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