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