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 = temp[cursor+1];
3275 if (ASCII.isDigit(ch)) {
3276 skip();
3277 int cmin = 0;
3278 do {
3279 cmin = cmin * 10 + (ch - '0');
3280 } while (ASCII.isDigit(ch = read()));
3281 int cmax = cmin;
3282 if (ch == ',') {
3283 ch = read();
3284 cmax = MAX_REPS;
3285 if (ch != '}') {
3286 cmax = 0;
3287 while (ASCII.isDigit(ch)) {
3288 cmax = cmax * 10 + (ch - '0');
3289 ch = read();
3290 }
3291 }
3292 }
3293 if (ch != '}')
3294 throw error("Unclosed counted closure");
3295 if (((cmin) | (cmax) | (cmax - cmin)) < 0)
3296 throw error("Illegal repetition range");
3297 Curly curly;
3298 ch = peek();
3299 if (ch == '?') {
3300 next();
3301 curly = new Curly(prev, cmin, cmax, Qtype.LAZY);
3302 } else if (ch == '+') {
3303 next();
3304 curly = new Curly(prev, cmin, cmax, Qtype.POSSESSIVE);
3305 } else {
3306 curly = new Curly(prev, cmin, cmax, Qtype.GREEDY);
3307 }
3308 return curly;
3309 } else {
3310 throw error("Illegal repetition");
3311 }
3312 default:
3313 return prev;
3314 }
3315 }
3316
3317 /**
3318 * Utility method for parsing control escape sequences.
3319 */
3320 private int c() {
3321 if (cursor < patternLength) {
3322 return read() ^ 64;
3323 }
3324 throw error("Illegal control escape sequence");
3325 }
3326
3327 /**
3328 * Utility method for parsing octal escape sequences.
3329 */
3330 private int o() {
3331 int n = read();
3332 if (((n-'0')|('7'-n)) >= 0) {
3333 int m = read();
3334 if (((m-'0')|('7'-m)) >= 0) {
3335 int o = read();
3336 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) {
3337 return (n - '0') * 64 + (m - '0') * 8 + (o - '0');
3338 }
3339 unread();
3340 return (n - '0') * 8 + (m - '0');
3341 }
3342 unread();
3343 return (n - '0');
3344 }
3345 throw error("Illegal octal escape sequence");
3346 }
3347
3348 /**
3349 * Utility method for parsing hexadecimal escape sequences.
3350 */
3351 private int x() {
3352 int n = read();
3353 if (ASCII.isHexDigit(n)) {
3354 int m = read();
3355 if (ASCII.isHexDigit(m)) {
3356 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m);
3357 }
3358 } else if (n == '{' && ASCII.isHexDigit(peek())) {
3359 int ch = 0;
3360 while (ASCII.isHexDigit(n = read())) {
3361 ch = (ch << 4) + ASCII.toDigit(n);
3362 if (ch > Character.MAX_CODE_POINT)
3363 throw error("Hexadecimal codepoint is too big");
3364 }
3365 if (n != '}')
3366 throw error("Unclosed hexadecimal escape sequence");
3367 return ch;
3368 }
3369 throw error("Illegal hexadecimal escape sequence");
3370 }
3371
3372 /**
3373 * Utility method for parsing unicode escape sequences.
3374 */
3375 private int cursor() {
3376 return cursor;
3377 }
3378
3379 private void setcursor(int pos) {
3380 cursor = pos;
3381 }
3382
3383 private int uxxxx() {
3384 int n = 0;
3385 for (int i = 0; i < 4; i++) {
3386 int ch = read();
3387 if (!ASCII.isHexDigit(ch)) {
3388 throw error("Illegal Unicode escape sequence");
3389 }
3390 n = n * 16 + ASCII.toDigit(ch);
3391 }
3392 return n;
3393 }
3394
3395 private int u() {
3396 int n = uxxxx();
3397 if (Character.isHighSurrogate((char)n)) {
3398 int cur = cursor();
3399 if (read() == '\\' && read() == 'u') {
3400 int n2 = uxxxx();
3401 if (Character.isLowSurrogate((char)n2))
3402 return Character.toCodePoint((char)n, (char)n2);
3403 }
3404 setcursor(cur);
3405 }
3406 return n;
3407 }
3408
3409 private int N() {
3410 if (read() == '{') {
3411 int i = cursor;
3412 while (cursor < patternLength && read() != '}') {}
3413 if (cursor > patternLength)
3414 throw error("Unclosed character name escape sequence");
3415 String name = new String(temp, i, cursor - i - 1);
3416 try {
3417 return Character.codePointOf(name);
3418 } catch (IllegalArgumentException x) {
3419 throw error("Unknown character name [" + name + "]");
3420 }
3421 }
3422 throw error("Illegal character name escape sequence");
3423 }
3424
3425 //
3426 // Utility methods for code point support
3427 //
3428 private static final int countChars(CharSequence seq, int index,
3429 int lengthInCodePoints) {
3430 // optimization
3431 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) {
3432 assert (index >= 0 && index < seq.length());
3433 return 1;
3434 }
3435 int length = seq.length();
3436 int x = index;
3437 if (lengthInCodePoints >= 0) {
3438 assert (index >= 0 && index < length);
3439 for (int i = 0; x < length && i < lengthInCodePoints; i++) {
3440 if (Character.isHighSurrogate(seq.charAt(x++))) {
3441 if (x < length && Character.isLowSurrogate(seq.charAt(x))) {
3442 x++;
3443 }
3444 }
3445 }
3446 return x - index;
3447 }
3448
3449 assert (index >= 0 && index <= length);
3450 if (index == 0) {
3451 return 0;
3452 }
3453 int len = -lengthInCodePoints;
3454 for (int i = 0; x > 0 && i < len; i++) {
3455 if (Character.isLowSurrogate(seq.charAt(--x))) {
3456 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) {
3457 x--;
3458 }
3459 }
3460 }
3461 return index - x;
3462 }
3463
3464 private static final int countCodePoints(CharSequence seq) {
3465 int length = seq.length();
3466 int n = 0;
3467 for (int i = 0; i < length; ) {
3468 n++;
3469 if (Character.isHighSurrogate(seq.charAt(i++))) {
3470 if (i < length && Character.isLowSurrogate(seq.charAt(i))) {
3471 i++;
3472 }
3473 }
3474 }
3475 return n;
3476 }
3477
3478 /**
3479 * Creates a bit vector for matching Latin-1 values. A normal BitClass
3480 * never matches values above Latin-1, and a complemented BitClass always
3481 * matches values above Latin-1.
3482 */
3483 static final class BitClass extends BmpCharProperty {
3484 final boolean[] bits;
3485 BitClass() {
3486 this(new boolean[256]);
3487 }
3488 private BitClass(boolean[] bits) {
3489 super( ch -> ch < 256 && bits[ch]);
3490 this.bits = bits;
3491 }
3492 BitClass add(int c, int flags) {
3493 assert c >= 0 && c <= 255;
3494 if ((flags & CASE_INSENSITIVE) != 0) {
3495 if (ASCII.isAscii(c)) {
3496 bits[ASCII.toUpper(c)] = true;
3497 bits[ASCII.toLower(c)] = true;
3498 } else if ((flags & UNICODE_CASE) != 0) {
3499 bits[Character.toLowerCase(c)] = true;
3500 bits[Character.toUpperCase(c)] = true;
3501 }
3502 }
3503 bits[c] = true;
3504 return this;
3505 }
3506 }
3507
3508 /**
3509 * Utility method for creating a string slice matcher.
3510 */
3511 private Node newSlice(int[] buf, int count, boolean hasSupplementary) {
3512 int[] tmp = new int[count];
3513 if (has(CASE_INSENSITIVE)) {
3514 if (has(UNICODE_CASE)) {
3515 for (int i = 0; i < count; i++) {
3516 tmp[i] = Character.toLowerCase(
3517 Character.toUpperCase(buf[i]));
3518 }
3519 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp);
3520 }
3521 for (int i = 0; i < count; i++) {
3522 tmp[i] = ASCII.toLower(buf[i]);
3523 }
3524 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp);
3525 }
3526 for (int i = 0; i < count; i++) {
3527 tmp[i] = buf[i];
3528 }
3529 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp);
3530 }
3531
3532 /**
3533 * The following classes are the building components of the object
3534 * tree that represents a compiled regular expression. The object tree
3535 * is made of individual elements that handle constructs in the Pattern.
3536 * Each type of object knows how to match its equivalent construct with
3537 * the match() method.
3538 */
3539
3540 /**
3541 * Base class for all node classes. Subclasses should override the match()
3542 * method as appropriate. This class is an accepting node, so its match()
3543 * always returns true.
3544 */
3545 static class Node extends Object {
3546 Node next;
3547 Node() {
3548 next = Pattern.accept;
3549 }
3550 /**
3551 * This method implements the classic accept node.
3552 */
3553 boolean match(Matcher matcher, int i, CharSequence seq) {
3554 matcher.last = i;
3555 matcher.groups[0] = matcher.first;
3556 matcher.groups[1] = matcher.last;
3557 return true;
3558 }
3559 /**
3560 * This method is good for all zero length assertions.
3561 */
3562 boolean study(TreeInfo info) {
3563 if (next != null) {
3564 return next.study(info);
3565 } else {
3566 return info.deterministic;
3567 }
3568 }
3569 }
3570
3571 static class LastNode extends Node {
3572 /**
3573 * This method implements the classic accept node with
3574 * the addition of a check to see if the match occurred
3575 * using all of the input.
3576 */
3577 boolean match(Matcher matcher, int i, CharSequence seq) {
3578 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to)
3579 return false;
3580 matcher.last = i;
3581 matcher.groups[0] = matcher.first;
3582 matcher.groups[1] = matcher.last;
3583 return true;
3584 }
3585 }
3586
3587 /**
3588 * Used for REs that can start anywhere within the input string.
3589 * This basically tries to match repeatedly at each spot in the
3590 * input string, moving forward after each try. An anchored search
3591 * or a BnM will bypass this node completely.
3592 */
3593 static class Start extends Node {
3594 int minLength;
3595 Start(Node node) {
3596 this.next = node;
3597 TreeInfo info = new TreeInfo();
3598 next.study(info);
3599 minLength = info.minLength;
3600 }
3601 boolean match(Matcher matcher, int i, CharSequence seq) {
3602 if (i > matcher.to - minLength) {
3603 matcher.hitEnd = true;
3604 return false;
3605 }
3606 int guard = matcher.to - minLength;
3607 for (; i <= guard; i++) {
3608 if (next.match(matcher, i, seq)) {
3609 matcher.first = i;
3610 matcher.groups[0] = matcher.first;
3611 matcher.groups[1] = matcher.last;
3612 return true;
3613 }
3614 }
3615 matcher.hitEnd = true;
3616 return false;
3617 }
3618 boolean study(TreeInfo info) {
3619 next.study(info);
3620 info.maxValid = false;
3621 info.deterministic = false;
3622 return false;
3623 }
3624 }
3625
3626 /*
3627 * StartS supports supplementary characters, including unpaired surrogates.
3628 */
3629 static final class StartS extends Start {
3630 StartS(Node node) {
3631 super(node);
3632 }
3633 boolean match(Matcher matcher, int i, CharSequence seq) {
3634 if (i > matcher.to - minLength) {
3635 matcher.hitEnd = true;
3636 return false;
3637 }
3638 int guard = matcher.to - minLength;
3639 while (i <= guard) {
3640 //if ((ret = next.match(matcher, i, seq)) || i == guard)
3641 if (next.match(matcher, i, seq)) {
3642 matcher.first = i;
3643 matcher.groups[0] = matcher.first;
3644 matcher.groups[1] = matcher.last;
3645 return true;
3646 }
3647 if (i == guard)
3648 break;
3649 // Optimization to move to the next character. This is
3650 // faster than countChars(seq, i, 1).
3651 if (Character.isHighSurrogate(seq.charAt(i++))) {
3652 if (i < seq.length() &&
3653 Character.isLowSurrogate(seq.charAt(i))) {
3654 i++;
3655 }
3656 }
3657 }
3658 matcher.hitEnd = true;
3659 return false;
3660 }
3661 }
3662
3663 /**
3664 * Node to anchor at the beginning of input. This object implements the
3665 * match for a \A sequence, and the caret anchor will use this if not in
3666 * multiline mode.
3667 */
3668 static final class Begin extends Node {
3669 boolean match(Matcher matcher, int i, CharSequence seq) {
3670 int fromIndex = (matcher.anchoringBounds) ?
3671 matcher.from : 0;
3672 if (i == fromIndex && next.match(matcher, i, seq)) {
3673 matcher.first = i;
3674 matcher.groups[0] = i;
3675 matcher.groups[1] = matcher.last;
3676 return true;
3677 } else {
3678 return false;
3679 }
3680 }
3681 }
3682
3683 /**
3684 * Node to anchor at the end of input. This is the absolute end, so this
3685 * should not match at the last newline before the end as $ will.
3686 */
3687 static final class End extends Node {
3688 boolean match(Matcher matcher, int i, CharSequence seq) {
3689 int endIndex = (matcher.anchoringBounds) ?
3690 matcher.to : matcher.getTextLength();
3691 if (i == endIndex) {
3692 matcher.hitEnd = true;
3693 return next.match(matcher, i, seq);
3694 }
3695 return false;
3696 }
3697 }
3698
3699 /**
3700 * Node to anchor at the beginning of a line. This is essentially the
3701 * object to match for the multiline ^.
3702 */
3703 static final class Caret extends Node {
3704 boolean match(Matcher matcher, int i, CharSequence seq) {
3705 int startIndex = matcher.from;
3706 int endIndex = matcher.to;
3707 if (!matcher.anchoringBounds) {
3708 startIndex = 0;
3709 endIndex = matcher.getTextLength();
3710 }
3711 // Perl does not match ^ at end of input even after newline
3712 if (i == endIndex) {
3713 matcher.hitEnd = true;
3714 return false;
3715 }
3716 if (i > startIndex) {
3717 char ch = seq.charAt(i-1);
3718 if (ch != '\n' && ch != '\r'
3719 && (ch|1) != '\u2029'
3720 && ch != '\u0085' ) {
3721 return false;
3722 }
3723 // Should treat /r/n as one newline
3724 if (ch == '\r' && seq.charAt(i) == '\n')
3725 return false;
3726 }
3727 return next.match(matcher, i, seq);
3728 }
3729 }
3730
3731 /**
3732 * Node to anchor at the beginning of a line when in unixdot mode.
3733 */
3734 static final class UnixCaret extends Node {
3735 boolean match(Matcher matcher, int i, CharSequence seq) {
3736 int startIndex = matcher.from;
3737 int endIndex = matcher.to;
3738 if (!matcher.anchoringBounds) {
3739 startIndex = 0;
3740 endIndex = matcher.getTextLength();
3741 }
3742 // Perl does not match ^ at end of input even after newline
3743 if (i == endIndex) {
3744 matcher.hitEnd = true;
3745 return false;
3746 }
3747 if (i > startIndex) {
3748 char ch = seq.charAt(i-1);
3749 if (ch != '\n') {
3750 return false;
3751 }
3752 }
3753 return next.match(matcher, i, seq);
3754 }
3755 }
3756
3757 /**
3758 * Node to match the location where the last match ended.
3759 * This is used for the \G construct.
3760 */
3761 static final class LastMatch extends Node {
3762 boolean match(Matcher matcher, int i, CharSequence seq) {
3763 if (i != matcher.oldLast)
3764 return false;
3765 return next.match(matcher, i, seq);
3766 }
3767 }
3768
3769 /**
3770 * Node to anchor at the end of a line or the end of input based on the
3771 * multiline mode.
3772 *
3773 * When not in multiline mode, the $ can only match at the very end
3774 * of the input, unless the input ends in a line terminator in which
3775 * it matches right before the last line terminator.
3776 *
3777 * Note that \r\n is considered an atomic line terminator.
3778 *
3779 * Like ^ the $ operator matches at a position, it does not match the
3780 * line terminators themselves.
3781 */
3782 static final class Dollar extends Node {
3783 boolean multiline;
3784 Dollar(boolean mul) {
3785 multiline = mul;
3786 }
3787 boolean match(Matcher matcher, int i, CharSequence seq) {
3788 int endIndex = (matcher.anchoringBounds) ?
3789 matcher.to : matcher.getTextLength();
3790 if (!multiline) {
3791 if (i < endIndex - 2)
3792 return false;
3793 if (i == endIndex - 2) {
3794 char ch = seq.charAt(i);
3795 if (ch != '\r')
3796 return false;
3797 ch = seq.charAt(i + 1);
3798 if (ch != '\n')
3799 return false;
3800 }
3801 }
3802 // Matches before any line terminator; also matches at the
3803 // end of input
3804 // Before line terminator:
3805 // If multiline, we match here no matter what
3806 // If not multiline, fall through so that the end
3807 // is marked as hit; this must be a /r/n or a /n
3808 // at the very end so the end was hit; more input
3809 // could make this not match here
3810 if (i < endIndex) {
3811 char ch = seq.charAt(i);
3812 if (ch == '\n') {
3813 // No match between \r\n
3814 if (i > 0 && seq.charAt(i-1) == '\r')
3815 return false;
3816 if (multiline)
3817 return next.match(matcher, i, seq);
3818 } else if (ch == '\r' || ch == '\u0085' ||
3819 (ch|1) == '\u2029') {
3820 if (multiline)
3821 return next.match(matcher, i, seq);
3822 } else { // No line terminator, no match
3823 return false;
3824 }
3825 }
3826 // Matched at current end so hit end
3827 matcher.hitEnd = true;
3828 // If a $ matches because of end of input, then more input
3829 // could cause it to fail!
3830 matcher.requireEnd = true;
3831 return next.match(matcher, i, seq);
3832 }
3833 boolean study(TreeInfo info) {
3834 next.study(info);
3835 return info.deterministic;
3836 }
3837 }
3838
3839 /**
3840 * Node to anchor at the end of a line or the end of input based on the
3841 * multiline mode when in unix lines mode.
3842 */
3843 static final class UnixDollar extends Node {
3844 boolean multiline;
3845 UnixDollar(boolean mul) {
3846 multiline = mul;
3847 }
3848 boolean match(Matcher matcher, int i, CharSequence seq) {
3849 int endIndex = (matcher.anchoringBounds) ?
3850 matcher.to : matcher.getTextLength();
3851 if (i < endIndex) {
3852 char ch = seq.charAt(i);
3853 if (ch == '\n') {
3854 // If not multiline, then only possible to
3855 // match at very end or one before end
3856 if (multiline == false && i != endIndex - 1)
3857 return false;
3858 // If multiline return next.match without setting
3859 // matcher.hitEnd
3860 if (multiline)
3861 return next.match(matcher, i, seq);
3862 } else {
3863 return false;
3864 }
3865 }
3866 // Matching because at the end or 1 before the end;
3867 // more input could change this so set hitEnd
3868 matcher.hitEnd = true;
3869 // If a $ matches because of end of input, then more input
3870 // could cause it to fail!
3871 matcher.requireEnd = true;
3872 return next.match(matcher, i, seq);
3873 }
3874 boolean study(TreeInfo info) {
3875 next.study(info);
3876 return info.deterministic;
3877 }
3878 }
3879
3880 /**
3881 * Node class that matches a Unicode line ending '\R'
3882 */
3883 static final class LineEnding extends Node {
3884 boolean match(Matcher matcher, int i, CharSequence seq) {
3885 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029])
3886 if (i < matcher.to) {
3887 int ch = seq.charAt(i);
3888 if (ch == 0x0A || ch == 0x0B || ch == 0x0C ||
3889 ch == 0x85 || ch == 0x2028 || ch == 0x2029)
3890 return next.match(matcher, i + 1, seq);
3891 if (ch == 0x0D) {
3892 i++;
3893 if (i < matcher.to) {
3894 if (seq.charAt(i) == 0x0A &&
3895 next.match(matcher, i + 1, seq)) {
3896 return true;
3897 }
3898 } else {
3899 matcher.hitEnd = true;
3900 }
3901 return next.match(matcher, i, seq);
3902 }
3903 } else {
3904 matcher.hitEnd = true;
3905 }
3906 return false;
3907 }
3908 boolean study(TreeInfo info) {
3909 info.minLength++;
3910 info.maxLength += 2;
3911 return next.study(info);
3912 }
3913 }
3914
3915 /**
3916 * Abstract node class to match one character satisfying some
3917 * boolean property.
3918 */
3919 static class CharProperty extends Node {
3920 CharPredicate predicate;
3921
3922 CharProperty (CharPredicate predicate) {
3923 this.predicate = predicate;
3924 }
3925 boolean match(Matcher matcher, int i, CharSequence seq) {
3926 if (i < matcher.to) {
3927 int ch = Character.codePointAt(seq, i);
3928 return predicate.is(ch) &&
3929 next.match(matcher, i + Character.charCount(ch), seq);
3930 } else {
3931 matcher.hitEnd = true;
3932 return false;
3933 }
3934 }
3935 boolean study(TreeInfo info) {
3936 info.minLength++;
3937 info.maxLength++;
3938 return next.study(info);
3939 }
3940 }
3941
3942 /**
3943 * Optimized version of CharProperty that works only for
3944 * properties never satisfied by Supplementary characters.
3945 */
3946 private static class BmpCharProperty extends CharProperty {
3947 BmpCharProperty (BmpCharPredicate predicate) {
3948 super(predicate);
3949 }
3950 boolean match(Matcher matcher, int i, CharSequence seq) {
3951 if (i < matcher.to) {
3952 return predicate.is(seq.charAt(i)) &&
3953 next.match(matcher, i + 1, seq);
3954 } else {
3955 matcher.hitEnd = true;
3956 return false;
3957 }
3958 }
3959 }
3960
3961 private static class NFCCharProperty extends Node {
3962 CharPredicate predicate;
3963 NFCCharProperty (CharPredicate predicate) {
3964 this.predicate = predicate;
3965 }
3966
3967 boolean match(Matcher matcher, int i, CharSequence seq) {
3968 if (i < matcher.to) {
3969 int ch0 = Character.codePointAt(seq, i);
3970 int n = Character.charCount(ch0);
3971 int j = Grapheme.nextBoundary(seq, i, matcher.to);
3972 if (i + n == j) { // single, assume nfc cp
3973 if (predicate.is(ch0))
3974 return next.match(matcher, j, seq);
3975 } else {
3976 while (i + n < j) {
3977 String nfc = Normalizer.normalize(
3978 seq.toString().substring(i, j), Normalizer.Form.NFC);
3979 if (nfc.codePointCount(0, nfc.length()) == 1) {
3980 if (predicate.is(nfc.codePointAt(0)) &&
3981 next.match(matcher, j, seq)) {
3982 return true;
3983 }
3984 }
3985
3986 ch0 = Character.codePointBefore(seq, j);
3987 j -= Character.charCount(ch0);
3988 }
3989 }
3990 if (j < matcher.to)
3991 return false;
3992 }
3993 matcher.hitEnd = true;
3994 return false;
3995 }
3996
3997 boolean study(TreeInfo info) {
3998 info.minLength++;
3999 info.deterministic = false;
4000 return next.study(info);
4001 }
4002 }
4003
4004 /**
4005 * Node class that matches an unicode extended grapheme cluster
4006 */
4007 static class XGrapheme extends Node {
4008 boolean match(Matcher matcher, int i, CharSequence seq) {
4009 if (i < matcher.to) {
4010 i = Grapheme.nextBoundary(seq, i, matcher.to);
4011 return next.match(matcher, i, seq);
4012 }
4013 matcher.hitEnd = true;
4014 return false;
4015 }
4016
4017 boolean study(TreeInfo info) {
4018 info.minLength++;
4019 info.deterministic = false;
4020 return next.study(info);
4021 }
4022 }
4023
4024 /**
4025 * Node class that handles grapheme boundaries
4026 */
4027 static class GraphemeBound extends Node {
4028 boolean match(Matcher matcher, int i, CharSequence seq) {
4029 int startIndex = matcher.from;
4030 int endIndex = matcher.to;
4031 if (matcher.transparentBounds) {
4032 startIndex = 0;
4033 endIndex = matcher.getTextLength();
4034 }
4035 if (i == startIndex) {
4036 return next.match(matcher, i, seq);
4037 }
4038 if (i < endIndex) {
4039 if (Character.isSurrogatePair(seq.charAt(i-1), seq.charAt(i)) ||
4040 Grapheme.nextBoundary(seq,
4041 i - Character.charCount(Character.codePointBefore(seq, i)),
4042 i + Character.charCount(Character.codePointAt(seq, i))) > i) {
4043 return false;
4044 }
4045 } else {
4046 matcher.hitEnd = true;
4047 matcher.requireEnd = true;
4048 }
4049 return next.match(matcher, i, seq);
4050 }
4051 }
4052
4053 /**
4054 * Base class for all Slice nodes
4055 */
4056 static class SliceNode extends Node {
4057 int[] buffer;
4058 SliceNode(int[] buf) {
4059 buffer = buf;
4060 }
4061 boolean study(TreeInfo info) {
4062 info.minLength += buffer.length;
4063 info.maxLength += buffer.length;
4064 return next.study(info);
4065 }
4066 }
4067
4068 /**
4069 * Node class for a case sensitive/BMP-only sequence of literal
4070 * characters.
4071 */
4072 static class Slice extends SliceNode {
4073 Slice(int[] buf) {
4074 super(buf);
4075 }
4076 boolean match(Matcher matcher, int i, CharSequence seq) {
4077 int[] buf = buffer;
4078 int len = buf.length;
4079 for (int j=0; j<len; j++) {
4080 if ((i+j) >= matcher.to) {
4081 matcher.hitEnd = true;
4082 return false;
4083 }
4084 if (buf[j] != seq.charAt(i+j))
4085 return false;
4086 }
4087 return next.match(matcher, i+len, seq);
4088 }
4089 }
4090
4091 /**
4092 * Node class for a case_insensitive/BMP-only sequence of literal
4093 * characters.
4094 */
4095 static class SliceI extends SliceNode {
4096 SliceI(int[] buf) {
4097 super(buf);
4098 }
4099 boolean match(Matcher matcher, int i, CharSequence seq) {
4100 int[] buf = buffer;
4101 int len = buf.length;
4102 for (int j=0; j<len; j++) {
4103 if ((i+j) >= matcher.to) {
4104 matcher.hitEnd = true;
4105 return false;
4106 }
4107 int c = seq.charAt(i+j);
4108 if (buf[j] != c &&
4109 buf[j] != ASCII.toLower(c))
4110 return false;
4111 }
4112 return next.match(matcher, i+len, seq);
4113 }
4114 }
4115
4116 /**
4117 * Node class for a unicode_case_insensitive/BMP-only sequence of
4118 * literal characters. Uses unicode case folding.
4119 */
4120 static final class SliceU extends SliceNode {
4121 SliceU(int[] buf) {
4122 super(buf);
4123 }
4124 boolean match(Matcher matcher, int i, CharSequence seq) {
4125 int[] buf = buffer;
4126 int len = buf.length;
4127 for (int j=0; j<len; j++) {
4128 if ((i+j) >= matcher.to) {
4129 matcher.hitEnd = true;
4130 return false;
4131 }
4132 int c = seq.charAt(i+j);
4133 if (buf[j] != c &&
4134 buf[j] != Character.toLowerCase(Character.toUpperCase(c)))
4135 return false;
4136 }
4137 return next.match(matcher, i+len, seq);
4138 }
4139 }
4140
4141 /**
4142 * Node class for a case sensitive sequence of literal characters
4143 * including supplementary characters.
4144 */
4145 static final class SliceS extends Slice {
4146 SliceS(int[] buf) {
4147 super(buf);
4148 }
4149 boolean match(Matcher matcher, int i, CharSequence seq) {
4150 int[] buf = buffer;
4151 int x = i;
4152 for (int j = 0; j < buf.length; j++) {
4153 if (x >= matcher.to) {
4154 matcher.hitEnd = true;
4155 return false;
4156 }
4157 int c = Character.codePointAt(seq, x);
4158 if (buf[j] != c)
4159 return false;
4160 x += Character.charCount(c);
4161 if (x > matcher.to) {
4162 matcher.hitEnd = true;
4163 return false;
4164 }
4165 }
4166 return next.match(matcher, x, seq);
4167 }
4168 }
4169
4170 /**
4171 * Node class for a case insensitive sequence of literal characters
4172 * including supplementary characters.
4173 */
4174 static class SliceIS extends SliceNode {
4175 SliceIS(int[] buf) {
4176 super(buf);
4177 }
4178 int toLower(int c) {
4179 return ASCII.toLower(c);
4180 }
4181 boolean match(Matcher matcher, int i, CharSequence seq) {
4182 int[] buf = buffer;
4183 int x = i;
4184 for (int j = 0; j < buf.length; j++) {
4185 if (x >= matcher.to) {
4186 matcher.hitEnd = true;
4187 return false;
4188 }
4189 int c = Character.codePointAt(seq, x);
4190 if (buf[j] != c && buf[j] != toLower(c))
4191 return false;
4192 x += Character.charCount(c);
4193 if (x > matcher.to) {
4194 matcher.hitEnd = true;
4195 return false;
4196 }
4197 }
4198 return next.match(matcher, x, seq);
4199 }
4200 }
4201
4202 /**
4203 * Node class for a case insensitive sequence of literal characters.
4204 * Uses unicode case folding.
4205 */
4206 static final class SliceUS extends SliceIS {
4207 SliceUS(int[] buf) {
4208 super(buf);
4209 }
4210 int toLower(int c) {
4211 return Character.toLowerCase(Character.toUpperCase(c));
4212 }
4213 }
4214
4215 /**
4216 * The 0 or 1 quantifier. This one class implements all three types.
4217 */
4218 static final class Ques extends Node {
4219 Node atom;
4220 Qtype type;
4221 Ques(Node node, Qtype type) {
4222 this.atom = node;
4223 this.type = type;
4224 }
4225 boolean match(Matcher matcher, int i, CharSequence seq) {
4226 switch (type) {
4227 case GREEDY:
4228 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq))
4229 || next.match(matcher, i, seq);
4230 case LAZY:
4231 return next.match(matcher, i, seq)
4232 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq));
4233 case POSSESSIVE:
4234 if (atom.match(matcher, i, seq)) i = matcher.last;
4235 return next.match(matcher, i, seq);
4236 default:
4237 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq);
4238 }
4239 }
4240 boolean study(TreeInfo info) {
4241 if (type != Qtype.INDEPENDENT) {
4242 int minL = info.minLength;
4243 atom.study(info);
4244 info.minLength = minL;
4245 info.deterministic = false;
4246 return next.study(info);
4247 } else {
4248 atom.study(info);
4249 return next.study(info);
4250 }
4251 }
4252 }
4253
4254 /**
4255 * Handles the greedy style repetition with the minimum either be
4256 * 0 or 1 and the maximum be MAX_REPS, for * and + quantifier.
4257 */
4258 static class CharPropertyGreedy extends Node {
4259 final CharPredicate predicate;
4260 final int cmin;
4261
4262 CharPropertyGreedy(CharProperty cp, int cmin) {
4263 this.predicate = cp.predicate;
4264 this.cmin = cmin;
4265 }
4266 boolean match(Matcher matcher, int i, CharSequence seq) {
4267 int n = 0;
4268 int to = matcher.to;
4269 // greedy, all the way down
4270 while (i < to) {
4271 int ch = Character.codePointAt(seq, i);
4272 if (!predicate.is(ch))
4273 break;
4274 i += Character.charCount(ch);
4275 n++;
4276 }
4277 if (i >= to) {
4278 matcher.hitEnd = true;
4279 }
4280 while (n >= cmin) {
4281 if (next.match(matcher, i, seq))
4282 return true;
4283 if (n == cmin)
4284 return false;
4285 // backing off if match fails
4286 int ch = Character.codePointBefore(seq, i);
4287 i -= Character.charCount(ch);
4288 n--;
4289 }
4290 return false;
4291 }
4292
4293 boolean study(TreeInfo info) {
4294 info.minLength += cmin;
4295 if (info.maxValid) {
4296 info.maxLength += MAX_REPS;
4297 }
4298 info.deterministic = false;
4299 return next.study(info);
4300 }
4301 }
4302
4303 static final class BmpCharPropertyGreedy extends CharPropertyGreedy {
4304
4305 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) {
4306 super(bcp, cmin);
4307 }
4308
4309 boolean match(Matcher matcher, int i, CharSequence seq) {
4310 int n = 0;
4311 int to = matcher.to;
4312 while (i < to && predicate.is(seq.charAt(i))) {
4313 i++; n++;
4314 }
4315 if (i >= to) {
4316 matcher.hitEnd = true;
4317 }
4318 while (n >= cmin) {
4319 if (next.match(matcher, i, seq))
4320 return true;
4321 i--; n--; // backing off if match fails
4322 }
4323 return false;
4324 }
4325 }
4326
4327 /**
4328 * Handles the curly-brace style repetition with a specified minimum and
4329 * maximum occurrences. The * quantifier is handled as a special case.
4330 * This class handles the three types.
4331 */
4332 static final class Curly extends Node {
4333 Node atom;
4334 Qtype type;
4335 int cmin;
4336 int cmax;
4337
4338 Curly(Node node, int cmin, int cmax, Qtype type) {
4339 this.atom = node;
4340 this.type = type;
4341 this.cmin = cmin;
4342 this.cmax = cmax;
4343 }
4344 boolean match(Matcher matcher, int i, CharSequence seq) {
4345 int j;
4346 for (j = 0; j < cmin; j++) {
4347 if (atom.match(matcher, i, seq)) {
4348 i = matcher.last;
4349 continue;
4350 }
4351 return false;
4352 }
4353 if (type == Qtype.GREEDY)
4354 return match0(matcher, i, j, seq);
4355 else if (type == Qtype.LAZY)
4356 return match1(matcher, i, j, seq);
4357 else
4358 return match2(matcher, i, j, seq);
4359 }
4360 // Greedy match.
4361 // i is the index to start matching at
4362 // j is the number of atoms that have matched
4363 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4364 if (j >= cmax) {
4365 // We have matched the maximum... continue with the rest of
4366 // the regular expression
4367 return next.match(matcher, i, seq);
4368 }
4369 int backLimit = j;
4370 while (atom.match(matcher, i, seq)) {
4371 // k is the length of this match
4372 int k = matcher.last - i;
4373 if (k == 0) // Zero length match
4374 break;
4375 // Move up index and number matched
4376 i = matcher.last;
4377 j++;
4378 // We are greedy so match as many as we can
4379 while (j < cmax) {
4380 if (!atom.match(matcher, i, seq))
4381 break;
4382 if (i + k != matcher.last) {
4383 if (match0(matcher, matcher.last, j+1, seq))
4384 return true;
4385 break;
4386 }
4387 i += k;
4388 j++;
4389 }
4390 // Handle backing off if match fails
4391 while (j >= backLimit) {
4392 if (next.match(matcher, i, seq))
4393 return true;
4394 i -= k;
4395 j--;
4396 }
4397 return false;
4398 }
4399 return next.match(matcher, i, seq);
4400 }
4401 // Reluctant match. At this point, the minimum has been satisfied.
4402 // i is the index to start matching at
4403 // j is the number of atoms that have matched
4404 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4405 for (;;) {
4406 // Try finishing match without consuming any more
4407 if (next.match(matcher, i, seq))
4408 return true;
4409 // At the maximum, no match found
4410 if (j >= cmax)
4411 return false;
4412 // Okay, must try one more atom
4413 if (!atom.match(matcher, i, seq))
4414 return false;
4415 // If we haven't moved forward then must break out
4416 if (i == matcher.last)
4417 return false;
4418 // Move up index and number matched
4419 i = matcher.last;
4420 j++;
4421 }
4422 }
4423 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4424 for (; j < cmax; j++) {
4425 if (!atom.match(matcher, i, seq))
4426 break;
4427 if (i == matcher.last)
4428 break;
4429 i = matcher.last;
4430 }
4431 return next.match(matcher, i, seq);
4432 }
4433 boolean study(TreeInfo info) {
4434 // Save original info
4435 int minL = info.minLength;
4436 int maxL = info.maxLength;
4437 boolean maxV = info.maxValid;
4438 boolean detm = info.deterministic;
4439 info.reset();
4440
4441 atom.study(info);
4442
4443 int temp = info.minLength * cmin + minL;
4444 if (temp < minL) {
4445 temp = 0xFFFFFFF; // arbitrary large number
4446 }
4447 info.minLength = temp;
4448
4449 if (maxV & info.maxValid) {
4450 temp = info.maxLength * cmax + maxL;
4451 info.maxLength = temp;
4452 if (temp < maxL) {
4453 info.maxValid = false;
4454 }
4455 } else {
4456 info.maxValid = false;
4457 }
4458
4459 if (info.deterministic && cmin == cmax)
4460 info.deterministic = detm;
4461 else
4462 info.deterministic = false;
4463 return next.study(info);
4464 }
4465 }
4466
4467 /**
4468 * Handles the curly-brace style repetition with a specified minimum and
4469 * maximum occurrences in deterministic cases. This is an iterative
4470 * optimization over the Prolog and Loop system which would handle this
4471 * in a recursive way. The * quantifier is handled as a special case.
4472 * If capture is true then this class saves group settings and ensures
4473 * that groups are unset when backing off of a group match.
4474 */
4475 static final class GroupCurly extends Node {
4476 Node atom;
4477 Qtype type;
4478 int cmin;
4479 int cmax;
4480 int localIndex;
4481 int groupIndex;
4482 boolean capture;
4483
4484 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local,
4485 int group, boolean capture) {
4486 this.atom = node;
4487 this.type = type;
4488 this.cmin = cmin;
4489 this.cmax = cmax;
4490 this.localIndex = local;
4491 this.groupIndex = group;
4492 this.capture = capture;
4493 }
4494 boolean match(Matcher matcher, int i, CharSequence seq) {
4495 int[] groups = matcher.groups;
4496 int[] locals = matcher.locals;
4497 int save0 = locals[localIndex];
4498 int save1 = 0;
4499 int save2 = 0;
4500
4501 if (capture) {
4502 save1 = groups[groupIndex];
4503 save2 = groups[groupIndex+1];
4504 }
4505
4506 // Notify GroupTail there is no need to setup group info
4507 // because it will be set here
4508 locals[localIndex] = -1;
4509
4510 boolean ret = true;
4511 for (int j = 0; j < cmin; j++) {
4512 if (atom.match(matcher, i, seq)) {
4513 if (capture) {
4514 groups[groupIndex] = i;
4515 groups[groupIndex+1] = matcher.last;
4516 }
4517 i = matcher.last;
4518 } else {
4519 ret = false;
4520 break;
4521 }
4522 }
4523 if (ret) {
4524 if (type == Qtype.GREEDY) {
4525 ret = match0(matcher, i, cmin, seq);
4526 } else if (type == Qtype.LAZY) {
4527 ret = match1(matcher, i, cmin, seq);
4528 } else {
4529 ret = match2(matcher, i, cmin, seq);
4530 }
4531 }
4532 if (!ret) {
4533 locals[localIndex] = save0;
4534 if (capture) {
4535 groups[groupIndex] = save1;
4536 groups[groupIndex+1] = save2;
4537 }
4538 }
4539 return ret;
4540 }
4541 // Aggressive group match
4542 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4543 // don't back off passing the starting "j"
4544 int min = j;
4545 int[] groups = matcher.groups;
4546 int save0 = 0;
4547 int save1 = 0;
4548 if (capture) {
4549 save0 = groups[groupIndex];
4550 save1 = groups[groupIndex+1];
4551 }
4552 for (;;) {
4553 if (j >= cmax)
4554 break;
4555 if (!atom.match(matcher, i, seq))
4556 break;
4557 int k = matcher.last - i;
4558 if (k <= 0) {
4559 if (capture) {
4560 groups[groupIndex] = i;
4561 groups[groupIndex+1] = i + k;
4562 }
4563 i = i + k;
4564 break;
4565 }
4566 for (;;) {
4567 if (capture) {
4568 groups[groupIndex] = i;
4569 groups[groupIndex+1] = i + k;
4570 }
4571 i = i + k;
4572 if (++j >= cmax)
4573 break;
4574 if (!atom.match(matcher, i, seq))
4575 break;
4576 if (i + k != matcher.last) {
4577 if (match0(matcher, i, j, seq))
4578 return true;
4579 break;
4580 }
4581 }
4582 while (j > min) {
4583 if (next.match(matcher, i, seq)) {
4584 if (capture) {
4585 groups[groupIndex+1] = i;
4586 groups[groupIndex] = i - k;
4587 }
4588 return true;
4589 }
4590 // backing off
4591 i = i - k;
4592 if (capture) {
4593 groups[groupIndex+1] = i;
4594 groups[groupIndex] = i - k;
4595 }
4596 j--;
4597
4598 }
4599 break;
4600 }
4601 if (capture) {
4602 groups[groupIndex] = save0;
4603 groups[groupIndex+1] = save1;
4604 }
4605 return next.match(matcher, i, seq);
4606 }
4607 // Reluctant matching
4608 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4609 for (;;) {
4610 if (next.match(matcher, i, seq))
4611 return true;
4612 if (j >= cmax)
4613 return false;
4614 if (!atom.match(matcher, i, seq))
4615 return false;
4616 if (i == matcher.last)
4617 return false;
4618 if (capture) {
4619 matcher.groups[groupIndex] = i;
4620 matcher.groups[groupIndex+1] = matcher.last;
4621 }
4622 i = matcher.last;
4623 j++;
4624 }
4625 }
4626 // Possessive matching
4627 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4628 for (; j < cmax; j++) {
4629 if (!atom.match(matcher, i, seq)) {
4630 break;
4631 }
4632 if (capture) {
4633 matcher.groups[groupIndex] = i;
4634 matcher.groups[groupIndex+1] = matcher.last;
4635 }
4636 if (i == matcher.last) {
4637 break;
4638 }
4639 i = matcher.last;
4640 }
4641 return next.match(matcher, i, seq);
4642 }
4643 boolean study(TreeInfo info) {
4644 // Save original info
4645 int minL = info.minLength;
4646 int maxL = info.maxLength;
4647 boolean maxV = info.maxValid;
4648 boolean detm = info.deterministic;
4649 info.reset();
4650
4651 atom.study(info);
4652
4653 int temp = info.minLength * cmin + minL;
4654 if (temp < minL) {
4655 temp = 0xFFFFFFF; // Arbitrary large number
4656 }
4657 info.minLength = temp;
4658
4659 if (maxV & info.maxValid) {
4660 temp = info.maxLength * cmax + maxL;
4661 info.maxLength = temp;
4662 if (temp < maxL) {
4663 info.maxValid = false;
4664 }
4665 } else {
4666 info.maxValid = false;
4667 }
4668
4669 if (info.deterministic && cmin == cmax) {
4670 info.deterministic = detm;
4671 } else {
4672 info.deterministic = false;
4673 }
4674 return next.study(info);
4675 }
4676 }
4677
4678 /**
4679 * A Guard node at the end of each atom node in a Branch. It
4680 * serves the purpose of chaining the "match" operation to
4681 * "next" but not the "study", so we can collect the TreeInfo
4682 * of each atom node without including the TreeInfo of the
4683 * "next".
4684 */
4685 static final class BranchConn extends Node {
4686 BranchConn() {};
4687 boolean match(Matcher matcher, int i, CharSequence seq) {
4688 return next.match(matcher, i, seq);
4689 }
4690 boolean study(TreeInfo info) {
4691 return info.deterministic;
4692 }
4693 }
4694
4695 /**
4696 * Handles the branching of alternations. Note this is also used for
4697 * the ? quantifier to branch between the case where it matches once
4698 * and where it does not occur.
4699 */
4700 static final class Branch extends Node {
4701 Node[] atoms = new Node[2];
4702 int size = 2;
4703 Node conn;
4704 Branch(Node first, Node second, Node branchConn) {
4705 conn = branchConn;
4706 atoms[0] = first;
4707 atoms[1] = second;
4708 }
4709
4710 void add(Node node) {
4711 if (size >= atoms.length) {
4712 Node[] tmp = new Node[atoms.length*2];
4713 System.arraycopy(atoms, 0, tmp, 0, atoms.length);
4714 atoms = tmp;
4715 }
4716 atoms[size++] = node;
4717 }
4718
4719 boolean match(Matcher matcher, int i, CharSequence seq) {
4720 for (int n = 0; n < size; n++) {
4721 if (atoms[n] == null) {
4722 if (conn.next.match(matcher, i, seq))
4723 return true;
4724 } else if (atoms[n].match(matcher, i, seq)) {
4725 return true;
4726 }
4727 }
4728 return false;
4729 }
4730
4731 boolean study(TreeInfo info) {
4732 int minL = info.minLength;
4733 int maxL = info.maxLength;
4734 boolean maxV = info.maxValid;
4735
4736 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num
4737 int maxL2 = -1;
4738 for (int n = 0; n < size; n++) {
4739 info.reset();
4740 if (atoms[n] != null)
4741 atoms[n].study(info);
4742 minL2 = Math.min(minL2, info.minLength);
4743 maxL2 = Math.max(maxL2, info.maxLength);
4744 maxV = (maxV & info.maxValid);
4745 }
4746
4747 minL += minL2;
4748 maxL += maxL2;
4749
4750 info.reset();
4751 conn.next.study(info);
4752
4753 info.minLength += minL;
4754 info.maxLength += maxL;
4755 info.maxValid &= maxV;
4756 info.deterministic = false;
4757 return false;
4758 }
4759 }
4760
4761 /**
4762 * The GroupHead saves the location where the group begins in the locals
4763 * and restores them when the match is done.
4764 *
4765 * The matchRef is used when a reference to this group is accessed later
4766 * in the expression. The locals will have a negative value in them to
4767 * indicate that we do not want to unset the group if the reference
4768 * doesn't match.
4769 */
4770 static final class GroupHead extends Node {
4771 int localIndex;
4772 GroupTail tail; // for debug/print only, match does not need to know
4773 GroupHead(int localCount) {
4774 localIndex = localCount;
4775 }
4776 boolean match(Matcher matcher, int i, CharSequence seq) {
4777 int save = matcher.locals[localIndex];
4778 matcher.locals[localIndex] = i;
4779 boolean ret = next.match(matcher, i, seq);
4780 matcher.locals[localIndex] = save;
4781 return ret;
4782 }
4783 boolean matchRef(Matcher matcher, int i, CharSequence seq) {
4784 int save = matcher.locals[localIndex];
4785 matcher.locals[localIndex] = ~i; // HACK
4786 boolean ret = next.match(matcher, i, seq);
4787 matcher.locals[localIndex] = save;
4788 return ret;
4789 }
4790 }
4791
4792 /**
4793 * Recursive reference to a group in the regular expression. It calls
4794 * matchRef because if the reference fails to match we would not unset
4795 * the group.
4796 */
4797 static final class GroupRef extends Node {
4798 GroupHead head;
4799 GroupRef(GroupHead head) {
4800 this.head = head;
4801 }
4802 boolean match(Matcher matcher, int i, CharSequence seq) {
4803 return head.matchRef(matcher, i, seq)
4804 && next.match(matcher, matcher.last, seq);
4805 }
4806 boolean study(TreeInfo info) {
4807 info.maxValid = false;
4808 info.deterministic = false;
4809 return next.study(info);
4810 }
4811 }
4812
4813 /**
4814 * The GroupTail handles the setting of group beginning and ending
4815 * locations when groups are successfully matched. It must also be able to
4816 * unset groups that have to be backed off of.
4817 *
4818 * The GroupTail node is also used when a previous group is referenced,
4819 * and in that case no group information needs to be set.
4820 */
4821 static final class GroupTail extends Node {
4822 int localIndex;
4823 int groupIndex;
4824 GroupTail(int localCount, int groupCount) {
4825 localIndex = localCount;
4826 groupIndex = groupCount + groupCount;
4827 }
4828 boolean match(Matcher matcher, int i, CharSequence seq) {
4829 int tmp = matcher.locals[localIndex];
4830 if (tmp >= 0) { // This is the normal group case.
4831 // Save the group so we can unset it if it
4832 // backs off of a match.
4833 int groupStart = matcher.groups[groupIndex];
4834 int groupEnd = matcher.groups[groupIndex+1];
4835
4836 matcher.groups[groupIndex] = tmp;
4837 matcher.groups[groupIndex+1] = i;
4838 if (next.match(matcher, i, seq)) {
4839 return true;
4840 }
4841 matcher.groups[groupIndex] = groupStart;
4842 matcher.groups[groupIndex+1] = groupEnd;
4843 return false;
4844 } else {
4845 // This is a group reference case. We don't need to save any
4846 // group info because it isn't really a group.
4847 matcher.last = i;
4848 return true;
4849 }
4850 }
4851 }
4852
4853 /**
4854 * This sets up a loop to handle a recursive quantifier structure.
4855 */
4856 static final class Prolog extends Node {
4857 Loop loop;
4858 Prolog(Loop loop) {
4859 this.loop = loop;
4860 }
4861 boolean match(Matcher matcher, int i, CharSequence seq) {
4862 return loop.matchInit(matcher, i, seq);
4863 }
4864 boolean study(TreeInfo info) {
4865 return loop.study(info);
4866 }
4867 }
4868
4869 /**
4870 * Handles the repetition count for a greedy Curly. The matchInit
4871 * is called from the Prolog to save the index of where the group
4872 * beginning is stored. A zero length group check occurs in the
4873 * normal match but is skipped in the matchInit.
4874 */
4875 static class Loop extends Node {
4876 Node body;
4877 int countIndex; // local count index in matcher locals
4878 int beginIndex; // group beginning index
4879 int cmin, cmax;
4880 int posIndex;
4881 Loop(int countIndex, int beginIndex) {
4882 this.countIndex = countIndex;
4883 this.beginIndex = beginIndex;
4884 this.posIndex = -1;
4885 }
4886 boolean match(Matcher matcher, int i, CharSequence seq) {
4887 // Avoid infinite loop in zero-length case.
4888 if (i > matcher.locals[beginIndex]) {
4889 int count = matcher.locals[countIndex];
4890
4891 // This block is for before we reach the minimum
4892 // iterations required for the loop to match
4893 if (count < cmin) {
4894 matcher.locals[countIndex] = count + 1;
4895 boolean b = body.match(matcher, i, seq);
4896 // If match failed we must backtrack, so
4897 // the loop count should NOT be incremented
4898 if (!b)
4899 matcher.locals[countIndex] = count;
4900 // Return success or failure since we are under
4901 // minimum
4902 return b;
4903 }
4904 // This block is for after we have the minimum
4905 // iterations required for the loop to match
4906 if (count < cmax) {
4907 // Let's check if we have already tried and failed
4908 // at this starting position "i" in the past.
4909 // If yes, then just return false wihtout trying
4910 // again, to stop the exponential backtracking.
4911 if (posIndex != -1 &&
4912 matcher.localsPos[posIndex].contains(i)) {
4913 return next.match(matcher, i, seq);
4914 }
4915 matcher.locals[countIndex] = count + 1;
4916 boolean b = body.match(matcher, i, seq);
4917 // If match failed we must backtrack, so
4918 // the loop count should NOT be incremented
4919 if (b)
4920 return true;
4921 matcher.locals[countIndex] = count;
4922 // save the failed position
4923 if (posIndex != -1) {
4924 matcher.localsPos[posIndex].add(i);
4925 }
4926 }
4927 }
4928 return next.match(matcher, i, seq);
4929 }
4930 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4931 int save = matcher.locals[countIndex];
4932 boolean ret = false;
4933 if (posIndex != -1 && matcher.localsPos[posIndex] == null) {
4934 matcher.localsPos[posIndex] = new IntHashSet();
4935 }
4936 if (0 < cmin) {
4937 matcher.locals[countIndex] = 1;
4938 ret = body.match(matcher, i, seq);
4939 } else if (0 < cmax) {
4940 matcher.locals[countIndex] = 1;
4941 ret = body.match(matcher, i, seq);
4942 if (ret == false)
4943 ret = next.match(matcher, i, seq);
4944 } else {
4945 ret = next.match(matcher, i, seq);
4946 }
4947 matcher.locals[countIndex] = save;
4948 return ret;
4949 }
4950 boolean study(TreeInfo info) {
4951 info.maxValid = false;
4952 info.deterministic = false;
4953 return false;
4954 }
4955 }
4956
4957 /**
4958 * Handles the repetition count for a reluctant Curly. The matchInit
4959 * is called from the Prolog to save the index of where the group
4960 * beginning is stored. A zero length group check occurs in the
4961 * normal match but is skipped in the matchInit.
4962 */
4963 static final class LazyLoop extends Loop {
4964 LazyLoop(int countIndex, int beginIndex) {
4965 super(countIndex, beginIndex);
4966 }
4967 boolean match(Matcher matcher, int i, CharSequence seq) {
4968 // Check for zero length group
4969 if (i > matcher.locals[beginIndex]) {
4970 int count = matcher.locals[countIndex];
4971 if (count < cmin) {
4972 matcher.locals[countIndex] = count + 1;
4973 boolean result = body.match(matcher, i, seq);
4974 // If match failed we must backtrack, so
4975 // the loop count should NOT be incremented
4976 if (!result)
4977 matcher.locals[countIndex] = count;
4978 return result;
4979 }
4980 if (next.match(matcher, i, seq))
4981 return true;
4982 if (count < cmax) {
4983 matcher.locals[countIndex] = count + 1;
4984 boolean result = body.match(matcher, i, seq);
4985 // If match failed we must backtrack, so
4986 // the loop count should NOT be incremented
4987 if (!result)
4988 matcher.locals[countIndex] = count;
4989 return result;
4990 }
4991 return false;
4992 }
4993 return next.match(matcher, i, seq);
4994 }
4995 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4996 int save = matcher.locals[countIndex];
4997 boolean ret = false;
4998 if (0 < cmin) {
4999 matcher.locals[countIndex] = 1;
5000 ret = body.match(matcher, i, seq);
5001 } else if (next.match(matcher, i, seq)) {
5002 ret = true;
5003 } else if (0 < cmax) {
5004 matcher.locals[countIndex] = 1;
5005 ret = body.match(matcher, i, seq);
5006 }
5007 matcher.locals[countIndex] = save;
5008 return ret;
5009 }
5010 boolean study(TreeInfo info) {
5011 info.maxValid = false;
5012 info.deterministic = false;
5013 return false;
5014 }
5015 }
5016
5017 /**
5018 * Refers to a group in the regular expression. Attempts to match
5019 * whatever the group referred to last matched.
5020 */
5021 static class BackRef extends Node {
5022 int groupIndex;
5023 BackRef(int groupCount) {
5024 super();
5025 groupIndex = groupCount + groupCount;
5026 }
5027 boolean match(Matcher matcher, int i, CharSequence seq) {
5028 int j = matcher.groups[groupIndex];
5029 int k = matcher.groups[groupIndex+1];
5030
5031 int groupSize = k - j;
5032 // If the referenced group didn't match, neither can this
5033 if (j < 0)
5034 return false;
5035
5036 // If there isn't enough input left no match
5037 if (i + groupSize > matcher.to) {
5038 matcher.hitEnd = true;
5039 return false;
5040 }
5041 // Check each new char to make sure it matches what the group
5042 // referenced matched last time around
5043 for (int index=0; index<groupSize; index++)
5044 if (seq.charAt(i+index) != seq.charAt(j+index))
5045 return false;
5046
5047 return next.match(matcher, i+groupSize, seq);
5048 }
5049 boolean study(TreeInfo info) {
5050 info.maxValid = false;
5051 return next.study(info);
5052 }
5053 }
5054
5055 static class CIBackRef extends Node {
5056 int groupIndex;
5057 boolean doUnicodeCase;
5058 CIBackRef(int groupCount, boolean doUnicodeCase) {
5059 super();
5060 groupIndex = groupCount + groupCount;
5061 this.doUnicodeCase = doUnicodeCase;
5062 }
5063 boolean match(Matcher matcher, int i, CharSequence seq) {
5064 int j = matcher.groups[groupIndex];
5065 int k = matcher.groups[groupIndex+1];
5066
5067 int groupSize = k - j;
5068
5069 // If the referenced group didn't match, neither can this
5070 if (j < 0)
5071 return false;
5072
5073 // If there isn't enough input left no match
5074 if (i + groupSize > matcher.to) {
5075 matcher.hitEnd = true;
5076 return false;
5077 }
5078
5079 // Check each new char to make sure it matches what the group
5080 // referenced matched last time around
5081 int x = i;
5082 for (int index=0; index<groupSize; index++) {
5083 int c1 = Character.codePointAt(seq, x);
5084 int c2 = Character.codePointAt(seq, j);
5085 if (c1 != c2) {
5086 if (doUnicodeCase) {
5087 int cc1 = Character.toUpperCase(c1);
5088 int cc2 = Character.toUpperCase(c2);
5089 if (cc1 != cc2 &&
5090 Character.toLowerCase(cc1) !=
5091 Character.toLowerCase(cc2))
5092 return false;
5093 } else {
5094 if (ASCII.toLower(c1) != ASCII.toLower(c2))
5095 return false;
5096 }
5097 }
5098 x += Character.charCount(c1);
5099 j += Character.charCount(c2);
5100 }
5101
5102 return next.match(matcher, i+groupSize, seq);
5103 }
5104 boolean study(TreeInfo info) {
5105 info.maxValid = false;
5106 return next.study(info);
5107 }
5108 }
5109
5110 /**
5111 * Searches until the next instance of its atom. This is useful for
5112 * finding the atom efficiently without passing an instance of it
5113 * (greedy problem) and without a lot of wasted search time (reluctant
5114 * problem).
5115 */
5116 static final class First extends Node {
5117 Node atom;
5118 First(Node node) {
5119 this.atom = BnM.optimize(node);
5120 }
5121 boolean match(Matcher matcher, int i, CharSequence seq) {
5122 if (atom instanceof BnM) {
5123 return atom.match(matcher, i, seq)
5124 && next.match(matcher, matcher.last, seq);
5125 }
5126 for (;;) {
5127 if (i > matcher.to) {
5128 matcher.hitEnd = true;
5129 return false;
5130 }
5131 if (atom.match(matcher, i, seq)) {
5132 return next.match(matcher, matcher.last, seq);
5133 }
5134 i += countChars(seq, i, 1);
5135 matcher.first++;
5136 }
5137 }
5138 boolean study(TreeInfo info) {
5139 atom.study(info);
5140 info.maxValid = false;
5141 info.deterministic = false;
5142 return next.study(info);
5143 }
5144 }
5145
5146 static final class Conditional extends Node {
5147 Node cond, yes, not;
5148 Conditional(Node cond, Node yes, Node not) {
5149 this.cond = cond;
5150 this.yes = yes;
5151 this.not = not;
5152 }
5153 boolean match(Matcher matcher, int i, CharSequence seq) {
5154 if (cond.match(matcher, i, seq)) {
5155 return yes.match(matcher, i, seq);
5156 } else {
5157 return not.match(matcher, i, seq);
5158 }
5159 }
5160 boolean study(TreeInfo info) {
5161 int minL = info.minLength;
5162 int maxL = info.maxLength;
5163 boolean maxV = info.maxValid;
5164 info.reset();
5165 yes.study(info);
5166
5167 int minL2 = info.minLength;
5168 int maxL2 = info.maxLength;
5169 boolean maxV2 = info.maxValid;
5170 info.reset();
5171 not.study(info);
5172
5173 info.minLength = minL + Math.min(minL2, info.minLength);
5174 info.maxLength = maxL + Math.max(maxL2, info.maxLength);
5175 info.maxValid = (maxV & maxV2 & info.maxValid);
5176 info.deterministic = false;
5177 return next.study(info);
5178 }
5179 }
5180
5181 /**
5182 * Zero width positive lookahead.
5183 */
5184 static final class Pos extends Node {
5185 Node cond;
5186 Pos(Node cond) {
5187 this.cond = cond;
5188 }
5189 boolean match(Matcher matcher, int i, CharSequence seq) {
5190 int savedTo = matcher.to;
5191 boolean conditionMatched = false;
5192
5193 // Relax transparent region boundaries for lookahead
5194 if (matcher.transparentBounds)
5195 matcher.to = matcher.getTextLength();
5196 try {
5197 conditionMatched = cond.match(matcher, i, seq);
5198 } finally {
5199 // Reinstate region boundaries
5200 matcher.to = savedTo;
5201 }
5202 return conditionMatched && next.match(matcher, i, seq);
5203 }
5204 }
5205
5206 /**
5207 * Zero width negative lookahead.
5208 */
5209 static final class Neg extends Node {
5210 Node cond;
5211 Neg(Node cond) {
5212 this.cond = cond;
5213 }
5214 boolean match(Matcher matcher, int i, CharSequence seq) {
5215 int savedTo = matcher.to;
5216 boolean conditionMatched = false;
5217
5218 // Relax transparent region boundaries for lookahead
5219 if (matcher.transparentBounds)
5220 matcher.to = matcher.getTextLength();
5221 try {
5222 if (i < matcher.to) {
5223 conditionMatched = !cond.match(matcher, i, seq);
5224 } else {
5225 // If a negative lookahead succeeds then more input
5226 // could cause it to fail!
5227 matcher.requireEnd = true;
5228 conditionMatched = !cond.match(matcher, i, seq);
5229 }
5230 } finally {
5231 // Reinstate region boundaries
5232 matcher.to = savedTo;
5233 }
5234 return conditionMatched && next.match(matcher, i, seq);
5235 }
5236 }
5237
5238 /**
5239 * For use with lookbehinds; matches the position where the lookbehind
5240 * was encountered.
5241 */
5242 static Node lookbehindEnd = new Node() {
5243 boolean match(Matcher matcher, int i, CharSequence seq) {
5244 return i == matcher.lookbehindTo;
5245 }
5246 };
5247
5248 /**
5249 * Zero width positive lookbehind.
5250 */
5251 static class Behind extends Node {
5252 Node cond;
5253 int rmax, rmin;
5254 Behind(Node cond, int rmax, int rmin) {
5255 this.cond = cond;
5256 this.rmax = rmax;
5257 this.rmin = rmin;
5258 }
5259
5260 boolean match(Matcher matcher, int i, CharSequence seq) {
5261 int savedFrom = matcher.from;
5262 boolean conditionMatched = false;
5263 int startIndex = (!matcher.transparentBounds) ?
5264 matcher.from : 0;
5265 int from = Math.max(i - rmax, startIndex);
5266 // Set end boundary
5267 int savedLBT = matcher.lookbehindTo;
5268 matcher.lookbehindTo = i;
5269 // Relax transparent region boundaries for lookbehind
5270 if (matcher.transparentBounds)
5271 matcher.from = 0;
5272 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
5273 conditionMatched = cond.match(matcher, j, seq);
5274 }
5275 matcher.from = savedFrom;
5276 matcher.lookbehindTo = savedLBT;
5277 return conditionMatched && next.match(matcher, i, seq);
5278 }
5279 }
5280
5281 /**
5282 * Zero width positive lookbehind, including supplementary
5283 * characters or unpaired surrogates.
5284 */
5285 static final class BehindS extends Behind {
5286 BehindS(Node cond, int rmax, int rmin) {
5287 super(cond, rmax, rmin);
5288 }
5289 boolean match(Matcher matcher, int i, CharSequence seq) {
5290 int rmaxChars = countChars(seq, i, -rmax);
5291 int rminChars = countChars(seq, i, -rmin);
5292 int savedFrom = matcher.from;
5293 int startIndex = (!matcher.transparentBounds) ?
5294 matcher.from : 0;
5295 boolean conditionMatched = false;
5296 int from = Math.max(i - rmaxChars, startIndex);
5297 // Set end boundary
5298 int savedLBT = matcher.lookbehindTo;
5299 matcher.lookbehindTo = i;
5300 // Relax transparent region boundaries for lookbehind
5301 if (matcher.transparentBounds)
5302 matcher.from = 0;
5303
5304 for (int j = i - rminChars;
5305 !conditionMatched && j >= from;
5306 j -= j>from ? countChars(seq, j, -1) : 1) {
5307 conditionMatched = cond.match(matcher, j, seq);
5308 }
5309 matcher.from = savedFrom;
5310 matcher.lookbehindTo = savedLBT;
5311 return conditionMatched && next.match(matcher, i, seq);
5312 }
5313 }
5314
5315 /**
5316 * Zero width negative lookbehind.
5317 */
5318 static class NotBehind extends Node {
5319 Node cond;
5320 int rmax, rmin;
5321 NotBehind(Node cond, int rmax, int rmin) {
5322 this.cond = cond;
5323 this.rmax = rmax;
5324 this.rmin = rmin;
5325 }
5326
5327 boolean match(Matcher matcher, int i, CharSequence seq) {
5328 int savedLBT = matcher.lookbehindTo;
5329 int savedFrom = matcher.from;
5330 boolean conditionMatched = false;
5331 int startIndex = (!matcher.transparentBounds) ?
5332 matcher.from : 0;
5333 int from = Math.max(i - rmax, startIndex);
5334 matcher.lookbehindTo = i;
5335 // Relax transparent region boundaries for lookbehind
5336 if (matcher.transparentBounds)
5337 matcher.from = 0;
5338 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
5339 conditionMatched = cond.match(matcher, j, seq);
5340 }
5341 // Reinstate region boundaries
5342 matcher.from = savedFrom;
5343 matcher.lookbehindTo = savedLBT;
5344 return !conditionMatched && next.match(matcher, i, seq);
5345 }
5346 }
5347
5348 /**
5349 * Zero width negative lookbehind, including supplementary
5350 * characters or unpaired surrogates.
5351 */
5352 static final class NotBehindS extends NotBehind {
5353 NotBehindS(Node cond, int rmax, int rmin) {
5354 super(cond, rmax, rmin);
5355 }
5356 boolean match(Matcher matcher, int i, CharSequence seq) {
5357 int rmaxChars = countChars(seq, i, -rmax);
5358 int rminChars = countChars(seq, i, -rmin);
5359 int savedFrom = matcher.from;
5360 int savedLBT = matcher.lookbehindTo;
5361 boolean conditionMatched = false;
5362 int startIndex = (!matcher.transparentBounds) ?
5363 matcher.from : 0;
5364 int from = Math.max(i - rmaxChars, startIndex);
5365 matcher.lookbehindTo = i;
5366 // Relax transparent region boundaries for lookbehind
5367 if (matcher.transparentBounds)
5368 matcher.from = 0;
5369 for (int j = i - rminChars;
5370 !conditionMatched && j >= from;
5371 j -= j>from ? countChars(seq, j, -1) : 1) {
5372 conditionMatched = cond.match(matcher, j, seq);
5373 }
5374 //Reinstate region boundaries
5375 matcher.from = savedFrom;
5376 matcher.lookbehindTo = savedLBT;
5377 return !conditionMatched && next.match(matcher, i, seq);
5378 }
5379 }
5380
5381 /**
5382 * Handles word boundaries. Includes a field to allow this one class to
5383 * deal with the different types of word boundaries we can match. The word
5384 * characters include underscores, letters, and digits. Non spacing marks
5385 * can are also part of a word if they have a base character, otherwise
5386 * they are ignored for purposes of finding word boundaries.
5387 */
5388 static final class Bound extends Node {
5389 static int LEFT = 0x1;
5390 static int RIGHT= 0x2;
5391 static int BOTH = 0x3;
5392 static int NONE = 0x4;
5393 int type;
5394 boolean useUWORD;
5395 Bound(int n, boolean useUWORD) {
5396 type = n;
5397 this.useUWORD = useUWORD;
5398 }
5399
5400 boolean isWord(int ch) {
5401 return useUWORD ? CharPredicates.WORD().is(ch)
5402 : (ch == '_' || Character.isLetterOrDigit(ch));
5403 }
5404
5405 int check(Matcher matcher, int i, CharSequence seq) {
5406 int ch;
5407 boolean left = false;
5408 int startIndex = matcher.from;
5409 int endIndex = matcher.to;
5410 if (matcher.transparentBounds) {
5411 startIndex = 0;
5412 endIndex = matcher.getTextLength();
5413 }
5414 if (i > startIndex) {
5415 ch = Character.codePointBefore(seq, i);
5416 left = (isWord(ch) ||
5417 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5418 && hasBaseCharacter(matcher, i-1, seq)));
5419 }
5420 boolean right = false;
5421 if (i < endIndex) {
5422 ch = Character.codePointAt(seq, i);
5423 right = (isWord(ch) ||
5424 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5425 && hasBaseCharacter(matcher, i, seq)));
5426 } else {
5427 // Tried to access char past the end
5428 matcher.hitEnd = true;
5429 // The addition of another char could wreck a boundary
5430 matcher.requireEnd = true;
5431 }
5432 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE);
5433 }
5434 boolean match(Matcher matcher, int i, CharSequence seq) {
5435 return (check(matcher, i, seq) & type) > 0
5436 && next.match(matcher, i, seq);
5437 }
5438 }
5439
5440 /**
5441 * Non spacing marks only count as word characters in bounds calculations
5442 * if they have a base character.
5443 */
5444 private static boolean hasBaseCharacter(Matcher matcher, int i,
5445 CharSequence seq)
5446 {
5447 int start = (!matcher.transparentBounds) ?
5448 matcher.from : 0;
5449 for (int x=i; x >= start; x--) {
5450 int ch = Character.codePointAt(seq, x);
5451 if (Character.isLetterOrDigit(ch))
5452 return true;
5453 if (Character.getType(ch) == Character.NON_SPACING_MARK)
5454 continue;
5455 return false;
5456 }
5457 return false;
5458 }
5459
5460 /**
5461 * Attempts to match a slice in the input using the Boyer-Moore string
5462 * matching algorithm. The algorithm is based on the idea that the
5463 * pattern can be shifted farther ahead in the search text if it is
5464 * matched right to left.
5465 * <p>
5466 * The pattern is compared to the input one character at a time, from
5467 * the rightmost character in the pattern to the left. If the characters
5468 * all match the pattern has been found. If a character does not match,
5469 * the pattern is shifted right a distance that is the maximum of two
5470 * functions, the bad character shift and the good suffix shift. This
5471 * shift moves the attempted match position through the input more
5472 * quickly than a naive one position at a time check.
5473 * <p>
5474 * The bad character shift is based on the character from the text that
5475 * did not match. If the character does not appear in the pattern, the
5476 * pattern can be shifted completely beyond the bad character. If the
5477 * character does occur in the pattern, the pattern can be shifted to
5478 * line the pattern up with the next occurrence of that character.
5479 * <p>
5480 * The good suffix shift is based on the idea that some subset on the right
5481 * side of the pattern has matched. When a bad character is found, the
5482 * pattern can be shifted right by the pattern length if the subset does
5483 * not occur again in pattern, or by the amount of distance to the
5484 * next occurrence of the subset in the pattern.
5485 *
5486 * Boyer-Moore search methods adapted from code by Amy Yu.
5487 */
5488 static class BnM extends Node {
5489 int[] buffer;
5490 int[] lastOcc;
5491 int[] optoSft;
5492
5493 /**
5494 * Pre calculates arrays needed to generate the bad character
5495 * shift and the good suffix shift. Only the last seven bits
5496 * are used to see if chars match; This keeps the tables small
5497 * and covers the heavily used ASCII range, but occasionally
5498 * results in an aliased match for the bad character shift.
5499 */
5500 static Node optimize(Node node) {
5501 if (!(node instanceof Slice)) {
5502 return node;
5503 }
5504
5505 int[] src = ((Slice) node).buffer;
5506 int patternLength = src.length;
5507 // The BM algorithm requires a bit of overhead;
5508 // If the pattern is short don't use it, since
5509 // a shift larger than the pattern length cannot
5510 // be used anyway.
5511 if (patternLength < 4) {
5512 return node;
5513 }
5514 int i, j, k;
5515 int[] lastOcc = new int[128];
5516 int[] optoSft = new int[patternLength];
5517 // Precalculate part of the bad character shift
5518 // It is a table for where in the pattern each
5519 // lower 7-bit value occurs
5520 for (i = 0; i < patternLength; i++) {
5521 lastOcc[src[i]&0x7F] = i + 1;
5522 }
5523 // Precalculate the good suffix shift
5524 // i is the shift amount being considered
5525 NEXT: for (i = patternLength; i > 0; i--) {
5526 // j is the beginning index of suffix being considered
5527 for (j = patternLength - 1; j >= i; j--) {
5528 // Testing for good suffix
5529 if (src[j] == src[j-i]) {
5530 // src[j..len] is a good suffix
5531 optoSft[j-1] = i;
5532 } else {
5533 // No match. The array has already been
5534 // filled up with correct values before.
5535 continue NEXT;
5536 }
5537 }
5538 // This fills up the remaining of optoSft
5539 // any suffix can not have larger shift amount
5540 // then its sub-suffix. Why???
5541 while (j > 0) {
5542 optoSft[--j] = i;
5543 }
5544 }
5545 // Set the guard value because of unicode compression
5546 optoSft[patternLength-1] = 1;
5547 if (node instanceof SliceS)
5548 return new BnMS(src, lastOcc, optoSft, node.next);
5549 return new BnM(src, lastOcc, optoSft, node.next);
5550 }
5551 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5552 this.buffer = src;
5553 this.lastOcc = lastOcc;
5554 this.optoSft = optoSft;
5555 this.next = next;
5556 }
5557 boolean match(Matcher matcher, int i, CharSequence seq) {
5558 int[] src = buffer;
5559 int patternLength = src.length;
5560 int last = matcher.to - patternLength;
5561
5562 // Loop over all possible match positions in text
5563 NEXT: while (i <= last) {
5564 // Loop over pattern from right to left
5565 for (int j = patternLength - 1; j >= 0; j--) {
5566 int ch = seq.charAt(i+j);
5567 if (ch != src[j]) {
5568 // Shift search to the right by the maximum of the
5569 // bad character shift and the good suffix shift
5570 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]);
5571 continue NEXT;
5572 }
5573 }
5574 // Entire pattern matched starting at i
5575 matcher.first = i;
5576 boolean ret = next.match(matcher, i + patternLength, seq);
5577 if (ret) {
5578 matcher.first = i;
5579 matcher.groups[0] = matcher.first;
5580 matcher.groups[1] = matcher.last;
5581 return true;
5582 }
5583 i++;
5584 }
5585 // BnM is only used as the leading node in the unanchored case,
5586 // and it replaced its Start() which always searches to the end
5587 // if it doesn't find what it's looking for, so hitEnd is true.
5588 matcher.hitEnd = true;
5589 return false;
5590 }
5591 boolean study(TreeInfo info) {
5592 info.minLength += buffer.length;
5593 info.maxValid = false;
5594 return next.study(info);
5595 }
5596 }
5597
5598 /**
5599 * Supplementary support version of BnM(). Unpaired surrogates are
5600 * also handled by this class.
5601 */
5602 static final class BnMS extends BnM {
5603 int lengthInChars;
5604
5605 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5606 super(src, lastOcc, optoSft, next);
5607 for (int cp : buffer) {
5608 lengthInChars += Character.charCount(cp);
5609 }
5610 }
5611 boolean match(Matcher matcher, int i, CharSequence seq) {
5612 int[] src = buffer;
5613 int patternLength = src.length;
5614 int last = matcher.to - lengthInChars;
5615
5616 // Loop over all possible match positions in text
5617 NEXT: while (i <= last) {
5618 // Loop over pattern from right to left
5619 int ch;
5620 for (int j = countChars(seq, i, patternLength), x = patternLength - 1;
5621 j > 0; j -= Character.charCount(ch), x--) {
5622 ch = Character.codePointBefore(seq, i+j);
5623 if (ch != src[x]) {
5624 // Shift search to the right by the maximum of the
5625 // bad character shift and the good suffix shift
5626 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]);
5627 i += countChars(seq, i, n);
5628 continue NEXT;
5629 }
5630 }
5631 // Entire pattern matched starting at i
5632 matcher.first = i;
5633 boolean ret = next.match(matcher, i + lengthInChars, seq);
5634 if (ret) {
5635 matcher.first = i;
5636 matcher.groups[0] = matcher.first;
5637 matcher.groups[1] = matcher.last;
5638 return true;
5639 }
5640 i += countChars(seq, i, 1);
5641 }
5642 matcher.hitEnd = true;
5643 return false;
5644 }
5645 }
5646
5647 @FunctionalInterface
5648 static interface CharPredicate {
5649 boolean is(int ch);
5650
5651 default CharPredicate and(CharPredicate p) {
5652 return ch -> is(ch) && p.is(ch);
5653 }
5654 default CharPredicate union(CharPredicate p) {
5655 return ch -> is(ch) || p.is(ch);
5656 }
5657 default CharPredicate union(CharPredicate p1,
5658 CharPredicate p2 ) {
5659 return ch -> is(ch) || p1.is(ch) || p2.is(ch);
5660 }
5661 default CharPredicate negate() {
5662 return ch -> !is(ch);
5663 }
5664 }
5665
5666 static interface BmpCharPredicate extends CharPredicate {
5667
5668 default CharPredicate and(CharPredicate p) {
5669 if(p instanceof BmpCharPredicate)
5670 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch));
5671 return ch -> is(ch) && p.is(ch);
5672 }
5673 default CharPredicate union(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 static CharPredicate union(CharPredicate... predicates) {
5679 CharPredicate cp = ch -> {
5680 for (CharPredicate p : predicates) {
5681 if (!p.is(ch))
5682 return false;
5683 }
5684 return true;
5685 };
5686 for (CharPredicate p : predicates) {
5687 if (! (p instanceof BmpCharPredicate))
5688 return cp;
5689 }
5690 return (BmpCharPredicate)cp;
5691 }
5692 }
5693
5694 /**
5695 * matches a Perl vertical whitespace
5696 */
5697 static BmpCharPredicate VertWS() {
5698 return cp -> (cp >= 0x0A && cp <= 0x0D) ||
5699 cp == 0x85 || cp == 0x2028 || cp == 0x2029;
5700 }
5701
5702 /**
5703 * matches a Perl horizontal whitespace
5704 */
5705 static BmpCharPredicate HorizWS() {
5706 return cp ->
5707 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 ||
5708 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f ||
5709 cp == 0x205f || cp == 0x3000;
5710 }
5711
5712 /**
5713 * for the Unicode category ALL and the dot metacharacter when
5714 * in dotall mode.
5715 */
5716 static CharPredicate ALL() {
5717 return ch -> true;
5718 }
5719
5720 /**
5721 * for the dot metacharacter when dotall is not enabled.
5722 */
5723 static CharPredicate DOT() {
5724 return ch ->
5725 (ch != '\n' && ch != '\r'
5726 && (ch|1) != '\u2029'
5727 && ch != '\u0085');
5728 }
5729
5730 /**
5731 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled.
5732 */
5733 static CharPredicate UNIXDOT() {
5734 return ch -> ch != '\n';
5735 }
5736
5737 /**
5738 * Indicate that matches a Supplementary Unicode character
5739 */
5740 static CharPredicate SingleS(int c) {
5741 return ch -> ch == c;
5742 }
5743
5744 /**
5745 * A bmp/optimized predicate of single
5746 */
5747 static BmpCharPredicate Single(int c) {
5748 return ch -> ch == c;
5749 }
5750
5751 /**
5752 * Case insensitive matches a given BMP character
5753 */
5754 static BmpCharPredicate SingleI(int lower, int upper) {
5755 return ch -> ch == lower || ch == upper;
5756 }
5757
5758 /**
5759 * Unicode case insensitive matches a given Unicode character
5760 */
5761 static CharPredicate SingleU(int lower) {
5762 return ch -> lower == ch ||
5763 lower == Character.toLowerCase(Character.toUpperCase(ch));
5764 }
5765
5766 private static boolean inRange(int lower, int ch, int upper) {
5767 return lower <= ch && ch <= upper;
5768 }
5769
5770 /**
5771 * Charactrs within a explicit value range
5772 */
5773 static CharPredicate Range(int lower, int upper) {
5774 if (upper < Character.MIN_HIGH_SURROGATE ||
5775 lower > Character.MAX_HIGH_SURROGATE &&
5776 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT)
5777 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper));
5778 return ch -> inRange(lower, ch, upper);
5779 }
5780
5781 /**
5782 * Charactrs within a explicit value range in a case insensitive manner.
5783 */
5784 static CharPredicate CIRange(int lower, int upper) {
5785 return ch -> inRange(lower, ch, upper) ||
5786 ASCII.isAscii(ch) &&
5787 (inRange(lower, ASCII.toUpper(ch), upper) ||
5788 inRange(lower, ASCII.toLower(ch), upper));
5789 }
5790
5791 static CharPredicate CIRangeU(int lower, int upper) {
5792 return ch -> {
5793 if (inRange(lower, ch, upper))
5794 return true;
5795 int up = Character.toUpperCase(ch);
5796 return inRange(lower, up, upper) ||
5797 inRange(lower, Character.toLowerCase(up), upper);
5798 };
5799 }
5800
5801 /**
5802 * This must be the very first initializer.
5803 */
5804 static final Node accept = new Node();
5805
5806 static final Node lastAccept = new LastNode();
5807
5808 /**
5809 * Creates a predicate that tests if this pattern is found in a given input
5810 * string.
5811 *
5812 * @apiNote
5813 * This method creates a predicate that behaves as if it creates a matcher
5814 * from the input sequence and then calls {@code find}, for example a
5815 * predicate of the form:
5816 * <pre>{@code
5817 * s -> matcher(s).find();
5818 * }</pre>
5819 *
5820 * @return The predicate which can be used for finding a match on a
5821 * subsequence of a string
5822 * @since 1.8
5823 * @see Matcher#find
5824 */
5825 public Predicate<String> asPredicate() {
5826 return s -> matcher(s).find();
5827 }
5828
5829 /**
5830 * Creates a predicate that tests if this pattern matches a given input string.
5831 *
5832 * @apiNote
5833 * This method creates a predicate that behaves as if it creates a matcher
5834 * from the input sequence and then calls {@code matches}, for example a
5835 * predicate of the form:
5836 * <pre>{@code
5837 * s -> matcher(s).matches();
5838 * }</pre>
5839 *
5840 * @return The predicate which can be used for matching an input string
5841 * against this pattern.
5842 * @since 11
5843 * @see Matcher#matches
5844 */
5845 public Predicate<String> asMatchPredicate() {
5846 return s -> matcher(s).matches();
5847 }
5848
5849 /**
5850 * Creates a stream from the given input sequence around matches of this
5851 * pattern.
5852 *
5853 * <p> The stream returned by this method contains each substring of the
5854 * input sequence that is terminated by another subsequence that matches
5855 * this pattern or is terminated by the end of the input sequence. The
5856 * substrings in the stream are in the order in which they occur in the
5857 * input. Trailing empty strings will be discarded and not encountered in
5858 * the stream.
5859 *
5860 * <p> If this pattern does not match any subsequence of the input then
5861 * the resulting stream has just one element, namely the input sequence in
5862 * string form.
5863 *
5864 * <p> When there is a positive-width match at the beginning of the input
5865 * sequence then an empty leading substring is included at the beginning
5866 * of the stream. A zero-width match at the beginning however never produces
5867 * such empty leading substring.
5868 *
5869 * <p> If the input sequence is mutable, it must remain constant during the
5870 * execution of the terminal stream operation. Otherwise, the result of the
5871 * terminal stream operation is undefined.
5872 *
5873 * @param input
5874 * The character sequence to be split
5875 *
5876 * @return The stream of strings computed by splitting the input
5877 * around matches of this pattern
5878 * @see #split(CharSequence)
5879 * @since 1.8
5880 */
5881 public Stream<String> splitAsStream(final CharSequence input) {
5882 class MatcherIterator implements Iterator<String> {
5883 private Matcher matcher;
5884 // The start position of the next sub-sequence of input
5885 // when current == input.length there are no more elements
5886 private int current;
5887 // null if the next element, if any, needs to obtained
5888 private String nextElement;
5889 // > 0 if there are N next empty elements
5890 private int emptyElementCount;
5891
5892 public String next() {
5893 if (!hasNext())
5894 throw new NoSuchElementException();
5895
5896 if (emptyElementCount == 0) {
5897 String n = nextElement;
5898 nextElement = null;
5899 return n;
5900 } else {
5901 emptyElementCount--;
5902 return "";
5903 }
5904 }
5905
5906 public boolean hasNext() {
5907 if (matcher == null) {
5908 matcher = matcher(input);
5909 // If the input is an empty string then the result can only be a
5910 // stream of the input. Induce that by setting the empty
5911 // element count to 1
5912 emptyElementCount = input.length() == 0 ? 1 : 0;
5913 }
5914 if (nextElement != null || emptyElementCount > 0)
5915 return true;
5916
5917 if (current == input.length())
5918 return false;
5919
5920 // Consume the next matching element
5921 // Count sequence of matching empty elements
5922 while (matcher.find()) {
5923 nextElement = input.subSequence(current, matcher.start()).toString();
5924 current = matcher.end();
5925 if (!nextElement.isEmpty()) {
5926 return true;
5927 } else if (current > 0) { // no empty leading substring for zero-width
5928 // match at the beginning of the input
5929 emptyElementCount++;
5930 }
5931 }
5932
5933 // Consume last matching element
5934 nextElement = input.subSequence(current, input.length()).toString();
5935 current = input.length();
5936 if (!nextElement.isEmpty()) {
5937 return true;
5938 } else {
5939 // Ignore a terminal sequence of matching empty elements
5940 emptyElementCount = 0;
5941 nextElement = null;
5942 return false;
5943 }
5944 }
5945 }
5946 return StreamSupport.stream(Spliterators.spliteratorUnknownSize(
5947 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false);
5948 }
5949 }