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