1 /*
2 * Copyright (c) 1999, 2013, 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.util.Locale;
30 import java.util.Iterator;
31 import java.util.Map;
32 import java.util.ArrayList;
33 import java.util.HashMap;
34 import java.util.Arrays;
35 import java.util.NoSuchElementException;
36 import java.util.Spliterator;
37 import java.util.Spliterators;
38 import java.util.function.Predicate;
39 import java.util.stream.Stream;
40 import java.util.stream.StreamSupport;
41
42
43 /**
44 * A compiled representation of a regular expression.
45 *
46 * <p> A regular expression, specified as a string, must first be compiled into
47 * an instance of this class. The resulting pattern can then be used to create
48 * a {@link Matcher} object that can match arbitrary {@linkplain
49 * java.lang.CharSequence character sequences} against the regular
50 * expression. All of the state involved in performing a match resides in the
51 * matcher, so many matchers can share the same pattern.
52 *
53 * <p> A typical invocation sequence is thus
54 *
55 * <blockquote><pre>
56 * Pattern p = Pattern.{@link #compile compile}("a*b");
57 * Matcher m = p.{@link #matcher matcher}("aaaaab");
58 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote>
59 *
60 * <p> A {@link #matches matches} method is defined by this class as a
61 * convenience for when a regular expression is used just once. This method
62 * compiles an expression and matches an input sequence against it in a single
63 * invocation. The statement
64 *
65 * <blockquote><pre>
66 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote>
67 *
68 * is equivalent to the three statements above, though for repeated matches it
69 * is less efficient since it does not allow the compiled pattern to be reused.
70 *
71 * <p> Instances of this class are immutable and are safe for use by multiple
72 * concurrent threads. Instances of the {@link Matcher} class are not safe for
73 * such use.
74 *
75 *
76 * <h3><a name="sum">Summary of regular-expression constructs</a></h3>
77 *
78 * <table border="0" cellpadding="1" cellspacing="0"
79 * summary="Regular expression constructs, and what they match">
80 *
81 * <tr align="left">
82 * <th align="left" id="construct">Construct</th>
83 * <th align="left" id="matches">Matches</th>
84 * </tr>
85 *
86 * <tr><th> </th></tr>
87 * <tr align="left"><th colspan="2" id="characters">Characters</th></tr>
88 *
89 * <tr><td valign="top" headers="construct characters"><i>x</i></td>
90 * <td headers="matches">The character <i>x</i></td></tr>
91 * <tr><td valign="top" headers="construct characters">{@code \\}</td>
92 * <td headers="matches">The backslash character</td></tr>
93 * <tr><td valign="top" headers="construct characters">{@code \0}<i>n</i></td>
94 * <td headers="matches">The character with octal value {@code 0}<i>n</i>
95 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr>
96 * <tr><td valign="top" headers="construct characters">{@code \0}<i>nn</i></td>
97 * <td headers="matches">The character with octal value {@code 0}<i>nn</i>
98 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr>
99 * <tr><td valign="top" headers="construct characters">{@code \0}<i>mnn</i></td>
100 * <td headers="matches">The character with octal value {@code 0}<i>mnn</i>
101 * (0 {@code <=} <i>m</i> {@code <=} 3,
102 * 0 {@code <=} <i>n</i> {@code <=} 7)</td></tr>
103 * <tr><td valign="top" headers="construct characters">{@code \x}<i>hh</i></td>
104 * <td headers="matches">The character with hexadecimal value {@code 0x}<i>hh</i></td></tr>
105 * <tr><td valign="top" headers="construct characters"><code>\u</code><i>hhhh</i></td>
106 * <td headers="matches">The character with hexadecimal value {@code 0x}<i>hhhh</i></td></tr>
107 * <tr><td valign="top" headers="construct characters"><code>\x</code><i>{h...h}</i></td>
108 * <td headers="matches">The character with hexadecimal value {@code 0x}<i>h...h</i>
109 * ({@link java.lang.Character#MIN_CODE_POINT Character.MIN_CODE_POINT}
110 * <= {@code 0x}<i>h...h</i> <=
111 * {@link java.lang.Character#MAX_CODE_POINT Character.MAX_CODE_POINT})</td></tr>
112 * <tr><td valign="top" headers="matches">{@code \t}</td>
113 * <td headers="matches">The tab character (<code>'\u0009'</code>)</td></tr>
114 * <tr><td valign="top" headers="construct characters">{@code \n}</td>
115 * <td headers="matches">The newline (line feed) character (<code>'\u000A'</code>)</td></tr>
116 * <tr><td valign="top" headers="construct characters">{@code \r}</td>
117 * <td headers="matches">The carriage-return character (<code>'\u000D'</code>)</td></tr>
118 * <tr><td valign="top" headers="construct characters">{@code \f}</td>
119 * <td headers="matches">The form-feed character (<code>'\u000C'</code>)</td></tr>
120 * <tr><td valign="top" headers="construct characters">{@code \a}</td>
121 * <td headers="matches">The alert (bell) character (<code>'\u0007'</code>)</td></tr>
122 * <tr><td valign="top" headers="construct characters">{@code \e}</td>
123 * <td headers="matches">The escape character (<code>'\u001B'</code>)</td></tr>
124 * <tr><td valign="top" headers="construct characters">{@code \c}<i>x</i></td>
125 * <td headers="matches">The control character corresponding to <i>x</i></td></tr>
126 *
127 * <tr><th> </th></tr>
128 * <tr align="left"><th colspan="2" id="classes">Character classes</th></tr>
129 *
130 * <tr><td valign="top" headers="construct classes">{@code [abc]}</td>
131 * <td headers="matches">{@code a}, {@code b}, or {@code c} (simple class)</td></tr>
132 * <tr><td valign="top" headers="construct classes">{@code [^abc]}</td>
133 * <td headers="matches">Any character except {@code a}, {@code b}, or {@code c} (negation)</td></tr>
134 * <tr><td valign="top" headers="construct classes">{@code [a-zA-Z]}</td>
135 * <td headers="matches">{@code a} through {@code z}
136 * or {@code A} through {@code Z}, inclusive (range)</td></tr>
137 * <tr><td valign="top" headers="construct classes">{@code [a-d[m-p]]}</td>
138 * <td headers="matches">{@code a} through {@code d},
139 * or {@code m} through {@code p}: {@code [a-dm-p]} (union)</td></tr>
140 * <tr><td valign="top" headers="construct classes">{@code [a-z&&[def]]}</td>
141 * <td headers="matches">{@code d}, {@code e}, or {@code f} (intersection)</tr>
142 * <tr><td valign="top" headers="construct classes">{@code [a-z&&[^bc]]}</td>
143 * <td headers="matches">{@code a} through {@code z},
144 * except for {@code b} and {@code c}: {@code [ad-z]} (subtraction)</td></tr>
145 * <tr><td valign="top" headers="construct classes">{@code [a-z&&[^m-p]]}</td>
146 * <td headers="matches">{@code a} through {@code z},
147 * and not {@code m} through {@code p}: {@code [a-lq-z]}(subtraction)</td></tr>
148 * <tr><th> </th></tr>
149 *
150 * <tr align="left"><th colspan="2" id="predef">Predefined character classes</th></tr>
151 *
152 * <tr><td valign="top" headers="construct predef">{@code .}</td>
153 * <td headers="matches">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr>
154 * <tr><td valign="top" headers="construct predef">{@code \d}</td>
155 * <td headers="matches">A digit: {@code [0-9]}</td></tr>
156 * <tr><td valign="top" headers="construct predef">{@code \D}</td>
157 * <td headers="matches">A non-digit: {@code [^0-9]}</td></tr>
158 * <tr><td valign="top" headers="construct predef">{@code \h}</td>
159 * <td headers="matches">A horizontal whitespace character:
160 * <code>[ \t\xA0\u1680\u180e\u2000-\u200a\u202f\u205f\u3000]</code></td></tr>
161 * <tr><td valign="top" headers="construct predef">{@code \H}</td>
162 * <td headers="matches">A non-horizontal whitespace character: {@code [^\h]}</td></tr>
163 * <tr><td valign="top" headers="construct predef">{@code \s}</td>
164 * <td headers="matches">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr>
165 * <tr><td valign="top" headers="construct predef">{@code \S}</td>
166 * <td headers="matches">A non-whitespace character: {@code [^\s]}</td></tr>
167 * <tr><td valign="top" headers="construct predef">{@code \v}</td>
168 * <td headers="matches">A vertical whitespace character: <code>[\n\x0B\f\r\x85\u2028\u2029]</code>
169 * </td></tr>
170 * <tr><td valign="top" headers="construct predef">{@code \V}</td>
171 * <td headers="matches">A non-vertical whitespace character: {@code [^\v]}</td></tr>
172 * <tr><td valign="top" headers="construct predef">{@code \w}</td>
173 * <td headers="matches">A word character: {@code [a-zA-Z_0-9]}</td></tr>
174 * <tr><td valign="top" headers="construct predef">{@code \W}</td>
175 * <td headers="matches">A non-word character: {@code [^\w]}</td></tr>
176 * <tr><th> </th></tr>
177 * <tr align="left"><th colspan="2" id="posix"><b>POSIX character classes (US-ASCII only)</b></th></tr>
178 *
179 * <tr><td valign="top" headers="construct posix">{@code \p{Lower}}</td>
180 * <td headers="matches">A lower-case alphabetic character: {@code [a-z]}</td></tr>
181 * <tr><td valign="top" headers="construct posix">{@code \p{Upper}}</td>
182 * <td headers="matches">An upper-case alphabetic character:{@code [A-Z]}</td></tr>
183 * <tr><td valign="top" headers="construct posix">{@code \p{ASCII}}</td>
184 * <td headers="matches">All ASCII:{@code [\x00-\x7F]}</td></tr>
185 * <tr><td valign="top" headers="construct posix">{@code \p{Alpha}}</td>
186 * <td headers="matches">An alphabetic character:{@code [\p{Lower}\p{Upper}]}</td></tr>
187 * <tr><td valign="top" headers="construct posix">{@code \p{Digit}}</td>
188 * <td headers="matches">A decimal digit: {@code [0-9]}</td></tr>
189 * <tr><td valign="top" headers="construct posix">{@code \p{Alnum}}</td>
190 * <td headers="matches">An alphanumeric character:{@code [\p{Alpha}\p{Digit}]}</td></tr>
191 * <tr><td valign="top" headers="construct posix">{@code \p{Punct}}</td>
192 * <td headers="matches">Punctuation: One of {@code !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~}</td></tr>
193 * <!-- {@code [\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]}
194 * {@code [\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]} -->
195 * <tr><td valign="top" headers="construct posix">{@code \p{Graph}}</td>
196 * <td headers="matches">A visible character: {@code [\p{Alnum}\p{Punct}]}</td></tr>
197 * <tr><td valign="top" headers="construct posix">{@code \p{Print}}</td>
198 * <td headers="matches">A printable character: {@code [\p{Graph}\x20]}</td></tr>
199 * <tr><td valign="top" headers="construct posix">{@code \p{Blank}}</td>
200 * <td headers="matches">A space or a tab: {@code [ \t]}</td></tr>
201 * <tr><td valign="top" headers="construct posix">{@code \p{Cntrl}}</td>
202 * <td headers="matches">A control character: {@code [\x00-\x1F\x7F]}</td></tr>
203 * <tr><td valign="top" headers="construct posix">{@code \p{XDigit}}</td>
204 * <td headers="matches">A hexadecimal digit: {@code [0-9a-fA-F]}</td></tr>
205 * <tr><td valign="top" headers="construct posix">{@code \p{Space}}</td>
206 * <td headers="matches">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr>
207 *
208 * <tr><th> </th></tr>
209 * <tr align="left"><th colspan="2">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr>
210 *
211 * <tr><td valign="top">{@code \p{javaLowerCase}}</td>
212 * <td>Equivalent to java.lang.Character.isLowerCase()</td></tr>
213 * <tr><td valign="top">{@code \p{javaUpperCase}}</td>
214 * <td>Equivalent to java.lang.Character.isUpperCase()</td></tr>
215 * <tr><td valign="top">{@code \p{javaWhitespace}}</td>
216 * <td>Equivalent to java.lang.Character.isWhitespace()</td></tr>
217 * <tr><td valign="top">{@code \p{javaMirrored}}</td>
218 * <td>Equivalent to java.lang.Character.isMirrored()</td></tr>
219 *
220 * <tr><th> </th></tr>
221 * <tr align="left"><th colspan="2" id="unicode">Classes for Unicode scripts, blocks, categories and binary properties</th></tr>
222 * <tr><td valign="top" headers="construct unicode">{@code \p{IsLatin}}</td>
223 * <td headers="matches">A Latin script character (<a href="#usc">script</a>)</td></tr>
224 * <tr><td valign="top" headers="construct unicode">{@code \p{InGreek}}</td>
225 * <td headers="matches">A character in the Greek block (<a href="#ubc">block</a>)</td></tr>
226 * <tr><td valign="top" headers="construct unicode">{@code \p{Lu}}</td>
227 * <td headers="matches">An uppercase letter (<a href="#ucc">category</a>)</td></tr>
228 * <tr><td valign="top" headers="construct unicode">{@code \p{IsAlphabetic}}</td>
229 * <td headers="matches">An alphabetic character (<a href="#ubpc">binary property</a>)</td></tr>
230 * <tr><td valign="top" headers="construct unicode">{@code \p{Sc}}</td>
231 * <td headers="matches">A currency symbol</td></tr>
232 * <tr><td valign="top" headers="construct unicode">{@code \P{InGreek}}</td>
233 * <td headers="matches">Any character except one in the Greek block (negation)</td></tr>
234 * <tr><td valign="top" headers="construct unicode">{@code [\p{L}&&[^\p{Lu}]]}</td>
235 * <td headers="matches">Any letter except an uppercase letter (subtraction)</td></tr>
236 *
237 * <tr><th> </th></tr>
238 * <tr align="left"><th colspan="2" id="bounds">Boundary matchers</th></tr>
239 *
240 * <tr><td valign="top" headers="construct bounds">{@code ^}</td>
241 * <td headers="matches">The beginning of a line</td></tr>
242 * <tr><td valign="top" headers="construct bounds">{@code $}</td>
243 * <td headers="matches">The end of a line</td></tr>
244 * <tr><td valign="top" headers="construct bounds">{@code \b}</td>
245 * <td headers="matches">A word boundary</td></tr>
246 * <tr><td valign="top" headers="construct bounds">{@code \B}</td>
247 * <td headers="matches">A non-word boundary</td></tr>
248 * <tr><td valign="top" headers="construct bounds">{@code \A}</td>
249 * <td headers="matches">The beginning of the input</td></tr>
250 * <tr><td valign="top" headers="construct bounds">{@code \G}</td>
251 * <td headers="matches">The end of the previous match</td></tr>
252 * <tr><td valign="top" headers="construct bounds">{@code \Z}</td>
253 * <td headers="matches">The end of the input but for the final
254 * <a href="#lt">terminator</a>, if any</td></tr>
255 * <tr><td valign="top" headers="construct bounds">{@code \z}</td>
256 * <td headers="matches">The end of the input</td></tr>
257 *
258 * <tr><th> </th></tr>
259 * <tr align="left"><th colspan="2" id="lineending">Linebreak matcher</th></tr>
260 * <tr><td valign="top" headers="construct lineending">{@code \R}</td>
261 * <td headers="matches">Any Unicode linebreak sequence, is equivalent to
262 * <code>\u000D\u000A|[\u000A\u000B\u000C\u000D\u0085\u2028\u2029]
263 * </code></td></tr>
264 *
265 * <tr><th> </th></tr>
266 * <tr align="left"><th colspan="2" id="greedy">Greedy quantifiers</th></tr>
267 *
268 * <tr><td valign="top" headers="construct greedy"><i>X</i>{@code ?}</td>
269 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
270 * <tr><td valign="top" headers="construct greedy"><i>X</i>{@code *}</td>
271 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
272 * <tr><td valign="top" headers="construct greedy"><i>X</i>{@code +}</td>
273 * <td headers="matches"><i>X</i>, one or more times</td></tr>
274 * <tr><td valign="top" headers="construct greedy"><i>X</i><code>{</code><i>n</i><code>}</code></td>
275 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
276 * <tr><td valign="top" headers="construct greedy"><i>X</i><code>{</code><i>n</i>{@code ,}}</td>
277 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
278 * <tr><td valign="top" headers="construct greedy"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}</code></td>
279 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
280 *
281 * <tr><th> </th></tr>
282 * <tr align="left"><th colspan="2" id="reluc">Reluctant quantifiers</th></tr>
283 *
284 * <tr><td valign="top" headers="construct reluc"><i>X</i>{@code ??}</td>
285 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
286 * <tr><td valign="top" headers="construct reluc"><i>X</i>{@code *?}</td>
287 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
288 * <tr><td valign="top" headers="construct reluc"><i>X</i>{@code +?}</td>
289 * <td headers="matches"><i>X</i>, one or more times</td></tr>
290 * <tr><td valign="top" headers="construct reluc"><i>X</i><code>{</code><i>n</i><code>}?</code></td>
291 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
292 * <tr><td valign="top" headers="construct reluc"><i>X</i><code>{</code><i>n</i><code>,}?</code></td>
293 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
294 * <tr><td valign="top" headers="construct reluc"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}?</code></td>
295 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
296 *
297 * <tr><th> </th></tr>
298 * <tr align="left"><th colspan="2" id="poss">Possessive quantifiers</th></tr>
299 *
300 * <tr><td valign="top" headers="construct poss"><i>X</i>{@code ?+}</td>
301 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
302 * <tr><td valign="top" headers="construct poss"><i>X</i>{@code *+}</td>
303 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
304 * <tr><td valign="top" headers="construct poss"><i>X</i>{@code ++}</td>
305 * <td headers="matches"><i>X</i>, one or more times</td></tr>
306 * <tr><td valign="top" headers="construct poss"><i>X</i><code>{</code><i>n</i><code>}+</code></td>
307 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
308 * <tr><td valign="top" headers="construct poss"><i>X</i><code>{</code><i>n</i><code>,}+</code></td>
309 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
310 * <tr><td valign="top" headers="construct poss"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}+</code></td>
311 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
312 *
313 * <tr><th> </th></tr>
314 * <tr align="left"><th colspan="2" id="logical">Logical operators</th></tr>
315 *
316 * <tr><td valign="top" headers="construct logical"><i>XY</i></td>
317 * <td headers="matches"><i>X</i> followed by <i>Y</i></td></tr>
318 * <tr><td valign="top" headers="construct logical"><i>X</i>{@code |}<i>Y</i></td>
319 * <td headers="matches">Either <i>X</i> or <i>Y</i></td></tr>
320 * <tr><td valign="top" headers="construct logical">{@code (}<i>X</i>{@code )}</td>
321 * <td headers="matches">X, as a <a href="#cg">capturing group</a></td></tr>
322 *
323 * <tr><th> </th></tr>
324 * <tr align="left"><th colspan="2" id="backref">Back references</th></tr>
325 *
326 * <tr><td valign="bottom" headers="construct backref">{@code \}<i>n</i></td>
327 * <td valign="bottom" headers="matches">Whatever the <i>n</i><sup>th</sup>
328 * <a href="#cg">capturing group</a> matched</td></tr>
329 *
330 * <tr><td valign="bottom" headers="construct backref">{@code \}<i>k</i><<i>name</i>></td>
331 * <td valign="bottom" headers="matches">Whatever the
332 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr>
333 *
334 * <tr><th> </th></tr>
335 * <tr align="left"><th colspan="2" id="quot">Quotation</th></tr>
336 *
337 * <tr><td valign="top" headers="construct quot">{@code \}</td>
338 * <td headers="matches">Nothing, but quotes the following character</td></tr>
339 * <tr><td valign="top" headers="construct quot">{@code \Q}</td>
340 * <td headers="matches">Nothing, but quotes all characters until {@code \E}</td></tr>
341 * <tr><td valign="top" headers="construct quot">{@code \E}</td>
342 * <td headers="matches">Nothing, but ends quoting started by {@code \Q}</td></tr>
343 * <!-- Metachars: !$()*+.<>?[\]^{|} -->
344 *
345 * <tr><th> </th></tr>
346 * <tr align="left"><th colspan="2" id="special">Special constructs (named-capturing and non-capturing)</th></tr>
347 *
348 * <tr><td valign="top" headers="construct special"><code>(?<<a href="#groupname">name</a>></code><i>X</i>{@code )}</td>
349 * <td headers="matches"><i>X</i>, as a named-capturing group</td></tr>
350 * <tr><td valign="top" headers="construct special">{@code (?:}<i>X</i>{@code )}</td>
351 * <td headers="matches"><i>X</i>, as a non-capturing group</td></tr>
352 * <tr><td valign="top" headers="construct special"><code>(?idmsuxU-idmsuxU) </code></td>
353 * <td headers="matches">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a>
354 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a>
355 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a>
356 * on - off</td></tr>
357 * <tr><td valign="top" headers="construct special"><code>(?idmsux-idmsux:</code><i>X</i>{@code )} </td>
358 * <td headers="matches"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the
359 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a>
360 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a >
361 * <a href="#COMMENTS">x</a> on - off</td></tr>
362 * <tr><td valign="top" headers="construct special">{@code (?=}<i>X</i>{@code )}</td>
363 * <td headers="matches"><i>X</i>, via zero-width positive lookahead</td></tr>
364 * <tr><td valign="top" headers="construct special">{@code (?!}<i>X</i>{@code )}</td>
365 * <td headers="matches"><i>X</i>, via zero-width negative lookahead</td></tr>
366 * <tr><td valign="top" headers="construct special">{@code (?<=}<i>X</i>{@code )}</td>
367 * <td headers="matches"><i>X</i>, via zero-width positive lookbehind</td></tr>
368 * <tr><td valign="top" headers="construct special">{@code (?<!}<i>X</i>{@code )}</td>
369 * <td headers="matches"><i>X</i>, via zero-width negative lookbehind</td></tr>
370 * <tr><td valign="top" headers="construct special">{@code (?>}<i>X</i>{@code )}</td>
371 * <td headers="matches"><i>X</i>, as an independent, non-capturing group</td></tr>
372 *
373 * </table>
374 *
375 * <hr>
376 *
377 *
378 * <h3><a name="bs">Backslashes, escapes, and quoting</a></h3>
379 *
380 * <p> The backslash character ({@code '\'}) serves to introduce escaped
381 * constructs, as defined in the table above, as well as to quote characters
382 * that otherwise would be interpreted as unescaped constructs. Thus the
383 * expression {@code \\} matches a single backslash and <code>\{</code> matches a
384 * left brace.
385 *
386 * <p> It is an error to use a backslash prior to any alphabetic character that
387 * does not denote an escaped construct; these are reserved for future
388 * extensions to the regular-expression language. A backslash may be used
389 * prior to a non-alphabetic character regardless of whether that character is
390 * part of an unescaped construct.
391 *
392 * <p> Backslashes within string literals in Java source code are interpreted
393 * as required by
394 * <cite>The Java™ Language Specification</cite>
395 * as either Unicode escapes (section 3.3) or other character escapes (section 3.10.6)
396 * It is therefore necessary to double backslashes in string
397 * literals that represent regular expressions to protect them from
398 * interpretation by the Java bytecode compiler. The string literal
399 * <code>"\b"</code>, for example, matches a single backspace character when
400 * interpreted as a regular expression, while {@code "\\b"} matches a
401 * word boundary. The string literal {@code "\(hello\)"} is illegal
402 * and leads to a compile-time error; in order to match the string
403 * {@code (hello)} the string literal {@code "\\(hello\\)"}
404 * must be used.
405 *
406 * <h3><a name="cc">Character Classes</a></h3>
407 *
408 * <p> Character classes may appear within other character classes, and
409 * may be composed by the union operator (implicit) and the intersection
410 * operator ({@code &&}).
411 * The union operator denotes a class that contains every character that is
412 * in at least one of its operand classes. The intersection operator
413 * denotes a class that contains every character that is in both of its
414 * operand classes.
415 *
416 * <p> The precedence of character-class operators is as follows, from
417 * highest to lowest:
418 *
419 * <blockquote><table border="0" cellpadding="1" cellspacing="0"
420 * summary="Precedence of character class operators.">
421 * <tr><th>1 </th>
422 * <td>Literal escape </td>
423 * <td>{@code \x}</td></tr>
424 * <tr><th>2 </th>
425 * <td>Grouping</td>
426 * <td>{@code [...]}</td></tr>
427 * <tr><th>3 </th>
428 * <td>Range</td>
429 * <td>{@code a-z}</td></tr>
430 * <tr><th>4 </th>
431 * <td>Union</td>
432 * <td>{@code [a-e][i-u]}</td></tr>
433 * <tr><th>5 </th>
434 * <td>Intersection</td>
435 * <td>{@code [a-z&&[aeiou]]}</td></tr>
436 * </table></blockquote>
437 *
438 * <p> Note that a different set of metacharacters are in effect inside
439 * a character class than outside a character class. For instance, the
440 * regular expression {@code .} loses its special meaning inside a
441 * character class, while the expression {@code -} becomes a range
442 * forming metacharacter.
443 *
444 * <h3><a name="lt">Line terminators</a></h3>
445 *
446 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks
447 * the end of a line of the input character sequence. The following are
448 * recognized as line terminators:
449 *
450 * <ul>
451 *
452 * <li> A newline (line feed) character ({@code '\n'}),
453 *
454 * <li> A carriage-return character followed immediately by a newline
455 * character ({@code "\r\n"}),
456 *
457 * <li> A standalone carriage-return character ({@code '\r'}),
458 *
459 * <li> A next-line character (<code>'\u0085'</code>),
460 *
461 * <li> A line-separator character (<code>'\u2028'</code>), or
462 *
463 * <li> A paragraph-separator character (<code>'\u2029</code>).
464 *
465 * </ul>
466 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators
467 * recognized are newline characters.
468 *
469 * <p> The regular expression {@code .} matches any character except a line
470 * terminator unless the {@link #DOTALL} flag is specified.
471 *
472 * <p> By default, the regular expressions {@code ^} and {@code $} ignore
473 * line terminators and only match at the beginning and the end, respectively,
474 * of the entire input sequence. If {@link #MULTILINE} mode is activated then
475 * {@code ^} matches at the beginning of input and after any line terminator
476 * except at the end of input. When in {@link #MULTILINE} mode {@code $}
477 * matches just before a line terminator or the end of the input sequence.
478 *
479 * <h3><a name="cg">Groups and capturing</a></h3>
480 *
481 * <h4><a name="gnumber">Group number</a></h4>
482 * <p> Capturing groups are numbered by counting their opening parentheses from
483 * left to right. In the expression {@code ((A)(B(C)))}, for example, there
484 * are four such groups: </p>
485 *
486 * <blockquote><table cellpadding=1 cellspacing=0 summary="Capturing group numberings">
487 * <tr><th>1 </th>
488 * <td>{@code ((A)(B(C)))}</td></tr>
489 * <tr><th>2 </th>
490 * <td>{@code (A)}</td></tr>
491 * <tr><th>3 </th>
492 * <td>{@code (B(C))}</td></tr>
493 * <tr><th>4 </th>
494 * <td>{@code (C)}</td></tr>
495 * </table></blockquote>
496 *
497 * <p> Group zero always stands for the entire expression.
498 *
499 * <p> Capturing groups are so named because, during a match, each subsequence
500 * of the input sequence that matches such a group is saved. The captured
501 * subsequence may be used later in the expression, via a back reference, and
502 * may also be retrieved from the matcher once the match operation is complete.
503 *
504 * <h4><a name="groupname">Group name</a></h4>
505 * <p>A capturing group can also be assigned a "name", a {@code named-capturing group},
506 * and then be back-referenced later by the "name". Group names are composed of
507 * the following characters. The first character must be a {@code letter}.
508 *
509 * <ul>
510 * <li> The uppercase letters {@code 'A'} through {@code 'Z'}
511 * (<code>'\u0041'</code> through <code>'\u005a'</code>),
512 * <li> The lowercase letters {@code 'a'} through {@code 'z'}
513 * (<code>'\u0061'</code> through <code>'\u007a'</code>),
514 * <li> The digits {@code '0'} through {@code '9'}
515 * (<code>'\u0030'</code> through <code>'\u0039'</code>),
516 * </ul>
517 *
518 * <p> A {@code named-capturing group} is still numbered as described in
519 * <a href="#gnumber">Group number</a>.
520 *
521 * <p> The captured input associated with a group is always the subsequence
522 * that the group most recently matched. If a group is evaluated a second time
523 * because of quantification then its previously-captured value, if any, will
524 * be retained if the second evaluation fails. Matching the string
525 * {@code "aba"} against the expression {@code (a(b)?)+}, for example, leaves
526 * group two set to {@code "b"}. All captured input is discarded at the
527 * beginning of each match.
528 *
529 * <p> Groups beginning with {@code (?} are either pure, <i>non-capturing</i> groups
530 * that do not capture text and do not count towards the group total, or
531 * <i>named-capturing</i> group.
532 *
533 * <h3> Unicode support </h3>
534 *
535 * <p> This class is in conformance with Level 1 of <a
536 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical
537 * Standard #18: Unicode Regular Expression</i></a>, plus RL2.1
538 * Canonical Equivalents.
539 * <p>
540 * <b>Unicode escape sequences</b> such as <code>\u2014</code> in Java source code
541 * are processed as described in section 3.3 of
542 * <cite>The Java™ Language Specification</cite>.
543 * Such escape sequences are also implemented directly by the regular-expression
544 * parser so that Unicode escapes can be used in expressions that are read from
545 * files or from the keyboard. Thus the strings <code>"\u2014"</code> and
546 * {@code "\\u2014"}, while not equal, compile into the same pattern, which
547 * matches the character with hexadecimal value {@code 0x2014}.
548 * <p>
549 * A Unicode character can also be represented in a regular-expression by
550 * using its <b>Hex notation</b>(hexadecimal code point value) directly as described in construct
551 * <code>\x{...}</code>, for example a supplementary character U+2011F
552 * can be specified as <code>\x{2011F}</code>, instead of two consecutive
553 * Unicode escape sequences of the surrogate pair
554 * <code>\uD840</code><code>\uDD1F</code>.
555 * <p>
556 * Unicode scripts, blocks, categories and binary properties are written with
557 * the {@code \p} and {@code \P} constructs as in Perl.
558 * <code>\p{</code><i>prop</i><code>}</code> matches if
559 * the input has the property <i>prop</i>, while <code>\P{</code><i>prop</i><code>}</code>
560 * does not match if the input has that property.
561 * <p>
562 * Scripts, blocks, categories and binary properties can be used both inside
563 * and outside of a character class.
564 *
565 * <p>
566 * <b><a name="usc">Scripts</a></b> are specified either with the prefix {@code Is}, as in
567 * {@code IsHiragana}, or by using the {@code script} keyword (or its short
568 * form {@code sc}) as in {@code script=Hiragana} or {@code sc=Hiragana}.
569 * <p>
570 * The script names supported by {@code Pattern} are the valid script names
571 * accepted and defined by
572 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}.
573 *
574 * <p>
575 * <b><a name="ubc">Blocks</a></b> are specified with the prefix {@code In}, as in
576 * {@code InMongolian}, or by using the keyword {@code block} (or its short
577 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}.
578 * <p>
579 * The block names supported by {@code Pattern} are the valid block names
580 * accepted and defined by
581 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}.
582 * <p>
583 *
584 * <b><a name="ucc">Categories</a></b> may be specified with the optional prefix {@code Is}:
585 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode
586 * letters. Same as scripts and blocks, categories can also be specified
587 * by using the keyword {@code general_category} (or its short form
588 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}.
589 * <p>
590 * The supported categories are those of
591 * <a href="http://www.unicode.org/unicode/standard/standard.html">
592 * <i>The Unicode Standard</i></a> in the version specified by the
593 * {@link java.lang.Character Character} class. The category names are those
594 * defined in the Standard, both normative and informative.
595 * <p>
596 *
597 * <b><a name="ubpc">Binary properties</a></b> are specified with the prefix {@code Is}, as in
598 * {@code IsAlphabetic}. The supported binary properties by {@code Pattern}
599 * are
600 * <ul>
601 * <li> Alphabetic
602 * <li> Ideographic
603 * <li> Letter
604 * <li> Lowercase
605 * <li> Uppercase
606 * <li> Titlecase
607 * <li> Punctuation
608 * <Li> Control
609 * <li> White_Space
610 * <li> Digit
611 * <li> Hex_Digit
612 * <li> Join_Control
613 * <li> Noncharacter_Code_Point
614 * <li> Assigned
615 * </ul>
616 * <p>
617 * The following <b>Predefined Character classes</b> and <b>POSIX character classes</b>
618 * are in conformance with the recommendation of <i>Annex C: Compatibility Properties</i>
619 * of <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Regular Expression
620 * </i></a>, when {@link #UNICODE_CHARACTER_CLASS} flag is specified.
621 *
622 * <table border="0" cellpadding="1" cellspacing="0"
623 * summary="predefined and posix character classes in Unicode mode">
624 * <tr align="left">
625 * <th align="left" id="predef_classes">Classes</th>
626 * <th align="left" id="predef_matches">Matches</th>
627 *</tr>
628 * <tr><td>{@code \p{Lower}}</td>
629 * <td>A lowercase character:{@code \p{IsLowercase}}</td></tr>
630 * <tr><td>{@code \p{Upper}}</td>
631 * <td>An uppercase character:{@code \p{IsUppercase}}</td></tr>
632 * <tr><td>{@code \p{ASCII}}</td>
633 * <td>All ASCII:{@code [\x00-\x7F]}</td></tr>
634 * <tr><td>{@code \p{Alpha}}</td>
635 * <td>An alphabetic character:{@code \p{IsAlphabetic}}</td></tr>
636 * <tr><td>{@code \p{Digit}}</td>
637 * <td>A decimal digit character:{@code p{IsDigit}}</td></tr>
638 * <tr><td>{@code \p{Alnum}}</td>
639 * <td>An alphanumeric character:{@code [\p{IsAlphabetic}\p{IsDigit}]}</td></tr>
640 * <tr><td>{@code \p{Punct}}</td>
641 * <td>A punctuation character:{@code p{IsPunctuation}}</td></tr>
642 * <tr><td>{@code \p{Graph}}</td>
643 * <td>A visible character: {@code [^\p{IsWhite_Space}\p{gc=Cc}\p{gc=Cs}\p{gc=Cn}]}</td></tr>
644 * <tr><td>{@code \p{Print}}</td>
645 * <td>A printable character: {@code [\p{Graph}\p{Blank}&&[^\p{Cntrl}]]}</td></tr>
646 * <tr><td>{@code \p{Blank}}</td>
647 * <td>A space or a tab: {@code [\p{IsWhite_Space}&&[^\p{gc=Zl}\p{gc=Zp}\x0a\x0b\x0c\x0d\x85]]}</td></tr>
648 * <tr><td>{@code \p{Cntrl}}</td>
649 * <td>A control character: {@code \p{gc=Cc}}</td></tr>
650 * <tr><td>{@code \p{XDigit}}</td>
651 * <td>A hexadecimal digit: {@code [\p{gc=Nd}\p{IsHex_Digit}]}</td></tr>
652 * <tr><td>{@code \p{Space}}</td>
653 * <td>A whitespace character:{@code \p{IsWhite_Space}}</td></tr>
654 * <tr><td>{@code \d}</td>
655 * <td>A digit: {@code \p{IsDigit}}</td></tr>
656 * <tr><td>{@code \D}</td>
657 * <td>A non-digit: {@code [^\d]}</td></tr>
658 * <tr><td>{@code \s}</td>
659 * <td>A whitespace character: {@code \p{IsWhite_Space}}</td></tr>
660 * <tr><td>{@code \S}</td>
661 * <td>A non-whitespace character: {@code [^\s]}</td></tr>
662 * <tr><td>{@code \w}</td>
663 * <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>
664 * <tr><td>{@code \W}</td>
665 * <td>A non-word character: {@code [^\w]}</td></tr>
666 * </table>
667 * <p>
668 * <a name="jcc">
669 * Categories that behave like the java.lang.Character
670 * boolean is<i>methodname</i> methods (except for the deprecated ones) are
671 * available through the same <code>\p{</code><i>prop</i><code>}</code> syntax where
672 * the specified property has the name <code>java<i>methodname</i></code></a>.
673 *
674 * <h3> Comparison to Perl 5 </h3>
675 *
676 * <p>The {@code Pattern} engine performs traditional NFA-based matching
677 * with ordered alternation as occurs in Perl 5.
678 *
679 * <p> Perl constructs not supported by this class: </p>
680 *
681 * <ul>
682 * <li><p> Predefined character classes (Unicode character)
683 * <p><code>\X </code>Match Unicode
684 * <a href="http://www.unicode.org/reports/tr18/#Default_Grapheme_Clusters">
685 * <i>extended grapheme cluster</i></a>
686 * </p></li>
687 *
688 * <li><p> The backreference constructs, <code>\g{</code><i>n</i><code>}</code> for
689 * the <i>n</i><sup>th</sup><a href="#cg">capturing group</a> and
690 * <code>\g{</code><i>name</i><code>}</code> for
691 * <a href="#groupname">named-capturing group</a>.
692 * </p></li>
693 *
694 * <li><p> The named character construct, <code>\N{</code><i>name</i><code>}</code>
695 * for a Unicode character by its name.
696 * </p></li>
697 *
698 * <li><p> The conditional constructs
699 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code )} and
700 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code |}<i>Y</i>{@code )},
701 * </p></li>
702 *
703 * <li><p> The embedded code constructs <code>(?{</code><i>code</i><code>})</code>
704 * and <code>(??{</code><i>code</i><code>})</code>,</p></li>
705 *
706 * <li><p> The embedded comment syntax {@code (?#comment)}, and </p></li>
707 *
708 * <li><p> The preprocessing operations {@code \l} <code>\u</code>,
709 * {@code \L}, and {@code \U}. </p></li>
710 *
711 * </ul>
712 *
713 * <p> Constructs supported by this class but not by Perl: </p>
714 *
715 * <ul>
716 *
717 * <li><p> Character-class union and intersection as described
718 * <a href="#cc">above</a>.</p></li>
719 *
720 * </ul>
721 *
722 * <p> Notable differences from Perl: </p>
723 *
724 * <ul>
725 *
726 * <li><p> In Perl, {@code \1} through {@code \9} are always interpreted
727 * as back references; a backslash-escaped number greater than {@code 9} is
728 * treated as a back reference if at least that many subexpressions exist,
729 * otherwise it is interpreted, if possible, as an octal escape. In this
730 * class octal escapes must always begin with a zero. In this class,
731 * {@code \1} through {@code \9} are always interpreted as back
732 * references, and a larger number is accepted as a back reference if at
733 * least that many subexpressions exist at that point in the regular
734 * expression, otherwise the parser will drop digits until the number is
735 * smaller or equal to the existing number of groups or it is one digit.
736 * </p></li>
737 *
738 * <li><p> Perl uses the {@code g} flag to request a match that resumes
739 * where the last match left off. This functionality is provided implicitly
740 * by the {@link Matcher} class: Repeated invocations of the {@link
741 * Matcher#find find} method will resume where the last match left off,
742 * unless the matcher is reset. </p></li>
743 *
744 * <li><p> In Perl, embedded flags at the top level of an expression affect
745 * the whole expression. In this class, embedded flags always take effect
746 * at the point at which they appear, whether they are at the top level or
747 * within a group; in the latter case, flags are restored at the end of the
748 * group just as in Perl. </p></li>
749 *
750 * </ul>
751 *
752 *
753 * <p> For a more precise description of the behavior of regular expression
754 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/">
755 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl,
756 * O'Reilly and Associates, 2006.</a>
757 * </p>
758 *
759 * @see java.lang.String#split(String, int)
760 * @see java.lang.String#split(String)
761 *
762 * @author Mike McCloskey
763 * @author Mark Reinhold
764 * @author JSR-51 Expert Group
765 * @since 1.4
766 * @spec JSR-51
767 */
768
769 public final class Pattern
770 implements java.io.Serializable
771 {
772
773 /**
774 * Regular expression modifier values. Instead of being passed as
775 * arguments, they can also be passed as inline modifiers.
776 * For example, the following statements have the same effect.
777 * <pre>
778 * RegExp r1 = RegExp.compile("abc", Pattern.I|Pattern.M);
779 * RegExp r2 = RegExp.compile("(?im)abc", 0);
780 * </pre>
781 *
782 * The flags are duplicated so that the familiar Perl match flag
783 * names are available.
784 */
785
786 /**
787 * Enables Unix lines mode.
788 *
789 * <p> In this mode, only the {@code '\n'} line terminator is recognized
790 * in the behavior of {@code .}, {@code ^}, and {@code $}.
791 *
792 * <p> Unix lines mode can also be enabled via the embedded flag
793 * expression {@code (?d)}.
794 */
795 public static final int UNIX_LINES = 0x01;
796
797 /**
798 * Enables case-insensitive matching.
799 *
800 * <p> By default, case-insensitive matching assumes that only characters
801 * in the US-ASCII charset are being matched. Unicode-aware
802 * case-insensitive matching can be enabled by specifying the {@link
803 * #UNICODE_CASE} flag in conjunction with this flag.
804 *
805 * <p> Case-insensitive matching can also be enabled via the embedded flag
806 * expression {@code (?i)}.
807 *
808 * <p> Specifying this flag may impose a slight performance penalty. </p>
809 */
810 public static final int CASE_INSENSITIVE = 0x02;
811
812 /**
813 * Permits whitespace and comments in pattern.
814 *
815 * <p> In this mode, whitespace is ignored, and embedded comments starting
816 * with {@code #} are ignored until the end of a line.
817 *
818 * <p> Comments mode can also be enabled via the embedded flag
819 * expression {@code (?x)}.
820 */
821 public static final int COMMENTS = 0x04;
822
823 /**
824 * Enables multiline mode.
825 *
826 * <p> In multiline mode the expressions {@code ^} and {@code $} match
827 * just after or just before, respectively, a line terminator or the end of
828 * the input sequence. By default these expressions only match at the
829 * beginning and the end of the entire input sequence.
830 *
831 * <p> Multiline mode can also be enabled via the embedded flag
832 * expression {@code (?m)}. </p>
833 */
834 public static final int MULTILINE = 0x08;
835
836 /**
837 * Enables literal parsing of the pattern.
838 *
839 * <p> When this flag is specified then the input string that specifies
840 * the pattern is treated as a sequence of literal characters.
841 * Metacharacters or escape sequences in the input sequence will be
842 * given no special meaning.
843 *
844 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on
845 * matching when used in conjunction with this flag. The other flags
846 * become superfluous.
847 *
848 * <p> There is no embedded flag character for enabling literal parsing.
849 * @since 1.5
850 */
851 public static final int LITERAL = 0x10;
852
853 /**
854 * Enables dotall mode.
855 *
856 * <p> In dotall mode, the expression {@code .} matches any character,
857 * including a line terminator. By default this expression does not match
858 * line terminators.
859 *
860 * <p> Dotall mode can also be enabled via the embedded flag
861 * expression {@code (?s)}. (The {@code s} is a mnemonic for
862 * "single-line" mode, which is what this is called in Perl.) </p>
863 */
864 public static final int DOTALL = 0x20;
865
866 /**
867 * Enables Unicode-aware case folding.
868 *
869 * <p> When this flag is specified then case-insensitive matching, when
870 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner
871 * consistent with the Unicode Standard. By default, case-insensitive
872 * matching assumes that only characters in the US-ASCII charset are being
873 * matched.
874 *
875 * <p> Unicode-aware case folding can also be enabled via the embedded flag
876 * expression {@code (?u)}.
877 *
878 * <p> Specifying this flag may impose a performance penalty. </p>
879 */
880 public static final int UNICODE_CASE = 0x40;
881
882 /**
883 * Enables canonical equivalence.
884 *
885 * <p> When this flag is specified then two characters will be considered
886 * to match if, and only if, their full canonical decompositions match.
887 * The expression <code>"a\u030A"</code>, for example, will match the
888 * string <code>"\u00E5"</code> when this flag is specified. By default,
889 * matching does not take canonical equivalence into account.
890 *
891 * <p> There is no embedded flag character for enabling canonical
892 * equivalence.
893 *
894 * <p> Specifying this flag may impose a performance penalty. </p>
895 */
896 public static final int CANON_EQ = 0x80;
897
898 /**
899 * Enables the Unicode version of <i>Predefined character classes</i> and
900 * <i>POSIX character classes</i>.
901 *
902 * <p> When this flag is specified then the (US-ASCII only)
903 * <i>Predefined character classes</i> and <i>POSIX character classes</i>
904 * are in conformance with
905 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical
906 * Standard #18: Unicode Regular Expression</i></a>
907 * <i>Annex C: Compatibility Properties</i>.
908 * <p>
909 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded
910 * flag expression {@code (?U)}.
911 * <p>
912 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case
913 * folding.
914 * <p>
915 * Specifying this flag may impose a performance penalty. </p>
916 * @since 1.7
917 */
918 public static final int UNICODE_CHARACTER_CLASS = 0x100;
919
920 /**
921 * Contains all possible flags for compile(regex, flags).
922 */
923 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE |
924 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL |
925 UNICODE_CHARACTER_CLASS | COMMENTS;
926
927 /* Pattern has only two serialized components: The pattern string
928 * and the flags, which are all that is needed to recompile the pattern
929 * when it is deserialized.
930 */
931
932 /** use serialVersionUID from Merlin b59 for interoperability */
933 private static final long serialVersionUID = 5073258162644648461L;
934
935 /**
936 * The original regular-expression pattern string.
937 *
938 * @serial
939 */
940 private String pattern;
941
942 /**
943 * The original pattern flags.
944 *
945 * @serial
946 */
947 private int flags;
948
949 /**
950 * Boolean indicating this Pattern is compiled; this is necessary in order
951 * to lazily compile deserialized Patterns.
952 */
953 private transient volatile boolean compiled = false;
954
955 /**
956 * The normalized pattern string.
957 */
958 private transient String normalizedPattern;
959
960 /**
961 * The starting point of state machine for the find operation. This allows
962 * a match to start anywhere in the input.
963 */
964 transient Node root;
965
966 /**
967 * The root of object tree for a match operation. The pattern is matched
968 * at the beginning. This may include a find that uses BnM or a First
969 * node.
970 */
971 transient Node matchRoot;
972
973 /**
974 * Temporary storage used by parsing pattern slice.
975 */
976 transient int[] buffer;
977
978 /**
979 * Map the "name" of the "named capturing group" to its group id
980 * node.
981 */
982 transient volatile Map<String, Integer> namedGroups;
983
984 /**
985 * Temporary storage used while parsing group references.
986 */
987 transient GroupHead[] groupNodes;
988
989 /**
990 * Temporary null terminated code point array used by pattern compiling.
991 */
992 private transient int[] temp;
993
994 /**
995 * The number of capturing groups in this Pattern. Used by matchers to
996 * allocate storage needed to perform a match.
997 */
998 transient int capturingGroupCount;
999
1000 /**
1001 * The local variable count used by parsing tree. Used by matchers to
1002 * allocate storage needed to perform a match.
1003 */
1004 transient int localCount;
1005
1006 /**
1007 * Index into the pattern string that keeps track of how much has been
1008 * parsed.
1009 */
1010 private transient int cursor;
1011
1012 /**
1013 * Holds the length of the pattern string.
1014 */
1015 private transient int patternLength;
1016
1017 /**
1018 * If the Start node might possibly match supplementary characters.
1019 * It is set to true during compiling if
1020 * (1) There is supplementary char in pattern, or
1021 * (2) There is complement node of Category or Block
1022 */
1023 private transient boolean hasSupplementary;
1024
1025 /**
1026 * Compiles the given regular expression into a pattern.
1027 *
1028 * @param regex
1029 * The expression to be compiled
1030 * @return the given regular expression compiled into a pattern
1031 * @throws PatternSyntaxException
1032 * If the expression's syntax is invalid
1033 */
1034 public static Pattern compile(String regex) {
1035 return new Pattern(regex, 0);
1036 }
1037
1038 /**
1039 * Compiles the given regular expression into a pattern with the given
1040 * flags.
1041 *
1042 * @param regex
1043 * The expression to be compiled
1044 *
1045 * @param flags
1046 * Match flags, a bit mask that may include
1047 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL},
1048 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES},
1049 * {@link #LITERAL}, {@link #UNICODE_CHARACTER_CLASS}
1050 * and {@link #COMMENTS}
1051 *
1052 * @return the given regular expression compiled into a pattern with the given flags
1053 * @throws IllegalArgumentException
1054 * If bit values other than those corresponding to the defined
1055 * match flags are set in {@code flags}
1056 *
1057 * @throws PatternSyntaxException
1058 * If the expression's syntax is invalid
1059 */
1060 public static Pattern compile(String regex, int flags) {
1061 return new Pattern(regex, flags);
1062 }
1063
1064 /**
1065 * Returns the regular expression from which this pattern was compiled.
1066 *
1067 * @return The source of this pattern
1068 */
1069 public String pattern() {
1070 return pattern;
1071 }
1072
1073 /**
1074 * <p>Returns the string representation of this pattern. This
1075 * is the regular expression from which this pattern was
1076 * compiled.</p>
1077 *
1078 * @return The string representation of this pattern
1079 * @since 1.5
1080 */
1081 public String toString() {
1082 return pattern;
1083 }
1084
1085 /**
1086 * Creates a matcher that will match the given input against this pattern.
1087 *
1088 * @param input
1089 * The character sequence to be matched
1090 *
1091 * @return A new matcher for this pattern
1092 */
1093 public Matcher matcher(CharSequence input) {
1094 if (!compiled) {
1095 synchronized(this) {
1096 if (!compiled)
1097 compile();
1098 }
1099 }
1100 Matcher m = new Matcher(this, input);
1101 return m;
1102 }
1103
1104 /**
1105 * Returns this pattern's match flags.
1106 *
1107 * @return The match flags specified when this pattern was compiled
1108 */
1109 public int flags() {
1110 return flags;
1111 }
1112
1113 /**
1114 * Compiles the given regular expression and attempts to match the given
1115 * input against it.
1116 *
1117 * <p> An invocation of this convenience method of the form
1118 *
1119 * <blockquote><pre>
1120 * Pattern.matches(regex, input);</pre></blockquote>
1121 *
1122 * behaves in exactly the same way as the expression
1123 *
1124 * <blockquote><pre>
1125 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote>
1126 *
1127 * <p> If a pattern is to be used multiple times, compiling it once and reusing
1128 * it will be more efficient than invoking this method each time. </p>
1129 *
1130 * @param regex
1131 * The expression to be compiled
1132 *
1133 * @param input
1134 * The character sequence to be matched
1135 * @return whether or not the regular expression matches on the input
1136 * @throws PatternSyntaxException
1137 * If the expression's syntax is invalid
1138 */
1139 public static boolean matches(String regex, CharSequence input) {
1140 Pattern p = Pattern.compile(regex);
1141 Matcher m = p.matcher(input);
1142 return m.matches();
1143 }
1144
1145 /**
1146 * Splits the given input sequence around matches of this pattern.
1147 *
1148 * <p> The array returned by this method contains each substring of the
1149 * input sequence that is terminated by another subsequence that matches
1150 * this pattern or is terminated by the end of the input sequence. The
1151 * substrings in the array are in the order in which they occur in the
1152 * input. If this pattern does not match any subsequence of the input then
1153 * the resulting array has just one element, namely the input sequence in
1154 * string form.
1155 *
1156 * <p> When there is a positive-width match at the beginning of the input
1157 * sequence then an empty leading substring is included at the beginning
1158 * of the resulting array. A zero-width match at the beginning however
1159 * never produces such empty leading substring.
1160 *
1161 * <p> The {@code limit} parameter controls the number of times the
1162 * pattern is applied and therefore affects the length of the resulting
1163 * array. If the limit <i>n</i> is greater than zero then the pattern
1164 * will be applied at most <i>n</i> - 1 times, the array's
1165 * length will be no greater than <i>n</i>, and the array's last entry
1166 * will contain all input beyond the last matched delimiter. If <i>n</i>
1167 * is non-positive then the pattern will be applied as many times as
1168 * possible and the array can have any length. If <i>n</i> is zero then
1169 * the pattern will be applied as many times as possible, the array can
1170 * have any length, and trailing empty strings will be discarded.
1171 *
1172 * <p> The input {@code "boo:and:foo"}, for example, yields the following
1173 * results with these parameters:
1174 *
1175 * <blockquote><table cellpadding=1 cellspacing=0
1176 * summary="Split examples showing regex, limit, and result">
1177 * <tr><th align="left"><i>Regex </i></th>
1178 * <th align="left"><i>Limit </i></th>
1179 * <th align="left"><i>Result </i></th></tr>
1180 * <tr><td align=center>:</td>
1181 * <td align=center>2</td>
1182 * <td>{@code { "boo", "and:foo" }}</td></tr>
1183 * <tr><td align=center>:</td>
1184 * <td align=center>5</td>
1185 * <td>{@code { "boo", "and", "foo" }}</td></tr>
1186 * <tr><td align=center>:</td>
1187 * <td align=center>-2</td>
1188 * <td>{@code { "boo", "and", "foo" }}</td></tr>
1189 * <tr><td align=center>o</td>
1190 * <td align=center>5</td>
1191 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr>
1192 * <tr><td align=center>o</td>
1193 * <td align=center>-2</td>
1194 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr>
1195 * <tr><td align=center>o</td>
1196 * <td align=center>0</td>
1197 * <td>{@code { "b", "", ":and:f" }}</td></tr>
1198 * </table></blockquote>
1199 *
1200 * @param input
1201 * The character sequence to be split
1202 *
1203 * @param limit
1204 * The result threshold, as described above
1205 *
1206 * @return The array of strings computed by splitting the input
1207 * around matches of this pattern
1208 */
1209 public String[] split(CharSequence input, int limit) {
1210 int index = 0;
1211 boolean matchLimited = limit > 0;
1212 ArrayList<String> matchList = new ArrayList<>();
1213 Matcher m = matcher(input);
1214
1215 // Add segments before each match found
1216 while(m.find()) {
1217 if (!matchLimited || matchList.size() < limit - 1) {
1218 if (index == 0 && index == m.start() && m.start() == m.end()) {
1219 // no empty leading substring included for zero-width match
1220 // at the beginning of the input char sequence.
1221 continue;
1222 }
1223 String match = input.subSequence(index, m.start()).toString();
1224 matchList.add(match);
1225 index = m.end();
1226 } else if (matchList.size() == limit - 1) { // last one
1227 String match = input.subSequence(index,
1228 input.length()).toString();
1229 matchList.add(match);
1230 index = m.end();
1231 }
1232 }
1233
1234 // If no match was found, return this
1235 if (index == 0)
1236 return new String[] {input.toString()};
1237
1238 // Add remaining segment
1239 if (!matchLimited || matchList.size() < limit)
1240 matchList.add(input.subSequence(index, input.length()).toString());
1241
1242 // Construct result
1243 int resultSize = matchList.size();
1244 if (limit == 0)
1245 while (resultSize > 0 && matchList.get(resultSize-1).equals(""))
1246 resultSize--;
1247 String[] result = new String[resultSize];
1248 return matchList.subList(0, resultSize).toArray(result);
1249 }
1250
1251 /**
1252 * Splits the given input sequence around matches of this pattern.
1253 *
1254 * <p> This method works as if by invoking the two-argument {@link
1255 * #split(java.lang.CharSequence, int) split} method with the given input
1256 * sequence and a limit argument of zero. Trailing empty strings are
1257 * therefore not included in the resulting array. </p>
1258 *
1259 * <p> The input {@code "boo:and:foo"}, for example, yields the following
1260 * results with these expressions:
1261 *
1262 * <blockquote><table cellpadding=1 cellspacing=0
1263 * summary="Split examples showing regex and result">
1264 * <tr><th align="left"><i>Regex </i></th>
1265 * <th align="left"><i>Result</i></th></tr>
1266 * <tr><td align=center>:</td>
1267 * <td>{@code { "boo", "and", "foo" }}</td></tr>
1268 * <tr><td align=center>o</td>
1269 * <td>{@code { "b", "", ":and:f" }}</td></tr>
1270 * </table></blockquote>
1271 *
1272 *
1273 * @param input
1274 * The character sequence to be split
1275 *
1276 * @return The array of strings computed by splitting the input
1277 * around matches of this pattern
1278 */
1279 public String[] split(CharSequence input) {
1280 return split(input, 0);
1281 }
1282
1283 /**
1284 * Returns a literal pattern {@code String} for the specified
1285 * {@code String}.
1286 *
1287 * <p>This method produces a {@code String} that can be used to
1288 * create a {@code Pattern} that would match the string
1289 * {@code s} as if it were a literal pattern.</p> Metacharacters
1290 * or escape sequences in the input sequence will be given no special
1291 * meaning.
1292 *
1293 * @param s The string to be literalized
1294 * @return A literal string replacement
1295 * @since 1.5
1296 */
1297 public static String quote(String s) {
1298 int slashEIndex = s.indexOf("\\E");
1299 if (slashEIndex == -1)
1300 return "\\Q" + s + "\\E";
1301
1302 int lenHint = s.length();
1303 lenHint = (lenHint < Integer.MAX_VALUE - 8 - lenHint) ?
1304 (lenHint << 1) : (Integer.MAX_VALUE - 8);
1305
1306 StringBuilder sb = new StringBuilder(lenHint);
1307 sb.append("\\Q");
1308 int current = 0;
1309 do {
1310 sb.append(s, current, slashEIndex)
1311 .append("\\E\\\\E\\Q");
1312 current = slashEIndex + 2;
1313 } while ((slashEIndex = s.indexOf("\\E", current)) != -1);
1314
1315 return sb.append(s, current, s.length())
1316 .append("\\E")
1317 .toString();
1318 }
1319
1320 /**
1321 * Recompile the Pattern instance from a stream. The original pattern
1322 * string is read in and the object tree is recompiled from it.
1323 */
1324 private void readObject(java.io.ObjectInputStream s)
1325 throws java.io.IOException, ClassNotFoundException {
1326
1327 // Read in all fields
1328 s.defaultReadObject();
1329
1330 // Initialize counts
1331 capturingGroupCount = 1;
1332 localCount = 0;
1333
1334 // if length > 0, the Pattern is lazily compiled
1335 compiled = false;
1336 if (pattern.length() == 0) {
1337 root = new Start(lastAccept);
1338 matchRoot = lastAccept;
1339 compiled = true;
1340 }
1341 }
1342
1343 /**
1344 * This private constructor is used to create all Patterns. The pattern
1345 * string and match flags are all that is needed to completely describe
1346 * a Pattern. An empty pattern string results in an object tree with
1347 * only a Start node and a LastNode node.
1348 */
1349 private Pattern(String p, int f) {
1350 if ((f & ~ALL_FLAGS) != 0) {
1351 throw new IllegalArgumentException("Unknown flag 0x"
1352 + Integer.toHexString(f));
1353 }
1354 pattern = p;
1355 flags = f;
1356
1357 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present
1358 if ((flags & UNICODE_CHARACTER_CLASS) != 0)
1359 flags |= UNICODE_CASE;
1360
1361 // Reset group index count
1362 capturingGroupCount = 1;
1363 localCount = 0;
1364
1365 if (pattern.length() > 0) {
1366 compile();
1367 } else {
1368 root = new Start(lastAccept);
1369 matchRoot = lastAccept;
1370 }
1371 }
1372
1373 /**
1374 * The pattern is converted to normalized form ({@linkplain
1375 * java.text.Normalizer.Form.NFD NFD}, canonical decomposition)
1376 * and then a pure group is constructed to match canonical
1377 * equivalences of the characters.
1378 */
1379 private void normalize() {
1380 boolean inCharClass = false;
1381 int lastCodePoint = -1;
1382
1383 // Convert pattern into normalized form
1384 normalizedPattern = Normalizer.normalize(pattern, Normalizer.Form.NFD);
1385 patternLength = normalizedPattern.length();
1386
1387 // Modify pattern to match canonical equivalences
1388 StringBuilder newPattern = new StringBuilder(patternLength);
1389 for(int i=0; i<patternLength; ) {
1390 int c = normalizedPattern.codePointAt(i);
1391 StringBuilder sequenceBuffer;
1392 if ((Character.getType(c) == Character.NON_SPACING_MARK)
1393 && (lastCodePoint != -1)) {
1394 sequenceBuffer = new StringBuilder();
1395 sequenceBuffer.appendCodePoint(lastCodePoint);
1396 sequenceBuffer.appendCodePoint(c);
1397 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1398 i += Character.charCount(c);
1399 if (i >= patternLength)
1400 break;
1401 c = normalizedPattern.codePointAt(i);
1402 sequenceBuffer.appendCodePoint(c);
1403 }
1404 String ea = produceEquivalentAlternation(
1405 sequenceBuffer.toString());
1406 newPattern.setLength(newPattern.length()-Character.charCount(lastCodePoint));
1407 newPattern.append("(?:").append(ea).append(")");
1408 } else if (c == '[' && lastCodePoint != '\\') {
1409 i = normalizeCharClass(newPattern, i);
1410 } else {
1411 newPattern.appendCodePoint(c);
1412 }
1413 lastCodePoint = c;
1414 i += Character.charCount(c);
1415 }
1416 normalizedPattern = newPattern.toString();
1417 }
1418
1419 /**
1420 * Complete the character class being parsed and add a set
1421 * of alternations to it that will match the canonical equivalences
1422 * of the characters within the class.
1423 */
1424 private int normalizeCharClass(StringBuilder newPattern, int i) {
1425 StringBuilder charClass = new StringBuilder();
1426 StringBuilder eq = null;
1427 int lastCodePoint = -1;
1428 String result;
1429
1430 i++;
1431 charClass.append("[");
1432 while(true) {
1433 int c = normalizedPattern.codePointAt(i);
1434 StringBuilder sequenceBuffer;
1435
1436 if (c == ']' && lastCodePoint != '\\') {
1437 charClass.append((char)c);
1438 break;
1439 } else if (Character.getType(c) == Character.NON_SPACING_MARK) {
1440 sequenceBuffer = new StringBuilder();
1441 sequenceBuffer.appendCodePoint(lastCodePoint);
1442 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1443 sequenceBuffer.appendCodePoint(c);
1444 i += Character.charCount(c);
1445 if (i >= normalizedPattern.length())
1446 break;
1447 c = normalizedPattern.codePointAt(i);
1448 }
1449 String ea = produceEquivalentAlternation(
1450 sequenceBuffer.toString());
1451
1452 charClass.setLength(charClass.length()-Character.charCount(lastCodePoint));
1453 if (eq == null)
1454 eq = new StringBuilder();
1455 eq.append('|');
1456 eq.append(ea);
1457 } else {
1458 charClass.appendCodePoint(c);
1459 i++;
1460 }
1461 if (i == normalizedPattern.length())
1462 throw error("Unclosed character class");
1463 lastCodePoint = c;
1464 }
1465
1466 if (eq != null) {
1467 result = "(?:"+charClass.toString()+eq.toString()+")";
1468 } else {
1469 result = charClass.toString();
1470 }
1471
1472 newPattern.append(result);
1473 return i;
1474 }
1475
1476 /**
1477 * Given a specific sequence composed of a regular character and
1478 * combining marks that follow it, produce the alternation that will
1479 * match all canonical equivalences of that sequence.
1480 */
1481 private String produceEquivalentAlternation(String source) {
1482 int len = countChars(source, 0, 1);
1483 if (source.length() == len)
1484 // source has one character.
1485 return source;
1486
1487 String base = source.substring(0,len);
1488 String combiningMarks = source.substring(len);
1489
1490 String[] perms = producePermutations(combiningMarks);
1491 StringBuilder result = new StringBuilder(source);
1492
1493 // Add combined permutations
1494 for(int x=0; x<perms.length; x++) {
1495 String next = base + perms[x];
1496 if (x>0)
1497 result.append("|"+next);
1498 next = composeOneStep(next);
1499 if (next != null)
1500 result.append("|"+produceEquivalentAlternation(next));
1501 }
1502 return result.toString();
1503 }
1504
1505 /**
1506 * Returns an array of strings that have all the possible
1507 * permutations of the characters in the input string.
1508 * This is used to get a list of all possible orderings
1509 * of a set of combining marks. Note that some of the permutations
1510 * are invalid because of combining class collisions, and these
1511 * possibilities must be removed because they are not canonically
1512 * equivalent.
1513 */
1514 private String[] producePermutations(String input) {
1515 if (input.length() == countChars(input, 0, 1))
1516 return new String[] {input};
1517
1518 if (input.length() == countChars(input, 0, 2)) {
1519 int c0 = Character.codePointAt(input, 0);
1520 int c1 = Character.codePointAt(input, Character.charCount(c0));
1521 if (getClass(c1) == getClass(c0)) {
1522 return new String[] {input};
1523 }
1524 String[] result = new String[2];
1525 result[0] = input;
1526 StringBuilder sb = new StringBuilder(2);
1527 sb.appendCodePoint(c1);
1528 sb.appendCodePoint(c0);
1529 result[1] = sb.toString();
1530 return result;
1531 }
1532
1533 int length = 1;
1534 int nCodePoints = countCodePoints(input);
1535 for(int x=1; x<nCodePoints; x++)
1536 length = length * (x+1);
1537
1538 String[] temp = new String[length];
1539
1540 int combClass[] = new int[nCodePoints];
1541 for(int x=0, i=0; x<nCodePoints; x++) {
1542 int c = Character.codePointAt(input, i);
1543 combClass[x] = getClass(c);
1544 i += Character.charCount(c);
1545 }
1546
1547 // For each char, take it out and add the permutations
1548 // of the remaining chars
1549 int index = 0;
1550 int len;
1551 // offset maintains the index in code units.
1552 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) {
1553 len = countChars(input, offset, 1);
1554 boolean skip = false;
1555 for(int y=x-1; y>=0; y--) {
1556 if (combClass[y] == combClass[x]) {
1557 continue loop;
1558 }
1559 }
1560 StringBuilder sb = new StringBuilder(input);
1561 String otherChars = sb.delete(offset, offset+len).toString();
1562 String[] subResult = producePermutations(otherChars);
1563
1564 String prefix = input.substring(offset, offset+len);
1565 for (String sre : subResult)
1566 temp[index++] = prefix + sre;
1567 }
1568 String[] result = new String[index];
1569 for (int x=0; x<index; x++)
1570 result[x] = temp[x];
1571 return result;
1572 }
1573
1574 private int getClass(int c) {
1575 return sun.text.Normalizer.getCombiningClass(c);
1576 }
1577
1578 /**
1579 * Attempts to compose input by combining the first character
1580 * with the first combining mark following it. Returns a String
1581 * that is the composition of the leading character with its first
1582 * combining mark followed by the remaining combining marks. Returns
1583 * null if the first two characters cannot be further composed.
1584 */
1585 private String composeOneStep(String input) {
1586 int len = countChars(input, 0, 2);
1587 String firstTwoCharacters = input.substring(0, len);
1588 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC);
1589
1590 if (result.equals(firstTwoCharacters))
1591 return null;
1592 else {
1593 String remainder = input.substring(len);
1594 return result + remainder;
1595 }
1596 }
1597
1598 /**
1599 * Preprocess any \Q...\E sequences in `temp', meta-quoting them.
1600 * See the description of `quotemeta' in perlfunc(1).
1601 */
1602 private void RemoveQEQuoting() {
1603 final int pLen = patternLength;
1604 int i = 0;
1605 while (i < pLen-1) {
1606 if (temp[i] != '\\')
1607 i += 1;
1608 else if (temp[i + 1] != 'Q')
1609 i += 2;
1610 else
1611 break;
1612 }
1613 if (i >= pLen - 1) // No \Q sequence found
1614 return;
1615 int j = i;
1616 i += 2;
1617 int[] newtemp = new int[j + 3*(pLen-i) + 2];
1618 System.arraycopy(temp, 0, newtemp, 0, j);
1619
1620 boolean inQuote = true;
1621 boolean beginQuote = true;
1622 while (i < pLen) {
1623 int c = temp[i++];
1624 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) {
1625 newtemp[j++] = c;
1626 } else if (ASCII.isDigit(c)) {
1627 if (beginQuote) {
1628 /*
1629 * A unicode escape \[0xu] could be before this quote,
1630 * and we don't want this numeric char to processed as
1631 * part of the escape.
1632 */
1633 newtemp[j++] = '\\';
1634 newtemp[j++] = 'x';
1635 newtemp[j++] = '3';
1636 }
1637 newtemp[j++] = c;
1638 } else if (c != '\\') {
1639 if (inQuote) newtemp[j++] = '\\';
1640 newtemp[j++] = c;
1641 } else if (inQuote) {
1642 if (temp[i] == 'E') {
1643 i++;
1644 inQuote = false;
1645 } else {
1646 newtemp[j++] = '\\';
1647 newtemp[j++] = '\\';
1648 }
1649 } else {
1650 if (temp[i] == 'Q') {
1651 i++;
1652 inQuote = true;
1653 beginQuote = true;
1654 continue;
1655 } else {
1656 newtemp[j++] = c;
1657 if (i != pLen)
1658 newtemp[j++] = temp[i++];
1659 }
1660 }
1661
1662 beginQuote = false;
1663 }
1664
1665 patternLength = j;
1666 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination
1667 }
1668
1669 /**
1670 * Copies regular expression to an int array and invokes the parsing
1671 * of the expression which will create the object tree.
1672 */
1673 private void compile() {
1674 // Handle canonical equivalences
1675 if (has(CANON_EQ) && !has(LITERAL)) {
1676 normalize();
1677 } else {
1678 normalizedPattern = pattern;
1679 }
1680 patternLength = normalizedPattern.length();
1681
1682 // Copy pattern to int array for convenience
1683 // Use double zero to terminate pattern
1684 temp = new int[patternLength + 2];
1685
1686 hasSupplementary = false;
1687 int c, count = 0;
1688 // Convert all chars into code points
1689 for (int x = 0; x < patternLength; x += Character.charCount(c)) {
1690 c = normalizedPattern.codePointAt(x);
1691 if (isSupplementary(c)) {
1692 hasSupplementary = true;
1693 }
1694 temp[count++] = c;
1695 }
1696
1697 patternLength = count; // patternLength now in code points
1698
1699 if (! has(LITERAL))
1700 RemoveQEQuoting();
1701
1702 // Allocate all temporary objects here.
1703 buffer = new int[32];
1704 groupNodes = new GroupHead[10];
1705 namedGroups = null;
1706
1707 if (has(LITERAL)) {
1708 // Literal pattern handling
1709 matchRoot = newSlice(temp, patternLength, hasSupplementary);
1710 matchRoot.next = lastAccept;
1711 } else {
1712 // Start recursive descent parsing
1713 matchRoot = expr(lastAccept);
1714 // Check extra pattern characters
1715 if (patternLength != cursor) {
1716 if (peek() == ')') {
1717 throw error("Unmatched closing ')'");
1718 } else {
1719 throw error("Unexpected internal error");
1720 }
1721 }
1722 }
1723
1724 // Peephole optimization
1725 if (matchRoot instanceof Slice) {
1726 root = BnM.optimize(matchRoot);
1727 if (root == matchRoot) {
1728 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1729 }
1730 } else if (matchRoot instanceof Begin || matchRoot instanceof First) {
1731 root = matchRoot;
1732 } else {
1733 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1734 }
1735
1736 // Release temporary storage
1737 temp = null;
1738 buffer = null;
1739 groupNodes = null;
1740 patternLength = 0;
1741 compiled = true;
1742 }
1743
1744 Map<String, Integer> namedGroups() {
1745 if (namedGroups == null)
1746 namedGroups = new HashMap<>(2);
1747 return namedGroups;
1748 }
1749
1750 /**
1751 * Used to print out a subtree of the Pattern to help with debugging.
1752 */
1753 private static void printObjectTree(Node node) {
1754 while(node != null) {
1755 if (node instanceof Prolog) {
1756 System.out.println(node);
1757 printObjectTree(((Prolog)node).loop);
1758 System.out.println("**** end contents prolog loop");
1759 } else if (node instanceof Loop) {
1760 System.out.println(node);
1761 printObjectTree(((Loop)node).body);
1762 System.out.println("**** end contents Loop body");
1763 } else if (node instanceof Curly) {
1764 System.out.println(node);
1765 printObjectTree(((Curly)node).atom);
1766 System.out.println("**** end contents Curly body");
1767 } else if (node instanceof GroupCurly) {
1768 System.out.println(node);
1769 printObjectTree(((GroupCurly)node).atom);
1770 System.out.println("**** end contents GroupCurly body");
1771 } else if (node instanceof GroupTail) {
1772 System.out.println(node);
1773 System.out.println("Tail next is "+node.next);
1774 return;
1775 } else {
1776 System.out.println(node);
1777 }
1778 node = node.next;
1779 if (node != null)
1780 System.out.println("->next:");
1781 if (node == Pattern.accept) {
1782 System.out.println("Accept Node");
1783 node = null;
1784 }
1785 }
1786 }
1787
1788 /**
1789 * Used to accumulate information about a subtree of the object graph
1790 * so that optimizations can be applied to the subtree.
1791 */
1792 static final class TreeInfo {
1793 int minLength;
1794 int maxLength;
1795 boolean maxValid;
1796 boolean deterministic;
1797
1798 TreeInfo() {
1799 reset();
1800 }
1801 void reset() {
1802 minLength = 0;
1803 maxLength = 0;
1804 maxValid = true;
1805 deterministic = true;
1806 }
1807 }
1808
1809 /*
1810 * The following private methods are mainly used to improve the
1811 * readability of the code. In order to let the Java compiler easily
1812 * inline them, we should not put many assertions or error checks in them.
1813 */
1814
1815 /**
1816 * Indicates whether a particular flag is set or not.
1817 */
1818 private boolean has(int f) {
1819 return (flags & f) != 0;
1820 }
1821
1822 /**
1823 * Match next character, signal error if failed.
1824 */
1825 private void accept(int ch, String s) {
1826 int testChar = temp[cursor++];
1827 if (has(COMMENTS))
1828 testChar = parsePastWhitespace(testChar);
1829 if (ch != testChar) {
1830 throw error(s);
1831 }
1832 }
1833
1834 /**
1835 * Mark the end of pattern with a specific character.
1836 */
1837 private void mark(int c) {
1838 temp[patternLength] = c;
1839 }
1840
1841 /**
1842 * Peek the next character, and do not advance the cursor.
1843 */
1844 private int peek() {
1845 int ch = temp[cursor];
1846 if (has(COMMENTS))
1847 ch = peekPastWhitespace(ch);
1848 return ch;
1849 }
1850
1851 /**
1852 * Read the next character, and advance the cursor by one.
1853 */
1854 private int read() {
1855 int ch = temp[cursor++];
1856 if (has(COMMENTS))
1857 ch = parsePastWhitespace(ch);
1858 return ch;
1859 }
1860
1861 /**
1862 * Read the next character, and advance the cursor by one,
1863 * ignoring the COMMENTS setting
1864 */
1865 private int readEscaped() {
1866 int ch = temp[cursor++];
1867 return ch;
1868 }
1869
1870 /**
1871 * Advance the cursor by one, and peek the next character.
1872 */
1873 private int next() {
1874 int ch = temp[++cursor];
1875 if (has(COMMENTS))
1876 ch = peekPastWhitespace(ch);
1877 return ch;
1878 }
1879
1880 /**
1881 * Advance the cursor by one, and peek the next character,
1882 * ignoring the COMMENTS setting
1883 */
1884 private int nextEscaped() {
1885 int ch = temp[++cursor];
1886 return ch;
1887 }
1888
1889 /**
1890 * If in xmode peek past whitespace and comments.
1891 */
1892 private int peekPastWhitespace(int ch) {
1893 while (ASCII.isSpace(ch) || ch == '#') {
1894 while (ASCII.isSpace(ch))
1895 ch = temp[++cursor];
1896 if (ch == '#') {
1897 ch = peekPastLine();
1898 }
1899 }
1900 return ch;
1901 }
1902
1903 /**
1904 * If in xmode parse past whitespace and comments.
1905 */
1906 private int parsePastWhitespace(int ch) {
1907 while (ASCII.isSpace(ch) || ch == '#') {
1908 while (ASCII.isSpace(ch))
1909 ch = temp[cursor++];
1910 if (ch == '#')
1911 ch = parsePastLine();
1912 }
1913 return ch;
1914 }
1915
1916 /**
1917 * xmode parse past comment to end of line.
1918 */
1919 private int parsePastLine() {
1920 int ch = temp[cursor++];
1921 while (ch != 0 && !isLineSeparator(ch))
1922 ch = temp[cursor++];
1923 return ch;
1924 }
1925
1926 /**
1927 * xmode peek past comment to end of line.
1928 */
1929 private int peekPastLine() {
1930 int ch = temp[++cursor];
1931 while (ch != 0 && !isLineSeparator(ch))
1932 ch = temp[++cursor];
1933 return ch;
1934 }
1935
1936 /**
1937 * Determines if character is a line separator in the current mode
1938 */
1939 private boolean isLineSeparator(int ch) {
1940 if (has(UNIX_LINES)) {
1941 return ch == '\n';
1942 } else {
1943 return (ch == '\n' ||
1944 ch == '\r' ||
1945 (ch|1) == '\u2029' ||
1946 ch == '\u0085');
1947 }
1948 }
1949
1950 /**
1951 * Read the character after the next one, and advance the cursor by two.
1952 */
1953 private int skip() {
1954 int i = cursor;
1955 int ch = temp[i+1];
1956 cursor = i + 2;
1957 return ch;
1958 }
1959
1960 /**
1961 * Unread one next character, and retreat cursor by one.
1962 */
1963 private void unread() {
1964 cursor--;
1965 }
1966
1967 /**
1968 * Internal method used for handling all syntax errors. The pattern is
1969 * displayed with a pointer to aid in locating the syntax error.
1970 */
1971 private PatternSyntaxException error(String s) {
1972 return new PatternSyntaxException(s, normalizedPattern, cursor - 1);
1973 }
1974
1975 /**
1976 * Determines if there is any supplementary character or unpaired
1977 * surrogate in the specified range.
1978 */
1979 private boolean findSupplementary(int start, int end) {
1980 for (int i = start; i < end; i++) {
1981 if (isSupplementary(temp[i]))
1982 return true;
1983 }
1984 return false;
1985 }
1986
1987 /**
1988 * Determines if the specified code point is a supplementary
1989 * character or unpaired surrogate.
1990 */
1991 private static final boolean isSupplementary(int ch) {
1992 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT ||
1993 Character.isSurrogate((char)ch);
1994 }
1995
1996 /**
1997 * The following methods handle the main parsing. They are sorted
1998 * according to their precedence order, the lowest one first.
1999 */
2000
2001 /**
2002 * The expression is parsed with branch nodes added for alternations.
2003 * This may be called recursively to parse sub expressions that may
2004 * contain alternations.
2005 */
2006 private Node expr(Node end) {
2007 Node prev = null;
2008 Node firstTail = null;
2009 Branch branch = null;
2010 Node branchConn = null;
2011
2012 for (;;) {
2013 Node node = sequence(end);
2014 Node nodeTail = root; //double return
2015 if (prev == null) {
2016 prev = node;
2017 firstTail = nodeTail;
2018 } else {
2019 // Branch
2020 if (branchConn == null) {
2021 branchConn = new BranchConn();
2022 branchConn.next = end;
2023 }
2024 if (node == end) {
2025 // if the node returned from sequence() is "end"
2026 // we have an empty expr, set a null atom into
2027 // the branch to indicate to go "next" directly.
2028 node = null;
2029 } else {
2030 // the "tail.next" of each atom goes to branchConn
2031 nodeTail.next = branchConn;
2032 }
2033 if (prev == branch) {
2034 branch.add(node);
2035 } else {
2036 if (prev == end) {
2037 prev = null;
2038 } else {
2039 // replace the "end" with "branchConn" at its tail.next
2040 // when put the "prev" into the branch as the first atom.
2041 firstTail.next = branchConn;
2042 }
2043 prev = branch = new Branch(prev, node, branchConn);
2044 }
2045 }
2046 if (peek() != '|') {
2047 return prev;
2048 }
2049 next();
2050 }
2051 }
2052
2053 @SuppressWarnings("fallthrough")
2054 /**
2055 * Parsing of sequences between alternations.
2056 */
2057 private Node sequence(Node end) {
2058 Node head = null;
2059 Node tail = null;
2060 Node node = null;
2061 LOOP:
2062 for (;;) {
2063 int ch = peek();
2064 switch (ch) {
2065 case '(':
2066 // Because group handles its own closure,
2067 // we need to treat it differently
2068 node = group0();
2069 // Check for comment or flag group
2070 if (node == null)
2071 continue;
2072 if (head == null)
2073 head = node;
2074 else
2075 tail.next = node;
2076 // Double return: Tail was returned in root
2077 tail = root;
2078 continue;
2079 case '[':
2080 node = clazz(true);
2081 break;
2082 case '\\':
2083 ch = nextEscaped();
2084 if (ch == 'p' || ch == 'P') {
2085 boolean oneLetter = true;
2086 boolean comp = (ch == 'P');
2087 ch = next(); // Consume { if present
2088 if (ch != '{') {
2089 unread();
2090 } else {
2091 oneLetter = false;
2092 }
2093 node = family(oneLetter, comp);
2094 } else {
2095 unread();
2096 node = atom();
2097 }
2098 break;
2099 case '^':
2100 next();
2101 if (has(MULTILINE)) {
2102 if (has(UNIX_LINES))
2103 node = new UnixCaret();
2104 else
2105 node = new Caret();
2106 } else {
2107 node = new Begin();
2108 }
2109 break;
2110 case '$':
2111 next();
2112 if (has(UNIX_LINES))
2113 node = new UnixDollar(has(MULTILINE));
2114 else
2115 node = new Dollar(has(MULTILINE));
2116 break;
2117 case '.':
2118 next();
2119 if (has(DOTALL)) {
2120 node = new All();
2121 } else {
2122 if (has(UNIX_LINES))
2123 node = new UnixDot();
2124 else {
2125 node = new Dot();
2126 }
2127 }
2128 break;
2129 case '|':
2130 case ')':
2131 break LOOP;
2132 case ']': // Now interpreting dangling ] and } as literals
2133 case '}':
2134 node = atom();
2135 break;
2136 case '?':
2137 case '*':
2138 case '+':
2139 next();
2140 throw error("Dangling meta character '" + ((char)ch) + "'");
2141 case 0:
2142 if (cursor >= patternLength) {
2143 break LOOP;
2144 }
2145 // Fall through
2146 default:
2147 node = atom();
2148 break;
2149 }
2150
2151 node = closure(node);
2152
2153 if (head == null) {
2154 head = tail = node;
2155 } else {
2156 tail.next = node;
2157 tail = node;
2158 }
2159 }
2160 if (head == null) {
2161 return end;
2162 }
2163 tail.next = end;
2164 root = tail; //double return
2165 return head;
2166 }
2167
2168 @SuppressWarnings("fallthrough")
2169 /**
2170 * Parse and add a new Single or Slice.
2171 */
2172 private Node atom() {
2173 int first = 0;
2174 int prev = -1;
2175 boolean hasSupplementary = false;
2176 int ch = peek();
2177 for (;;) {
2178 switch (ch) {
2179 case '*':
2180 case '+':
2181 case '?':
2182 case '{':
2183 if (first > 1) {
2184 cursor = prev; // Unwind one character
2185 first--;
2186 }
2187 break;
2188 case '$':
2189 case '.':
2190 case '^':
2191 case '(':
2192 case '[':
2193 case '|':
2194 case ')':
2195 break;
2196 case '\\':
2197 ch = nextEscaped();
2198 if (ch == 'p' || ch == 'P') { // Property
2199 if (first > 0) { // Slice is waiting; handle it first
2200 unread();
2201 break;
2202 } else { // No slice; just return the family node
2203 boolean comp = (ch == 'P');
2204 boolean oneLetter = true;
2205 ch = next(); // Consume { if present
2206 if (ch != '{')
2207 unread();
2208 else
2209 oneLetter = false;
2210 return family(oneLetter, comp);
2211 }
2212 }
2213 unread();
2214 prev = cursor;
2215 ch = escape(false, first == 0, false);
2216 if (ch >= 0) {
2217 append(ch, first);
2218 first++;
2219 if (isSupplementary(ch)) {
2220 hasSupplementary = true;
2221 }
2222 ch = peek();
2223 continue;
2224 } else if (first == 0) {
2225 return root;
2226 }
2227 // Unwind meta escape sequence
2228 cursor = prev;
2229 break;
2230 case 0:
2231 if (cursor >= patternLength) {
2232 break;
2233 }
2234 // Fall through
2235 default:
2236 prev = cursor;
2237 append(ch, first);
2238 first++;
2239 if (isSupplementary(ch)) {
2240 hasSupplementary = true;
2241 }
2242 ch = next();
2243 continue;
2244 }
2245 break;
2246 }
2247 if (first == 1) {
2248 return newSingle(buffer[0]);
2249 } else {
2250 return newSlice(buffer, first, hasSupplementary);
2251 }
2252 }
2253
2254 private void append(int ch, int len) {
2255 if (len >= buffer.length) {
2256 int[] tmp = new int[len+len];
2257 System.arraycopy(buffer, 0, tmp, 0, len);
2258 buffer = tmp;
2259 }
2260 buffer[len] = ch;
2261 }
2262
2263 /**
2264 * Parses a backref greedily, taking as many numbers as it
2265 * can. The first digit is always treated as a backref, but
2266 * multi digit numbers are only treated as a backref if at
2267 * least that many backrefs exist at this point in the regex.
2268 */
2269 private Node ref(int refNum) {
2270 boolean done = false;
2271 while(!done) {
2272 int ch = peek();
2273 switch(ch) {
2274 case '0':
2275 case '1':
2276 case '2':
2277 case '3':
2278 case '4':
2279 case '5':
2280 case '6':
2281 case '7':
2282 case '8':
2283 case '9':
2284 int newRefNum = (refNum * 10) + (ch - '0');
2285 // Add another number if it doesn't make a group
2286 // that doesn't exist
2287 if (capturingGroupCount - 1 < newRefNum) {
2288 done = true;
2289 break;
2290 }
2291 refNum = newRefNum;
2292 read();
2293 break;
2294 default:
2295 done = true;
2296 break;
2297 }
2298 }
2299 if (has(CASE_INSENSITIVE))
2300 return new CIBackRef(refNum, has(UNICODE_CASE));
2301 else
2302 return new BackRef(refNum);
2303 }
2304
2305 /**
2306 * Parses an escape sequence to determine the actual value that needs
2307 * to be matched.
2308 * If -1 is returned and create was true a new object was added to the tree
2309 * to handle the escape sequence.
2310 * If the returned value is greater than zero, it is the value that
2311 * matches the escape sequence.
2312 */
2313 private int escape(boolean inclass, boolean create, boolean isrange) {
2314 int ch = skip();
2315 switch (ch) {
2316 case '0':
2317 return o();
2318 case '1':
2319 case '2':
2320 case '3':
2321 case '4':
2322 case '5':
2323 case '6':
2324 case '7':
2325 case '8':
2326 case '9':
2327 if (inclass) break;
2328 if (create) {
2329 root = ref((ch - '0'));
2330 }
2331 return -1;
2332 case 'A':
2333 if (inclass) break;
2334 if (create) root = new Begin();
2335 return -1;
2336 case 'B':
2337 if (inclass) break;
2338 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS));
2339 return -1;
2340 case 'C':
2341 break;
2342 case 'D':
2343 if (create) root = has(UNICODE_CHARACTER_CLASS)
2344 ? new Utype(UnicodeProp.DIGIT).complement()
2345 : new Ctype(ASCII.DIGIT).complement();
2346 return -1;
2347 case 'E':
2348 case 'F':
2349 break;
2350 case 'G':
2351 if (inclass) break;
2352 if (create) root = new LastMatch();
2353 return -1;
2354 case 'H':
2355 if (create) root = new HorizWS().complement();
2356 return -1;
2357 case 'I':
2358 case 'J':
2359 case 'K':
2360 case 'L':
2361 case 'M':
2362 case 'N':
2363 case 'O':
2364 case 'P':
2365 case 'Q':
2366 break;
2367 case 'R':
2368 if (inclass) break;
2369 if (create) root = new LineEnding();
2370 return -1;
2371 case 'S':
2372 if (create) root = has(UNICODE_CHARACTER_CLASS)
2373 ? new Utype(UnicodeProp.WHITE_SPACE).complement()
2374 : new Ctype(ASCII.SPACE).complement();
2375 return -1;
2376 case 'T':
2377 case 'U':
2378 break;
2379 case 'V':
2380 if (create) root = new VertWS().complement();
2381 return -1;
2382 case 'W':
2383 if (create) root = has(UNICODE_CHARACTER_CLASS)
2384 ? new Utype(UnicodeProp.WORD).complement()
2385 : new Ctype(ASCII.WORD).complement();
2386 return -1;
2387 case 'X':
2388 case 'Y':
2389 break;
2390 case 'Z':
2391 if (inclass) break;
2392 if (create) {
2393 if (has(UNIX_LINES))
2394 root = new UnixDollar(false);
2395 else
2396 root = new Dollar(false);
2397 }
2398 return -1;
2399 case 'a':
2400 return '\007';
2401 case 'b':
2402 if (inclass) break;
2403 if (create) root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS));
2404 return -1;
2405 case 'c':
2406 return c();
2407 case 'd':
2408 if (create) root = has(UNICODE_CHARACTER_CLASS)
2409 ? new Utype(UnicodeProp.DIGIT)
2410 : new Ctype(ASCII.DIGIT);
2411 return -1;
2412 case 'e':
2413 return '\033';
2414 case 'f':
2415 return '\f';
2416 case 'g':
2417 break;
2418 case 'h':
2419 if (create) root = new HorizWS();
2420 return -1;
2421 case 'i':
2422 case 'j':
2423 break;
2424 case 'k':
2425 if (inclass)
2426 break;
2427 if (read() != '<')
2428 throw error("\\k is not followed by '<' for named capturing group");
2429 String name = groupname(read());
2430 if (!namedGroups().containsKey(name))
2431 throw error("(named capturing group <"+ name+"> does not exit");
2432 if (create) {
2433 if (has(CASE_INSENSITIVE))
2434 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE));
2435 else
2436 root = new BackRef(namedGroups().get(name));
2437 }
2438 return -1;
2439 case 'l':
2440 case 'm':
2441 break;
2442 case 'n':
2443 return '\n';
2444 case 'o':
2445 case 'p':
2446 case 'q':
2447 break;
2448 case 'r':
2449 return '\r';
2450 case 's':
2451 if (create) root = has(UNICODE_CHARACTER_CLASS)
2452 ? new Utype(UnicodeProp.WHITE_SPACE)
2453 : new Ctype(ASCII.SPACE);
2454 return -1;
2455 case 't':
2456 return '\t';
2457 case 'u':
2458 return u();
2459 case 'v':
2460 // '\v' was implemented as VT/0x0B in releases < 1.8 (though
2461 // undocumented). In JDK8 '\v' is specified as a predefined
2462 // character class for all vertical whitespace characters.
2463 // So [-1, root=VertWS node] pair is returned (instead of a
2464 // single 0x0B). This breaks the range if '\v' is used as
2465 // the start or end value, such as [\v-...] or [...-\v], in
2466 // which a single definite value (0x0B) is expected. For
2467 // compatibility concern '\013'/0x0B is returned if isrange.
2468 if (isrange)
2469 return '\013';
2470 if (create) root = new VertWS();
2471 return -1;
2472 case 'w':
2473 if (create) root = has(UNICODE_CHARACTER_CLASS)
2474 ? new Utype(UnicodeProp.WORD)
2475 : new Ctype(ASCII.WORD);
2476 return -1;
2477 case 'x':
2478 return x();
2479 case 'y':
2480 break;
2481 case 'z':
2482 if (inclass) break;
2483 if (create) root = new End();
2484 return -1;
2485 default:
2486 return ch;
2487 }
2488 throw error("Illegal/unsupported escape sequence");
2489 }
2490
2491 /**
2492 * Parse a character class, and return the node that matches it.
2493 *
2494 * Consumes a ] on the way out if consume is true. Usually consume
2495 * is true except for the case of [abc&&def] where def is a separate
2496 * right hand node with "understood" brackets.
2497 */
2498 private CharProperty clazz(boolean consume) {
2499 CharProperty prev = null;
2500 CharProperty node = null;
2501 BitClass bits = new BitClass();
2502 boolean include = true;
2503 boolean firstInClass = true;
2504 int ch = next();
2505 for (;;) {
2506 switch (ch) {
2507 case '^':
2508 // Negates if first char in a class, otherwise literal
2509 if (firstInClass) {
2510 if (temp[cursor-1] != '[')
2511 break;
2512 ch = next();
2513 include = !include;
2514 continue;
2515 } else {
2516 // ^ not first in class, treat as literal
2517 break;
2518 }
2519 case '[':
2520 firstInClass = false;
2521 node = clazz(true);
2522 if (prev == null)
2523 prev = node;
2524 else
2525 prev = union(prev, node);
2526 ch = peek();
2527 continue;
2528 case '&':
2529 firstInClass = false;
2530 ch = next();
2531 if (ch == '&') {
2532 ch = next();
2533 CharProperty rightNode = null;
2534 while (ch != ']' && ch != '&') {
2535 if (ch == '[') {
2536 if (rightNode == null)
2537 rightNode = clazz(true);
2538 else
2539 rightNode = union(rightNode, clazz(true));
2540 } else { // abc&&def
2541 unread();
2542 rightNode = clazz(false);
2543 }
2544 ch = peek();
2545 }
2546 if (rightNode != null)
2547 node = rightNode;
2548 if (prev == null) {
2549 if (rightNode == null)
2550 throw error("Bad class syntax");
2551 else
2552 prev = rightNode;
2553 } else {
2554 prev = intersection(prev, node);
2555 }
2556 } else {
2557 // treat as a literal &
2558 unread();
2559 break;
2560 }
2561 continue;
2562 case 0:
2563 firstInClass = false;
2564 if (cursor >= patternLength)
2565 throw error("Unclosed character class");
2566 break;
2567 case ']':
2568 firstInClass = false;
2569 if (prev != null) {
2570 if (consume)
2571 next();
2572 return prev;
2573 }
2574 break;
2575 default:
2576 firstInClass = false;
2577 break;
2578 }
2579 node = range(bits);
2580 if (include) {
2581 if (prev == null) {
2582 prev = node;
2583 } else {
2584 if (prev != node)
2585 prev = union(prev, node);
2586 }
2587 } else {
2588 if (prev == null) {
2589 prev = node.complement();
2590 } else {
2591 if (prev != node)
2592 prev = setDifference(prev, node);
2593 }
2594 }
2595 ch = peek();
2596 }
2597 }
2598
2599 private CharProperty bitsOrSingle(BitClass bits, int ch) {
2600 /* Bits can only handle codepoints in [u+0000-u+00ff] range.
2601 Use "single" node instead of bits when dealing with unicode
2602 case folding for codepoints listed below.
2603 (1)Uppercase out of range: u+00ff, u+00b5
2604 toUpperCase(u+00ff) -> u+0178
2605 toUpperCase(u+00b5) -> u+039c
2606 (2)LatinSmallLetterLongS u+17f
2607 toUpperCase(u+017f) -> u+0053
2608 (3)LatinSmallLetterDotlessI u+131
2609 toUpperCase(u+0131) -> u+0049
2610 (4)LatinCapitalLetterIWithDotAbove u+0130
2611 toLowerCase(u+0130) -> u+0069
2612 (5)KelvinSign u+212a
2613 toLowerCase(u+212a) ==> u+006B
2614 (6)AngstromSign u+212b
2615 toLowerCase(u+212b) ==> u+00e5
2616 */
2617 int d;
2618 if (ch < 256 &&
2619 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) &&
2620 (ch == 0xff || ch == 0xb5 ||
2621 ch == 0x49 || ch == 0x69 || //I and i
2622 ch == 0x53 || ch == 0x73 || //S and s
2623 ch == 0x4b || ch == 0x6b || //K and k
2624 ch == 0xc5 || ch == 0xe5))) //A+ring
2625 return bits.add(ch, flags());
2626 return newSingle(ch);
2627 }
2628
2629 /**
2630 * Parse a single character or a character range in a character class
2631 * and return its representative node.
2632 */
2633 private CharProperty range(BitClass bits) {
2634 int ch = peek();
2635 if (ch == '\\') {
2636 ch = nextEscaped();
2637 if (ch == 'p' || ch == 'P') { // A property
2638 boolean comp = (ch == 'P');
2639 boolean oneLetter = true;
2640 // Consume { if present
2641 ch = next();
2642 if (ch != '{')
2643 unread();
2644 else
2645 oneLetter = false;
2646 return family(oneLetter, comp);
2647 } else { // ordinary escape
2648 boolean isrange = temp[cursor+1] == '-';
2649 unread();
2650 ch = escape(true, true, isrange);
2651 if (ch == -1)
2652 return (CharProperty) root;
2653 }
2654 } else {
2655 next();
2656 }
2657 if (ch >= 0) {
2658 if (peek() == '-') {
2659 int endRange = temp[cursor+1];
2660 if (endRange == '[') {
2661 return bitsOrSingle(bits, ch);
2662 }
2663 if (endRange != ']') {
2664 next();
2665 int m = peek();
2666 if (m == '\\') {
2667 m = escape(true, false, true);
2668 } else {
2669 next();
2670 }
2671 if (m < ch) {
2672 throw error("Illegal character range");
2673 }
2674 if (has(CASE_INSENSITIVE))
2675 return caseInsensitiveRangeFor(ch, m);
2676 else
2677 return rangeFor(ch, m);
2678 }
2679 }
2680 return bitsOrSingle(bits, ch);
2681 }
2682 throw error("Unexpected character '"+((char)ch)+"'");
2683 }
2684
2685 /**
2686 * Parses a Unicode character family and returns its representative node.
2687 */
2688 private CharProperty family(boolean singleLetter,
2689 boolean maybeComplement)
2690 {
2691 next();
2692 String name;
2693 CharProperty node = null;
2694
2695 if (singleLetter) {
2696 int c = temp[cursor];
2697 if (!Character.isSupplementaryCodePoint(c)) {
2698 name = String.valueOf((char)c);
2699 } else {
2700 name = new String(temp, cursor, 1);
2701 }
2702 read();
2703 } else {
2704 int i = cursor;
2705 mark('}');
2706 while(read() != '}') {
2707 }
2708 mark('\000');
2709 int j = cursor;
2710 if (j > patternLength)
2711 throw error("Unclosed character family");
2712 if (i + 1 >= j)
2713 throw error("Empty character family");
2714 name = new String(temp, i, j-i-1);
2715 }
2716
2717 int i = name.indexOf('=');
2718 if (i != -1) {
2719 // property construct \p{name=value}
2720 String value = name.substring(i + 1);
2721 name = name.substring(0, i).toLowerCase(Locale.ENGLISH);
2722 switch (name) {
2723 case "sc":
2724 case "script":
2725 node = unicodeScriptPropertyFor(value);
2726 break;
2727 case "blk":
2728 case "block":
2729 node = unicodeBlockPropertyFor(value);
2730 break;
2731 case "gc":
2732 case "general_category":
2733 node = charPropertyNodeFor(value);
2734 break;
2735 default:
2736 throw error("Unknown Unicode property {name=<" + name + ">, "
2737 + "value=<" + value + ">}");
2738 }
2739 } else {
2740 if (name.startsWith("In")) {
2741 // \p{inBlockName}
2742 node = unicodeBlockPropertyFor(name.substring(2));
2743 } else if (name.startsWith("Is")) {
2744 // \p{isGeneralCategory} and \p{isScriptName}
2745 name = name.substring(2);
2746 UnicodeProp uprop = UnicodeProp.forName(name);
2747 if (uprop != null)
2748 node = new Utype(uprop);
2749 if (node == null)
2750 node = CharPropertyNames.charPropertyFor(name);
2751 if (node == null)
2752 node = unicodeScriptPropertyFor(name);
2753 } else {
2754 if (has(UNICODE_CHARACTER_CLASS)) {
2755 UnicodeProp uprop = UnicodeProp.forPOSIXName(name);
2756 if (uprop != null)
2757 node = new Utype(uprop);
2758 }
2759 if (node == null)
2760 node = charPropertyNodeFor(name);
2761 }
2762 }
2763 if (maybeComplement) {
2764 if (node instanceof Category || node instanceof Block)
2765 hasSupplementary = true;
2766 node = node.complement();
2767 }
2768 return node;
2769 }
2770
2771
2772 /**
2773 * Returns a CharProperty matching all characters belong to
2774 * a UnicodeScript.
2775 */
2776 private CharProperty unicodeScriptPropertyFor(String name) {
2777 final Character.UnicodeScript script;
2778 try {
2779 script = Character.UnicodeScript.forName(name);
2780 } catch (IllegalArgumentException iae) {
2781 throw error("Unknown character script name {" + name + "}");
2782 }
2783 return new Script(script);
2784 }
2785
2786 /**
2787 * Returns a CharProperty matching all characters in a UnicodeBlock.
2788 */
2789 private CharProperty unicodeBlockPropertyFor(String name) {
2790 final Character.UnicodeBlock block;
2791 try {
2792 block = Character.UnicodeBlock.forName(name);
2793 } catch (IllegalArgumentException iae) {
2794 throw error("Unknown character block name {" + name + "}");
2795 }
2796 return new Block(block);
2797 }
2798
2799 /**
2800 * Returns a CharProperty matching all characters in a named property.
2801 */
2802 private CharProperty charPropertyNodeFor(String name) {
2803 CharProperty p = CharPropertyNames.charPropertyFor(name);
2804 if (p == null)
2805 throw error("Unknown character property name {" + name + "}");
2806 return p;
2807 }
2808
2809 /**
2810 * Parses and returns the name of a "named capturing group", the trailing
2811 * ">" is consumed after parsing.
2812 */
2813 private String groupname(int ch) {
2814 StringBuilder sb = new StringBuilder();
2815 sb.append(Character.toChars(ch));
2816 while (ASCII.isLower(ch=read()) || ASCII.isUpper(ch) ||
2817 ASCII.isDigit(ch)) {
2818 sb.append(Character.toChars(ch));
2819 }
2820 if (sb.length() == 0)
2821 throw error("named capturing group has 0 length name");
2822 if (ch != '>')
2823 throw error("named capturing group is missing trailing '>'");
2824 return sb.toString();
2825 }
2826
2827 /**
2828 * Parses a group and returns the head node of a set of nodes that process
2829 * the group. Sometimes a double return system is used where the tail is
2830 * returned in root.
2831 */
2832 private Node group0() {
2833 boolean capturingGroup = false;
2834 Node head = null;
2835 Node tail = null;
2836 int save = flags;
2837 root = null;
2838 int ch = next();
2839 if (ch == '?') {
2840 ch = skip();
2841 switch (ch) {
2842 case ':': // (?:xxx) pure group
2843 head = createGroup(true);
2844 tail = root;
2845 head.next = expr(tail);
2846 break;
2847 case '=': // (?=xxx) and (?!xxx) lookahead
2848 case '!':
2849 head = createGroup(true);
2850 tail = root;
2851 head.next = expr(tail);
2852 if (ch == '=') {
2853 head = tail = new Pos(head);
2854 } else {
2855 head = tail = new Neg(head);
2856 }
2857 break;
2858 case '>': // (?>xxx) independent group
2859 head = createGroup(true);
2860 tail = root;
2861 head.next = expr(tail);
2862 head = tail = new Ques(head, INDEPENDENT);
2863 break;
2864 case '<': // (?<xxx) look behind
2865 ch = read();
2866 if (ASCII.isLower(ch) || ASCII.isUpper(ch)) {
2867 // named captured group
2868 String name = groupname(ch);
2869 if (namedGroups().containsKey(name))
2870 throw error("Named capturing group <" + name
2871 + "> is already defined");
2872 capturingGroup = true;
2873 head = createGroup(false);
2874 tail = root;
2875 namedGroups().put(name, capturingGroupCount-1);
2876 head.next = expr(tail);
2877 break;
2878 }
2879 int start = cursor;
2880 head = createGroup(true);
2881 tail = root;
2882 head.next = expr(tail);
2883 tail.next = lookbehindEnd;
2884 TreeInfo info = new TreeInfo();
2885 head.study(info);
2886 if (info.maxValid == false) {
2887 throw error("Look-behind group does not have "
2888 + "an obvious maximum length");
2889 }
2890 boolean hasSupplementary = findSupplementary(start, patternLength);
2891 if (ch == '=') {
2892 head = tail = (hasSupplementary ?
2893 new BehindS(head, info.maxLength,
2894 info.minLength) :
2895 new Behind(head, info.maxLength,
2896 info.minLength));
2897 } else if (ch == '!') {
2898 head = tail = (hasSupplementary ?
2899 new NotBehindS(head, info.maxLength,
2900 info.minLength) :
2901 new NotBehind(head, info.maxLength,
2902 info.minLength));
2903 } else {
2904 throw error("Unknown look-behind group");
2905 }
2906 break;
2907 case '$':
2908 case '@':
2909 throw error("Unknown group type");
2910 default: // (?xxx:) inlined match flags
2911 unread();
2912 addFlag();
2913 ch = read();
2914 if (ch == ')') {
2915 return null; // Inline modifier only
2916 }
2917 if (ch != ':') {
2918 throw error("Unknown inline modifier");
2919 }
2920 head = createGroup(true);
2921 tail = root;
2922 head.next = expr(tail);
2923 break;
2924 }
2925 } else { // (xxx) a regular group
2926 capturingGroup = true;
2927 head = createGroup(false);
2928 tail = root;
2929 head.next = expr(tail);
2930 }
2931
2932 accept(')', "Unclosed group");
2933 flags = save;
2934
2935 // Check for quantifiers
2936 Node node = closure(head);
2937 if (node == head) { // No closure
2938 root = tail;
2939 return node; // Dual return
2940 }
2941 if (head == tail) { // Zero length assertion
2942 root = node;
2943 return node; // Dual return
2944 }
2945
2946 if (node instanceof Ques) {
2947 Ques ques = (Ques) node;
2948 if (ques.type == POSSESSIVE) {
2949 root = node;
2950 return node;
2951 }
2952 tail.next = new BranchConn();
2953 tail = tail.next;
2954 if (ques.type == GREEDY) {
2955 head = new Branch(head, null, tail);
2956 } else { // Reluctant quantifier
2957 head = new Branch(null, head, tail);
2958 }
2959 root = tail;
2960 return head;
2961 } else if (node instanceof Curly) {
2962 Curly curly = (Curly) node;
2963 if (curly.type == POSSESSIVE) {
2964 root = node;
2965 return node;
2966 }
2967 // Discover if the group is deterministic
2968 TreeInfo info = new TreeInfo();
2969 if (head.study(info)) { // Deterministic
2970 GroupTail temp = (GroupTail) tail;
2971 head = root = new GroupCurly(head.next, curly.cmin,
2972 curly.cmax, curly.type,
2973 ((GroupTail)tail).localIndex,
2974 ((GroupTail)tail).groupIndex,
2975 capturingGroup);
2976 return head;
2977 } else { // Non-deterministic
2978 int temp = ((GroupHead) head).localIndex;
2979 Loop loop;
2980 if (curly.type == GREEDY)
2981 loop = new Loop(this.localCount, temp);
2982 else // Reluctant Curly
2983 loop = new LazyLoop(this.localCount, temp);
2984 Prolog prolog = new Prolog(loop);
2985 this.localCount += 1;
2986 loop.cmin = curly.cmin;
2987 loop.cmax = curly.cmax;
2988 loop.body = head;
2989 tail.next = loop;
2990 root = loop;
2991 return prolog; // Dual return
2992 }
2993 }
2994 throw error("Internal logic error");
2995 }
2996
2997 /**
2998 * Create group head and tail nodes using double return. If the group is
2999 * created with anonymous true then it is a pure group and should not
3000 * affect group counting.
3001 */
3002 private Node createGroup(boolean anonymous) {
3003 int localIndex = localCount++;
3004 int groupIndex = 0;
3005 if (!anonymous)
3006 groupIndex = capturingGroupCount++;
3007 GroupHead head = new GroupHead(localIndex);
3008 root = new GroupTail(localIndex, groupIndex);
3009 if (!anonymous && groupIndex < 10)
3010 groupNodes[groupIndex] = head;
3011 return head;
3012 }
3013
3014 @SuppressWarnings("fallthrough")
3015 /**
3016 * Parses inlined match flags and set them appropriately.
3017 */
3018 private void addFlag() {
3019 int ch = peek();
3020 for (;;) {
3021 switch (ch) {
3022 case 'i':
3023 flags |= CASE_INSENSITIVE;
3024 break;
3025 case 'm':
3026 flags |= MULTILINE;
3027 break;
3028 case 's':
3029 flags |= DOTALL;
3030 break;
3031 case 'd':
3032 flags |= UNIX_LINES;
3033 break;
3034 case 'u':
3035 flags |= UNICODE_CASE;
3036 break;
3037 case 'c':
3038 flags |= CANON_EQ;
3039 break;
3040 case 'x':
3041 flags |= COMMENTS;
3042 break;
3043 case 'U':
3044 flags |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE);
3045 break;
3046 case '-': // subFlag then fall through
3047 ch = next();
3048 subFlag();
3049 default:
3050 return;
3051 }
3052 ch = next();
3053 }
3054 }
3055
3056 @SuppressWarnings("fallthrough")
3057 /**
3058 * Parses the second part of inlined match flags and turns off
3059 * flags appropriately.
3060 */
3061 private void subFlag() {
3062 int ch = peek();
3063 for (;;) {
3064 switch (ch) {
3065 case 'i':
3066 flags &= ~CASE_INSENSITIVE;
3067 break;
3068 case 'm':
3069 flags &= ~MULTILINE;
3070 break;
3071 case 's':
3072 flags &= ~DOTALL;
3073 break;
3074 case 'd':
3075 flags &= ~UNIX_LINES;
3076 break;
3077 case 'u':
3078 flags &= ~UNICODE_CASE;
3079 break;
3080 case 'c':
3081 flags &= ~CANON_EQ;
3082 break;
3083 case 'x':
3084 flags &= ~COMMENTS;
3085 break;
3086 case 'U':
3087 flags &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE);
3088 default:
3089 return;
3090 }
3091 ch = next();
3092 }
3093 }
3094
3095 static final int MAX_REPS = 0x7FFFFFFF;
3096
3097 static final int GREEDY = 0;
3098
3099 static final int LAZY = 1;
3100
3101 static final int POSSESSIVE = 2;
3102
3103 static final int INDEPENDENT = 3;
3104
3105 /**
3106 * Processes repetition. If the next character peeked is a quantifier
3107 * then new nodes must be appended to handle the repetition.
3108 * Prev could be a single or a group, so it could be a chain of nodes.
3109 */
3110 private Node closure(Node prev) {
3111 Node atom;
3112 int ch = peek();
3113 switch (ch) {
3114 case '?':
3115 ch = next();
3116 if (ch == '?') {
3117 next();
3118 return new Ques(prev, LAZY);
3119 } else if (ch == '+') {
3120 next();
3121 return new Ques(prev, POSSESSIVE);
3122 }
3123 return new Ques(prev, GREEDY);
3124 case '*':
3125 ch = next();
3126 if (ch == '?') {
3127 next();
3128 return new Curly(prev, 0, MAX_REPS, LAZY);
3129 } else if (ch == '+') {
3130 next();
3131 return new Curly(prev, 0, MAX_REPS, POSSESSIVE);
3132 }
3133 return new Curly(prev, 0, MAX_REPS, GREEDY);
3134 case '+':
3135 ch = next();
3136 if (ch == '?') {
3137 next();
3138 return new Curly(prev, 1, MAX_REPS, LAZY);
3139 } else if (ch == '+') {
3140 next();
3141 return new Curly(prev, 1, MAX_REPS, POSSESSIVE);
3142 }
3143 return new Curly(prev, 1, MAX_REPS, GREEDY);
3144 case '{':
3145 ch = temp[cursor+1];
3146 if (ASCII.isDigit(ch)) {
3147 skip();
3148 int cmin = 0;
3149 do {
3150 cmin = cmin * 10 + (ch - '0');
3151 } while (ASCII.isDigit(ch = read()));
3152 int cmax = cmin;
3153 if (ch == ',') {
3154 ch = read();
3155 cmax = MAX_REPS;
3156 if (ch != '}') {
3157 cmax = 0;
3158 while (ASCII.isDigit(ch)) {
3159 cmax = cmax * 10 + (ch - '0');
3160 ch = read();
3161 }
3162 }
3163 }
3164 if (ch != '}')
3165 throw error("Unclosed counted closure");
3166 if (((cmin) | (cmax) | (cmax - cmin)) < 0)
3167 throw error("Illegal repetition range");
3168 Curly curly;
3169 ch = peek();
3170 if (ch == '?') {
3171 next();
3172 curly = new Curly(prev, cmin, cmax, LAZY);
3173 } else if (ch == '+') {
3174 next();
3175 curly = new Curly(prev, cmin, cmax, POSSESSIVE);
3176 } else {
3177 curly = new Curly(prev, cmin, cmax, GREEDY);
3178 }
3179 return curly;
3180 } else {
3181 throw error("Illegal repetition");
3182 }
3183 default:
3184 return prev;
3185 }
3186 }
3187
3188 /**
3189 * Utility method for parsing control escape sequences.
3190 */
3191 private int c() {
3192 if (cursor < patternLength) {
3193 return read() ^ 64;
3194 }
3195 throw error("Illegal control escape sequence");
3196 }
3197
3198 /**
3199 * Utility method for parsing octal escape sequences.
3200 */
3201 private int o() {
3202 int n = read();
3203 if (((n-'0')|('7'-n)) >= 0) {
3204 int m = read();
3205 if (((m-'0')|('7'-m)) >= 0) {
3206 int o = read();
3207 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) {
3208 return (n - '0') * 64 + (m - '0') * 8 + (o - '0');
3209 }
3210 unread();
3211 return (n - '0') * 8 + (m - '0');
3212 }
3213 unread();
3214 return (n - '0');
3215 }
3216 throw error("Illegal octal escape sequence");
3217 }
3218
3219 /**
3220 * Utility method for parsing hexadecimal escape sequences.
3221 */
3222 private int x() {
3223 int n = read();
3224 if (ASCII.isHexDigit(n)) {
3225 int m = read();
3226 if (ASCII.isHexDigit(m)) {
3227 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m);
3228 }
3229 } else if (n == '{' && ASCII.isHexDigit(peek())) {
3230 int ch = 0;
3231 while (ASCII.isHexDigit(n = read())) {
3232 ch = (ch << 4) + ASCII.toDigit(n);
3233 if (ch > Character.MAX_CODE_POINT)
3234 throw error("Hexadecimal codepoint is too big");
3235 }
3236 if (n != '}')
3237 throw error("Unclosed hexadecimal escape sequence");
3238 return ch;
3239 }
3240 throw error("Illegal hexadecimal escape sequence");
3241 }
3242
3243 /**
3244 * Utility method for parsing unicode escape sequences.
3245 */
3246 private int cursor() {
3247 return cursor;
3248 }
3249
3250 private void setcursor(int pos) {
3251 cursor = pos;
3252 }
3253
3254 private int uxxxx() {
3255 int n = 0;
3256 for (int i = 0; i < 4; i++) {
3257 int ch = read();
3258 if (!ASCII.isHexDigit(ch)) {
3259 throw error("Illegal Unicode escape sequence");
3260 }
3261 n = n * 16 + ASCII.toDigit(ch);
3262 }
3263 return n;
3264 }
3265
3266 private int u() {
3267 int n = uxxxx();
3268 if (Character.isHighSurrogate((char)n)) {
3269 int cur = cursor();
3270 if (read() == '\\' && read() == 'u') {
3271 int n2 = uxxxx();
3272 if (Character.isLowSurrogate((char)n2))
3273 return Character.toCodePoint((char)n, (char)n2);
3274 }
3275 setcursor(cur);
3276 }
3277 return n;
3278 }
3279
3280 //
3281 // Utility methods for code point support
3282 //
3283
3284 private static final int countChars(CharSequence seq, int index,
3285 int lengthInCodePoints) {
3286 // optimization
3287 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) {
3288 assert (index >= 0 && index < seq.length());
3289 return 1;
3290 }
3291 int length = seq.length();
3292 int x = index;
3293 if (lengthInCodePoints >= 0) {
3294 assert (index >= 0 && index < length);
3295 for (int i = 0; x < length && i < lengthInCodePoints; i++) {
3296 if (Character.isHighSurrogate(seq.charAt(x++))) {
3297 if (x < length && Character.isLowSurrogate(seq.charAt(x))) {
3298 x++;
3299 }
3300 }
3301 }
3302 return x - index;
3303 }
3304
3305 assert (index >= 0 && index <= length);
3306 if (index == 0) {
3307 return 0;
3308 }
3309 int len = -lengthInCodePoints;
3310 for (int i = 0; x > 0 && i < len; i++) {
3311 if (Character.isLowSurrogate(seq.charAt(--x))) {
3312 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) {
3313 x--;
3314 }
3315 }
3316 }
3317 return index - x;
3318 }
3319
3320 private static final int countCodePoints(CharSequence seq) {
3321 int length = seq.length();
3322 int n = 0;
3323 for (int i = 0; i < length; ) {
3324 n++;
3325 if (Character.isHighSurrogate(seq.charAt(i++))) {
3326 if (i < length && Character.isLowSurrogate(seq.charAt(i))) {
3327 i++;
3328 }
3329 }
3330 }
3331 return n;
3332 }
3333
3334 /**
3335 * Creates a bit vector for matching Latin-1 values. A normal BitClass
3336 * never matches values above Latin-1, and a complemented BitClass always
3337 * matches values above Latin-1.
3338 */
3339 private static final class BitClass extends BmpCharProperty {
3340 final boolean[] bits;
3341 BitClass() { bits = new boolean[256]; }
3342 private BitClass(boolean[] bits) { this.bits = bits; }
3343 BitClass add(int c, int flags) {
3344 assert c >= 0 && c <= 255;
3345 if ((flags & CASE_INSENSITIVE) != 0) {
3346 if (ASCII.isAscii(c)) {
3347 bits[ASCII.toUpper(c)] = true;
3348 bits[ASCII.toLower(c)] = true;
3349 } else if ((flags & UNICODE_CASE) != 0) {
3350 bits[Character.toLowerCase(c)] = true;
3351 bits[Character.toUpperCase(c)] = true;
3352 }
3353 }
3354 bits[c] = true;
3355 return this;
3356 }
3357 boolean isSatisfiedBy(int ch) {
3358 return ch < 256 && bits[ch];
3359 }
3360 }
3361
3362 /**
3363 * Returns a suitably optimized, single character matcher.
3364 */
3365 private CharProperty newSingle(final int ch) {
3366 if (has(CASE_INSENSITIVE)) {
3367 int lower, upper;
3368 if (has(UNICODE_CASE)) {
3369 upper = Character.toUpperCase(ch);
3370 lower = Character.toLowerCase(upper);
3371 if (upper != lower)
3372 return new SingleU(lower);
3373 } else if (ASCII.isAscii(ch)) {
3374 lower = ASCII.toLower(ch);
3375 upper = ASCII.toUpper(ch);
3376 if (lower != upper)
3377 return new SingleI(lower, upper);
3378 }
3379 }
3380 if (isSupplementary(ch))
3381 return new SingleS(ch); // Match a given Unicode character
3382 return new Single(ch); // Match a given BMP character
3383 }
3384
3385 /**
3386 * Utility method for creating a string slice matcher.
3387 */
3388 private Node newSlice(int[] buf, int count, boolean hasSupplementary) {
3389 int[] tmp = new int[count];
3390 if (has(CASE_INSENSITIVE)) {
3391 if (has(UNICODE_CASE)) {
3392 for (int i = 0; i < count; i++) {
3393 tmp[i] = Character.toLowerCase(
3394 Character.toUpperCase(buf[i]));
3395 }
3396 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp);
3397 }
3398 for (int i = 0; i < count; i++) {
3399 tmp[i] = ASCII.toLower(buf[i]);
3400 }
3401 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp);
3402 }
3403 for (int i = 0; i < count; i++) {
3404 tmp[i] = buf[i];
3405 }
3406 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp);
3407 }
3408
3409 /**
3410 * The following classes are the building components of the object
3411 * tree that represents a compiled regular expression. The object tree
3412 * is made of individual elements that handle constructs in the Pattern.
3413 * Each type of object knows how to match its equivalent construct with
3414 * the match() method.
3415 */
3416
3417 /**
3418 * Base class for all node classes. Subclasses should override the match()
3419 * method as appropriate. This class is an accepting node, so its match()
3420 * always returns true.
3421 */
3422 static class Node extends Object {
3423 Node next;
3424 Node() {
3425 next = Pattern.accept;
3426 }
3427 /**
3428 * This method implements the classic accept node.
3429 */
3430 boolean match(Matcher matcher, int i, CharSequence seq) {
3431 matcher.last = i;
3432 matcher.groups[0] = matcher.first;
3433 matcher.groups[1] = matcher.last;
3434 return true;
3435 }
3436 /**
3437 * This method is good for all zero length assertions.
3438 */
3439 boolean study(TreeInfo info) {
3440 if (next != null) {
3441 return next.study(info);
3442 } else {
3443 return info.deterministic;
3444 }
3445 }
3446 }
3447
3448 static class LastNode extends Node {
3449 /**
3450 * This method implements the classic accept node with
3451 * the addition of a check to see if the match occurred
3452 * using all of the input.
3453 */
3454 boolean match(Matcher matcher, int i, CharSequence seq) {
3455 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to)
3456 return false;
3457 matcher.last = i;
3458 matcher.groups[0] = matcher.first;
3459 matcher.groups[1] = matcher.last;
3460 return true;
3461 }
3462 }
3463
3464 /**
3465 * Used for REs that can start anywhere within the input string.
3466 * This basically tries to match repeatedly at each spot in the
3467 * input string, moving forward after each try. An anchored search
3468 * or a BnM will bypass this node completely.
3469 */
3470 static class Start extends Node {
3471 int minLength;
3472 Start(Node node) {
3473 this.next = node;
3474 TreeInfo info = new TreeInfo();
3475 next.study(info);
3476 minLength = info.minLength;
3477 }
3478 boolean match(Matcher matcher, int i, CharSequence seq) {
3479 if (i > matcher.to - minLength) {
3480 matcher.hitEnd = true;
3481 return false;
3482 }
3483 int guard = matcher.to - minLength;
3484 for (; i <= guard; i++) {
3485 if (next.match(matcher, i, seq)) {
3486 matcher.first = i;
3487 matcher.groups[0] = matcher.first;
3488 matcher.groups[1] = matcher.last;
3489 return true;
3490 }
3491 }
3492 matcher.hitEnd = true;
3493 return false;
3494 }
3495 boolean study(TreeInfo info) {
3496 next.study(info);
3497 info.maxValid = false;
3498 info.deterministic = false;
3499 return false;
3500 }
3501 }
3502
3503 /*
3504 * StartS supports supplementary characters, including unpaired surrogates.
3505 */
3506 static final class StartS extends Start {
3507 StartS(Node node) {
3508 super(node);
3509 }
3510 boolean match(Matcher matcher, int i, CharSequence seq) {
3511 if (i > matcher.to - minLength) {
3512 matcher.hitEnd = true;
3513 return false;
3514 }
3515 int guard = matcher.to - minLength;
3516 while (i <= guard) {
3517 //if ((ret = next.match(matcher, i, seq)) || i == guard)
3518 if (next.match(matcher, i, seq)) {
3519 matcher.first = i;
3520 matcher.groups[0] = matcher.first;
3521 matcher.groups[1] = matcher.last;
3522 return true;
3523 }
3524 if (i == guard)
3525 break;
3526 // Optimization to move to the next character. This is
3527 // faster than countChars(seq, i, 1).
3528 if (Character.isHighSurrogate(seq.charAt(i++))) {
3529 if (i < seq.length() &&
3530 Character.isLowSurrogate(seq.charAt(i))) {
3531 i++;
3532 }
3533 }
3534 }
3535 matcher.hitEnd = true;
3536 return false;
3537 }
3538 }
3539
3540 /**
3541 * Node to anchor at the beginning of input. This object implements the
3542 * match for a \A sequence, and the caret anchor will use this if not in
3543 * multiline mode.
3544 */
3545 static final class Begin extends Node {
3546 boolean match(Matcher matcher, int i, CharSequence seq) {
3547 int fromIndex = (matcher.anchoringBounds) ?
3548 matcher.from : 0;
3549 if (i == fromIndex && next.match(matcher, i, seq)) {
3550 matcher.first = i;
3551 matcher.groups[0] = i;
3552 matcher.groups[1] = matcher.last;
3553 return true;
3554 } else {
3555 return false;
3556 }
3557 }
3558 }
3559
3560 /**
3561 * Node to anchor at the end of input. This is the absolute end, so this
3562 * should not match at the last newline before the end as $ will.
3563 */
3564 static final class End extends Node {
3565 boolean match(Matcher matcher, int i, CharSequence seq) {
3566 int endIndex = (matcher.anchoringBounds) ?
3567 matcher.to : matcher.getTextLength();
3568 if (i == endIndex) {
3569 matcher.hitEnd = true;
3570 return next.match(matcher, i, seq);
3571 }
3572 return false;
3573 }
3574 }
3575
3576 /**
3577 * Node to anchor at the beginning of a line. This is essentially the
3578 * object to match for the multiline ^.
3579 */
3580 static final class Caret extends Node {
3581 boolean match(Matcher matcher, int i, CharSequence seq) {
3582 int startIndex = matcher.from;
3583 int endIndex = matcher.to;
3584 if (!matcher.anchoringBounds) {
3585 startIndex = 0;
3586 endIndex = matcher.getTextLength();
3587 }
3588 // Perl does not match ^ at end of input even after newline
3589 if (i == endIndex) {
3590 matcher.hitEnd = true;
3591 return false;
3592 }
3593 if (i > startIndex) {
3594 char ch = seq.charAt(i-1);
3595 if (ch != '\n' && ch != '\r'
3596 && (ch|1) != '\u2029'
3597 && ch != '\u0085' ) {
3598 return false;
3599 }
3600 // Should treat /r/n as one newline
3601 if (ch == '\r' && seq.charAt(i) == '\n')
3602 return false;
3603 }
3604 return next.match(matcher, i, seq);
3605 }
3606 }
3607
3608 /**
3609 * Node to anchor at the beginning of a line when in unixdot mode.
3610 */
3611 static final class UnixCaret extends Node {
3612 boolean match(Matcher matcher, int i, CharSequence seq) {
3613 int startIndex = matcher.from;
3614 int endIndex = matcher.to;
3615 if (!matcher.anchoringBounds) {
3616 startIndex = 0;
3617 endIndex = matcher.getTextLength();
3618 }
3619 // Perl does not match ^ at end of input even after newline
3620 if (i == endIndex) {
3621 matcher.hitEnd = true;
3622 return false;
3623 }
3624 if (i > startIndex) {
3625 char ch = seq.charAt(i-1);
3626 if (ch != '\n') {
3627 return false;
3628 }
3629 }
3630 return next.match(matcher, i, seq);
3631 }
3632 }
3633
3634 /**
3635 * Node to match the location where the last match ended.
3636 * This is used for the \G construct.
3637 */
3638 static final class LastMatch extends Node {
3639 boolean match(Matcher matcher, int i, CharSequence seq) {
3640 if (i != matcher.oldLast)
3641 return false;
3642 return next.match(matcher, i, seq);
3643 }
3644 }
3645
3646 /**
3647 * Node to anchor at the end of a line or the end of input based on the
3648 * multiline mode.
3649 *
3650 * When not in multiline mode, the $ can only match at the very end
3651 * of the input, unless the input ends in a line terminator in which
3652 * it matches right before the last line terminator.
3653 *
3654 * Note that \r\n is considered an atomic line terminator.
3655 *
3656 * Like ^ the $ operator matches at a position, it does not match the
3657 * line terminators themselves.
3658 */
3659 static final class Dollar extends Node {
3660 boolean multiline;
3661 Dollar(boolean mul) {
3662 multiline = mul;
3663 }
3664 boolean match(Matcher matcher, int i, CharSequence seq) {
3665 int endIndex = (matcher.anchoringBounds) ?
3666 matcher.to : matcher.getTextLength();
3667 if (!multiline) {
3668 if (i < endIndex - 2)
3669 return false;
3670 if (i == endIndex - 2) {
3671 char ch = seq.charAt(i);
3672 if (ch != '\r')
3673 return false;
3674 ch = seq.charAt(i + 1);
3675 if (ch != '\n')
3676 return false;
3677 }
3678 }
3679 // Matches before any line terminator; also matches at the
3680 // end of input
3681 // Before line terminator:
3682 // If multiline, we match here no matter what
3683 // If not multiline, fall through so that the end
3684 // is marked as hit; this must be a /r/n or a /n
3685 // at the very end so the end was hit; more input
3686 // could make this not match here
3687 if (i < endIndex) {
3688 char ch = seq.charAt(i);
3689 if (ch == '\n') {
3690 // No match between \r\n
3691 if (i > 0 && seq.charAt(i-1) == '\r')
3692 return false;
3693 if (multiline)
3694 return next.match(matcher, i, seq);
3695 } else if (ch == '\r' || ch == '\u0085' ||
3696 (ch|1) == '\u2029') {
3697 if (multiline)
3698 return next.match(matcher, i, seq);
3699 } else { // No line terminator, no match
3700 return false;
3701 }
3702 }
3703 // Matched at current end so hit end
3704 matcher.hitEnd = true;
3705 // If a $ matches because of end of input, then more input
3706 // could cause it to fail!
3707 matcher.requireEnd = true;
3708 return next.match(matcher, i, seq);
3709 }
3710 boolean study(TreeInfo info) {
3711 next.study(info);
3712 return info.deterministic;
3713 }
3714 }
3715
3716 /**
3717 * Node to anchor at the end of a line or the end of input based on the
3718 * multiline mode when in unix lines mode.
3719 */
3720 static final class UnixDollar extends Node {
3721 boolean multiline;
3722 UnixDollar(boolean mul) {
3723 multiline = mul;
3724 }
3725 boolean match(Matcher matcher, int i, CharSequence seq) {
3726 int endIndex = (matcher.anchoringBounds) ?
3727 matcher.to : matcher.getTextLength();
3728 if (i < endIndex) {
3729 char ch = seq.charAt(i);
3730 if (ch == '\n') {
3731 // If not multiline, then only possible to
3732 // match at very end or one before end
3733 if (multiline == false && i != endIndex - 1)
3734 return false;
3735 // If multiline return next.match without setting
3736 // matcher.hitEnd
3737 if (multiline)
3738 return next.match(matcher, i, seq);
3739 } else {
3740 return false;
3741 }
3742 }
3743 // Matching because at the end or 1 before the end;
3744 // more input could change this so set hitEnd
3745 matcher.hitEnd = true;
3746 // If a $ matches because of end of input, then more input
3747 // could cause it to fail!
3748 matcher.requireEnd = true;
3749 return next.match(matcher, i, seq);
3750 }
3751 boolean study(TreeInfo info) {
3752 next.study(info);
3753 return info.deterministic;
3754 }
3755 }
3756
3757 /**
3758 * Node class that matches a Unicode line ending '\R'
3759 */
3760 static final class LineEnding extends Node {
3761 boolean match(Matcher matcher, int i, CharSequence seq) {
3762 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029])
3763 if (i < matcher.to) {
3764 int ch = seq.charAt(i);
3765 if (ch == 0x0A || ch == 0x0B || ch == 0x0C ||
3766 ch == 0x85 || ch == 0x2028 || ch == 0x2029)
3767 return next.match(matcher, i + 1, seq);
3768 if (ch == 0x0D) {
3769 i++;
3770 if (i < matcher.to && seq.charAt(i) == 0x0A)
3771 i++;
3772 return next.match(matcher, i, seq);
3773 }
3774 } else {
3775 matcher.hitEnd = true;
3776 }
3777 return false;
3778 }
3779 boolean study(TreeInfo info) {
3780 info.minLength++;
3781 info.maxLength += 2;
3782 return next.study(info);
3783 }
3784 }
3785
3786 /**
3787 * Abstract node class to match one character satisfying some
3788 * boolean property.
3789 */
3790 private static abstract class CharProperty extends Node {
3791 abstract boolean isSatisfiedBy(int ch);
3792 CharProperty complement() {
3793 return new CharProperty() {
3794 boolean isSatisfiedBy(int ch) {
3795 return ! CharProperty.this.isSatisfiedBy(ch);}};
3796 }
3797 boolean match(Matcher matcher, int i, CharSequence seq) {
3798 if (i < matcher.to) {
3799 int ch = Character.codePointAt(seq, i);
3800 return isSatisfiedBy(ch)
3801 && next.match(matcher, i+Character.charCount(ch), seq);
3802 } else {
3803 matcher.hitEnd = true;
3804 return false;
3805 }
3806 }
3807 boolean study(TreeInfo info) {
3808 info.minLength++;
3809 info.maxLength++;
3810 return next.study(info);
3811 }
3812 }
3813
3814 /**
3815 * Optimized version of CharProperty that works only for
3816 * properties never satisfied by Supplementary characters.
3817 */
3818 private static abstract class BmpCharProperty extends CharProperty {
3819 boolean match(Matcher matcher, int i, CharSequence seq) {
3820 if (i < matcher.to) {
3821 return isSatisfiedBy(seq.charAt(i))
3822 && next.match(matcher, i+1, seq);
3823 } else {
3824 matcher.hitEnd = true;
3825 return false;
3826 }
3827 }
3828 }
3829
3830 /**
3831 * Node class that matches a Supplementary Unicode character
3832 */
3833 static final class SingleS extends CharProperty {
3834 final int c;
3835 SingleS(int c) { this.c = c; }
3836 boolean isSatisfiedBy(int ch) {
3837 return ch == c;
3838 }
3839 }
3840
3841 /**
3842 * Optimization -- matches a given BMP character
3843 */
3844 static final class Single extends BmpCharProperty {
3845 final int c;
3846 Single(int c) { this.c = c; }
3847 boolean isSatisfiedBy(int ch) {
3848 return ch == c;
3849 }
3850 }
3851
3852 /**
3853 * Case insensitive matches a given BMP character
3854 */
3855 static final class SingleI extends BmpCharProperty {
3856 final int lower;
3857 final int upper;
3858 SingleI(int lower, int upper) {
3859 this.lower = lower;
3860 this.upper = upper;
3861 }
3862 boolean isSatisfiedBy(int ch) {
3863 return ch == lower || ch == upper;
3864 }
3865 }
3866
3867 /**
3868 * Unicode case insensitive matches a given Unicode character
3869 */
3870 static final class SingleU extends CharProperty {
3871 final int lower;
3872 SingleU(int lower) {
3873 this.lower = lower;
3874 }
3875 boolean isSatisfiedBy(int ch) {
3876 return lower == ch ||
3877 lower == Character.toLowerCase(Character.toUpperCase(ch));
3878 }
3879 }
3880
3881 /**
3882 * Node class that matches a Unicode block.
3883 */
3884 static final class Block extends CharProperty {
3885 final Character.UnicodeBlock block;
3886 Block(Character.UnicodeBlock block) {
3887 this.block = block;
3888 }
3889 boolean isSatisfiedBy(int ch) {
3890 return block == Character.UnicodeBlock.of(ch);
3891 }
3892 }
3893
3894 /**
3895 * Node class that matches a Unicode script
3896 */
3897 static final class Script extends CharProperty {
3898 final Character.UnicodeScript script;
3899 Script(Character.UnicodeScript script) {
3900 this.script = script;
3901 }
3902 boolean isSatisfiedBy(int ch) {
3903 return script == Character.UnicodeScript.of(ch);
3904 }
3905 }
3906
3907 /**
3908 * Node class that matches a Unicode category.
3909 */
3910 static final class Category extends CharProperty {
3911 final int typeMask;
3912 Category(int typeMask) { this.typeMask = typeMask; }
3913 boolean isSatisfiedBy(int ch) {
3914 return (typeMask & (1 << Character.getType(ch))) != 0;
3915 }
3916 }
3917
3918 /**
3919 * Node class that matches a Unicode "type"
3920 */
3921 static final class Utype extends CharProperty {
3922 final UnicodeProp uprop;
3923 Utype(UnicodeProp uprop) { this.uprop = uprop; }
3924 boolean isSatisfiedBy(int ch) {
3925 return uprop.is(ch);
3926 }
3927 }
3928
3929 /**
3930 * Node class that matches a POSIX type.
3931 */
3932 static final class Ctype extends BmpCharProperty {
3933 final int ctype;
3934 Ctype(int ctype) { this.ctype = ctype; }
3935 boolean isSatisfiedBy(int ch) {
3936 return ch < 128 && ASCII.isType(ch, ctype);
3937 }
3938 }
3939
3940 /**
3941 * Node class that matches a Perl vertical whitespace
3942 */
3943 static final class VertWS extends BmpCharProperty {
3944 boolean isSatisfiedBy(int cp) {
3945 return (cp >= 0x0A && cp <= 0x0D) ||
3946 cp == 0x85 || cp == 0x2028 || cp == 0x2029;
3947 }
3948 }
3949
3950 /**
3951 * Node class that matches a Perl horizontal whitespace
3952 */
3953 static final class HorizWS extends BmpCharProperty {
3954 boolean isSatisfiedBy(int cp) {
3955 return cp == 0x09 || cp == 0x20 || cp == 0xa0 ||
3956 cp == 0x1680 || cp == 0x180e ||
3957 cp >= 0x2000 && cp <= 0x200a ||
3958 cp == 0x202f || cp == 0x205f || cp == 0x3000;
3959 }
3960 }
3961
3962 /**
3963 * Base class for all Slice nodes
3964 */
3965 static class SliceNode extends Node {
3966 int[] buffer;
3967 SliceNode(int[] buf) {
3968 buffer = buf;
3969 }
3970 boolean study(TreeInfo info) {
3971 info.minLength += buffer.length;
3972 info.maxLength += buffer.length;
3973 return next.study(info);
3974 }
3975 }
3976
3977 /**
3978 * Node class for a case sensitive/BMP-only sequence of literal
3979 * characters.
3980 */
3981 static class Slice extends SliceNode {
3982 Slice(int[] buf) {
3983 super(buf);
3984 }
3985 boolean match(Matcher matcher, int i, CharSequence seq) {
3986 int[] buf = buffer;
3987 int len = buf.length;
3988 for (int j=0; j<len; j++) {
3989 if ((i+j) >= matcher.to) {
3990 matcher.hitEnd = true;
3991 return false;
3992 }
3993 if (buf[j] != seq.charAt(i+j))
3994 return false;
3995 }
3996 return next.match(matcher, i+len, seq);
3997 }
3998 }
3999
4000 /**
4001 * Node class for a case_insensitive/BMP-only sequence of literal
4002 * characters.
4003 */
4004 static class SliceI extends SliceNode {
4005 SliceI(int[] buf) {
4006 super(buf);
4007 }
4008 boolean match(Matcher matcher, int i, CharSequence seq) {
4009 int[] buf = buffer;
4010 int len = buf.length;
4011 for (int j=0; j<len; j++) {
4012 if ((i+j) >= matcher.to) {
4013 matcher.hitEnd = true;
4014 return false;
4015 }
4016 int c = seq.charAt(i+j);
4017 if (buf[j] != c &&
4018 buf[j] != ASCII.toLower(c))
4019 return false;
4020 }
4021 return next.match(matcher, i+len, seq);
4022 }
4023 }
4024
4025 /**
4026 * Node class for a unicode_case_insensitive/BMP-only sequence of
4027 * literal characters. Uses unicode case folding.
4028 */
4029 static final class SliceU extends SliceNode {
4030 SliceU(int[] buf) {
4031 super(buf);
4032 }
4033 boolean match(Matcher matcher, int i, CharSequence seq) {
4034 int[] buf = buffer;
4035 int len = buf.length;
4036 for (int j=0; j<len; j++) {
4037 if ((i+j) >= matcher.to) {
4038 matcher.hitEnd = true;
4039 return false;
4040 }
4041 int c = seq.charAt(i+j);
4042 if (buf[j] != c &&
4043 buf[j] != Character.toLowerCase(Character.toUpperCase(c)))
4044 return false;
4045 }
4046 return next.match(matcher, i+len, seq);
4047 }
4048 }
4049
4050 /**
4051 * Node class for a case sensitive sequence of literal characters
4052 * including supplementary characters.
4053 */
4054 static final class SliceS extends Slice {
4055 SliceS(int[] buf) {
4056 super(buf);
4057 }
4058 boolean match(Matcher matcher, int i, CharSequence seq) {
4059 int[] buf = buffer;
4060 int x = i;
4061 for (int j = 0; j < buf.length; j++) {
4062 if (x >= matcher.to) {
4063 matcher.hitEnd = true;
4064 return false;
4065 }
4066 int c = Character.codePointAt(seq, x);
4067 if (buf[j] != c)
4068 return false;
4069 x += Character.charCount(c);
4070 if (x > matcher.to) {
4071 matcher.hitEnd = true;
4072 return false;
4073 }
4074 }
4075 return next.match(matcher, x, seq);
4076 }
4077 }
4078
4079 /**
4080 * Node class for a case insensitive sequence of literal characters
4081 * including supplementary characters.
4082 */
4083 static class SliceIS extends SliceNode {
4084 SliceIS(int[] buf) {
4085 super(buf);
4086 }
4087 int toLower(int c) {
4088 return ASCII.toLower(c);
4089 }
4090 boolean match(Matcher matcher, int i, CharSequence seq) {
4091 int[] buf = buffer;
4092 int x = i;
4093 for (int j = 0; j < buf.length; j++) {
4094 if (x >= matcher.to) {
4095 matcher.hitEnd = true;
4096 return false;
4097 }
4098 int c = Character.codePointAt(seq, x);
4099 if (buf[j] != c && buf[j] != toLower(c))
4100 return false;
4101 x += Character.charCount(c);
4102 if (x > matcher.to) {
4103 matcher.hitEnd = true;
4104 return false;
4105 }
4106 }
4107 return next.match(matcher, x, seq);
4108 }
4109 }
4110
4111 /**
4112 * Node class for a case insensitive sequence of literal characters.
4113 * Uses unicode case folding.
4114 */
4115 static final class SliceUS extends SliceIS {
4116 SliceUS(int[] buf) {
4117 super(buf);
4118 }
4119 int toLower(int c) {
4120 return Character.toLowerCase(Character.toUpperCase(c));
4121 }
4122 }
4123
4124 private static boolean inRange(int lower, int ch, int upper) {
4125 return lower <= ch && ch <= upper;
4126 }
4127
4128 /**
4129 * Returns node for matching characters within an explicit value range.
4130 */
4131 private static CharProperty rangeFor(final int lower,
4132 final int upper) {
4133 return new CharProperty() {
4134 boolean isSatisfiedBy(int ch) {
4135 return inRange(lower, ch, upper);}};
4136 }
4137
4138 /**
4139 * Returns node for matching characters within an explicit value
4140 * range in a case insensitive manner.
4141 */
4142 private CharProperty caseInsensitiveRangeFor(final int lower,
4143 final int upper) {
4144 if (has(UNICODE_CASE))
4145 return new CharProperty() {
4146 boolean isSatisfiedBy(int ch) {
4147 if (inRange(lower, ch, upper))
4148 return true;
4149 int up = Character.toUpperCase(ch);
4150 return inRange(lower, up, upper) ||
4151 inRange(lower, Character.toLowerCase(up), upper);}};
4152 return new CharProperty() {
4153 boolean isSatisfiedBy(int ch) {
4154 return inRange(lower, ch, upper) ||
4155 ASCII.isAscii(ch) &&
4156 (inRange(lower, ASCII.toUpper(ch), upper) ||
4157 inRange(lower, ASCII.toLower(ch), upper));
4158 }};
4159 }
4160
4161 /**
4162 * Implements the Unicode category ALL and the dot metacharacter when
4163 * in dotall mode.
4164 */
4165 static final class All extends CharProperty {
4166 boolean isSatisfiedBy(int ch) {
4167 return true;
4168 }
4169 }
4170
4171 /**
4172 * Node class for the dot metacharacter when dotall is not enabled.
4173 */
4174 static final class Dot extends CharProperty {
4175 boolean isSatisfiedBy(int ch) {
4176 return (ch != '\n' && ch != '\r'
4177 && (ch|1) != '\u2029'
4178 && ch != '\u0085');
4179 }
4180 }
4181
4182 /**
4183 * Node class for the dot metacharacter when dotall is not enabled
4184 * but UNIX_LINES is enabled.
4185 */
4186 static final class UnixDot extends CharProperty {
4187 boolean isSatisfiedBy(int ch) {
4188 return ch != '\n';
4189 }
4190 }
4191
4192 /**
4193 * The 0 or 1 quantifier. This one class implements all three types.
4194 */
4195 static final class Ques extends Node {
4196 Node atom;
4197 int type;
4198 Ques(Node node, int type) {
4199 this.atom = node;
4200 this.type = type;
4201 }
4202 boolean match(Matcher matcher, int i, CharSequence seq) {
4203 switch (type) {
4204 case GREEDY:
4205 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq))
4206 || next.match(matcher, i, seq);
4207 case LAZY:
4208 return next.match(matcher, i, seq)
4209 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq));
4210 case POSSESSIVE:
4211 if (atom.match(matcher, i, seq)) i = matcher.last;
4212 return next.match(matcher, i, seq);
4213 default:
4214 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq);
4215 }
4216 }
4217 boolean study(TreeInfo info) {
4218 if (type != INDEPENDENT) {
4219 int minL = info.minLength;
4220 atom.study(info);
4221 info.minLength = minL;
4222 info.deterministic = false;
4223 return next.study(info);
4224 } else {
4225 atom.study(info);
4226 return next.study(info);
4227 }
4228 }
4229 }
4230
4231 /**
4232 * Handles the curly-brace style repetition with a specified minimum and
4233 * maximum occurrences. The * quantifier is handled as a special case.
4234 * This class handles the three types.
4235 */
4236 static final class Curly extends Node {
4237 Node atom;
4238 int type;
4239 int cmin;
4240 int cmax;
4241
4242 Curly(Node node, int cmin, int cmax, int type) {
4243 this.atom = node;
4244 this.type = type;
4245 this.cmin = cmin;
4246 this.cmax = cmax;
4247 }
4248 boolean match(Matcher matcher, int i, CharSequence seq) {
4249 int j;
4250 for (j = 0; j < cmin; j++) {
4251 if (atom.match(matcher, i, seq)) {
4252 i = matcher.last;
4253 continue;
4254 }
4255 return false;
4256 }
4257 if (type == GREEDY)
4258 return match0(matcher, i, j, seq);
4259 else if (type == LAZY)
4260 return match1(matcher, i, j, seq);
4261 else
4262 return match2(matcher, i, j, seq);
4263 }
4264 // Greedy match.
4265 // i is the index to start matching at
4266 // j is the number of atoms that have matched
4267 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4268 if (j >= cmax) {
4269 // We have matched the maximum... continue with the rest of
4270 // the regular expression
4271 return next.match(matcher, i, seq);
4272 }
4273 int backLimit = j;
4274 while (atom.match(matcher, i, seq)) {
4275 // k is the length of this match
4276 int k = matcher.last - i;
4277 if (k == 0) // Zero length match
4278 break;
4279 // Move up index and number matched
4280 i = matcher.last;
4281 j++;
4282 // We are greedy so match as many as we can
4283 while (j < cmax) {
4284 if (!atom.match(matcher, i, seq))
4285 break;
4286 if (i + k != matcher.last) {
4287 if (match0(matcher, matcher.last, j+1, seq))
4288 return true;
4289 break;
4290 }
4291 i += k;
4292 j++;
4293 }
4294 // Handle backing off if match fails
4295 while (j >= backLimit) {
4296 if (next.match(matcher, i, seq))
4297 return true;
4298 i -= k;
4299 j--;
4300 }
4301 return false;
4302 }
4303 return next.match(matcher, i, seq);
4304 }
4305 // Reluctant match. At this point, the minimum has been satisfied.
4306 // i is the index to start matching at
4307 // j is the number of atoms that have matched
4308 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4309 for (;;) {
4310 // Try finishing match without consuming any more
4311 if (next.match(matcher, i, seq))
4312 return true;
4313 // At the maximum, no match found
4314 if (j >= cmax)
4315 return false;
4316 // Okay, must try one more atom
4317 if (!atom.match(matcher, i, seq))
4318 return false;
4319 // If we haven't moved forward then must break out
4320 if (i == matcher.last)
4321 return false;
4322 // Move up index and number matched
4323 i = matcher.last;
4324 j++;
4325 }
4326 }
4327 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4328 for (; j < cmax; j++) {
4329 if (!atom.match(matcher, i, seq))
4330 break;
4331 if (i == matcher.last)
4332 break;
4333 i = matcher.last;
4334 }
4335 return next.match(matcher, i, seq);
4336 }
4337 boolean study(TreeInfo info) {
4338 // Save original info
4339 int minL = info.minLength;
4340 int maxL = info.maxLength;
4341 boolean maxV = info.maxValid;
4342 boolean detm = info.deterministic;
4343 info.reset();
4344
4345 atom.study(info);
4346
4347 int temp = info.minLength * cmin + minL;
4348 if (temp < minL) {
4349 temp = 0xFFFFFFF; // arbitrary large number
4350 }
4351 info.minLength = temp;
4352
4353 if (maxV & info.maxValid) {
4354 temp = info.maxLength * cmax + maxL;
4355 info.maxLength = temp;
4356 if (temp < maxL) {
4357 info.maxValid = false;
4358 }
4359 } else {
4360 info.maxValid = false;
4361 }
4362
4363 if (info.deterministic && cmin == cmax)
4364 info.deterministic = detm;
4365 else
4366 info.deterministic = false;
4367 return next.study(info);
4368 }
4369 }
4370
4371 /**
4372 * Handles the curly-brace style repetition with a specified minimum and
4373 * maximum occurrences in deterministic cases. This is an iterative
4374 * optimization over the Prolog and Loop system which would handle this
4375 * in a recursive way. The * quantifier is handled as a special case.
4376 * If capture is true then this class saves group settings and ensures
4377 * that groups are unset when backing off of a group match.
4378 */
4379 static final class GroupCurly extends Node {
4380 Node atom;
4381 int type;
4382 int cmin;
4383 int cmax;
4384 int localIndex;
4385 int groupIndex;
4386 boolean capture;
4387
4388 GroupCurly(Node node, int cmin, int cmax, int type, int local,
4389 int group, boolean capture) {
4390 this.atom = node;
4391 this.type = type;
4392 this.cmin = cmin;
4393 this.cmax = cmax;
4394 this.localIndex = local;
4395 this.groupIndex = group;
4396 this.capture = capture;
4397 }
4398 boolean match(Matcher matcher, int i, CharSequence seq) {
4399 int[] groups = matcher.groups;
4400 int[] locals = matcher.locals;
4401 int save0 = locals[localIndex];
4402 int save1 = 0;
4403 int save2 = 0;
4404
4405 if (capture) {
4406 save1 = groups[groupIndex];
4407 save2 = groups[groupIndex+1];
4408 }
4409
4410 // Notify GroupTail there is no need to setup group info
4411 // because it will be set here
4412 locals[localIndex] = -1;
4413
4414 boolean ret = true;
4415 for (int j = 0; j < cmin; j++) {
4416 if (atom.match(matcher, i, seq)) {
4417 if (capture) {
4418 groups[groupIndex] = i;
4419 groups[groupIndex+1] = matcher.last;
4420 }
4421 i = matcher.last;
4422 } else {
4423 ret = false;
4424 break;
4425 }
4426 }
4427 if (ret) {
4428 if (type == GREEDY) {
4429 ret = match0(matcher, i, cmin, seq);
4430 } else if (type == LAZY) {
4431 ret = match1(matcher, i, cmin, seq);
4432 } else {
4433 ret = match2(matcher, i, cmin, seq);
4434 }
4435 }
4436 if (!ret) {
4437 locals[localIndex] = save0;
4438 if (capture) {
4439 groups[groupIndex] = save1;
4440 groups[groupIndex+1] = save2;
4441 }
4442 }
4443 return ret;
4444 }
4445 // Aggressive group match
4446 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4447 // don't back off passing the starting "j"
4448 int min = j;
4449 int[] groups = matcher.groups;
4450 int save0 = 0;
4451 int save1 = 0;
4452 if (capture) {
4453 save0 = groups[groupIndex];
4454 save1 = groups[groupIndex+1];
4455 }
4456 for (;;) {
4457 if (j >= cmax)
4458 break;
4459 if (!atom.match(matcher, i, seq))
4460 break;
4461 int k = matcher.last - i;
4462 if (k <= 0) {
4463 if (capture) {
4464 groups[groupIndex] = i;
4465 groups[groupIndex+1] = i + k;
4466 }
4467 i = i + k;
4468 break;
4469 }
4470 for (;;) {
4471 if (capture) {
4472 groups[groupIndex] = i;
4473 groups[groupIndex+1] = i + k;
4474 }
4475 i = i + k;
4476 if (++j >= cmax)
4477 break;
4478 if (!atom.match(matcher, i, seq))
4479 break;
4480 if (i + k != matcher.last) {
4481 if (match0(matcher, i, j, seq))
4482 return true;
4483 break;
4484 }
4485 }
4486 while (j > min) {
4487 if (next.match(matcher, i, seq)) {
4488 if (capture) {
4489 groups[groupIndex+1] = i;
4490 groups[groupIndex] = i - k;
4491 }
4492 return true;
4493 }
4494 // backing off
4495 i = i - k;
4496 if (capture) {
4497 groups[groupIndex+1] = i;
4498 groups[groupIndex] = i - k;
4499 }
4500 j--;
4501
4502 }
4503 break;
4504 }
4505 if (capture) {
4506 groups[groupIndex] = save0;
4507 groups[groupIndex+1] = save1;
4508 }
4509 return next.match(matcher, i, seq);
4510 }
4511 // Reluctant matching
4512 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4513 for (;;) {
4514 if (next.match(matcher, i, seq))
4515 return true;
4516 if (j >= cmax)
4517 return false;
4518 if (!atom.match(matcher, i, seq))
4519 return false;
4520 if (i == matcher.last)
4521 return false;
4522 if (capture) {
4523 matcher.groups[groupIndex] = i;
4524 matcher.groups[groupIndex+1] = matcher.last;
4525 }
4526 i = matcher.last;
4527 j++;
4528 }
4529 }
4530 // Possessive matching
4531 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4532 for (; j < cmax; j++) {
4533 if (!atom.match(matcher, i, seq)) {
4534 break;
4535 }
4536 if (capture) {
4537 matcher.groups[groupIndex] = i;
4538 matcher.groups[groupIndex+1] = matcher.last;
4539 }
4540 if (i == matcher.last) {
4541 break;
4542 }
4543 i = matcher.last;
4544 }
4545 return next.match(matcher, i, seq);
4546 }
4547 boolean study(TreeInfo info) {
4548 // Save original info
4549 int minL = info.minLength;
4550 int maxL = info.maxLength;
4551 boolean maxV = info.maxValid;
4552 boolean detm = info.deterministic;
4553 info.reset();
4554
4555 atom.study(info);
4556
4557 int temp = info.minLength * cmin + minL;
4558 if (temp < minL) {
4559 temp = 0xFFFFFFF; // Arbitrary large number
4560 }
4561 info.minLength = temp;
4562
4563 if (maxV & info.maxValid) {
4564 temp = info.maxLength * cmax + maxL;
4565 info.maxLength = temp;
4566 if (temp < maxL) {
4567 info.maxValid = false;
4568 }
4569 } else {
4570 info.maxValid = false;
4571 }
4572
4573 if (info.deterministic && cmin == cmax) {
4574 info.deterministic = detm;
4575 } else {
4576 info.deterministic = false;
4577 }
4578 return next.study(info);
4579 }
4580 }
4581
4582 /**
4583 * A Guard node at the end of each atom node in a Branch. It
4584 * serves the purpose of chaining the "match" operation to
4585 * "next" but not the "study", so we can collect the TreeInfo
4586 * of each atom node without including the TreeInfo of the
4587 * "next".
4588 */
4589 static final class BranchConn extends Node {
4590 BranchConn() {};
4591 boolean match(Matcher matcher, int i, CharSequence seq) {
4592 return next.match(matcher, i, seq);
4593 }
4594 boolean study(TreeInfo info) {
4595 return info.deterministic;
4596 }
4597 }
4598
4599 /**
4600 * Handles the branching of alternations. Note this is also used for
4601 * the ? quantifier to branch between the case where it matches once
4602 * and where it does not occur.
4603 */
4604 static final class Branch extends Node {
4605 Node[] atoms = new Node[2];
4606 int size = 2;
4607 Node conn;
4608 Branch(Node first, Node second, Node branchConn) {
4609 conn = branchConn;
4610 atoms[0] = first;
4611 atoms[1] = second;
4612 }
4613
4614 void add(Node node) {
4615 if (size >= atoms.length) {
4616 Node[] tmp = new Node[atoms.length*2];
4617 System.arraycopy(atoms, 0, tmp, 0, atoms.length);
4618 atoms = tmp;
4619 }
4620 atoms[size++] = node;
4621 }
4622
4623 boolean match(Matcher matcher, int i, CharSequence seq) {
4624 for (int n = 0; n < size; n++) {
4625 if (atoms[n] == null) {
4626 if (conn.next.match(matcher, i, seq))
4627 return true;
4628 } else if (atoms[n].match(matcher, i, seq)) {
4629 return true;
4630 }
4631 }
4632 return false;
4633 }
4634
4635 boolean study(TreeInfo info) {
4636 int minL = info.minLength;
4637 int maxL = info.maxLength;
4638 boolean maxV = info.maxValid;
4639
4640 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num
4641 int maxL2 = -1;
4642 for (int n = 0; n < size; n++) {
4643 info.reset();
4644 if (atoms[n] != null)
4645 atoms[n].study(info);
4646 minL2 = Math.min(minL2, info.minLength);
4647 maxL2 = Math.max(maxL2, info.maxLength);
4648 maxV = (maxV & info.maxValid);
4649 }
4650
4651 minL += minL2;
4652 maxL += maxL2;
4653
4654 info.reset();
4655 conn.next.study(info);
4656
4657 info.minLength += minL;
4658 info.maxLength += maxL;
4659 info.maxValid &= maxV;
4660 info.deterministic = false;
4661 return false;
4662 }
4663 }
4664
4665 /**
4666 * The GroupHead saves the location where the group begins in the locals
4667 * and restores them when the match is done.
4668 *
4669 * The matchRef is used when a reference to this group is accessed later
4670 * in the expression. The locals will have a negative value in them to
4671 * indicate that we do not want to unset the group if the reference
4672 * doesn't match.
4673 */
4674 static final class GroupHead extends Node {
4675 int localIndex;
4676 GroupHead(int localCount) {
4677 localIndex = localCount;
4678 }
4679 boolean match(Matcher matcher, int i, CharSequence seq) {
4680 int save = matcher.locals[localIndex];
4681 matcher.locals[localIndex] = i;
4682 boolean ret = next.match(matcher, i, seq);
4683 matcher.locals[localIndex] = save;
4684 return ret;
4685 }
4686 boolean matchRef(Matcher matcher, int i, CharSequence seq) {
4687 int save = matcher.locals[localIndex];
4688 matcher.locals[localIndex] = ~i; // HACK
4689 boolean ret = next.match(matcher, i, seq);
4690 matcher.locals[localIndex] = save;
4691 return ret;
4692 }
4693 }
4694
4695 /**
4696 * Recursive reference to a group in the regular expression. It calls
4697 * matchRef because if the reference fails to match we would not unset
4698 * the group.
4699 */
4700 static final class GroupRef extends Node {
4701 GroupHead head;
4702 GroupRef(GroupHead head) {
4703 this.head = head;
4704 }
4705 boolean match(Matcher matcher, int i, CharSequence seq) {
4706 return head.matchRef(matcher, i, seq)
4707 && next.match(matcher, matcher.last, seq);
4708 }
4709 boolean study(TreeInfo info) {
4710 info.maxValid = false;
4711 info.deterministic = false;
4712 return next.study(info);
4713 }
4714 }
4715
4716 /**
4717 * The GroupTail handles the setting of group beginning and ending
4718 * locations when groups are successfully matched. It must also be able to
4719 * unset groups that have to be backed off of.
4720 *
4721 * The GroupTail node is also used when a previous group is referenced,
4722 * and in that case no group information needs to be set.
4723 */
4724 static final class GroupTail extends Node {
4725 int localIndex;
4726 int groupIndex;
4727 GroupTail(int localCount, int groupCount) {
4728 localIndex = localCount;
4729 groupIndex = groupCount + groupCount;
4730 }
4731 boolean match(Matcher matcher, int i, CharSequence seq) {
4732 int tmp = matcher.locals[localIndex];
4733 if (tmp >= 0) { // This is the normal group case.
4734 // Save the group so we can unset it if it
4735 // backs off of a match.
4736 int groupStart = matcher.groups[groupIndex];
4737 int groupEnd = matcher.groups[groupIndex+1];
4738
4739 matcher.groups[groupIndex] = tmp;
4740 matcher.groups[groupIndex+1] = i;
4741 if (next.match(matcher, i, seq)) {
4742 return true;
4743 }
4744 matcher.groups[groupIndex] = groupStart;
4745 matcher.groups[groupIndex+1] = groupEnd;
4746 return false;
4747 } else {
4748 // This is a group reference case. We don't need to save any
4749 // group info because it isn't really a group.
4750 matcher.last = i;
4751 return true;
4752 }
4753 }
4754 }
4755
4756 /**
4757 * This sets up a loop to handle a recursive quantifier structure.
4758 */
4759 static final class Prolog extends Node {
4760 Loop loop;
4761 Prolog(Loop loop) {
4762 this.loop = loop;
4763 }
4764 boolean match(Matcher matcher, int i, CharSequence seq) {
4765 return loop.matchInit(matcher, i, seq);
4766 }
4767 boolean study(TreeInfo info) {
4768 return loop.study(info);
4769 }
4770 }
4771
4772 /**
4773 * Handles the repetition count for a greedy Curly. The matchInit
4774 * is called from the Prolog to save the index of where the group
4775 * beginning is stored. A zero length group check occurs in the
4776 * normal match but is skipped in the matchInit.
4777 */
4778 static class Loop extends Node {
4779 Node body;
4780 int countIndex; // local count index in matcher locals
4781 int beginIndex; // group beginning index
4782 int cmin, cmax;
4783 Loop(int countIndex, int beginIndex) {
4784 this.countIndex = countIndex;
4785 this.beginIndex = beginIndex;
4786 }
4787 boolean match(Matcher matcher, int i, CharSequence seq) {
4788 // Avoid infinite loop in zero-length case.
4789 if (i > matcher.locals[beginIndex]) {
4790 int count = matcher.locals[countIndex];
4791
4792 // This block is for before we reach the minimum
4793 // iterations required for the loop to match
4794 if (count < cmin) {
4795 matcher.locals[countIndex] = count + 1;
4796 boolean b = body.match(matcher, i, seq);
4797 // If match failed we must backtrack, so
4798 // the loop count should NOT be incremented
4799 if (!b)
4800 matcher.locals[countIndex] = count;
4801 // Return success or failure since we are under
4802 // minimum
4803 return b;
4804 }
4805 // This block is for after we have the minimum
4806 // iterations required for the loop to match
4807 if (count < cmax) {
4808 matcher.locals[countIndex] = count + 1;
4809 boolean b = body.match(matcher, i, seq);
4810 // If match failed we must backtrack, so
4811 // the loop count should NOT be incremented
4812 if (!b)
4813 matcher.locals[countIndex] = count;
4814 else
4815 return true;
4816 }
4817 }
4818 return next.match(matcher, i, seq);
4819 }
4820 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4821 int save = matcher.locals[countIndex];
4822 boolean ret = false;
4823 if (0 < cmin) {
4824 matcher.locals[countIndex] = 1;
4825 ret = body.match(matcher, i, seq);
4826 } else if (0 < cmax) {
4827 matcher.locals[countIndex] = 1;
4828 ret = body.match(matcher, i, seq);
4829 if (ret == false)
4830 ret = next.match(matcher, i, seq);
4831 } else {
4832 ret = next.match(matcher, i, seq);
4833 }
4834 matcher.locals[countIndex] = save;
4835 return ret;
4836 }
4837 boolean study(TreeInfo info) {
4838 info.maxValid = false;
4839 info.deterministic = false;
4840 return false;
4841 }
4842 }
4843
4844 /**
4845 * Handles the repetition count for a reluctant Curly. The matchInit
4846 * is called from the Prolog to save the index of where the group
4847 * beginning is stored. A zero length group check occurs in the
4848 * normal match but is skipped in the matchInit.
4849 */
4850 static final class LazyLoop extends Loop {
4851 LazyLoop(int countIndex, int beginIndex) {
4852 super(countIndex, beginIndex);
4853 }
4854 boolean match(Matcher matcher, int i, CharSequence seq) {
4855 // Check for zero length group
4856 if (i > matcher.locals[beginIndex]) {
4857 int count = matcher.locals[countIndex];
4858 if (count < cmin) {
4859 matcher.locals[countIndex] = count + 1;
4860 boolean result = body.match(matcher, i, seq);
4861 // If match failed we must backtrack, so
4862 // the loop count should NOT be incremented
4863 if (!result)
4864 matcher.locals[countIndex] = count;
4865 return result;
4866 }
4867 if (next.match(matcher, i, seq))
4868 return true;
4869 if (count < cmax) {
4870 matcher.locals[countIndex] = count + 1;
4871 boolean result = body.match(matcher, i, seq);
4872 // If match failed we must backtrack, so
4873 // the loop count should NOT be incremented
4874 if (!result)
4875 matcher.locals[countIndex] = count;
4876 return result;
4877 }
4878 return false;
4879 }
4880 return next.match(matcher, i, seq);
4881 }
4882 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4883 int save = matcher.locals[countIndex];
4884 boolean ret = false;
4885 if (0 < cmin) {
4886 matcher.locals[countIndex] = 1;
4887 ret = body.match(matcher, i, seq);
4888 } else if (next.match(matcher, i, seq)) {
4889 ret = true;
4890 } else if (0 < cmax) {
4891 matcher.locals[countIndex] = 1;
4892 ret = body.match(matcher, i, seq);
4893 }
4894 matcher.locals[countIndex] = save;
4895 return ret;
4896 }
4897 boolean study(TreeInfo info) {
4898 info.maxValid = false;
4899 info.deterministic = false;
4900 return false;
4901 }
4902 }
4903
4904 /**
4905 * Refers to a group in the regular expression. Attempts to match
4906 * whatever the group referred to last matched.
4907 */
4908 static class BackRef extends Node {
4909 int groupIndex;
4910 BackRef(int groupCount) {
4911 super();
4912 groupIndex = groupCount + groupCount;
4913 }
4914 boolean match(Matcher matcher, int i, CharSequence seq) {
4915 int j = matcher.groups[groupIndex];
4916 int k = matcher.groups[groupIndex+1];
4917
4918 int groupSize = k - j;
4919 // If the referenced group didn't match, neither can this
4920 if (j < 0)
4921 return false;
4922
4923 // If there isn't enough input left no match
4924 if (i + groupSize > matcher.to) {
4925 matcher.hitEnd = true;
4926 return false;
4927 }
4928 // Check each new char to make sure it matches what the group
4929 // referenced matched last time around
4930 for (int index=0; index<groupSize; index++)
4931 if (seq.charAt(i+index) != seq.charAt(j+index))
4932 return false;
4933
4934 return next.match(matcher, i+groupSize, seq);
4935 }
4936 boolean study(TreeInfo info) {
4937 info.maxValid = false;
4938 return next.study(info);
4939 }
4940 }
4941
4942 static class CIBackRef extends Node {
4943 int groupIndex;
4944 boolean doUnicodeCase;
4945 CIBackRef(int groupCount, boolean doUnicodeCase) {
4946 super();
4947 groupIndex = groupCount + groupCount;
4948 this.doUnicodeCase = doUnicodeCase;
4949 }
4950 boolean match(Matcher matcher, int i, CharSequence seq) {
4951 int j = matcher.groups[groupIndex];
4952 int k = matcher.groups[groupIndex+1];
4953
4954 int groupSize = k - j;
4955
4956 // If the referenced group didn't match, neither can this
4957 if (j < 0)
4958 return false;
4959
4960 // If there isn't enough input left no match
4961 if (i + groupSize > matcher.to) {
4962 matcher.hitEnd = true;
4963 return false;
4964 }
4965
4966 // Check each new char to make sure it matches what the group
4967 // referenced matched last time around
4968 int x = i;
4969 for (int index=0; index<groupSize; index++) {
4970 int c1 = Character.codePointAt(seq, x);
4971 int c2 = Character.codePointAt(seq, j);
4972 if (c1 != c2) {
4973 if (doUnicodeCase) {
4974 int cc1 = Character.toUpperCase(c1);
4975 int cc2 = Character.toUpperCase(c2);
4976 if (cc1 != cc2 &&
4977 Character.toLowerCase(cc1) !=
4978 Character.toLowerCase(cc2))
4979 return false;
4980 } else {
4981 if (ASCII.toLower(c1) != ASCII.toLower(c2))
4982 return false;
4983 }
4984 }
4985 x += Character.charCount(c1);
4986 j += Character.charCount(c2);
4987 }
4988
4989 return next.match(matcher, i+groupSize, seq);
4990 }
4991 boolean study(TreeInfo info) {
4992 info.maxValid = false;
4993 return next.study(info);
4994 }
4995 }
4996
4997 /**
4998 * Searches until the next instance of its atom. This is useful for
4999 * finding the atom efficiently without passing an instance of it
5000 * (greedy problem) and without a lot of wasted search time (reluctant
5001 * problem).
5002 */
5003 static final class First extends Node {
5004 Node atom;
5005 First(Node node) {
5006 this.atom = BnM.optimize(node);
5007 }
5008 boolean match(Matcher matcher, int i, CharSequence seq) {
5009 if (atom instanceof BnM) {
5010 return atom.match(matcher, i, seq)
5011 && next.match(matcher, matcher.last, seq);
5012 }
5013 for (;;) {
5014 if (i > matcher.to) {
5015 matcher.hitEnd = true;
5016 return false;
5017 }
5018 if (atom.match(matcher, i, seq)) {
5019 return next.match(matcher, matcher.last, seq);
5020 }
5021 i += countChars(seq, i, 1);
5022 matcher.first++;
5023 }
5024 }
5025 boolean study(TreeInfo info) {
5026 atom.study(info);
5027 info.maxValid = false;
5028 info.deterministic = false;
5029 return next.study(info);
5030 }
5031 }
5032
5033 static final class Conditional extends Node {
5034 Node cond, yes, not;
5035 Conditional(Node cond, Node yes, Node not) {
5036 this.cond = cond;
5037 this.yes = yes;
5038 this.not = not;
5039 }
5040 boolean match(Matcher matcher, int i, CharSequence seq) {
5041 if (cond.match(matcher, i, seq)) {
5042 return yes.match(matcher, i, seq);
5043 } else {
5044 return not.match(matcher, i, seq);
5045 }
5046 }
5047 boolean study(TreeInfo info) {
5048 int minL = info.minLength;
5049 int maxL = info.maxLength;
5050 boolean maxV = info.maxValid;
5051 info.reset();
5052 yes.study(info);
5053
5054 int minL2 = info.minLength;
5055 int maxL2 = info.maxLength;
5056 boolean maxV2 = info.maxValid;
5057 info.reset();
5058 not.study(info);
5059
5060 info.minLength = minL + Math.min(minL2, info.minLength);
5061 info.maxLength = maxL + Math.max(maxL2, info.maxLength);
5062 info.maxValid = (maxV & maxV2 & info.maxValid);
5063 info.deterministic = false;
5064 return next.study(info);
5065 }
5066 }
5067
5068 /**
5069 * Zero width positive lookahead.
5070 */
5071 static final class Pos extends Node {
5072 Node cond;
5073 Pos(Node cond) {
5074 this.cond = cond;
5075 }
5076 boolean match(Matcher matcher, int i, CharSequence seq) {
5077 int savedTo = matcher.to;
5078 boolean conditionMatched = false;
5079
5080 // Relax transparent region boundaries for lookahead
5081 if (matcher.transparentBounds)
5082 matcher.to = matcher.getTextLength();
5083 try {
5084 conditionMatched = cond.match(matcher, i, seq);
5085 } finally {
5086 // Reinstate region boundaries
5087 matcher.to = savedTo;
5088 }
5089 return conditionMatched && next.match(matcher, i, seq);
5090 }
5091 }
5092
5093 /**
5094 * Zero width negative lookahead.
5095 */
5096 static final class Neg extends Node {
5097 Node cond;
5098 Neg(Node cond) {
5099 this.cond = cond;
5100 }
5101 boolean match(Matcher matcher, int i, CharSequence seq) {
5102 int savedTo = matcher.to;
5103 boolean conditionMatched = false;
5104
5105 // Relax transparent region boundaries for lookahead
5106 if (matcher.transparentBounds)
5107 matcher.to = matcher.getTextLength();
5108 try {
5109 if (i < matcher.to) {
5110 conditionMatched = !cond.match(matcher, i, seq);
5111 } else {
5112 // If a negative lookahead succeeds then more input
5113 // could cause it to fail!
5114 matcher.requireEnd = true;
5115 conditionMatched = !cond.match(matcher, i, seq);
5116 }
5117 } finally {
5118 // Reinstate region boundaries
5119 matcher.to = savedTo;
5120 }
5121 return conditionMatched && next.match(matcher, i, seq);
5122 }
5123 }
5124
5125 /**
5126 * For use with lookbehinds; matches the position where the lookbehind
5127 * was encountered.
5128 */
5129 static Node lookbehindEnd = new Node() {
5130 boolean match(Matcher matcher, int i, CharSequence seq) {
5131 return i == matcher.lookbehindTo;
5132 }
5133 };
5134
5135 /**
5136 * Zero width positive lookbehind.
5137 */
5138 static class Behind extends Node {
5139 Node cond;
5140 int rmax, rmin;
5141 Behind(Node cond, int rmax, int rmin) {
5142 this.cond = cond;
5143 this.rmax = rmax;
5144 this.rmin = rmin;
5145 }
5146
5147 boolean match(Matcher matcher, int i, CharSequence seq) {
5148 int savedFrom = matcher.from;
5149 boolean conditionMatched = false;
5150 int startIndex = (!matcher.transparentBounds) ?
5151 matcher.from : 0;
5152 int from = Math.max(i - rmax, startIndex);
5153 // Set end boundary
5154 int savedLBT = matcher.lookbehindTo;
5155 matcher.lookbehindTo = i;
5156 // Relax transparent region boundaries for lookbehind
5157 if (matcher.transparentBounds)
5158 matcher.from = 0;
5159 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
5160 conditionMatched = cond.match(matcher, j, seq);
5161 }
5162 matcher.from = savedFrom;
5163 matcher.lookbehindTo = savedLBT;
5164 return conditionMatched && next.match(matcher, i, seq);
5165 }
5166 }
5167
5168 /**
5169 * Zero width positive lookbehind, including supplementary
5170 * characters or unpaired surrogates.
5171 */
5172 static final class BehindS extends Behind {
5173 BehindS(Node cond, int rmax, int rmin) {
5174 super(cond, rmax, rmin);
5175 }
5176 boolean match(Matcher matcher, int i, CharSequence seq) {
5177 int rmaxChars = countChars(seq, i, -rmax);
5178 int rminChars = countChars(seq, i, -rmin);
5179 int savedFrom = matcher.from;
5180 int startIndex = (!matcher.transparentBounds) ?
5181 matcher.from : 0;
5182 boolean conditionMatched = false;
5183 int from = Math.max(i - rmaxChars, startIndex);
5184 // Set end boundary
5185 int savedLBT = matcher.lookbehindTo;
5186 matcher.lookbehindTo = i;
5187 // Relax transparent region boundaries for lookbehind
5188 if (matcher.transparentBounds)
5189 matcher.from = 0;
5190
5191 for (int j = i - rminChars;
5192 !conditionMatched && j >= from;
5193 j -= j>from ? countChars(seq, j, -1) : 1) {
5194 conditionMatched = cond.match(matcher, j, seq);
5195 }
5196 matcher.from = savedFrom;
5197 matcher.lookbehindTo = savedLBT;
5198 return conditionMatched && next.match(matcher, i, seq);
5199 }
5200 }
5201
5202 /**
5203 * Zero width negative lookbehind.
5204 */
5205 static class NotBehind extends Node {
5206 Node cond;
5207 int rmax, rmin;
5208 NotBehind(Node cond, int rmax, int rmin) {
5209 this.cond = cond;
5210 this.rmax = rmax;
5211 this.rmin = rmin;
5212 }
5213
5214 boolean match(Matcher matcher, int i, CharSequence seq) {
5215 int savedLBT = matcher.lookbehindTo;
5216 int savedFrom = matcher.from;
5217 boolean conditionMatched = false;
5218 int startIndex = (!matcher.transparentBounds) ?
5219 matcher.from : 0;
5220 int from = Math.max(i - rmax, startIndex);
5221 matcher.lookbehindTo = i;
5222 // Relax transparent region boundaries for lookbehind
5223 if (matcher.transparentBounds)
5224 matcher.from = 0;
5225 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
5226 conditionMatched = cond.match(matcher, j, seq);
5227 }
5228 // Reinstate region boundaries
5229 matcher.from = savedFrom;
5230 matcher.lookbehindTo = savedLBT;
5231 return !conditionMatched && next.match(matcher, i, seq);
5232 }
5233 }
5234
5235 /**
5236 * Zero width negative lookbehind, including supplementary
5237 * characters or unpaired surrogates.
5238 */
5239 static final class NotBehindS extends NotBehind {
5240 NotBehindS(Node cond, int rmax, int rmin) {
5241 super(cond, rmax, rmin);
5242 }
5243 boolean match(Matcher matcher, int i, CharSequence seq) {
5244 int rmaxChars = countChars(seq, i, -rmax);
5245 int rminChars = countChars(seq, i, -rmin);
5246 int savedFrom = matcher.from;
5247 int savedLBT = matcher.lookbehindTo;
5248 boolean conditionMatched = false;
5249 int startIndex = (!matcher.transparentBounds) ?
5250 matcher.from : 0;
5251 int from = Math.max(i - rmaxChars, startIndex);
5252 matcher.lookbehindTo = i;
5253 // Relax transparent region boundaries for lookbehind
5254 if (matcher.transparentBounds)
5255 matcher.from = 0;
5256 for (int j = i - rminChars;
5257 !conditionMatched && j >= from;
5258 j -= j>from ? countChars(seq, j, -1) : 1) {
5259 conditionMatched = cond.match(matcher, j, seq);
5260 }
5261 //Reinstate region boundaries
5262 matcher.from = savedFrom;
5263 matcher.lookbehindTo = savedLBT;
5264 return !conditionMatched && next.match(matcher, i, seq);
5265 }
5266 }
5267
5268 /**
5269 * Returns the set union of two CharProperty nodes.
5270 */
5271 private static CharProperty union(final CharProperty lhs,
5272 final CharProperty rhs) {
5273 return new CharProperty() {
5274 boolean isSatisfiedBy(int ch) {
5275 return lhs.isSatisfiedBy(ch) || rhs.isSatisfiedBy(ch);}};
5276 }
5277
5278 /**
5279 * Returns the set intersection of two CharProperty nodes.
5280 */
5281 private static CharProperty intersection(final CharProperty lhs,
5282 final CharProperty rhs) {
5283 return new CharProperty() {
5284 boolean isSatisfiedBy(int ch) {
5285 return lhs.isSatisfiedBy(ch) && rhs.isSatisfiedBy(ch);}};
5286 }
5287
5288 /**
5289 * Returns the set difference of two CharProperty nodes.
5290 */
5291 private static CharProperty setDifference(final CharProperty lhs,
5292 final CharProperty rhs) {
5293 return new CharProperty() {
5294 boolean isSatisfiedBy(int ch) {
5295 return ! rhs.isSatisfiedBy(ch) && lhs.isSatisfiedBy(ch);}};
5296 }
5297
5298 /**
5299 * Handles word boundaries. Includes a field to allow this one class to
5300 * deal with the different types of word boundaries we can match. The word
5301 * characters include underscores, letters, and digits. Non spacing marks
5302 * can are also part of a word if they have a base character, otherwise
5303 * they are ignored for purposes of finding word boundaries.
5304 */
5305 static final class Bound extends Node {
5306 static int LEFT = 0x1;
5307 static int RIGHT= 0x2;
5308 static int BOTH = 0x3;
5309 static int NONE = 0x4;
5310 int type;
5311 boolean useUWORD;
5312 Bound(int n, boolean useUWORD) {
5313 type = n;
5314 this.useUWORD = useUWORD;
5315 }
5316
5317 boolean isWord(int ch) {
5318 return useUWORD ? UnicodeProp.WORD.is(ch)
5319 : (ch == '_' || Character.isLetterOrDigit(ch));
5320 }
5321
5322 int check(Matcher matcher, int i, CharSequence seq) {
5323 int ch;
5324 boolean left = false;
5325 int startIndex = matcher.from;
5326 int endIndex = matcher.to;
5327 if (matcher.transparentBounds) {
5328 startIndex = 0;
5329 endIndex = matcher.getTextLength();
5330 }
5331 if (i > startIndex) {
5332 ch = Character.codePointBefore(seq, i);
5333 left = (isWord(ch) ||
5334 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5335 && hasBaseCharacter(matcher, i-1, seq)));
5336 }
5337 boolean right = false;
5338 if (i < endIndex) {
5339 ch = Character.codePointAt(seq, i);
5340 right = (isWord(ch) ||
5341 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5342 && hasBaseCharacter(matcher, i, seq)));
5343 } else {
5344 // Tried to access char past the end
5345 matcher.hitEnd = true;
5346 // The addition of another char could wreck a boundary
5347 matcher.requireEnd = true;
5348 }
5349 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE);
5350 }
5351 boolean match(Matcher matcher, int i, CharSequence seq) {
5352 return (check(matcher, i, seq) & type) > 0
5353 && next.match(matcher, i, seq);
5354 }
5355 }
5356
5357 /**
5358 * Non spacing marks only count as word characters in bounds calculations
5359 * if they have a base character.
5360 */
5361 private static boolean hasBaseCharacter(Matcher matcher, int i,
5362 CharSequence seq)
5363 {
5364 int start = (!matcher.transparentBounds) ?
5365 matcher.from : 0;
5366 for (int x=i; x >= start; x--) {
5367 int ch = Character.codePointAt(seq, x);
5368 if (Character.isLetterOrDigit(ch))
5369 return true;
5370 if (Character.getType(ch) == Character.NON_SPACING_MARK)
5371 continue;
5372 return false;
5373 }
5374 return false;
5375 }
5376
5377 /**
5378 * Attempts to match a slice in the input using the Boyer-Moore string
5379 * matching algorithm. The algorithm is based on the idea that the
5380 * pattern can be shifted farther ahead in the search text if it is
5381 * matched right to left.
5382 * <p>
5383 * The pattern is compared to the input one character at a time, from
5384 * the rightmost character in the pattern to the left. If the characters
5385 * all match the pattern has been found. If a character does not match,
5386 * the pattern is shifted right a distance that is the maximum of two
5387 * functions, the bad character shift and the good suffix shift. This
5388 * shift moves the attempted match position through the input more
5389 * quickly than a naive one position at a time check.
5390 * <p>
5391 * The bad character shift is based on the character from the text that
5392 * did not match. If the character does not appear in the pattern, the
5393 * pattern can be shifted completely beyond the bad character. If the
5394 * character does occur in the pattern, the pattern can be shifted to
5395 * line the pattern up with the next occurrence of that character.
5396 * <p>
5397 * The good suffix shift is based on the idea that some subset on the right
5398 * side of the pattern has matched. When a bad character is found, the
5399 * pattern can be shifted right by the pattern length if the subset does
5400 * not occur again in pattern, or by the amount of distance to the
5401 * next occurrence of the subset in the pattern.
5402 *
5403 * Boyer-Moore search methods adapted from code by Amy Yu.
5404 */
5405 static class BnM extends Node {
5406 int[] buffer;
5407 int[] lastOcc;
5408 int[] optoSft;
5409
5410 /**
5411 * Pre calculates arrays needed to generate the bad character
5412 * shift and the good suffix shift. Only the last seven bits
5413 * are used to see if chars match; This keeps the tables small
5414 * and covers the heavily used ASCII range, but occasionally
5415 * results in an aliased match for the bad character shift.
5416 */
5417 static Node optimize(Node node) {
5418 if (!(node instanceof Slice)) {
5419 return node;
5420 }
5421
5422 int[] src = ((Slice) node).buffer;
5423 int patternLength = src.length;
5424 // The BM algorithm requires a bit of overhead;
5425 // If the pattern is short don't use it, since
5426 // a shift larger than the pattern length cannot
5427 // be used anyway.
5428 if (patternLength < 4) {
5429 return node;
5430 }
5431 int i, j, k;
5432 int[] lastOcc = new int[128];
5433 int[] optoSft = new int[patternLength];
5434 // Precalculate part of the bad character shift
5435 // It is a table for where in the pattern each
5436 // lower 7-bit value occurs
5437 for (i = 0; i < patternLength; i++) {
5438 lastOcc[src[i]&0x7F] = i + 1;
5439 }
5440 // Precalculate the good suffix shift
5441 // i is the shift amount being considered
5442 NEXT: for (i = patternLength; i > 0; i--) {
5443 // j is the beginning index of suffix being considered
5444 for (j = patternLength - 1; j >= i; j--) {
5445 // Testing for good suffix
5446 if (src[j] == src[j-i]) {
5447 // src[j..len] is a good suffix
5448 optoSft[j-1] = i;
5449 } else {
5450 // No match. The array has already been
5451 // filled up with correct values before.
5452 continue NEXT;
5453 }
5454 }
5455 // This fills up the remaining of optoSft
5456 // any suffix can not have larger shift amount
5457 // then its sub-suffix. Why???
5458 while (j > 0) {
5459 optoSft[--j] = i;
5460 }
5461 }
5462 // Set the guard value because of unicode compression
5463 optoSft[patternLength-1] = 1;
5464 if (node instanceof SliceS)
5465 return new BnMS(src, lastOcc, optoSft, node.next);
5466 return new BnM(src, lastOcc, optoSft, node.next);
5467 }
5468 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5469 this.buffer = src;
5470 this.lastOcc = lastOcc;
5471 this.optoSft = optoSft;
5472 this.next = next;
5473 }
5474 boolean match(Matcher matcher, int i, CharSequence seq) {
5475 int[] src = buffer;
5476 int patternLength = src.length;
5477 int last = matcher.to - patternLength;
5478
5479 // Loop over all possible match positions in text
5480 NEXT: while (i <= last) {
5481 // Loop over pattern from right to left
5482 for (int j = patternLength - 1; j >= 0; j--) {
5483 int ch = seq.charAt(i+j);
5484 if (ch != src[j]) {
5485 // Shift search to the right by the maximum of the
5486 // bad character shift and the good suffix shift
5487 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]);
5488 continue NEXT;
5489 }
5490 }
5491 // Entire pattern matched starting at i
5492 matcher.first = i;
5493 boolean ret = next.match(matcher, i + patternLength, seq);
5494 if (ret) {
5495 matcher.first = i;
5496 matcher.groups[0] = matcher.first;
5497 matcher.groups[1] = matcher.last;
5498 return true;
5499 }
5500 i++;
5501 }
5502 // BnM is only used as the leading node in the unanchored case,
5503 // and it replaced its Start() which always searches to the end
5504 // if it doesn't find what it's looking for, so hitEnd is true.
5505 matcher.hitEnd = true;
5506 return false;
5507 }
5508 boolean study(TreeInfo info) {
5509 info.minLength += buffer.length;
5510 info.maxValid = false;
5511 return next.study(info);
5512 }
5513 }
5514
5515 /**
5516 * Supplementary support version of BnM(). Unpaired surrogates are
5517 * also handled by this class.
5518 */
5519 static final class BnMS extends BnM {
5520 int lengthInChars;
5521
5522 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5523 super(src, lastOcc, optoSft, next);
5524 for (int cp : buffer) {
5525 lengthInChars += Character.charCount(cp);
5526 }
5527 }
5528 boolean match(Matcher matcher, int i, CharSequence seq) {
5529 int[] src = buffer;
5530 int patternLength = src.length;
5531 int last = matcher.to - lengthInChars;
5532
5533 // Loop over all possible match positions in text
5534 NEXT: while (i <= last) {
5535 // Loop over pattern from right to left
5536 int ch;
5537 for (int j = countChars(seq, i, patternLength), x = patternLength - 1;
5538 j > 0; j -= Character.charCount(ch), x--) {
5539 ch = Character.codePointBefore(seq, i+j);
5540 if (ch != src[x]) {
5541 // Shift search to the right by the maximum of the
5542 // bad character shift and the good suffix shift
5543 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]);
5544 i += countChars(seq, i, n);
5545 continue NEXT;
5546 }
5547 }
5548 // Entire pattern matched starting at i
5549 matcher.first = i;
5550 boolean ret = next.match(matcher, i + lengthInChars, seq);
5551 if (ret) {
5552 matcher.first = i;
5553 matcher.groups[0] = matcher.first;
5554 matcher.groups[1] = matcher.last;
5555 return true;
5556 }
5557 i += countChars(seq, i, 1);
5558 }
5559 matcher.hitEnd = true;
5560 return false;
5561 }
5562 }
5563
5564 ///////////////////////////////////////////////////////////////////////////////
5565 ///////////////////////////////////////////////////////////////////////////////
5566
5567 /**
5568 * This must be the very first initializer.
5569 */
5570 static Node accept = new Node();
5571
5572 static Node lastAccept = new LastNode();
5573
5574 private static class CharPropertyNames {
5575
5576 static CharProperty charPropertyFor(String name) {
5577 CharPropertyFactory m = map.get(name);
5578 return m == null ? null : m.make();
5579 }
5580
5581 private static abstract class CharPropertyFactory {
5582 abstract CharProperty make();
5583 }
5584
5585 private static void defCategory(String name,
5586 final int typeMask) {
5587 map.put(name, new CharPropertyFactory() {
5588 CharProperty make() { return new Category(typeMask);}});
5589 }
5590
5591 private static void defRange(String name,
5592 final int lower, final int upper) {
5593 map.put(name, new CharPropertyFactory() {
5594 CharProperty make() { return rangeFor(lower, upper);}});
5595 }
5596
5597 private static void defCtype(String name,
5598 final int ctype) {
5599 map.put(name, new CharPropertyFactory() {
5600 CharProperty make() { return new Ctype(ctype);}});
5601 }
5602
5603 private static abstract class CloneableProperty
5604 extends CharProperty implements Cloneable
5605 {
5606 public CloneableProperty clone() {
5607 try {
5608 return (CloneableProperty) super.clone();
5609 } catch (CloneNotSupportedException e) {
5610 throw new AssertionError(e);
5611 }
5612 }
5613 }
5614
5615 private static void defClone(String name,
5616 final CloneableProperty p) {
5617 map.put(name, new CharPropertyFactory() {
5618 CharProperty make() { return p.clone();}});
5619 }
5620
5621 private static final HashMap<String, CharPropertyFactory> map
5622 = new HashMap<>();
5623
5624 static {
5625 // Unicode character property aliases, defined in
5626 // http://www.unicode.org/Public/UNIDATA/PropertyValueAliases.txt
5627 defCategory("Cn", 1<<Character.UNASSIGNED);
5628 defCategory("Lu", 1<<Character.UPPERCASE_LETTER);
5629 defCategory("Ll", 1<<Character.LOWERCASE_LETTER);
5630 defCategory("Lt", 1<<Character.TITLECASE_LETTER);
5631 defCategory("Lm", 1<<Character.MODIFIER_LETTER);
5632 defCategory("Lo", 1<<Character.OTHER_LETTER);
5633 defCategory("Mn", 1<<Character.NON_SPACING_MARK);
5634 defCategory("Me", 1<<Character.ENCLOSING_MARK);
5635 defCategory("Mc", 1<<Character.COMBINING_SPACING_MARK);
5636 defCategory("Nd", 1<<Character.DECIMAL_DIGIT_NUMBER);
5637 defCategory("Nl", 1<<Character.LETTER_NUMBER);
5638 defCategory("No", 1<<Character.OTHER_NUMBER);
5639 defCategory("Zs", 1<<Character.SPACE_SEPARATOR);
5640 defCategory("Zl", 1<<Character.LINE_SEPARATOR);
5641 defCategory("Zp", 1<<Character.PARAGRAPH_SEPARATOR);
5642 defCategory("Cc", 1<<Character.CONTROL);
5643 defCategory("Cf", 1<<Character.FORMAT);
5644 defCategory("Co", 1<<Character.PRIVATE_USE);
5645 defCategory("Cs", 1<<Character.SURROGATE);
5646 defCategory("Pd", 1<<Character.DASH_PUNCTUATION);
5647 defCategory("Ps", 1<<Character.START_PUNCTUATION);
5648 defCategory("Pe", 1<<Character.END_PUNCTUATION);
5649 defCategory("Pc", 1<<Character.CONNECTOR_PUNCTUATION);
5650 defCategory("Po", 1<<Character.OTHER_PUNCTUATION);
5651 defCategory("Sm", 1<<Character.MATH_SYMBOL);
5652 defCategory("Sc", 1<<Character.CURRENCY_SYMBOL);
5653 defCategory("Sk", 1<<Character.MODIFIER_SYMBOL);
5654 defCategory("So", 1<<Character.OTHER_SYMBOL);
5655 defCategory("Pi", 1<<Character.INITIAL_QUOTE_PUNCTUATION);
5656 defCategory("Pf", 1<<Character.FINAL_QUOTE_PUNCTUATION);
5657 defCategory("L", ((1<<Character.UPPERCASE_LETTER) |
5658 (1<<Character.LOWERCASE_LETTER) |
5659 (1<<Character.TITLECASE_LETTER) |
5660 (1<<Character.MODIFIER_LETTER) |
5661 (1<<Character.OTHER_LETTER)));
5662 defCategory("M", ((1<<Character.NON_SPACING_MARK) |
5663 (1<<Character.ENCLOSING_MARK) |
5664 (1<<Character.COMBINING_SPACING_MARK)));
5665 defCategory("N", ((1<<Character.DECIMAL_DIGIT_NUMBER) |
5666 (1<<Character.LETTER_NUMBER) |
5667 (1<<Character.OTHER_NUMBER)));
5668 defCategory("Z", ((1<<Character.SPACE_SEPARATOR) |
5669 (1<<Character.LINE_SEPARATOR) |
5670 (1<<Character.PARAGRAPH_SEPARATOR)));
5671 defCategory("C", ((1<<Character.CONTROL) |
5672 (1<<Character.FORMAT) |
5673 (1<<Character.PRIVATE_USE) |
5674 (1<<Character.SURROGATE))); // Other
5675 defCategory("P", ((1<<Character.DASH_PUNCTUATION) |
5676 (1<<Character.START_PUNCTUATION) |
5677 (1<<Character.END_PUNCTUATION) |
5678 (1<<Character.CONNECTOR_PUNCTUATION) |
5679 (1<<Character.OTHER_PUNCTUATION) |
5680 (1<<Character.INITIAL_QUOTE_PUNCTUATION) |
5681 (1<<Character.FINAL_QUOTE_PUNCTUATION)));
5682 defCategory("S", ((1<<Character.MATH_SYMBOL) |
5683 (1<<Character.CURRENCY_SYMBOL) |
5684 (1<<Character.MODIFIER_SYMBOL) |
5685 (1<<Character.OTHER_SYMBOL)));
5686 defCategory("LC", ((1<<Character.UPPERCASE_LETTER) |
5687 (1<<Character.LOWERCASE_LETTER) |
5688 (1<<Character.TITLECASE_LETTER)));
5689 defCategory("LD", ((1<<Character.UPPERCASE_LETTER) |
5690 (1<<Character.LOWERCASE_LETTER) |
5691 (1<<Character.TITLECASE_LETTER) |
5692 (1<<Character.MODIFIER_LETTER) |
5693 (1<<Character.OTHER_LETTER) |
5694 (1<<Character.DECIMAL_DIGIT_NUMBER)));
5695 defRange("L1", 0x00, 0xFF); // Latin-1
5696 map.put("all", new CharPropertyFactory() {
5697 CharProperty make() { return new All(); }});
5698
5699 // Posix regular expression character classes, defined in
5700 // http://www.unix.org/onlinepubs/009695399/basedefs/xbd_chap09.html
5701 defRange("ASCII", 0x00, 0x7F); // ASCII
5702 defCtype("Alnum", ASCII.ALNUM); // Alphanumeric characters
5703 defCtype("Alpha", ASCII.ALPHA); // Alphabetic characters
5704 defCtype("Blank", ASCII.BLANK); // Space and tab characters
5705 defCtype("Cntrl", ASCII.CNTRL); // Control characters
5706 defRange("Digit", '0', '9'); // Numeric characters
5707 defCtype("Graph", ASCII.GRAPH); // printable and visible
5708 defRange("Lower", 'a', 'z'); // Lower-case alphabetic
5709 defRange("Print", 0x20, 0x7E); // Printable characters
5710 defCtype("Punct", ASCII.PUNCT); // Punctuation characters
5711 defCtype("Space", ASCII.SPACE); // Space characters
5712 defRange("Upper", 'A', 'Z'); // Upper-case alphabetic
5713 defCtype("XDigit",ASCII.XDIGIT); // hexadecimal digits
5714
5715 // Java character properties, defined by methods in Character.java
5716 defClone("javaLowerCase", new CloneableProperty() {
5717 boolean isSatisfiedBy(int ch) {
5718 return Character.isLowerCase(ch);}});
5719 defClone("javaUpperCase", new CloneableProperty() {
5720 boolean isSatisfiedBy(int ch) {
5721 return Character.isUpperCase(ch);}});
5722 defClone("javaAlphabetic", new CloneableProperty() {
5723 boolean isSatisfiedBy(int ch) {
5724 return Character.isAlphabetic(ch);}});
5725 defClone("javaIdeographic", new CloneableProperty() {
5726 boolean isSatisfiedBy(int ch) {
5727 return Character.isIdeographic(ch);}});
5728 defClone("javaTitleCase", new CloneableProperty() {
5729 boolean isSatisfiedBy(int ch) {
5730 return Character.isTitleCase(ch);}});
5731 defClone("javaDigit", new CloneableProperty() {
5732 boolean isSatisfiedBy(int ch) {
5733 return Character.isDigit(ch);}});
5734 defClone("javaDefined", new CloneableProperty() {
5735 boolean isSatisfiedBy(int ch) {
5736 return Character.isDefined(ch);}});
5737 defClone("javaLetter", new CloneableProperty() {
5738 boolean isSatisfiedBy(int ch) {
5739 return Character.isLetter(ch);}});
5740 defClone("javaLetterOrDigit", new CloneableProperty() {
5741 boolean isSatisfiedBy(int ch) {
5742 return Character.isLetterOrDigit(ch);}});
5743 defClone("javaJavaIdentifierStart", new CloneableProperty() {
5744 boolean isSatisfiedBy(int ch) {
5745 return Character.isJavaIdentifierStart(ch);}});
5746 defClone("javaJavaIdentifierPart", new CloneableProperty() {
5747 boolean isSatisfiedBy(int ch) {
5748 return Character.isJavaIdentifierPart(ch);}});
5749 defClone("javaUnicodeIdentifierStart", new CloneableProperty() {
5750 boolean isSatisfiedBy(int ch) {
5751 return Character.isUnicodeIdentifierStart(ch);}});
5752 defClone("javaUnicodeIdentifierPart", new CloneableProperty() {
5753 boolean isSatisfiedBy(int ch) {
5754 return Character.isUnicodeIdentifierPart(ch);}});
5755 defClone("javaIdentifierIgnorable", new CloneableProperty() {
5756 boolean isSatisfiedBy(int ch) {
5757 return Character.isIdentifierIgnorable(ch);}});
5758 defClone("javaSpaceChar", new CloneableProperty() {
5759 boolean isSatisfiedBy(int ch) {
5760 return Character.isSpaceChar(ch);}});
5761 defClone("javaWhitespace", new CloneableProperty() {
5762 boolean isSatisfiedBy(int ch) {
5763 return Character.isWhitespace(ch);}});
5764 defClone("javaISOControl", new CloneableProperty() {
5765 boolean isSatisfiedBy(int ch) {
5766 return Character.isISOControl(ch);}});
5767 defClone("javaMirrored", new CloneableProperty() {
5768 boolean isSatisfiedBy(int ch) {
5769 return Character.isMirrored(ch);}});
5770 }
5771 }
5772
5773 /**
5774 * Creates a predicate which can be used to match a string.
5775 *
5776 * @return The predicate which can be used for matching on a string
5777 * @since 1.8
5778 */
5779 public Predicate<String> asPredicate() {
5780 return s -> matcher(s).find();
5781 }
5782
5783 /**
5784 * Creates a stream from the given input sequence around matches of this
5785 * pattern.
5786 *
5787 * <p> The stream returned by this method contains each substring of the
5788 * input sequence that is terminated by another subsequence that matches
5789 * this pattern or is terminated by the end of the input sequence. The
5790 * substrings in the stream are in the order in which they occur in the
5791 * input. Trailing empty strings will be discarded and not encountered in
5792 * the stream.
5793 *
5794 * <p> If this pattern does not match any subsequence of the input then
5795 * the resulting stream has just one element, namely the input sequence in
5796 * string form.
5797 *
5798 * <p> When there is a positive-width match at the beginning of the input
5799 * sequence then an empty leading substring is included at the beginning
5800 * of the stream. A zero-width match at the beginning however never produces
5801 * such empty leading substring.
5802 *
5803 * <p> If the input sequence is mutable, it must remain constant during the
5804 * execution of the terminal stream operation. Otherwise, the result of the
5805 * terminal stream operation is undefined.
5806 *
5807 * @param input
5808 * The character sequence to be split
5809 *
5810 * @return The stream of strings computed by splitting the input
5811 * around matches of this pattern
5812 * @see #split(CharSequence)
5813 * @since 1.8
5814 */
5815 public Stream<String> splitAsStream(final CharSequence input) {
5816 class MatcherIterator implements Iterator<String> {
5817 private final Matcher matcher;
5818 // The start position of the next sub-sequence of input
5819 // when current == input.length there are no more elements
5820 private int current;
5821 // null if the next element, if any, needs to obtained
5822 private String nextElement;
5823 // > 0 if there are N next empty elements
5824 private int emptyElementCount;
5825
5826 MatcherIterator() {
5827 this.matcher = matcher(input);
5828 // If the input is an empty string then the result can only be a
5829 // stream of the input. Induce that by setting the empty
5830 // element count to 1
5831 this.emptyElementCount = input.length() == 0 ? 1 : 0;
5832 }
5833
5834 public String next() {
5835 if (!hasNext())
5836 throw new NoSuchElementException();
5837
5838 if (emptyElementCount == 0) {
5839 String n = nextElement;
5840 nextElement = null;
5841 return n;
5842 } else {
5843 emptyElementCount--;
5844 return "";
5845 }
5846 }
5847
5848 public boolean hasNext() {
5849 if (nextElement != null || emptyElementCount > 0)
5850 return true;
5851
5852 if (current == input.length())
5853 return false;
5854
5855 // Consume the next matching element
5856 // Count sequence of matching empty elements
5857 while (matcher.find()) {
5858 nextElement = input.subSequence(current, matcher.start()).toString();
5859 current = matcher.end();
5860 if (!nextElement.isEmpty()) {
5861 return true;
5862 } else if (current > 0) { // no empty leading substring for zero-width
5863 // match at the beginning of the input
5864 emptyElementCount++;
5865 }
5866 }
5867
5868 // Consume last matching element
5869 nextElement = input.subSequence(current, input.length()).toString();
5870 current = input.length();
5871 if (!nextElement.isEmpty()) {
5872 return true;
5873 } else {
5874 // Ignore a terminal sequence of matching empty elements
5875 emptyElementCount = 0;
5876 nextElement = null;
5877 return false;
5878 }
5879 }
5880 }
5881 return StreamSupport.stream(Spliterators.spliteratorUnknownSize(
5882 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false);
5883 }
5884 }