 2453 case 'W': 2454 if (create) { 2455 predicate = has(UNICODE_CHARACTER_CLASS) ? 2456 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2457 predicate = predicate.negate(); 2458 if (!inclass) 2459 root = newCharProperty(predicate); 2460 } 2461 return -1; 2462 case 'X': 2463 if (inclass) break; 2464 if (create) { 2465 root = new XGrapheme(); 2466 } 2467 return -1; 2468 case 'Y': 2469 break; 2470 case 'Z': 2471 if (inclass) break; 2472 if (create) { 2473 if (has(UNIX_LINES)) 2474 root = new UnixDollar(false); 2475 else 2476 root = new Dollar(false); 2477 } 2478 return -1; 2479 case 'a': 2480 return '\007'; 2481 case 'b': 2482 if (inclass) break; 2483 if (create) { 2484 if (peek() == '{') { 2485 if (skip() == 'g') { 2486 if (read() == '}') { 2487 root = new GraphemeBound(); 2488 return -1; 2489 } 2490 break; // error missing trailing } 2491 } 2492 unread(); unread(); 2493 } 2494 root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS)); 2495 } 2496 return -1; 2497 case 'c': 2498 return c(); 2499 case 'd': 2500 if (create) { 2501 predicate = has(UNICODE_CHARACTER_CLASS) ? 2502 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2503 if (!inclass) 2504 root = newCharProperty(predicate); 2505 } 2506 return -1; 2507 case 'e': 2508 return '\033'; 2509 case 'f': 2510 return '\f'; 2511 case 'g': 2512 break; 2513 case 'h': 2514 if (create) { 2515 predicate = HorizWS(); 2516 if (!inclass) 2517 root = newCharProperty(predicate); 2518 } 2519 return -1; 2520 case 'i': 2521 case 'j': 2522 break; 2523 case 'k': 2524 if (inclass) 2525 break; 2526 if (read() != '<') 2527 throw error("\\k is not followed by '<' for named capturing group"); 2528 String name = groupname(read()); 2529 if (!namedGroups().containsKey(name)) 2530 throw error("(named capturing group <"+ name+"> does not exit"); 2531 if (create) { 2532 hasGroupRef = true; 2533 if (has(CASE_INSENSITIVE)) 2534 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE)); 2535 else 2536 root = new BackRef(namedGroups().get(name)); 2537 } 2538 return -1; 2539 case 'l': 2540 case 'm': 2541 break; 2542 case 'n': 2543 return '\n'; 2544 case 'o': 2545 case 'p': 2546 case 'q': 2547 break; 2548 case 'r': 2549 return '\r'; 2550 case 's': 2551 if (create) { 2552 predicate = has(UNICODE_CHARACTER_CLASS) ? 2553 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2554 if (!inclass) 2555 root = newCharProperty(predicate); 2556 } 2557 return -1; 2558 case 't': 2559 return '\t'; 2560 case 'u': 2561 return u(); 2562 case 'v': 2563 // '\v' was implemented as VT/0x0B in releases < 1.8 (though 2564 // undocumented). In JDK8 '\v' is specified as a predefined 2565 // character class for all vertical whitespace characters. 2566 // So [-1, root=VertWS node] pair is returned (instead of a 2567 // single 0x0B). This breaks the range if '\v' is used as 2568 // the start or end value, such as [\v-...] or [...-\v], in 2569 // which a single definite value (0x0B) is expected. For 2570 // compatibility concern '\013'/0x0B is returned if isrange. 2571 if (isrange) 2572 return '\013'; 2573 if (create) { 2574 predicate = VertWS(); 2575 if (!inclass) 2576 root = newCharProperty(predicate); 2577 } 2578 return -1; 2579 case 'w': 2580 if (create) { 2581 predicate = has(UNICODE_CHARACTER_CLASS) ? 2582 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2583 if (!inclass) 2584 root = newCharProperty(predicate); 2585 } 2586 return -1; 2587 case 'x': 2588 return x(); 2589 case 'y': 2590 break; 2591 case 'z': 2592 if (inclass) break; 2593 if (create) root = new End(); 2594 return -1; 2595 default: 2596 return ch; 2597 } 2598 throw error("Illegal/unsupported escape sequence"); 2599 } 2600 2601 /** 2602 * Parse a character class, and return the node that matches it. 2603 * 2604 * Consumes a ] on the way out if consume is true. Usually consume 2605 * is true except for the case of [abc&&def] where def is a separate 2606 * right hand node with "understood" brackets. 2607 */ 2608 private CharPredicate clazz(boolean consume) { 2609 CharPredicate prev = null; 2610 CharPredicate curr = null; 2611 BitClass bits = new BitClass(); 2612 BmpCharPredicate bitsP = ch -> ch < 256 && bits.bits[ch]; 2613 2614 boolean isNeg = false; 2615 boolean hasBits = false; 2616 int ch = next(); 2617 2618 // Negates if first char in a class, otherwise literal 2619 if (ch == '^' && temp[cursor-1] == '[') { 2620 ch = next(); 2621 isNeg = true; 2622 } 2623 for (;;) { 2624 switch (ch) { 2625 case '[': 2626 curr = clazz(true); 2627 if (prev == null) 2628 prev = curr; 2629 else 2630 prev = prev.union(curr); 2631 ch = peek(); 2632 continue; 2633 case '&': 2634 ch = next(); 2635 if (ch == '&') { 2636 ch = next(); 2637 CharPredicate right = null; 2638 while (ch != ']' && ch != '&') { 2639 if (ch == '[') { 2640 if (right == null) 2641 right = clazz(true); 2642 else 2643 right = right.union(clazz(true)); 2644 } else { // abc&&def 2645 unread(); 2646 right = clazz(false); 2647 } 2648 ch = peek(); 2649 } 2650 if (hasBits) { 2651 // bits used, union has high precedence 2652 if (prev == null) { 2653 prev = curr = bitsP; 2654 } else { 2655 prev = prev.union(bitsP); 2656 } 2657 hasBits = false; 2658 } 2659 if (right != null) 2660 curr = right; 2661 if (prev == null) { 2662 if (right == null) 2663 throw error("Bad class syntax"); 2664 else 2665 prev = right; 2666 } else { 2667 prev = prev.and(curr); 2668 } 2669 } else { 2670 // treat as a literal & 2671 unread(); 2672 break; 2673 } 2674 continue; 2675 case 0: 2676 if (cursor >= patternLength) 2677 throw error("Unclosed character class"); 2678 break; 2679 case ']': 2680 if (prev != null || hasBits) { 2681 if (consume) 2682 next(); 2683 if (prev == null) 2684 prev = bitsP; 2685 else if (hasBits) 2686 prev = prev.union(bitsP); 2687 if (isNeg) 2688 return prev.negate(); 2689 return prev; 2690 } 2691 break; 2692 default: 2693 break; 2694 } 2695 curr = range(bits); 2696 if (curr == null) { // the bits used 2697 hasBits = true; 2698 } else { 2699 if (prev == null) 2700 prev = curr; 2701 else if (prev != curr) 2702 prev = prev.union(curr); 2703 } 2704 ch = peek(); 2705 } 2706 } 2707 2708 private CharPredicate bitsOrSingle(BitClass bits, int ch) { 2709 /* Bits can only handle codepoints in [u+0000-u+00ff] range. 2710 Use "single" node instead of bits when dealing with unicode 2711 case folding for codepoints listed below. 2712 (1)Uppercase out of range: u+00ff, u+00b5 2713 toUpperCase(u+00ff) -> u+0178 2714 toUpperCase(u+00b5) -> u+039c 2715 (2)LatinSmallLetterLongS u+17f 2716 toUpperCase(u+017f) -> u+0053 2717 (3)LatinSmallLetterDotlessI u+131 2718 toUpperCase(u+0131) -> u+0049 2719 (4)LatinCapitalLetterIWithDotAbove u+0130 2720 toLowerCase(u+0130) -> u+0069 2721 (5)KelvinSign u+212a 2722 toLowerCase(u+212a) ==> u+006B 2723 (6)AngstromSign u+212b 2724 toLowerCase(u+212b) ==> u+00e5 2725 */ 2726 if (ch < 256 && 2727 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) && 2728 (ch == 0xff || ch == 0xb5 || 2729 ch == 0x49 || ch == 0x69 || //I and i 2730 ch == 0x53 || ch == 0x73 || //S and s 2731 ch == 0x4b || ch == 0x6b || //K and k 2732 ch == 0xc5 || ch == 0xe5))) { //A+ring 2733 bits.add(ch, flags0); 2734 return null; 2735 } 2736 return single(ch); 2737 } 2738 2739 /** 2740 * Returns a suitably optimized, single character predicate 2741 */ 2742 private CharPredicate single(final int ch) { 2743 if (has(CASE_INSENSITIVE)) { 2744 int lower, upper; 2745 if (has(UNICODE_CASE)) { 2746 upper = Character.toUpperCase(ch); 2747 lower = Character.toLowerCase(upper); 2748 // Unicode case insensitive matches 2749 if (upper != lower) 2750 return SingleU(lower); 2751 } else if (ASCII.isAscii(ch)) { 2752 lower = ASCII.toLower(ch); 2753 upper = ASCII.toUpper(ch); 2754 // Case insensitive matches a given BMP character 2755 if (lower != upper) 2756 return SingleI(lower, upper); 2757 } 2758 } 2759 if (isSupplementary(ch)) 2760 return SingleS(ch); 2761 return Single(ch); // Match a given BMP character 2762 } 2763 2764 /** 2765 * Parse a single character or a character range in a character class 2766 * and return its representative node. 2767 */ 2768 private CharPredicate range(BitClass bits) { 2769 int ch = peek(); 2770 if (ch == '\\') { 2771 ch = nextEscaped(); 2772 if (ch == 'p' || ch == 'P') { // A property 2773 boolean comp = (ch == 'P'); 2774 boolean oneLetter = true; 2775 // Consume { if present 2776 ch = next(); 2777 if (ch != '{') 2778 unread(); 2779 else 2780 oneLetter = false; 2781 return family(oneLetter, comp); 2782 } else { // ordinary escape 2783 boolean isrange = temp[cursor+1] == '-'; 2784 unread(); 2785 ch = escape(true, true, isrange); 2786 if (ch == -1) 2787 return predicate; 2788 } 2789 } else { 2790 next(); 2791 } 2792 if (ch >= 0) { 2793 if (peek() == '-') { 2794 int endRange = temp[cursor+1]; 2795 if (endRange == '[') { 2796 return bitsOrSingle(bits, ch); 2797 } 2798 if (endRange != ']') { 2799 next(); 2800 int m = peek(); 2801 if (m == '\\') { 2802 m = escape(true, false, true); 2803 } else { 2804 next(); 2805 } 2806 if (m < ch) { 2807 throw error("Illegal character range"); 2808 } 2809 if (has(CASE_INSENSITIVE)) { 2810 if (has(UNICODE_CASE)) 2811 return CIRangeU(ch, m); 2812 return CIRange(ch, m); 2813 } else { 2814 return Range(ch, m); 2815 } 2816 } 2817 } 2818 return bitsOrSingle(bits, ch); 2819 } 2820 throw error("Unexpected character '"+((char)ch)+"'"); 2821 } 2822 2823 /** 2824 * Parses a Unicode character family and returns its representative node. 2825 */ 2826 private CharPredicate family(boolean singleLetter, boolean isComplement) { 2827 next(); 2828 String name; 2829 CharPredicate p = null; 2830 2831 if (singleLetter) { 2832 int c = temp[cursor]; 2833 if (!Character.isSupplementaryCodePoint(c)) { 2834 name = String.valueOf((char)c); 2835 } else { 2836 name = new String(temp, cursor, 1); 2837 } 2838 read(); 2839 } else { 2840 int i = cursor; 2841 mark('}'); 2842 while(read() != '}') { 2843 } 2844 mark('\000'); 2845 int j = cursor; 2846 if (j > patternLength) 2847 throw error("Unclosed character family"); 2848 if (i + 1 >= j) 2849 throw error("Empty character family"); 2850 name = new String(temp, i, j-i-1); 2851 } 2852 2853 int i = name.indexOf('='); 2854 if (i != -1) { 2855 // property construct \p{name=value} 2856 String value = name.substring(i + 1); 2857 name = name.substring(0, i).toLowerCase(Locale.ENGLISH); 2858 switch (name) { 2859 case "sc": 2860 case "script": 2861 p = CharPredicates.forUnicodeScript(value); 2862 break; 2863 case "blk": 2864 case "block": 2865 p = CharPredicates.forUnicodeBlock(value); 2866 break; 2867 case "gc": 2868 case "general_category": 2869 p = CharPredicates.forProperty(value); 2870 break; 2871 default: 2872 break; 2873 } 2874 if (p == null) 2875 throw error("Unknown Unicode property {name=<" + name + ">, " 2876 + "value=<" + value + ">}"); 2877 2878 } else { 2879 if (name.startsWith("In")) { 2880 // \p{InBlockName} 2881 p = CharPredicates.forUnicodeBlock(name.substring(2)); 2882 } else if (name.startsWith("Is")) { 2883 // \p{IsGeneralCategory} and \p{IsScriptName} 2884 name = name.substring(2); 2885 p = CharPredicates.forUnicodeProperty(name); 2886 if (p == null) 2887 p = CharPredicates.forProperty(name); 2888 if (p == null) 2889 p = CharPredicates.forUnicodeScript(name); 2890 } else { 2891 if (has(UNICODE_CHARACTER_CLASS)) { 2892 p = CharPredicates.forPOSIXName(name); 2893 } 2894 if (p == null) 2895 p = CharPredicates.forProperty(name); 2896 } 2897 if (p == null) 2898 throw error("Unknown character property name {In/Is" + name + "}"); 2899 } 2900 if (isComplement) { 2901 // it might be too expensive to detect if a complement of 2902 // CharProperty can match "certain" supplementary. So just 2903 // go with StartS. 2904 hasSupplementary = true; 2905 p = p.negate(); 2906 } 2907 return p; 2908 } 2909 2910 private CharProperty newCharProperty(CharPredicate p) { 2911 if (p == null) 2912 return null; 2913 if (p instanceof BmpCharPredicate) 2914 return new BmpCharProperty((BmpCharPredicate)p); 2915 else 2916 return new CharProperty(p); 2917 } 2918 2919 /** 2920 * Parses and returns the name of a "named capturing group", the trailing 2921 * ">" is consumed after parsing. 2922 */ 2923 private String groupname(int ch) { 2924 StringBuilder sb = new StringBuilder(); 2925 sb.append(Character.toChars(ch)); 2926 while (ASCII.isLower(ch=read()) || ASCII.isUpper(ch) || 2927 ASCII.isDigit(ch)) { 2928 sb.append(Character.toChars(ch)); 2929 } 2930 if (sb.length() == 0) 2931 throw error("named capturing group has 0 length name"); 2932 if (ch != '>') 2933 throw error("named capturing group is missing trailing '>'"); 2934 return sb.toString(); 2935 } 2936 2937 /** 2938 * Parses a group and returns the head node of a set of nodes that process 2939 * the group. Sometimes a double return system is used where the tail is 2940 * returned in root. 2941 */ 2942 private Node group0() { 2943 boolean capturingGroup = false; 2944 Node head = null; 2945 Node tail = null; 2946 int save = flags0; 2947 int saveTCNCount = topClosureNodes.size(); 2948 root = null; 2949 int ch = next(); 2950 if (ch == '?') { 2951 ch = skip(); 2952 switch (ch) { 2953 case ':': // (?:xxx) pure group 2954 head = createGroup(true); 2955 tail = root; 2956 head.next = expr(tail); 2957 break; 2958 case '=': // (?=xxx) and (?!xxx) lookahead 2959 case '!': 2960 head = createGroup(true); 2961 tail = root; 2962 head.next = expr(tail); 2963 if (ch == '=') { 2964 head = tail = new Pos(head); 2965 } else { 2966 head = tail = new Neg(head); 2967 } 2968 break; 2969 case '>': // (?>xxx) independent group 2970 head = createGroup(true); 2971 tail = root; 2972 head.next = expr(tail); 2973 head = tail = new Ques(head, Qtype.INDEPENDENT); 2974 break; 2975 case '<': // (?<xxx) look behind 2976 ch = read(); 2977 if (ASCII.isLower(ch) || ASCII.isUpper(ch)) { 2978 // named captured group 2979 String name = groupname(ch); 2980 if (namedGroups().containsKey(name)) 2981 throw error("Named capturing group <" + name 2982 + "> is already defined"); 2983 capturingGroup = true; 2984 head = createGroup(false); 2985 tail = root; 2986 namedGroups().put(name, capturingGroupCount-1); 2987 head.next = expr(tail); 2988 break; 2989 } 2990 int start = cursor; 2991 head = createGroup(true); 2992 tail = root; 2993 head.next = expr(tail); 2994 tail.next = lookbehindEnd; 2995 TreeInfo info = new TreeInfo(); 2996 head.study(info); 2997 if (info.maxValid == false) { 2998 throw error("Look-behind group does not have " 2999 + "an obvious maximum length"); 3000 } 3001 boolean hasSupplementary = findSupplementary(start, patternLength); 3002 if (ch == '=') { 3003 head = tail = (hasSupplementary ? 3004 new BehindS(head, info.maxLength, 3005 info.minLength) : 3006 new Behind(head, info.maxLength, 3007 info.minLength)); 3008 } else if (ch == '!') { 3009 head = tail = (hasSupplementary ? 3010 new NotBehindS(head, info.maxLength, 3011 info.minLength) : 3012 new NotBehind(head, info.maxLength, 3013 info.minLength)); 3014 } else { 3015 throw error("Unknown look-behind group"); 3016 } 3017 // clear all top-closure-nodes inside lookbehind 3018 if (saveTCNCount < topClosureNodes.size()) 3019 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3020 break; 3021 case '$': 3022 case '@': 3023 throw error("Unknown group type"); 3024 default: // (?xxx:) inlined match flags 3025 unread(); 3026 addFlag(); 3027 ch = read(); 3028 if (ch == ')') { 3029 return null; // Inline modifier only 3030 } 3031 if (ch != ':') { 3032 throw error("Unknown inline modifier"); 3033 } 3034 head = createGroup(true); 3035 tail = root; 3036 head.next = expr(tail); 3037 break; 3038 } 3039 } else { // (xxx) a regular group 3040 capturingGroup = true; 3041 head = createGroup(false); 3042 tail = root; 3043 head.next = expr(tail); 3044 } 3045 3046 accept(')', "Unclosed group"); 3047 flags0 = save; 3048 3049 // Check for quantifiers 3050 Node node = closure(head); 3051 if (node == head) { // No closure 3052 root = tail; 3053 return node; // Dual return 3054 } 3055 if (head == tail) { // Zero length assertion 3056 root = node; 3057 return node; // Dual return 3058 } 3059 3060 // have group closure, clear all inner closure nodes from the 3061 // top list (no backtracking stopper optimization for inner 3062 if (saveTCNCount < topClosureNodes.size()) 3063 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3064 3065 if (node instanceof Ques) { 3066 Ques ques = (Ques) node; 3067 if (ques.type == Qtype.POSSESSIVE) { 3068 root = node; 3069 return node; 3070 } 3071 tail.next = new BranchConn(); 3072 tail = tail.next; 3073 if (ques.type == Qtype.GREEDY) { 3074 head = new Branch(head, null, tail); 3075 } else { // Reluctant quantifier 3076 head = new Branch(null, head, tail); 3077 } 3078 root = tail; 3079 return head; 3080 } else if (node instanceof Curly) { 3081 Curly curly = (Curly) node; 3082 if (curly.type == Qtype.POSSESSIVE) { 3083 root = node; 3084 return node; 3085 } 3086 // Discover if the group is deterministic 3087 TreeInfo info = new TreeInfo(); 3088 if (head.study(info)) { // Deterministic 3089 GroupTail temp = (GroupTail) tail; 3090 head = root = new GroupCurly(head.next, curly.cmin, 3091 curly.cmax, curly.type, 3092 ((GroupTail)tail).localIndex, 3093 ((GroupTail)tail).groupIndex, 3094 capturingGroup); 3095 return head; 3096 } else { // Non-deterministic 3097 int temp = ((GroupHead) head).localIndex; 3098 Loop loop; 3099 if (curly.type == Qtype.GREEDY) { 3100 loop = new Loop(this.localCount, temp); 3101 // add the max_reps greedy to the top-closure-node list 3102 if (curly.cmax == MAX_REPS) 3103 topClosureNodes.add(loop); 3104 } else { // Reluctant Curly 3105 loop = new LazyLoop(this.localCount, temp); 3106 } 3107 Prolog prolog = new Prolog(loop); 3108 this.localCount += 1; 3109 loop.cmin = curly.cmin; 3110 loop.cmax = curly.cmax; 3111 loop.body = head; 3112 tail.next = loop; 3113 root = loop; 3114 return prolog; // Dual return 3115 } 3116 } 3117 throw error("Internal logic error"); 3118 } 3119 3120 /** 3121 * Create group head and tail nodes using double return. If the group is 3122 * created with anonymous true then it is a pure group and should not 3123 * affect group counting. 3124 */ 3125 private Node createGroup(boolean anonymous) { 3126 int localIndex = localCount++; 3127 int groupIndex = 0; 3128 if (!anonymous) 3129 groupIndex = capturingGroupCount++; 3130 GroupHead head = new GroupHead(localIndex); 3131 root = new GroupTail(localIndex, groupIndex); 3132 3133 // for debug/print only, head.match does NOT need the "tail" info 3134 head.tail = (GroupTail)root; 3135 3136 if (!anonymous && groupIndex < 10) 3137 groupNodes[groupIndex] = head; 3138 return head; 3139 } 3140 3141 @SuppressWarnings("fallthrough") 3142 /** 3143 * Parses inlined match flags and set them appropriately. 3144 */ 3145 private void addFlag() { 3146 int ch = peek(); 3147 for (;;) { 3148 switch (ch) { 3149 case 'i': 3150 flags0 |= CASE_INSENSITIVE; 3151 break; 3152 case 'm': 3153 flags0 |= MULTILINE; 3154 break; 3155 case 's': 3156 flags0 |= DOTALL; 3157 break; 3158 case 'd': 3159 flags0 |= UNIX_LINES; 3160 break; 3161 case 'u': 3162 flags0 |= UNICODE_CASE; 3163 break; 3164 case 'c': 3165 flags0 |= CANON_EQ; 3166 break; 3167 case 'x': 3168 flags0 |= COMMENTS; 3169 break; 3170 case 'U': 3171 flags0 |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3172 break; 3173 case '-': // subFlag then fall through 3174 ch = next(); 3175 subFlag(); 3176 default: 3177 return; 3178 } 3179 ch = next(); 3180 } 3181 } 3182 3183 @SuppressWarnings("fallthrough") 3184 /** 3185 * Parses the second part of inlined match flags and turns off 3186 * flags appropriately. 3187 */ 3188 private void subFlag() { 3189 int ch = peek(); 3190 for (;;) { 3191 switch (ch) { 3192 case 'i': 3193 flags0 &= ~CASE_INSENSITIVE; 3194 break; 3195 case 'm': 3196 flags0 &= ~MULTILINE; 3197 break; 3198 case 's': 3199 flags0 &= ~DOTALL; 3200 break; 3201 case 'd': 3202 flags0 &= ~UNIX_LINES; 3203 break; 3204 case 'u': 3205 flags0 &= ~UNICODE_CASE; 3206 break; 3207 case 'c': 3208 flags0 &= ~CANON_EQ; 3209 break; 3210 case 'x': 3211 flags0 &= ~COMMENTS; 3212 break; 3213 case 'U': 3214 flags0 &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3215 break; 3216 default: 3217 return; 3218 } 3219 ch = next(); 3220 } 3221 } 3222 3223 static final int MAX_REPS = 0x7FFFFFFF; 3224 3225 static enum Qtype { 3226 GREEDY, LAZY, POSSESSIVE, INDEPENDENT 3227 } 3228 3229 private Node curly(Node prev, int cmin) { 3230 int ch = next(); 3231 if (ch == '?') { 3232 next(); 3233 return new Curly(prev, cmin, MAX_REPS, Qtype.LAZY); 3234 } else if (ch == '+') { 3235 next(); 3236 return new Curly(prev, cmin, MAX_REPS, Qtype.POSSESSIVE); 3237 } 3238 if (prev instanceof BmpCharProperty) { 3239 return new BmpCharPropertyGreedy((BmpCharProperty)prev, cmin); 3240 } else if (prev instanceof CharProperty) { 3241 return new CharPropertyGreedy((CharProperty)prev, cmin); 3242 } 3243 return new Curly(prev, cmin, MAX_REPS, Qtype.GREEDY); 3244 } 3245 3246 /** 3247 * Processes repetition. If the next character peeked is a quantifier 3248 * then new nodes must be appended to handle the repetition. 3249 * Prev could be a single or a group, so it could be a chain of nodes. 3250 */ 3251 private Node closure(Node prev) { 3252 Node atom; 3253 int ch = peek(); 3254 switch (ch) { 3255 case '?': 3256 ch = next(); 3257 if (ch == '?') { 3258 next(); 3259 return new Ques(prev, Qtype.LAZY); 3260 } else if (ch == '+') { 3261 next(); 3262 return new Ques(prev, Qtype.POSSESSIVE); 3263 } 3264 return new Ques(prev, Qtype.GREEDY); 3265 case '*': 3266 return curly(prev, 0); 3267 case '+': 3268 return curly(prev, 1); 3269 case '{': 3270 ch = temp[cursor+1]; 3271 if (ASCII.isDigit(ch)) { 3272 skip(); 3273 int cmin = 0; 3274 do { 3275 cmin = cmin * 10 + (ch - '0'); 3276 } while (ASCII.isDigit(ch = read())); 3277 int cmax = cmin; 3278 if (ch == ',') { 3279 ch = read(); 3280 cmax = MAX_REPS; 3281 if (ch != '}') { 3282 cmax = 0; 3283 while (ASCII.isDigit(ch)) { 3284 cmax = cmax * 10 + (ch - '0'); 3285 ch = read(); 3286 } 3287 } 3288 } 3289 if (ch != '}') 3290 throw error("Unclosed counted closure"); 3291 if (((cmin) | (cmax) | (cmax - cmin)) < 0) 3292 throw error("Illegal repetition range"); 3293 Curly curly; 3294 ch = peek(); 3295 if (ch == '?') { 3296 next(); 3297 curly = new Curly(prev, cmin, cmax, Qtype.LAZY); 3298 } else if (ch == '+') { 3299 next(); 3300 curly = new Curly(prev, cmin, cmax, Qtype.POSSESSIVE); 3301 } else { 3302 curly = new Curly(prev, cmin, cmax, Qtype.GREEDY); 3303 } 3304 return curly; 3305 } else { 3306 throw error("Illegal repetition"); 3307 } 3308 default: 3309 return prev; 3310 } 3311 } 3312 3313 /** 3314 * Utility method for parsing control escape sequences. 3315 */ 3316 private int c() { 3317 if (cursor < patternLength) { 3318 return read() ^ 64; 3319 } 3320 throw error("Illegal control escape sequence"); 3321 } 3322 3323 /** 3324 * Utility method for parsing octal escape sequences. 3325 */ 3326 private int o() { 3327 int n = read(); 3328 if (((n-'0')|('7'-n)) >= 0) { 3329 int m = read(); 3330 if (((m-'0')|('7'-m)) >= 0) { 3331 int o = read(); 3332 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) { 3333 return (n - '0') * 64 + (m - '0') * 8 + (o - '0'); 3334 } 3335 unread(); 3336 return (n - '0') * 8 + (m - '0'); 3337 } 3338 unread(); 3339 return (n - '0'); 3340 } 3341 throw error("Illegal octal escape sequence"); 3342 } 3343 3344 /** 3345 * Utility method for parsing hexadecimal escape sequences. 3346 */ 3347 private int x() { 3348 int n = read(); 3349 if (ASCII.isHexDigit(n)) { 3350 int m = read(); 3351 if (ASCII.isHexDigit(m)) { 3352 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m); 3353 } 3354 } else if (n == '{' && ASCII.isHexDigit(peek())) { 3355 int ch = 0; 3356 while (ASCII.isHexDigit(n = read())) { 3357 ch = (ch << 4) + ASCII.toDigit(n); 3358 if (ch > Character.MAX_CODE_POINT) 3359 throw error("Hexadecimal codepoint is too big"); 3360 } 3361 if (n != '}') 3362 throw error("Unclosed hexadecimal escape sequence"); 3363 return ch; 3364 } 3365 throw error("Illegal hexadecimal escape sequence"); 3366 } 3367 3368 /** 3369 * Utility method for parsing unicode escape sequences. 3370 */ 3371 private int cursor() { 3372 return cursor; 3373 } 3374 3375 private void setcursor(int pos) { 3376 cursor = pos; 3377 } 3378 3379 private int uxxxx() { 3380 int n = 0; 3381 for (int i = 0; i < 4; i++) { 3382 int ch = read(); 3383 if (!ASCII.isHexDigit(ch)) { 3384 throw error("Illegal Unicode escape sequence"); 3385 } 3386 n = n * 16 + ASCII.toDigit(ch); 3387 } 3388 return n; 3389 } 3390 3391 private int u() { 3392 int n = uxxxx(); 3393 if (Character.isHighSurrogate((char)n)) { 3394 int cur = cursor(); 3395 if (read() == '\\' && read() == 'u') { 3396 int n2 = uxxxx(); 3397 if (Character.isLowSurrogate((char)n2)) 3398 return Character.toCodePoint((char)n, (char)n2); 3399 } 3400 setcursor(cur); 3401 } 3402 return n; 3403 } 3404 3405 private int N() { 3406 if (read() == '{') { 3407 int i = cursor; 3408 while (cursor < patternLength && read() != '}') {} 3409 if (cursor > patternLength) 3410 throw error("Unclosed character name escape sequence"); 3411 String name = new String(temp, i, cursor - i - 1); 3412 try { 3413 return Character.codePointOf(name); 3414 } catch (IllegalArgumentException x) { 3415 throw error("Unknown character name [" + name + "]"); 3416 } 3417 } 3418 throw error("Illegal character name escape sequence"); 3419 } 3420 3421 // 3422 // Utility methods for code point support 3423 // 3424 private static final int countChars(CharSequence seq, int index, 3425 int lengthInCodePoints) { 3426 // optimization 3427 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) { 3428 assert (index >= 0 && index < seq.length()); 3429 return 1; 3430 } 3431 int length = seq.length(); 3432 int x = index; 3433 if (lengthInCodePoints >= 0) { 3434 assert (index >= 0 && index < length); 3435 for (int i = 0; x < length && i < lengthInCodePoints; i++) { 3436 if (Character.isHighSurrogate(seq.charAt(x++))) { 3437 if (x < length && Character.isLowSurrogate(seq.charAt(x))) { 3438 x++; 3439 } 3440 } 3441 } 3442 return x - index; 3443 } 3444 3445 assert (index >= 0 && index <= length); 3446 if (index == 0) { 3447 return 0; 3448 } 3449 int len = -lengthInCodePoints; 3450 for (int i = 0; x > 0 && i < len; i++) { 3451 if (Character.isLowSurrogate(seq.charAt(--x))) { 3452 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) { 3453 x--; 3454 } 3455 } 3456 } 3457 return index - x; 3458 } 3459 3460 private static final int countCodePoints(CharSequence seq) { 3461 int length = seq.length(); 3462 int n = 0; 3463 for (int i = 0; i < length; ) { 3464 n++; 3465 if (Character.isHighSurrogate(seq.charAt(i++))) { 3466 if (i < length && Character.isLowSurrogate(seq.charAt(i))) { 3467 i++; 3468 } 3469 } 3470 } 3471 return n; 3472 } 3473 3474 /** 3475 * Creates a bit vector for matching Latin-1 values. A normal BitClass 3476 * never matches values above Latin-1, and a complemented BitClass always 3477 * matches values above Latin-1. 3478 */ 3479 static final class BitClass extends BmpCharProperty { 3480 final boolean[] bits; 3481 BitClass() { 3482 this(new boolean[256]); 3483 } 3484 private BitClass(boolean[] bits) { 3485 super( ch -> ch < 256 && bits[ch]); 3486 this.bits = bits; 3487 } 3488 BitClass add(int c, int flags) { 3489 assert c >= 0 && c <= 255; 3490 if ((flags & CASE_INSENSITIVE) != 0) { 3491 if (ASCII.isAscii(c)) { 3492 bits[ASCII.toUpper(c)] = true; 3493 bits[ASCII.toLower(c)] = true; 3494 } else if ((flags & UNICODE_CASE) != 0) { 3495 bits[Character.toLowerCase(c)] = true; 3496 bits[Character.toUpperCase(c)] = true; 3497 } 3498 } 3499 bits[c] = true; 3500 return this; 3501 } 3502 } 3503 3504 /** 3505 * Utility method for creating a string slice matcher. 3506 */ 3507 private Node newSlice(int[] buf, int count, boolean hasSupplementary) { 3508 int[] tmp = new int[count]; 3509 if (has(CASE_INSENSITIVE)) { 3510 if (has(UNICODE_CASE)) { 3511 for (int i = 0; i < count; i++) { 3512 tmp[i] = Character.toLowerCase( 3513 Character.toUpperCase(buf[i])); 3514 } 3515 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp); 3516 } 3517 for (int i = 0; i < count; i++) { 3518 tmp[i] = ASCII.toLower(buf[i]); 3519 } 3520 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp); 3521 } 3522 for (int i = 0; i < count; i++) { 3523 tmp[i] = buf[i]; 3524 } 3525 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp); 3526 } 3527 3528 /** 3529 * The following classes are the building components of the object 3530 * tree that represents a compiled regular expression. The object tree 3531 * is made of individual elements that handle constructs in the Pattern. 3532 * Each type of object knows how to match its equivalent construct with 3533 * the match() method. 3534 */ 3535 3536 /** 3537 * Base class for all node classes. Subclasses should override the match() 3538 * method as appropriate. This class is an accepting node, so its match() 3539 * always returns true. 3540 */ 3541 static class Node extends Object { 3542 Node next; 3543 Node() { 3544 next = Pattern.accept; 3545 } 3546 /** 3547 * This method implements the classic accept node. 3548 */ 3549 boolean match(Matcher matcher, int i, CharSequence seq) { 3550 matcher.last = i; 3551 matcher.groups[0] = matcher.first; 3552 matcher.groups[1] = matcher.last; 3553 return true; 3554 } 3555 /** 3556 * This method is good for all zero length assertions. 3557 */ 3558 boolean study(TreeInfo info) { 3559 if (next != null) { 3560 return next.study(info); 3561 } else { 3562 return info.deterministic; 3563 } 3564 } 3565 } 3566 3567 static class LastNode extends Node { 3568 /** 3569 * This method implements the classic accept node with 3570 * the addition of a check to see if the match occurred 3571 * using all of the input. 3572 */ 3573 boolean match(Matcher matcher, int i, CharSequence seq) { 3574 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to) 3575 return false; 3576 matcher.last = i; 3577 matcher.groups[0] = matcher.first; 3578 matcher.groups[1] = matcher.last; 3579 return true; 3580 } 3581 } 3582 3583 /** 3584 * Used for REs that can start anywhere within the input string. 3585 * This basically tries to match repeatedly at each spot in the 3586 * input string, moving forward after each try. An anchored search 3587 * or a BnM will bypass this node completely. 3588 */ 3589 static class Start extends Node { 3590 int minLength; 3591 Start(Node node) { 3592 this.next = node; 3593 TreeInfo info = new TreeInfo(); 3594 next.study(info); 3595 minLength = info.minLength; 3596 } 3597 boolean match(Matcher matcher, int i, CharSequence seq) { 3598 if (i > matcher.to - minLength) { 3599 matcher.hitEnd = true; 3600 return false; 3601 } 3602 int guard = matcher.to - minLength; 3603 for (; i <= guard; i++) { 3604 if (next.match(matcher, i, seq)) { 3605 matcher.first = i; 3606 matcher.groups[0] = matcher.first; 3607 matcher.groups[1] = matcher.last; 3608 return true; 3609 } 3610 } 3611 matcher.hitEnd = true; 3612 return false; 3613 } 3614 boolean study(TreeInfo info) { 3615 next.study(info); 3616 info.maxValid = false; 3617 info.deterministic = false; 3618 return false; 3619 } 3620 } 3621 3622 /* 3623 * StartS supports supplementary characters, including unpaired surrogates. 3624 */ 3625 static final class StartS extends Start { 3626 StartS(Node node) { 3627 super(node); 3628 } 3629 boolean match(Matcher matcher, int i, CharSequence seq) { 3630 if (i > matcher.to - minLength) { 3631 matcher.hitEnd = true; 3632 return false; 3633 } 3634 int guard = matcher.to - minLength; 3635 while (i <= guard) { 3636 //if ((ret = next.match(matcher, i, seq)) || i == guard) 3637 if (next.match(matcher, i, seq)) { 3638 matcher.first = i; 3639 matcher.groups[0] = matcher.first; 3640 matcher.groups[1] = matcher.last; 3641 return true; 3642 } 3643 if (i == guard) 3644 break; 3645 // Optimization to move to the next character. This is 3646 // faster than countChars(seq, i, 1). 3647 if (Character.isHighSurrogate(seq.charAt(i++))) { 3648 if (i < seq.length() && 3649 Character.isLowSurrogate(seq.charAt(i))) { 3650 i++; 3651 } 3652 } 3653 } 3654 matcher.hitEnd = true; 3655 return false; 3656 } 3657 } 3658 3659 /** 3660 * Node to anchor at the beginning of input. This object implements the 3661 * match for a \A sequence, and the caret anchor will use this if not in 3662 * multiline mode. 3663 */ 3664 static final class Begin extends Node { 3665 boolean match(Matcher matcher, int i, CharSequence seq) { 3666 int fromIndex = (matcher.anchoringBounds) ? 3667 matcher.from : 0; 3668 if (i == fromIndex && next.match(matcher, i, seq)) { 3669 matcher.first = i; 3670 matcher.groups[0] = i; 3671 matcher.groups[1] = matcher.last; 3672 return true; 3673 } else { 3674 return false; 3675 } 3676 } 3677 } 3678 3679 /** 3680 * Node to anchor at the end of input. This is the absolute end, so this 3681 * should not match at the last newline before the end as $ will. 3682 */ 3683 static final class End extends Node { 3684 boolean match(Matcher matcher, int i, CharSequence seq) { 3685 int endIndex = (matcher.anchoringBounds) ? 3686 matcher.to : matcher.getTextLength(); 3687 if (i == endIndex) { 3688 matcher.hitEnd = true; 3689 return next.match(matcher, i, seq); 3690 } 3691 return false; 3692 } 3693 } 3694 3695 /** 3696 * Node to anchor at the beginning of a line. This is essentially the 3697 * object to match for the multiline ^. 3698 */ 3699 static final class Caret extends Node { 3700 boolean match(Matcher matcher, int i, CharSequence seq) { 3701 int startIndex = matcher.from; 3702 int endIndex = matcher.to; 3703 if (!matcher.anchoringBounds) { 3704 startIndex = 0; 3705 endIndex = matcher.getTextLength(); 3706 } 3707 // Perl does not match ^ at end of input even after newline 3708 if (i == endIndex) { 3709 matcher.hitEnd = true; 3710 return false; 3711 } 3712 if (i > startIndex) { 3713 char ch = seq.charAt(i-1); 3714 if (ch != '\n' && ch != '\r' 3715 && (ch|1) != '\u2029' 3716 && ch != '\u0085' ) { 3717 return false; 3718 } 3719 // Should treat /r/n as one newline 3720 if (ch == '\r' && seq.charAt(i) == '\n') 3721 return false; 3722 } 3723 return next.match(matcher, i, seq); 3724 } 3725 } 3726 3727 /** 3728 * Node to anchor at the beginning of a line when in unixdot mode. 3729 */ 3730 static final class UnixCaret extends Node { 3731 boolean match(Matcher matcher, int i, CharSequence seq) { 3732 int startIndex = matcher.from; 3733 int endIndex = matcher.to; 3734 if (!matcher.anchoringBounds) { 3735 startIndex = 0; 3736 endIndex = matcher.getTextLength(); 3737 } 3738 // Perl does not match ^ at end of input even after newline 3739 if (i == endIndex) { 3740 matcher.hitEnd = true; 3741 return false; 3742 } 3743 if (i > startIndex) { 3744 char ch = seq.charAt(i-1); 3745 if (ch != '\n') { 3746 return false; 3747 } 3748 } 3749 return next.match(matcher, i, seq); 3750 } 3751 } 3752 3753 /** 3754 * Node to match the location where the last match ended. 3755 * This is used for the \G construct. 3756 */ 3757 static final class LastMatch extends Node { 3758 boolean match(Matcher matcher, int i, CharSequence seq) { 3759 if (i != matcher.oldLast) 3760 return false; 3761 return next.match(matcher, i, seq); 3762 } 3763 } 3764 3765 /** 3766 * Node to anchor at the end of a line or the end of input based on the 3767 * multiline mode. 3768 * 3769 * When not in multiline mode, the $ can only match at the very end 3770 * of the input, unless the input ends in a line terminator in which 3771 * it matches right before the last line terminator. 3772 * 3773 * Note that \r\n is considered an atomic line terminator. 3774 * 3775 * Like ^ the $ operator matches at a position, it does not match the 3776 * line terminators themselves. 3777 */ 3778 static final class Dollar extends Node { 3779 boolean multiline; 3780 Dollar(boolean mul) { 3781 multiline = mul; 3782 } 3783 boolean match(Matcher matcher, int i, CharSequence seq) { 3784 int endIndex = (matcher.anchoringBounds) ? 3785 matcher.to : matcher.getTextLength(); 3786 if (!multiline) { 3787 if (i < endIndex - 2) 3788 return false; 3789 if (i == endIndex - 2) { 3790 char ch = seq.charAt(i); 3791 if (ch != '\r') 3792 return false; 3793 ch = seq.charAt(i + 1); 3794 if (ch != '\n') 3795 return false; 3796 } 3797 } 3798 // Matches before any line terminator; also matches at the 3799 // end of input 3800 // Before line terminator: 3801 // If multiline, we match here no matter what 3802 // If not multiline, fall through so that the end 3803 // is marked as hit; this must be a /r/n or a /n 3804 // at the very end so the end was hit; more input 3805 // could make this not match here 3806 if (i < endIndex) { 3807 char ch = seq.charAt(i); 3808 if (ch == '\n') { 3809 // No match between \r\n 3810 if (i > 0 && seq.charAt(i-1) == '\r') 3811 return false; 3812 if (multiline) 3813 return next.match(matcher, i, seq); 3814 } else if (ch == '\r' || ch == '\u0085' || 3815 (ch|1) == '\u2029') { 3816 if (multiline) 3817 return next.match(matcher, i, seq); 3818 } else { // No line terminator, no match 3819 return false; 3820 } 3821 } 3822 // Matched at current end so hit end 3823 matcher.hitEnd = true; 3824 // If a $ matches because of end of input, then more input 3825 // could cause it to fail! 3826 matcher.requireEnd = true; 3827 return next.match(matcher, i, seq); 3828 } 3829 boolean study(TreeInfo info) { 3830 next.study(info); 3831 return info.deterministic; 3832 } 3833 } 3834 3835 /** 3836 * Node to anchor at the end of a line or the end of input based on the 3837 * multiline mode when in unix lines mode. 3838 */ 3839 static final class UnixDollar extends Node { 3840 boolean multiline; 3841 UnixDollar(boolean mul) { 3842 multiline = mul; 3843 } 3844 boolean match(Matcher matcher, int i, CharSequence seq) { 3845 int endIndex = (matcher.anchoringBounds) ? 3846 matcher.to : matcher.getTextLength(); 3847 if (i < endIndex) { 3848 char ch = seq.charAt(i); 3849 if (ch == '\n') { 3850 // If not multiline, then only possible to 3851 // match at very end or one before end 3852 if (multiline == false && i != endIndex - 1) 3853 return false; 3854 // If multiline return next.match without setting 3855 // matcher.hitEnd 3856 if (multiline) 3857 return next.match(matcher, i, seq); 3858 } else { 3859 return false; 3860 } 3861 } 3862 // Matching because at the end or 1 before the end; 3863 // more input could change this so set hitEnd 3864 matcher.hitEnd = true; 3865 // If a $ matches because of end of input, then more input 3866 // could cause it to fail! 3867 matcher.requireEnd = true; 3868 return next.match(matcher, i, seq); 3869 } 3870 boolean study(TreeInfo info) { 3871 next.study(info); 3872 return info.deterministic; 3873 } 3874 } 3875 3876 /** 3877 * Node class that matches a Unicode line ending '\R' 3878 */ 3879 static final class LineEnding extends Node { 3880 boolean match(Matcher matcher, int i, CharSequence seq) { 3881 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029]) 3882 if (i < matcher.to) { 3883 int ch = seq.charAt(i); 3884 if (ch == 0x0A || ch == 0x0B || ch == 0x0C || 3885 ch == 0x85 || ch == 0x2028 || ch == 0x2029) 3886 return next.match(matcher, i + 1, seq); 3887 if (ch == 0x0D) { 3888 i++; 3889 if (i < matcher.to) { 3890 if (seq.charAt(i) == 0x0A && 3891 next.match(matcher, i + 1, seq)) { 3892 return true; 3893 } 3894 } else { 3895 matcher.hitEnd = true; 3896 } 3897 return next.match(matcher, i, seq); 3898 } 3899 } else { 3900 matcher.hitEnd = true; 3901 } 3902 return false; 3903 } 3904 boolean study(TreeInfo info) { 3905 info.minLength++; 3906 info.maxLength += 2; 3907 return next.study(info); 3908 } 3909 } 3910 3911 /** 3912 * Abstract node class to match one character satisfying some 3913 * boolean property. 3914 */ 3915 static class CharProperty extends Node { 3916 CharPredicate predicate; 3917 3918 CharProperty (CharPredicate predicate) { 3919 this.predicate = predicate; 3920 } 3921 boolean match(Matcher matcher, int i, CharSequence seq) { 3922 if (i < matcher.to) { 3923 int ch = Character.codePointAt(seq, i); 3924 return predicate.is(ch) && 3925 next.match(matcher, i + Character.charCount(ch), seq); 3926 } else { 3927 matcher.hitEnd = true; 3928 return false; 3929 } 3930 } 3931 boolean study(TreeInfo info) { 3932 info.minLength++; 3933 info.maxLength++; 3934 return next.study(info); 3935 } 3936 } 3937 3938 /** 3939 * Optimized version of CharProperty that works only for 3940 * properties never satisfied by Supplementary characters. 3941 */ 3942 private static class BmpCharProperty extends CharProperty { 3943 BmpCharProperty (BmpCharPredicate predicate) { 3944 super(predicate); 3945 } 3946 boolean match(Matcher matcher, int i, CharSequence seq) { 3947 if (i < matcher.to) { 3948 return predicate.is(seq.charAt(i)) && 3949 next.match(matcher, i + 1, seq); 3950 } else { 3951 matcher.hitEnd = true; 3952 return false; 3953 } 3954 } 3955 } 3956 3957 private static class NFCCharProperty extends Node { 3958 CharPredicate predicate; 3959 NFCCharProperty (CharPredicate predicate) { 3960 this.predicate = predicate; 3961 } 3962 3963 boolean match(Matcher matcher, int i, CharSequence seq) { 3964 if (i < matcher.to) { 3965 int ch0 = Character.codePointAt(seq, i); 3966 int n = Character.charCount(ch0); 3967 int j = i + n; 3968 while (j < matcher.to) { 3969 int ch1 = Character.codePointAt(seq, j); 3970 if (Grapheme.isBoundary(ch0, ch1)) 3971 break; 3972 ch0 = ch1; 3973 j += Character.charCount(ch1); 3974 } 3975 if (i + n == j) { // single, assume nfc cp 3976 if (predicate.is(ch0)) 3977 return next.match(matcher, j, seq); 3978 } else { 3979 while (i + n < j) { 3980 String nfc = Normalizer.normalize( 3981 seq.toString().substring(i, j), Normalizer.Form.NFC); 3982 if (nfc.codePointCount(0, nfc.length()) == 1) { 3983 if (predicate.is(nfc.codePointAt(0)) && 3984 next.match(matcher, j, seq)) { 3985 return true; 3986 } 3987 } 3988 3989 ch0 = Character.codePointBefore(seq, j); 3990 j -= Character.charCount(ch0); 3991 } 3992 } 3993 if (j < matcher.to) 3994 return false; 3995 } 3996 matcher.hitEnd = true; 3997 return false; 3998 } 3999 4000 boolean study(TreeInfo info) { 4001 info.minLength++; 4002 info.deterministic = false; 4003 return next.study(info); 4004 } 4005 } 4006 4007 /** 4008 * Node class that matches an unicode extended grapheme cluster 4009 */ 4010 static class XGrapheme extends Node { 4011 boolean match(Matcher matcher, int i, CharSequence seq) { 4012 if (i < matcher.to) { 4013 int ch0 = Character.codePointAt(seq, i); 4014 i += Character.charCount(ch0); 4015 while (i < matcher.to) { 4016 int ch1 = Character.codePointAt(seq, i); 4017 if (Grapheme.isBoundary(ch0, ch1)) 4018 break; 4019 ch0 = ch1; 4020 i += Character.charCount(ch1); 4021 } 4022 return next.match(matcher, i, seq); 4023 } 4024 matcher.hitEnd = true; 4025 return false; 4026 } 4027 4028 boolean study(TreeInfo info) { 4029 info.minLength++; 4030 info.deterministic = false; 4031 return next.study(info); 4032 } 4033 } 4034 4035 /** 4036 * Node class that handles grapheme boundaries 4037 */ 4038 static class GraphemeBound extends Node { 4039 boolean match(Matcher matcher, int i, CharSequence seq) { 4040 int startIndex = matcher.from; 4041 int endIndex = matcher.to; 4042 if (matcher.transparentBounds) { 4043 startIndex = 0; 4044 endIndex = matcher.getTextLength(); 4045 } 4046 if (i == startIndex) { 4047 return next.match(matcher, i, seq); 4048 } 4049 if (i < endIndex) { 4050 if (Character.isSurrogatePair(seq.charAt(i-1), seq.charAt(i)) || 4051 !Grapheme.isBoundary(Character.codePointBefore(seq, i), 4052 Character.codePointAt(seq, i))) { 4053 return false; 4054 } 4055 } else { 4056 matcher.hitEnd = true; 4057 matcher.requireEnd = true; 4058 } 4059 return next.match(matcher, i, seq); 4060 } 4061 } 4062 4063 /** 4064 * Base class for all Slice nodes 4065 */ 4066 static class SliceNode extends Node { 4067 int[] buffer; 4068 SliceNode(int[] buf) { 4069 buffer = buf; 4070 } 4071 boolean study(TreeInfo info) { 4072 info.minLength += buffer.length; 4073 info.maxLength += buffer.length; 4074 return next.study(info); 4075 } 4076 } 4077 4078 /** 4079 * Node class for a case sensitive/BMP-only sequence of literal 4080 * characters. 4081 */ 4082 static class Slice extends SliceNode { 4083 Slice(int[] buf) { 4084 super(buf); 4085 } 4086 boolean match(Matcher matcher, int i, CharSequence seq) { 4087 int[] buf = buffer; 4088 int len = buf.length; 4089 for (int j=0; j<len; j++) { 4090 if ((i+j) >= matcher.to) { 4091 matcher.hitEnd = true; 4092 return false; 4093 } 4094 if (buf[j] != seq.charAt(i+j)) 4095 return false; 4096 } 4097 return next.match(matcher, i+len, seq); 4098 } 4099 } 4100 4101 /** 4102 * Node class for a case_insensitive/BMP-only sequence of literal 4103 * characters. 4104 */ 4105 static class SliceI extends SliceNode { 4106 SliceI(int[] buf) { 4107 super(buf); 4108 } 4109 boolean match(Matcher matcher, int i, CharSequence seq) { 4110 int[] buf = buffer; 4111 int len = buf.length; 4112 for (int j=0; j<len; j++) { 4113 if ((i+j) >= matcher.to) { 4114 matcher.hitEnd = true; 4115 return false; 4116 } 4117 int c = seq.charAt(i+j); 4118 if (buf[j] != c && 4119 buf[j] != ASCII.toLower(c)) 4120 return false; 4121 } 4122 return next.match(matcher, i+len, seq); 4123 } 4124 } 4125 4126 /** 4127 * Node class for a unicode_case_insensitive/BMP-only sequence of 4128 * literal characters. Uses unicode case folding. 4129 */ 4130 static final class SliceU extends SliceNode { 4131 SliceU(int[] buf) { 4132 super(buf); 4133 } 4134 boolean match(Matcher matcher, int i, CharSequence seq) { 4135 int[] buf = buffer; 4136 int len = buf.length; 4137 for (int j=0; j<len; j++) { 4138 if ((i+j) >= matcher.to) { 4139 matcher.hitEnd = true; 4140 return false; 4141 } 4142 int c = seq.charAt(i+j); 4143 if (buf[j] != c && 4144 buf[j] != Character.toLowerCase(Character.toUpperCase(c))) 4145 return false; 4146 } 4147 return next.match(matcher, i+len, seq); 4148 } 4149 } 4150 4151 /** 4152 * Node class for a case sensitive sequence of literal characters 4153 * including supplementary characters. 4154 */ 4155 static final class SliceS extends Slice { 4156 SliceS(int[] buf) { 4157 super(buf); 4158 } 4159 boolean match(Matcher matcher, int i, CharSequence seq) { 4160 int[] buf = buffer; 4161 int x = i; 4162 for (int j = 0; j < buf.length; j++) { 4163 if (x >= matcher.to) { 4164 matcher.hitEnd = true; 4165 return false; 4166 } 4167 int c = Character.codePointAt(seq, x); 4168 if (buf[j] != c) 4169 return false; 4170 x += Character.charCount(c); 4171 if (x > matcher.to) { 4172 matcher.hitEnd = true; 4173 return false; 4174 } 4175 } 4176 return next.match(matcher, x, seq); 4177 } 4178 } 4179 4180 /** 4181 * Node class for a case insensitive sequence of literal characters 4182 * including supplementary characters. 4183 */ 4184 static class SliceIS extends SliceNode { 4185 SliceIS(int[] buf) { 4186 super(buf); 4187 } 4188 int toLower(int c) { 4189 return ASCII.toLower(c); 4190 } 4191 boolean match(Matcher matcher, int i, CharSequence seq) { 4192 int[] buf = buffer; 4193 int x = i; 4194 for (int j = 0; j < buf.length; j++) { 4195 if (x >= matcher.to) { 4196 matcher.hitEnd = true; 4197 return false; 4198 } 4199 int c = Character.codePointAt(seq, x); 4200 if (buf[j] != c && buf[j] != toLower(c)) 4201 return false; 4202 x += Character.charCount(c); 4203 if (x > matcher.to) { 4204 matcher.hitEnd = true; 4205 return false; 4206 } 4207 } 4208 return next.match(matcher, x, seq); 4209 } 4210 } 4211 4212 /** 4213 * Node class for a case insensitive sequence of literal characters. 4214 * Uses unicode case folding. 4215 */ 4216 static final class SliceUS extends SliceIS { 4217 SliceUS(int[] buf) { 4218 super(buf); 4219 } 4220 int toLower(int c) { 4221 return Character.toLowerCase(Character.toUpperCase(c)); 4222 } 4223 } 4224 4225 /** 4226 * The 0 or 1 quantifier. This one class implements all three types. 4227 */ 4228 static final class Ques extends Node { 4229 Node atom; 4230 Qtype type; 4231 Ques(Node node, Qtype type) { 4232 this.atom = node; 4233 this.type = type; 4234 } 4235 boolean match(Matcher matcher, int i, CharSequence seq) { 4236 switch (type) { 4237 case GREEDY: 4238 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)) 4239 || next.match(matcher, i, seq); 4240 case LAZY: 4241 return next.match(matcher, i, seq) 4242 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)); 4243 case POSSESSIVE: 4244 if (atom.match(matcher, i, seq)) i = matcher.last; 4245 return next.match(matcher, i, seq); 4246 default: 4247 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq); 4248 } 4249 } 4250 boolean study(TreeInfo info) { 4251 if (type != Qtype.INDEPENDENT) { 4252 int minL = info.minLength; 4253 atom.study(info); 4254 info.minLength = minL; 4255 info.deterministic = false; 4256 return next.study(info); 4257 } else { 4258 atom.study(info); 4259 return next.study(info); 4260 } 4261 } 4262 } 4263 4264 /** 4265 * Handles the greedy style repetition with the minimum either be 4266 * 0 or 1 and the maximum be MAX_REPS, for * and + quantifier. 4267 */ 4268 static class CharPropertyGreedy extends Node { 4269 final CharPredicate predicate; 4270 final int cmin; 4271 4272 CharPropertyGreedy(CharProperty cp, int cmin) { 4273 this.predicate = cp.predicate; 4274 this.cmin = cmin; 4275 } 4276 boolean match(Matcher matcher, int i, CharSequence seq) { 4277 int n = 0; 4278 int to = matcher.to; 4279 // greedy, all the way down 4280 while (i < to) { 4281 int ch = Character.codePointAt(seq, i); 4282 if (!predicate.is(ch)) 4283 break; 4284 i += Character.charCount(ch); 4285 n++; 4286 } 4287 if (i >= to) { 4288 matcher.hitEnd = true; 4289 } 4290 while (n >= cmin) { 4291 if (next.match(matcher, i, seq)) 4292 return true; 4293 if (n == cmin) 4294 return false; 4295 // backing off if match fails 4296 int ch = Character.codePointBefore(seq, i); 4297 i -= Character.charCount(ch); 4298 n--; 4299 } 4300 return false; 4301 } 4302 4303 boolean study(TreeInfo info) { 4304 info.minLength += cmin; 4305 if (info.maxValid) { 4306 info.maxLength += MAX_REPS; 4307 } 4308 info.deterministic = false; 4309 return next.study(info); 4310 } 4311 } 4312 4313 static final class BmpCharPropertyGreedy extends CharPropertyGreedy { 4314 4315 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) { 4316 super(bcp, cmin); 4317 } 4318 4319 boolean match(Matcher matcher, int i, CharSequence seq) { 4320 int n = 0; 4321 int to = matcher.to; 4322 while (i < to && predicate.is(seq.charAt(i))) { 4323 i++; n++; 4324 } 4325 if (i >= to) { 4326 matcher.hitEnd = true; 4327 } 4328 while (n >= cmin) { 4329 if (next.match(matcher, i, seq)) 4330 return true; 4331 i--; n--; // backing off if match fails 4332 } 4333 return false; 4334 } 4335 } 4336 4337 /** 4338 * Handles the curly-brace style repetition with a specified minimum and 4339 * maximum occurrences. The * quantifier is handled as a special case. 4340 * This class handles the three types. 4341 */ 4342 static final class Curly extends Node { 4343 Node atom; 4344 Qtype type; 4345 int cmin; 4346 int cmax; 4347 4348 Curly(Node node, int cmin, int cmax, Qtype type) { 4349 this.atom = node; 4350 this.type = type; 4351 this.cmin = cmin; 4352 this.cmax = cmax; 4353 } 4354 boolean match(Matcher matcher, int i, CharSequence seq) { 4355 int j; 4356 for (j = 0; j < cmin; j++) { 4357 if (atom.match(matcher, i, seq)) { 4358 i = matcher.last; 4359 continue; 4360 } 4361 return false; 4362 } 4363 if (type == Qtype.GREEDY) 4364 return match0(matcher, i, j, seq); 4365 else if (type == Qtype.LAZY) 4366 return match1(matcher, i, j, seq); 4367 else 4368 return match2(matcher, i, j, seq); 4369 } 4370 // Greedy match. 4371 // i is the index to start matching at 4372 // j is the number of atoms that have matched 4373 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4374 if (j >= cmax) { 4375 // We have matched the maximum... continue with the rest of 4376 // the regular expression 4377 return next.match(matcher, i, seq); 4378 } 4379 int backLimit = j; 4380 while (atom.match(matcher, i, seq)) { 4381 // k is the length of this match 4382 int k = matcher.last - i; 4383 if (k == 0) // Zero length match 4384 break; 4385 // Move up index and number matched 4386 i = matcher.last; 4387 j++; 4388 // We are greedy so match as many as we can 4389 while (j < cmax) { 4390 if (!atom.match(matcher, i, seq)) 4391 break; 4392 if (i + k != matcher.last) { 4393 if (match0(matcher, matcher.last, j+1, seq)) 4394 return true; 4395 break; 4396 } 4397 i += k; 4398 j++; 4399 } 4400 // Handle backing off if match fails 4401 while (j >= backLimit) { 4402 if (next.match(matcher, i, seq)) 4403 return true; 4404 i -= k; 4405 j--; 4406 } 4407 return false; 4408 } 4409 return next.match(matcher, i, seq); 4410 } 4411 // Reluctant match. At this point, the minimum has been satisfied. 4412 // i is the index to start matching at 4413 // j is the number of atoms that have matched 4414 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4415 for (;;) { 4416 // Try finishing match without consuming any more 4417 if (next.match(matcher, i, seq)) 4418 return true; 4419 // At the maximum, no match found 4420 if (j >= cmax) 4421 return false; 4422 // Okay, must try one more atom 4423 if (!atom.match(matcher, i, seq)) 4424 return false; 4425 // If we haven't moved forward then must break out 4426 if (i == matcher.last) 4427 return false; 4428 // Move up index and number matched 4429 i = matcher.last; 4430 j++; 4431 } 4432 } 4433 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4434 for (; j < cmax; j++) { 4435 if (!atom.match(matcher, i, seq)) 4436 break; 4437 if (i == matcher.last) 4438 break; 4439 i = matcher.last; 4440 } 4441 return next.match(matcher, i, seq); 4442 } 4443 boolean study(TreeInfo info) { 4444 // Save original info 4445 int minL = info.minLength; 4446 int maxL = info.maxLength; 4447 boolean maxV = info.maxValid; 4448 boolean detm = info.deterministic; 4449 info.reset(); 4450 4451 atom.study(info); 4452 4453 int temp = info.minLength * cmin + minL; 4454 if (temp < minL) { 4455 temp = 0xFFFFFFF; // arbitrary large number 4456 } 4457 info.minLength = temp; 4458 4459 if (maxV & info.maxValid) { 4460 temp = info.maxLength * cmax + maxL; 4461 info.maxLength = temp; 4462 if (temp < maxL) { 4463 info.maxValid = false; 4464 } 4465 } else { 4466 info.maxValid = false; 4467 } 4468 4469 if (info.deterministic && cmin == cmax) 4470 info.deterministic = detm; 4471 else 4472 info.deterministic = false; 4473 return next.study(info); 4474 } 4475 } 4476 4477 /** 4478 * Handles the curly-brace style repetition with a specified minimum and 4479 * maximum occurrences in deterministic cases. This is an iterative 4480 * optimization over the Prolog and Loop system which would handle this 4481 * in a recursive way. The * quantifier is handled as a special case. 4482 * If capture is true then this class saves group settings and ensures 4483 * that groups are unset when backing off of a group match. 4484 */ 4485 static final class GroupCurly extends Node { 4486 Node atom; 4487 Qtype type; 4488 int cmin; 4489 int cmax; 4490 int localIndex; 4491 int groupIndex; 4492 boolean capture; 4493 4494 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local, 4495 int group, boolean capture) { 4496 this.atom = node; 4497 this.type = type; 4498 this.cmin = cmin; 4499 this.cmax = cmax; 4500 this.localIndex = local; 4501 this.groupIndex = group; 4502 this.capture = capture; 4503 } 4504 boolean match(Matcher matcher, int i, CharSequence seq) { 4505 int[] groups = matcher.groups; 4506 int[] locals = matcher.locals; 4507 int save0 = locals[localIndex]; 4508 int save1 = 0; 4509 int save2 = 0; 4510 4511 if (capture) { 4512 save1 = groups[groupIndex]; 4513 save2 = groups[groupIndex+1]; 4514 } 4515 4516 // Notify GroupTail there is no need to setup group info 4517 // because it will be set here 4518 locals[localIndex] = -1; 4519 4520 boolean ret = true; 4521 for (int j = 0; j < cmin; j++) { 4522 if (atom.match(matcher, i, seq)) { 4523 if (capture) { 4524 groups[groupIndex] = i; 4525 groups[groupIndex+1] = matcher.last; 4526 } 4527 i = matcher.last; 4528 } else { 4529 ret = false; 4530 break; 4531 } 4532 } 4533 if (ret) { 4534 if (type == Qtype.GREEDY) { 4535 ret = match0(matcher, i, cmin, seq); 4536 } else if (type == Qtype.LAZY) { 4537 ret = match1(matcher, i, cmin, seq); 4538 } else { 4539 ret = match2(matcher, i, cmin, seq); 4540 } 4541 } 4542 if (!ret) { 4543 locals[localIndex] = save0; 4544 if (capture) { 4545 groups[groupIndex] = save1; 4546 groups[groupIndex+1] = save2; 4547 } 4548 } 4549 return ret; 4550 } 4551 // Aggressive group match 4552 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4553 // don't back off passing the starting "j" 4554 int min = j; 4555 int[] groups = matcher.groups; 4556 int save0 = 0; 4557 int save1 = 0; 4558 if (capture) { 4559 save0 = groups[groupIndex]; 4560 save1 = groups[groupIndex+1]; 4561 } 4562 for (;;) { 4563 if (j >= cmax) 4564 break; 4565 if (!atom.match(matcher, i, seq)) 4566 break; 4567 int k = matcher.last - i; 4568 if (k <= 0) { 4569 if (capture) { 4570 groups[groupIndex] = i; 4571 groups[groupIndex+1] = i + k; 4572 } 4573 i = i + k; 4574 break; 4575 } 4576 for (;;) { 4577 if (capture) { 4578 groups[groupIndex] = i; 4579 groups[groupIndex+1] = i + k; 4580 } 4581 i = i + k; 4582 if (++j >= cmax) 4583 break; 4584 if (!atom.match(matcher, i, seq)) 4585 break; 4586 if (i + k != matcher.last) { 4587 if (match0(matcher, i, j, seq)) 4588 return true; 4589 break; 4590 } 4591 } 4592 while (j > min) { 4593 if (next.match(matcher, i, seq)) { 4594 if (capture) { 4595 groups[groupIndex+1] = i; 4596 groups[groupIndex] = i - k; 4597 } 4598 return true; 4599 } 4600 // backing off 4601 i = i - k; 4602 if (capture) { 4603 groups[groupIndex+1] = i; 4604 groups[groupIndex] = i - k; 4605 } 4606 j--; 4607 4608 } 4609 break; 4610 } 4611 if (capture) { 4612 groups[groupIndex] = save0; 4613 groups[groupIndex+1] = save1; 4614 } 4615 return next.match(matcher, i, seq); 4616 } 4617 // Reluctant matching 4618 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4619 for (;;) { 4620 if (next.match(matcher, i, seq)) 4621 return true; 4622 if (j >= cmax) 4623 return false; 4624 if (!atom.match(matcher, i, seq)) 4625 return false; 4626 if (i == matcher.last) 4627 return false; 4628 if (capture) { 4629 matcher.groups[groupIndex] = i; 4630 matcher.groups[groupIndex+1] = matcher.last; 4631 } 4632 i = matcher.last; 4633 j++; 4634 } 4635 } 4636 // Possessive matching 4637 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4638 for (; j < cmax; j++) { 4639 if (!atom.match(matcher, i, seq)) { 4640 break; 4641 } 4642 if (capture) { 4643 matcher.groups[groupIndex] = i; 4644 matcher.groups[groupIndex+1] = matcher.last; 4645 } 4646 if (i == matcher.last) { 4647 break; 4648 } 4649 i = matcher.last; 4650 } 4651 return next.match(matcher, i, seq); 4652 } 4653 boolean study(TreeInfo info) { 4654 // Save original info 4655 int minL = info.minLength; 4656 int maxL = info.maxLength; 4657 boolean maxV = info.maxValid; 4658 boolean detm = info.deterministic; 4659 info.reset(); 4660 4661 atom.study(info); 4662 4663 int temp = info.minLength * cmin + minL; 4664 if (temp < minL) { 4665 temp = 0xFFFFFFF; // Arbitrary large number 4666 } 4667 info.minLength = temp; 4668 4669 if (maxV & info.maxValid) { 4670 temp = info.maxLength * cmax + maxL; 4671 info.maxLength = temp; 4672 if (temp < maxL) { 4673 info.maxValid = false; 4674 } 4675 } else { 4676 info.maxValid = false; 4677 } 4678 4679 if (info.deterministic && cmin == cmax) { 4680 info.deterministic = detm; 4681 } else { 4682 info.deterministic = false; 4683 } 4684 return next.study(info); 4685 } 4686 } 4687 4688 /** 4689 * A Guard node at the end of each atom node in a Branch. It 4690 * serves the purpose of chaining the "match" operation to 4691 * "next" but not the "study", so we can collect the TreeInfo 4692 * of each atom node without including the TreeInfo of the 4693 * "next". 4694 */ 4695 static final class BranchConn extends Node { 4696 BranchConn() {}; 4697 boolean match(Matcher matcher, int i, CharSequence seq) { 4698 return next.match(matcher, i, seq); 4699 } 4700 boolean study(TreeInfo info) { 4701 return info.deterministic; 4702 } 4703 } 4704 4705 /** 4706 * Handles the branching of alternations. Note this is also used for 4707 * the ? quantifier to branch between the case where it matches once 4708 * and where it does not occur. 4709 */ 4710 static final class Branch extends Node { 4711 Node[] atoms = new Node[2]; 4712 int size = 2; 4713 Node conn; 4714 Branch(Node first, Node second, Node branchConn) { 4715 conn = branchConn; 4716 atoms[0] = first; 4717 atoms[1] = second; 4718 } 4719 4720 void add(Node node) { 4721 if (size >= atoms.length) { 4722 Node[] tmp = new Node[atoms.length*2]; 4723 System.arraycopy(atoms, 0, tmp, 0, atoms.length); 4724 atoms = tmp; 4725 } 4726 atoms[size++] = node; 4727 } 4728 4729 boolean match(Matcher matcher, int i, CharSequence seq) { 4730 for (int n = 0; n < size; n++) { 4731 if (atoms[n] == null) { 4732 if (conn.next.match(matcher, i, seq)) 4733 return true; 4734 } else if (atoms[n].match(matcher, i, seq)) { 4735 return true; 4736 } 4737 } 4738 return false; 4739 } 4740 4741 boolean study(TreeInfo info) { 4742 int minL = info.minLength; 4743 int maxL = info.maxLength; 4744 boolean maxV = info.maxValid; 4745 4746 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num 4747 int maxL2 = -1; 4748 for (int n = 0; n < size; n++) { 4749 info.reset(); 4750 if (atoms[n] != null) 4751 atoms[n].study(info); 4752 minL2 = Math.min(minL2, info.minLength); 4753 maxL2 = Math.max(maxL2, info.maxLength); 4754 maxV = (maxV & info.maxValid); 4755 } 4756 4757 minL += minL2; 4758 maxL += maxL2; 4759 4760 info.reset(); 4761 conn.next.study(info); 4762 4763 info.minLength += minL; 4764 info.maxLength += maxL; 4765 info.maxValid &= maxV; 4766 info.deterministic = false; 4767 return false; 4768 } 4769 } 4770 4771 /** 4772 * The GroupHead saves the location where the group begins in the locals 4773 * and restores them when the match is done. 4774 * 4775 * The matchRef is used when a reference to this group is accessed later 4776 * in the expression. The locals will have a negative value in them to 4777 * indicate that we do not want to unset the group if the reference 4778 * doesn't match. 4779 */ 4780 static final class GroupHead extends Node { 4781 int localIndex; 4782 GroupTail tail; // for debug/print only, match does not need to know 4783 GroupHead(int localCount) { 4784 localIndex = localCount; 4785 } 4786 boolean match(Matcher matcher, int i, CharSequence seq) { 4787 int save = matcher.locals[localIndex]; 4788 matcher.locals[localIndex] = i; 4789 boolean ret = next.match(matcher, i, seq); 4790 matcher.locals[localIndex] = save; 4791 return ret; 4792 } 4793 boolean matchRef(Matcher matcher, int i, CharSequence seq) { 4794 int save = matcher.locals[localIndex]; 4795 matcher.locals[localIndex] = ~i; // HACK 4796 boolean ret = next.match(matcher, i, seq); 4797 matcher.locals[localIndex] = save; 4798 return ret; 4799 } 4800 } 4801 4802 /** 4803 * Recursive reference to a group in the regular expression. It calls 4804 * matchRef because if the reference fails to match we would not unset 4805 * the group. 4806 */ 4807 static final class GroupRef extends Node { 4808 GroupHead head; 4809 GroupRef(GroupHead head) { 4810 this.head = head; 4811 } 4812 boolean match(Matcher matcher, int i, CharSequence seq) { 4813 return head.matchRef(matcher, i, seq) 4814 && next.match(matcher, matcher.last, seq); 4815 } 4816 boolean study(TreeInfo info) { 4817 info.maxValid = false; 4818 info.deterministic = false; 4819 return next.study(info); 4820 } 4821 } 4822 4823 /** 4824 * The GroupTail handles the setting of group beginning and ending 4825 * locations when groups are successfully matched. It must also be able to 4826 * unset groups that have to be backed off of. 4827 * 4828 * The GroupTail node is also used when a previous group is referenced, 4829 * and in that case no group information needs to be set. 4830 */ 4831 static final class GroupTail extends Node { 4832 int localIndex; 4833 int groupIndex; 4834 GroupTail(int localCount, int groupCount) { 4835 localIndex = localCount; 4836 groupIndex = groupCount + groupCount; 4837 } 4838 boolean match(Matcher matcher, int i, CharSequence seq) { 4839 int tmp = matcher.locals[localIndex]; 4840 if (tmp >= 0) { // This is the normal group case. 4841 // Save the group so we can unset it if it 4842 // backs off of a match. 4843 int groupStart = matcher.groups[groupIndex]; 4844 int groupEnd = matcher.groups[groupIndex+1]; 4845 4846 matcher.groups[groupIndex] = tmp; 4847 matcher.groups[groupIndex+1] = i; 4848 if (next.match(matcher, i, seq)) { 4849 return true; 4850 } 4851 matcher.groups[groupIndex] = groupStart; 4852 matcher.groups[groupIndex+1] = groupEnd; 4853 return false; 4854 } else { 4855 // This is a group reference case. We don't need to save any 4856 // group info because it isn't really a group. 4857 matcher.last = i; 4858 return true; 4859 } 4860 } 4861 } 4862 4863 /** 4864 * This sets up a loop to handle a recursive quantifier structure. 4865 */ 4866 static final class Prolog extends Node { 4867 Loop loop; 4868 Prolog(Loop loop) { 4869 this.loop = loop; 4870 } 4871 boolean match(Matcher matcher, int i, CharSequence seq) { 4872 return loop.matchInit(matcher, i, seq); 4873 } 4874 boolean study(TreeInfo info) { 4875 return loop.study(info); 4876 } 4877 } 4878 4879 /** 4880 * Handles the repetition count for a greedy Curly. The matchInit 4881 * is called from the Prolog to save the index of where the group 4882 * beginning is stored. A zero length group check occurs in the 4883 * normal match but is skipped in the matchInit. 4884 */ 4885 static class Loop extends Node { 4886 Node body; 4887 int countIndex; // local count index in matcher locals 4888 int beginIndex; // group beginning index 4889 int cmin, cmax; 4890 int posIndex; 4891 Loop(int countIndex, int beginIndex) { 4892 this.countIndex = countIndex; 4893 this.beginIndex = beginIndex; 4894 this.posIndex = -1; 4895 } 4896 boolean match(Matcher matcher, int i, CharSequence seq) { 4897 // Avoid infinite loop in zero-length case. 4898 if (i > matcher.locals[beginIndex]) { 4899 int count = matcher.locals[countIndex]; 4900 4901 // This block is for before we reach the minimum 4902 // iterations required for the loop to match 4903 if (count < cmin) { 4904 matcher.locals[countIndex] = count + 1; 4905 boolean b = body.match(matcher, i, seq); 4906 // If match failed we must backtrack, so 4907 // the loop count should NOT be incremented 4908 if (!b) 4909 matcher.locals[countIndex] = count; 4910 // Return success or failure since we are under 4911 // minimum 4912 return b; 4913 } 4914 // This block is for after we have the minimum 4915 // iterations required for the loop to match 4916 if (count < cmax) { 4917 // Let's check if we have already tried and failed 4918 // at this starting position "i" in the past. 4919 // If yes, then just return false wihtout trying 4920 // again, to stop the exponential backtracking. 4921 if (posIndex != -1 && 4922 matcher.localsPos[posIndex].contains(i)) { 4923 return next.match(matcher, i, seq); 4924 } 4925 matcher.locals[countIndex] = count + 1; 4926 boolean b = body.match(matcher, i, seq); 4927 // If match failed we must backtrack, so 4928 // the loop count should NOT be incremented 4929 if (b) 4930 return true; 4931 matcher.locals[countIndex] = count; 4932 // save the failed position 4933 if (posIndex != -1) { 4934 matcher.localsPos[posIndex].add(i); 4935 } 4936 } 4937 } 4938 return next.match(matcher, i, seq); 4939 } 4940 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4941 int save = matcher.locals[countIndex]; 4942 boolean ret = false; 4943 if (posIndex != -1 && matcher.localsPos[posIndex] == null) { 4944 matcher.localsPos[posIndex] = new IntHashSet(); 4945 } 4946 if (0 < cmin) { 4947 matcher.locals[countIndex] = 1; 4948 ret = body.match(matcher, i, seq); 4949 } else if (0 < cmax) { 4950 matcher.locals[countIndex] = 1; 4951 ret = body.match(matcher, i, seq); 4952 if (ret == false) 4953 ret = next.match(matcher, i, seq); 4954 } else { 4955 ret = next.match(matcher, i, seq); 4956 } 4957 matcher.locals[countIndex] = save; 4958 return ret; 4959 } 4960 boolean study(TreeInfo info) { 4961 info.maxValid = false; 4962 info.deterministic = false; 4963 return false; 4964 } 4965 } 4966 4967 /** 4968 * Handles the repetition count for a reluctant Curly. The matchInit 4969 * is called from the Prolog to save the index of where the group 4970 * beginning is stored. A zero length group check occurs in the 4971 * normal match but is skipped in the matchInit. 4972 */ 4973 static final class LazyLoop extends Loop { 4974 LazyLoop(int countIndex, int beginIndex) { 4975 super(countIndex, beginIndex); 4976 } 4977 boolean match(Matcher matcher, int i, CharSequence seq) { 4978 // Check for zero length group 4979 if (i > matcher.locals[beginIndex]) { 4980 int count = matcher.locals[countIndex]; 4981 if (count < cmin) { 4982 matcher.locals[countIndex] = count + 1; 4983 boolean result = body.match(matcher, i, seq); 4984 // If match failed we must backtrack, so 4985 // the loop count should NOT be incremented 4986 if (!result) 4987 matcher.locals[countIndex] = count; 4988 return result; 4989 } 4990 if (next.match(matcher, i, seq)) 4991 return true; 4992 if (count < cmax) { 4993 matcher.locals[countIndex] = count + 1; 4994 boolean result = body.match(matcher, i, seq); 4995 // If match failed we must backtrack, so 4996 // the loop count should NOT be incremented 4997 if (!result) 4998 matcher.locals[countIndex] = count; 4999 return result; 5000 } 5001 return false; 5002 } 5003 return next.match(matcher, i, seq); 5004 } 5005 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 5006 int save = matcher.locals[countIndex]; 5007 boolean ret = false; 5008 if (0 < cmin) { 5009 matcher.locals[countIndex] = 1; 5010 ret = body.match(matcher, i, seq); 5011 } else if (next.match(matcher, i, seq)) { 5012 ret = true; 5013 } else if (0 < cmax) { 5014 matcher.locals[countIndex] = 1; 5015 ret = body.match(matcher, i, seq); 5016 } 5017 matcher.locals[countIndex] = save; 5018 return ret; 5019 } 5020 boolean study(TreeInfo info) { 5021 info.maxValid = false; 5022 info.deterministic = false; 5023 return false; 5024 } 5025 } 5026 5027 /** 5028 * Refers to a group in the regular expression. Attempts to match 5029 * whatever the group referred to last matched. 5030 */ 5031 static class BackRef extends Node { 5032 int groupIndex; 5033 BackRef(int groupCount) { 5034 super(); 5035 groupIndex = groupCount + groupCount; 5036 } 5037 boolean match(Matcher matcher, int i, CharSequence seq) { 5038 int j = matcher.groups[groupIndex]; 5039 int k = matcher.groups[groupIndex+1]; 5040 5041 int groupSize = k - j; 5042 // If the referenced group didn't match, neither can this 5043 if (j < 0) 5044 return false; 5045 5046 // If there isn't enough input left no match 5047 if (i + groupSize > matcher.to) { 5048 matcher.hitEnd = true; 5049 return false; 5050 } 5051 // Check each new char to make sure it matches what the group 5052 // referenced matched last time around 5053 for (int index=0; index<groupSize; index++) 5054 if (seq.charAt(i+index) != seq.charAt(j+index)) 5055 return false; 5056 5057 return next.match(matcher, i+groupSize, seq); 5058 } 5059 boolean study(TreeInfo info) { 5060 info.maxValid = false; 5061 return next.study(info); 5062 } 5063 } 5064 5065 static class CIBackRef extends Node { 5066 int groupIndex; 5067 boolean doUnicodeCase; 5068 CIBackRef(int groupCount, boolean doUnicodeCase) { 5069 super(); 5070 groupIndex = groupCount + groupCount; 5071 this.doUnicodeCase = doUnicodeCase; 5072 } 5073 boolean match(Matcher matcher, int i, CharSequence seq) { 5074 int j = matcher.groups[groupIndex]; 5075 int k = matcher.groups[groupIndex+1]; 5076 5077 int groupSize = k - j; 5078 5079 // If the referenced group didn't match, neither can this 5080 if (j < 0) 5081 return false; 5082 5083 // If there isn't enough input left no match 5084 if (i + groupSize > matcher.to) { 5085 matcher.hitEnd = true; 5086 return false; 5087 } 5088 5089 // Check each new char to make sure it matches what the group 5090 // referenced matched last time around 5091 int x = i; 5092 for (int index=0; index<groupSize; index++) { 5093 int c1 = Character.codePointAt(seq, x); 5094 int c2 = Character.codePointAt(seq, j); 5095 if (c1 != c2) { 5096 if (doUnicodeCase) { 5097 int cc1 = Character.toUpperCase(c1); 5098 int cc2 = Character.toUpperCase(c2); 5099 if (cc1 != cc2 && 5100 Character.toLowerCase(cc1) != 5101 Character.toLowerCase(cc2)) 5102 return false; 5103 } else { 5104 if (ASCII.toLower(c1) != ASCII.toLower(c2)) 5105 return false; 5106 } 5107 } 5108 x += Character.charCount(c1); 5109 j += Character.charCount(c2); 5110 } 5111 5112 return next.match(matcher, i+groupSize, seq); 5113 } 5114 boolean study(TreeInfo info) { 5115 info.maxValid = false; 5116 return next.study(info); 5117 } 5118 } 5119 5120 /** 5121 * Searches until the next instance of its atom. This is useful for 5122 * finding the atom efficiently without passing an instance of it 5123 * (greedy problem) and without a lot of wasted search time (reluctant 5124 * problem). 5125 */ 5126 static final class First extends Node { 5127 Node atom; 5128 First(Node node) { 5129 this.atom = BnM.optimize(node); 5130 } 5131 boolean match(Matcher matcher, int i, CharSequence seq) { 5132 if (atom instanceof BnM) { 5133 return atom.match(matcher, i, seq) 5134 && next.match(matcher, matcher.last, seq); 5135 } 5136 for (;;) { 5137 if (i > matcher.to) { 5138 matcher.hitEnd = true; 5139 return false; 5140 } 5141 if (atom.match(matcher, i, seq)) { 5142 return next.match(matcher, matcher.last, seq); 5143 } 5144 i += countChars(seq, i, 1); 5145 matcher.first++; 5146 } 5147 } 5148 boolean study(TreeInfo info) { 5149 atom.study(info); 5150 info.maxValid = false; 5151 info.deterministic = false; 5152 return next.study(info); 5153 } 5154 } 5155 5156 static final class Conditional extends Node { 5157 Node cond, yes, not; 5158 Conditional(Node cond, Node yes, Node not) { 5159 this.cond = cond; 5160 this.yes = yes; 5161 this.not = not; 5162 } 5163 boolean match(Matcher matcher, int i, CharSequence seq) { 5164 if (cond.match(matcher, i, seq)) { 5165 return yes.match(matcher, i, seq); 5166 } else { 5167 return not.match(matcher, i, seq); 5168 } 5169 } 5170 boolean study(TreeInfo info) { 5171 int minL = info.minLength; 5172 int maxL = info.maxLength; 5173 boolean maxV = info.maxValid; 5174 info.reset(); 5175 yes.study(info); 5176 5177 int minL2 = info.minLength; 5178 int maxL2 = info.maxLength; 5179 boolean maxV2 = info.maxValid; 5180 info.reset(); 5181 not.study(info); 5182 5183 info.minLength = minL + Math.min(minL2, info.minLength); 5184 info.maxLength = maxL + Math.max(maxL2, info.maxLength); 5185 info.maxValid = (maxV & maxV2 & info.maxValid); 5186 info.deterministic = false; 5187 return next.study(info); 5188 } 5189 } 5190 5191 /** 5192 * Zero width positive lookahead. 5193 */ 5194 static final class Pos extends Node { 5195 Node cond; 5196 Pos(Node cond) { 5197 this.cond = cond; 5198 } 5199 boolean match(Matcher matcher, int i, CharSequence seq) { 5200 int savedTo = matcher.to; 5201 boolean conditionMatched = false; 5202 5203 // Relax transparent region boundaries for lookahead 5204 if (matcher.transparentBounds) 5205 matcher.to = matcher.getTextLength(); 5206 try { 5207 conditionMatched = cond.match(matcher, i, seq); 5208 } finally { 5209 // Reinstate region boundaries 5210 matcher.to = savedTo; 5211 } 5212 return conditionMatched && next.match(matcher, i, seq); 5213 } 5214 } 5215 5216 /** 5217 * Zero width negative lookahead. 5218 */ 5219 static final class Neg extends Node { 5220 Node cond; 5221 Neg(Node cond) { 5222 this.cond = cond; 5223 } 5224 boolean match(Matcher matcher, int i, CharSequence seq) { 5225 int savedTo = matcher.to; 5226 boolean conditionMatched = false; 5227 5228 // Relax transparent region boundaries for lookahead 5229 if (matcher.transparentBounds) 5230 matcher.to = matcher.getTextLength(); 5231 try { 5232 if (i < matcher.to) { 5233 conditionMatched = !cond.match(matcher, i, seq); 5234 } else { 5235 // If a negative lookahead succeeds then more input 5236 // could cause it to fail! 5237 matcher.requireEnd = true; 5238 conditionMatched = !cond.match(matcher, i, seq); 5239 } 5240 } finally { 5241 // Reinstate region boundaries 5242 matcher.to = savedTo; 5243 } 5244 return conditionMatched && next.match(matcher, i, seq); 5245 } 5246 } 5247 5248 /** 5249 * For use with lookbehinds; matches the position where the lookbehind 5250 * was encountered. 5251 */ 5252 static Node lookbehindEnd = new Node() { 5253 boolean match(Matcher matcher, int i, CharSequence seq) { 5254 return i == matcher.lookbehindTo; 5255 } 5256 }; 5257 5258 /** 5259 * Zero width positive lookbehind. 5260 */ 5261 static class Behind extends Node { 5262 Node cond; 5263 int rmax, rmin; 5264 Behind(Node cond, int rmax, int rmin) { 5265 this.cond = cond; 5266 this.rmax = rmax; 5267 this.rmin = rmin; 5268 } 5269 5270 boolean match(Matcher matcher, int i, CharSequence seq) { 5271 int savedFrom = matcher.from; 5272 boolean conditionMatched = false; 5273 int startIndex = (!matcher.transparentBounds) ? 5274 matcher.from : 0; 5275 int from = Math.max(i - rmax, startIndex); 5276 // Set end boundary 5277 int savedLBT = matcher.lookbehindTo; 5278 matcher.lookbehindTo = i; 5279 // Relax transparent region boundaries for lookbehind 5280 if (matcher.transparentBounds) 5281 matcher.from = 0; 5282 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5283 conditionMatched = cond.match(matcher, j, seq); 5284 } 5285 matcher.from = savedFrom; 5286 matcher.lookbehindTo = savedLBT; 5287 return conditionMatched && next.match(matcher, i, seq); 5288 } 5289 } 5290 5291 /** 5292 * Zero width positive lookbehind, including supplementary 5293 * characters or unpaired surrogates. 5294 */ 5295 static final class BehindS extends Behind { 5296 BehindS(Node cond, int rmax, int rmin) { 5297 super(cond, rmax, rmin); 5298 } 5299 boolean match(Matcher matcher, int i, CharSequence seq) { 5300 int rmaxChars = countChars(seq, i, -rmax); 5301 int rminChars = countChars(seq, i, -rmin); 5302 int savedFrom = matcher.from; 5303 int startIndex = (!matcher.transparentBounds) ? 5304 matcher.from : 0; 5305 boolean conditionMatched = false; 5306 int from = Math.max(i - rmaxChars, startIndex); 5307 // Set end boundary 5308 int savedLBT = matcher.lookbehindTo; 5309 matcher.lookbehindTo = i; 5310 // Relax transparent region boundaries for lookbehind 5311 if (matcher.transparentBounds) 5312 matcher.from = 0; 5313 5314 for (int j = i - rminChars; 5315 !conditionMatched && j >= from; 5316 j -= j>from ? countChars(seq, j, -1) : 1) { 5317 conditionMatched = cond.match(matcher, j, seq); 5318 } 5319 matcher.from = savedFrom; 5320 matcher.lookbehindTo = savedLBT; 5321 return conditionMatched && next.match(matcher, i, seq); 5322 } 5323 } 5324 5325 /** 5326 * Zero width negative lookbehind. 5327 */ 5328 static class NotBehind extends Node { 5329 Node cond; 5330 int rmax, rmin; 5331 NotBehind(Node cond, int rmax, int rmin) { 5332 this.cond = cond; 5333 this.rmax = rmax; 5334 this.rmin = rmin; 5335 } 5336 5337 boolean match(Matcher matcher, int i, CharSequence seq) { 5338 int savedLBT = matcher.lookbehindTo; 5339 int savedFrom = matcher.from; 5340 boolean conditionMatched = false; 5341 int startIndex = (!matcher.transparentBounds) ? 5342 matcher.from : 0; 5343 int from = Math.max(i - rmax, startIndex); 5344 matcher.lookbehindTo = i; 5345 // Relax transparent region boundaries for lookbehind 5346 if (matcher.transparentBounds) 5347 matcher.from = 0; 5348 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5349 conditionMatched = cond.match(matcher, j, seq); 5350 } 5351 // Reinstate region boundaries 5352 matcher.from = savedFrom; 5353 matcher.lookbehindTo = savedLBT; 5354 return !conditionMatched && next.match(matcher, i, seq); 5355 } 5356 } 5357 5358 /** 5359 * Zero width negative lookbehind, including supplementary 5360 * characters or unpaired surrogates. 5361 */ 5362 static final class NotBehindS extends NotBehind { 5363 NotBehindS(Node cond, int rmax, int rmin) { 5364 super(cond, rmax, rmin); 5365 } 5366 boolean match(Matcher matcher, int i, CharSequence seq) { 5367 int rmaxChars = countChars(seq, i, -rmax); 5368 int rminChars = countChars(seq, i, -rmin); 5369 int savedFrom = matcher.from; 5370 int savedLBT = matcher.lookbehindTo; 5371 boolean conditionMatched = false; 5372 int startIndex = (!matcher.transparentBounds) ? 5373 matcher.from : 0; 5374 int from = Math.max(i - rmaxChars, startIndex); 5375 matcher.lookbehindTo = i; 5376 // Relax transparent region boundaries for lookbehind 5377 if (matcher.transparentBounds) 5378 matcher.from = 0; 5379 for (int j = i - rminChars; 5380 !conditionMatched && j >= from; 5381 j -= j>from ? countChars(seq, j, -1) : 1) { 5382 conditionMatched = cond.match(matcher, j, seq); 5383 } 5384 //Reinstate region boundaries 5385 matcher.from = savedFrom; 5386 matcher.lookbehindTo = savedLBT; 5387 return !conditionMatched && next.match(matcher, i, seq); 5388 } 5389 } 5390 5391 /** 5392 * Handles word boundaries. Includes a field to allow this one class to 5393 * deal with the different types of word boundaries we can match. The word 5394 * characters include underscores, letters, and digits. Non spacing marks 5395 * can are also part of a word if they have a base character, otherwise 5396 * they are ignored for purposes of finding word boundaries. 5397 */ 5398 static final class Bound extends Node { 5399 static int LEFT = 0x1; 5400 static int RIGHT= 0x2; 5401 static int BOTH = 0x3; 5402 static int NONE = 0x4; 5403 int type; 5404 boolean useUWORD; 5405 Bound(int n, boolean useUWORD) { 5406 type = n; 5407 this.useUWORD = useUWORD; 5408 } 5409 5410 boolean isWord(int ch) { 5411 return useUWORD ? CharPredicates.WORD().is(ch) 5412 : (ch == '_' || Character.isLetterOrDigit(ch)); 5413 } 5414 5415 int check(Matcher matcher, int i, CharSequence seq) { 5416 int ch; 5417 boolean left = false; 5418 int startIndex = matcher.from; 5419 int endIndex = matcher.to; 5420 if (matcher.transparentBounds) { 5421 startIndex = 0; 5422 endIndex = matcher.getTextLength(); 5423 } 5424 if (i > startIndex) { 5425 ch = Character.codePointBefore(seq, i); 5426 left = (isWord(ch) || 5427 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5428 && hasBaseCharacter(matcher, i-1, seq))); 5429 } 5430 boolean right = false; 5431 if (i < endIndex) { 5432 ch = Character.codePointAt(seq, i); 5433 right = (isWord(ch) || 5434 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5435 && hasBaseCharacter(matcher, i, seq))); 5436 } else { 5437 // Tried to access char past the end 5438 matcher.hitEnd = true; 5439 // The addition of another char could wreck a boundary 5440 matcher.requireEnd = true; 5441 } 5442 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE); 5443 } 5444 boolean match(Matcher matcher, int i, CharSequence seq) { 5445 return (check(matcher, i, seq) & type) > 0 5446 && next.match(matcher, i, seq); 5447 } 5448 } 5449 5450 /** 5451 * Non spacing marks only count as word characters in bounds calculations 5452 * if they have a base character. 5453 */ 5454 private static boolean hasBaseCharacter(Matcher matcher, int i, 5455 CharSequence seq) 5456 { 5457 int start = (!matcher.transparentBounds) ? 5458 matcher.from : 0; 5459 for (int x=i; x >= start; x--) { 5460 int ch = Character.codePointAt(seq, x); 5461 if (Character.isLetterOrDigit(ch)) 5462 return true; 5463 if (Character.getType(ch) == Character.NON_SPACING_MARK) 5464 continue; 5465 return false; 5466 } 5467 return false; 5468 } 5469 5470 /** 5471 * Attempts to match a slice in the input using the Boyer-Moore string 5472 * matching algorithm. The algorithm is based on the idea that the 5473 * pattern can be shifted farther ahead in the search text if it is 5474 * matched right to left. 5475 * <p> 5476 * The pattern is compared to the input one character at a time, from 5477 * the rightmost character in the pattern to the left. If the characters 5478 * all match the pattern has been found. If a character does not match, 5479 * the pattern is shifted right a distance that is the maximum of two 5480 * functions, the bad character shift and the good suffix shift. This 5481 * shift moves the attempted match position through the input more 5482 * quickly than a naive one position at a time check. 5483 * <p> 5484 * The bad character shift is based on the character from the text that 5485 * did not match. If the character does not appear in the pattern, the 5486 * pattern can be shifted completely beyond the bad character. If the 5487 * character does occur in the pattern, the pattern can be shifted to 5488 * line the pattern up with the next occurrence of that character. 5489 * <p> 5490 * The good suffix shift is based on the idea that some subset on the right 5491 * side of the pattern has matched. When a bad character is found, the 5492 * pattern can be shifted right by the pattern length if the subset does 5493 * not occur again in pattern, or by the amount of distance to the 5494 * next occurrence of the subset in the pattern. 5495 * 5496 * Boyer-Moore search methods adapted from code by Amy Yu. 5497 */ 5498 static class BnM extends Node { 5499 int[] buffer; 5500 int[] lastOcc; 5501 int[] optoSft; 5502 5503 /** 5504 * Pre calculates arrays needed to generate the bad character 5505 * shift and the good suffix shift. Only the last seven bits 5506 * are used to see if chars match; This keeps the tables small 5507 * and covers the heavily used ASCII range, but occasionally 5508 * results in an aliased match for the bad character shift. 5509 */ 5510 static Node optimize(Node node) { 5511 if (!(node instanceof Slice)) { 5512 return node; 5513 } 5514 5515 int[] src = ((Slice) node).buffer; 5516 int patternLength = src.length; 5517 // The BM algorithm requires a bit of overhead; 5518 // If the pattern is short don't use it, since 5519 // a shift larger than the pattern length cannot 5520 // be used anyway. 5521 if (patternLength < 4) { 5522 return node; 5523 } 5524 int i, j, k; 5525 int[] lastOcc = new int[128]; 5526 int[] optoSft = new int[patternLength]; 5527 // Precalculate part of the bad character shift 5528 // It is a table for where in the pattern each 5529 // lower 7-bit value occurs 5530 for (i = 0; i < patternLength; i++) { 5531 lastOcc[src[i]&0x7F] = i + 1; 5532 } 5533 // Precalculate the good suffix shift 5534 // i is the shift amount being considered 5535 NEXT: for (i = patternLength; i > 0; i--) { 5536 // j is the beginning index of suffix being considered 5537 for (j = patternLength - 1; j >= i; j--) { 5538 // Testing for good suffix 5539 if (src[j] == src[j-i]) { 5540 // src[j..len] is a good suffix 5541 optoSft[j-1] = i; 5542 } else { 5543 // No match. The array has already been 5544 // filled up with correct values before. 5545 continue NEXT; 5546 } 5547 } 5548 // This fills up the remaining of optoSft 5549 // any suffix can not have larger shift amount 5550 // then its sub-suffix. Why??? 5551 while (j > 0) { 5552 optoSft[--j] = i; 5553 } 5554 } 5555 // Set the guard value because of unicode compression 5556 optoSft[patternLength-1] = 1; 5557 if (node instanceof SliceS) 5558 return new BnMS(src, lastOcc, optoSft, node.next); 5559 return new BnM(src, lastOcc, optoSft, node.next); 5560 } 5561 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5562 this.buffer = src; 5563 this.lastOcc = lastOcc; 5564 this.optoSft = optoSft; 5565 this.next = next; 5566 } 5567 boolean match(Matcher matcher, int i, CharSequence seq) { 5568 int[] src = buffer; 5569 int patternLength = src.length; 5570 int last = matcher.to - patternLength; 5571 5572 // Loop over all possible match positions in text 5573 NEXT: while (i <= last) { 5574 // Loop over pattern from right to left 5575 for (int j = patternLength - 1; j >= 0; j--) { 5576 int ch = seq.charAt(i+j); 5577 if (ch != src[j]) { 5578 // Shift search to the right by the maximum of the 5579 // bad character shift and the good suffix shift 5580 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]); 5581 continue NEXT; 5582 } 5583 } 5584 // Entire pattern matched starting at i 5585 matcher.first = i; 5586 boolean ret = next.match(matcher, i + patternLength, seq); 5587 if (ret) { 5588 matcher.first = i; 5589 matcher.groups[0] = matcher.first; 5590 matcher.groups[1] = matcher.last; 5591 return true; 5592 } 5593 i++; 5594 } 5595 // BnM is only used as the leading node in the unanchored case, 5596 // and it replaced its Start() which always searches to the end 5597 // if it doesn't find what it's looking for, so hitEnd is true. 5598 matcher.hitEnd = true; 5599 return false; 5600 } 5601 boolean study(TreeInfo info) { 5602 info.minLength += buffer.length; 5603 info.maxValid = false; 5604 return next.study(info); 5605 } 5606 } 5607 5608 /** 5609 * Supplementary support version of BnM(). Unpaired surrogates are 5610 * also handled by this class. 5611 */ 5612 static final class BnMS extends BnM { 5613 int lengthInChars; 5614 5615 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5616 super(src, lastOcc, optoSft, next); 5617 for (int cp : buffer) { 5618 lengthInChars += Character.charCount(cp); 5619 } 5620 } 5621 boolean match(Matcher matcher, int i, CharSequence seq) { 5622 int[] src = buffer; 5623 int patternLength = src.length; 5624 int last = matcher.to - lengthInChars; 5625 5626 // Loop over all possible match positions in text 5627 NEXT: while (i <= last) { 5628 // Loop over pattern from right to left 5629 int ch; 5630 for (int j = countChars(seq, i, patternLength), x = patternLength - 1; 5631 j > 0; j -= Character.charCount(ch), x--) { 5632 ch = Character.codePointBefore(seq, i+j); 5633 if (ch != src[x]) { 5634 // Shift search to the right by the maximum of the 5635 // bad character shift and the good suffix shift 5636 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]); 5637 i += countChars(seq, i, n); 5638 continue NEXT; 5639 } 5640 } 5641 // Entire pattern matched starting at i 5642 matcher.first = i; 5643 boolean ret = next.match(matcher, i + lengthInChars, seq); 5644 if (ret) { 5645 matcher.first = i; 5646 matcher.groups[0] = matcher.first; 5647 matcher.groups[1] = matcher.last; 5648 return true; 5649 } 5650 i += countChars(seq, i, 1); 5651 } 5652 matcher.hitEnd = true; 5653 return false; 5654 } 5655 } 5656 5657 @FunctionalInterface 5658 static interface CharPredicate { 5659 boolean is(int ch); 5660 5661 default CharPredicate and(CharPredicate p) { 5662 return ch -> is(ch) && p.is(ch); 5663 } 5664 default CharPredicate union(CharPredicate p) { 5665 return ch -> is(ch) || p.is(ch); 5666 } 5667 default CharPredicate union(CharPredicate p1, 5668 CharPredicate p2 ) { 5669 return ch -> is(ch) || p1.is(ch) || p2.is(ch); 5670 } 5671 default CharPredicate negate() { 5672 return ch -> !is(ch); 5673 } 5674 } 5675 5676 static interface BmpCharPredicate extends CharPredicate { 5677 5678 default CharPredicate and(CharPredicate p) { 5679 if(p instanceof BmpCharPredicate) 5680 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch)); 5681 return ch -> is(ch) && p.is(ch); 5682 } 5683 default CharPredicate union(CharPredicate p) { 5684 if (p instanceof BmpCharPredicate) 5685 return (BmpCharPredicate)(ch -> is(ch) || p.is(ch)); 5686 return ch -> is(ch) || p.is(ch); 5687 } 5688 static CharPredicate union(CharPredicate... predicates) { 5689 CharPredicate cp = ch -> { 5690 for (CharPredicate p : predicates) { 5691 if (!p.is(ch)) 5692 return false; 5693 } 5694 return true; 5695 }; 5696 for (CharPredicate p : predicates) { 5697 if (! (p instanceof BmpCharPredicate)) 5698 return cp; 5699 } 5700 return (BmpCharPredicate)cp; 5701 } 5702 } 5703 5704 /** 5705 * matches a Perl vertical whitespace 5706 */ 5707 static BmpCharPredicate VertWS() { 5708 return cp -> (cp >= 0x0A && cp <= 0x0D) || 5709 cp == 0x85 || cp == 0x2028 || cp == 0x2029; 5710 } 5711 5712 /** 5713 * matches a Perl horizontal whitespace 5714 */ 5715 static BmpCharPredicate HorizWS() { 5716 return cp -> 5717 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 || 5718 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f || 5719 cp == 0x205f || cp == 0x3000; 5720 } 5721 5722 /** 5723 * for the Unicode category ALL and the dot metacharacter when 5724 * in dotall mode. 5725 */ 5726 static CharPredicate ALL() { 5727 return ch -> true; 5728 } 5729 5730 /** 5731 * for the dot metacharacter when dotall is not enabled. 5732 */ 5733 static CharPredicate DOT() { 5734 return ch -> 5735 (ch != '\n' && ch != '\r' 5736 && (ch|1) != '\u2029' 5737 && ch != '\u0085'); 5738 } 5739 5740 /** 5741 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled. 5742 */ 5743 static CharPredicate UNIXDOT() { 5744 return ch -> ch != '\n'; 5745 } 5746 5747 /** 5748 * Indicate that matches a Supplementary Unicode character 5749 */ 5750 static CharPredicate SingleS(int c) { 5751 return ch -> ch == c; 5752 } 5753 5754 /** 5755 * A bmp/optimized predicate of single 5756 */ 5757 static BmpCharPredicate Single(int c) { 5758 return ch -> ch == c; 5759 } 5760 5761 /** 5762 * Case insensitive matches a given BMP character 5763 */ 5764 static BmpCharPredicate SingleI(int lower, int upper) { 5765 return ch -> ch == lower || ch == upper; 5766 } 5767 5768 /** 5769 * Unicode case insensitive matches a given Unicode character 5770 */ 5771 static CharPredicate SingleU(int lower) { 5772 return ch -> lower == ch || 5773 lower == Character.toLowerCase(Character.toUpperCase(ch)); 5774 } 5775 5776 private static boolean inRange(int lower, int ch, int upper) { 5777 return lower <= ch && ch <= upper; 5778 } 5779 5780 /** 5781 * Charactrs within a explicit value range 5782 */ 5783 static CharPredicate Range(int lower, int upper) { 5784 if (upper < Character.MIN_HIGH_SURROGATE || 5785 lower > Character.MAX_HIGH_SURROGATE && 5786 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT) 5787 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper)); 5788 return ch -> inRange(lower, ch, upper); 5789 } 5790 5791 /** 5792 * Charactrs within a explicit value range in a case insensitive manner. 5793 */ 5794 static CharPredicate CIRange(int lower, int upper) { 5795 return ch -> inRange(lower, ch, upper) || 5796 ASCII.isAscii(ch) && 5797 (inRange(lower, ASCII.toUpper(ch), upper) || 5798 inRange(lower, ASCII.toLower(ch), upper)); 5799 } 5800 5801 static CharPredicate CIRangeU(int lower, int upper) { 5802 return ch -> { 5803 if (inRange(lower, ch, upper)) 5804 return true; 5805 int up = Character.toUpperCase(ch); 5806 return inRange(lower, up, upper) || 5807 inRange(lower, Character.toLowerCase(up), upper); 5808 }; 5809 } 5810 5811 /** 5812 * This must be the very first initializer. 5813 */ 5814 static final Node accept = new Node(); 5815 5816 static final Node lastAccept = new LastNode(); 5817 5818 /** 5819 * Creates a predicate which can be used to match a string. 5820 * 5821 * @return The predicate which can be used for matching on a string 5822 * @since 1.8 5823 */ 5824 public Predicate<String> asPredicate() { 5825 return s -> matcher(s).find(); 5826 } 5827 5828 /** 5829 * Creates a stream from the given input sequence around matches of this 5830 * pattern. 5831 * 5832 * <p> The stream returned by this method contains each substring of the 5833 * input sequence that is terminated by another subsequence that matches 5834 * this pattern or is terminated by the end of the input sequence. The 5835 * substrings in the stream are in the order in which they occur in the 5836 * input. Trailing empty strings will be discarded and not encountered in 5837 * the stream. 5838 * 5839 * <p> If this pattern does not match any subsequence of the input then 5840 * the resulting stream has just one element, namely the input sequence in 5841 * string form. 5842 * 5843 * <p> When there is a positive-width match at the beginning of the input 5844 * sequence then an empty leading substring is included at the beginning 5845 * of the stream. A zero-width match at the beginning however never produces 5846 * such empty leading substring. 5847 * 5848 * <p> If the input sequence is mutable, it must remain constant during the 5849 * execution of the terminal stream operation. Otherwise, the result of the 5850 * terminal stream operation is undefined. 5851 * 5852 * @param input 5853 * The character sequence to be split 5854 * 5855 * @return The stream of strings computed by splitting the input 5856 * around matches of this pattern 5857 * @see #split(CharSequence) 5858 * @since 1.8 5859 */ 5860 public Stream<String> splitAsStream(final CharSequence input) { 5861 class MatcherIterator implements Iterator<String> { 5862 private Matcher matcher; 5863 // The start position of the next sub-sequence of input 5864 // when current == input.length there are no more elements 5865 private int current; 5866 // null if the next element, if any, needs to obtained 5867 private String nextElement; 5868 // > 0 if there are N next empty elements 5869 private int emptyElementCount; 5870 5871 public String next() { 5872 if (!hasNext()) 5873 throw new NoSuchElementException(); 5874 5875 if (emptyElementCount == 0) { 5876 String n = nextElement; 5877 nextElement = null; 5878 return n; 5879 } else { 5880 emptyElementCount--; 5881 return ""; 5882 } 5883 } 5884 5885 public boolean hasNext() { 5886 if (matcher == null) { 5887 matcher = matcher(input); 5888 // If the input is an empty string then the result can only be a 5889 // stream of the input. Induce that by setting the empty 5890 // element count to 1 5891 emptyElementCount = input.length() == 0 ? 1 : 0; 5892 } 5893 if (nextElement != null || emptyElementCount > 0) 5894 return true; 5895 5896 if (current == input.length()) 5897 return false; 5898 5899 // Consume the next matching element 5900 // Count sequence of matching empty elements 5901 while (matcher.find()) { 5902 nextElement = input.subSequence(current, matcher.start()).toString(); 5903 current = matcher.end(); 5904 if (!nextElement.isEmpty()) { 5905 return true; 5906 } else if (current > 0) { // no empty leading substring for zero-width 5907 // match at the beginning of the input 5908 emptyElementCount++; 5909 } 5910 } 5911 5912 // Consume last matching element 5913 nextElement = input.subSequence(current, input.length()).toString(); 5914 current = input.length(); 5915 if (!nextElement.isEmpty()) { 5916 return true; 5917 } else { 5918 // Ignore a terminal sequence of matching empty elements 5919 emptyElementCount = 0; 5920 nextElement = null; 5921 return false; 5922 } 5923 } 5924 } 5925 return StreamSupport.stream(Spliterators.spliteratorUnknownSize( 5926 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false); 5927 } 5928 }