1 /* 2 * Copyright (c) 1999, 2019, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. Oracle designates this 8 * particular file as subject to the "Classpath" exception as provided 9 * by Oracle in the LICENSE file that accompanied this code. 10 * 11 * This code is distributed in the hope that it will be useful, but WITHOUT 12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 14 * version 2 for more details (a copy is included in the LICENSE file that 15 * accompanied this code). 16 * 17 * You should have received a copy of the GNU General Public License version 18 * 2 along with this work; if not, write to the Free Software Foundation, 19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 20 * 21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 22 * or visit www.oracle.com if you need additional information or have any 23 * questions. 24 */ 25 26 package java.util.regex; 27 28 import java.text.Normalizer; 29 import java.text.Normalizer.Form; 30 import java.util.Locale; 31 import java.util.Iterator; 32 import java.util.Map; 33 import java.util.ArrayList; 34 import java.util.HashMap; 35 import java.util.LinkedHashSet; 36 import java.util.List; 37 import java.util.Set; 38 import java.util.Arrays; 39 import java.util.NoSuchElementException; 40 import java.util.Spliterator; 41 import java.util.Spliterators; 42 import java.util.function.Predicate; 43 import java.util.stream.Stream; 44 import java.util.stream.StreamSupport; 45 46 import jdk.internal.util.ArraysSupport; 47 48 /** 49 * A compiled representation of a regular expression. 50 * 51 * <p> A regular expression, specified as a string, must first be compiled into 52 * an instance of this class. The resulting pattern can then be used to create 53 * a {@link Matcher} object that can match arbitrary {@linkplain 54 * java.lang.CharSequence character sequences} against the regular 55 * expression. All of the state involved in performing a match resides in the 56 * matcher, so many matchers can share the same pattern. 57 * 58 * <p> A typical invocation sequence is thus 59 * 60 * <blockquote><pre> 61 * Pattern p = Pattern.{@link #compile compile}("a*b"); 62 * Matcher m = p.{@link #matcher matcher}("aaaaab"); 63 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote> 64 * 65 * <p> A {@link #matches matches} method is defined by this class as a 66 * convenience for when a regular expression is used just once. This method 67 * compiles an expression and matches an input sequence against it in a single 68 * invocation. The statement 69 * 70 * <blockquote><pre> 71 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote> 72 * 73 * is equivalent to the three statements above, though for repeated matches it 74 * is less efficient since it does not allow the compiled pattern to be reused. 75 * 76 * <p> Instances of this class are immutable and are safe for use by multiple 77 * concurrent threads. Instances of the {@link Matcher} class are not safe for 78 * such use. 79 * 80 * 81 * <h2><a id="sum">Summary of regular-expression constructs</a></h2> 82 * 83 * <table class="borderless"> 84 * <caption style="display:none">Regular expression constructs, and what they match</caption> 85 * <thead style="text-align:left"> 86 * <tr> 87 * <th id="construct">Construct</th> 88 * <th id="matches">Matches</th> 89 * </tr> 90 * </thead> 91 * <tbody style="text-align:left"> 92 * 93 * <tr><th colspan="2" style="padding-top:20px" id="characters">Characters</th></tr> 94 * 95 * <tr><th style="vertical-align:top; font-weight: normal" id="x"><i>x</i></th> 96 * <td headers="matches characters x">The character <i>x</i></td></tr> 97 * <tr><th style="vertical-align:top; font-weight: normal" id="backslash">{@code \\}</th> 98 * <td headers="matches characters backslash">The backslash character</td></tr> 99 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_n">{@code \0}<i>n</i></th> 100 * <td headers="matches characters octal_n">The character with octal value {@code 0}<i>n</i> 101 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 102 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nn">{@code \0}<i>nn</i></th> 103 * <td headers="matches characters octal_nn">The character with octal value {@code 0}<i>nn</i> 104 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 105 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nnn">{@code \0}<i>mnn</i></th> 106 * <td headers="matches characters octal_nnn">The character with octal value {@code 0}<i>mnn</i> 107 * (0 {@code <=} <i>m</i> {@code <=} 3, 108 * 0 {@code <=} <i>n</i> {@code <=} 7)</td></tr> 109 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hh">{@code \x}<i>hh</i></th> 110 * <td headers="matches characters hex_hh">The character with hexadecimal value {@code 0x}<i>hh</i></td></tr> 111 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hhhh"><code>\u</code><i>hhhh</i></th> 112 * <td headers="matches characters hex_hhhh">The character with hexadecimal value {@code 0x}<i>hhhh</i></td></tr> 113 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_h_h"><code>\x</code><i>{h...h}</i></th> 114 * <td headers="matches characters hex_h_h">The character with hexadecimal value {@code 0x}<i>h...h</i> 115 * ({@link java.lang.Character#MIN_CODE_POINT Character.MIN_CODE_POINT} 116 * <= {@code 0x}<i>h...h</i> <= 117 * {@link java.lang.Character#MAX_CODE_POINT Character.MAX_CODE_POINT})</td></tr> 118 * <tr><th style="vertical-align:top; font-weight: normal" id="unicode_name"><code>\N{</code><i>name</i><code>}</code></th> 119 * <td headers="matches characters unicode_name">The character with Unicode character name <i>'name'</i></td></tr> 120 * <tr><th style="vertical-align:top; font-weight:normal" id="tab">{@code \t}</th> 121 * <td headers="matches characters tab">The tab character (<code>'\u0009'</code>)</td></tr> 122 * <tr><th style="vertical-align:top; font-weight:normal" id="newline">{@code \n}</th> 123 * <td headers="matches characters newline">The newline (line feed) character (<code>'\u000A'</code>)</td></tr> 124 * <tr><th style="vertical-align:top; font-weight:normal" id="return">{@code \r}</th> 125 * <td headers="matches characters return">The carriage-return character (<code>'\u000D'</code>)</td></tr> 126 * <tr><th style="vertical-align:top; font-weight:normal" id="form_feed">{@code \f}</th> 127 * <td headers="matches characters form_feed">The form-feed character (<code>'\u000C'</code>)</td></tr> 128 * <tr><th style="vertical-align:top; font-weight:normal" id="bell">{@code \a}</th> 129 * <td headers="matches characters bell">The alert (bell) character (<code>'\u0007'</code>)</td></tr> 130 * <tr><th style="vertical-align:top; font-weight:normal" id="escape">{@code \e}</th> 131 * <td headers="matches characters escape">The escape character (<code>'\u001B'</code>)</td></tr> 132 * <tr><th style="vertical-align:top; font-weight:normal" id="ctrl_x">{@code \c}<i>x</i></th> 133 * <td headers="matches characters ctrl_x">The control character corresponding to <i>x</i></td></tr> 134 * 135 * <tr><th colspan="2" style="padding-top:20px" id="classes">Character classes</th></tr> 136 * 137 * <tr><th style="vertical-align:top; font-weight:normal" id="simple">{@code [abc]}</th> 138 * <td headers="matches classes simple">{@code a}, {@code b}, or {@code c} (simple class)</td></tr> 139 * <tr><th style="vertical-align:top; font-weight:normal" id="negation">{@code [^abc]}</th> 140 * <td headers="matches classes negation">Any character except {@code a}, {@code b}, or {@code c} (negation)</td></tr> 141 * <tr><th style="vertical-align:top; font-weight:normal" id="range">{@code [a-zA-Z]}</th> 142 * <td headers="matches classes range">{@code a} through {@code z} 143 * or {@code A} through {@code Z}, inclusive (range)</td></tr> 144 * <tr><th style="vertical-align:top; font-weight:normal" id="union">{@code [a-d[m-p]]}</th> 145 * <td headers="matches classes union">{@code a} through {@code d}, 146 * or {@code m} through {@code p}: {@code [a-dm-p]} (union)</td></tr> 147 * <tr><th style="vertical-align:top; font-weight:normal" id="intersection">{@code [a-z&&[def]]}</th> 148 * <td headers="matches classes intersection">{@code d}, {@code e}, or {@code f} (intersection)</tr> 149 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction1">{@code [a-z&&[^bc]]}</th> 150 * <td headers="matches classes subtraction1">{@code a} through {@code z}, 151 * except for {@code b} and {@code c}: {@code [ad-z]} (subtraction)</td></tr> 152 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction2">{@code [a-z&&[^m-p]]}</th> 153 * <td headers="matches classes subtraction2">{@code a} through {@code z}, 154 * and not {@code m} through {@code p}: {@code [a-lq-z]}(subtraction)</td></tr> 155 * 156 * <tr><th colspan="2" style="padding-top:20px" id="predef">Predefined character classes</th></tr> 157 * 158 * <tr><th style="vertical-align:top; font-weight:normal" id="any">{@code .}</th> 159 * <td headers="matches predef any">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr> 160 * <tr><th style="vertical-align:top; font-weight:normal" id="digit">{@code \d}</th> 161 * <td headers="matches predef digit">A digit: {@code [0-9]}</td></tr> 162 * <tr><th style="vertical-align:top; font-weight:normal" id="non_digit">{@code \D}</th> 163 * <td headers="matches predef non_digit">A non-digit: {@code [^0-9]}</td></tr> 164 * <tr><th style="vertical-align:top; font-weight:normal" id="horiz_white">{@code \h}</th> 165 * <td headers="matches predef horiz_white">A horizontal whitespace character: 166 * <code>[ \t\xA0\u1680\u180e\u2000-\u200a\u202f\u205f\u3000]</code></td></tr> 167 * <tr><th style="vertical-align:top; font-weight:normal" id="non_horiz_white">{@code \H}</th> 168 * <td headers="matches predef non_horiz_white">A non-horizontal whitespace character: {@code [^\h]}</td></tr> 169 * <tr><th style="vertical-align:top; font-weight:normal" id="white">{@code \s}</th> 170 * <td headers="matches predef white">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 171 * <tr><th style="vertical-align:top; font-weight:normal" id="non_white">{@code \S}</th> 172 * <td headers="matches predef non_white">A non-whitespace character: {@code [^\s]}</td></tr> 173 * <tr><th style="vertical-align:top; font-weight:normal" id="vert_white">{@code \v}</th> 174 * <td headers="matches predef vert_white">A vertical whitespace character: <code>[\n\x0B\f\r\x85\u2028\u2029]</code> 175 * </td></tr> 176 * <tr><th style="vertical-align:top; font-weight:normal" id="non_vert_white">{@code \V}</th> 177 * <td headers="matches predef non_vert_white">A non-vertical whitespace character: {@code [^\v]}</td></tr> 178 * <tr><th style="vertical-align:top; font-weight:normal" id="word">{@code \w}</th> 179 * <td headers="matches predef word">A word character: {@code [a-zA-Z_0-9]}</td></tr> 180 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word">{@code \W}</th> 181 * <td headers="matches predef non_word">A non-word character: {@code [^\w]}</td></tr> 182 * 183 * <tr><th colspan="2" style="padding-top:20px" id="posix"><b>POSIX character classes (US-ASCII only)</b></th></tr> 184 * 185 * <tr><th style="vertical-align:top; font-weight:normal" id="Lower">{@code \p{Lower}}</th> 186 * <td headers="matches posix Lower">A lower-case alphabetic character: {@code [a-z]}</td></tr> 187 * <tr><th style="vertical-align:top; font-weight:normal" id="Upper">{@code \p{Upper}}</th> 188 * <td headers="matches posix Upper">An upper-case alphabetic character:{@code [A-Z]}</td></tr> 189 * <tr><th style="vertical-align:top; font-weight:normal" id="ASCII">{@code \p{ASCII}}</th> 190 * <td headers="matches posix ASCII">All ASCII:{@code [\x00-\x7F]}</td></tr> 191 * <tr><th style="vertical-align:top; font-weight:normal" id="Alpha">{@code \p{Alpha}}</th> 192 * <td headers="matches posix Alpha">An alphabetic character:{@code [\p{Lower}\p{Upper}]}</td></tr> 193 * <tr><th style="vertical-align:top; font-weight:normal" id="Digit">{@code \p{Digit}}</th> 194 * <td headers="matches posix Digit">A decimal digit: {@code [0-9]}</td></tr> 195 * <tr><th style="vertical-align:top; font-weight:normal" id="Alnum">{@code \p{Alnum}}</th> 196 * <td headers="matches posix Alnum">An alphanumeric character:{@code [\p{Alpha}\p{Digit}]}</td></tr> 197 * <tr><th style="vertical-align:top; font-weight:normal" id="Punct">{@code \p{Punct}}</th> 198 * <td headers="matches posix Punct">Punctuation: One of {@code !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~}</td></tr> 199 * <!-- {@code [\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]} 200 * {@code [\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]} --> 201 * <tr><th style="vertical-align:top; font-weight:normal" id="Graph">{@code \p{Graph}}</th> 202 * <td headers="matches posix Graph">A visible character: {@code [\p{Alnum}\p{Punct}]}</td></tr> 203 * <tr><th style="vertical-align:top; font-weight:normal" id="Print">{@code \p{Print}}</th> 204 * <td headers="matches posix Print">A printable character: {@code [\p{Graph}\x20]}</td></tr> 205 * <tr><th style="vertical-align:top; font-weight:normal" id="Blank">{@code \p{Blank}}</th> 206 * <td headers="matches posix Blank">A space or a tab: {@code [ \t]}</td></tr> 207 * <tr><th style="vertical-align:top; font-weight:normal" id="Cntrl">{@code \p{Cntrl}}</th> 208 * <td headers="matches posix Cntrl">A control character: {@code [\x00-\x1F\x7F]}</td></tr> 209 * <tr><th style="vertical-align:top; font-weight:normal" id="XDigit">{@code \p{XDigit}}</th> 210 * <td headers="matches posix XDigit">A hexadecimal digit: {@code [0-9a-fA-F]}</td></tr> 211 * <tr><th style="vertical-align:top; font-weight:normal" id="Space">{@code \p{Space}}</th> 212 * <td headers="matches posix Space">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr> 213 * 214 * <tr><th colspan="2" style="padding-top:20px" id="java">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr> 215 * 216 * <tr><th style="vertical-align:top; font-weight:normal" id="javaLowerCase">{@code \p{javaLowerCase}}</th> 217 * <td headers="matches java javaLowerCase">Equivalent to java.lang.Character.isLowerCase()</td></tr> 218 * <tr><th style="vertical-align:top; font-weight:normal" id="javaUpperCase">{@code \p{javaUpperCase}}</th> 219 * <td headers="matches java javaUpperCase">Equivalent to java.lang.Character.isUpperCase()</td></tr> 220 * <tr><th style="vertical-align:top; font-weight:normal" id="javaWhitespace">{@code \p{javaWhitespace}}</th> 221 * <td headers="matches java javaWhitespace">Equivalent to java.lang.Character.isWhitespace()</td></tr> 222 * <tr><th style="vertical-align:top; font-weight:normal" id="javaMirrored">{@code \p{javaMirrored}}</th> 223 * <td headers="matches java javaMirrored">Equivalent to java.lang.Character.isMirrored()</td></tr> 224 * 225 * <tr><th colspan="2" style="padding-top:20px" id="unicode">Classes for Unicode scripts, blocks, categories and binary properties</th></tr> 226 * 227 * <tr><th style="vertical-align:top; font-weight:normal" id="IsLatin">{@code \p{IsLatin}}</th> 228 * <td headers="matches unicode IsLatin">A Latin script character (<a href="#usc">script</a>)</td></tr> 229 * <tr><th style="vertical-align:top; font-weight:normal" id="InGreek">{@code \p{InGreek}}</th> 230 * <td headers="matches unicode InGreek">A character in the Greek block (<a href="#ubc">block</a>)</td></tr> 231 * <tr><th style="vertical-align:top; font-weight:normal" id="Lu">{@code \p{Lu}}</th> 232 * <td headers="matches unicode Lu">An uppercase letter (<a href="#ucc">category</a>)</td></tr> 233 * <tr><th style="vertical-align:top; font-weight:normal" id="IsAlphabetic">{@code \p{IsAlphabetic}}</th> 234 * <td headers="matches unicode IsAlphabetic">An alphabetic character (<a href="#ubpc">binary property</a>)</td></tr> 235 * <tr><th style="vertical-align:top; font-weight:normal" id="Sc">{@code \p{Sc}}</th> 236 * <td headers="matches unicode Sc">A currency symbol</td></tr> 237 * <tr><th style="vertical-align:top; font-weight:normal" id="not_InGreek">{@code \P{InGreek}}</th> 238 * <td headers="matches unicode not_InGreek">Any character except one in the Greek block (negation)</td></tr> 239 * <tr><th style="vertical-align:top; font-weight:normal" id="not_uppercase">{@code [\p{L}&&[^\p{Lu}]]}</th> 240 * <td headers="matches unicode not_uppercase">Any letter except an uppercase letter (subtraction)</td></tr> 241 * 242 * <tr><th colspan="2" style="padding-top:20px" id="bounds">Boundary matchers</th></tr> 243 * 244 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_line">{@code ^}</th> 245 * <td headers="matches bounds begin_line">The beginning of a line</td></tr> 246 * <tr><th style="vertical-align:top; font-weight:normal" id="end_line">{@code $}</th> 247 * <td headers="matches bounds end_line">The end of a line</td></tr> 248 * <tr><th style="vertical-align:top; font-weight:normal" id="word_boundary">{@code \b}</th> 249 * <td headers="matches bounds word_boundary">A word boundary</td></tr> 250 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_cluster_boundary">{@code \b{g}}</th> 251 * <td headers="matches bounds grapheme_cluster_boundary">A Unicode extended grapheme cluster boundary</td></tr> 252 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word_boundary">{@code \B}</th> 253 * <td headers="matches bounds non_word_boundary">A non-word boundary</td></tr> 254 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_input">{@code \A}</th> 255 * <td headers="matches bounds begin_input">The beginning of the input</td></tr> 256 * <tr><th style="vertical-align:top; font-weight:normal" id="end_prev_match">{@code \G}</th> 257 * <td headers="matches bounds end_prev_match">The end of the previous match</td></tr> 258 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input_except_term">{@code \Z}</th> 259 * <td headers="matches bounds end_input_except_term">The end of the input but for the final 260 * <a href="#lt">terminator</a>, if any</td></tr> 261 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input">{@code \z}</th> 262 * <td headers="matches bounds end_input">The end of the input</td></tr> 263 * 264 * <tr><th colspan="2" style="padding-top:20px" id="linebreak">Linebreak matcher</th></tr> 265 * 266 * <tr><th style="vertical-align:top; font-weight:normal" id="any_unicode_linebreak">{@code \R}</th> 267 * <td headers="matches linebreak any_unicode_linebreak">Any Unicode linebreak sequence, is equivalent to 268 * <code>\u000D\u000A|[\u000A\u000B\u000C\u000D\u0085\u2028\u2029] 269 * </code></td></tr> 270 * 271 * <tr><th colspan="2" style="padding-top:20px" id="grapheme">Unicode Extended Grapheme matcher</th></tr> 272 * 273 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_any">{@code \X}</th> 274 * <td headers="matches grapheme grapheme_any">Any Unicode extended grapheme cluster</td></tr> 275 * 276 * <tr><th colspan="2" style="padding-top:20px" id="greedy">Greedy quantifiers</th></tr> 277 * 278 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_once_or_not"><i>X</i>{@code ?}</th> 279 * <td headers="matches greedy greedy_once_or_not"><i>X</i>, once or not at all</td></tr> 280 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_zero_or_more"><i>X</i>{@code *}</th> 281 * <td headers="matches greedy greedy_zero_or_more"><i>X</i>, zero or more times</td></tr> 282 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_one_or_more"><i>X</i>{@code +}</th> 283 * <td headers="matches greedy greedy_one_or_more"><i>X</i>, one or more times</td></tr> 284 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_exactly"><i>X</i><code>{</code><i>n</i><code>}</code></th> 285 * <td headers="matches greedy greedy_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 286 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least"><i>X</i><code>{</code><i>n</i>{@code ,}}</th> 287 * <td headers="matches greedy greedy_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 288 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}</code></th> 289 * <td headers="matches greedy greedy_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 290 * 291 * <tr><th colspan="2" style="padding-top:20px" id="reluc">Reluctant quantifiers</th></tr> 292 * 293 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_once_or_not"><i>X</i>{@code ??}</th> 294 * <td headers="matches reluc reluc_once_or_not"><i>X</i>, once or not at all</td></tr> 295 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_zero_or_more"><i>X</i>{@code *?}</th> 296 * <td headers="matches reluc reluc_zero_or_more"><i>X</i>, zero or more times</td></tr> 297 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_one_or_more"><i>X</i>{@code +?}</th> 298 * <td headers="matches reluc reluc_one_or_more"><i>X</i>, one or more times</td></tr> 299 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_exactly"><i>X</i><code>{</code><i>n</i><code>}?</code></th> 300 * <td headers="matches reluc reluc_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 301 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least"><i>X</i><code>{</code><i>n</i><code>,}?</code></th> 302 * <td headers="matches reluc reluc_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 303 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}?</code></th> 304 * <td headers="matches reluc reluc_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 305 * 306 * <tr><th colspan="2" style="padding-top:20px" id="poss">Possessive quantifiers</th></tr> 307 * 308 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_once_or_not"><i>X</i>{@code ?+}</th> 309 * <td headers="matches poss poss_once_or_not"><i>X</i>, once or not at all</td></tr> 310 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_zero_or_more"><i>X</i>{@code *+}</th> 311 * <td headers="matches poss poss_zero_or_more"><i>X</i>, zero or more times</td></tr> 312 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_one_or_more"><i>X</i>{@code ++}</th> 313 * <td headers="matches poss poss_one_or_more"><i>X</i>, one or more times</td></tr> 314 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_exactly"><i>X</i><code>{</code><i>n</i><code>}+</code></th> 315 * <td headers="matches poss poss_exactly"><i>X</i>, exactly <i>n</i> times</td></tr> 316 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least"><i>X</i><code>{</code><i>n</i><code>,}+</code></th> 317 * <td headers="matches poss poss_at_least"><i>X</i>, at least <i>n</i> times</td></tr> 318 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}+</code></th> 319 * <td headers="matches poss poss_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr> 320 * 321 * <tr><th colspan="2" style="padding-top:20px" id="logical">Logical operators</th></tr> 322 * 323 * <tr><th style="vertical-align:top; font-weight:normal" id="concat"><i>XY</i></th> 324 * <td headers="matches logical concat"><i>X</i> followed by <i>Y</i></td></tr> 325 * <tr><th style="vertical-align:top; font-weight:normal" id="alternate"><i>X</i>{@code |}<i>Y</i></th> 326 * <td headers="matches logical alternate">Either <i>X</i> or <i>Y</i></td></tr> 327 * <tr><th style="vertical-align:top; font-weight:normal" id="group">{@code (}<i>X</i>{@code )}</th> 328 * <td headers="matches logical group">X, as a <a href="#cg">capturing group</a></td></tr> 329 * 330 * <tr><th colspan="2" style="padding-top:20px" id="backref">Back references</th></tr> 331 * 332 * <tr><th style="vertical-align:top; font-weight:normal" id="back_nth">{@code \}<i>n</i></th> 333 * <td headers="matches backref back_nth">Whatever the <i>n</i><sup>th</sup> 334 * <a href="#cg">capturing group</a> matched</td></tr> 335 * <tr><th style="vertical-align:top; font-weight:normal" id="back_named">{@code \}<i>k</i><<i>name</i>></th> 336 * <td headers="matches backref back_named">Whatever the 337 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr> 338 * 339 * <tr><th colspan="2" style="padding-top:20px" id="quote">Quotation</th></tr> 340 * 341 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_follow">{@code \}</th> 342 * <td headers="matches quote quote_follow">Nothing, but quotes the following character</td></tr> 343 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_begin">{@code \Q}</th> 344 * <td headers="matches quote quote_begin">Nothing, but quotes all characters until {@code \E}</td></tr> 345 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_end">{@code \E}</th> 346 * <td headers="matches quote quote_end">Nothing, but ends quoting started by {@code \Q}</td></tr> 347 * <!-- Metachars: !$()*+.<>?[\]^{|} --> 348 * 349 * <tr><th colspan="2" style="padding-top:20px" id="special">Special constructs (named-capturing and non-capturing)</th></tr> 350 * 351 * <tr><th style="vertical-align:top; font-weight:normal" id="named_group"><code>(?<<a href="#groupname">name</a>></code><i>X</i>{@code )}</th> 352 * <td headers="matches special named_group"><i>X</i>, as a named-capturing group</td></tr> 353 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group">{@code (?:}<i>X</i>{@code )}</th> 354 * <td headers="matches special non_capture_group"><i>X</i>, as a non-capturing group</td></tr> 355 * <tr><th style="vertical-align:top; font-weight:normal" id="flags"><code>(?idmsuxU-idmsuxU) </code></th> 356 * <td headers="matches special flags">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a> 357 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> 358 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a> 359 * on - off</td></tr> 360 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group_flags"><code>(?idmsux-idmsux:</code><i>X</i>{@code )} </th> 361 * <td headers="matches special non_capture_group_flags"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the 362 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a> 363 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a > 364 * <a href="#COMMENTS">x</a> on - off</td></tr> 365 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookahead">{@code (?=}<i>X</i>{@code )}</th> 366 * <td headers="matches special pos_lookahead"><i>X</i>, via zero-width positive lookahead</td></tr> 367 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookahead">{@code (?!}<i>X</i>{@code )}</th> 368 * <td headers="matches special neg_lookahead"><i>X</i>, via zero-width negative lookahead</td></tr> 369 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookbehind">{@code (?<=}<i>X</i>{@code )}</th> 370 * <td headers="matches special pos_lookbehind"><i>X</i>, via zero-width positive lookbehind</td></tr> 371 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookbehind">{@code (?<!}<i>X</i>{@code )}</th> 372 * <td headers="matches special neg_lookbehind"><i>X</i>, via zero-width negative lookbehind</td></tr> 373 * <tr><th style="vertical-align:top; font-weight:normal" id="indep_non_capture_group">{@code (?>}<i>X</i>{@code )}</th> 374 * <td headers="matches special indep_non_capture_group"><i>X</i>, as an independent, non-capturing group</td></tr> 375 * 376 * </tbody> 377 * </table> 378 * 379 * <hr> 380 * 381 * 382 * <h2><a id="bs">Backslashes, escapes, and quoting</a></h2> 383 * 384 * <p> The backslash character ({@code '\'}) serves to introduce escaped 385 * constructs, as defined in the table above, as well as to quote characters 386 * that otherwise would be interpreted as unescaped constructs. Thus the 387 * expression {@code \\} matches a single backslash and <code>\{</code> matches a 388 * left brace. 389 * 390 * <p> It is an error to use a backslash prior to any alphabetic character that 391 * does not denote an escaped construct; these are reserved for future 392 * extensions to the regular-expression language. A backslash may be used 393 * prior to a non-alphabetic character regardless of whether that character is 394 * part of an unescaped construct. 395 * 396 * <p> Backslashes within string literals in Java source code are interpreted 397 * as required by 398 * <cite>The Java™ Language Specification</cite> 399 * as either Unicode escapes (section 3.3) or other character escapes (section 3.10.6) 400 * It is therefore necessary to double backslashes in string 401 * literals that represent regular expressions to protect them from 402 * interpretation by the Java bytecode compiler. The string literal 403 * <code>"\b"</code>, for example, matches a single backspace character when 404 * interpreted as a regular expression, while {@code "\\b"} matches a 405 * word boundary. The string literal {@code "\(hello\)"} is illegal 406 * and leads to a compile-time error; in order to match the string 407 * {@code (hello)} the string literal {@code "\\(hello\\)"} 408 * must be used. 409 * 410 * <h2><a id="cc">Character Classes</a></h2> 411 * 412 * <p> Character classes may appear within other character classes, and 413 * may be composed by the union operator (implicit) and the intersection 414 * operator ({@code &&}). 415 * The union operator denotes a class that contains every character that is 416 * in at least one of its operand classes. The intersection operator 417 * denotes a class that contains every character that is in both of its 418 * operand classes. 419 * 420 * <p> The precedence of character-class operators is as follows, from 421 * highest to lowest: 422 * 423 * <table class="striped" style="margin-left: 2em;"> 424 * <caption style="display:none">Precedence of character class operators.</caption> 425 * <thead> 426 * <tr><th scope="col">Precedence<th scope="col">Name<th scope="col">Example 427 * </thead> 428 * <tbody> 429 * <tr><th scope="row">1</th> 430 * <td>Literal escape </td> 431 * <td>{@code \x}</td></tr> 432 * <tr><th scope="row">2</th> 433 * <td>Grouping</td> 434 * <td>{@code [...]}</td></tr> 435 * <tr><th scope="row">3</th> 436 * <td>Range</td> 437 * <td>{@code a-z}</td></tr> 438 * <tr><th scope="row">4</th> 439 * <td>Union</td> 440 * <td>{@code [a-e][i-u]}</td></tr> 441 * <tr><th scope="row">5</th> 442 * <td>Intersection</td> 443 * <td>{@code [a-z&&[aeiou]]}</td></tr> 444 * </tbody> 445 * </table> 446 * 447 * <p> Note that a different set of metacharacters are in effect inside 448 * a character class than outside a character class. For instance, the 449 * regular expression {@code .} loses its special meaning inside a 450 * character class, while the expression {@code -} becomes a range 451 * forming metacharacter. 452 * 453 * <h2><a id="lt">Line terminators</a></h2> 454 * 455 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks 456 * the end of a line of the input character sequence. The following are 457 * recognized as line terminators: 458 * 459 * <ul> 460 * 461 * <li> A newline (line feed) character ({@code '\n'}), 462 * 463 * <li> A carriage-return character followed immediately by a newline 464 * character ({@code "\r\n"}), 465 * 466 * <li> A standalone carriage-return character ({@code '\r'}), 467 * 468 * <li> A next-line character (<code>'\u0085'</code>), 469 * 470 * <li> A line-separator character (<code>'\u2028'</code>), or 471 * 472 * <li> A paragraph-separator character (<code>'\u2029'</code>). 473 * 474 * </ul> 475 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators 476 * recognized are newline characters. 477 * 478 * <p> The regular expression {@code .} matches any character except a line 479 * terminator unless the {@link #DOTALL} flag is specified. 480 * 481 * <p> By default, the regular expressions {@code ^} and {@code $} ignore 482 * line terminators and only match at the beginning and the end, respectively, 483 * of the entire input sequence. If {@link #MULTILINE} mode is activated then 484 * {@code ^} matches at the beginning of input and after any line terminator 485 * except at the end of input. When in {@link #MULTILINE} mode {@code $} 486 * matches just before a line terminator or the end of the input sequence. 487 * 488 * <h2><a id="cg">Groups and capturing</a></h2> 489 * 490 * <h3><a id="gnumber">Group number</a></h3> 491 * <p> Capturing groups are numbered by counting their opening parentheses from 492 * left to right. In the expression {@code ((A)(B(C)))}, for example, there 493 * are four such groups: </p> 494 * 495 * <ol style="margin-left:2em;"> 496 * <li> {@code ((A)(B(C)))} 497 * <li> {@code (A)} 498 * <li> {@code (B(C))} 499 * <li> {@code (C)} 500 * </ol> 501 * 502 * <p> Group zero always stands for the entire expression. 503 * 504 * <p> Capturing groups are so named because, during a match, each subsequence 505 * of the input sequence that matches such a group is saved. The captured 506 * subsequence may be used later in the expression, via a back reference, and 507 * may also be retrieved from the matcher once the match operation is complete. 508 * 509 * <h3><a id="groupname">Group name</a></h3> 510 * <p>A capturing group can also be assigned a "name", a {@code named-capturing group}, 511 * and then be back-referenced later by the "name". Group names are composed of 512 * the following characters. The first character must be a {@code letter}. 513 * 514 * <ul> 515 * <li> The uppercase letters {@code 'A'} through {@code 'Z'} 516 * (<code>'\u0041'</code> through <code>'\u005a'</code>), 517 * <li> The lowercase letters {@code 'a'} through {@code 'z'} 518 * (<code>'\u0061'</code> through <code>'\u007a'</code>), 519 * <li> The digits {@code '0'} through {@code '9'} 520 * (<code>'\u0030'</code> through <code>'\u0039'</code>), 521 * </ul> 522 * 523 * <p> A {@code named-capturing group} is still numbered as described in 524 * <a href="#gnumber">Group number</a>. 525 * 526 * <p> The captured input associated with a group is always the subsequence 527 * that the group most recently matched. If a group is evaluated a second time 528 * because of quantification then its previously-captured value, if any, will 529 * be retained if the second evaluation fails. Matching the string 530 * {@code "aba"} against the expression {@code (a(b)?)+}, for example, leaves 531 * group two set to {@code "b"}. All captured input is discarded at the 532 * beginning of each match. 533 * 534 * <p> Groups beginning with {@code (?} are either pure, <i>non-capturing</i> groups 535 * that do not capture text and do not count towards the group total, or 536 * <i>named-capturing</i> group. 537 * 538 * <h2> Unicode support </h2> 539 * 540 * <p> This class is in conformance with Level 1 of <a 541 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 542 * Standard #18: Unicode Regular Expression</i></a>, plus RL2.1 543 * Canonical Equivalents. 544 * <p> 545 * <b>Unicode escape sequences</b> such as <code>\u2014</code> in Java source code 546 * are processed as described in section 3.3 of 547 * <cite>The Java™ Language Specification</cite>. 548 * Such escape sequences are also implemented directly by the regular-expression 549 * parser so that Unicode escapes can be used in expressions that are read from 550 * files or from the keyboard. Thus the strings <code>"\u2014"</code> and 551 * {@code "\\u2014"}, while not equal, compile into the same pattern, which 552 * matches the character with hexadecimal value {@code 0x2014}. 553 * <p> 554 * A Unicode character can also be represented by using its <b>Hex notation</b> 555 * (hexadecimal code point value) directly as described in construct 556 * <code>\x{...}</code>, for example a supplementary character U+2011F can be 557 * specified as <code>\x{2011F}</code>, instead of two consecutive Unicode escape 558 * sequences of the surrogate pair <code>\uD840</code><code>\uDD1F</code>. 559 * <p> 560 * <b>Unicode character names</b> are supported by the named character construct 561 * <code>\N{</code>...<code>}</code>, for example, <code>\N{WHITE SMILING FACE}</code> 562 * specifies character <code>\u263A</code>. The character names supported 563 * by this class are the valid Unicode character names matched by 564 * {@link java.lang.Character#codePointOf(String) Character.codePointOf(name)}. 565 * <p> 566 * <a href="http://www.unicode.org/reports/tr18/#Default_Grapheme_Clusters"> 567 * <b>Unicode extended grapheme clusters</b></a> are supported by the grapheme 568 * cluster matcher {@code \X} and the corresponding boundary matcher {@code \b{g}}. 569 * <p> 570 * Unicode scripts, blocks, categories and binary properties are written with 571 * the {@code \p} and {@code \P} constructs as in Perl. 572 * <code>\p{</code><i>prop</i><code>}</code> matches if 573 * the input has the property <i>prop</i>, while <code>\P{</code><i>prop</i><code>}</code> 574 * does not match if the input has that property. 575 * <p> 576 * Scripts, blocks, categories and binary properties can be used both inside 577 * and outside of a character class. 578 * 579 * <p> 580 * <b><a id="usc">Scripts</a></b> are specified either with the prefix {@code Is}, as in 581 * {@code IsHiragana}, or by using the {@code script} keyword (or its short 582 * form {@code sc}) as in {@code script=Hiragana} or {@code sc=Hiragana}. 583 * <p> 584 * The script names supported by {@code Pattern} are the valid script names 585 * accepted and defined by 586 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}. 587 * 588 * <p> 589 * <b><a id="ubc">Blocks</a></b> are specified with the prefix {@code In}, as in 590 * {@code InMongolian}, or by using the keyword {@code block} (or its short 591 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}. 592 * <p> 593 * The block names supported by {@code Pattern} are the valid block names 594 * accepted and defined by 595 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}. 596 * <p> 597 * 598 * <b><a id="ucc">Categories</a></b> may be specified with the optional prefix {@code Is}: 599 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode 600 * letters. Same as scripts and blocks, categories can also be specified 601 * by using the keyword {@code general_category} (or its short form 602 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}. 603 * <p> 604 * The supported categories are those of 605 * <a href="http://www.unicode.org/unicode/standard/standard.html"> 606 * <i>The Unicode Standard</i></a> in the version specified by the 607 * {@link java.lang.Character Character} class. The category names are those 608 * defined in the Standard, both normative and informative. 609 * <p> 610 * 611 * <b><a id="ubpc">Binary properties</a></b> are specified with the prefix {@code Is}, as in 612 * {@code IsAlphabetic}. The supported binary properties by {@code Pattern} 613 * are 614 * <ul> 615 * <li> Alphabetic 616 * <li> Ideographic 617 * <li> Letter 618 * <li> Lowercase 619 * <li> Uppercase 620 * <li> Titlecase 621 * <li> Punctuation 622 * <Li> Control 623 * <li> White_Space 624 * <li> Digit 625 * <li> Hex_Digit 626 * <li> Join_Control 627 * <li> Noncharacter_Code_Point 628 * <li> Assigned 629 * </ul> 630 * <p> 631 * The following <b>Predefined Character classes</b> and <b>POSIX character classes</b> 632 * are in conformance with the recommendation of <i>Annex C: Compatibility Properties</i> 633 * of <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Regular Expression 634 * </i></a>, when {@link #UNICODE_CHARACTER_CLASS} flag is specified. 635 * 636 * <table class="striped"> 637 * <caption style="display:none">predefined and posix character classes in Unicode mode</caption> 638 * <thead> 639 * <tr> 640 * <th scope="col" id="predef_classes">Classes</th> 641 * <th scope="col" id="predef_matches">Matches</th> 642 * </tr> 643 * </thead> 644 * <tbody> 645 * <tr><th scope="row">{@code \p{Lower}}</th> 646 * <td>A lowercase character:{@code \p{IsLowercase}}</td></tr> 647 * <tr><th scope="row">{@code \p{Upper}}</th> 648 * <td>An uppercase character:{@code \p{IsUppercase}}</td></tr> 649 * <tr><th scope="row">{@code \p{ASCII}}</th> 650 * <td>All ASCII:{@code [\x00-\x7F]}</td></tr> 651 * <tr><th scope="row">{@code \p{Alpha}}</th> 652 * <td>An alphabetic character:{@code \p{IsAlphabetic}}</td></tr> 653 * <tr><th scope="row">{@code \p{Digit}}</th> 654 * <td>A decimal digit character:{@code \p{IsDigit}}</td></tr> 655 * <tr><th scope="row">{@code \p{Alnum}}</th> 656 * <td>An alphanumeric character:{@code [\p{IsAlphabetic}\p{IsDigit}]}</td></tr> 657 * <tr><th scope="row">{@code \p{Punct}}</th> 658 * <td>A punctuation character:{@code \p{IsPunctuation}}</td></tr> 659 * <tr><th scope="row">{@code \p{Graph}}</th> 660 * <td>A visible character: {@code [^\p{IsWhite_Space}\p{gc=Cc}\p{gc=Cs}\p{gc=Cn}]}</td></tr> 661 * <tr><th scope="row">{@code \p{Print}}</th> 662 * <td>A printable character: {@code [\p{Graph}\p{Blank}&&[^\p{Cntrl}]]}</td></tr> 663 * <tr><th scope="row">{@code \p{Blank}}</th> 664 * <td>A space or a tab: {@code [\p{IsWhite_Space}&&[^\p{gc=Zl}\p{gc=Zp}\x0a\x0b\x0c\x0d\x85]]}</td></tr> 665 * <tr><th scope="row">{@code \p{Cntrl}}</th> 666 * <td>A control character: {@code \p{gc=Cc}}</td></tr> 667 * <tr><th scope="row">{@code \p{XDigit}}</th> 668 * <td>A hexadecimal digit: {@code [\p{gc=Nd}\p{IsHex_Digit}]}</td></tr> 669 * <tr><th scope="row">{@code \p{Space}}</th> 670 * <td>A whitespace character:{@code \p{IsWhite_Space}}</td></tr> 671 * <tr><th scope="row">{@code \d}</th> 672 * <td>A digit: {@code \p{IsDigit}}</td></tr> 673 * <tr><th scope="row">{@code \D}</th> 674 * <td>A non-digit: {@code [^\d]}</td></tr> 675 * <tr><th scope="row">{@code \s}</th> 676 * <td>A whitespace character: {@code \p{IsWhite_Space}}</td></tr> 677 * <tr><th scope="row">{@code \S}</th> 678 * <td>A non-whitespace character: {@code [^\s]}</td></tr> 679 * <tr><th scope="row">{@code \w}</th> 680 * <td>A word character: {@code [\p{Alpha}\p{gc=Mn}\p{gc=Me}\p{gc=Mc}\p{Digit}\p{gc=Pc}\p{IsJoin_Control}]}</td></tr> 681 * <tr><th scope="row">{@code \W}</th> 682 * <td>A non-word character: {@code [^\w]}</td></tr> 683 * </tbody> 684 * </table> 685 * <p> 686 * <a id="jcc"> 687 * Categories that behave like the java.lang.Character 688 * boolean is<i>methodname</i> methods (except for the deprecated ones) are 689 * available through the same <code>\p{</code><i>prop</i><code>}</code> syntax where 690 * the specified property has the name <code>java<i>methodname</i></code></a>. 691 * 692 * <h2> Comparison to Perl 5 </h2> 693 * 694 * <p>The {@code Pattern} engine performs traditional NFA-based matching 695 * with ordered alternation as occurs in Perl 5. 696 * 697 * <p> Perl constructs not supported by this class: </p> 698 * 699 * <ul> 700 * <li><p> The backreference constructs, <code>\g{</code><i>n</i><code>}</code> for 701 * the <i>n</i><sup>th</sup><a href="#cg">capturing group</a> and 702 * <code>\g{</code><i>name</i><code>}</code> for 703 * <a href="#groupname">named-capturing group</a>. 704 * </p></li> 705 * 706 * <li><p> The conditional constructs 707 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code )} and 708 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code |}<i>Y</i>{@code )}, 709 * </p></li> 710 * 711 * <li><p> The embedded code constructs <code>(?{</code><i>code</i><code>})</code> 712 * and <code>(??{</code><i>code</i><code>})</code>,</p></li> 713 * 714 * <li><p> The embedded comment syntax {@code (?#comment)}, and </p></li> 715 * 716 * <li><p> The preprocessing operations {@code \l} <code>\u</code>, 717 * {@code \L}, and {@code \U}. </p></li> 718 * 719 * </ul> 720 * 721 * <p> Constructs supported by this class but not by Perl: </p> 722 * 723 * <ul> 724 * 725 * <li><p> Character-class union and intersection as described 726 * <a href="#cc">above</a>.</p></li> 727 * 728 * </ul> 729 * 730 * <p> Notable differences from Perl: </p> 731 * 732 * <ul> 733 * 734 * <li><p> In Perl, {@code \1} through {@code \9} are always interpreted 735 * as back references; a backslash-escaped number greater than {@code 9} is 736 * treated as a back reference if at least that many subexpressions exist, 737 * otherwise it is interpreted, if possible, as an octal escape. In this 738 * class octal escapes must always begin with a zero. In this class, 739 * {@code \1} through {@code \9} are always interpreted as back 740 * references, and a larger number is accepted as a back reference if at 741 * least that many subexpressions exist at that point in the regular 742 * expression, otherwise the parser will drop digits until the number is 743 * smaller or equal to the existing number of groups or it is one digit. 744 * </p></li> 745 * 746 * <li><p> Perl uses the {@code g} flag to request a match that resumes 747 * where the last match left off. This functionality is provided implicitly 748 * by the {@link Matcher} class: Repeated invocations of the {@link 749 * Matcher#find find} method will resume where the last match left off, 750 * unless the matcher is reset. </p></li> 751 * 752 * <li><p> In Perl, embedded flags at the top level of an expression affect 753 * the whole expression. In this class, embedded flags always take effect 754 * at the point at which they appear, whether they are at the top level or 755 * within a group; in the latter case, flags are restored at the end of the 756 * group just as in Perl. </p></li> 757 * 758 * </ul> 759 * 760 * 761 * <p> For a more precise description of the behavior of regular expression 762 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/"> 763 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl, 764 * O'Reilly and Associates, 2006.</a> 765 * </p> 766 * 767 * @see java.lang.String#split(String, int) 768 * @see java.lang.String#split(String) 769 * 770 * @author Mike McCloskey 771 * @author Mark Reinhold 772 * @author JSR-51 Expert Group 773 * @since 1.4 774 * @spec JSR-51 775 */ 776 777 public final class Pattern 778 implements java.io.Serializable 779 { 780 781 /** 782 * Regular expression modifier values. Instead of being passed as 783 * arguments, they can also be passed as inline modifiers. 784 * For example, the following statements have the same effect. 785 * <pre> 786 * Pattern p1 = Pattern.compile("abc", Pattern.CASE_INSENSITIVE|Pattern.MULTILINE); 787 * Pattern p2 = Pattern.compile("(?im)abc", 0); 788 * </pre> 789 */ 790 791 /** 792 * Enables Unix lines mode. 793 * 794 * <p> In this mode, only the {@code '\n'} line terminator is recognized 795 * in the behavior of {@code .}, {@code ^}, and {@code $}. 796 * 797 * <p> Unix lines mode can also be enabled via the embedded flag 798 * expression {@code (?d)}. 799 */ 800 public static final int UNIX_LINES = 0x01; 801 802 /** 803 * Enables case-insensitive matching. 804 * 805 * <p> By default, case-insensitive matching assumes that only characters 806 * in the US-ASCII charset are being matched. Unicode-aware 807 * case-insensitive matching can be enabled by specifying the {@link 808 * #UNICODE_CASE} flag in conjunction with this flag. 809 * 810 * <p> Case-insensitive matching can also be enabled via the embedded flag 811 * expression {@code (?i)}. 812 * 813 * <p> Specifying this flag may impose a slight performance penalty. </p> 814 */ 815 public static final int CASE_INSENSITIVE = 0x02; 816 817 /** 818 * Permits whitespace and comments in pattern. 819 * 820 * <p> In this mode, whitespace is ignored, and embedded comments starting 821 * with {@code #} are ignored until the end of a line. 822 * 823 * <p> Comments mode can also be enabled via the embedded flag 824 * expression {@code (?x)}. 825 */ 826 public static final int COMMENTS = 0x04; 827 828 /** 829 * Enables multiline mode. 830 * 831 * <p> In multiline mode the expressions {@code ^} and {@code $} match 832 * just after or just before, respectively, a line terminator or the end of 833 * the input sequence. By default these expressions only match at the 834 * beginning and the end of the entire input sequence. 835 * 836 * <p> Multiline mode can also be enabled via the embedded flag 837 * expression {@code (?m)}. </p> 838 */ 839 public static final int MULTILINE = 0x08; 840 841 /** 842 * Enables literal parsing of the pattern. 843 * 844 * <p> When this flag is specified then the input string that specifies 845 * the pattern is treated as a sequence of literal characters. 846 * Metacharacters or escape sequences in the input sequence will be 847 * given no special meaning. 848 * 849 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on 850 * matching when used in conjunction with this flag. The other flags 851 * become superfluous. 852 * 853 * <p> There is no embedded flag character for enabling literal parsing. 854 * @since 1.5 855 */ 856 public static final int LITERAL = 0x10; 857 858 /** 859 * Enables dotall mode. 860 * 861 * <p> In dotall mode, the expression {@code .} matches any character, 862 * including a line terminator. By default this expression does not match 863 * line terminators. 864 * 865 * <p> Dotall mode can also be enabled via the embedded flag 866 * expression {@code (?s)}. (The {@code s} is a mnemonic for 867 * "single-line" mode, which is what this is called in Perl.) </p> 868 */ 869 public static final int DOTALL = 0x20; 870 871 /** 872 * Enables Unicode-aware case folding. 873 * 874 * <p> When this flag is specified then case-insensitive matching, when 875 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner 876 * consistent with the Unicode Standard. By default, case-insensitive 877 * matching assumes that only characters in the US-ASCII charset are being 878 * matched. 879 * 880 * <p> Unicode-aware case folding can also be enabled via the embedded flag 881 * expression {@code (?u)}. 882 * 883 * <p> Specifying this flag may impose a performance penalty. </p> 884 */ 885 public static final int UNICODE_CASE = 0x40; 886 887 /** 888 * Enables canonical equivalence. 889 * 890 * <p> When this flag is specified then two characters will be considered 891 * to match if, and only if, their full canonical decompositions match. 892 * The expression <code>"a\u030A"</code>, for example, will match the 893 * string <code>"\u00E5"</code> when this flag is specified. By default, 894 * matching does not take canonical equivalence into account. 895 * 896 * <p> There is no embedded flag character for enabling canonical 897 * equivalence. 898 * 899 * <p> Specifying this flag may impose a performance penalty. </p> 900 */ 901 public static final int CANON_EQ = 0x80; 902 903 /** 904 * Enables the Unicode version of <i>Predefined character classes</i> and 905 * <i>POSIX character classes</i>. 906 * 907 * <p> When this flag is specified then the (US-ASCII only) 908 * <i>Predefined character classes</i> and <i>POSIX character classes</i> 909 * are in conformance with 910 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical 911 * Standard #18: Unicode Regular Expression</i></a> 912 * <i>Annex C: Compatibility Properties</i>. 913 * <p> 914 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded 915 * flag expression {@code (?U)}. 916 * <p> 917 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case 918 * folding. 919 * <p> 920 * Specifying this flag may impose a performance penalty. </p> 921 * @since 1.7 922 */ 923 public static final int UNICODE_CHARACTER_CLASS = 0x100; 924 925 /** 926 * Contains all possible flags for compile(regex, flags). 927 */ 928 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE | 929 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL | 930 UNICODE_CHARACTER_CLASS | COMMENTS; 931 932 /* Pattern has only two serialized components: The pattern string 933 * and the flags, which are all that is needed to recompile the pattern 934 * when it is deserialized. 935 */ 936 937 /** use serialVersionUID from Merlin b59 for interoperability */ 938 private static final long serialVersionUID = 5073258162644648461L; 939 940 /** 941 * The original regular-expression pattern string. 942 * 943 * @serial 944 */ 945 private String pattern; 946 947 /** 948 * The original pattern flags. 949 * 950 * @serial 951 */ 952 private int flags; 953 954 /** 955 * The temporary pattern flags used during compiling. The flags might be turn 956 * on and off by embedded flag. 957 */ 958 private transient int flags0; 959 960 /** 961 * Boolean indicating this Pattern is compiled; this is necessary in order 962 * to lazily compile deserialized Patterns. 963 */ 964 private transient volatile boolean compiled; 965 966 /** 967 * The normalized pattern string. 968 */ 969 private transient String normalizedPattern; 970 971 /** 972 * The starting point of state machine for the find operation. This allows 973 * a match to start anywhere in the input. 974 */ 975 transient Node root; 976 977 /** 978 * The root of object tree for a match operation. The pattern is matched 979 * at the beginning. This may include a find that uses BnM or a First 980 * node. 981 */ 982 transient Node matchRoot; 983 984 /** 985 * Temporary storage used by parsing pattern slice. 986 */ 987 transient int[] buffer; 988 989 /** 990 * A temporary storage used for predicate for double return. 991 */ 992 transient CharPredicate predicate; 993 994 /** 995 * Map the "name" of the "named capturing group" to its group id 996 * node. 997 */ 998 transient volatile Map<String, Integer> namedGroups; 999 1000 /** 1001 * Temporary storage used while parsing group references. 1002 */ 1003 transient GroupHead[] groupNodes; 1004 1005 /** 1006 * Temporary storage used to store the top level closure nodes. 1007 */ 1008 transient List<Node> topClosureNodes; 1009 1010 /** 1011 * The number of top greedy closure nodes in this Pattern. Used by 1012 * matchers to allocate storage needed for a IntHashSet to keep the 1013 * beginning pos {@code i} of all failed match. 1014 */ 1015 transient int localTCNCount; 1016 1017 /* 1018 * Turn off the stop-exponential-backtracking optimization if there 1019 * is a group ref in the pattern. 1020 */ 1021 transient boolean hasGroupRef; 1022 1023 /** 1024 * Temporary null terminated code point array used by pattern compiling. 1025 */ 1026 private transient int[] temp; 1027 1028 /** 1029 * The number of capturing groups in this Pattern. Used by matchers to 1030 * allocate storage needed to perform a match. 1031 */ 1032 transient int capturingGroupCount; 1033 1034 /** 1035 * The local variable count used by parsing tree. Used by matchers to 1036 * allocate storage needed to perform a match. 1037 */ 1038 transient int localCount; 1039 1040 /** 1041 * Index into the pattern string that keeps track of how much has been 1042 * parsed. 1043 */ 1044 private transient int cursor; 1045 1046 /** 1047 * Holds the length of the pattern string. 1048 */ 1049 private transient int patternLength; 1050 1051 /** 1052 * If the Start node might possibly match supplementary characters. 1053 * It is set to true during compiling if 1054 * (1) There is supplementary char in pattern, or 1055 * (2) There is complement node of a "family" CharProperty 1056 */ 1057 private transient boolean hasSupplementary; 1058 1059 /** 1060 * Compiles the given regular expression into a pattern. 1061 * 1062 * @param regex 1063 * The expression to be compiled 1064 * @return the given regular expression compiled into a pattern 1065 * @throws PatternSyntaxException 1066 * If the expression's syntax is invalid 1067 */ 1068 public static Pattern compile(String regex) { 1069 return new Pattern(regex, 0); 1070 } 1071 1072 /** 1073 * Compiles the given regular expression into a pattern with the given 1074 * flags. 1075 * 1076 * @param regex 1077 * The expression to be compiled 1078 * 1079 * @param flags 1080 * Match flags, a bit mask that may include 1081 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL}, 1082 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES}, 1083 * {@link #LITERAL}, {@link #UNICODE_CHARACTER_CLASS} 1084 * and {@link #COMMENTS} 1085 * 1086 * @return the given regular expression compiled into a pattern with the given flags 1087 * @throws IllegalArgumentException 1088 * If bit values other than those corresponding to the defined 1089 * match flags are set in {@code flags} 1090 * 1091 * @throws PatternSyntaxException 1092 * If the expression's syntax is invalid 1093 */ 1094 public static Pattern compile(String regex, int flags) { 1095 return new Pattern(regex, flags); 1096 } 1097 1098 /** 1099 * Returns the regular expression from which this pattern was compiled. 1100 * 1101 * @return The source of this pattern 1102 */ 1103 public String pattern() { 1104 return pattern; 1105 } 1106 1107 /** 1108 * <p>Returns the string representation of this pattern. This 1109 * is the regular expression from which this pattern was 1110 * compiled.</p> 1111 * 1112 * @return The string representation of this pattern 1113 * @since 1.5 1114 */ 1115 public String toString() { 1116 return pattern; 1117 } 1118 1119 /** 1120 * Creates a matcher that will match the given input against this pattern. 1121 * 1122 * @param input 1123 * The character sequence to be matched 1124 * 1125 * @return A new matcher for this pattern 1126 */ 1127 public Matcher matcher(CharSequence input) { 1128 if (!compiled) { 1129 synchronized(this) { 1130 if (!compiled) 1131 compile(); 1132 } 1133 } 1134 Matcher m = new Matcher(this, input); 1135 return m; 1136 } 1137 1138 /** 1139 * Returns this pattern's match flags. 1140 * 1141 * @return The match flags specified when this pattern was compiled 1142 */ 1143 public int flags() { 1144 return flags0; 1145 } 1146 1147 /** 1148 * Compiles the given regular expression and attempts to match the given 1149 * input against it. 1150 * 1151 * <p> An invocation of this convenience method of the form 1152 * 1153 * <blockquote><pre> 1154 * Pattern.matches(regex, input);</pre></blockquote> 1155 * 1156 * behaves in exactly the same way as the expression 1157 * 1158 * <blockquote><pre> 1159 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote> 1160 * 1161 * <p> If a pattern is to be used multiple times, compiling it once and reusing 1162 * it will be more efficient than invoking this method each time. </p> 1163 * 1164 * @param regex 1165 * The expression to be compiled 1166 * 1167 * @param input 1168 * The character sequence to be matched 1169 * @return whether or not the regular expression matches on the input 1170 * @throws PatternSyntaxException 1171 * If the expression's syntax is invalid 1172 */ 1173 public static boolean matches(String regex, CharSequence input) { 1174 Pattern p = Pattern.compile(regex); 1175 Matcher m = p.matcher(input); 1176 return m.matches(); 1177 } 1178 1179 /** 1180 * Splits the given input sequence around matches of this pattern. 1181 * 1182 * <p> The array returned by this method contains each substring of the 1183 * input sequence that is terminated by another subsequence that matches 1184 * this pattern or is terminated by the end of the input sequence. The 1185 * substrings in the array are in the order in which they occur in the 1186 * input. If this pattern does not match any subsequence of the input then 1187 * the resulting array has just one element, namely the input sequence in 1188 * string form. 1189 * 1190 * <p> When there is a positive-width match at the beginning of the input 1191 * sequence then an empty leading substring is included at the beginning 1192 * of the resulting array. A zero-width match at the beginning however 1193 * never produces such empty leading substring. 1194 * 1195 * <p> The {@code limit} parameter controls the number of times the 1196 * pattern is applied and therefore affects the length of the resulting 1197 * array. 1198 * <ul> 1199 * <li><p> 1200 * If the <i>limit</i> is positive then the pattern will be applied 1201 * at most <i>limit</i> - 1 times, the array's length will be 1202 * no greater than <i>limit</i>, and the array's last entry will contain 1203 * all input beyond the last matched delimiter.</p></li> 1204 * 1205 * <li><p> 1206 * If the <i>limit</i> is zero then the pattern will be applied as 1207 * many times as possible, the array can have any length, and trailing 1208 * empty strings will be discarded.</p></li> 1209 * 1210 * <li><p> 1211 * If the <i>limit</i> is negative then the pattern will be applied 1212 * as many times as possible and the array can have any length.</p></li> 1213 * </ul> 1214 * 1215 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1216 * results with these parameters: 1217 * 1218 * <table class="plain" style="margin-left:2em;"> 1219 * <caption style="display:none">Split example showing regex, limit, and result</caption> 1220 * <thead> 1221 * <tr> 1222 * <th scope="col">Regex</th> 1223 * <th scope="col">Limit</th> 1224 * <th scope="col">Result</th> 1225 * </tr> 1226 * </thead> 1227 * <tbody> 1228 * <tr><th scope="row" rowspan="3" style="font-weight:normal">:</th> 1229 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">2</th> 1230 * <td>{@code { "boo", "and:foo" }}</td></tr> 1231 * <tr><!-- : --> 1232 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1233 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1234 * <tr><!-- : --> 1235 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1236 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1237 * <tr><th scope="row" rowspan="3" style="font-weight:normal">o</th> 1238 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th> 1239 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1240 * <tr><!-- o --> 1241 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th> 1242 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr> 1243 * <tr><!-- o --> 1244 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">0</th> 1245 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1246 * </tbody> 1247 * </table> 1248 * 1249 * @param input 1250 * The character sequence to be split 1251 * 1252 * @param limit 1253 * The result threshold, as described above 1254 * 1255 * @return The array of strings computed by splitting the input 1256 * around matches of this pattern 1257 */ 1258 public String[] split(CharSequence input, int limit) { 1259 int index = 0; 1260 boolean matchLimited = limit > 0; 1261 ArrayList<String> matchList = new ArrayList<>(); 1262 Matcher m = matcher(input); 1263 1264 // Add segments before each match found 1265 while(m.find()) { 1266 if (!matchLimited || matchList.size() < limit - 1) { 1267 if (index == 0 && index == m.start() && m.start() == m.end()) { 1268 // no empty leading substring included for zero-width match 1269 // at the beginning of the input char sequence. 1270 continue; 1271 } 1272 String match = input.subSequence(index, m.start()).toString(); 1273 matchList.add(match); 1274 index = m.end(); 1275 } else if (matchList.size() == limit - 1) { // last one 1276 String match = input.subSequence(index, 1277 input.length()).toString(); 1278 matchList.add(match); 1279 index = m.end(); 1280 } 1281 } 1282 1283 // If no match was found, return this 1284 if (index == 0) 1285 return new String[] {input.toString()}; 1286 1287 // Add remaining segment 1288 if (!matchLimited || matchList.size() < limit) 1289 matchList.add(input.subSequence(index, input.length()).toString()); 1290 1291 // Construct result 1292 int resultSize = matchList.size(); 1293 if (limit == 0) 1294 while (resultSize > 0 && matchList.get(resultSize-1).isEmpty()) 1295 resultSize--; 1296 String[] result = new String[resultSize]; 1297 return matchList.subList(0, resultSize).toArray(result); 1298 } 1299 1300 /** 1301 * Splits the given input sequence around matches of this pattern. 1302 * 1303 * <p> This method works as if by invoking the two-argument {@link 1304 * #split(java.lang.CharSequence, int) split} method with the given input 1305 * sequence and a limit argument of zero. Trailing empty strings are 1306 * therefore not included in the resulting array. </p> 1307 * 1308 * <p> The input {@code "boo:and:foo"}, for example, yields the following 1309 * results with these expressions: 1310 * 1311 * <table class="plain" style="margin-left:2em"> 1312 * <caption style="display:none">Split examples showing regex and result</caption> 1313 * <thead> 1314 * <tr> 1315 * <th scope="col">Regex</th> 1316 * <th scope="col">Result</th> 1317 * </tr> 1318 * </thead> 1319 * <tbody> 1320 * <tr><th scope="row" style="text-weight:normal">:</th> 1321 * <td>{@code { "boo", "and", "foo" }}</td></tr> 1322 * <tr><th scope="row" style="text-weight:normal">o</th> 1323 * <td>{@code { "b", "", ":and:f" }}</td></tr> 1324 * </tbody> 1325 * </table> 1326 * 1327 * 1328 * @param input 1329 * The character sequence to be split 1330 * 1331 * @return The array of strings computed by splitting the input 1332 * around matches of this pattern 1333 */ 1334 public String[] split(CharSequence input) { 1335 return split(input, 0); 1336 } 1337 1338 /** 1339 * Returns a literal pattern {@code String} for the specified 1340 * {@code String}. 1341 * 1342 * <p>This method produces a {@code String} that can be used to 1343 * create a {@code Pattern} that would match the string 1344 * {@code s} as if it were a literal pattern.</p> Metacharacters 1345 * or escape sequences in the input sequence will be given no special 1346 * meaning. 1347 * 1348 * @param s The string to be literalized 1349 * @return A literal string replacement 1350 * @since 1.5 1351 */ 1352 public static String quote(String s) { 1353 int slashEIndex = s.indexOf("\\E"); 1354 if (slashEIndex == -1) 1355 return "\\Q" + s + "\\E"; 1356 1357 int lenHint = s.length(); 1358 lenHint = (lenHint < Integer.MAX_VALUE - 8 - lenHint) ? 1359 (lenHint << 1) : (Integer.MAX_VALUE - 8); 1360 1361 StringBuilder sb = new StringBuilder(lenHint); 1362 sb.append("\\Q"); 1363 int current = 0; 1364 do { 1365 sb.append(s, current, slashEIndex) 1366 .append("\\E\\\\E\\Q"); 1367 current = slashEIndex + 2; 1368 } while ((slashEIndex = s.indexOf("\\E", current)) != -1); 1369 1370 return sb.append(s, current, s.length()) 1371 .append("\\E") 1372 .toString(); 1373 } 1374 1375 /** 1376 * Recompile the Pattern instance from a stream. The original pattern 1377 * string is read in and the object tree is recompiled from it. 1378 */ 1379 private void readObject(java.io.ObjectInputStream s) 1380 throws java.io.IOException, ClassNotFoundException { 1381 1382 // Read in all fields 1383 s.defaultReadObject(); 1384 1385 // reset the flags 1386 flags0 = flags; 1387 1388 // Initialize counts 1389 capturingGroupCount = 1; 1390 localCount = 0; 1391 localTCNCount = 0; 1392 1393 // if length > 0, the Pattern is lazily compiled 1394 if (pattern.isEmpty()) { 1395 root = new Start(lastAccept); 1396 matchRoot = lastAccept; 1397 compiled = true; 1398 } 1399 } 1400 1401 /** 1402 * This private constructor is used to create all Patterns. The pattern 1403 * string and match flags are all that is needed to completely describe 1404 * a Pattern. An empty pattern string results in an object tree with 1405 * only a Start node and a LastNode node. 1406 */ 1407 private Pattern(String p, int f) { 1408 if ((f & ~ALL_FLAGS) != 0) { 1409 throw new IllegalArgumentException("Unknown flag 0x" 1410 + Integer.toHexString(f)); 1411 } 1412 pattern = p; 1413 flags = f; 1414 1415 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present 1416 if ((flags & UNICODE_CHARACTER_CLASS) != 0) 1417 flags |= UNICODE_CASE; 1418 1419 // 'flags' for compiling 1420 flags0 = flags; 1421 1422 // Reset group index count 1423 capturingGroupCount = 1; 1424 localCount = 0; 1425 localTCNCount = 0; 1426 1427 if (!pattern.isEmpty()) { 1428 compile(); 1429 } else { 1430 root = new Start(lastAccept); 1431 matchRoot = lastAccept; 1432 } 1433 } 1434 1435 /** 1436 * The pattern is converted to normalized form ({@link 1437 * java.text.Normalizer.Form#NFC NFC}, canonical decomposition, 1438 * followed by canonical composition for the character class 1439 * part, and {@link java.text.Normalizer.Form#NFD NFD}, 1440 * canonical decomposition for the rest), and then a pure 1441 * group is constructed to match canonical equivalences of the 1442 * characters. 1443 */ 1444 private static String normalize(String pattern) { 1445 int plen = pattern.length(); 1446 StringBuilder pbuf = new StringBuilder(plen); 1447 char last = 0; 1448 int lastStart = 0; 1449 char cc = 0; 1450 for (int i = 0; i < plen;) { 1451 char c = pattern.charAt(i); 1452 if (cc == 0 && // top level 1453 c == '\\' && i + 1 < plen && pattern.charAt(i + 1) == '\\') { 1454 i += 2; last = 0; 1455 continue; 1456 } 1457 if (c == '[' && last != '\\') { 1458 if (cc == 0) { 1459 if (lastStart < i) 1460 normalizeSlice(pattern, lastStart, i, pbuf); 1461 lastStart = i; 1462 } 1463 cc++; 1464 } else if (c == ']' && last != '\\') { 1465 cc--; 1466 if (cc == 0) { 1467 normalizeClazz(pattern, lastStart, i + 1, pbuf); 1468 lastStart = i + 1; 1469 } 1470 } 1471 last = c; 1472 i++; 1473 } 1474 assert (cc == 0); 1475 if (lastStart < plen) 1476 normalizeSlice(pattern, lastStart, plen, pbuf); 1477 return pbuf.toString(); 1478 } 1479 1480 private static void normalizeSlice(String src, int off, int limit, 1481 StringBuilder dst) 1482 { 1483 int len = src.length(); 1484 int off0 = off; 1485 while (off < limit && ASCII.isAscii(src.charAt(off))) { 1486 off++; 1487 } 1488 if (off == limit) { 1489 dst.append(src, off0, limit); 1490 return; 1491 } 1492 off--; 1493 if (off < off0) 1494 off = off0; 1495 else 1496 dst.append(src, off0, off); 1497 while (off < limit) { 1498 int ch0 = src.codePointAt(off); 1499 if (".$|()[]{}^?*+\\".indexOf(ch0) != -1) { 1500 dst.append((char)ch0); 1501 off++; 1502 continue; 1503 } 1504 int j = off + Character.charCount(ch0); 1505 int ch1; 1506 while (j < limit) { 1507 ch1 = src.codePointAt(j); 1508 if (Grapheme.isBoundary(ch0, ch1)) 1509 break; 1510 ch0 = ch1; 1511 j += Character.charCount(ch1); 1512 } 1513 String seq = src.substring(off, j); 1514 String nfd = Normalizer.normalize(seq, Normalizer.Form.NFD); 1515 off = j; 1516 if (nfd.length() > 1) { 1517 ch0 = nfd.codePointAt(0); 1518 ch1 = nfd.codePointAt(Character.charCount(ch0)); 1519 if (Character.getType(ch1) == Character.NON_SPACING_MARK) { 1520 Set<String> altns = new LinkedHashSet<>(); 1521 altns.add(seq); 1522 produceEquivalentAlternation(nfd, altns); 1523 dst.append("(?:"); 1524 altns.forEach( s -> dst.append(s).append('|')); 1525 dst.delete(dst.length() - 1, dst.length()); 1526 dst.append(")"); 1527 continue; 1528 } 1529 } 1530 String nfc = Normalizer.normalize(seq, Normalizer.Form.NFC); 1531 if (!seq.equals(nfc) && !nfd.equals(nfc)) 1532 dst.append("(?:" + seq + "|" + nfd + "|" + nfc + ")"); 1533 else if (!seq.equals(nfd)) 1534 dst.append("(?:" + seq + "|" + nfd + ")"); 1535 else 1536 dst.append(seq); 1537 } 1538 } 1539 1540 private static void normalizeClazz(String src, int off, int limit, 1541 StringBuilder dst) 1542 { 1543 dst.append(Normalizer.normalize(src.substring(off, limit), Form.NFC)); 1544 } 1545 1546 /** 1547 * Given a specific sequence composed of a regular character and 1548 * combining marks that follow it, produce the alternation that will 1549 * match all canonical equivalences of that sequence. 1550 */ 1551 private static void produceEquivalentAlternation(String src, 1552 Set<String> dst) 1553 { 1554 int len = countChars(src, 0, 1); 1555 if (src.length() == len) { 1556 dst.add(src); // source has one character. 1557 return; 1558 } 1559 String base = src.substring(0,len); 1560 String combiningMarks = src.substring(len); 1561 String[] perms = producePermutations(combiningMarks); 1562 // Add combined permutations 1563 for(int x = 0; x < perms.length; x++) { 1564 String next = base + perms[x]; 1565 dst.add(next); 1566 next = composeOneStep(next); 1567 if (next != null) { 1568 produceEquivalentAlternation(next, dst); 1569 } 1570 } 1571 } 1572 1573 /** 1574 * Returns an array of strings that have all the possible 1575 * permutations of the characters in the input string. 1576 * This is used to get a list of all possible orderings 1577 * of a set of combining marks. Note that some of the permutations 1578 * are invalid because of combining class collisions, and these 1579 * possibilities must be removed because they are not canonically 1580 * equivalent. 1581 */ 1582 private static String[] producePermutations(String input) { 1583 if (input.length() == countChars(input, 0, 1)) 1584 return new String[] {input}; 1585 1586 if (input.length() == countChars(input, 0, 2)) { 1587 int c0 = Character.codePointAt(input, 0); 1588 int c1 = Character.codePointAt(input, Character.charCount(c0)); 1589 if (getClass(c1) == getClass(c0)) { 1590 return new String[] {input}; 1591 } 1592 String[] result = new String[2]; 1593 result[0] = input; 1594 StringBuilder sb = new StringBuilder(2); 1595 sb.appendCodePoint(c1); 1596 sb.appendCodePoint(c0); 1597 result[1] = sb.toString(); 1598 return result; 1599 } 1600 1601 int length = 1; 1602 int nCodePoints = countCodePoints(input); 1603 for(int x=1; x<nCodePoints; x++) 1604 length = length * (x+1); 1605 1606 String[] temp = new String[length]; 1607 1608 int combClass[] = new int[nCodePoints]; 1609 for(int x=0, i=0; x<nCodePoints; x++) { 1610 int c = Character.codePointAt(input, i); 1611 combClass[x] = getClass(c); 1612 i += Character.charCount(c); 1613 } 1614 1615 // For each char, take it out and add the permutations 1616 // of the remaining chars 1617 int index = 0; 1618 int len; 1619 // offset maintains the index in code units. 1620 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) { 1621 len = countChars(input, offset, 1); 1622 for(int y=x-1; y>=0; y--) { 1623 if (combClass[y] == combClass[x]) { 1624 continue loop; 1625 } 1626 } 1627 StringBuilder sb = new StringBuilder(input); 1628 String otherChars = sb.delete(offset, offset+len).toString(); 1629 String[] subResult = producePermutations(otherChars); 1630 1631 String prefix = input.substring(offset, offset+len); 1632 for (String sre : subResult) 1633 temp[index++] = prefix + sre; 1634 } 1635 String[] result = new String[index]; 1636 System.arraycopy(temp, 0, result, 0, index); 1637 return result; 1638 } 1639 1640 private static int getClass(int c) { 1641 return sun.text.Normalizer.getCombiningClass(c); 1642 } 1643 1644 /** 1645 * Attempts to compose input by combining the first character 1646 * with the first combining mark following it. Returns a String 1647 * that is the composition of the leading character with its first 1648 * combining mark followed by the remaining combining marks. Returns 1649 * null if the first two characters cannot be further composed. 1650 */ 1651 private static String composeOneStep(String input) { 1652 int len = countChars(input, 0, 2); 1653 String firstTwoCharacters = input.substring(0, len); 1654 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC); 1655 if (result.equals(firstTwoCharacters)) 1656 return null; 1657 else { 1658 String remainder = input.substring(len); 1659 return result + remainder; 1660 } 1661 } 1662 1663 /** 1664 * Preprocess any \Q...\E sequences in `temp', meta-quoting them. 1665 * See the description of `quotemeta' in perlfunc(1). 1666 */ 1667 private void RemoveQEQuoting() { 1668 final int pLen = patternLength; 1669 int i = 0; 1670 while (i < pLen-1) { 1671 if (temp[i] != '\\') 1672 i += 1; 1673 else if (temp[i + 1] != 'Q') 1674 i += 2; 1675 else 1676 break; 1677 } 1678 if (i >= pLen - 1) // No \Q sequence found 1679 return; 1680 int j = i; 1681 i += 2; 1682 int newTempLen; 1683 try { 1684 newTempLen = Math.addExact(j + 2, Math.multiplyExact(3, pLen - i)); 1685 } catch (ArithmeticException ae) { 1686 throw new OutOfMemoryError(); 1687 } 1688 int[] newtemp = new int[newTempLen]; 1689 System.arraycopy(temp, 0, newtemp, 0, j); 1690 1691 boolean inQuote = true; 1692 boolean beginQuote = true; 1693 while (i < pLen) { 1694 int c = temp[i++]; 1695 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) { 1696 newtemp[j++] = c; 1697 } else if (ASCII.isDigit(c)) { 1698 if (beginQuote) { 1699 /* 1700 * A unicode escape \[0xu] could be before this quote, 1701 * and we don't want this numeric char to processed as 1702 * part of the escape. 1703 */ 1704 newtemp[j++] = '\\'; 1705 newtemp[j++] = 'x'; 1706 newtemp[j++] = '3'; 1707 } 1708 newtemp[j++] = c; 1709 } else if (c != '\\') { 1710 if (inQuote) newtemp[j++] = '\\'; 1711 newtemp[j++] = c; 1712 } else if (inQuote) { 1713 if (temp[i] == 'E') { 1714 i++; 1715 inQuote = false; 1716 } else { 1717 newtemp[j++] = '\\'; 1718 newtemp[j++] = '\\'; 1719 } 1720 } else { 1721 if (temp[i] == 'Q') { 1722 i++; 1723 inQuote = true; 1724 beginQuote = true; 1725 continue; 1726 } else { 1727 newtemp[j++] = c; 1728 if (i != pLen) 1729 newtemp[j++] = temp[i++]; 1730 } 1731 } 1732 1733 beginQuote = false; 1734 } 1735 1736 patternLength = j; 1737 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination 1738 } 1739 1740 /** 1741 * Copies regular expression to an int array and invokes the parsing 1742 * of the expression which will create the object tree. 1743 */ 1744 private void compile() { 1745 // Handle canonical equivalences 1746 if (has(CANON_EQ) && !has(LITERAL)) { 1747 normalizedPattern = normalize(pattern); 1748 } else { 1749 normalizedPattern = pattern; 1750 } 1751 patternLength = normalizedPattern.length(); 1752 1753 // Copy pattern to int array for convenience 1754 // Use double zero to terminate pattern 1755 temp = new int[patternLength + 2]; 1756 1757 hasSupplementary = false; 1758 int c, count = 0; 1759 // Convert all chars into code points 1760 for (int x = 0; x < patternLength; x += Character.charCount(c)) { 1761 c = normalizedPattern.codePointAt(x); 1762 if (isSupplementary(c)) { 1763 hasSupplementary = true; 1764 } 1765 temp[count++] = c; 1766 } 1767 1768 patternLength = count; // patternLength now in code points 1769 1770 if (! has(LITERAL)) 1771 RemoveQEQuoting(); 1772 1773 // Allocate all temporary objects here. 1774 buffer = new int[32]; 1775 groupNodes = new GroupHead[10]; 1776 namedGroups = null; 1777 topClosureNodes = new ArrayList<>(10); 1778 1779 if (has(LITERAL)) { 1780 // Literal pattern handling 1781 matchRoot = newSlice(temp, patternLength, hasSupplementary); 1782 matchRoot.next = lastAccept; 1783 } else { 1784 // Start recursive descent parsing 1785 matchRoot = expr(lastAccept); 1786 // Check extra pattern characters 1787 if (patternLength != cursor) { 1788 if (peek() == ')') { 1789 throw error("Unmatched closing ')'"); 1790 } else { 1791 throw error("Unexpected internal error"); 1792 } 1793 } 1794 } 1795 1796 // Peephole optimization 1797 if (matchRoot instanceof Slice) { 1798 root = BnM.optimize(matchRoot); 1799 if (root == matchRoot) { 1800 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1801 } 1802 } else if (matchRoot instanceof Begin || matchRoot instanceof First) { 1803 root = matchRoot; 1804 } else { 1805 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot); 1806 } 1807 1808 // Optimize the greedy Loop to prevent exponential backtracking, IF there 1809 // is no group ref in this pattern. With a non-negative localTCNCount value, 1810 // the greedy type Loop, Curly will skip the backtracking for any starting 1811 // position "i" that failed in the past. 1812 if (!hasGroupRef) { 1813 for (Node node : topClosureNodes) { 1814 if (node instanceof Loop) { 1815 // non-deterministic-greedy-group 1816 ((Loop)node).posIndex = localTCNCount++; 1817 } 1818 } 1819 } 1820 1821 // Release temporary storage 1822 temp = null; 1823 buffer = null; 1824 groupNodes = null; 1825 patternLength = 0; 1826 compiled = true; 1827 topClosureNodes = null; 1828 } 1829 1830 Map<String, Integer> namedGroups() { 1831 Map<String, Integer> groups = namedGroups; 1832 if (groups == null) { 1833 namedGroups = groups = new HashMap<>(2); 1834 } 1835 return groups; 1836 } 1837 1838 /** 1839 * Used to accumulate information about a subtree of the object graph 1840 * so that optimizations can be applied to the subtree. 1841 */ 1842 static final class TreeInfo { 1843 int minLength; 1844 int maxLength; 1845 boolean maxValid; 1846 boolean deterministic; 1847 1848 TreeInfo() { 1849 reset(); 1850 } 1851 void reset() { 1852 minLength = 0; 1853 maxLength = 0; 1854 maxValid = true; 1855 deterministic = true; 1856 } 1857 } 1858 1859 /* 1860 * The following private methods are mainly used to improve the 1861 * readability of the code. In order to let the Java compiler easily 1862 * inline them, we should not put many assertions or error checks in them. 1863 */ 1864 1865 /** 1866 * Indicates whether a particular flag is set or not. 1867 */ 1868 private boolean has(int f) { 1869 return (flags0 & f) != 0; 1870 } 1871 1872 /** 1873 * Match next character, signal error if failed. 1874 */ 1875 private void accept(int ch, String s) { 1876 int testChar = temp[cursor++]; 1877 if (has(COMMENTS)) 1878 testChar = parsePastWhitespace(testChar); 1879 if (ch != testChar) { 1880 throw error(s); 1881 } 1882 } 1883 1884 /** 1885 * Mark the end of pattern with a specific character. 1886 */ 1887 private void mark(int c) { 1888 temp[patternLength] = c; 1889 } 1890 1891 /** 1892 * Peek the next character, and do not advance the cursor. 1893 */ 1894 private int peek() { 1895 int ch = temp[cursor]; 1896 if (has(COMMENTS)) 1897 ch = peekPastWhitespace(ch); 1898 return ch; 1899 } 1900 1901 /** 1902 * Read the next character, and advance the cursor by one. 1903 */ 1904 private int read() { 1905 int ch = temp[cursor++]; 1906 if (has(COMMENTS)) 1907 ch = parsePastWhitespace(ch); 1908 return ch; 1909 } 1910 1911 /** 1912 * Read the next character, and advance the cursor by one, 1913 * ignoring the COMMENTS setting 1914 */ 1915 private int readEscaped() { 1916 int ch = temp[cursor++]; 1917 return ch; 1918 } 1919 1920 /** 1921 * Advance the cursor by one, and peek the next character. 1922 */ 1923 private int next() { 1924 int ch = temp[++cursor]; 1925 if (has(COMMENTS)) 1926 ch = peekPastWhitespace(ch); 1927 return ch; 1928 } 1929 1930 /** 1931 * Advance the cursor by one, and peek the next character, 1932 * ignoring the COMMENTS setting 1933 */ 1934 private int nextEscaped() { 1935 int ch = temp[++cursor]; 1936 return ch; 1937 } 1938 1939 /** 1940 * If in xmode peek past whitespace and comments. 1941 */ 1942 private int peekPastWhitespace(int ch) { 1943 while (ASCII.isSpace(ch) || ch == '#') { 1944 while (ASCII.isSpace(ch)) 1945 ch = temp[++cursor]; 1946 if (ch == '#') { 1947 ch = peekPastLine(); 1948 } 1949 } 1950 return ch; 1951 } 1952 1953 /** 1954 * If in xmode parse past whitespace and comments. 1955 */ 1956 private int parsePastWhitespace(int ch) { 1957 while (ASCII.isSpace(ch) || ch == '#') { 1958 while (ASCII.isSpace(ch)) 1959 ch = temp[cursor++]; 1960 if (ch == '#') 1961 ch = parsePastLine(); 1962 } 1963 return ch; 1964 } 1965 1966 /** 1967 * xmode parse past comment to end of line. 1968 */ 1969 private int parsePastLine() { 1970 int ch = temp[cursor++]; 1971 while (ch != 0 && !isLineSeparator(ch)) 1972 ch = temp[cursor++]; 1973 return ch; 1974 } 1975 1976 /** 1977 * xmode peek past comment to end of line. 1978 */ 1979 private int peekPastLine() { 1980 int ch = temp[++cursor]; 1981 while (ch != 0 && !isLineSeparator(ch)) 1982 ch = temp[++cursor]; 1983 return ch; 1984 } 1985 1986 /** 1987 * Determines if character is a line separator in the current mode 1988 */ 1989 private boolean isLineSeparator(int ch) { 1990 if (has(UNIX_LINES)) { 1991 return ch == '\n'; 1992 } else { 1993 return (ch == '\n' || 1994 ch == '\r' || 1995 (ch|1) == '\u2029' || 1996 ch == '\u0085'); 1997 } 1998 } 1999 2000 /** 2001 * Read the character after the next one, and advance the cursor by two. 2002 */ 2003 private int skip() { 2004 int i = cursor; 2005 int ch = temp[i+1]; 2006 cursor = i + 2; 2007 return ch; 2008 } 2009 2010 /** 2011 * Unread one next character, and retreat cursor by one. 2012 */ 2013 private void unread() { 2014 cursor--; 2015 } 2016 2017 /** 2018 * Internal method used for handling all syntax errors. The pattern is 2019 * displayed with a pointer to aid in locating the syntax error. 2020 */ 2021 private PatternSyntaxException error(String s) { 2022 return new PatternSyntaxException(s, normalizedPattern, cursor - 1); 2023 } 2024 2025 /** 2026 * Determines if there is any supplementary character or unpaired 2027 * surrogate in the specified range. 2028 */ 2029 private boolean findSupplementary(int start, int end) { 2030 for (int i = start; i < end; i++) { 2031 if (isSupplementary(temp[i])) 2032 return true; 2033 } 2034 return false; 2035 } 2036 2037 /** 2038 * Determines if the specified code point is a supplementary 2039 * character or unpaired surrogate. 2040 */ 2041 private static final boolean isSupplementary(int ch) { 2042 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || 2043 Character.isSurrogate((char)ch); 2044 } 2045 2046 /** 2047 * The following methods handle the main parsing. They are sorted 2048 * according to their precedence order, the lowest one first. 2049 */ 2050 2051 /** 2052 * The expression is parsed with branch nodes added for alternations. 2053 * This may be called recursively to parse sub expressions that may 2054 * contain alternations. 2055 */ 2056 private Node expr(Node end) { 2057 Node prev = null; 2058 Node firstTail = null; 2059 Branch branch = null; 2060 Node branchConn = null; 2061 2062 for (;;) { 2063 Node node = sequence(end); 2064 Node nodeTail = root; //double return 2065 if (prev == null) { 2066 prev = node; 2067 firstTail = nodeTail; 2068 } else { 2069 // Branch 2070 if (branchConn == null) { 2071 branchConn = new BranchConn(); 2072 branchConn.next = end; 2073 } 2074 if (node == end) { 2075 // if the node returned from sequence() is "end" 2076 // we have an empty expr, set a null atom into 2077 // the branch to indicate to go "next" directly. 2078 node = null; 2079 } else { 2080 // the "tail.next" of each atom goes to branchConn 2081 nodeTail.next = branchConn; 2082 } 2083 if (prev == branch) { 2084 branch.add(node); 2085 } else { 2086 if (prev == end) { 2087 prev = null; 2088 } else { 2089 // replace the "end" with "branchConn" at its tail.next 2090 // when put the "prev" into the branch as the first atom. 2091 firstTail.next = branchConn; 2092 } 2093 prev = branch = new Branch(prev, node, branchConn); 2094 } 2095 } 2096 if (peek() != '|') { 2097 return prev; 2098 } 2099 next(); 2100 } 2101 } 2102 2103 @SuppressWarnings("fallthrough") 2104 /** 2105 * Parsing of sequences between alternations. 2106 */ 2107 private Node sequence(Node end) { 2108 Node head = null; 2109 Node tail = null; 2110 Node node = null; 2111 LOOP: 2112 for (;;) { 2113 int ch = peek(); 2114 switch (ch) { 2115 case '(': 2116 // Because group handles its own closure, 2117 // we need to treat it differently 2118 node = group0(); 2119 // Check for comment or flag group 2120 if (node == null) 2121 continue; 2122 if (head == null) 2123 head = node; 2124 else 2125 tail.next = node; 2126 // Double return: Tail was returned in root 2127 tail = root; 2128 continue; 2129 case '[': 2130 if (has(CANON_EQ) && !has(LITERAL)) 2131 node = new NFCCharProperty(clazz(true)); 2132 else 2133 node = newCharProperty(clazz(true)); 2134 break; 2135 case '\\': 2136 ch = nextEscaped(); 2137 if (ch == 'p' || ch == 'P') { 2138 boolean oneLetter = true; 2139 boolean comp = (ch == 'P'); 2140 ch = next(); // Consume { if present 2141 if (ch != '{') { 2142 unread(); 2143 } else { 2144 oneLetter = false; 2145 } 2146 // node = newCharProperty(family(oneLetter, comp)); 2147 if (has(CANON_EQ) && !has(LITERAL)) 2148 node = new NFCCharProperty(family(oneLetter, comp)); 2149 else 2150 node = newCharProperty(family(oneLetter, comp)); 2151 } else { 2152 unread(); 2153 node = atom(); 2154 } 2155 break; 2156 case '^': 2157 next(); 2158 if (has(MULTILINE)) { 2159 if (has(UNIX_LINES)) 2160 node = new UnixCaret(); 2161 else 2162 node = new Caret(); 2163 } else { 2164 node = new Begin(); 2165 } 2166 break; 2167 case '$': 2168 next(); 2169 if (has(UNIX_LINES)) 2170 node = new UnixDollar(has(MULTILINE)); 2171 else 2172 node = new Dollar(has(MULTILINE)); 2173 break; 2174 case '.': 2175 next(); 2176 if (has(DOTALL)) { 2177 node = new CharProperty(ALL()); 2178 } else { 2179 if (has(UNIX_LINES)) { 2180 node = new CharProperty(UNIXDOT()); 2181 } else { 2182 node = new CharProperty(DOT()); 2183 } 2184 } 2185 break; 2186 case '|': 2187 case ')': 2188 break LOOP; 2189 case ']': // Now interpreting dangling ] and } as literals 2190 case '}': 2191 node = atom(); 2192 break; 2193 case '?': 2194 case '*': 2195 case '+': 2196 next(); 2197 throw error("Dangling meta character '" + ((char)ch) + "'"); 2198 case 0: 2199 if (cursor >= patternLength) { 2200 break LOOP; 2201 } 2202 // Fall through 2203 default: 2204 node = atom(); 2205 break; 2206 } 2207 2208 node = closure(node); 2209 /* save the top dot-greedy nodes (.*, .+) as well 2210 if (node instanceof GreedyCharProperty && 2211 ((GreedyCharProperty)node).cp instanceof Dot) { 2212 topClosureNodes.add(node); 2213 } 2214 */ 2215 if (head == null) { 2216 head = tail = node; 2217 } else { 2218 tail.next = node; 2219 tail = node; 2220 } 2221 } 2222 if (head == null) { 2223 return end; 2224 } 2225 tail.next = end; 2226 root = tail; //double return 2227 return head; 2228 } 2229 2230 @SuppressWarnings("fallthrough") 2231 /** 2232 * Parse and add a new Single or Slice. 2233 */ 2234 private Node atom() { 2235 int first = 0; 2236 int prev = -1; 2237 boolean hasSupplementary = false; 2238 int ch = peek(); 2239 for (;;) { 2240 switch (ch) { 2241 case '*': 2242 case '+': 2243 case '?': 2244 case '{': 2245 if (first > 1) { 2246 cursor = prev; // Unwind one character 2247 first--; 2248 } 2249 break; 2250 case '$': 2251 case '.': 2252 case '^': 2253 case '(': 2254 case '[': 2255 case '|': 2256 case ')': 2257 break; 2258 case '\\': 2259 ch = nextEscaped(); 2260 if (ch == 'p' || ch == 'P') { // Property 2261 if (first > 0) { // Slice is waiting; handle it first 2262 unread(); 2263 break; 2264 } else { // No slice; just return the family node 2265 boolean comp = (ch == 'P'); 2266 boolean oneLetter = true; 2267 ch = next(); // Consume { if present 2268 if (ch != '{') 2269 unread(); 2270 else 2271 oneLetter = false; 2272 if (has(CANON_EQ) && !has(LITERAL)) 2273 return new NFCCharProperty(family(oneLetter, comp)); 2274 else 2275 return newCharProperty(family(oneLetter, comp)); 2276 } 2277 } 2278 unread(); 2279 prev = cursor; 2280 ch = escape(false, first == 0, false); 2281 if (ch >= 0) { 2282 append(ch, first); 2283 first++; 2284 if (isSupplementary(ch)) { 2285 hasSupplementary = true; 2286 } 2287 ch = peek(); 2288 continue; 2289 } else if (first == 0) { 2290 return root; 2291 } 2292 // Unwind meta escape sequence 2293 cursor = prev; 2294 break; 2295 case 0: 2296 if (cursor >= patternLength) { 2297 break; 2298 } 2299 // Fall through 2300 default: 2301 prev = cursor; 2302 append(ch, first); 2303 first++; 2304 if (isSupplementary(ch)) { 2305 hasSupplementary = true; 2306 } 2307 ch = next(); 2308 continue; 2309 } 2310 break; 2311 } 2312 if (first == 1) { 2313 return newCharProperty(single(buffer[0])); 2314 } else { 2315 return newSlice(buffer, first, hasSupplementary); 2316 } 2317 } 2318 2319 private void append(int ch, int index) { 2320 int oldCapacity = buffer.length; 2321 if (index - oldCapacity >= 0) { 2322 int newCapacity = ArraysSupport.newCapacity(oldCapacity, 2323 1 + index - oldCapacity, oldCapacity); 2324 buffer = Arrays.copyOf(buffer, newCapacity); 2325 } 2326 buffer[index] = ch; 2327 } 2328 2329 /** 2330 * Parses a backref greedily, taking as many numbers as it 2331 * can. The first digit is always treated as a backref, but 2332 * multi digit numbers are only treated as a backref if at 2333 * least that many backrefs exist at this point in the regex. 2334 */ 2335 private Node ref(int refNum) { 2336 boolean done = false; 2337 while(!done) { 2338 int ch = peek(); 2339 switch(ch) { 2340 case '0': 2341 case '1': 2342 case '2': 2343 case '3': 2344 case '4': 2345 case '5': 2346 case '6': 2347 case '7': 2348 case '8': 2349 case '9': 2350 int newRefNum = (refNum * 10) + (ch - '0'); 2351 // Add another number if it doesn't make a group 2352 // that doesn't exist 2353 if (capturingGroupCount - 1 < newRefNum) { 2354 done = true; 2355 break; 2356 } 2357 refNum = newRefNum; 2358 read(); 2359 break; 2360 default: 2361 done = true; 2362 break; 2363 } 2364 } 2365 hasGroupRef = true; 2366 if (has(CASE_INSENSITIVE)) 2367 return new CIBackRef(refNum, has(UNICODE_CASE)); 2368 else 2369 return new BackRef(refNum); 2370 } 2371 2372 /** 2373 * Parses an escape sequence to determine the actual value that needs 2374 * to be matched. 2375 * If -1 is returned and create was true a new object was added to the tree 2376 * to handle the escape sequence. 2377 * If the returned value is greater than zero, it is the value that 2378 * matches the escape sequence. 2379 */ 2380 private int escape(boolean inclass, boolean create, boolean isrange) { 2381 int ch = skip(); 2382 switch (ch) { 2383 case '0': 2384 return o(); 2385 case '1': 2386 case '2': 2387 case '3': 2388 case '4': 2389 case '5': 2390 case '6': 2391 case '7': 2392 case '8': 2393 case '9': 2394 if (inclass) break; 2395 if (create) { 2396 root = ref((ch - '0')); 2397 } 2398 return -1; 2399 case 'A': 2400 if (inclass) break; 2401 if (create) root = new Begin(); 2402 return -1; 2403 case 'B': 2404 if (inclass) break; 2405 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS)); 2406 return -1; 2407 case 'C': 2408 break; 2409 case 'D': 2410 if (create) { 2411 predicate = has(UNICODE_CHARACTER_CLASS) ? 2412 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2413 predicate = predicate.negate(); 2414 if (!inclass) 2415 root = newCharProperty(predicate); 2416 } 2417 return -1; 2418 case 'E': 2419 case 'F': 2420 break; 2421 case 'G': 2422 if (inclass) break; 2423 if (create) root = new LastMatch(); 2424 return -1; 2425 case 'H': 2426 if (create) { 2427 predicate = HorizWS().negate(); 2428 if (!inclass) 2429 root = newCharProperty(predicate); 2430 } 2431 return -1; 2432 case 'I': 2433 case 'J': 2434 case 'K': 2435 case 'L': 2436 case 'M': 2437 break; 2438 case 'N': 2439 return N(); 2440 case 'O': 2441 case 'P': 2442 case 'Q': 2443 break; 2444 case 'R': 2445 if (inclass) break; 2446 if (create) root = new LineEnding(); 2447 return -1; 2448 case 'S': 2449 if (create) { 2450 predicate = has(UNICODE_CHARACTER_CLASS) ? 2451 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2452 predicate = predicate.negate(); 2453 if (!inclass) 2454 root = newCharProperty(predicate); 2455 } 2456 return -1; 2457 case 'T': 2458 case 'U': 2459 break; 2460 case 'V': 2461 if (create) { 2462 predicate = VertWS().negate(); 2463 if (!inclass) 2464 root = newCharProperty(predicate); 2465 } 2466 return -1; 2467 case 'W': 2468 if (create) { 2469 predicate = has(UNICODE_CHARACTER_CLASS) ? 2470 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2471 predicate = predicate.negate(); 2472 if (!inclass) 2473 root = newCharProperty(predicate); 2474 } 2475 return -1; 2476 case 'X': 2477 if (inclass) break; 2478 if (create) { 2479 root = new XGrapheme(); 2480 } 2481 return -1; 2482 case 'Y': 2483 break; 2484 case 'Z': 2485 if (inclass) break; 2486 if (create) { 2487 if (has(UNIX_LINES)) 2488 root = new UnixDollar(false); 2489 else 2490 root = new Dollar(false); 2491 } 2492 return -1; 2493 case 'a': 2494 return '\007'; 2495 case 'b': 2496 if (inclass) break; 2497 if (create) { 2498 if (peek() == '{') { 2499 if (skip() == 'g') { 2500 if (read() == '}') { 2501 root = new GraphemeBound(); 2502 return -1; 2503 } 2504 break; // error missing trailing } 2505 } 2506 unread(); unread(); 2507 } 2508 root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS)); 2509 } 2510 return -1; 2511 case 'c': 2512 return c(); 2513 case 'd': 2514 if (create) { 2515 predicate = has(UNICODE_CHARACTER_CLASS) ? 2516 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT(); 2517 if (!inclass) 2518 root = newCharProperty(predicate); 2519 } 2520 return -1; 2521 case 'e': 2522 return '\033'; 2523 case 'f': 2524 return '\f'; 2525 case 'g': 2526 break; 2527 case 'h': 2528 if (create) { 2529 predicate = HorizWS(); 2530 if (!inclass) 2531 root = newCharProperty(predicate); 2532 } 2533 return -1; 2534 case 'i': 2535 case 'j': 2536 break; 2537 case 'k': 2538 if (inclass) 2539 break; 2540 if (read() != '<') 2541 throw error("\\k is not followed by '<' for named capturing group"); 2542 String name = groupname(read()); 2543 if (!namedGroups().containsKey(name)) 2544 throw error("named capturing group <" + name + "> does not exist"); 2545 if (create) { 2546 hasGroupRef = true; 2547 if (has(CASE_INSENSITIVE)) 2548 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE)); 2549 else 2550 root = new BackRef(namedGroups().get(name)); 2551 } 2552 return -1; 2553 case 'l': 2554 case 'm': 2555 break; 2556 case 'n': 2557 return '\n'; 2558 case 'o': 2559 case 'p': 2560 case 'q': 2561 break; 2562 case 'r': 2563 return '\r'; 2564 case 's': 2565 if (create) { 2566 predicate = has(UNICODE_CHARACTER_CLASS) ? 2567 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE(); 2568 if (!inclass) 2569 root = newCharProperty(predicate); 2570 } 2571 return -1; 2572 case 't': 2573 return '\t'; 2574 case 'u': 2575 return u(); 2576 case 'v': 2577 // '\v' was implemented as VT/0x0B in releases < 1.8 (though 2578 // undocumented). In JDK8 '\v' is specified as a predefined 2579 // character class for all vertical whitespace characters. 2580 // So [-1, root=VertWS node] pair is returned (instead of a 2581 // single 0x0B). This breaks the range if '\v' is used as 2582 // the start or end value, such as [\v-...] or [...-\v], in 2583 // which a single definite value (0x0B) is expected. For 2584 // compatibility concern '\013'/0x0B is returned if isrange. 2585 if (isrange) 2586 return '\013'; 2587 if (create) { 2588 predicate = VertWS(); 2589 if (!inclass) 2590 root = newCharProperty(predicate); 2591 } 2592 return -1; 2593 case 'w': 2594 if (create) { 2595 predicate = has(UNICODE_CHARACTER_CLASS) ? 2596 CharPredicates.WORD() : CharPredicates.ASCII_WORD(); 2597 if (!inclass) 2598 root = newCharProperty(predicate); 2599 } 2600 return -1; 2601 case 'x': 2602 return x(); 2603 case 'y': 2604 break; 2605 case 'z': 2606 if (inclass) break; 2607 if (create) root = new End(); 2608 return -1; 2609 default: 2610 return ch; 2611 } 2612 throw error("Illegal/unsupported escape sequence"); 2613 } 2614 2615 /** 2616 * Parse a character class, and return the node that matches it. 2617 * 2618 * Consumes a ] on the way out if consume is true. Usually consume 2619 * is true except for the case of [abc&&def] where def is a separate 2620 * right hand node with "understood" brackets. 2621 */ 2622 private CharPredicate clazz(boolean consume) { 2623 CharPredicate prev = null; 2624 CharPredicate curr = null; 2625 BitClass bits = new BitClass(); 2626 BmpCharPredicate bitsP = ch -> ch < 256 && bits.bits[ch]; 2627 2628 boolean isNeg = false; 2629 boolean hasBits = false; 2630 int ch = next(); 2631 2632 // Negates if first char in a class, otherwise literal 2633 if (ch == '^' && temp[cursor-1] == '[') { 2634 ch = next(); 2635 isNeg = true; 2636 } 2637 for (;;) { 2638 switch (ch) { 2639 case '[': 2640 curr = clazz(true); 2641 if (prev == null) 2642 prev = curr; 2643 else 2644 prev = prev.union(curr); 2645 ch = peek(); 2646 continue; 2647 case '&': 2648 ch = next(); 2649 if (ch == '&') { 2650 ch = next(); 2651 CharPredicate right = null; 2652 while (ch != ']' && ch != '&') { 2653 if (ch == '[') { 2654 if (right == null) 2655 right = clazz(true); 2656 else 2657 right = right.union(clazz(true)); 2658 } else { // abc&&def 2659 unread(); 2660 right = clazz(false); 2661 } 2662 ch = peek(); 2663 } 2664 if (hasBits) { 2665 // bits used, union has high precedence 2666 if (prev == null) { 2667 prev = curr = bitsP; 2668 } else { 2669 prev = prev.union(bitsP); 2670 } 2671 hasBits = false; 2672 } 2673 if (right != null) 2674 curr = right; 2675 if (prev == null) { 2676 if (right == null) 2677 throw error("Bad class syntax"); 2678 else 2679 prev = right; 2680 } else { 2681 prev = prev.and(curr); 2682 } 2683 } else { 2684 // treat as a literal & 2685 unread(); 2686 break; 2687 } 2688 continue; 2689 case 0: 2690 if (cursor >= patternLength) 2691 throw error("Unclosed character class"); 2692 break; 2693 case ']': 2694 if (prev != null || hasBits) { 2695 if (consume) 2696 next(); 2697 if (prev == null) 2698 prev = bitsP; 2699 else if (hasBits) 2700 prev = prev.union(bitsP); 2701 if (isNeg) 2702 return prev.negate(); 2703 return prev; 2704 } 2705 break; 2706 default: 2707 break; 2708 } 2709 curr = range(bits); 2710 if (curr == null) { // the bits used 2711 hasBits = true; 2712 } else { 2713 if (prev == null) 2714 prev = curr; 2715 else if (prev != curr) 2716 prev = prev.union(curr); 2717 } 2718 ch = peek(); 2719 } 2720 } 2721 2722 private CharPredicate bitsOrSingle(BitClass bits, int ch) { 2723 /* Bits can only handle codepoints in [u+0000-u+00ff] range. 2724 Use "single" node instead of bits when dealing with unicode 2725 case folding for codepoints listed below. 2726 (1)Uppercase out of range: u+00ff, u+00b5 2727 toUpperCase(u+00ff) -> u+0178 2728 toUpperCase(u+00b5) -> u+039c 2729 (2)LatinSmallLetterLongS u+17f 2730 toUpperCase(u+017f) -> u+0053 2731 (3)LatinSmallLetterDotlessI u+131 2732 toUpperCase(u+0131) -> u+0049 2733 (4)LatinCapitalLetterIWithDotAbove u+0130 2734 toLowerCase(u+0130) -> u+0069 2735 (5)KelvinSign u+212a 2736 toLowerCase(u+212a) ==> u+006B 2737 (6)AngstromSign u+212b 2738 toLowerCase(u+212b) ==> u+00e5 2739 */ 2740 if (ch < 256 && 2741 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) && 2742 (ch == 0xff || ch == 0xb5 || 2743 ch == 0x49 || ch == 0x69 || //I and i 2744 ch == 0x53 || ch == 0x73 || //S and s 2745 ch == 0x4b || ch == 0x6b || //K and k 2746 ch == 0xc5 || ch == 0xe5))) { //A+ring 2747 bits.add(ch, flags0); 2748 return null; 2749 } 2750 return single(ch); 2751 } 2752 2753 /** 2754 * Returns a suitably optimized, single character predicate 2755 */ 2756 private CharPredicate single(final int ch) { 2757 if (has(CASE_INSENSITIVE)) { 2758 int lower, upper; 2759 if (has(UNICODE_CASE)) { 2760 upper = Character.toUpperCase(ch); 2761 lower = Character.toLowerCase(upper); 2762 // Unicode case insensitive matches 2763 if (upper != lower) 2764 return SingleU(lower); 2765 } else if (ASCII.isAscii(ch)) { 2766 lower = ASCII.toLower(ch); 2767 upper = ASCII.toUpper(ch); 2768 // Case insensitive matches a given BMP character 2769 if (lower != upper) 2770 return SingleI(lower, upper); 2771 } 2772 } 2773 if (isSupplementary(ch)) 2774 return SingleS(ch); 2775 return Single(ch); // Match a given BMP character 2776 } 2777 2778 /** 2779 * Parse a single character or a character range in a character class 2780 * and return its representative node. 2781 */ 2782 private CharPredicate range(BitClass bits) { 2783 int ch = peek(); 2784 if (ch == '\\') { 2785 ch = nextEscaped(); 2786 if (ch == 'p' || ch == 'P') { // A property 2787 boolean comp = (ch == 'P'); 2788 boolean oneLetter = true; 2789 // Consume { if present 2790 ch = next(); 2791 if (ch != '{') 2792 unread(); 2793 else 2794 oneLetter = false; 2795 return family(oneLetter, comp); 2796 } else { // ordinary escape 2797 boolean isrange = temp[cursor+1] == '-'; 2798 unread(); 2799 ch = escape(true, true, isrange); 2800 if (ch == -1) 2801 return predicate; 2802 } 2803 } else { 2804 next(); 2805 } 2806 if (ch >= 0) { 2807 if (peek() == '-') { 2808 int endRange = temp[cursor+1]; 2809 if (endRange == '[') { 2810 return bitsOrSingle(bits, ch); 2811 } 2812 if (endRange != ']') { 2813 next(); 2814 int m = peek(); 2815 if (m == '\\') { 2816 m = escape(true, false, true); 2817 } else { 2818 next(); 2819 } 2820 if (m < ch) { 2821 throw error("Illegal character range"); 2822 } 2823 if (has(CASE_INSENSITIVE)) { 2824 if (has(UNICODE_CASE)) 2825 return CIRangeU(ch, m); 2826 return CIRange(ch, m); 2827 } else { 2828 return Range(ch, m); 2829 } 2830 } 2831 } 2832 return bitsOrSingle(bits, ch); 2833 } 2834 throw error("Unexpected character '"+((char)ch)+"'"); 2835 } 2836 2837 /** 2838 * Parses a Unicode character family and returns its representative node. 2839 */ 2840 private CharPredicate family(boolean singleLetter, boolean isComplement) { 2841 next(); 2842 String name; 2843 CharPredicate p = null; 2844 2845 if (singleLetter) { 2846 int c = temp[cursor]; 2847 if (!Character.isSupplementaryCodePoint(c)) { 2848 name = String.valueOf((char)c); 2849 } else { 2850 name = new String(temp, cursor, 1); 2851 } 2852 read(); 2853 } else { 2854 int i = cursor; 2855 mark('}'); 2856 while(read() != '}') { 2857 } 2858 mark('\000'); 2859 int j = cursor; 2860 if (j > patternLength) 2861 throw error("Unclosed character family"); 2862 if (i + 1 >= j) 2863 throw error("Empty character family"); 2864 name = new String(temp, i, j-i-1); 2865 } 2866 2867 int i = name.indexOf('='); 2868 if (i != -1) { 2869 // property construct \p{name=value} 2870 String value = name.substring(i + 1); 2871 name = name.substring(0, i).toLowerCase(Locale.ENGLISH); 2872 switch (name) { 2873 case "sc": 2874 case "script": 2875 p = CharPredicates.forUnicodeScript(value); 2876 break; 2877 case "blk": 2878 case "block": 2879 p = CharPredicates.forUnicodeBlock(value); 2880 break; 2881 case "gc": 2882 case "general_category": 2883 p = CharPredicates.forProperty(value); 2884 break; 2885 default: 2886 break; 2887 } 2888 if (p == null) 2889 throw error("Unknown Unicode property {name=<" + name + ">, " 2890 + "value=<" + value + ">}"); 2891 2892 } else { 2893 if (name.startsWith("In")) { 2894 // \p{InBlockName} 2895 p = CharPredicates.forUnicodeBlock(name.substring(2)); 2896 } else if (name.startsWith("Is")) { 2897 // \p{IsGeneralCategory} and \p{IsScriptName} 2898 name = name.substring(2); 2899 p = CharPredicates.forUnicodeProperty(name); 2900 if (p == null) 2901 p = CharPredicates.forProperty(name); 2902 if (p == null) 2903 p = CharPredicates.forUnicodeScript(name); 2904 } else { 2905 if (has(UNICODE_CHARACTER_CLASS)) { 2906 p = CharPredicates.forPOSIXName(name); 2907 } 2908 if (p == null) 2909 p = CharPredicates.forProperty(name); 2910 } 2911 if (p == null) 2912 throw error("Unknown character property name {In/Is" + name + "}"); 2913 } 2914 if (isComplement) { 2915 // it might be too expensive to detect if a complement of 2916 // CharProperty can match "certain" supplementary. So just 2917 // go with StartS. 2918 hasSupplementary = true; 2919 p = p.negate(); 2920 } 2921 return p; 2922 } 2923 2924 private CharProperty newCharProperty(CharPredicate p) { 2925 if (p == null) 2926 return null; 2927 if (p instanceof BmpCharPredicate) 2928 return new BmpCharProperty((BmpCharPredicate)p); 2929 else 2930 return new CharProperty(p); 2931 } 2932 2933 /** 2934 * Parses and returns the name of a "named capturing group", the trailing 2935 * ">" is consumed after parsing. 2936 */ 2937 private String groupname(int ch) { 2938 StringBuilder sb = new StringBuilder(); 2939 if (!ASCII.isAlpha(ch)) 2940 throw error("capturing group name does not start with a Latin letter"); 2941 do { 2942 sb.append((char) ch); 2943 } while (ASCII.isAlnum(ch=read())); 2944 if (ch != '>') 2945 throw error("named capturing group is missing trailing '>'"); 2946 return sb.toString(); 2947 } 2948 2949 /** 2950 * Parses a group and returns the head node of a set of nodes that process 2951 * the group. Sometimes a double return system is used where the tail is 2952 * returned in root. 2953 */ 2954 private Node group0() { 2955 boolean capturingGroup = false; 2956 Node head = null; 2957 Node tail = null; 2958 int save = flags0; 2959 int saveTCNCount = topClosureNodes.size(); 2960 root = null; 2961 int ch = next(); 2962 if (ch == '?') { 2963 ch = skip(); 2964 switch (ch) { 2965 case ':': // (?:xxx) pure group 2966 head = createGroup(true); 2967 tail = root; 2968 head.next = expr(tail); 2969 break; 2970 case '=': // (?=xxx) and (?!xxx) lookahead 2971 case '!': 2972 head = createGroup(true); 2973 tail = root; 2974 head.next = expr(tail); 2975 if (ch == '=') { 2976 head = tail = new Pos(head); 2977 } else { 2978 head = tail = new Neg(head); 2979 } 2980 break; 2981 case '>': // (?>xxx) independent group 2982 head = createGroup(true); 2983 tail = root; 2984 head.next = expr(tail); 2985 head = tail = new Ques(head, Qtype.INDEPENDENT); 2986 break; 2987 case '<': // (?<xxx) look behind 2988 ch = read(); 2989 if (ch != '=' && ch != '!') { 2990 // named captured group 2991 String name = groupname(ch); 2992 if (namedGroups().containsKey(name)) 2993 throw error("Named capturing group <" + name 2994 + "> is already defined"); 2995 capturingGroup = true; 2996 head = createGroup(false); 2997 tail = root; 2998 namedGroups().put(name, capturingGroupCount-1); 2999 head.next = expr(tail); 3000 break; 3001 } 3002 int start = cursor; 3003 head = createGroup(true); 3004 tail = root; 3005 head.next = expr(tail); 3006 tail.next = lookbehindEnd; 3007 TreeInfo info = new TreeInfo(); 3008 head.study(info); 3009 if (info.maxValid == false) { 3010 throw error("Look-behind group does not have " 3011 + "an obvious maximum length"); 3012 } 3013 boolean hasSupplementary = findSupplementary(start, patternLength); 3014 if (ch == '=') { 3015 head = tail = (hasSupplementary ? 3016 new BehindS(head, info.maxLength, 3017 info.minLength) : 3018 new Behind(head, info.maxLength, 3019 info.minLength)); 3020 } else { // if (ch == '!') 3021 head = tail = (hasSupplementary ? 3022 new NotBehindS(head, info.maxLength, 3023 info.minLength) : 3024 new NotBehind(head, info.maxLength, 3025 info.minLength)); 3026 } 3027 // clear all top-closure-nodes inside lookbehind 3028 if (saveTCNCount < topClosureNodes.size()) 3029 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3030 break; 3031 case '$': 3032 case '@': 3033 throw error("Unknown group type"); 3034 default: // (?xxx:) inlined match flags 3035 unread(); 3036 addFlag(); 3037 ch = read(); 3038 if (ch == ')') { 3039 return null; // Inline modifier only 3040 } 3041 if (ch != ':') { 3042 throw error("Unknown inline modifier"); 3043 } 3044 head = createGroup(true); 3045 tail = root; 3046 head.next = expr(tail); 3047 break; 3048 } 3049 } else { // (xxx) a regular group 3050 capturingGroup = true; 3051 head = createGroup(false); 3052 tail = root; 3053 head.next = expr(tail); 3054 } 3055 3056 accept(')', "Unclosed group"); 3057 flags0 = save; 3058 3059 // Check for quantifiers 3060 Node node = closure(head); 3061 if (node == head) { // No closure 3062 root = tail; 3063 return node; // Dual return 3064 } 3065 if (head == tail) { // Zero length assertion 3066 root = node; 3067 return node; // Dual return 3068 } 3069 3070 // have group closure, clear all inner closure nodes from the 3071 // top list (no backtracking stopper optimization for inner 3072 if (saveTCNCount < topClosureNodes.size()) 3073 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear(); 3074 3075 if (node instanceof Ques) { 3076 Ques ques = (Ques) node; 3077 if (ques.type == Qtype.POSSESSIVE) { 3078 root = node; 3079 return node; 3080 } 3081 tail.next = new BranchConn(); 3082 tail = tail.next; 3083 if (ques.type == Qtype.GREEDY) { 3084 head = new Branch(head, null, tail); 3085 } else { // Reluctant quantifier 3086 head = new Branch(null, head, tail); 3087 } 3088 root = tail; 3089 return head; 3090 } else if (node instanceof Curly) { 3091 Curly curly = (Curly) node; 3092 if (curly.type == Qtype.POSSESSIVE) { 3093 root = node; 3094 return node; 3095 } 3096 // Discover if the group is deterministic 3097 TreeInfo info = new TreeInfo(); 3098 if (head.study(info)) { // Deterministic 3099 GroupTail temp = (GroupTail) tail; 3100 head = root = new GroupCurly(head.next, curly.cmin, 3101 curly.cmax, curly.type, 3102 ((GroupTail)tail).localIndex, 3103 ((GroupTail)tail).groupIndex, 3104 capturingGroup); 3105 return head; 3106 } else { // Non-deterministic 3107 int temp = ((GroupHead) head).localIndex; 3108 Loop loop; 3109 if (curly.type == Qtype.GREEDY) { 3110 loop = new Loop(this.localCount, temp); 3111 // add the max_reps greedy to the top-closure-node list 3112 if (curly.cmax == MAX_REPS) 3113 topClosureNodes.add(loop); 3114 } else { // Reluctant Curly 3115 loop = new LazyLoop(this.localCount, temp); 3116 } 3117 Prolog prolog = new Prolog(loop); 3118 this.localCount += 1; 3119 loop.cmin = curly.cmin; 3120 loop.cmax = curly.cmax; 3121 loop.body = head; 3122 tail.next = loop; 3123 root = loop; 3124 return prolog; // Dual return 3125 } 3126 } 3127 throw error("Internal logic error"); 3128 } 3129 3130 /** 3131 * Create group head and tail nodes using double return. If the group is 3132 * created with anonymous true then it is a pure group and should not 3133 * affect group counting. 3134 */ 3135 private Node createGroup(boolean anonymous) { 3136 int localIndex = localCount++; 3137 int groupIndex = 0; 3138 if (!anonymous) 3139 groupIndex = capturingGroupCount++; 3140 GroupHead head = new GroupHead(localIndex); 3141 root = new GroupTail(localIndex, groupIndex); 3142 3143 // for debug/print only, head.match does NOT need the "tail" info 3144 head.tail = (GroupTail)root; 3145 3146 if (!anonymous && groupIndex < 10) 3147 groupNodes[groupIndex] = head; 3148 return head; 3149 } 3150 3151 @SuppressWarnings("fallthrough") 3152 /** 3153 * Parses inlined match flags and set them appropriately. 3154 */ 3155 private void addFlag() { 3156 int ch = peek(); 3157 for (;;) { 3158 switch (ch) { 3159 case 'i': 3160 flags0 |= CASE_INSENSITIVE; 3161 break; 3162 case 'm': 3163 flags0 |= MULTILINE; 3164 break; 3165 case 's': 3166 flags0 |= DOTALL; 3167 break; 3168 case 'd': 3169 flags0 |= UNIX_LINES; 3170 break; 3171 case 'u': 3172 flags0 |= UNICODE_CASE; 3173 break; 3174 case 'c': 3175 flags0 |= CANON_EQ; 3176 break; 3177 case 'x': 3178 flags0 |= COMMENTS; 3179 break; 3180 case 'U': 3181 flags0 |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3182 break; 3183 case '-': // subFlag then fall through 3184 ch = next(); 3185 subFlag(); 3186 default: 3187 return; 3188 } 3189 ch = next(); 3190 } 3191 } 3192 3193 @SuppressWarnings("fallthrough") 3194 /** 3195 * Parses the second part of inlined match flags and turns off 3196 * flags appropriately. 3197 */ 3198 private void subFlag() { 3199 int ch = peek(); 3200 for (;;) { 3201 switch (ch) { 3202 case 'i': 3203 flags0 &= ~CASE_INSENSITIVE; 3204 break; 3205 case 'm': 3206 flags0 &= ~MULTILINE; 3207 break; 3208 case 's': 3209 flags0 &= ~DOTALL; 3210 break; 3211 case 'd': 3212 flags0 &= ~UNIX_LINES; 3213 break; 3214 case 'u': 3215 flags0 &= ~UNICODE_CASE; 3216 break; 3217 case 'c': 3218 flags0 &= ~CANON_EQ; 3219 break; 3220 case 'x': 3221 flags0 &= ~COMMENTS; 3222 break; 3223 case 'U': 3224 flags0 &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE); 3225 break; 3226 default: 3227 return; 3228 } 3229 ch = next(); 3230 } 3231 } 3232 3233 static final int MAX_REPS = 0x7FFFFFFF; 3234 3235 static enum Qtype { 3236 GREEDY, LAZY, POSSESSIVE, INDEPENDENT 3237 } 3238 3239 private Node curly(Node prev, int cmin) { 3240 int ch = next(); 3241 if (ch == '?') { 3242 next(); 3243 return new Curly(prev, cmin, MAX_REPS, Qtype.LAZY); 3244 } else if (ch == '+') { 3245 next(); 3246 return new Curly(prev, cmin, MAX_REPS, Qtype.POSSESSIVE); 3247 } 3248 if (prev instanceof BmpCharProperty) { 3249 return new BmpCharPropertyGreedy((BmpCharProperty)prev, cmin); 3250 } else if (prev instanceof CharProperty) { 3251 return new CharPropertyGreedy((CharProperty)prev, cmin); 3252 } 3253 return new Curly(prev, cmin, MAX_REPS, Qtype.GREEDY); 3254 } 3255 3256 /** 3257 * Processes repetition. If the next character peeked is a quantifier 3258 * then new nodes must be appended to handle the repetition. 3259 * Prev could be a single or a group, so it could be a chain of nodes. 3260 */ 3261 private Node closure(Node prev) { 3262 Node atom; 3263 int ch = peek(); 3264 switch (ch) { 3265 case '?': 3266 ch = next(); 3267 if (ch == '?') { 3268 next(); 3269 return new Ques(prev, Qtype.LAZY); 3270 } else if (ch == '+') { 3271 next(); 3272 return new Ques(prev, Qtype.POSSESSIVE); 3273 } 3274 return new Ques(prev, Qtype.GREEDY); 3275 case '*': 3276 return curly(prev, 0); 3277 case '+': 3278 return curly(prev, 1); 3279 case '{': 3280 ch = temp[cursor+1]; 3281 if (ASCII.isDigit(ch)) { 3282 skip(); 3283 int cmin = 0; 3284 do { 3285 cmin = cmin * 10 + (ch - '0'); 3286 } while (ASCII.isDigit(ch = read())); 3287 int cmax = cmin; 3288 if (ch == ',') { 3289 ch = read(); 3290 cmax = MAX_REPS; 3291 if (ch != '}') { 3292 cmax = 0; 3293 while (ASCII.isDigit(ch)) { 3294 cmax = cmax * 10 + (ch - '0'); 3295 ch = read(); 3296 } 3297 } 3298 } 3299 if (ch != '}') 3300 throw error("Unclosed counted closure"); 3301 if (((cmin) | (cmax) | (cmax - cmin)) < 0) 3302 throw error("Illegal repetition range"); 3303 Curly curly; 3304 ch = peek(); 3305 if (ch == '?') { 3306 next(); 3307 curly = new Curly(prev, cmin, cmax, Qtype.LAZY); 3308 } else if (ch == '+') { 3309 next(); 3310 curly = new Curly(prev, cmin, cmax, Qtype.POSSESSIVE); 3311 } else { 3312 curly = new Curly(prev, cmin, cmax, Qtype.GREEDY); 3313 } 3314 return curly; 3315 } else { 3316 throw error("Illegal repetition"); 3317 } 3318 default: 3319 return prev; 3320 } 3321 } 3322 3323 /** 3324 * Utility method for parsing control escape sequences. 3325 */ 3326 private int c() { 3327 if (cursor < patternLength) { 3328 return read() ^ 64; 3329 } 3330 throw error("Illegal control escape sequence"); 3331 } 3332 3333 /** 3334 * Utility method for parsing octal escape sequences. 3335 */ 3336 private int o() { 3337 int n = read(); 3338 if (((n-'0')|('7'-n)) >= 0) { 3339 int m = read(); 3340 if (((m-'0')|('7'-m)) >= 0) { 3341 int o = read(); 3342 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) { 3343 return (n - '0') * 64 + (m - '0') * 8 + (o - '0'); 3344 } 3345 unread(); 3346 return (n - '0') * 8 + (m - '0'); 3347 } 3348 unread(); 3349 return (n - '0'); 3350 } 3351 throw error("Illegal octal escape sequence"); 3352 } 3353 3354 /** 3355 * Utility method for parsing hexadecimal escape sequences. 3356 */ 3357 private int x() { 3358 int n = read(); 3359 if (ASCII.isHexDigit(n)) { 3360 int m = read(); 3361 if (ASCII.isHexDigit(m)) { 3362 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m); 3363 } 3364 } else if (n == '{' && ASCII.isHexDigit(peek())) { 3365 int ch = 0; 3366 while (ASCII.isHexDigit(n = read())) { 3367 ch = (ch << 4) + ASCII.toDigit(n); 3368 if (ch > Character.MAX_CODE_POINT) 3369 throw error("Hexadecimal codepoint is too big"); 3370 } 3371 if (n != '}') 3372 throw error("Unclosed hexadecimal escape sequence"); 3373 return ch; 3374 } 3375 throw error("Illegal hexadecimal escape sequence"); 3376 } 3377 3378 /** 3379 * Utility method for parsing unicode escape sequences. 3380 */ 3381 private int cursor() { 3382 return cursor; 3383 } 3384 3385 private void setcursor(int pos) { 3386 cursor = pos; 3387 } 3388 3389 private int uxxxx() { 3390 int n = 0; 3391 for (int i = 0; i < 4; i++) { 3392 int ch = read(); 3393 if (!ASCII.isHexDigit(ch)) { 3394 throw error("Illegal Unicode escape sequence"); 3395 } 3396 n = n * 16 + ASCII.toDigit(ch); 3397 } 3398 return n; 3399 } 3400 3401 private int u() { 3402 int n = uxxxx(); 3403 if (Character.isHighSurrogate((char)n)) { 3404 int cur = cursor(); 3405 if (read() == '\\' && read() == 'u') { 3406 int n2 = uxxxx(); 3407 if (Character.isLowSurrogate((char)n2)) 3408 return Character.toCodePoint((char)n, (char)n2); 3409 } 3410 setcursor(cur); 3411 } 3412 return n; 3413 } 3414 3415 private int N() { 3416 if (read() == '{') { 3417 int i = cursor; 3418 while (cursor < patternLength && read() != '}') {} 3419 if (cursor > patternLength) 3420 throw error("Unclosed character name escape sequence"); 3421 String name = new String(temp, i, cursor - i - 1); 3422 try { 3423 return Character.codePointOf(name); 3424 } catch (IllegalArgumentException x) { 3425 throw error("Unknown character name [" + name + "]"); 3426 } 3427 } 3428 throw error("Illegal character name escape sequence"); 3429 } 3430 3431 // 3432 // Utility methods for code point support 3433 // 3434 private static final int countChars(CharSequence seq, int index, 3435 int lengthInCodePoints) { 3436 // optimization 3437 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) { 3438 assert (index >= 0 && index < seq.length()); 3439 return 1; 3440 } 3441 int length = seq.length(); 3442 int x = index; 3443 if (lengthInCodePoints >= 0) { 3444 assert (index >= 0 && index < length); 3445 for (int i = 0; x < length && i < lengthInCodePoints; i++) { 3446 if (Character.isHighSurrogate(seq.charAt(x++))) { 3447 if (x < length && Character.isLowSurrogate(seq.charAt(x))) { 3448 x++; 3449 } 3450 } 3451 } 3452 return x - index; 3453 } 3454 3455 assert (index >= 0 && index <= length); 3456 if (index == 0) { 3457 return 0; 3458 } 3459 int len = -lengthInCodePoints; 3460 for (int i = 0; x > 0 && i < len; i++) { 3461 if (Character.isLowSurrogate(seq.charAt(--x))) { 3462 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) { 3463 x--; 3464 } 3465 } 3466 } 3467 return index - x; 3468 } 3469 3470 private static final int countCodePoints(CharSequence seq) { 3471 int length = seq.length(); 3472 int n = 0; 3473 for (int i = 0; i < length; ) { 3474 n++; 3475 if (Character.isHighSurrogate(seq.charAt(i++))) { 3476 if (i < length && Character.isLowSurrogate(seq.charAt(i))) { 3477 i++; 3478 } 3479 } 3480 } 3481 return n; 3482 } 3483 3484 /** 3485 * Creates a bit vector for matching Latin-1 values. A normal BitClass 3486 * never matches values above Latin-1, and a complemented BitClass always 3487 * matches values above Latin-1. 3488 */ 3489 static final class BitClass extends BmpCharProperty { 3490 final boolean[] bits; 3491 BitClass() { 3492 this(new boolean[256]); 3493 } 3494 private BitClass(boolean[] bits) { 3495 super( ch -> ch < 256 && bits[ch]); 3496 this.bits = bits; 3497 } 3498 BitClass add(int c, int flags) { 3499 assert c >= 0 && c <= 255; 3500 if ((flags & CASE_INSENSITIVE) != 0) { 3501 if (ASCII.isAscii(c)) { 3502 bits[ASCII.toUpper(c)] = true; 3503 bits[ASCII.toLower(c)] = true; 3504 } else if ((flags & UNICODE_CASE) != 0) { 3505 bits[Character.toLowerCase(c)] = true; 3506 bits[Character.toUpperCase(c)] = true; 3507 } 3508 } 3509 bits[c] = true; 3510 return this; 3511 } 3512 } 3513 3514 /** 3515 * Utility method for creating a string slice matcher. 3516 */ 3517 private Node newSlice(int[] buf, int count, boolean hasSupplementary) { 3518 int[] tmp = new int[count]; 3519 if (has(CASE_INSENSITIVE)) { 3520 if (has(UNICODE_CASE)) { 3521 for (int i = 0; i < count; i++) { 3522 tmp[i] = Character.toLowerCase( 3523 Character.toUpperCase(buf[i])); 3524 } 3525 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp); 3526 } 3527 for (int i = 0; i < count; i++) { 3528 tmp[i] = ASCII.toLower(buf[i]); 3529 } 3530 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp); 3531 } 3532 for (int i = 0; i < count; i++) { 3533 tmp[i] = buf[i]; 3534 } 3535 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp); 3536 } 3537 3538 /** 3539 * The following classes are the building components of the object 3540 * tree that represents a compiled regular expression. The object tree 3541 * is made of individual elements that handle constructs in the Pattern. 3542 * Each type of object knows how to match its equivalent construct with 3543 * the match() method. 3544 */ 3545 3546 /** 3547 * Base class for all node classes. Subclasses should override the match() 3548 * method as appropriate. This class is an accepting node, so its match() 3549 * always returns true. 3550 */ 3551 static class Node extends Object { 3552 Node next; 3553 Node() { 3554 next = Pattern.accept; 3555 } 3556 /** 3557 * This method implements the classic accept node. 3558 */ 3559 boolean match(Matcher matcher, int i, CharSequence seq) { 3560 matcher.last = i; 3561 matcher.groups[0] = matcher.first; 3562 matcher.groups[1] = matcher.last; 3563 return true; 3564 } 3565 /** 3566 * This method is good for all zero length assertions. 3567 */ 3568 boolean study(TreeInfo info) { 3569 if (next != null) { 3570 return next.study(info); 3571 } else { 3572 return info.deterministic; 3573 } 3574 } 3575 } 3576 3577 static class LastNode extends Node { 3578 /** 3579 * This method implements the classic accept node with 3580 * the addition of a check to see if the match occurred 3581 * using all of the input. 3582 */ 3583 boolean match(Matcher matcher, int i, CharSequence seq) { 3584 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to) 3585 return false; 3586 matcher.last = i; 3587 matcher.groups[0] = matcher.first; 3588 matcher.groups[1] = matcher.last; 3589 return true; 3590 } 3591 } 3592 3593 /** 3594 * Used for REs that can start anywhere within the input string. 3595 * This basically tries to match repeatedly at each spot in the 3596 * input string, moving forward after each try. An anchored search 3597 * or a BnM will bypass this node completely. 3598 */ 3599 static class Start extends Node { 3600 int minLength; 3601 Start(Node node) { 3602 this.next = node; 3603 TreeInfo info = new TreeInfo(); 3604 next.study(info); 3605 minLength = info.minLength; 3606 } 3607 boolean match(Matcher matcher, int i, CharSequence seq) { 3608 if (i > matcher.to - minLength) { 3609 matcher.hitEnd = true; 3610 return false; 3611 } 3612 int guard = matcher.to - minLength; 3613 for (; i <= guard; i++) { 3614 if (next.match(matcher, i, seq)) { 3615 matcher.first = i; 3616 matcher.groups[0] = matcher.first; 3617 matcher.groups[1] = matcher.last; 3618 return true; 3619 } 3620 } 3621 matcher.hitEnd = true; 3622 return false; 3623 } 3624 boolean study(TreeInfo info) { 3625 next.study(info); 3626 info.maxValid = false; 3627 info.deterministic = false; 3628 return false; 3629 } 3630 } 3631 3632 /* 3633 * StartS supports supplementary characters, including unpaired surrogates. 3634 */ 3635 static final class StartS extends Start { 3636 StartS(Node node) { 3637 super(node); 3638 } 3639 boolean match(Matcher matcher, int i, CharSequence seq) { 3640 if (i > matcher.to - minLength) { 3641 matcher.hitEnd = true; 3642 return false; 3643 } 3644 int guard = matcher.to - minLength; 3645 while (i <= guard) { 3646 //if ((ret = next.match(matcher, i, seq)) || i == guard) 3647 if (next.match(matcher, i, seq)) { 3648 matcher.first = i; 3649 matcher.groups[0] = matcher.first; 3650 matcher.groups[1] = matcher.last; 3651 return true; 3652 } 3653 if (i == guard) 3654 break; 3655 // Optimization to move to the next character. This is 3656 // faster than countChars(seq, i, 1). 3657 if (Character.isHighSurrogate(seq.charAt(i++))) { 3658 if (i < seq.length() && 3659 Character.isLowSurrogate(seq.charAt(i))) { 3660 i++; 3661 } 3662 } 3663 } 3664 matcher.hitEnd = true; 3665 return false; 3666 } 3667 } 3668 3669 /** 3670 * Node to anchor at the beginning of input. This object implements the 3671 * match for a \A sequence, and the caret anchor will use this if not in 3672 * multiline mode. 3673 */ 3674 static final class Begin extends Node { 3675 boolean match(Matcher matcher, int i, CharSequence seq) { 3676 int fromIndex = (matcher.anchoringBounds) ? 3677 matcher.from : 0; 3678 if (i == fromIndex && next.match(matcher, i, seq)) { 3679 matcher.first = i; 3680 matcher.groups[0] = i; 3681 matcher.groups[1] = matcher.last; 3682 return true; 3683 } else { 3684 return false; 3685 } 3686 } 3687 } 3688 3689 /** 3690 * Node to anchor at the end of input. This is the absolute end, so this 3691 * should not match at the last newline before the end as $ will. 3692 */ 3693 static final class End extends Node { 3694 boolean match(Matcher matcher, int i, CharSequence seq) { 3695 int endIndex = (matcher.anchoringBounds) ? 3696 matcher.to : matcher.getTextLength(); 3697 if (i == endIndex) { 3698 matcher.hitEnd = true; 3699 return next.match(matcher, i, seq); 3700 } 3701 return false; 3702 } 3703 } 3704 3705 /** 3706 * Node to anchor at the beginning of a line. This is essentially the 3707 * object to match for the multiline ^. 3708 */ 3709 static final class Caret extends Node { 3710 boolean match(Matcher matcher, int i, CharSequence seq) { 3711 int startIndex = matcher.from; 3712 int endIndex = matcher.to; 3713 if (!matcher.anchoringBounds) { 3714 startIndex = 0; 3715 endIndex = matcher.getTextLength(); 3716 } 3717 // Perl does not match ^ at end of input even after newline 3718 if (i == endIndex) { 3719 matcher.hitEnd = true; 3720 return false; 3721 } 3722 if (i > startIndex) { 3723 char ch = seq.charAt(i-1); 3724 if (ch != '\n' && ch != '\r' 3725 && (ch|1) != '\u2029' 3726 && ch != '\u0085' ) { 3727 return false; 3728 } 3729 // Should treat /r/n as one newline 3730 if (ch == '\r' && seq.charAt(i) == '\n') 3731 return false; 3732 } 3733 return next.match(matcher, i, seq); 3734 } 3735 } 3736 3737 /** 3738 * Node to anchor at the beginning of a line when in unixdot mode. 3739 */ 3740 static final class UnixCaret extends Node { 3741 boolean match(Matcher matcher, int i, CharSequence seq) { 3742 int startIndex = matcher.from; 3743 int endIndex = matcher.to; 3744 if (!matcher.anchoringBounds) { 3745 startIndex = 0; 3746 endIndex = matcher.getTextLength(); 3747 } 3748 // Perl does not match ^ at end of input even after newline 3749 if (i == endIndex) { 3750 matcher.hitEnd = true; 3751 return false; 3752 } 3753 if (i > startIndex) { 3754 char ch = seq.charAt(i-1); 3755 if (ch != '\n') { 3756 return false; 3757 } 3758 } 3759 return next.match(matcher, i, seq); 3760 } 3761 } 3762 3763 /** 3764 * Node to match the location where the last match ended. 3765 * This is used for the \G construct. 3766 */ 3767 static final class LastMatch extends Node { 3768 boolean match(Matcher matcher, int i, CharSequence seq) { 3769 if (i != matcher.oldLast) 3770 return false; 3771 return next.match(matcher, i, seq); 3772 } 3773 } 3774 3775 /** 3776 * Node to anchor at the end of a line or the end of input based on the 3777 * multiline mode. 3778 * 3779 * When not in multiline mode, the $ can only match at the very end 3780 * of the input, unless the input ends in a line terminator in which 3781 * it matches right before the last line terminator. 3782 * 3783 * Note that \r\n is considered an atomic line terminator. 3784 * 3785 * Like ^ the $ operator matches at a position, it does not match the 3786 * line terminators themselves. 3787 */ 3788 static final class Dollar extends Node { 3789 boolean multiline; 3790 Dollar(boolean mul) { 3791 multiline = mul; 3792 } 3793 boolean match(Matcher matcher, int i, CharSequence seq) { 3794 int endIndex = (matcher.anchoringBounds) ? 3795 matcher.to : matcher.getTextLength(); 3796 if (!multiline) { 3797 if (i < endIndex - 2) 3798 return false; 3799 if (i == endIndex - 2) { 3800 char ch = seq.charAt(i); 3801 if (ch != '\r') 3802 return false; 3803 ch = seq.charAt(i + 1); 3804 if (ch != '\n') 3805 return false; 3806 } 3807 } 3808 // Matches before any line terminator; also matches at the 3809 // end of input 3810 // Before line terminator: 3811 // If multiline, we match here no matter what 3812 // If not multiline, fall through so that the end 3813 // is marked as hit; this must be a /r/n or a /n 3814 // at the very end so the end was hit; more input 3815 // could make this not match here 3816 if (i < endIndex) { 3817 char ch = seq.charAt(i); 3818 if (ch == '\n') { 3819 // No match between \r\n 3820 if (i > 0 && seq.charAt(i-1) == '\r') 3821 return false; 3822 if (multiline) 3823 return next.match(matcher, i, seq); 3824 } else if (ch == '\r' || ch == '\u0085' || 3825 (ch|1) == '\u2029') { 3826 if (multiline) 3827 return next.match(matcher, i, seq); 3828 } else { // No line terminator, no match 3829 return false; 3830 } 3831 } 3832 // Matched at current end so hit end 3833 matcher.hitEnd = true; 3834 // If a $ matches because of end of input, then more input 3835 // could cause it to fail! 3836 matcher.requireEnd = true; 3837 return next.match(matcher, i, seq); 3838 } 3839 boolean study(TreeInfo info) { 3840 next.study(info); 3841 return info.deterministic; 3842 } 3843 } 3844 3845 /** 3846 * Node to anchor at the end of a line or the end of input based on the 3847 * multiline mode when in unix lines mode. 3848 */ 3849 static final class UnixDollar extends Node { 3850 boolean multiline; 3851 UnixDollar(boolean mul) { 3852 multiline = mul; 3853 } 3854 boolean match(Matcher matcher, int i, CharSequence seq) { 3855 int endIndex = (matcher.anchoringBounds) ? 3856 matcher.to : matcher.getTextLength(); 3857 if (i < endIndex) { 3858 char ch = seq.charAt(i); 3859 if (ch == '\n') { 3860 // If not multiline, then only possible to 3861 // match at very end or one before end 3862 if (multiline == false && i != endIndex - 1) 3863 return false; 3864 // If multiline return next.match without setting 3865 // matcher.hitEnd 3866 if (multiline) 3867 return next.match(matcher, i, seq); 3868 } else { 3869 return false; 3870 } 3871 } 3872 // Matching because at the end or 1 before the end; 3873 // more input could change this so set hitEnd 3874 matcher.hitEnd = true; 3875 // If a $ matches because of end of input, then more input 3876 // could cause it to fail! 3877 matcher.requireEnd = true; 3878 return next.match(matcher, i, seq); 3879 } 3880 boolean study(TreeInfo info) { 3881 next.study(info); 3882 return info.deterministic; 3883 } 3884 } 3885 3886 /** 3887 * Node class that matches a Unicode line ending '\R' 3888 */ 3889 static final class LineEnding extends Node { 3890 boolean match(Matcher matcher, int i, CharSequence seq) { 3891 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029]) 3892 if (i < matcher.to) { 3893 int ch = seq.charAt(i); 3894 if (ch == 0x0A || ch == 0x0B || ch == 0x0C || 3895 ch == 0x85 || ch == 0x2028 || ch == 0x2029) 3896 return next.match(matcher, i + 1, seq); 3897 if (ch == 0x0D) { 3898 i++; 3899 if (i < matcher.to) { 3900 if (seq.charAt(i) == 0x0A && 3901 next.match(matcher, i + 1, seq)) { 3902 return true; 3903 } 3904 } else { 3905 matcher.hitEnd = true; 3906 } 3907 return next.match(matcher, i, seq); 3908 } 3909 } else { 3910 matcher.hitEnd = true; 3911 } 3912 return false; 3913 } 3914 boolean study(TreeInfo info) { 3915 info.minLength++; 3916 info.maxLength += 2; 3917 return next.study(info); 3918 } 3919 } 3920 3921 /** 3922 * Abstract node class to match one character satisfying some 3923 * boolean property. 3924 */ 3925 static class CharProperty extends Node { 3926 CharPredicate predicate; 3927 3928 CharProperty (CharPredicate predicate) { 3929 this.predicate = predicate; 3930 } 3931 boolean match(Matcher matcher, int i, CharSequence seq) { 3932 if (i < matcher.to) { 3933 int ch = Character.codePointAt(seq, i); 3934 return predicate.is(ch) && 3935 next.match(matcher, i + Character.charCount(ch), seq); 3936 } else { 3937 matcher.hitEnd = true; 3938 return false; 3939 } 3940 } 3941 boolean study(TreeInfo info) { 3942 info.minLength++; 3943 info.maxLength++; 3944 return next.study(info); 3945 } 3946 } 3947 3948 /** 3949 * Optimized version of CharProperty that works only for 3950 * properties never satisfied by Supplementary characters. 3951 */ 3952 private static class BmpCharProperty extends CharProperty { 3953 BmpCharProperty (BmpCharPredicate predicate) { 3954 super(predicate); 3955 } 3956 boolean match(Matcher matcher, int i, CharSequence seq) { 3957 if (i < matcher.to) { 3958 return predicate.is(seq.charAt(i)) && 3959 next.match(matcher, i + 1, seq); 3960 } else { 3961 matcher.hitEnd = true; 3962 return false; 3963 } 3964 } 3965 } 3966 3967 private static class NFCCharProperty extends Node { 3968 CharPredicate predicate; 3969 NFCCharProperty (CharPredicate predicate) { 3970 this.predicate = predicate; 3971 } 3972 3973 boolean match(Matcher matcher, int i, CharSequence seq) { 3974 if (i < matcher.to) { 3975 int ch0 = Character.codePointAt(seq, i); 3976 int n = Character.charCount(ch0); 3977 int j = i + n; 3978 while (j < matcher.to) { 3979 int ch1 = Character.codePointAt(seq, j); 3980 if (Grapheme.isBoundary(ch0, ch1)) 3981 break; 3982 ch0 = ch1; 3983 j += Character.charCount(ch1); 3984 } 3985 if (i + n == j) { // single, assume nfc cp 3986 if (predicate.is(ch0)) 3987 return next.match(matcher, j, seq); 3988 } else { 3989 while (i + n < j) { 3990 String nfc = Normalizer.normalize( 3991 seq.toString().substring(i, j), Normalizer.Form.NFC); 3992 if (nfc.codePointCount(0, nfc.length()) == 1) { 3993 if (predicate.is(nfc.codePointAt(0)) && 3994 next.match(matcher, j, seq)) { 3995 return true; 3996 } 3997 } 3998 3999 ch0 = Character.codePointBefore(seq, j); 4000 j -= Character.charCount(ch0); 4001 } 4002 } 4003 if (j < matcher.to) 4004 return false; 4005 } 4006 matcher.hitEnd = true; 4007 return false; 4008 } 4009 4010 boolean study(TreeInfo info) { 4011 info.minLength++; 4012 info.deterministic = false; 4013 return next.study(info); 4014 } 4015 } 4016 4017 /** 4018 * Node class that matches an unicode extended grapheme cluster 4019 */ 4020 static class XGrapheme extends Node { 4021 boolean match(Matcher matcher, int i, CharSequence seq) { 4022 if (i < matcher.to) { 4023 int ch0 = Character.codePointAt(seq, i); 4024 i += Character.charCount(ch0); 4025 while (i < matcher.to) { 4026 int ch1 = Character.codePointAt(seq, i); 4027 if (Grapheme.isBoundary(ch0, ch1)) 4028 break; 4029 ch0 = ch1; 4030 i += Character.charCount(ch1); 4031 } 4032 return next.match(matcher, i, seq); 4033 } 4034 matcher.hitEnd = true; 4035 return false; 4036 } 4037 4038 boolean study(TreeInfo info) { 4039 info.minLength++; 4040 info.deterministic = false; 4041 return next.study(info); 4042 } 4043 } 4044 4045 /** 4046 * Node class that handles grapheme boundaries 4047 */ 4048 static class GraphemeBound extends Node { 4049 boolean match(Matcher matcher, int i, CharSequence seq) { 4050 int startIndex = matcher.from; 4051 int endIndex = matcher.to; 4052 if (matcher.transparentBounds) { 4053 startIndex = 0; 4054 endIndex = matcher.getTextLength(); 4055 } 4056 if (i == startIndex) { 4057 return next.match(matcher, i, seq); 4058 } 4059 if (i < endIndex) { 4060 if (Character.isSurrogatePair(seq.charAt(i-1), seq.charAt(i)) || 4061 !Grapheme.isBoundary(Character.codePointBefore(seq, i), 4062 Character.codePointAt(seq, i))) { 4063 return false; 4064 } 4065 } else { 4066 matcher.hitEnd = true; 4067 matcher.requireEnd = true; 4068 } 4069 return next.match(matcher, i, seq); 4070 } 4071 } 4072 4073 /** 4074 * Base class for all Slice nodes 4075 */ 4076 static class SliceNode extends Node { 4077 int[] buffer; 4078 SliceNode(int[] buf) { 4079 buffer = buf; 4080 } 4081 boolean study(TreeInfo info) { 4082 info.minLength += buffer.length; 4083 info.maxLength += buffer.length; 4084 return next.study(info); 4085 } 4086 } 4087 4088 /** 4089 * Node class for a case sensitive/BMP-only sequence of literal 4090 * characters. 4091 */ 4092 static class Slice extends SliceNode { 4093 Slice(int[] buf) { 4094 super(buf); 4095 } 4096 boolean match(Matcher matcher, int i, CharSequence seq) { 4097 int[] buf = buffer; 4098 int len = buf.length; 4099 for (int j=0; j<len; j++) { 4100 if ((i+j) >= matcher.to) { 4101 matcher.hitEnd = true; 4102 return false; 4103 } 4104 if (buf[j] != seq.charAt(i+j)) 4105 return false; 4106 } 4107 return next.match(matcher, i+len, seq); 4108 } 4109 } 4110 4111 /** 4112 * Node class for a case_insensitive/BMP-only sequence of literal 4113 * characters. 4114 */ 4115 static class SliceI extends SliceNode { 4116 SliceI(int[] buf) { 4117 super(buf); 4118 } 4119 boolean match(Matcher matcher, int i, CharSequence seq) { 4120 int[] buf = buffer; 4121 int len = buf.length; 4122 for (int j=0; j<len; j++) { 4123 if ((i+j) >= matcher.to) { 4124 matcher.hitEnd = true; 4125 return false; 4126 } 4127 int c = seq.charAt(i+j); 4128 if (buf[j] != c && 4129 buf[j] != ASCII.toLower(c)) 4130 return false; 4131 } 4132 return next.match(matcher, i+len, seq); 4133 } 4134 } 4135 4136 /** 4137 * Node class for a unicode_case_insensitive/BMP-only sequence of 4138 * literal characters. Uses unicode case folding. 4139 */ 4140 static final class SliceU extends SliceNode { 4141 SliceU(int[] buf) { 4142 super(buf); 4143 } 4144 boolean match(Matcher matcher, int i, CharSequence seq) { 4145 int[] buf = buffer; 4146 int len = buf.length; 4147 for (int j=0; j<len; j++) { 4148 if ((i+j) >= matcher.to) { 4149 matcher.hitEnd = true; 4150 return false; 4151 } 4152 int c = seq.charAt(i+j); 4153 if (buf[j] != c && 4154 buf[j] != Character.toLowerCase(Character.toUpperCase(c))) 4155 return false; 4156 } 4157 return next.match(matcher, i+len, seq); 4158 } 4159 } 4160 4161 /** 4162 * Node class for a case sensitive sequence of literal characters 4163 * including supplementary characters. 4164 */ 4165 static final class SliceS extends Slice { 4166 SliceS(int[] buf) { 4167 super(buf); 4168 } 4169 boolean match(Matcher matcher, int i, CharSequence seq) { 4170 int[] buf = buffer; 4171 int x = i; 4172 for (int j = 0; j < buf.length; j++) { 4173 if (x >= matcher.to) { 4174 matcher.hitEnd = true; 4175 return false; 4176 } 4177 int c = Character.codePointAt(seq, x); 4178 if (buf[j] != c) 4179 return false; 4180 x += Character.charCount(c); 4181 if (x > matcher.to) { 4182 matcher.hitEnd = true; 4183 return false; 4184 } 4185 } 4186 return next.match(matcher, x, seq); 4187 } 4188 } 4189 4190 /** 4191 * Node class for a case insensitive sequence of literal characters 4192 * including supplementary characters. 4193 */ 4194 static class SliceIS extends SliceNode { 4195 SliceIS(int[] buf) { 4196 super(buf); 4197 } 4198 int toLower(int c) { 4199 return ASCII.toLower(c); 4200 } 4201 boolean match(Matcher matcher, int i, CharSequence seq) { 4202 int[] buf = buffer; 4203 int x = i; 4204 for (int j = 0; j < buf.length; j++) { 4205 if (x >= matcher.to) { 4206 matcher.hitEnd = true; 4207 return false; 4208 } 4209 int c = Character.codePointAt(seq, x); 4210 if (buf[j] != c && buf[j] != toLower(c)) 4211 return false; 4212 x += Character.charCount(c); 4213 if (x > matcher.to) { 4214 matcher.hitEnd = true; 4215 return false; 4216 } 4217 } 4218 return next.match(matcher, x, seq); 4219 } 4220 } 4221 4222 /** 4223 * Node class for a case insensitive sequence of literal characters. 4224 * Uses unicode case folding. 4225 */ 4226 static final class SliceUS extends SliceIS { 4227 SliceUS(int[] buf) { 4228 super(buf); 4229 } 4230 int toLower(int c) { 4231 return Character.toLowerCase(Character.toUpperCase(c)); 4232 } 4233 } 4234 4235 /** 4236 * The 0 or 1 quantifier. This one class implements all three types. 4237 */ 4238 static final class Ques extends Node { 4239 Node atom; 4240 Qtype type; 4241 Ques(Node node, Qtype type) { 4242 this.atom = node; 4243 this.type = type; 4244 } 4245 boolean match(Matcher matcher, int i, CharSequence seq) { 4246 switch (type) { 4247 case GREEDY: 4248 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)) 4249 || next.match(matcher, i, seq); 4250 case LAZY: 4251 return next.match(matcher, i, seq) 4252 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq)); 4253 case POSSESSIVE: 4254 if (atom.match(matcher, i, seq)) i = matcher.last; 4255 return next.match(matcher, i, seq); 4256 default: 4257 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq); 4258 } 4259 } 4260 boolean study(TreeInfo info) { 4261 if (type != Qtype.INDEPENDENT) { 4262 int minL = info.minLength; 4263 atom.study(info); 4264 info.minLength = minL; 4265 info.deterministic = false; 4266 return next.study(info); 4267 } else { 4268 atom.study(info); 4269 return next.study(info); 4270 } 4271 } 4272 } 4273 4274 /** 4275 * Handles the greedy style repetition with the minimum either be 4276 * 0 or 1 and the maximum be MAX_REPS, for * and + quantifier. 4277 */ 4278 static class CharPropertyGreedy extends Node { 4279 final CharPredicate predicate; 4280 final int cmin; 4281 4282 CharPropertyGreedy(CharProperty cp, int cmin) { 4283 this.predicate = cp.predicate; 4284 this.cmin = cmin; 4285 } 4286 boolean match(Matcher matcher, int i, CharSequence seq) { 4287 int n = 0; 4288 int to = matcher.to; 4289 // greedy, all the way down 4290 while (i < to) { 4291 int ch = Character.codePointAt(seq, i); 4292 if (!predicate.is(ch)) 4293 break; 4294 i += Character.charCount(ch); 4295 n++; 4296 } 4297 if (i >= to) { 4298 matcher.hitEnd = true; 4299 } 4300 while (n >= cmin) { 4301 if (next.match(matcher, i, seq)) 4302 return true; 4303 if (n == cmin) 4304 return false; 4305 // backing off if match fails 4306 int ch = Character.codePointBefore(seq, i); 4307 i -= Character.charCount(ch); 4308 n--; 4309 } 4310 return false; 4311 } 4312 4313 boolean study(TreeInfo info) { 4314 info.minLength += cmin; 4315 if (info.maxValid) { 4316 info.maxLength += MAX_REPS; 4317 } 4318 info.deterministic = false; 4319 return next.study(info); 4320 } 4321 } 4322 4323 static final class BmpCharPropertyGreedy extends CharPropertyGreedy { 4324 4325 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) { 4326 super(bcp, cmin); 4327 } 4328 4329 boolean match(Matcher matcher, int i, CharSequence seq) { 4330 int n = 0; 4331 int to = matcher.to; 4332 while (i < to && predicate.is(seq.charAt(i))) { 4333 i++; n++; 4334 } 4335 if (i >= to) { 4336 matcher.hitEnd = true; 4337 } 4338 while (n >= cmin) { 4339 if (next.match(matcher, i, seq)) 4340 return true; 4341 i--; n--; // backing off if match fails 4342 } 4343 return false; 4344 } 4345 } 4346 4347 /** 4348 * Handles the curly-brace style repetition with a specified minimum and 4349 * maximum occurrences. The * quantifier is handled as a special case. 4350 * This class handles the three types. 4351 */ 4352 static final class Curly extends Node { 4353 Node atom; 4354 Qtype type; 4355 int cmin; 4356 int cmax; 4357 4358 Curly(Node node, int cmin, int cmax, Qtype type) { 4359 this.atom = node; 4360 this.type = type; 4361 this.cmin = cmin; 4362 this.cmax = cmax; 4363 } 4364 boolean match(Matcher matcher, int i, CharSequence seq) { 4365 int j; 4366 for (j = 0; j < cmin; j++) { 4367 if (atom.match(matcher, i, seq)) { 4368 i = matcher.last; 4369 continue; 4370 } 4371 return false; 4372 } 4373 if (type == Qtype.GREEDY) 4374 return match0(matcher, i, j, seq); 4375 else if (type == Qtype.LAZY) 4376 return match1(matcher, i, j, seq); 4377 else 4378 return match2(matcher, i, j, seq); 4379 } 4380 // Greedy match. 4381 // i is the index to start matching at 4382 // j is the number of atoms that have matched 4383 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4384 if (j >= cmax) { 4385 // We have matched the maximum... continue with the rest of 4386 // the regular expression 4387 return next.match(matcher, i, seq); 4388 } 4389 int backLimit = j; 4390 while (atom.match(matcher, i, seq)) { 4391 // k is the length of this match 4392 int k = matcher.last - i; 4393 if (k == 0) // Zero length match 4394 break; 4395 // Move up index and number matched 4396 i = matcher.last; 4397 j++; 4398 // We are greedy so match as many as we can 4399 while (j < cmax) { 4400 if (!atom.match(matcher, i, seq)) 4401 break; 4402 if (i + k != matcher.last) { 4403 if (match0(matcher, matcher.last, j+1, seq)) 4404 return true; 4405 break; 4406 } 4407 i += k; 4408 j++; 4409 } 4410 // Handle backing off if match fails 4411 while (j >= backLimit) { 4412 if (next.match(matcher, i, seq)) 4413 return true; 4414 i -= k; 4415 j--; 4416 } 4417 return false; 4418 } 4419 return next.match(matcher, i, seq); 4420 } 4421 // Reluctant match. At this point, the minimum has been satisfied. 4422 // i is the index to start matching at 4423 // j is the number of atoms that have matched 4424 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4425 for (;;) { 4426 // Try finishing match without consuming any more 4427 if (next.match(matcher, i, seq)) 4428 return true; 4429 // At the maximum, no match found 4430 if (j >= cmax) 4431 return false; 4432 // Okay, must try one more atom 4433 if (!atom.match(matcher, i, seq)) 4434 return false; 4435 // If we haven't moved forward then must break out 4436 if (i == matcher.last) 4437 return false; 4438 // Move up index and number matched 4439 i = matcher.last; 4440 j++; 4441 } 4442 } 4443 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4444 for (; j < cmax; j++) { 4445 if (!atom.match(matcher, i, seq)) 4446 break; 4447 if (i == matcher.last) 4448 break; 4449 i = matcher.last; 4450 } 4451 return next.match(matcher, i, seq); 4452 } 4453 boolean study(TreeInfo info) { 4454 // Save original info 4455 int minL = info.minLength; 4456 int maxL = info.maxLength; 4457 boolean maxV = info.maxValid; 4458 boolean detm = info.deterministic; 4459 info.reset(); 4460 4461 atom.study(info); 4462 4463 int temp = info.minLength * cmin + minL; 4464 if (temp < minL) { 4465 temp = 0xFFFFFFF; // arbitrary large number 4466 } 4467 info.minLength = temp; 4468 4469 if (maxV & info.maxValid) { 4470 temp = info.maxLength * cmax + maxL; 4471 info.maxLength = temp; 4472 if (temp < maxL) { 4473 info.maxValid = false; 4474 } 4475 } else { 4476 info.maxValid = false; 4477 } 4478 4479 if (info.deterministic && cmin == cmax) 4480 info.deterministic = detm; 4481 else 4482 info.deterministic = false; 4483 return next.study(info); 4484 } 4485 } 4486 4487 /** 4488 * Handles the curly-brace style repetition with a specified minimum and 4489 * maximum occurrences in deterministic cases. This is an iterative 4490 * optimization over the Prolog and Loop system which would handle this 4491 * in a recursive way. The * quantifier is handled as a special case. 4492 * If capture is true then this class saves group settings and ensures 4493 * that groups are unset when backing off of a group match. 4494 */ 4495 static final class GroupCurly extends Node { 4496 Node atom; 4497 Qtype type; 4498 int cmin; 4499 int cmax; 4500 int localIndex; 4501 int groupIndex; 4502 boolean capture; 4503 4504 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local, 4505 int group, boolean capture) { 4506 this.atom = node; 4507 this.type = type; 4508 this.cmin = cmin; 4509 this.cmax = cmax; 4510 this.localIndex = local; 4511 this.groupIndex = group; 4512 this.capture = capture; 4513 } 4514 boolean match(Matcher matcher, int i, CharSequence seq) { 4515 int[] groups = matcher.groups; 4516 int[] locals = matcher.locals; 4517 int save0 = locals[localIndex]; 4518 int save1 = 0; 4519 int save2 = 0; 4520 4521 if (capture) { 4522 save1 = groups[groupIndex]; 4523 save2 = groups[groupIndex+1]; 4524 } 4525 4526 // Notify GroupTail there is no need to setup group info 4527 // because it will be set here 4528 locals[localIndex] = -1; 4529 4530 boolean ret = true; 4531 for (int j = 0; j < cmin; j++) { 4532 if (atom.match(matcher, i, seq)) { 4533 if (capture) { 4534 groups[groupIndex] = i; 4535 groups[groupIndex+1] = matcher.last; 4536 } 4537 i = matcher.last; 4538 } else { 4539 ret = false; 4540 break; 4541 } 4542 } 4543 if (ret) { 4544 if (type == Qtype.GREEDY) { 4545 ret = match0(matcher, i, cmin, seq); 4546 } else if (type == Qtype.LAZY) { 4547 ret = match1(matcher, i, cmin, seq); 4548 } else { 4549 ret = match2(matcher, i, cmin, seq); 4550 } 4551 } 4552 if (!ret) { 4553 locals[localIndex] = save0; 4554 if (capture) { 4555 groups[groupIndex] = save1; 4556 groups[groupIndex+1] = save2; 4557 } 4558 } 4559 return ret; 4560 } 4561 // Aggressive group match 4562 boolean match0(Matcher matcher, int i, int j, CharSequence seq) { 4563 // don't back off passing the starting "j" 4564 int min = j; 4565 int[] groups = matcher.groups; 4566 int save0 = 0; 4567 int save1 = 0; 4568 if (capture) { 4569 save0 = groups[groupIndex]; 4570 save1 = groups[groupIndex+1]; 4571 } 4572 for (;;) { 4573 if (j >= cmax) 4574 break; 4575 if (!atom.match(matcher, i, seq)) 4576 break; 4577 int k = matcher.last - i; 4578 if (k <= 0) { 4579 if (capture) { 4580 groups[groupIndex] = i; 4581 groups[groupIndex+1] = i + k; 4582 } 4583 i = i + k; 4584 break; 4585 } 4586 for (;;) { 4587 if (capture) { 4588 groups[groupIndex] = i; 4589 groups[groupIndex+1] = i + k; 4590 } 4591 i = i + k; 4592 if (++j >= cmax) 4593 break; 4594 if (!atom.match(matcher, i, seq)) 4595 break; 4596 if (i + k != matcher.last) { 4597 if (match0(matcher, i, j, seq)) 4598 return true; 4599 break; 4600 } 4601 } 4602 while (j > min) { 4603 if (next.match(matcher, i, seq)) { 4604 if (capture) { 4605 groups[groupIndex+1] = i; 4606 groups[groupIndex] = i - k; 4607 } 4608 return true; 4609 } 4610 // backing off 4611 i = i - k; 4612 if (capture) { 4613 groups[groupIndex+1] = i; 4614 groups[groupIndex] = i - k; 4615 } 4616 j--; 4617 4618 } 4619 break; 4620 } 4621 if (capture) { 4622 groups[groupIndex] = save0; 4623 groups[groupIndex+1] = save1; 4624 } 4625 return next.match(matcher, i, seq); 4626 } 4627 // Reluctant matching 4628 boolean match1(Matcher matcher, int i, int j, CharSequence seq) { 4629 for (;;) { 4630 if (next.match(matcher, i, seq)) 4631 return true; 4632 if (j >= cmax) 4633 return false; 4634 if (!atom.match(matcher, i, seq)) 4635 return false; 4636 if (i == matcher.last) 4637 return false; 4638 if (capture) { 4639 matcher.groups[groupIndex] = i; 4640 matcher.groups[groupIndex+1] = matcher.last; 4641 } 4642 i = matcher.last; 4643 j++; 4644 } 4645 } 4646 // Possessive matching 4647 boolean match2(Matcher matcher, int i, int j, CharSequence seq) { 4648 for (; j < cmax; j++) { 4649 if (!atom.match(matcher, i, seq)) { 4650 break; 4651 } 4652 if (capture) { 4653 matcher.groups[groupIndex] = i; 4654 matcher.groups[groupIndex+1] = matcher.last; 4655 } 4656 if (i == matcher.last) { 4657 break; 4658 } 4659 i = matcher.last; 4660 } 4661 return next.match(matcher, i, seq); 4662 } 4663 boolean study(TreeInfo info) { 4664 // Save original info 4665 int minL = info.minLength; 4666 int maxL = info.maxLength; 4667 boolean maxV = info.maxValid; 4668 boolean detm = info.deterministic; 4669 info.reset(); 4670 4671 atom.study(info); 4672 4673 int temp = info.minLength * cmin + minL; 4674 if (temp < minL) { 4675 temp = 0xFFFFFFF; // Arbitrary large number 4676 } 4677 info.minLength = temp; 4678 4679 if (maxV & info.maxValid) { 4680 temp = info.maxLength * cmax + maxL; 4681 info.maxLength = temp; 4682 if (temp < maxL) { 4683 info.maxValid = false; 4684 } 4685 } else { 4686 info.maxValid = false; 4687 } 4688 4689 if (info.deterministic && cmin == cmax) { 4690 info.deterministic = detm; 4691 } else { 4692 info.deterministic = false; 4693 } 4694 return next.study(info); 4695 } 4696 } 4697 4698 /** 4699 * A Guard node at the end of each atom node in a Branch. It 4700 * serves the purpose of chaining the "match" operation to 4701 * "next" but not the "study", so we can collect the TreeInfo 4702 * of each atom node without including the TreeInfo of the 4703 * "next". 4704 */ 4705 static final class BranchConn extends Node { 4706 BranchConn() {}; 4707 boolean match(Matcher matcher, int i, CharSequence seq) { 4708 return next.match(matcher, i, seq); 4709 } 4710 boolean study(TreeInfo info) { 4711 return info.deterministic; 4712 } 4713 } 4714 4715 /** 4716 * Handles the branching of alternations. Note this is also used for 4717 * the ? quantifier to branch between the case where it matches once 4718 * and where it does not occur. 4719 */ 4720 static final class Branch extends Node { 4721 Node[] atoms = new Node[2]; 4722 int size = 2; 4723 Node conn; 4724 Branch(Node first, Node second, Node branchConn) { 4725 conn = branchConn; 4726 atoms[0] = first; 4727 atoms[1] = second; 4728 } 4729 4730 void add(Node node) { 4731 if (size >= atoms.length) { 4732 Node[] tmp = new Node[atoms.length*2]; 4733 System.arraycopy(atoms, 0, tmp, 0, atoms.length); 4734 atoms = tmp; 4735 } 4736 atoms[size++] = node; 4737 } 4738 4739 boolean match(Matcher matcher, int i, CharSequence seq) { 4740 for (int n = 0; n < size; n++) { 4741 if (atoms[n] == null) { 4742 if (conn.next.match(matcher, i, seq)) 4743 return true; 4744 } else if (atoms[n].match(matcher, i, seq)) { 4745 return true; 4746 } 4747 } 4748 return false; 4749 } 4750 4751 boolean study(TreeInfo info) { 4752 int minL = info.minLength; 4753 int maxL = info.maxLength; 4754 boolean maxV = info.maxValid; 4755 4756 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num 4757 int maxL2 = -1; 4758 for (int n = 0; n < size; n++) { 4759 info.reset(); 4760 if (atoms[n] != null) 4761 atoms[n].study(info); 4762 minL2 = Math.min(minL2, info.minLength); 4763 maxL2 = Math.max(maxL2, info.maxLength); 4764 maxV = (maxV & info.maxValid); 4765 } 4766 4767 minL += minL2; 4768 maxL += maxL2; 4769 4770 info.reset(); 4771 conn.next.study(info); 4772 4773 info.minLength += minL; 4774 info.maxLength += maxL; 4775 info.maxValid &= maxV; 4776 info.deterministic = false; 4777 return false; 4778 } 4779 } 4780 4781 /** 4782 * The GroupHead saves the location where the group begins in the locals 4783 * and restores them when the match is done. 4784 * 4785 * The matchRef is used when a reference to this group is accessed later 4786 * in the expression. The locals will have a negative value in them to 4787 * indicate that we do not want to unset the group if the reference 4788 * doesn't match. 4789 */ 4790 static final class GroupHead extends Node { 4791 int localIndex; 4792 GroupTail tail; // for debug/print only, match does not need to know 4793 GroupHead(int localCount) { 4794 localIndex = localCount; 4795 } 4796 boolean match(Matcher matcher, int i, CharSequence seq) { 4797 int save = matcher.locals[localIndex]; 4798 matcher.locals[localIndex] = i; 4799 boolean ret = next.match(matcher, i, seq); 4800 matcher.locals[localIndex] = save; 4801 return ret; 4802 } 4803 boolean matchRef(Matcher matcher, int i, CharSequence seq) { 4804 int save = matcher.locals[localIndex]; 4805 matcher.locals[localIndex] = ~i; // HACK 4806 boolean ret = next.match(matcher, i, seq); 4807 matcher.locals[localIndex] = save; 4808 return ret; 4809 } 4810 } 4811 4812 /** 4813 * Recursive reference to a group in the regular expression. It calls 4814 * matchRef because if the reference fails to match we would not unset 4815 * the group. 4816 */ 4817 static final class GroupRef extends Node { 4818 GroupHead head; 4819 GroupRef(GroupHead head) { 4820 this.head = head; 4821 } 4822 boolean match(Matcher matcher, int i, CharSequence seq) { 4823 return head.matchRef(matcher, i, seq) 4824 && next.match(matcher, matcher.last, seq); 4825 } 4826 boolean study(TreeInfo info) { 4827 info.maxValid = false; 4828 info.deterministic = false; 4829 return next.study(info); 4830 } 4831 } 4832 4833 /** 4834 * The GroupTail handles the setting of group beginning and ending 4835 * locations when groups are successfully matched. It must also be able to 4836 * unset groups that have to be backed off of. 4837 * 4838 * The GroupTail node is also used when a previous group is referenced, 4839 * and in that case no group information needs to be set. 4840 */ 4841 static final class GroupTail extends Node { 4842 int localIndex; 4843 int groupIndex; 4844 GroupTail(int localCount, int groupCount) { 4845 localIndex = localCount; 4846 groupIndex = groupCount + groupCount; 4847 } 4848 boolean match(Matcher matcher, int i, CharSequence seq) { 4849 int tmp = matcher.locals[localIndex]; 4850 if (tmp >= 0) { // This is the normal group case. 4851 // Save the group so we can unset it if it 4852 // backs off of a match. 4853 int groupStart = matcher.groups[groupIndex]; 4854 int groupEnd = matcher.groups[groupIndex+1]; 4855 4856 matcher.groups[groupIndex] = tmp; 4857 matcher.groups[groupIndex+1] = i; 4858 if (next.match(matcher, i, seq)) { 4859 return true; 4860 } 4861 matcher.groups[groupIndex] = groupStart; 4862 matcher.groups[groupIndex+1] = groupEnd; 4863 return false; 4864 } else { 4865 // This is a group reference case. We don't need to save any 4866 // group info because it isn't really a group. 4867 matcher.last = i; 4868 return true; 4869 } 4870 } 4871 } 4872 4873 /** 4874 * This sets up a loop to handle a recursive quantifier structure. 4875 */ 4876 static final class Prolog extends Node { 4877 Loop loop; 4878 Prolog(Loop loop) { 4879 this.loop = loop; 4880 } 4881 boolean match(Matcher matcher, int i, CharSequence seq) { 4882 return loop.matchInit(matcher, i, seq); 4883 } 4884 boolean study(TreeInfo info) { 4885 return loop.study(info); 4886 } 4887 } 4888 4889 /** 4890 * Handles the repetition count for a greedy Curly. The matchInit 4891 * is called from the Prolog to save the index of where the group 4892 * beginning is stored. A zero length group check occurs in the 4893 * normal match but is skipped in the matchInit. 4894 */ 4895 static class Loop extends Node { 4896 Node body; 4897 int countIndex; // local count index in matcher locals 4898 int beginIndex; // group beginning index 4899 int cmin, cmax; 4900 int posIndex; 4901 Loop(int countIndex, int beginIndex) { 4902 this.countIndex = countIndex; 4903 this.beginIndex = beginIndex; 4904 this.posIndex = -1; 4905 } 4906 boolean match(Matcher matcher, int i, CharSequence seq) { 4907 // Avoid infinite loop in zero-length case. 4908 if (i > matcher.locals[beginIndex]) { 4909 int count = matcher.locals[countIndex]; 4910 4911 // This block is for before we reach the minimum 4912 // iterations required for the loop to match 4913 if (count < cmin) { 4914 matcher.locals[countIndex] = count + 1; 4915 boolean b = body.match(matcher, i, seq); 4916 // If match failed we must backtrack, so 4917 // the loop count should NOT be incremented 4918 if (!b) 4919 matcher.locals[countIndex] = count; 4920 // Return success or failure since we are under 4921 // minimum 4922 return b; 4923 } 4924 // This block is for after we have the minimum 4925 // iterations required for the loop to match 4926 if (count < cmax) { 4927 // Let's check if we have already tried and failed 4928 // at this starting position "i" in the past. 4929 // If yes, then just return false wihtout trying 4930 // again, to stop the exponential backtracking. 4931 if (posIndex != -1 && 4932 matcher.localsPos[posIndex].contains(i)) { 4933 return next.match(matcher, i, seq); 4934 } 4935 matcher.locals[countIndex] = count + 1; 4936 boolean b = body.match(matcher, i, seq); 4937 // If match failed we must backtrack, so 4938 // the loop count should NOT be incremented 4939 if (b) 4940 return true; 4941 matcher.locals[countIndex] = count; 4942 // save the failed position 4943 if (posIndex != -1) { 4944 matcher.localsPos[posIndex].add(i); 4945 } 4946 } 4947 } 4948 return next.match(matcher, i, seq); 4949 } 4950 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 4951 int save = matcher.locals[countIndex]; 4952 boolean ret = false; 4953 if (posIndex != -1 && matcher.localsPos[posIndex] == null) { 4954 matcher.localsPos[posIndex] = new IntHashSet(); 4955 } 4956 if (0 < cmin) { 4957 matcher.locals[countIndex] = 1; 4958 ret = body.match(matcher, i, seq); 4959 } else if (0 < cmax) { 4960 matcher.locals[countIndex] = 1; 4961 ret = body.match(matcher, i, seq); 4962 if (ret == false) 4963 ret = next.match(matcher, i, seq); 4964 } else { 4965 ret = next.match(matcher, i, seq); 4966 } 4967 matcher.locals[countIndex] = save; 4968 return ret; 4969 } 4970 boolean study(TreeInfo info) { 4971 info.maxValid = false; 4972 info.deterministic = false; 4973 return false; 4974 } 4975 } 4976 4977 /** 4978 * Handles the repetition count for a reluctant Curly. The matchInit 4979 * is called from the Prolog to save the index of where the group 4980 * beginning is stored. A zero length group check occurs in the 4981 * normal match but is skipped in the matchInit. 4982 */ 4983 static final class LazyLoop extends Loop { 4984 LazyLoop(int countIndex, int beginIndex) { 4985 super(countIndex, beginIndex); 4986 } 4987 boolean match(Matcher matcher, int i, CharSequence seq) { 4988 // Check for zero length group 4989 if (i > matcher.locals[beginIndex]) { 4990 int count = matcher.locals[countIndex]; 4991 if (count < cmin) { 4992 matcher.locals[countIndex] = count + 1; 4993 boolean result = body.match(matcher, i, seq); 4994 // If match failed we must backtrack, so 4995 // the loop count should NOT be incremented 4996 if (!result) 4997 matcher.locals[countIndex] = count; 4998 return result; 4999 } 5000 if (next.match(matcher, i, seq)) 5001 return true; 5002 if (count < cmax) { 5003 matcher.locals[countIndex] = count + 1; 5004 boolean result = body.match(matcher, i, seq); 5005 // If match failed we must backtrack, so 5006 // the loop count should NOT be incremented 5007 if (!result) 5008 matcher.locals[countIndex] = count; 5009 return result; 5010 } 5011 return false; 5012 } 5013 return next.match(matcher, i, seq); 5014 } 5015 boolean matchInit(Matcher matcher, int i, CharSequence seq) { 5016 int save = matcher.locals[countIndex]; 5017 boolean ret = false; 5018 if (0 < cmin) { 5019 matcher.locals[countIndex] = 1; 5020 ret = body.match(matcher, i, seq); 5021 } else if (next.match(matcher, i, seq)) { 5022 ret = true; 5023 } else if (0 < cmax) { 5024 matcher.locals[countIndex] = 1; 5025 ret = body.match(matcher, i, seq); 5026 } 5027 matcher.locals[countIndex] = save; 5028 return ret; 5029 } 5030 boolean study(TreeInfo info) { 5031 info.maxValid = false; 5032 info.deterministic = false; 5033 return false; 5034 } 5035 } 5036 5037 /** 5038 * Refers to a group in the regular expression. Attempts to match 5039 * whatever the group referred to last matched. 5040 */ 5041 static class BackRef extends Node { 5042 int groupIndex; 5043 BackRef(int groupCount) { 5044 super(); 5045 groupIndex = groupCount + groupCount; 5046 } 5047 boolean match(Matcher matcher, int i, CharSequence seq) { 5048 int j = matcher.groups[groupIndex]; 5049 int k = matcher.groups[groupIndex+1]; 5050 5051 int groupSize = k - j; 5052 // If the referenced group didn't match, neither can this 5053 if (j < 0) 5054 return false; 5055 5056 // If there isn't enough input left no match 5057 if (i + groupSize > matcher.to) { 5058 matcher.hitEnd = true; 5059 return false; 5060 } 5061 // Check each new char to make sure it matches what the group 5062 // referenced matched last time around 5063 for (int index=0; index<groupSize; index++) 5064 if (seq.charAt(i+index) != seq.charAt(j+index)) 5065 return false; 5066 5067 return next.match(matcher, i+groupSize, seq); 5068 } 5069 boolean study(TreeInfo info) { 5070 info.maxValid = false; 5071 return next.study(info); 5072 } 5073 } 5074 5075 static class CIBackRef extends Node { 5076 int groupIndex; 5077 boolean doUnicodeCase; 5078 CIBackRef(int groupCount, boolean doUnicodeCase) { 5079 super(); 5080 groupIndex = groupCount + groupCount; 5081 this.doUnicodeCase = doUnicodeCase; 5082 } 5083 boolean match(Matcher matcher, int i, CharSequence seq) { 5084 int j = matcher.groups[groupIndex]; 5085 int k = matcher.groups[groupIndex+1]; 5086 5087 int groupSize = k - j; 5088 5089 // If the referenced group didn't match, neither can this 5090 if (j < 0) 5091 return false; 5092 5093 // If there isn't enough input left no match 5094 if (i + groupSize > matcher.to) { 5095 matcher.hitEnd = true; 5096 return false; 5097 } 5098 5099 // Check each new char to make sure it matches what the group 5100 // referenced matched last time around 5101 int x = i; 5102 for (int index=0; index<groupSize; index++) { 5103 int c1 = Character.codePointAt(seq, x); 5104 int c2 = Character.codePointAt(seq, j); 5105 if (c1 != c2) { 5106 if (doUnicodeCase) { 5107 int cc1 = Character.toUpperCase(c1); 5108 int cc2 = Character.toUpperCase(c2); 5109 if (cc1 != cc2 && 5110 Character.toLowerCase(cc1) != 5111 Character.toLowerCase(cc2)) 5112 return false; 5113 } else { 5114 if (ASCII.toLower(c1) != ASCII.toLower(c2)) 5115 return false; 5116 } 5117 } 5118 x += Character.charCount(c1); 5119 j += Character.charCount(c2); 5120 } 5121 5122 return next.match(matcher, i+groupSize, seq); 5123 } 5124 boolean study(TreeInfo info) { 5125 info.maxValid = false; 5126 return next.study(info); 5127 } 5128 } 5129 5130 /** 5131 * Searches until the next instance of its atom. This is useful for 5132 * finding the atom efficiently without passing an instance of it 5133 * (greedy problem) and without a lot of wasted search time (reluctant 5134 * problem). 5135 */ 5136 static final class First extends Node { 5137 Node atom; 5138 First(Node node) { 5139 this.atom = BnM.optimize(node); 5140 } 5141 boolean match(Matcher matcher, int i, CharSequence seq) { 5142 if (atom instanceof BnM) { 5143 return atom.match(matcher, i, seq) 5144 && next.match(matcher, matcher.last, seq); 5145 } 5146 for (;;) { 5147 if (i > matcher.to) { 5148 matcher.hitEnd = true; 5149 return false; 5150 } 5151 if (atom.match(matcher, i, seq)) { 5152 return next.match(matcher, matcher.last, seq); 5153 } 5154 i += countChars(seq, i, 1); 5155 matcher.first++; 5156 } 5157 } 5158 boolean study(TreeInfo info) { 5159 atom.study(info); 5160 info.maxValid = false; 5161 info.deterministic = false; 5162 return next.study(info); 5163 } 5164 } 5165 5166 static final class Conditional extends Node { 5167 Node cond, yes, not; 5168 Conditional(Node cond, Node yes, Node not) { 5169 this.cond = cond; 5170 this.yes = yes; 5171 this.not = not; 5172 } 5173 boolean match(Matcher matcher, int i, CharSequence seq) { 5174 if (cond.match(matcher, i, seq)) { 5175 return yes.match(matcher, i, seq); 5176 } else { 5177 return not.match(matcher, i, seq); 5178 } 5179 } 5180 boolean study(TreeInfo info) { 5181 int minL = info.minLength; 5182 int maxL = info.maxLength; 5183 boolean maxV = info.maxValid; 5184 info.reset(); 5185 yes.study(info); 5186 5187 int minL2 = info.minLength; 5188 int maxL2 = info.maxLength; 5189 boolean maxV2 = info.maxValid; 5190 info.reset(); 5191 not.study(info); 5192 5193 info.minLength = minL + Math.min(minL2, info.minLength); 5194 info.maxLength = maxL + Math.max(maxL2, info.maxLength); 5195 info.maxValid = (maxV & maxV2 & info.maxValid); 5196 info.deterministic = false; 5197 return next.study(info); 5198 } 5199 } 5200 5201 /** 5202 * Zero width positive lookahead. 5203 */ 5204 static final class Pos extends Node { 5205 Node cond; 5206 Pos(Node cond) { 5207 this.cond = cond; 5208 } 5209 boolean match(Matcher matcher, int i, CharSequence seq) { 5210 int savedTo = matcher.to; 5211 boolean conditionMatched = false; 5212 5213 // Relax transparent region boundaries for lookahead 5214 if (matcher.transparentBounds) 5215 matcher.to = matcher.getTextLength(); 5216 try { 5217 conditionMatched = cond.match(matcher, i, seq); 5218 } finally { 5219 // Reinstate region boundaries 5220 matcher.to = savedTo; 5221 } 5222 return conditionMatched && next.match(matcher, i, seq); 5223 } 5224 } 5225 5226 /** 5227 * Zero width negative lookahead. 5228 */ 5229 static final class Neg extends Node { 5230 Node cond; 5231 Neg(Node cond) { 5232 this.cond = cond; 5233 } 5234 boolean match(Matcher matcher, int i, CharSequence seq) { 5235 int savedTo = matcher.to; 5236 boolean conditionMatched = false; 5237 5238 // Relax transparent region boundaries for lookahead 5239 if (matcher.transparentBounds) 5240 matcher.to = matcher.getTextLength(); 5241 try { 5242 if (i < matcher.to) { 5243 conditionMatched = !cond.match(matcher, i, seq); 5244 } else { 5245 // If a negative lookahead succeeds then more input 5246 // could cause it to fail! 5247 matcher.requireEnd = true; 5248 conditionMatched = !cond.match(matcher, i, seq); 5249 } 5250 } finally { 5251 // Reinstate region boundaries 5252 matcher.to = savedTo; 5253 } 5254 return conditionMatched && next.match(matcher, i, seq); 5255 } 5256 } 5257 5258 /** 5259 * For use with lookbehinds; matches the position where the lookbehind 5260 * was encountered. 5261 */ 5262 static Node lookbehindEnd = new Node() { 5263 boolean match(Matcher matcher, int i, CharSequence seq) { 5264 return i == matcher.lookbehindTo; 5265 } 5266 }; 5267 5268 /** 5269 * Zero width positive lookbehind. 5270 */ 5271 static class Behind extends Node { 5272 Node cond; 5273 int rmax, rmin; 5274 Behind(Node cond, int rmax, int rmin) { 5275 this.cond = cond; 5276 this.rmax = rmax; 5277 this.rmin = rmin; 5278 } 5279 5280 boolean match(Matcher matcher, int i, CharSequence seq) { 5281 int savedFrom = matcher.from; 5282 boolean conditionMatched = false; 5283 int startIndex = (!matcher.transparentBounds) ? 5284 matcher.from : 0; 5285 int from = Math.max(i - rmax, startIndex); 5286 // Set end boundary 5287 int savedLBT = matcher.lookbehindTo; 5288 matcher.lookbehindTo = i; 5289 // Relax transparent region boundaries for lookbehind 5290 if (matcher.transparentBounds) 5291 matcher.from = 0; 5292 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5293 conditionMatched = cond.match(matcher, j, seq); 5294 } 5295 matcher.from = savedFrom; 5296 matcher.lookbehindTo = savedLBT; 5297 return conditionMatched && next.match(matcher, i, seq); 5298 } 5299 } 5300 5301 /** 5302 * Zero width positive lookbehind, including supplementary 5303 * characters or unpaired surrogates. 5304 */ 5305 static final class BehindS extends Behind { 5306 BehindS(Node cond, int rmax, int rmin) { 5307 super(cond, rmax, rmin); 5308 } 5309 boolean match(Matcher matcher, int i, CharSequence seq) { 5310 int rmaxChars = countChars(seq, i, -rmax); 5311 int rminChars = countChars(seq, i, -rmin); 5312 int savedFrom = matcher.from; 5313 int startIndex = (!matcher.transparentBounds) ? 5314 matcher.from : 0; 5315 boolean conditionMatched = false; 5316 int from = Math.max(i - rmaxChars, startIndex); 5317 // Set end boundary 5318 int savedLBT = matcher.lookbehindTo; 5319 matcher.lookbehindTo = i; 5320 // Relax transparent region boundaries for lookbehind 5321 if (matcher.transparentBounds) 5322 matcher.from = 0; 5323 5324 for (int j = i - rminChars; 5325 !conditionMatched && j >= from; 5326 j -= j>from ? countChars(seq, j, -1) : 1) { 5327 conditionMatched = cond.match(matcher, j, seq); 5328 } 5329 matcher.from = savedFrom; 5330 matcher.lookbehindTo = savedLBT; 5331 return conditionMatched && next.match(matcher, i, seq); 5332 } 5333 } 5334 5335 /** 5336 * Zero width negative lookbehind. 5337 */ 5338 static class NotBehind extends Node { 5339 Node cond; 5340 int rmax, rmin; 5341 NotBehind(Node cond, int rmax, int rmin) { 5342 this.cond = cond; 5343 this.rmax = rmax; 5344 this.rmin = rmin; 5345 } 5346 5347 boolean match(Matcher matcher, int i, CharSequence seq) { 5348 int savedLBT = matcher.lookbehindTo; 5349 int savedFrom = matcher.from; 5350 boolean conditionMatched = false; 5351 int startIndex = (!matcher.transparentBounds) ? 5352 matcher.from : 0; 5353 int from = Math.max(i - rmax, startIndex); 5354 matcher.lookbehindTo = i; 5355 // Relax transparent region boundaries for lookbehind 5356 if (matcher.transparentBounds) 5357 matcher.from = 0; 5358 for (int j = i - rmin; !conditionMatched && j >= from; j--) { 5359 conditionMatched = cond.match(matcher, j, seq); 5360 } 5361 // Reinstate region boundaries 5362 matcher.from = savedFrom; 5363 matcher.lookbehindTo = savedLBT; 5364 return !conditionMatched && next.match(matcher, i, seq); 5365 } 5366 } 5367 5368 /** 5369 * Zero width negative lookbehind, including supplementary 5370 * characters or unpaired surrogates. 5371 */ 5372 static final class NotBehindS extends NotBehind { 5373 NotBehindS(Node cond, int rmax, int rmin) { 5374 super(cond, rmax, rmin); 5375 } 5376 boolean match(Matcher matcher, int i, CharSequence seq) { 5377 int rmaxChars = countChars(seq, i, -rmax); 5378 int rminChars = countChars(seq, i, -rmin); 5379 int savedFrom = matcher.from; 5380 int savedLBT = matcher.lookbehindTo; 5381 boolean conditionMatched = false; 5382 int startIndex = (!matcher.transparentBounds) ? 5383 matcher.from : 0; 5384 int from = Math.max(i - rmaxChars, startIndex); 5385 matcher.lookbehindTo = i; 5386 // Relax transparent region boundaries for lookbehind 5387 if (matcher.transparentBounds) 5388 matcher.from = 0; 5389 for (int j = i - rminChars; 5390 !conditionMatched && j >= from; 5391 j -= j>from ? countChars(seq, j, -1) : 1) { 5392 conditionMatched = cond.match(matcher, j, seq); 5393 } 5394 //Reinstate region boundaries 5395 matcher.from = savedFrom; 5396 matcher.lookbehindTo = savedLBT; 5397 return !conditionMatched && next.match(matcher, i, seq); 5398 } 5399 } 5400 5401 /** 5402 * Handles word boundaries. Includes a field to allow this one class to 5403 * deal with the different types of word boundaries we can match. The word 5404 * characters include underscores, letters, and digits. Non spacing marks 5405 * can are also part of a word if they have a base character, otherwise 5406 * they are ignored for purposes of finding word boundaries. 5407 */ 5408 static final class Bound extends Node { 5409 static int LEFT = 0x1; 5410 static int RIGHT= 0x2; 5411 static int BOTH = 0x3; 5412 static int NONE = 0x4; 5413 int type; 5414 boolean useUWORD; 5415 Bound(int n, boolean useUWORD) { 5416 type = n; 5417 this.useUWORD = useUWORD; 5418 } 5419 5420 boolean isWord(int ch) { 5421 return useUWORD ? CharPredicates.WORD().is(ch) 5422 : (ch == '_' || Character.isLetterOrDigit(ch)); 5423 } 5424 5425 int check(Matcher matcher, int i, CharSequence seq) { 5426 int ch; 5427 boolean left = false; 5428 int startIndex = matcher.from; 5429 int endIndex = matcher.to; 5430 if (matcher.transparentBounds) { 5431 startIndex = 0; 5432 endIndex = matcher.getTextLength(); 5433 } 5434 if (i > startIndex) { 5435 ch = Character.codePointBefore(seq, i); 5436 left = (isWord(ch) || 5437 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5438 && hasBaseCharacter(matcher, i-1, seq))); 5439 } 5440 boolean right = false; 5441 if (i < endIndex) { 5442 ch = Character.codePointAt(seq, i); 5443 right = (isWord(ch) || 5444 ((Character.getType(ch) == Character.NON_SPACING_MARK) 5445 && hasBaseCharacter(matcher, i, seq))); 5446 } else { 5447 // Tried to access char past the end 5448 matcher.hitEnd = true; 5449 // The addition of another char could wreck a boundary 5450 matcher.requireEnd = true; 5451 } 5452 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE); 5453 } 5454 boolean match(Matcher matcher, int i, CharSequence seq) { 5455 return (check(matcher, i, seq) & type) > 0 5456 && next.match(matcher, i, seq); 5457 } 5458 } 5459 5460 /** 5461 * Non spacing marks only count as word characters in bounds calculations 5462 * if they have a base character. 5463 */ 5464 private static boolean hasBaseCharacter(Matcher matcher, int i, 5465 CharSequence seq) 5466 { 5467 int start = (!matcher.transparentBounds) ? 5468 matcher.from : 0; 5469 for (int x=i; x >= start; x--) { 5470 int ch = Character.codePointAt(seq, x); 5471 if (Character.isLetterOrDigit(ch)) 5472 return true; 5473 if (Character.getType(ch) == Character.NON_SPACING_MARK) 5474 continue; 5475 return false; 5476 } 5477 return false; 5478 } 5479 5480 /** 5481 * Attempts to match a slice in the input using the Boyer-Moore string 5482 * matching algorithm. The algorithm is based on the idea that the 5483 * pattern can be shifted farther ahead in the search text if it is 5484 * matched right to left. 5485 * <p> 5486 * The pattern is compared to the input one character at a time, from 5487 * the rightmost character in the pattern to the left. If the characters 5488 * all match the pattern has been found. If a character does not match, 5489 * the pattern is shifted right a distance that is the maximum of two 5490 * functions, the bad character shift and the good suffix shift. This 5491 * shift moves the attempted match position through the input more 5492 * quickly than a naive one position at a time check. 5493 * <p> 5494 * The bad character shift is based on the character from the text that 5495 * did not match. If the character does not appear in the pattern, the 5496 * pattern can be shifted completely beyond the bad character. If the 5497 * character does occur in the pattern, the pattern can be shifted to 5498 * line the pattern up with the next occurrence of that character. 5499 * <p> 5500 * The good suffix shift is based on the idea that some subset on the right 5501 * side of the pattern has matched. When a bad character is found, the 5502 * pattern can be shifted right by the pattern length if the subset does 5503 * not occur again in pattern, or by the amount of distance to the 5504 * next occurrence of the subset in the pattern. 5505 * 5506 * Boyer-Moore search methods adapted from code by Amy Yu. 5507 */ 5508 static class BnM extends Node { 5509 int[] buffer; 5510 int[] lastOcc; 5511 int[] optoSft; 5512 5513 /** 5514 * Pre calculates arrays needed to generate the bad character 5515 * shift and the good suffix shift. Only the last seven bits 5516 * are used to see if chars match; This keeps the tables small 5517 * and covers the heavily used ASCII range, but occasionally 5518 * results in an aliased match for the bad character shift. 5519 */ 5520 static Node optimize(Node node) { 5521 if (!(node instanceof Slice)) { 5522 return node; 5523 } 5524 5525 int[] src = ((Slice) node).buffer; 5526 int patternLength = src.length; 5527 // The BM algorithm requires a bit of overhead; 5528 // If the pattern is short don't use it, since 5529 // a shift larger than the pattern length cannot 5530 // be used anyway. 5531 if (patternLength < 4) { 5532 return node; 5533 } 5534 int i, j, k; 5535 int[] lastOcc = new int[128]; 5536 int[] optoSft = new int[patternLength]; 5537 // Precalculate part of the bad character shift 5538 // It is a table for where in the pattern each 5539 // lower 7-bit value occurs 5540 for (i = 0; i < patternLength; i++) { 5541 lastOcc[src[i]&0x7F] = i + 1; 5542 } 5543 // Precalculate the good suffix shift 5544 // i is the shift amount being considered 5545 NEXT: for (i = patternLength; i > 0; i--) { 5546 // j is the beginning index of suffix being considered 5547 for (j = patternLength - 1; j >= i; j--) { 5548 // Testing for good suffix 5549 if (src[j] == src[j-i]) { 5550 // src[j..len] is a good suffix 5551 optoSft[j-1] = i; 5552 } else { 5553 // No match. The array has already been 5554 // filled up with correct values before. 5555 continue NEXT; 5556 } 5557 } 5558 // This fills up the remaining of optoSft 5559 // any suffix can not have larger shift amount 5560 // then its sub-suffix. Why??? 5561 while (j > 0) { 5562 optoSft[--j] = i; 5563 } 5564 } 5565 // Set the guard value because of unicode compression 5566 optoSft[patternLength-1] = 1; 5567 if (node instanceof SliceS) 5568 return new BnMS(src, lastOcc, optoSft, node.next); 5569 return new BnM(src, lastOcc, optoSft, node.next); 5570 } 5571 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5572 this.buffer = src; 5573 this.lastOcc = lastOcc; 5574 this.optoSft = optoSft; 5575 this.next = next; 5576 } 5577 boolean match(Matcher matcher, int i, CharSequence seq) { 5578 int[] src = buffer; 5579 int patternLength = src.length; 5580 int last = matcher.to - patternLength; 5581 5582 // Loop over all possible match positions in text 5583 NEXT: while (i <= last) { 5584 // Loop over pattern from right to left 5585 for (int j = patternLength - 1; j >= 0; j--) { 5586 int ch = seq.charAt(i+j); 5587 if (ch != src[j]) { 5588 // Shift search to the right by the maximum of the 5589 // bad character shift and the good suffix shift 5590 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]); 5591 continue NEXT; 5592 } 5593 } 5594 // Entire pattern matched starting at i 5595 matcher.first = i; 5596 boolean ret = next.match(matcher, i + patternLength, seq); 5597 if (ret) { 5598 matcher.first = i; 5599 matcher.groups[0] = matcher.first; 5600 matcher.groups[1] = matcher.last; 5601 return true; 5602 } 5603 i++; 5604 } 5605 // BnM is only used as the leading node in the unanchored case, 5606 // and it replaced its Start() which always searches to the end 5607 // if it doesn't find what it's looking for, so hitEnd is true. 5608 matcher.hitEnd = true; 5609 return false; 5610 } 5611 boolean study(TreeInfo info) { 5612 info.minLength += buffer.length; 5613 info.maxValid = false; 5614 return next.study(info); 5615 } 5616 } 5617 5618 /** 5619 * Supplementary support version of BnM(). Unpaired surrogates are 5620 * also handled by this class. 5621 */ 5622 static final class BnMS extends BnM { 5623 int lengthInChars; 5624 5625 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) { 5626 super(src, lastOcc, optoSft, next); 5627 for (int cp : buffer) { 5628 lengthInChars += Character.charCount(cp); 5629 } 5630 } 5631 boolean match(Matcher matcher, int i, CharSequence seq) { 5632 int[] src = buffer; 5633 int patternLength = src.length; 5634 int last = matcher.to - lengthInChars; 5635 5636 // Loop over all possible match positions in text 5637 NEXT: while (i <= last) { 5638 // Loop over pattern from right to left 5639 int ch; 5640 for (int j = countChars(seq, i, patternLength), x = patternLength - 1; 5641 j > 0; j -= Character.charCount(ch), x--) { 5642 ch = Character.codePointBefore(seq, i+j); 5643 if (ch != src[x]) { 5644 // Shift search to the right by the maximum of the 5645 // bad character shift and the good suffix shift 5646 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]); 5647 i += countChars(seq, i, n); 5648 continue NEXT; 5649 } 5650 } 5651 // Entire pattern matched starting at i 5652 matcher.first = i; 5653 boolean ret = next.match(matcher, i + lengthInChars, seq); 5654 if (ret) { 5655 matcher.first = i; 5656 matcher.groups[0] = matcher.first; 5657 matcher.groups[1] = matcher.last; 5658 return true; 5659 } 5660 i += countChars(seq, i, 1); 5661 } 5662 matcher.hitEnd = true; 5663 return false; 5664 } 5665 } 5666 5667 @FunctionalInterface 5668 static interface CharPredicate { 5669 boolean is(int ch); 5670 5671 default CharPredicate and(CharPredicate p) { 5672 return ch -> is(ch) && p.is(ch); 5673 } 5674 default CharPredicate union(CharPredicate p) { 5675 return ch -> is(ch) || p.is(ch); 5676 } 5677 default CharPredicate union(CharPredicate p1, 5678 CharPredicate p2 ) { 5679 return ch -> is(ch) || p1.is(ch) || p2.is(ch); 5680 } 5681 default CharPredicate negate() { 5682 return ch -> !is(ch); 5683 } 5684 } 5685 5686 static interface BmpCharPredicate extends CharPredicate { 5687 5688 default CharPredicate and(CharPredicate p) { 5689 if(p instanceof BmpCharPredicate) 5690 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch)); 5691 return ch -> is(ch) && p.is(ch); 5692 } 5693 default CharPredicate union(CharPredicate p) { 5694 if (p instanceof BmpCharPredicate) 5695 return (BmpCharPredicate)(ch -> is(ch) || p.is(ch)); 5696 return ch -> is(ch) || p.is(ch); 5697 } 5698 static CharPredicate union(CharPredicate... predicates) { 5699 CharPredicate cp = ch -> { 5700 for (CharPredicate p : predicates) { 5701 if (!p.is(ch)) 5702 return false; 5703 } 5704 return true; 5705 }; 5706 for (CharPredicate p : predicates) { 5707 if (! (p instanceof BmpCharPredicate)) 5708 return cp; 5709 } 5710 return (BmpCharPredicate)cp; 5711 } 5712 } 5713 5714 /** 5715 * matches a Perl vertical whitespace 5716 */ 5717 static BmpCharPredicate VertWS() { 5718 return cp -> (cp >= 0x0A && cp <= 0x0D) || 5719 cp == 0x85 || cp == 0x2028 || cp == 0x2029; 5720 } 5721 5722 /** 5723 * matches a Perl horizontal whitespace 5724 */ 5725 static BmpCharPredicate HorizWS() { 5726 return cp -> 5727 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 || 5728 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f || 5729 cp == 0x205f || cp == 0x3000; 5730 } 5731 5732 /** 5733 * for the Unicode category ALL and the dot metacharacter when 5734 * in dotall mode. 5735 */ 5736 static CharPredicate ALL() { 5737 return ch -> true; 5738 } 5739 5740 /** 5741 * for the dot metacharacter when dotall is not enabled. 5742 */ 5743 static CharPredicate DOT() { 5744 return ch -> 5745 (ch != '\n' && ch != '\r' 5746 && (ch|1) != '\u2029' 5747 && ch != '\u0085'); 5748 } 5749 5750 /** 5751 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled. 5752 */ 5753 static CharPredicate UNIXDOT() { 5754 return ch -> ch != '\n'; 5755 } 5756 5757 /** 5758 * Indicate that matches a Supplementary Unicode character 5759 */ 5760 static CharPredicate SingleS(int c) { 5761 return ch -> ch == c; 5762 } 5763 5764 /** 5765 * A bmp/optimized predicate of single 5766 */ 5767 static BmpCharPredicate Single(int c) { 5768 return ch -> ch == c; 5769 } 5770 5771 /** 5772 * Case insensitive matches a given BMP character 5773 */ 5774 static BmpCharPredicate SingleI(int lower, int upper) { 5775 return ch -> ch == lower || ch == upper; 5776 } 5777 5778 /** 5779 * Unicode case insensitive matches a given Unicode character 5780 */ 5781 static CharPredicate SingleU(int lower) { 5782 return ch -> lower == ch || 5783 lower == Character.toLowerCase(Character.toUpperCase(ch)); 5784 } 5785 5786 private static boolean inRange(int lower, int ch, int upper) { 5787 return lower <= ch && ch <= upper; 5788 } 5789 5790 /** 5791 * Charactrs within a explicit value range 5792 */ 5793 static CharPredicate Range(int lower, int upper) { 5794 if (upper < Character.MIN_HIGH_SURROGATE || 5795 lower > Character.MAX_HIGH_SURROGATE && 5796 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT) 5797 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper)); 5798 return ch -> inRange(lower, ch, upper); 5799 } 5800 5801 /** 5802 * Charactrs within a explicit value range in a case insensitive manner. 5803 */ 5804 static CharPredicate CIRange(int lower, int upper) { 5805 return ch -> inRange(lower, ch, upper) || 5806 ASCII.isAscii(ch) && 5807 (inRange(lower, ASCII.toUpper(ch), upper) || 5808 inRange(lower, ASCII.toLower(ch), upper)); 5809 } 5810 5811 static CharPredicate CIRangeU(int lower, int upper) { 5812 return ch -> { 5813 if (inRange(lower, ch, upper)) 5814 return true; 5815 int up = Character.toUpperCase(ch); 5816 return inRange(lower, up, upper) || 5817 inRange(lower, Character.toLowerCase(up), upper); 5818 }; 5819 } 5820 5821 /** 5822 * This must be the very first initializer. 5823 */ 5824 static final Node accept = new Node(); 5825 5826 static final Node lastAccept = new LastNode(); 5827 5828 /** 5829 * Creates a predicate that tests if this pattern is found in a given input 5830 * string. 5831 * 5832 * @apiNote 5833 * This method creates a predicate that behaves as if it creates a matcher 5834 * from the input sequence and then calls {@code find}, for example a 5835 * predicate of the form: 5836 * <pre>{@code 5837 * s -> matcher(s).find(); 5838 * }</pre> 5839 * 5840 * @return The predicate which can be used for finding a match on a 5841 * subsequence of a string 5842 * @since 1.8 5843 * @see Matcher#find 5844 */ 5845 public Predicate<String> asPredicate() { 5846 return s -> matcher(s).find(); 5847 } 5848 5849 /** 5850 * Creates a predicate that tests if this pattern matches a given input string. 5851 * 5852 * @apiNote 5853 * This method creates a predicate that behaves as if it creates a matcher 5854 * from the input sequence and then calls {@code matches}, for example a 5855 * predicate of the form: 5856 * <pre>{@code 5857 * s -> matcher(s).matches(); 5858 * }</pre> 5859 * 5860 * @return The predicate which can be used for matching an input string 5861 * against this pattern. 5862 * @since 11 5863 * @see Matcher#matches 5864 */ 5865 public Predicate<String> asMatchPredicate() { 5866 return s -> matcher(s).matches(); 5867 } 5868 5869 /** 5870 * Creates a stream from the given input sequence around matches of this 5871 * pattern. 5872 * 5873 * <p> The stream returned by this method contains each substring of the 5874 * input sequence that is terminated by another subsequence that matches 5875 * this pattern or is terminated by the end of the input sequence. The 5876 * substrings in the stream are in the order in which they occur in the 5877 * input. Trailing empty strings will be discarded and not encountered in 5878 * the stream. 5879 * 5880 * <p> If this pattern does not match any subsequence of the input then 5881 * the resulting stream has just one element, namely the input sequence in 5882 * string form. 5883 * 5884 * <p> When there is a positive-width match at the beginning of the input 5885 * sequence then an empty leading substring is included at the beginning 5886 * of the stream. A zero-width match at the beginning however never produces 5887 * such empty leading substring. 5888 * 5889 * <p> If the input sequence is mutable, it must remain constant during the 5890 * execution of the terminal stream operation. Otherwise, the result of the 5891 * terminal stream operation is undefined. 5892 * 5893 * @param input 5894 * The character sequence to be split 5895 * 5896 * @return The stream of strings computed by splitting the input 5897 * around matches of this pattern 5898 * @see #split(CharSequence) 5899 * @since 1.8 5900 */ 5901 public Stream<String> splitAsStream(final CharSequence input) { 5902 class MatcherIterator implements Iterator<String> { 5903 private Matcher matcher; 5904 // The start position of the next sub-sequence of input 5905 // when current == input.length there are no more elements 5906 private int current; 5907 // null if the next element, if any, needs to obtained 5908 private String nextElement; 5909 // > 0 if there are N next empty elements 5910 private int emptyElementCount; 5911 5912 public String next() { 5913 if (!hasNext()) 5914 throw new NoSuchElementException(); 5915 5916 if (emptyElementCount == 0) { 5917 String n = nextElement; 5918 nextElement = null; 5919 return n; 5920 } else { 5921 emptyElementCount--; 5922 return ""; 5923 } 5924 } 5925 5926 public boolean hasNext() { 5927 if (matcher == null) { 5928 matcher = matcher(input); 5929 // If the input is an empty string then the result can only be a 5930 // stream of the input. Induce that by setting the empty 5931 // element count to 1 5932 emptyElementCount = input.length() == 0 ? 1 : 0; 5933 } 5934 if (nextElement != null || emptyElementCount > 0) 5935 return true; 5936 5937 if (current == input.length()) 5938 return false; 5939 5940 // Consume the next matching element 5941 // Count sequence of matching empty elements 5942 while (matcher.find()) { 5943 nextElement = input.subSequence(current, matcher.start()).toString(); 5944 current = matcher.end(); 5945 if (!nextElement.isEmpty()) { 5946 return true; 5947 } else if (current > 0) { // no empty leading substring for zero-width 5948 // match at the beginning of the input 5949 emptyElementCount++; 5950 } 5951 } 5952 5953 // Consume last matching element 5954 nextElement = input.subSequence(current, input.length()).toString(); 5955 current = input.length(); 5956 if (!nextElement.isEmpty()) { 5957 return true; 5958 } else { 5959 // Ignore a terminal sequence of matching empty elements 5960 emptyElementCount = 0; 5961 nextElement = null; 5962 return false; 5963 } 5964 } 5965 } 5966 return StreamSupport.stream(Spliterators.spliteratorUnknownSize( 5967 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false); 5968 } 5969 }