1 /*
2 * Copyright (c) 1999, 2009, 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.security.AccessController;
29 import java.security.PrivilegedAction;
30 import java.text.CharacterIterator;
31 import java.text.Normalizer;
32 import java.util.Locale;
33 import java.util.Map;
34 import java.util.ArrayList;
35 import java.util.HashMap;
36 import java.util.Arrays;
37
38
39 /**
40 * A compiled representation of a regular expression.
41 *
42 * <p> A regular expression, specified as a string, must first be compiled into
43 * an instance of this class. The resulting pattern can then be used to create
44 * a {@link Matcher} object that can match arbitrary {@link
45 * java.lang.CharSequence </code>character sequences<code>} against the regular
46 * expression. All of the state involved in performing a match resides in the
47 * matcher, so many matchers can share the same pattern.
48 *
49 * <p> A typical invocation sequence is thus
50 *
51 * <blockquote><pre>
52 * Pattern p = Pattern.{@link #compile compile}("a*b");
53 * Matcher m = p.{@link #matcher matcher}("aaaaab");
54 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote>
55 *
56 * <p> A {@link #matches matches} method is defined by this class as a
57 * convenience for when a regular expression is used just once. This method
58 * compiles an expression and matches an input sequence against it in a single
59 * invocation. The statement
60 *
61 * <blockquote><pre>
62 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote>
63 *
64 * is equivalent to the three statements above, though for repeated matches it
65 * is less efficient since it does not allow the compiled pattern to be reused.
66 *
67 * <p> Instances of this class are immutable and are safe for use by multiple
68 * concurrent threads. Instances of the {@link Matcher} class are not safe for
69 * such use.
70 *
71 *
72 * <a name="sum">
73 * <h4> Summary of regular-expression constructs </h4>
74 *
75 * <table border="0" cellpadding="1" cellspacing="0"
76 * summary="Regular expression constructs, and what they match">
77 *
78 * <tr align="left">
79 * <th bgcolor="#CCCCFF" align="left" id="construct">Construct</th>
80 * <th bgcolor="#CCCCFF" align="left" id="matches">Matches</th>
81 * </tr>
82 *
83 * <tr><th> </th></tr>
84 * <tr align="left"><th colspan="2" id="characters">Characters</th></tr>
85 *
86 * <tr><td valign="top" headers="construct characters"><i>x</i></td>
87 * <td headers="matches">The character <i>x</i></td></tr>
88 * <tr><td valign="top" headers="construct characters"><tt>\\</tt></td>
89 * <td headers="matches">The backslash character</td></tr>
90 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>n</i></td>
91 * <td headers="matches">The character with octal value <tt>0</tt><i>n</i>
92 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
93 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>nn</i></td>
94 * <td headers="matches">The character with octal value <tt>0</tt><i>nn</i>
95 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
96 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>mnn</i></td>
97 * <td headers="matches">The character with octal value <tt>0</tt><i>mnn</i>
98 * (0 <tt><=</tt> <i>m</i> <tt><=</tt> 3,
99 * 0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
100 * <tr><td valign="top" headers="construct characters"><tt>\x</tt><i>hh</i></td>
101 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hh</i></td></tr>
102 * <tr><td valign="top" headers="construct characters"><tt>\u</tt><i>hhhh</i></td>
103 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hhhh</i></td></tr>
104 * <tr><td valign="top" headers="matches"><tt>\t</tt></td>
105 * <td headers="matches">The tab character (<tt>'\u0009'</tt>)</td></tr>
106 * <tr><td valign="top" headers="construct characters"><tt>\n</tt></td>
107 * <td headers="matches">The newline (line feed) character (<tt>'\u000A'</tt>)</td></tr>
108 * <tr><td valign="top" headers="construct characters"><tt>\r</tt></td>
109 * <td headers="matches">The carriage-return character (<tt>'\u000D'</tt>)</td></tr>
110 * <tr><td valign="top" headers="construct characters"><tt>\f</tt></td>
111 * <td headers="matches">The form-feed character (<tt>'\u000C'</tt>)</td></tr>
112 * <tr><td valign="top" headers="construct characters"><tt>\a</tt></td>
113 * <td headers="matches">The alert (bell) character (<tt>'\u0007'</tt>)</td></tr>
114 * <tr><td valign="top" headers="construct characters"><tt>\e</tt></td>
115 * <td headers="matches">The escape character (<tt>'\u001B'</tt>)</td></tr>
116 * <tr><td valign="top" headers="construct characters"><tt>\c</tt><i>x</i></td>
117 * <td headers="matches">The control character corresponding to <i>x</i></td></tr>
118 *
119 * <tr><th> </th></tr>
120 * <tr align="left"><th colspan="2" id="classes">Character classes</th></tr>
121 *
122 * <tr><td valign="top" headers="construct classes"><tt>[abc]</tt></td>
123 * <td headers="matches"><tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (simple class)</td></tr>
124 * <tr><td valign="top" headers="construct classes"><tt>[^abc]</tt></td>
125 * <td headers="matches">Any character except <tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (negation)</td></tr>
126 * <tr><td valign="top" headers="construct classes"><tt>[a-zA-Z]</tt></td>
127 * <td headers="matches"><tt>a</tt> through <tt>z</tt>
128 * or <tt>A</tt> through <tt>Z</tt>, inclusive (range)</td></tr>
129 * <tr><td valign="top" headers="construct classes"><tt>[a-d[m-p]]</tt></td>
130 * <td headers="matches"><tt>a</tt> through <tt>d</tt>,
131 * or <tt>m</tt> through <tt>p</tt>: <tt>[a-dm-p]</tt> (union)</td></tr>
132 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[def]]</tt></td>
133 * <td headers="matches"><tt>d</tt>, <tt>e</tt>, or <tt>f</tt> (intersection)</tr>
134 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^bc]]</tt></td>
135 * <td headers="matches"><tt>a</tt> through <tt>z</tt>,
136 * except for <tt>b</tt> and <tt>c</tt>: <tt>[ad-z]</tt> (subtraction)</td></tr>
137 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^m-p]]</tt></td>
138 * <td headers="matches"><tt>a</tt> through <tt>z</tt>,
139 * and not <tt>m</tt> through <tt>p</tt>: <tt>[a-lq-z]</tt>(subtraction)</td></tr>
140 * <tr><th> </th></tr>
141 *
142 * <tr align="left"><th colspan="2" id="predef">Predefined character classes</th></tr>
143 *
144 * <tr><td valign="top" headers="construct predef"><tt>.</tt></td>
145 * <td headers="matches">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr>
146 * <tr><td valign="top" headers="construct predef"><tt>\d</tt></td>
147 * <td headers="matches">A digit: <tt>[0-9]</tt></td></tr>
148 * <tr><td valign="top" headers="construct predef"><tt>\D</tt></td>
149 * <td headers="matches">A non-digit: <tt>[^0-9]</tt></td></tr>
150 * <tr><td valign="top" headers="construct predef"><tt>\s</tt></td>
151 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr>
152 * <tr><td valign="top" headers="construct predef"><tt>\S</tt></td>
153 * <td headers="matches">A non-whitespace character: <tt>[^\s]</tt></td></tr>
154 * <tr><td valign="top" headers="construct predef"><tt>\w</tt></td>
155 * <td headers="matches">A word character: <tt>[a-zA-Z_0-9]</tt></td></tr>
156 * <tr><td valign="top" headers="construct predef"><tt>\W</tt></td>
157 * <td headers="matches">A non-word character: <tt>[^\w]</tt></td></tr>
158 *
159 * <tr><th> </th></tr>
160 * <tr align="left"><th colspan="2" id="posix">POSIX character classes</b> (US-ASCII only)<b></th></tr>
161 *
162 * <tr><td valign="top" headers="construct posix"><tt>\p{Lower}</tt></td>
163 * <td headers="matches">A lower-case alphabetic character: <tt>[a-z]</tt></td></tr>
164 * <tr><td valign="top" headers="construct posix"><tt>\p{Upper}</tt></td>
165 * <td headers="matches">An upper-case alphabetic character:<tt>[A-Z]</tt></td></tr>
166 * <tr><td valign="top" headers="construct posix"><tt>\p{ASCII}</tt></td>
167 * <td headers="matches">All ASCII:<tt>[\x00-\x7F]</tt></td></tr>
168 * <tr><td valign="top" headers="construct posix"><tt>\p{Alpha}</tt></td>
169 * <td headers="matches">An alphabetic character:<tt>[\p{Lower}\p{Upper}]</tt></td></tr>
170 * <tr><td valign="top" headers="construct posix"><tt>\p{Digit}</tt></td>
171 * <td headers="matches">A decimal digit: <tt>[0-9]</tt></td></tr>
172 * <tr><td valign="top" headers="construct posix"><tt>\p{Alnum}</tt></td>
173 * <td headers="matches">An alphanumeric character:<tt>[\p{Alpha}\p{Digit}]</tt></td></tr>
174 * <tr><td valign="top" headers="construct posix"><tt>\p{Punct}</tt></td>
175 * <td headers="matches">Punctuation: One of <tt>!"#$%&'()*+,-./:;<=>?@[\]^_`{|}~</tt></td></tr>
176 * <!-- <tt>[\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]</tt>
177 * <tt>[\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]</tt> -->
178 * <tr><td valign="top" headers="construct posix"><tt>\p{Graph}</tt></td>
179 * <td headers="matches">A visible character: <tt>[\p{Alnum}\p{Punct}]</tt></td></tr>
180 * <tr><td valign="top" headers="construct posix"><tt>\p{Print}</tt></td>
181 * <td headers="matches">A printable character: <tt>[\p{Graph}\x20]</tt></td></tr>
182 * <tr><td valign="top" headers="construct posix"><tt>\p{Blank}</tt></td>
183 * <td headers="matches">A space or a tab: <tt>[ \t]</tt></td></tr>
184 * <tr><td valign="top" headers="construct posix"><tt>\p{Cntrl}</tt></td>
185 * <td headers="matches">A control character: <tt>[\x00-\x1F\x7F]</tt></td></tr>
186 * <tr><td valign="top" headers="construct posix"><tt>\p{XDigit}</tt></td>
187 * <td headers="matches">A hexadecimal digit: <tt>[0-9a-fA-F]</tt></td></tr>
188 * <tr><td valign="top" headers="construct posix"><tt>\p{Space}</tt></td>
189 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr>
190 *
191 * <tr><th> </th></tr>
192 * <tr align="left"><th colspan="2">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr>
193 *
194 * <tr><td valign="top"><tt>\p{javaLowerCase}</tt></td>
195 * <td>Equivalent to java.lang.Character.isLowerCase()</td></tr>
196 * <tr><td valign="top"><tt>\p{javaUpperCase}</tt></td>
197 * <td>Equivalent to java.lang.Character.isUpperCase()</td></tr>
198 * <tr><td valign="top"><tt>\p{javaWhitespace}</tt></td>
199 * <td>Equivalent to java.lang.Character.isWhitespace()</td></tr>
200 * <tr><td valign="top"><tt>\p{javaMirrored}</tt></td>
201 * <td>Equivalent to java.lang.Character.isMirrored()</td></tr>
202 *
203 * <tr><th> </th></tr>
204 * <tr align="left"><th colspan="2" id="unicode">Classes for Unicode scripts, blocks and categories</th></tr>
205 * * <tr><td valign="top" headers="construct unicode"><tt>\p{IsLatin}</tt></td>
206 * <td headers="matches">A Latin script character (simple <a href="#ubc">script</a>)</td></tr>
207 * <tr><td valign="top" headers="construct unicode"><tt>\p{InGreek}</tt></td>
208 * <td headers="matches">A character in the Greek block (simple <a href="#ubc">block</a>)</td></tr>
209 * <tr><td valign="top" headers="construct unicode"><tt>\p{Lu}</tt></td>
210 * <td headers="matches">An uppercase letter (simple <a href="#ubc">category</a>)</td></tr>
211 * <tr><td valign="top" headers="construct unicode"><tt>\p{Sc}</tt></td>
212 * <td headers="matches">A currency symbol</td></tr>
213 * <tr><td valign="top" headers="construct unicode"><tt>\P{InGreek}</tt></td>
214 * <td headers="matches">Any character except one in the Greek block (negation)</td></tr>
215 * <tr><td valign="top" headers="construct unicode"><tt>[\p{L}&&[^\p{Lu}]] </tt></td>
216 * <td headers="matches">Any letter except an uppercase letter (subtraction)</td></tr>
217 *
218 * <tr><th> </th></tr>
219 * <tr align="left"><th colspan="2" id="bounds">Boundary matchers</th></tr>
220 *
221 * <tr><td valign="top" headers="construct bounds"><tt>^</tt></td>
222 * <td headers="matches">The beginning of a line</td></tr>
223 * <tr><td valign="top" headers="construct bounds"><tt>$</tt></td>
224 * <td headers="matches">The end of a line</td></tr>
225 * <tr><td valign="top" headers="construct bounds"><tt>\b</tt></td>
226 * <td headers="matches">A word boundary</td></tr>
227 * <tr><td valign="top" headers="construct bounds"><tt>\B</tt></td>
228 * <td headers="matches">A non-word boundary</td></tr>
229 * <tr><td valign="top" headers="construct bounds"><tt>\A</tt></td>
230 * <td headers="matches">The beginning of the input</td></tr>
231 * <tr><td valign="top" headers="construct bounds"><tt>\G</tt></td>
232 * <td headers="matches">The end of the previous match</td></tr>
233 * <tr><td valign="top" headers="construct bounds"><tt>\Z</tt></td>
234 * <td headers="matches">The end of the input but for the final
235 * <a href="#lt">terminator</a>, if any</td></tr>
236 * <tr><td valign="top" headers="construct bounds"><tt>\z</tt></td>
237 * <td headers="matches">The end of the input</td></tr>
238 *
239 * <tr><th> </th></tr>
240 * <tr align="left"><th colspan="2" id="greedy">Greedy quantifiers</th></tr>
241 *
242 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>?</tt></td>
243 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
244 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>*</tt></td>
245 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
246 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>+</tt></td>
247 * <td headers="matches"><i>X</i>, one or more times</td></tr>
248 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>}</tt></td>
249 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
250 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>,}</tt></td>
251 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
252 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}</tt></td>
253 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
254 *
255 * <tr><th> </th></tr>
256 * <tr align="left"><th colspan="2" id="reluc">Reluctant quantifiers</th></tr>
257 *
258 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>??</tt></td>
259 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
260 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>*?</tt></td>
261 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
262 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>+?</tt></td>
263 * <td headers="matches"><i>X</i>, one or more times</td></tr>
264 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>}?</tt></td>
265 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
266 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>,}?</tt></td>
267 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
268 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}?</tt></td>
269 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
270 *
271 * <tr><th> </th></tr>
272 * <tr align="left"><th colspan="2" id="poss">Possessive quantifiers</th></tr>
273 *
274 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>?+</tt></td>
275 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
276 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>*+</tt></td>
277 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
278 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>++</tt></td>
279 * <td headers="matches"><i>X</i>, one or more times</td></tr>
280 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>}+</tt></td>
281 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
282 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>,}+</tt></td>
283 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
284 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}+</tt></td>
285 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
286 *
287 * <tr><th> </th></tr>
288 * <tr align="left"><th colspan="2" id="logical">Logical operators</th></tr>
289 *
290 * <tr><td valign="top" headers="construct logical"><i>XY</i></td>
291 * <td headers="matches"><i>X</i> followed by <i>Y</i></td></tr>
292 * <tr><td valign="top" headers="construct logical"><i>X</i><tt>|</tt><i>Y</i></td>
293 * <td headers="matches">Either <i>X</i> or <i>Y</i></td></tr>
294 * <tr><td valign="top" headers="construct logical"><tt>(</tt><i>X</i><tt>)</tt></td>
295 * <td headers="matches">X, as a <a href="#cg">capturing group</a></td></tr>
296 *
297 * <tr><th> </th></tr>
298 * <tr align="left"><th colspan="2" id="backref">Back references</th></tr>
299 *
300 * <tr><td valign="bottom" headers="construct backref"><tt>\</tt><i>n</i></td>
301 * <td valign="bottom" headers="matches">Whatever the <i>n</i><sup>th</sup>
302 * <a href="#cg">capturing group</a> matched</td></tr>
303 *
304 * <tr><td valign="bottom" headers="construct backref"><tt>\</tt><i>k</i><<i>name</i>></td>
305 * <td valign="bottom" headers="matches">Whatever the
306 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr>
307 *
308 * <tr><th> </th></tr>
309 * <tr align="left"><th colspan="2" id="quot">Quotation</th></tr>
310 *
311 * <tr><td valign="top" headers="construct quot"><tt>\</tt></td>
312 * <td headers="matches">Nothing, but quotes the following character</td></tr>
313 * <tr><td valign="top" headers="construct quot"><tt>\Q</tt></td>
314 * <td headers="matches">Nothing, but quotes all characters until <tt>\E</tt></td></tr>
315 * <tr><td valign="top" headers="construct quot"><tt>\E</tt></td>
316 * <td headers="matches">Nothing, but ends quoting started by <tt>\Q</tt></td></tr>
317 * <!-- Metachars: !$()*+.<>?[\]^{|} -->
318 *
319 * <tr><th> </th></tr>
320 * <tr align="left"><th colspan="2" id="special">Special constructs (named-capturing and non-capturing)</th></tr>
321 *
322 * <tr><td valign="top" headers="construct special"><tt>(?<<a href="#groupname">name</a>></tt><i>X</i><tt>)</tt></td>
323 * <td headers="matches"><i>X</i>, as a named-capturing group</td></tr>
324 * <tr><td valign="top" headers="construct special"><tt>(?:</tt><i>X</i><tt>)</tt></td>
325 * <td headers="matches"><i>X</i>, as a non-capturing group</td></tr>
326 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux) </tt></td>
327 * <td headers="matches">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a>
328 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a>
329 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> on - off</td></tr>
330 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux:</tt><i>X</i><tt>)</tt> </td>
331 * <td headers="matches"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the
332 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a>
333 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a >
334 * <a href="#COMMENTS">x</a> on - off</td></tr>
335 * <tr><td valign="top" headers="construct special"><tt>(?=</tt><i>X</i><tt>)</tt></td>
336 * <td headers="matches"><i>X</i>, via zero-width positive lookahead</td></tr>
337 * <tr><td valign="top" headers="construct special"><tt>(?!</tt><i>X</i><tt>)</tt></td>
338 * <td headers="matches"><i>X</i>, via zero-width negative lookahead</td></tr>
339 * <tr><td valign="top" headers="construct special"><tt>(?<=</tt><i>X</i><tt>)</tt></td>
340 * <td headers="matches"><i>X</i>, via zero-width positive lookbehind</td></tr>
341 * <tr><td valign="top" headers="construct special"><tt>(?<!</tt><i>X</i><tt>)</tt></td>
342 * <td headers="matches"><i>X</i>, via zero-width negative lookbehind</td></tr>
343 * <tr><td valign="top" headers="construct special"><tt>(?></tt><i>X</i><tt>)</tt></td>
344 * <td headers="matches"><i>X</i>, as an independent, non-capturing group</td></tr>
345 *
346 * </table>
347 *
348 * <hr>
349 *
350 *
351 * <a name="bs">
352 * <h4> Backslashes, escapes, and quoting </h4>
353 *
354 * <p> The backslash character (<tt>'\'</tt>) serves to introduce escaped
355 * constructs, as defined in the table above, as well as to quote characters
356 * that otherwise would be interpreted as unescaped constructs. Thus the
357 * expression <tt>\\</tt> matches a single backslash and <tt>\{</tt> matches a
358 * left brace.
359 *
360 * <p> It is an error to use a backslash prior to any alphabetic character that
361 * does not denote an escaped construct; these are reserved for future
362 * extensions to the regular-expression language. A backslash may be used
363 * prior to a non-alphabetic character regardless of whether that character is
364 * part of an unescaped construct.
365 *
366 * <p> Backslashes within string literals in Java source code are interpreted
367 * as required by the <a
368 * href="http://java.sun.com/docs/books/jls">Java Language
369 * Specification</a> as either <a
370 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">Unicode
371 * escapes</a> or other <a
372 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#101089">character
373 * escapes</a>. It is therefore necessary to double backslashes in string
374 * literals that represent regular expressions to protect them from
375 * interpretation by the Java bytecode compiler. The string literal
376 * <tt>"\b"</tt>, for example, matches a single backspace character when
377 * interpreted as a regular expression, while <tt>"\\b"</tt> matches a
378 * word boundary. The string literal <tt>"\(hello\)"</tt> is illegal
379 * and leads to a compile-time error; in order to match the string
380 * <tt>(hello)</tt> the string literal <tt>"\\(hello\\)"</tt>
381 * must be used.
382 *
383 * <a name="cc">
384 * <h4> Character Classes </h4>
385 *
386 * <p> Character classes may appear within other character classes, and
387 * may be composed by the union operator (implicit) and the intersection
388 * operator (<tt>&&</tt>).
389 * The union operator denotes a class that contains every character that is
390 * in at least one of its operand classes. The intersection operator
391 * denotes a class that contains every character that is in both of its
392 * operand classes.
393 *
394 * <p> The precedence of character-class operators is as follows, from
395 * highest to lowest:
396 *
397 * <blockquote><table border="0" cellpadding="1" cellspacing="0"
398 * summary="Precedence of character class operators.">
399 * <tr><th>1 </th>
400 * <td>Literal escape </td>
401 * <td><tt>\x</tt></td></tr>
402 * <tr><th>2 </th>
403 * <td>Grouping</td>
404 * <td><tt>[...]</tt></td></tr>
405 * <tr><th>3 </th>
406 * <td>Range</td>
407 * <td><tt>a-z</tt></td></tr>
408 * <tr><th>4 </th>
409 * <td>Union</td>
410 * <td><tt>[a-e][i-u]</tt></td></tr>
411 * <tr><th>5 </th>
412 * <td>Intersection</td>
413 * <td><tt>[a-z&&[aeiou]]</tt></td></tr>
414 * </table></blockquote>
415 *
416 * <p> Note that a different set of metacharacters are in effect inside
417 * a character class than outside a character class. For instance, the
418 * regular expression <tt>.</tt> loses its special meaning inside a
419 * character class, while the expression <tt>-</tt> becomes a range
420 * forming metacharacter.
421 *
422 * <a name="lt">
423 * <h4> Line terminators </h4>
424 *
425 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks
426 * the end of a line of the input character sequence. The following are
427 * recognized as line terminators:
428 *
429 * <ul>
430 *
431 * <li> A newline (line feed) character (<tt>'\n'</tt>),
432 *
433 * <li> A carriage-return character followed immediately by a newline
434 * character (<tt>"\r\n"</tt>),
435 *
436 * <li> A standalone carriage-return character (<tt>'\r'</tt>),
437 *
438 * <li> A next-line character (<tt>'\u0085'</tt>),
439 *
440 * <li> A line-separator character (<tt>'\u2028'</tt>), or
441 *
442 * <li> A paragraph-separator character (<tt>'\u2029</tt>).
443 *
444 * </ul>
445 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators
446 * recognized are newline characters.
447 *
448 * <p> The regular expression <tt>.</tt> matches any character except a line
449 * terminator unless the {@link #DOTALL} flag is specified.
450 *
451 * <p> By default, the regular expressions <tt>^</tt> and <tt>$</tt> ignore
452 * line terminators and only match at the beginning and the end, respectively,
453 * of the entire input sequence. If {@link #MULTILINE} mode is activated then
454 * <tt>^</tt> matches at the beginning of input and after any line terminator
455 * except at the end of input. When in {@link #MULTILINE} mode <tt>$</tt>
456 * matches just before a line terminator or the end of the input sequence.
457 *
458 * <a name="cg">
459 * <h4> Groups and capturing </h4>
460 *
461 * <a name="gnumber">
462 * <h5> Group number </h5>
463 * <p> Capturing groups are numbered by counting their opening parentheses from
464 * left to right. In the expression <tt>((A)(B(C)))</tt>, for example, there
465 * are four such groups: </p>
466 *
467 * <blockquote><table cellpadding=1 cellspacing=0 summary="Capturing group numberings">
468 * <tr><th>1 </th>
469 * <td><tt>((A)(B(C)))</tt></td></tr>
470 * <tr><th>2 </th>
471 * <td><tt>(A)</tt></td></tr>
472 * <tr><th>3 </th>
473 * <td><tt>(B(C))</tt></td></tr>
474 * <tr><th>4 </th>
475 * <td><tt>(C)</tt></td></tr>
476 * </table></blockquote>
477 *
478 * <p> Group zero always stands for the entire expression.
479 *
480 * <p> Capturing groups are so named because, during a match, each subsequence
481 * of the input sequence that matches such a group is saved. The captured
482 * subsequence may be used later in the expression, via a back reference, and
483 * may also be retrieved from the matcher once the match operation is complete.
484 *
485 * <a name="groupname">
486 * <h5> Group name </h5>
487 * <p>A capturing group can also be assigned a "name", a <tt>named-capturing group</tt>,
488 * and then be back-referenced later by the "name". Group names are composed of
489 * the following characters. The first character must be a <tt>letter</tt>.
490 *
491 * <ul>
492 * <li> The uppercase letters <tt>'A'</tt> through <tt>'Z'</tt>
493 * (<tt>'\u0041'</tt> through <tt>'\u005a'</tt>),
494 * <li> The lowercase letters <tt>'a'</tt> through <tt>'z'</tt>
495 * (<tt>'\u0061'</tt> through <tt>'\u007a'</tt>),
496 * <li> The digits <tt>'0'</tt> through <tt>'9'</tt>
497 * (<tt>'\u0030'</tt> through <tt>'\u0039'</tt>),
498 * </ul>
499 *
500 * <p> A <tt>named-capturing group</tt> is still numbered as described in
501 * <a href="#gnumber">Group number</a>.
502 *
503 * <p> The captured input associated with a group is always the subsequence
504 * that the group most recently matched. If a group is evaluated a second time
505 * because of quantification then its previously-captured value, if any, will
506 * be retained if the second evaluation fails. Matching the string
507 * <tt>"aba"</tt> against the expression <tt>(a(b)?)+</tt>, for example, leaves
508 * group two set to <tt>"b"</tt>. All captured input is discarded at the
509 * beginning of each match.
510 *
511 * <p> Groups beginning with <tt>(?</tt> are either pure, <i>non-capturing</i> groups
512 * that do not capture text and do not count towards the group total, or
513 * <i>named-capturing</i> group.
514 *
515 * <h4> Unicode support </h4>
516 *
517 * <p> This class is in conformance with Level 1 of <a
518 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical
519 * Standard #18: Unicode Regular Expression Guidelines</i></a>, plus RL2.1
520 * Canonical Equivalents.
521 *
522 * <p> Unicode escape sequences such as <tt>\u2014</tt> in Java source code
523 * are processed as described in <a
524 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">\u00A73.3</a>
525 * of the Java Language Specification. Such escape sequences are also
526 * implemented directly by the regular-expression parser so that Unicode
527 * escapes can be used in expressions that are read from files or from the
528 * keyboard. Thus the strings <tt>"\u2014"</tt> and <tt>"\\u2014"</tt>,
529 * while not equal, compile into the same pattern, which matches the character
530 * with hexadecimal value <tt>0x2014</tt>.
531 *
532 * <a name="ubc">
533 * <p>Unicode scripts, blocks and categories are written with the <tt>\p</tt> and
534 * <tt>\P</tt> constructs as in Perl. <tt>\p{</tt><i>prop</i><tt>}</tt> matches if
535 * the input has the property <i>prop</i>, while <tt>\P{</tt><i>prop</i><tt>}</tt>
536 * does not match if the input has that property.
537 * <p>
538 * Scripts are specified either with the prefix {@code Is}, as in
539 * {@code IsHiragana}, or by using the {@code script} keyword (or its short
540 * form {@code sc})as in {@code script=Hiragana} or {@code sc=Hiragana}.
541 * <p>
542 * Blocks are specified with the prefix {@code In}, as in
543 * {@code InMongolian}, or by using the keyword {@code block} (or its short
544 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}.
545 * <p>
546 * Categories may be specified with the optional prefix {@code Is}:
547 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode
548 * letters. Same as scripts and blocks, categories can also be specified
549 * by using the keyword {@code general_category} (or its short form
550 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}.
551 * <p>
552 * Scripts, blocks and categories can be used both inside and outside of a
553 * character class.
554 * <p> The supported categories are those of
555 * <a href="http://www.unicode.org/unicode/standard/standard.html">
556 * <i>The Unicode Standard</i></a> in the version specified by the
557 * {@link java.lang.Character Character} class. The category names are those
558 * defined in the Standard, both normative and informative.
559 * The script names supported by <code>Pattern</code> are the valid script names
560 * accepted and defined by
561 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}.
562 * The block names supported by <code>Pattern</code> are the valid block names
563 * accepted and defined by
564 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}.
565 * <p>
566 * <a name="jcc"> <p>Categories that behave like the java.lang.Character
567 * boolean is<i>methodname</i> methods (except for the deprecated ones) are
568 * available through the same <tt>\p{</tt><i>prop</i><tt>}</tt> syntax where
569 * the specified property has the name <tt>java<i>methodname</i></tt>.
570 *
571 * <h4> Comparison to Perl 5 </h4>
572 *
573 * <p>The <code>Pattern</code> engine performs traditional NFA-based matching
574 * with ordered alternation as occurs in Perl 5.
575 *
576 * <p> Perl constructs not supported by this class: </p>
577 *
578 * <ul>
579 *
580 * <li><p> The conditional constructs <tt>(?{</tt><i>X</i><tt>})</tt> and
581 * <tt>(?(</tt><i>condition</i><tt>)</tt><i>X</i><tt>|</tt><i>Y</i><tt>)</tt>,
582 * </p></li>
583 *
584 * <li><p> The embedded code constructs <tt>(?{</tt><i>code</i><tt>})</tt>
585 * and <tt>(??{</tt><i>code</i><tt>})</tt>,</p></li>
586 *
587 * <li><p> The embedded comment syntax <tt>(?#comment)</tt>, and </p></li>
588 *
589 * <li><p> The preprocessing operations <tt>\l</tt> <tt>\u</tt>,
590 * <tt>\L</tt>, and <tt>\U</tt>. </p></li>
591 *
592 * </ul>
593 *
594 * <p> Constructs supported by this class but not by Perl: </p>
595 *
596 * <ul>
597 *
598 * <li><p> Possessive quantifiers, which greedily match as much as they can
599 * and do not back off, even when doing so would allow the overall match to
600 * succeed. </p></li>
601 *
602 * <li><p> Character-class union and intersection as described
603 * <a href="#cc">above</a>.</p></li>
604 *
605 * </ul>
606 *
607 * <p> Notable differences from Perl: </p>
608 *
609 * <ul>
610 *
611 * <li><p> In Perl, <tt>\1</tt> through <tt>\9</tt> are always interpreted
612 * as back references; a backslash-escaped number greater than <tt>9</tt> is
613 * treated as a back reference if at least that many subexpressions exist,
614 * otherwise it is interpreted, if possible, as an octal escape. In this
615 * class octal escapes must always begin with a zero. In this class,
616 * <tt>\1</tt> through <tt>\9</tt> are always interpreted as back
617 * references, and a larger number is accepted as a back reference if at
618 * least that many subexpressions exist at that point in the regular
619 * expression, otherwise the parser will drop digits until the number is
620 * smaller or equal to the existing number of groups or it is one digit.
621 * </p></li>
622 *
623 * <li><p> Perl uses the <tt>g</tt> flag to request a match that resumes
624 * where the last match left off. This functionality is provided implicitly
625 * by the {@link Matcher} class: Repeated invocations of the {@link
626 * Matcher#find find} method will resume where the last match left off,
627 * unless the matcher is reset. </p></li>
628 *
629 * <li><p> In Perl, embedded flags at the top level of an expression affect
630 * the whole expression. In this class, embedded flags always take effect
631 * at the point at which they appear, whether they are at the top level or
632 * within a group; in the latter case, flags are restored at the end of the
633 * group just as in Perl. </p></li>
634 *
635 * <li><p> Perl is forgiving about malformed matching constructs, as in the
636 * expression <tt>*a</tt>, as well as dangling brackets, as in the
637 * expression <tt>abc]</tt>, and treats them as literals. This
638 * class also accepts dangling brackets but is strict about dangling
639 * metacharacters like +, ? and *, and will throw a
640 * {@link PatternSyntaxException} if it encounters them. </p></li>
641 *
642 * </ul>
643 *
644 *
645 * <p> For a more precise description of the behavior of regular expression
646 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/">
647 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl,
648 * O'Reilly and Associates, 2006.</a>
649 * </p>
650 *
651 * @see java.lang.String#split(String, int)
652 * @see java.lang.String#split(String)
653 *
654 * @author Mike McCloskey
655 * @author Mark Reinhold
656 * @author JSR-51 Expert Group
657 * @since 1.4
658 * @spec JSR-51
659 */
660
661 public final class Pattern
662 implements java.io.Serializable
663 {
664
665 /**
666 * Regular expression modifier values. Instead of being passed as
667 * arguments, they can also be passed as inline modifiers.
668 * For example, the following statements have the same effect.
669 * <pre>
670 * RegExp r1 = RegExp.compile("abc", Pattern.I|Pattern.M);
671 * RegExp r2 = RegExp.compile("(?im)abc", 0);
672 * </pre>
673 *
674 * The flags are duplicated so that the familiar Perl match flag
675 * names are available.
676 */
677
678 /**
679 * Enables Unix lines mode.
680 *
681 * <p> In this mode, only the <tt>'\n'</tt> line terminator is recognized
682 * in the behavior of <tt>.</tt>, <tt>^</tt>, and <tt>$</tt>.
683 *
684 * <p> Unix lines mode can also be enabled via the embedded flag
685 * expression <tt>(?d)</tt>.
686 */
687 public static final int UNIX_LINES = 0x01;
688
689 /**
690 * Enables case-insensitive matching.
691 *
692 * <p> By default, case-insensitive matching assumes that only characters
693 * in the US-ASCII charset are being matched. Unicode-aware
694 * case-insensitive matching can be enabled by specifying the {@link
695 * #UNICODE_CASE} flag in conjunction with this flag.
696 *
697 * <p> Case-insensitive matching can also be enabled via the embedded flag
698 * expression <tt>(?i)</tt>.
699 *
700 * <p> Specifying this flag may impose a slight performance penalty. </p>
701 */
702 public static final int CASE_INSENSITIVE = 0x02;
703
704 /**
705 * Permits whitespace and comments in pattern.
706 *
707 * <p> In this mode, whitespace is ignored, and embedded comments starting
708 * with <tt>#</tt> are ignored until the end of a line.
709 *
710 * <p> Comments mode can also be enabled via the embedded flag
711 * expression <tt>(?x)</tt>.
712 */
713 public static final int COMMENTS = 0x04;
714
715 /**
716 * Enables multiline mode.
717 *
718 * <p> In multiline mode the expressions <tt>^</tt> and <tt>$</tt> match
719 * just after or just before, respectively, a line terminator or the end of
720 * the input sequence. By default these expressions only match at the
721 * beginning and the end of the entire input sequence.
722 *
723 * <p> Multiline mode can also be enabled via the embedded flag
724 * expression <tt>(?m)</tt>. </p>
725 */
726 public static final int MULTILINE = 0x08;
727
728 /**
729 * Enables literal parsing of the pattern.
730 *
731 * <p> When this flag is specified then the input string that specifies
732 * the pattern is treated as a sequence of literal characters.
733 * Metacharacters or escape sequences in the input sequence will be
734 * given no special meaning.
735 *
736 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on
737 * matching when used in conjunction with this flag. The other flags
738 * become superfluous.
739 *
740 * <p> There is no embedded flag character for enabling literal parsing.
741 * @since 1.5
742 */
743 public static final int LITERAL = 0x10;
744
745 /**
746 * Enables dotall mode.
747 *
748 * <p> In dotall mode, the expression <tt>.</tt> matches any character,
749 * including a line terminator. By default this expression does not match
750 * line terminators.
751 *
752 * <p> Dotall mode can also be enabled via the embedded flag
753 * expression <tt>(?s)</tt>. (The <tt>s</tt> is a mnemonic for
754 * "single-line" mode, which is what this is called in Perl.) </p>
755 */
756 public static final int DOTALL = 0x20;
757
758 /**
759 * Enables Unicode-aware case folding.
760 *
761 * <p> When this flag is specified then case-insensitive matching, when
762 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner
763 * consistent with the Unicode Standard. By default, case-insensitive
764 * matching assumes that only characters in the US-ASCII charset are being
765 * matched.
766 *
767 * <p> Unicode-aware case folding can also be enabled via the embedded flag
768 * expression <tt>(?u)</tt>.
769 *
770 * <p> Specifying this flag may impose a performance penalty. </p>
771 */
772 public static final int UNICODE_CASE = 0x40;
773
774 /**
775 * Enables canonical equivalence.
776 *
777 * <p> When this flag is specified then two characters will be considered
778 * to match if, and only if, their full canonical decompositions match.
779 * The expression <tt>"a\u030A"</tt>, for example, will match the
780 * string <tt>"\u00E5"</tt> when this flag is specified. By default,
781 * matching does not take canonical equivalence into account.
782 *
783 * <p> There is no embedded flag character for enabling canonical
784 * equivalence.
785 *
786 * <p> Specifying this flag may impose a performance penalty. </p>
787 */
788 public static final int CANON_EQ = 0x80;
789
790 /* Pattern has only two serialized components: The pattern string
791 * and the flags, which are all that is needed to recompile the pattern
792 * when it is deserialized.
793 */
794
795 /** use serialVersionUID from Merlin b59 for interoperability */
796 private static final long serialVersionUID = 5073258162644648461L;
797
798 /**
799 * The original regular-expression pattern string.
800 *
801 * @serial
802 */
803 private String pattern;
804
805 /**
806 * The original pattern flags.
807 *
808 * @serial
809 */
810 private int flags;
811
812 /**
813 * Boolean indicating this Pattern is compiled; this is necessary in order
814 * to lazily compile deserialized Patterns.
815 */
816 private transient volatile boolean compiled = false;
817
818 /**
819 * The normalized pattern string.
820 */
821 private transient String normalizedPattern;
822
823 /**
824 * The starting point of state machine for the find operation. This allows
825 * a match to start anywhere in the input.
826 */
827 transient Node root;
828
829 /**
830 * The root of object tree for a match operation. The pattern is matched
831 * at the beginning. This may include a find that uses BnM or a First
832 * node.
833 */
834 transient Node matchRoot;
835
836 /**
837 * Temporary storage used by parsing pattern slice.
838 */
839 transient int[] buffer;
840
841 /**
842 * Map the "name" of the "named capturing group" to its group id
843 * node.
844 */
845 transient volatile Map<String, Integer> namedGroups;
846
847 /**
848 * Temporary storage used while parsing group references.
849 */
850 transient GroupHead[] groupNodes;
851
852 /**
853 * Temporary null terminated code point array used by pattern compiling.
854 */
855 private transient int[] temp;
856
857 /**
858 * The number of capturing groups in this Pattern. Used by matchers to
859 * allocate storage needed to perform a match.
860 */
861 transient int capturingGroupCount;
862
863 /**
864 * The local variable count used by parsing tree. Used by matchers to
865 * allocate storage needed to perform a match.
866 */
867 transient int localCount;
868
869 /**
870 * Index into the pattern string that keeps track of how much has been
871 * parsed.
872 */
873 private transient int cursor;
874
875 /**
876 * Holds the length of the pattern string.
877 */
878 private transient int patternLength;
879
880 /**
881 * If the Start node might possibly match supplementary characters.
882 * It is set to true during compiling if
883 * (1) There is supplementary char in pattern, or
884 * (2) There is complement node of Category or Block
885 */
886 private transient boolean hasSupplementary;
887
888 /**
889 * Compiles the given regular expression into a pattern. </p>
890 *
891 * @param regex
892 * The expression to be compiled
893 *
894 * @throws PatternSyntaxException
895 * If the expression's syntax is invalid
896 */
897 public static Pattern compile(String regex) {
898 return new Pattern(regex, 0);
899 }
900
901 /**
902 * Compiles the given regular expression into a pattern with the given
903 * flags. </p>
904 *
905 * @param regex
906 * The expression to be compiled
907 *
908 * @param flags
909 * Match flags, a bit mask that may include
910 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL},
911 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES},
912 * {@link #LITERAL} and {@link #COMMENTS}
913 *
914 * @throws IllegalArgumentException
915 * If bit values other than those corresponding to the defined
916 * match flags are set in <tt>flags</tt>
917 *
918 * @throws PatternSyntaxException
919 * If the expression's syntax is invalid
920 */
921 public static Pattern compile(String regex, int flags) {
922 return new Pattern(regex, flags);
923 }
924
925 /**
926 * Returns the regular expression from which this pattern was compiled.
927 * </p>
928 *
929 * @return The source of this pattern
930 */
931 public String pattern() {
932 return pattern;
933 }
934
935 /**
936 * <p>Returns the string representation of this pattern. This
937 * is the regular expression from which this pattern was
938 * compiled.</p>
939 *
940 * @return The string representation of this pattern
941 * @since 1.5
942 */
943 public String toString() {
944 return pattern;
945 }
946
947 /**
948 * Creates a matcher that will match the given input against this pattern.
949 * </p>
950 *
951 * @param input
952 * The character sequence to be matched
953 *
954 * @return A new matcher for this pattern
955 */
956 public Matcher matcher(CharSequence input) {
957 if (!compiled) {
958 synchronized(this) {
959 if (!compiled)
960 compile();
961 }
962 }
963 Matcher m = new Matcher(this, input);
964 return m;
965 }
966
967 /**
968 * Returns this pattern's match flags. </p>
969 *
970 * @return The match flags specified when this pattern was compiled
971 */
972 public int flags() {
973 return flags;
974 }
975
976 /**
977 * Compiles the given regular expression and attempts to match the given
978 * input against it.
979 *
980 * <p> An invocation of this convenience method of the form
981 *
982 * <blockquote><pre>
983 * Pattern.matches(regex, input);</pre></blockquote>
984 *
985 * behaves in exactly the same way as the expression
986 *
987 * <blockquote><pre>
988 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote>
989 *
990 * <p> If a pattern is to be used multiple times, compiling it once and reusing
991 * it will be more efficient than invoking this method each time. </p>
992 *
993 * @param regex
994 * The expression to be compiled
995 *
996 * @param input
997 * The character sequence to be matched
998 *
999 * @throws PatternSyntaxException
1000 * If the expression's syntax is invalid
1001 */
1002 public static boolean matches(String regex, CharSequence input) {
1003 Pattern p = Pattern.compile(regex);
1004 Matcher m = p.matcher(input);
1005 return m.matches();
1006 }
1007
1008 /**
1009 * Splits the given input sequence around matches of this pattern.
1010 *
1011 * <p> The array returned by this method contains each substring of the
1012 * input sequence that is terminated by another subsequence that matches
1013 * this pattern or is terminated by the end of the input sequence. The
1014 * substrings in the array are in the order in which they occur in the
1015 * input. If this pattern does not match any subsequence of the input then
1016 * the resulting array has just one element, namely the input sequence in
1017 * string form.
1018 *
1019 * <p> The <tt>limit</tt> parameter controls the number of times the
1020 * pattern is applied and therefore affects the length of the resulting
1021 * array. If the limit <i>n</i> is greater than zero then the pattern
1022 * will be applied at most <i>n</i> - 1 times, the array's
1023 * length will be no greater than <i>n</i>, and the array's last entry
1024 * will contain all input beyond the last matched delimiter. If <i>n</i>
1025 * is non-positive then the pattern will be applied as many times as
1026 * possible and the array can have any length. If <i>n</i> is zero then
1027 * the pattern will be applied as many times as possible, the array can
1028 * have any length, and trailing empty strings will be discarded.
1029 *
1030 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following
1031 * results with these parameters:
1032 *
1033 * <blockquote><table cellpadding=1 cellspacing=0
1034 * summary="Split examples showing regex, limit, and result">
1035 * <tr><th><P align="left"><i>Regex </i></th>
1036 * <th><P align="left"><i>Limit </i></th>
1037 * <th><P align="left"><i>Result </i></th></tr>
1038 * <tr><td align=center>:</td>
1039 * <td align=center>2</td>
1040 * <td><tt>{ "boo", "and:foo" }</tt></td></tr>
1041 * <tr><td align=center>:</td>
1042 * <td align=center>5</td>
1043 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
1044 * <tr><td align=center>:</td>
1045 * <td align=center>-2</td>
1046 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
1047 * <tr><td align=center>o</td>
1048 * <td align=center>5</td>
1049 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr>
1050 * <tr><td align=center>o</td>
1051 * <td align=center>-2</td>
1052 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr>
1053 * <tr><td align=center>o</td>
1054 * <td align=center>0</td>
1055 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr>
1056 * </table></blockquote>
1057 *
1058 *
1059 * @param input
1060 * The character sequence to be split
1061 *
1062 * @param limit
1063 * The result threshold, as described above
1064 *
1065 * @return The array of strings computed by splitting the input
1066 * around matches of this pattern
1067 */
1068 public String[] split(CharSequence input, int limit) {
1069 int index = 0;
1070 boolean matchLimited = limit > 0;
1071 ArrayList<String> matchList = new ArrayList<String>();
1072 Matcher m = matcher(input);
1073
1074 // Add segments before each match found
1075 while(m.find()) {
1076 if (!matchLimited || matchList.size() < limit - 1) {
1077 String match = input.subSequence(index, m.start()).toString();
1078 matchList.add(match);
1079 index = m.end();
1080 } else if (matchList.size() == limit - 1) { // last one
1081 String match = input.subSequence(index,
1082 input.length()).toString();
1083 matchList.add(match);
1084 index = m.end();
1085 }
1086 }
1087
1088 // If no match was found, return this
1089 if (index == 0)
1090 return new String[] {input.toString()};
1091
1092 // Add remaining segment
1093 if (!matchLimited || matchList.size() < limit)
1094 matchList.add(input.subSequence(index, input.length()).toString());
1095
1096 // Construct result
1097 int resultSize = matchList.size();
1098 if (limit == 0)
1099 while (resultSize > 0 && matchList.get(resultSize-1).equals(""))
1100 resultSize--;
1101 String[] result = new String[resultSize];
1102 return matchList.subList(0, resultSize).toArray(result);
1103 }
1104
1105 /**
1106 * Splits the given input sequence around matches of this pattern.
1107 *
1108 * <p> This method works as if by invoking the two-argument {@link
1109 * #split(java.lang.CharSequence, int) split} method with the given input
1110 * sequence and a limit argument of zero. Trailing empty strings are
1111 * therefore not included in the resulting array. </p>
1112 *
1113 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following
1114 * results with these expressions:
1115 *
1116 * <blockquote><table cellpadding=1 cellspacing=0
1117 * summary="Split examples showing regex and result">
1118 * <tr><th><P align="left"><i>Regex </i></th>
1119 * <th><P align="left"><i>Result</i></th></tr>
1120 * <tr><td align=center>:</td>
1121 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
1122 * <tr><td align=center>o</td>
1123 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr>
1124 * </table></blockquote>
1125 *
1126 *
1127 * @param input
1128 * The character sequence to be split
1129 *
1130 * @return The array of strings computed by splitting the input
1131 * around matches of this pattern
1132 */
1133 public String[] split(CharSequence input) {
1134 return split(input, 0);
1135 }
1136
1137 /**
1138 * Returns a literal pattern <code>String</code> for the specified
1139 * <code>String</code>.
1140 *
1141 * <p>This method produces a <code>String</code> that can be used to
1142 * create a <code>Pattern</code> that would match the string
1143 * <code>s</code> as if it were a literal pattern.</p> Metacharacters
1144 * or escape sequences in the input sequence will be given no special
1145 * meaning.
1146 *
1147 * @param s The string to be literalized
1148 * @return A literal string replacement
1149 * @since 1.5
1150 */
1151 public static String quote(String s) {
1152 int slashEIndex = s.indexOf("\\E");
1153 if (slashEIndex == -1)
1154 return "\\Q" + s + "\\E";
1155
1156 StringBuilder sb = new StringBuilder(s.length() * 2);
1157 sb.append("\\Q");
1158 slashEIndex = 0;
1159 int current = 0;
1160 while ((slashEIndex = s.indexOf("\\E", current)) != -1) {
1161 sb.append(s.substring(current, slashEIndex));
1162 current = slashEIndex + 2;
1163 sb.append("\\E\\\\E\\Q");
1164 }
1165 sb.append(s.substring(current, s.length()));
1166 sb.append("\\E");
1167 return sb.toString();
1168 }
1169
1170 /**
1171 * Recompile the Pattern instance from a stream. The original pattern
1172 * string is read in and the object tree is recompiled from it.
1173 */
1174 private void readObject(java.io.ObjectInputStream s)
1175 throws java.io.IOException, ClassNotFoundException {
1176
1177 // Read in all fields
1178 s.defaultReadObject();
1179
1180 // Initialize counts
1181 capturingGroupCount = 1;
1182 localCount = 0;
1183
1184 // if length > 0, the Pattern is lazily compiled
1185 compiled = false;
1186 if (pattern.length() == 0) {
1187 root = new Start(lastAccept);
1188 matchRoot = lastAccept;
1189 compiled = true;
1190 }
1191 }
1192
1193 /**
1194 * This private constructor is used to create all Patterns. The pattern
1195 * string and match flags are all that is needed to completely describe
1196 * a Pattern. An empty pattern string results in an object tree with
1197 * only a Start node and a LastNode node.
1198 */
1199 private Pattern(String p, int f) {
1200 pattern = p;
1201 flags = f;
1202
1203 // Reset group index count
1204 capturingGroupCount = 1;
1205 localCount = 0;
1206
1207 if (pattern.length() > 0) {
1208 compile();
1209 } else {
1210 root = new Start(lastAccept);
1211 matchRoot = lastAccept;
1212 }
1213 }
1214
1215 /**
1216 * The pattern is converted to normalizedD form and then a pure group
1217 * is constructed to match canonical equivalences of the characters.
1218 */
1219 private void normalize() {
1220 boolean inCharClass = false;
1221 int lastCodePoint = -1;
1222
1223 // Convert pattern into normalizedD form
1224 normalizedPattern = Normalizer.normalize(pattern, Normalizer.Form.NFD);
1225 patternLength = normalizedPattern.length();
1226
1227 // Modify pattern to match canonical equivalences
1228 StringBuilder newPattern = new StringBuilder(patternLength);
1229 for(int i=0; i<patternLength; ) {
1230 int c = normalizedPattern.codePointAt(i);
1231 StringBuilder sequenceBuffer;
1232 if ((Character.getType(c) == Character.NON_SPACING_MARK)
1233 && (lastCodePoint != -1)) {
1234 sequenceBuffer = new StringBuilder();
1235 sequenceBuffer.appendCodePoint(lastCodePoint);
1236 sequenceBuffer.appendCodePoint(c);
1237 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1238 i += Character.charCount(c);
1239 if (i >= patternLength)
1240 break;
1241 c = normalizedPattern.codePointAt(i);
1242 sequenceBuffer.appendCodePoint(c);
1243 }
1244 String ea = produceEquivalentAlternation(
1245 sequenceBuffer.toString());
1246 newPattern.setLength(newPattern.length()-Character.charCount(lastCodePoint));
1247 newPattern.append("(?:").append(ea).append(")");
1248 } else if (c == '[' && lastCodePoint != '\\') {
1249 i = normalizeCharClass(newPattern, i);
1250 } else {
1251 newPattern.appendCodePoint(c);
1252 }
1253 lastCodePoint = c;
1254 i += Character.charCount(c);
1255 }
1256 normalizedPattern = newPattern.toString();
1257 }
1258
1259 /**
1260 * Complete the character class being parsed and add a set
1261 * of alternations to it that will match the canonical equivalences
1262 * of the characters within the class.
1263 */
1264 private int normalizeCharClass(StringBuilder newPattern, int i) {
1265 StringBuilder charClass = new StringBuilder();
1266 StringBuilder eq = null;
1267 int lastCodePoint = -1;
1268 String result;
1269
1270 i++;
1271 charClass.append("[");
1272 while(true) {
1273 int c = normalizedPattern.codePointAt(i);
1274 StringBuilder sequenceBuffer;
1275
1276 if (c == ']' && lastCodePoint != '\\') {
1277 charClass.append((char)c);
1278 break;
1279 } else if (Character.getType(c) == Character.NON_SPACING_MARK) {
1280 sequenceBuffer = new StringBuilder();
1281 sequenceBuffer.appendCodePoint(lastCodePoint);
1282 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1283 sequenceBuffer.appendCodePoint(c);
1284 i += Character.charCount(c);
1285 if (i >= normalizedPattern.length())
1286 break;
1287 c = normalizedPattern.codePointAt(i);
1288 }
1289 String ea = produceEquivalentAlternation(
1290 sequenceBuffer.toString());
1291
1292 charClass.setLength(charClass.length()-Character.charCount(lastCodePoint));
1293 if (eq == null)
1294 eq = new StringBuilder();
1295 eq.append('|');
1296 eq.append(ea);
1297 } else {
1298 charClass.appendCodePoint(c);
1299 i++;
1300 }
1301 if (i == normalizedPattern.length())
1302 throw error("Unclosed character class");
1303 lastCodePoint = c;
1304 }
1305
1306 if (eq != null) {
1307 result = "(?:"+charClass.toString()+eq.toString()+")";
1308 } else {
1309 result = charClass.toString();
1310 }
1311
1312 newPattern.append(result);
1313 return i;
1314 }
1315
1316 /**
1317 * Given a specific sequence composed of a regular character and
1318 * combining marks that follow it, produce the alternation that will
1319 * match all canonical equivalences of that sequence.
1320 */
1321 private String produceEquivalentAlternation(String source) {
1322 int len = countChars(source, 0, 1);
1323 if (source.length() == len)
1324 // source has one character.
1325 return source;
1326
1327 String base = source.substring(0,len);
1328 String combiningMarks = source.substring(len);
1329
1330 String[] perms = producePermutations(combiningMarks);
1331 StringBuilder result = new StringBuilder(source);
1332
1333 // Add combined permutations
1334 for(int x=0; x<perms.length; x++) {
1335 String next = base + perms[x];
1336 if (x>0)
1337 result.append("|"+next);
1338 next = composeOneStep(next);
1339 if (next != null)
1340 result.append("|"+produceEquivalentAlternation(next));
1341 }
1342 return result.toString();
1343 }
1344
1345 /**
1346 * Returns an array of strings that have all the possible
1347 * permutations of the characters in the input string.
1348 * This is used to get a list of all possible orderings
1349 * of a set of combining marks. Note that some of the permutations
1350 * are invalid because of combining class collisions, and these
1351 * possibilities must be removed because they are not canonically
1352 * equivalent.
1353 */
1354 private String[] producePermutations(String input) {
1355 if (input.length() == countChars(input, 0, 1))
1356 return new String[] {input};
1357
1358 if (input.length() == countChars(input, 0, 2)) {
1359 int c0 = Character.codePointAt(input, 0);
1360 int c1 = Character.codePointAt(input, Character.charCount(c0));
1361 if (getClass(c1) == getClass(c0)) {
1362 return new String[] {input};
1363 }
1364 String[] result = new String[2];
1365 result[0] = input;
1366 StringBuilder sb = new StringBuilder(2);
1367 sb.appendCodePoint(c1);
1368 sb.appendCodePoint(c0);
1369 result[1] = sb.toString();
1370 return result;
1371 }
1372
1373 int length = 1;
1374 int nCodePoints = countCodePoints(input);
1375 for(int x=1; x<nCodePoints; x++)
1376 length = length * (x+1);
1377
1378 String[] temp = new String[length];
1379
1380 int combClass[] = new int[nCodePoints];
1381 for(int x=0, i=0; x<nCodePoints; x++) {
1382 int c = Character.codePointAt(input, i);
1383 combClass[x] = getClass(c);
1384 i += Character.charCount(c);
1385 }
1386
1387 // For each char, take it out and add the permutations
1388 // of the remaining chars
1389 int index = 0;
1390 int len;
1391 // offset maintains the index in code units.
1392 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) {
1393 len = countChars(input, offset, 1);
1394 boolean skip = false;
1395 for(int y=x-1; y>=0; y--) {
1396 if (combClass[y] == combClass[x]) {
1397 continue loop;
1398 }
1399 }
1400 StringBuilder sb = new StringBuilder(input);
1401 String otherChars = sb.delete(offset, offset+len).toString();
1402 String[] subResult = producePermutations(otherChars);
1403
1404 String prefix = input.substring(offset, offset+len);
1405 for(int y=0; y<subResult.length; y++)
1406 temp[index++] = prefix + subResult[y];
1407 }
1408 String[] result = new String[index];
1409 for (int x=0; x<index; x++)
1410 result[x] = temp[x];
1411 return result;
1412 }
1413
1414 private int getClass(int c) {
1415 return sun.text.Normalizer.getCombiningClass(c);
1416 }
1417
1418 /**
1419 * Attempts to compose input by combining the first character
1420 * with the first combining mark following it. Returns a String
1421 * that is the composition of the leading character with its first
1422 * combining mark followed by the remaining combining marks. Returns
1423 * null if the first two characters cannot be further composed.
1424 */
1425 private String composeOneStep(String input) {
1426 int len = countChars(input, 0, 2);
1427 String firstTwoCharacters = input.substring(0, len);
1428 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC);
1429
1430 if (result.equals(firstTwoCharacters))
1431 return null;
1432 else {
1433 String remainder = input.substring(len);
1434 return result + remainder;
1435 }
1436 }
1437
1438 /**
1439 * Preprocess any \Q...\E sequences in `temp', meta-quoting them.
1440 * See the description of `quotemeta' in perlfunc(1).
1441 */
1442 private void RemoveQEQuoting() {
1443 final int pLen = patternLength;
1444 int i = 0;
1445 while (i < pLen-1) {
1446 if (temp[i] != '\\')
1447 i += 1;
1448 else if (temp[i + 1] != 'Q')
1449 i += 2;
1450 else
1451 break;
1452 }
1453 if (i >= pLen - 1) // No \Q sequence found
1454 return;
1455 int j = i;
1456 i += 2;
1457 int[] newtemp = new int[j + 2*(pLen-i) + 2];
1458 System.arraycopy(temp, 0, newtemp, 0, j);
1459
1460 boolean inQuote = true;
1461 while (i < pLen) {
1462 int c = temp[i++];
1463 if (! ASCII.isAscii(c) || ASCII.isAlnum(c)) {
1464 newtemp[j++] = c;
1465 } else if (c != '\\') {
1466 if (inQuote) newtemp[j++] = '\\';
1467 newtemp[j++] = c;
1468 } else if (inQuote) {
1469 if (temp[i] == 'E') {
1470 i++;
1471 inQuote = false;
1472 } else {
1473 newtemp[j++] = '\\';
1474 newtemp[j++] = '\\';
1475 }
1476 } else {
1477 if (temp[i] == 'Q') {
1478 i++;
1479 inQuote = true;
1480 } else {
1481 newtemp[j++] = c;
1482 if (i != pLen)
1483 newtemp[j++] = temp[i++];
1484 }
1485 }
1486 }
1487
1488 patternLength = j;
1489 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination
1490 }
1491
1492 /**
1493 * Copies regular expression to an int array and invokes the parsing
1494 * of the expression which will create the object tree.
1495 */
1496 private void compile() {
1497 // Handle canonical equivalences
1498 if (has(CANON_EQ) && !has(LITERAL)) {
1499 normalize();
1500 } else {
1501 normalizedPattern = pattern;
1502 }
1503 patternLength = normalizedPattern.length();
1504
1505 // Copy pattern to int array for convenience
1506 // Use double zero to terminate pattern
1507 temp = new int[patternLength + 2];
1508
1509 hasSupplementary = false;
1510 int c, count = 0;
1511 // Convert all chars into code points
1512 for (int x = 0; x < patternLength; x += Character.charCount(c)) {
1513 c = normalizedPattern.codePointAt(x);
1514 if (isSupplementary(c)) {
1515 hasSupplementary = true;
1516 }
1517 temp[count++] = c;
1518 }
1519
1520 patternLength = count; // patternLength now in code points
1521
1522 if (! has(LITERAL))
1523 RemoveQEQuoting();
1524
1525 // Allocate all temporary objects here.
1526 buffer = new int[32];
1527 groupNodes = new GroupHead[10];
1528 namedGroups = null;
1529
1530 if (has(LITERAL)) {
1531 // Literal pattern handling
1532 matchRoot = newSlice(temp, patternLength, hasSupplementary);
1533 matchRoot.next = lastAccept;
1534 } else {
1535 // Start recursive descent parsing
1536 matchRoot = expr(lastAccept);
1537 // Check extra pattern characters
1538 if (patternLength != cursor) {
1539 if (peek() == ')') {
1540 throw error("Unmatched closing ')'");
1541 } else {
1542 throw error("Unexpected internal error");
1543 }
1544 }
1545 }
1546
1547 // Peephole optimization
1548 if (matchRoot instanceof Slice) {
1549 root = BnM.optimize(matchRoot);
1550 if (root == matchRoot) {
1551 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1552 }
1553 } else if (matchRoot instanceof Begin || matchRoot instanceof First) {
1554 root = matchRoot;
1555 } else {
1556 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1557 }
1558
1559 // Release temporary storage
1560 temp = null;
1561 buffer = null;
1562 groupNodes = null;
1563 patternLength = 0;
1564 compiled = true;
1565 }
1566
1567 Map<String, Integer> namedGroups() {
1568 if (namedGroups == null)
1569 namedGroups = new HashMap<String, Integer>(2);
1570 return namedGroups;
1571 }
1572
1573 /**
1574 * Used to print out a subtree of the Pattern to help with debugging.
1575 */
1576 private static void printObjectTree(Node node) {
1577 while(node != null) {
1578 if (node instanceof Prolog) {
1579 System.out.println(node);
1580 printObjectTree(((Prolog)node).loop);
1581 System.out.println("**** end contents prolog loop");
1582 } else if (node instanceof Loop) {
1583 System.out.println(node);
1584 printObjectTree(((Loop)node).body);
1585 System.out.println("**** end contents Loop body");
1586 } else if (node instanceof Curly) {
1587 System.out.println(node);
1588 printObjectTree(((Curly)node).atom);
1589 System.out.println("**** end contents Curly body");
1590 } else if (node instanceof GroupCurly) {
1591 System.out.println(node);
1592 printObjectTree(((GroupCurly)node).atom);
1593 System.out.println("**** end contents GroupCurly body");
1594 } else if (node instanceof GroupTail) {
1595 System.out.println(node);
1596 System.out.println("Tail next is "+node.next);
1597 return;
1598 } else {
1599 System.out.println(node);
1600 }
1601 node = node.next;
1602 if (node != null)
1603 System.out.println("->next:");
1604 if (node == Pattern.accept) {
1605 System.out.println("Accept Node");
1606 node = null;
1607 }
1608 }
1609 }
1610
1611 /**
1612 * Used to accumulate information about a subtree of the object graph
1613 * so that optimizations can be applied to the subtree.
1614 */
1615 static final class TreeInfo {
1616 int minLength;
1617 int maxLength;
1618 boolean maxValid;
1619 boolean deterministic;
1620
1621 TreeInfo() {
1622 reset();
1623 }
1624 void reset() {
1625 minLength = 0;
1626 maxLength = 0;
1627 maxValid = true;
1628 deterministic = true;
1629 }
1630 }
1631
1632 /*
1633 * The following private methods are mainly used to improve the
1634 * readability of the code. In order to let the Java compiler easily
1635 * inline them, we should not put many assertions or error checks in them.
1636 */
1637
1638 /**
1639 * Indicates whether a particular flag is set or not.
1640 */
1641 private boolean has(int f) {
1642 return (flags & f) != 0;
1643 }
1644
1645 /**
1646 * Match next character, signal error if failed.
1647 */
1648 private void accept(int ch, String s) {
1649 int testChar = temp[cursor++];
1650 if (has(COMMENTS))
1651 testChar = parsePastWhitespace(testChar);
1652 if (ch != testChar) {
1653 throw error(s);
1654 }
1655 }
1656
1657 /**
1658 * Mark the end of pattern with a specific character.
1659 */
1660 private void mark(int c) {
1661 temp[patternLength] = c;
1662 }
1663
1664 /**
1665 * Peek the next character, and do not advance the cursor.
1666 */
1667 private int peek() {
1668 int ch = temp[cursor];
1669 if (has(COMMENTS))
1670 ch = peekPastWhitespace(ch);
1671 return ch;
1672 }
1673
1674 /**
1675 * Read the next character, and advance the cursor by one.
1676 */
1677 private int read() {
1678 int ch = temp[cursor++];
1679 if (has(COMMENTS))
1680 ch = parsePastWhitespace(ch);
1681 return ch;
1682 }
1683
1684 /**
1685 * Read the next character, and advance the cursor by one,
1686 * ignoring the COMMENTS setting
1687 */
1688 private int readEscaped() {
1689 int ch = temp[cursor++];
1690 return ch;
1691 }
1692
1693 /**
1694 * Advance the cursor by one, and peek the next character.
1695 */
1696 private int next() {
1697 int ch = temp[++cursor];
1698 if (has(COMMENTS))
1699 ch = peekPastWhitespace(ch);
1700 return ch;
1701 }
1702
1703 /**
1704 * Advance the cursor by one, and peek the next character,
1705 * ignoring the COMMENTS setting
1706 */
1707 private int nextEscaped() {
1708 int ch = temp[++cursor];
1709 return ch;
1710 }
1711
1712 /**
1713 * If in xmode peek past whitespace and comments.
1714 */
1715 private int peekPastWhitespace(int ch) {
1716 while (ASCII.isSpace(ch) || ch == '#') {
1717 while (ASCII.isSpace(ch))
1718 ch = temp[++cursor];
1719 if (ch == '#') {
1720 ch = peekPastLine();
1721 }
1722 }
1723 return ch;
1724 }
1725
1726 /**
1727 * If in xmode parse past whitespace and comments.
1728 */
1729 private int parsePastWhitespace(int ch) {
1730 while (ASCII.isSpace(ch) || ch == '#') {
1731 while (ASCII.isSpace(ch))
1732 ch = temp[cursor++];
1733 if (ch == '#')
1734 ch = parsePastLine();
1735 }
1736 return ch;
1737 }
1738
1739 /**
1740 * xmode parse past comment to end of line.
1741 */
1742 private int parsePastLine() {
1743 int ch = temp[cursor++];
1744 while (ch != 0 && !isLineSeparator(ch))
1745 ch = temp[cursor++];
1746 return ch;
1747 }
1748
1749 /**
1750 * xmode peek past comment to end of line.
1751 */
1752 private int peekPastLine() {
1753 int ch = temp[++cursor];
1754 while (ch != 0 && !isLineSeparator(ch))
1755 ch = temp[++cursor];
1756 return ch;
1757 }
1758
1759 /**
1760 * Determines if character is a line separator in the current mode
1761 */
1762 private boolean isLineSeparator(int ch) {
1763 if (has(UNIX_LINES)) {
1764 return ch == '\n';
1765 } else {
1766 return (ch == '\n' ||
1767 ch == '\r' ||
1768 (ch|1) == '\u2029' ||
1769 ch == '\u0085');
1770 }
1771 }
1772
1773 /**
1774 * Read the character after the next one, and advance the cursor by two.
1775 */
1776 private int skip() {
1777 int i = cursor;
1778 int ch = temp[i+1];
1779 cursor = i + 2;
1780 return ch;
1781 }
1782
1783 /**
1784 * Unread one next character, and retreat cursor by one.
1785 */
1786 private void unread() {
1787 cursor--;
1788 }
1789
1790 /**
1791 * Internal method used for handling all syntax errors. The pattern is
1792 * displayed with a pointer to aid in locating the syntax error.
1793 */
1794 private PatternSyntaxException error(String s) {
1795 return new PatternSyntaxException(s, normalizedPattern, cursor - 1);
1796 }
1797
1798 /**
1799 * Determines if there is any supplementary character or unpaired
1800 * surrogate in the specified range.
1801 */
1802 private boolean findSupplementary(int start, int end) {
1803 for (int i = start; i < end; i++) {
1804 if (isSupplementary(temp[i]))
1805 return true;
1806 }
1807 return false;
1808 }
1809
1810 /**
1811 * Determines if the specified code point is a supplementary
1812 * character or unpaired surrogate.
1813 */
1814 private static final boolean isSupplementary(int ch) {
1815 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT ||
1816 Character.isSurrogate((char)ch);
1817 }
1818
1819 /**
1820 * The following methods handle the main parsing. They are sorted
1821 * according to their precedence order, the lowest one first.
1822 */
1823
1824 /**
1825 * The expression is parsed with branch nodes added for alternations.
1826 * This may be called recursively to parse sub expressions that may
1827 * contain alternations.
1828 */
1829 private Node expr(Node end) {
1830 Node prev = null;
1831 Node firstTail = null;
1832 Node branchConn = null;
1833
1834 for (;;) {
1835 Node node = sequence(end);
1836 Node nodeTail = root; //double return
1837 if (prev == null) {
1838 prev = node;
1839 firstTail = nodeTail;
1840 } else {
1841 // Branch
1842 if (branchConn == null) {
1843 branchConn = new BranchConn();
1844 branchConn.next = end;
1845 }
1846 if (node == end) {
1847 // if the node returned from sequence() is "end"
1848 // we have an empty expr, set a null atom into
1849 // the branch to indicate to go "next" directly.
1850 node = null;
1851 } else {
1852 // the "tail.next" of each atom goes to branchConn
1853 nodeTail.next = branchConn;
1854 }
1855 if (prev instanceof Branch) {
1856 ((Branch)prev).add(node);
1857 } else {
1858 if (prev == end) {
1859 prev = null;
1860 } else {
1861 // replace the "end" with "branchConn" at its tail.next
1862 // when put the "prev" into the branch as the first atom.
1863 firstTail.next = branchConn;
1864 }
1865 prev = new Branch(prev, node, branchConn);
1866 }
1867 }
1868 if (peek() != '|') {
1869 return prev;
1870 }
1871 next();
1872 }
1873 }
1874
1875 /**
1876 * Parsing of sequences between alternations.
1877 */
1878 private Node sequence(Node end) {
1879 Node head = null;
1880 Node tail = null;
1881 Node node = null;
1882 LOOP:
1883 for (;;) {
1884 int ch = peek();
1885 switch (ch) {
1886 case '(':
1887 // Because group handles its own closure,
1888 // we need to treat it differently
1889 node = group0();
1890 // Check for comment or flag group
1891 if (node == null)
1892 continue;
1893 if (head == null)
1894 head = node;
1895 else
1896 tail.next = node;
1897 // Double return: Tail was returned in root
1898 tail = root;
1899 continue;
1900 case '[':
1901 node = clazz(true);
1902 break;
1903 case '\\':
1904 ch = nextEscaped();
1905 if (ch == 'p' || ch == 'P') {
1906 boolean oneLetter = true;
1907 boolean comp = (ch == 'P');
1908 ch = next(); // Consume { if present
1909 if (ch != '{') {
1910 unread();
1911 } else {
1912 oneLetter = false;
1913 }
1914 node = family(oneLetter, comp);
1915 } else {
1916 unread();
1917 node = atom();
1918 }
1919 break;
1920 case '^':
1921 next();
1922 if (has(MULTILINE)) {
1923 if (has(UNIX_LINES))
1924 node = new UnixCaret();
1925 else
1926 node = new Caret();
1927 } else {
1928 node = new Begin();
1929 }
1930 break;
1931 case '$':
1932 next();
1933 if (has(UNIX_LINES))
1934 node = new UnixDollar(has(MULTILINE));
1935 else
1936 node = new Dollar(has(MULTILINE));
1937 break;
1938 case '.':
1939 next();
1940 if (has(DOTALL)) {
1941 node = new All();
1942 } else {
1943 if (has(UNIX_LINES))
1944 node = new UnixDot();
1945 else {
1946 node = new Dot();
1947 }
1948 }
1949 break;
1950 case '|':
1951 case ')':
1952 break LOOP;
1953 case ']': // Now interpreting dangling ] and } as literals
1954 case '}':
1955 node = atom();
1956 break;
1957 case '?':
1958 case '*':
1959 case '+':
1960 next();
1961 throw error("Dangling meta character '" + ((char)ch) + "'");
1962 case 0:
1963 if (cursor >= patternLength) {
1964 break LOOP;
1965 }
1966 // Fall through
1967 default:
1968 node = atom();
1969 break;
1970 }
1971
1972 node = closure(node);
1973
1974 if (head == null) {
1975 head = tail = node;
1976 } else {
1977 tail.next = node;
1978 tail = node;
1979 }
1980 }
1981 if (head == null) {
1982 return end;
1983 }
1984 tail.next = end;
1985 root = tail; //double return
1986 return head;
1987 }
1988
1989 /**
1990 * Parse and add a new Single or Slice.
1991 */
1992 private Node atom() {
1993 int first = 0;
1994 int prev = -1;
1995 boolean hasSupplementary = false;
1996 int ch = peek();
1997 for (;;) {
1998 switch (ch) {
1999 case '*':
2000 case '+':
2001 case '?':
2002 case '{':
2003 if (first > 1) {
2004 cursor = prev; // Unwind one character
2005 first--;
2006 }
2007 break;
2008 case '$':
2009 case '.':
2010 case '^':
2011 case '(':
2012 case '[':
2013 case '|':
2014 case ')':
2015 break;
2016 case '\\':
2017 ch = nextEscaped();
2018 if (ch == 'p' || ch == 'P') { // Property
2019 if (first > 0) { // Slice is waiting; handle it first
2020 unread();
2021 break;
2022 } else { // No slice; just return the family node
2023 boolean comp = (ch == 'P');
2024 boolean oneLetter = true;
2025 ch = next(); // Consume { if present
2026 if (ch != '{')
2027 unread();
2028 else
2029 oneLetter = false;
2030 return family(oneLetter, comp);
2031 }
2032 }
2033 unread();
2034 prev = cursor;
2035 ch = escape(false, first == 0);
2036 if (ch >= 0) {
2037 append(ch, first);
2038 first++;
2039 if (isSupplementary(ch)) {
2040 hasSupplementary = true;
2041 }
2042 ch = peek();
2043 continue;
2044 } else if (first == 0) {
2045 return root;
2046 }
2047 // Unwind meta escape sequence
2048 cursor = prev;
2049 break;
2050 case 0:
2051 if (cursor >= patternLength) {
2052 break;
2053 }
2054 // Fall through
2055 default:
2056 prev = cursor;
2057 append(ch, first);
2058 first++;
2059 if (isSupplementary(ch)) {
2060 hasSupplementary = true;
2061 }
2062 ch = next();
2063 continue;
2064 }
2065 break;
2066 }
2067 if (first == 1) {
2068 return newSingle(buffer[0]);
2069 } else {
2070 return newSlice(buffer, first, hasSupplementary);
2071 }
2072 }
2073
2074 private void append(int ch, int len) {
2075 if (len >= buffer.length) {
2076 int[] tmp = new int[len+len];
2077 System.arraycopy(buffer, 0, tmp, 0, len);
2078 buffer = tmp;
2079 }
2080 buffer[len] = ch;
2081 }
2082
2083 /**
2084 * Parses a backref greedily, taking as many numbers as it
2085 * can. The first digit is always treated as a backref, but
2086 * multi digit numbers are only treated as a backref if at
2087 * least that many backrefs exist at this point in the regex.
2088 */
2089 private Node ref(int refNum) {
2090 boolean done = false;
2091 while(!done) {
2092 int ch = peek();
2093 switch(ch) {
2094 case '0':
2095 case '1':
2096 case '2':
2097 case '3':
2098 case '4':
2099 case '5':
2100 case '6':
2101 case '7':
2102 case '8':
2103 case '9':
2104 int newRefNum = (refNum * 10) + (ch - '0');
2105 // Add another number if it doesn't make a group
2106 // that doesn't exist
2107 if (capturingGroupCount - 1 < newRefNum) {
2108 done = true;
2109 break;
2110 }
2111 refNum = newRefNum;
2112 read();
2113 break;
2114 default:
2115 done = true;
2116 break;
2117 }
2118 }
2119 if (has(CASE_INSENSITIVE))
2120 return new CIBackRef(refNum, has(UNICODE_CASE));
2121 else
2122 return new BackRef(refNum);
2123 }
2124
2125 /**
2126 * Parses an escape sequence to determine the actual value that needs
2127 * to be matched.
2128 * If -1 is returned and create was true a new object was added to the tree
2129 * to handle the escape sequence.
2130 * If the returned value is greater than zero, it is the value that
2131 * matches the escape sequence.
2132 */
2133 private int escape(boolean inclass, boolean create) {
2134 int ch = skip();
2135 switch (ch) {
2136 case '0':
2137 return o();
2138 case '1':
2139 case '2':
2140 case '3':
2141 case '4':
2142 case '5':
2143 case '6':
2144 case '7':
2145 case '8':
2146 case '9':
2147 if (inclass) break;
2148 if (create) {
2149 root = ref((ch - '0'));
2150 }
2151 return -1;
2152 case 'A':
2153 if (inclass) break;
2154 if (create) root = new Begin();
2155 return -1;
2156 case 'B':
2157 if (inclass) break;
2158 if (create) root = new Bound(Bound.NONE);
2159 return -1;
2160 case 'C':
2161 break;
2162 case 'D':
2163 if (create) root = new Ctype(ASCII.DIGIT).complement();
2164 return -1;
2165 case 'E':
2166 case 'F':
2167 break;
2168 case 'G':
2169 if (inclass) break;
2170 if (create) root = new LastMatch();
2171 return -1;
2172 case 'H':
2173 case 'I':
2174 case 'J':
2175 case 'K':
2176 case 'L':
2177 case 'M':
2178 case 'N':
2179 case 'O':
2180 case 'P':
2181 case 'Q':
2182 case 'R':
2183 break;
2184 case 'S':
2185 if (create) root = new Ctype(ASCII.SPACE).complement();
2186 return -1;
2187 case 'T':
2188 case 'U':
2189 case 'V':
2190 break;
2191 case 'W':
2192 if (create) root = new Ctype(ASCII.WORD).complement();
2193 return -1;
2194 case 'X':
2195 case 'Y':
2196 break;
2197 case 'Z':
2198 if (inclass) break;
2199 if (create) {
2200 if (has(UNIX_LINES))
2201 root = new UnixDollar(false);
2202 else
2203 root = new Dollar(false);
2204 }
2205 return -1;
2206 case 'a':
2207 return '\007';
2208 case 'b':
2209 if (inclass) break;
2210 if (create) root = new Bound(Bound.BOTH);
2211 return -1;
2212 case 'c':
2213 return c();
2214 case 'd':
2215 if (create) root = new Ctype(ASCII.DIGIT);
2216 return -1;
2217 case 'e':
2218 return '\033';
2219 case 'f':
2220 return '\f';
2221 case 'g':
2222 case 'h':
2223 case 'i':
2224 case 'j':
2225 break;
2226 case 'k':
2227 if (inclass)
2228 break;
2229 if (read() != '<')
2230 throw error("\\k is not followed by '<' for named capturing group");
2231 String name = groupname(read());
2232 if (!namedGroups().containsKey(name))
2233 throw error("(named capturing group <"+ name+"> does not exit");
2234 if (create) {
2235 if (has(CASE_INSENSITIVE))
2236 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE));
2237 else
2238 root = new BackRef(namedGroups().get(name));
2239 }
2240 return -1;
2241 case 'l':
2242 case 'm':
2243 break;
2244 case 'n':
2245 return '\n';
2246 case 'o':
2247 case 'p':
2248 case 'q':
2249 break;
2250 case 'r':
2251 return '\r';
2252 case 's':
2253 if (create) root = new Ctype(ASCII.SPACE);
2254 return -1;
2255 case 't':
2256 return '\t';
2257 case 'u':
2258 return u();
2259 case 'v':
2260 return '\013';
2261 case 'w':
2262 if (create) root = new Ctype(ASCII.WORD);
2263 return -1;
2264 case 'x':
2265 return x();
2266 case 'y':
2267 break;
2268 case 'z':
2269 if (inclass) break;
2270 if (create) root = new End();
2271 return -1;
2272 default:
2273 return ch;
2274 }
2275 throw error("Illegal/unsupported escape sequence");
2276 }
2277
2278 /**
2279 * Parse a character class, and return the node that matches it.
2280 *
2281 * Consumes a ] on the way out if consume is true. Usually consume
2282 * is true except for the case of [abc&&def] where def is a separate
2283 * right hand node with "understood" brackets.
2284 */
2285 private CharProperty clazz(boolean consume) {
2286 CharProperty prev = null;
2287 CharProperty node = null;
2288 BitClass bits = new BitClass();
2289 boolean include = true;
2290 boolean firstInClass = true;
2291 int ch = next();
2292 for (;;) {
2293 switch (ch) {
2294 case '^':
2295 // Negates if first char in a class, otherwise literal
2296 if (firstInClass) {
2297 if (temp[cursor-1] != '[')
2298 break;
2299 ch = next();
2300 include = !include;
2301 continue;
2302 } else {
2303 // ^ not first in class, treat as literal
2304 break;
2305 }
2306 case '[':
2307 firstInClass = false;
2308 node = clazz(true);
2309 if (prev == null)
2310 prev = node;
2311 else
2312 prev = union(prev, node);
2313 ch = peek();
2314 continue;
2315 case '&':
2316 firstInClass = false;
2317 ch = next();
2318 if (ch == '&') {
2319 ch = next();
2320 CharProperty rightNode = null;
2321 while (ch != ']' && ch != '&') {
2322 if (ch == '[') {
2323 if (rightNode == null)
2324 rightNode = clazz(true);
2325 else
2326 rightNode = union(rightNode, clazz(true));
2327 } else { // abc&&def
2328 unread();
2329 rightNode = clazz(false);
2330 }
2331 ch = peek();
2332 }
2333 if (rightNode != null)
2334 node = rightNode;
2335 if (prev == null) {
2336 if (rightNode == null)
2337 throw error("Bad class syntax");
2338 else
2339 prev = rightNode;
2340 } else {
2341 prev = intersection(prev, node);
2342 }
2343 } else {
2344 // treat as a literal &
2345 unread();
2346 break;
2347 }
2348 continue;
2349 case 0:
2350 firstInClass = false;
2351 if (cursor >= patternLength)
2352 throw error("Unclosed character class");
2353 break;
2354 case ']':
2355 firstInClass = false;
2356 if (prev != null) {
2357 if (consume)
2358 next();
2359 return prev;
2360 }
2361 break;
2362 default:
2363 firstInClass = false;
2364 break;
2365 }
2366 node = range(bits);
2367 if (include) {
2368 if (prev == null) {
2369 prev = node;
2370 } else {
2371 if (prev != node)
2372 prev = union(prev, node);
2373 }
2374 } else {
2375 if (prev == null) {
2376 prev = node.complement();
2377 } else {
2378 if (prev != node)
2379 prev = setDifference(prev, node);
2380 }
2381 }
2382 ch = peek();
2383 }
2384 }
2385
2386 private CharProperty bitsOrSingle(BitClass bits, int ch) {
2387 /* Bits can only handle codepoints in [u+0000-u+00ff] range.
2388 Use "single" node instead of bits when dealing with unicode
2389 case folding for codepoints listed below.
2390 (1)Uppercase out of range: u+00ff, u+00b5
2391 toUpperCase(u+00ff) -> u+0178
2392 toUpperCase(u+00b5) -> u+039c
2393 (2)LatinSmallLetterLongS u+17f
2394 toUpperCase(u+017f) -> u+0053
2395 (3)LatinSmallLetterDotlessI u+131
2396 toUpperCase(u+0131) -> u+0049
2397 (4)LatinCapitalLetterIWithDotAbove u+0130
2398 toLowerCase(u+0130) -> u+0069
2399 (5)KelvinSign u+212a
2400 toLowerCase(u+212a) ==> u+006B
2401 (6)AngstromSign u+212b
2402 toLowerCase(u+212b) ==> u+00e5
2403 */
2404 int d;
2405 if (ch < 256 &&
2406 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) &&
2407 (ch == 0xff || ch == 0xb5 ||
2408 ch == 0x49 || ch == 0x69 || //I and i
2409 ch == 0x53 || ch == 0x73 || //S and s
2410 ch == 0x4b || ch == 0x6b || //K and k
2411 ch == 0xc5 || ch == 0xe5))) //A+ring
2412 return bits.add(ch, flags());
2413 return newSingle(ch);
2414 }
2415
2416 /**
2417 * Parse a single character or a character range in a character class
2418 * and return its representative node.
2419 */
2420 private CharProperty range(BitClass bits) {
2421 int ch = peek();
2422 if (ch == '\\') {
2423 ch = nextEscaped();
2424 if (ch == 'p' || ch == 'P') { // A property
2425 boolean comp = (ch == 'P');
2426 boolean oneLetter = true;
2427 // Consume { if present
2428 ch = next();
2429 if (ch != '{')
2430 unread();
2431 else
2432 oneLetter = false;
2433 return family(oneLetter, comp);
2434 } else { // ordinary escape
2435 unread();
2436 ch = escape(true, true);
2437 if (ch == -1)
2438 return (CharProperty) root;
2439 }
2440 } else {
2441 ch = single();
2442 }
2443 if (ch >= 0) {
2444 if (peek() == '-') {
2445 int endRange = temp[cursor+1];
2446 if (endRange == '[') {
2447 return bitsOrSingle(bits, ch);
2448 }
2449 if (endRange != ']') {
2450 next();
2451 int m = single();
2452 if (m < ch)
2453 throw error("Illegal character range");
2454 if (has(CASE_INSENSITIVE))
2455 return caseInsensitiveRangeFor(ch, m);
2456 else
2457 return rangeFor(ch, m);
2458 }
2459 }
2460 return bitsOrSingle(bits, ch);
2461 }
2462 throw error("Unexpected character '"+((char)ch)+"'");
2463 }
2464
2465 private int single() {
2466 int ch = peek();
2467 switch (ch) {
2468 case '\\':
2469 return escape(true, false);
2470 default:
2471 next();
2472 return ch;
2473 }
2474 }
2475
2476 /**
2477 * Parses a Unicode character family and returns its representative node.
2478 */
2479 private CharProperty family(boolean singleLetter,
2480 boolean maybeComplement)
2481 {
2482 next();
2483 String name;
2484 CharProperty node;
2485
2486 if (singleLetter) {
2487 int c = temp[cursor];
2488 if (!Character.isSupplementaryCodePoint(c)) {
2489 name = String.valueOf((char)c);
2490 } else {
2491 name = new String(temp, cursor, 1);
2492 }
2493 read();
2494 } else {
2495 int i = cursor;
2496 mark('}');
2497 while(read() != '}') {
2498 }
2499 mark('\000');
2500 int j = cursor;
2501 if (j > patternLength)
2502 throw error("Unclosed character family");
2503 if (i + 1 >= j)
2504 throw error("Empty character family");
2505 name = new String(temp, i, j-i-1);
2506 }
2507
2508 int i = name.indexOf('=');
2509 if (i != -1) {
2510 // property construct \p{name=value}
2511 String value = name.substring(i + 1);
2512 name = name.substring(0, i).toLowerCase(Locale.ENGLISH);
2513 if ("sc".equals(name) || "script".equals(name)) {
2514 node = unicodeScriptPropertyFor(value);
2515 } else if ("blk".equals(name) || "block".equals(name)) {
2516 node = unicodeBlockPropertyFor(value);
2517 } else if ("gc".equals(name) || "general_category".equals(name)) {
2518 node = charPropertyNodeFor(value);
2519 } else {
2520 throw error("Unknown Unicode property {name=<" + name + ">, "
2521 + "value=<" + value + ">}");
2522 }
2523 } else {
2524 if (name.startsWith("In")) {
2525 // \p{inBlockName}
2526 node = unicodeBlockPropertyFor(name.substring(2));
2527 } else if (name.startsWith("Is")) {
2528 // \p{isGeneralCategory} and \p{isScriptName}
2529 name = name.substring(2);
2530 node = CharPropertyNames.charPropertyFor(name);
2531 if (node == null)
2532 node = unicodeScriptPropertyFor(name);
2533 } else {
2534 node = charPropertyNodeFor(name);
2535 }
2536 }
2537 if (maybeComplement) {
2538 if (node instanceof Category || node instanceof Block)
2539 hasSupplementary = true;
2540 node = node.complement();
2541 }
2542 return node;
2543 }
2544
2545
2546 /**
2547 * Returns a CharProperty matching all characters belong to
2548 * a UnicodeScript.
2549 */
2550 private CharProperty unicodeScriptPropertyFor(String name) {
2551 final Character.UnicodeScript script;
2552 try {
2553 script = Character.UnicodeScript.forName(name);
2554 } catch (IllegalArgumentException iae) {
2555 throw error("Unknown character script name {" + name + "}");
2556 }
2557 return new Script(script);
2558 }
2559
2560 /**
2561 * Returns a CharProperty matching all characters in a UnicodeBlock.
2562 */
2563 private CharProperty unicodeBlockPropertyFor(String name) {
2564 final Character.UnicodeBlock block;
2565 try {
2566 block = Character.UnicodeBlock.forName(name);
2567 } catch (IllegalArgumentException iae) {
2568 throw error("Unknown character block name {" + name + "}");
2569 }
2570 return new Block(block);
2571 }
2572
2573 /**
2574 * Returns a CharProperty matching all characters in a named property.
2575 */
2576 private CharProperty charPropertyNodeFor(String name) {
2577 CharProperty p = CharPropertyNames.charPropertyFor(name);
2578 if (p == null)
2579 throw error("Unknown character property name {" + name + "}");
2580 return p;
2581 }
2582
2583 /**
2584 * Parses and returns the name of a "named capturing group", the trailing
2585 * ">" is consumed after parsing.
2586 */
2587 private String groupname(int ch) {
2588 StringBuilder sb = new StringBuilder();
2589 sb.append(Character.toChars(ch));
2590 while (ASCII.isLower(ch=read()) || ASCII.isUpper(ch) ||
2591 ASCII.isDigit(ch)) {
2592 sb.append(Character.toChars(ch));
2593 }
2594 if (sb.length() == 0)
2595 throw error("named capturing group has 0 length name");
2596 if (ch != '>')
2597 throw error("named capturing group is missing trailing '>'");
2598 return sb.toString();
2599 }
2600
2601 /**
2602 * Parses a group and returns the head node of a set of nodes that process
2603 * the group. Sometimes a double return system is used where the tail is
2604 * returned in root.
2605 */
2606 private Node group0() {
2607 boolean capturingGroup = false;
2608 Node head = null;
2609 Node tail = null;
2610 int save = flags;
2611 root = null;
2612 int ch = next();
2613 if (ch == '?') {
2614 ch = skip();
2615 switch (ch) {
2616 case ':': // (?:xxx) pure group
2617 head = createGroup(true);
2618 tail = root;
2619 head.next = expr(tail);
2620 break;
2621 case '=': // (?=xxx) and (?!xxx) lookahead
2622 case '!':
2623 head = createGroup(true);
2624 tail = root;
2625 head.next = expr(tail);
2626 if (ch == '=') {
2627 head = tail = new Pos(head);
2628 } else {
2629 head = tail = new Neg(head);
2630 }
2631 break;
2632 case '>': // (?>xxx) independent group
2633 head = createGroup(true);
2634 tail = root;
2635 head.next = expr(tail);
2636 head = tail = new Ques(head, INDEPENDENT);
2637 break;
2638 case '<': // (?<xxx) look behind
2639 ch = read();
2640 if (ASCII.isLower(ch) || ASCII.isUpper(ch)) {
2641 // named captured group
2642 String name = groupname(ch);
2643 if (namedGroups().containsKey(name))
2644 throw error("Named capturing group <" + name
2645 + "> is already defined");
2646 capturingGroup = true;
2647 head = createGroup(false);
2648 tail = root;
2649 namedGroups().put(name, capturingGroupCount-1);
2650 head.next = expr(tail);
2651 break;
2652 }
2653 int start = cursor;
2654 head = createGroup(true);
2655 tail = root;
2656 head.next = expr(tail);
2657 tail.next = lookbehindEnd;
2658 TreeInfo info = new TreeInfo();
2659 head.study(info);
2660 if (info.maxValid == false) {
2661 throw error("Look-behind group does not have "
2662 + "an obvious maximum length");
2663 }
2664 boolean hasSupplementary = findSupplementary(start, patternLength);
2665 if (ch == '=') {
2666 head = tail = (hasSupplementary ?
2667 new BehindS(head, info.maxLength,
2668 info.minLength) :
2669 new Behind(head, info.maxLength,
2670 info.minLength));
2671 } else if (ch == '!') {
2672 head = tail = (hasSupplementary ?
2673 new NotBehindS(head, info.maxLength,
2674 info.minLength) :
2675 new NotBehind(head, info.maxLength,
2676 info.minLength));
2677 } else {
2678 throw error("Unknown look-behind group");
2679 }
2680 break;
2681 case '$':
2682 case '@':
2683 throw error("Unknown group type");
2684 default: // (?xxx:) inlined match flags
2685 unread();
2686 addFlag();
2687 ch = read();
2688 if (ch == ')') {
2689 return null; // Inline modifier only
2690 }
2691 if (ch != ':') {
2692 throw error("Unknown inline modifier");
2693 }
2694 head = createGroup(true);
2695 tail = root;
2696 head.next = expr(tail);
2697 break;
2698 }
2699 } else { // (xxx) a regular group
2700 capturingGroup = true;
2701 head = createGroup(false);
2702 tail = root;
2703 head.next = expr(tail);
2704 }
2705
2706 accept(')', "Unclosed group");
2707 flags = save;
2708
2709 // Check for quantifiers
2710 Node node = closure(head);
2711 if (node == head) { // No closure
2712 root = tail;
2713 return node; // Dual return
2714 }
2715 if (head == tail) { // Zero length assertion
2716 root = node;
2717 return node; // Dual return
2718 }
2719
2720 if (node instanceof Ques) {
2721 Ques ques = (Ques) node;
2722 if (ques.type == POSSESSIVE) {
2723 root = node;
2724 return node;
2725 }
2726 tail.next = new BranchConn();
2727 tail = tail.next;
2728 if (ques.type == GREEDY) {
2729 head = new Branch(head, null, tail);
2730 } else { // Reluctant quantifier
2731 head = new Branch(null, head, tail);
2732 }
2733 root = tail;
2734 return head;
2735 } else if (node instanceof Curly) {
2736 Curly curly = (Curly) node;
2737 if (curly.type == POSSESSIVE) {
2738 root = node;
2739 return node;
2740 }
2741 // Discover if the group is deterministic
2742 TreeInfo info = new TreeInfo();
2743 if (head.study(info)) { // Deterministic
2744 GroupTail temp = (GroupTail) tail;
2745 head = root = new GroupCurly(head.next, curly.cmin,
2746 curly.cmax, curly.type,
2747 ((GroupTail)tail).localIndex,
2748 ((GroupTail)tail).groupIndex,
2749 capturingGroup);
2750 return head;
2751 } else { // Non-deterministic
2752 int temp = ((GroupHead) head).localIndex;
2753 Loop loop;
2754 if (curly.type == GREEDY)
2755 loop = new Loop(this.localCount, temp);
2756 else // Reluctant Curly
2757 loop = new LazyLoop(this.localCount, temp);
2758 Prolog prolog = new Prolog(loop);
2759 this.localCount += 1;
2760 loop.cmin = curly.cmin;
2761 loop.cmax = curly.cmax;
2762 loop.body = head;
2763 tail.next = loop;
2764 root = loop;
2765 return prolog; // Dual return
2766 }
2767 }
2768 throw error("Internal logic error");
2769 }
2770
2771 /**
2772 * Create group head and tail nodes using double return. If the group is
2773 * created with anonymous true then it is a pure group and should not
2774 * affect group counting.
2775 */
2776 private Node createGroup(boolean anonymous) {
2777 int localIndex = localCount++;
2778 int groupIndex = 0;
2779 if (!anonymous)
2780 groupIndex = capturingGroupCount++;
2781 GroupHead head = new GroupHead(localIndex);
2782 root = new GroupTail(localIndex, groupIndex);
2783 if (!anonymous && groupIndex < 10)
2784 groupNodes[groupIndex] = head;
2785 return head;
2786 }
2787
2788 /**
2789 * Parses inlined match flags and set them appropriately.
2790 */
2791 private void addFlag() {
2792 int ch = peek();
2793 for (;;) {
2794 switch (ch) {
2795 case 'i':
2796 flags |= CASE_INSENSITIVE;
2797 break;
2798 case 'm':
2799 flags |= MULTILINE;
2800 break;
2801 case 's':
2802 flags |= DOTALL;
2803 break;
2804 case 'd':
2805 flags |= UNIX_LINES;
2806 break;
2807 case 'u':
2808 flags |= UNICODE_CASE;
2809 break;
2810 case 'c':
2811 flags |= CANON_EQ;
2812 break;
2813 case 'x':
2814 flags |= COMMENTS;
2815 break;
2816 case '-': // subFlag then fall through
2817 ch = next();
2818 subFlag();
2819 default:
2820 return;
2821 }
2822 ch = next();
2823 }
2824 }
2825
2826 /**
2827 * Parses the second part of inlined match flags and turns off
2828 * flags appropriately.
2829 */
2830 private void subFlag() {
2831 int ch = peek();
2832 for (;;) {
2833 switch (ch) {
2834 case 'i':
2835 flags &= ~CASE_INSENSITIVE;
2836 break;
2837 case 'm':
2838 flags &= ~MULTILINE;
2839 break;
2840 case 's':
2841 flags &= ~DOTALL;
2842 break;
2843 case 'd':
2844 flags &= ~UNIX_LINES;
2845 break;
2846 case 'u':
2847 flags &= ~UNICODE_CASE;
2848 break;
2849 case 'c':
2850 flags &= ~CANON_EQ;
2851 break;
2852 case 'x':
2853 flags &= ~COMMENTS;
2854 break;
2855 default:
2856 return;
2857 }
2858 ch = next();
2859 }
2860 }
2861
2862 static final int MAX_REPS = 0x7FFFFFFF;
2863
2864 static final int GREEDY = 0;
2865
2866 static final int LAZY = 1;
2867
2868 static final int POSSESSIVE = 2;
2869
2870 static final int INDEPENDENT = 3;
2871
2872 /**
2873 * Processes repetition. If the next character peeked is a quantifier
2874 * then new nodes must be appended to handle the repetition.
2875 * Prev could be a single or a group, so it could be a chain of nodes.
2876 */
2877 private Node closure(Node prev) {
2878 Node atom;
2879 int ch = peek();
2880 switch (ch) {
2881 case '?':
2882 ch = next();
2883 if (ch == '?') {
2884 next();
2885 return new Ques(prev, LAZY);
2886 } else if (ch == '+') {
2887 next();
2888 return new Ques(prev, POSSESSIVE);
2889 }
2890 return new Ques(prev, GREEDY);
2891 case '*':
2892 ch = next();
2893 if (ch == '?') {
2894 next();
2895 return new Curly(prev, 0, MAX_REPS, LAZY);
2896 } else if (ch == '+') {
2897 next();
2898 return new Curly(prev, 0, MAX_REPS, POSSESSIVE);
2899 }
2900 return new Curly(prev, 0, MAX_REPS, GREEDY);
2901 case '+':
2902 ch = next();
2903 if (ch == '?') {
2904 next();
2905 return new Curly(prev, 1, MAX_REPS, LAZY);
2906 } else if (ch == '+') {
2907 next();
2908 return new Curly(prev, 1, MAX_REPS, POSSESSIVE);
2909 }
2910 return new Curly(prev, 1, MAX_REPS, GREEDY);
2911 case '{':
2912 ch = temp[cursor+1];
2913 if (ASCII.isDigit(ch)) {
2914 skip();
2915 int cmin = 0;
2916 do {
2917 cmin = cmin * 10 + (ch - '0');
2918 } while (ASCII.isDigit(ch = read()));
2919 int cmax = cmin;
2920 if (ch == ',') {
2921 ch = read();
2922 cmax = MAX_REPS;
2923 if (ch != '}') {
2924 cmax = 0;
2925 while (ASCII.isDigit(ch)) {
2926 cmax = cmax * 10 + (ch - '0');
2927 ch = read();
2928 }
2929 }
2930 }
2931 if (ch != '}')
2932 throw error("Unclosed counted closure");
2933 if (((cmin) | (cmax) | (cmax - cmin)) < 0)
2934 throw error("Illegal repetition range");
2935 Curly curly;
2936 ch = peek();
2937 if (ch == '?') {
2938 next();
2939 curly = new Curly(prev, cmin, cmax, LAZY);
2940 } else if (ch == '+') {
2941 next();
2942 curly = new Curly(prev, cmin, cmax, POSSESSIVE);
2943 } else {
2944 curly = new Curly(prev, cmin, cmax, GREEDY);
2945 }
2946 return curly;
2947 } else {
2948 throw error("Illegal repetition");
2949 }
2950 default:
2951 return prev;
2952 }
2953 }
2954
2955 /**
2956 * Utility method for parsing control escape sequences.
2957 */
2958 private int c() {
2959 if (cursor < patternLength) {
2960 return read() ^ 64;
2961 }
2962 throw error("Illegal control escape sequence");
2963 }
2964
2965 /**
2966 * Utility method for parsing octal escape sequences.
2967 */
2968 private int o() {
2969 int n = read();
2970 if (((n-'0')|('7'-n)) >= 0) {
2971 int m = read();
2972 if (((m-'0')|('7'-m)) >= 0) {
2973 int o = read();
2974 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) {
2975 return (n - '0') * 64 + (m - '0') * 8 + (o - '0');
2976 }
2977 unread();
2978 return (n - '0') * 8 + (m - '0');
2979 }
2980 unread();
2981 return (n - '0');
2982 }
2983 throw error("Illegal octal escape sequence");
2984 }
2985
2986 /**
2987 * Utility method for parsing hexadecimal escape sequences.
2988 */
2989 private int x() {
2990 int n = read();
2991 if (ASCII.isHexDigit(n)) {
2992 int m = read();
2993 if (ASCII.isHexDigit(m)) {
2994 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m);
2995 }
2996 }
2997 throw error("Illegal hexadecimal escape sequence");
2998 }
2999
3000 /**
3001 * Utility method for parsing unicode escape sequences.
3002 */
3003 private int cursor() {
3004 return cursor;
3005 }
3006
3007 private void setcursor(int pos) {
3008 cursor = pos;
3009 }
3010
3011 private int uxxxx() {
3012 int n = 0;
3013 for (int i = 0; i < 4; i++) {
3014 int ch = read();
3015 if (!ASCII.isHexDigit(ch)) {
3016 throw error("Illegal Unicode escape sequence");
3017 }
3018 n = n * 16 + ASCII.toDigit(ch);
3019 }
3020 return n;
3021 }
3022
3023 private int u() {
3024 int n = uxxxx();
3025 if (Character.isHighSurrogate((char)n)) {
3026 int cur = cursor();
3027 if (read() == '\\' && read() == 'u') {
3028 int n2 = uxxxx();
3029 if (Character.isLowSurrogate((char)n2))
3030 return Character.toCodePoint((char)n, (char)n2);
3031 }
3032 setcursor(cur);
3033 }
3034 return n;
3035 }
3036
3037 //
3038 // Utility methods for code point support
3039 //
3040
3041 private static final int countChars(CharSequence seq, int index,
3042 int lengthInCodePoints) {
3043 // optimization
3044 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) {
3045 assert (index >= 0 && index < seq.length());
3046 return 1;
3047 }
3048 int length = seq.length();
3049 int x = index;
3050 if (lengthInCodePoints >= 0) {
3051 assert (index >= 0 && index < length);
3052 for (int i = 0; x < length && i < lengthInCodePoints; i++) {
3053 if (Character.isHighSurrogate(seq.charAt(x++))) {
3054 if (x < length && Character.isLowSurrogate(seq.charAt(x))) {
3055 x++;
3056 }
3057 }
3058 }
3059 return x - index;
3060 }
3061
3062 assert (index >= 0 && index <= length);
3063 if (index == 0) {
3064 return 0;
3065 }
3066 int len = -lengthInCodePoints;
3067 for (int i = 0; x > 0 && i < len; i++) {
3068 if (Character.isLowSurrogate(seq.charAt(--x))) {
3069 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) {
3070 x--;
3071 }
3072 }
3073 }
3074 return index - x;
3075 }
3076
3077 private static final int countCodePoints(CharSequence seq) {
3078 int length = seq.length();
3079 int n = 0;
3080 for (int i = 0; i < length; ) {
3081 n++;
3082 if (Character.isHighSurrogate(seq.charAt(i++))) {
3083 if (i < length && Character.isLowSurrogate(seq.charAt(i))) {
3084 i++;
3085 }
3086 }
3087 }
3088 return n;
3089 }
3090
3091 /**
3092 * Creates a bit vector for matching Latin-1 values. A normal BitClass
3093 * never matches values above Latin-1, and a complemented BitClass always
3094 * matches values above Latin-1.
3095 */
3096 private static final class BitClass extends BmpCharProperty {
3097 final boolean[] bits;
3098 BitClass() { bits = new boolean[256]; }
3099 private BitClass(boolean[] bits) { this.bits = bits; }
3100 BitClass add(int c, int flags) {
3101 assert c >= 0 && c <= 255;
3102 if ((flags & CASE_INSENSITIVE) != 0) {
3103 if (ASCII.isAscii(c)) {
3104 bits[ASCII.toUpper(c)] = true;
3105 bits[ASCII.toLower(c)] = true;
3106 } else if ((flags & UNICODE_CASE) != 0) {
3107 bits[Character.toLowerCase(c)] = true;
3108 bits[Character.toUpperCase(c)] = true;
3109 }
3110 }
3111 bits[c] = true;
3112 return this;
3113 }
3114 boolean isSatisfiedBy(int ch) {
3115 return ch < 256 && bits[ch];
3116 }
3117 }
3118
3119 /**
3120 * Returns a suitably optimized, single character matcher.
3121 */
3122 private CharProperty newSingle(final int ch) {
3123 if (has(CASE_INSENSITIVE)) {
3124 int lower, upper;
3125 if (has(UNICODE_CASE)) {
3126 upper = Character.toUpperCase(ch);
3127 lower = Character.toLowerCase(upper);
3128 if (upper != lower)
3129 return new SingleU(lower);
3130 } else if (ASCII.isAscii(ch)) {
3131 lower = ASCII.toLower(ch);
3132 upper = ASCII.toUpper(ch);
3133 if (lower != upper)
3134 return new SingleI(lower, upper);
3135 }
3136 }
3137 if (isSupplementary(ch))
3138 return new SingleS(ch); // Match a given Unicode character
3139 return new Single(ch); // Match a given BMP character
3140 }
3141
3142 /**
3143 * Utility method for creating a string slice matcher.
3144 */
3145 private Node newSlice(int[] buf, int count, boolean hasSupplementary) {
3146 int[] tmp = new int[count];
3147 if (has(CASE_INSENSITIVE)) {
3148 if (has(UNICODE_CASE)) {
3149 for (int i = 0; i < count; i++) {
3150 tmp[i] = Character.toLowerCase(
3151 Character.toUpperCase(buf[i]));
3152 }
3153 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp);
3154 }
3155 for (int i = 0; i < count; i++) {
3156 tmp[i] = ASCII.toLower(buf[i]);
3157 }
3158 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp);
3159 }
3160 for (int i = 0; i < count; i++) {
3161 tmp[i] = buf[i];
3162 }
3163 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp);
3164 }
3165
3166 /**
3167 * The following classes are the building components of the object
3168 * tree that represents a compiled regular expression. The object tree
3169 * is made of individual elements that handle constructs in the Pattern.
3170 * Each type of object knows how to match its equivalent construct with
3171 * the match() method.
3172 */
3173
3174 /**
3175 * Base class for all node classes. Subclasses should override the match()
3176 * method as appropriate. This class is an accepting node, so its match()
3177 * always returns true.
3178 */
3179 static class Node extends Object {
3180 Node next;
3181 Node() {
3182 next = Pattern.accept;
3183 }
3184 /**
3185 * This method implements the classic accept node.
3186 */
3187 boolean match(Matcher matcher, int i, CharSequence seq) {
3188 matcher.last = i;
3189 matcher.groups[0] = matcher.first;
3190 matcher.groups[1] = matcher.last;
3191 return true;
3192 }
3193 /**
3194 * This method is good for all zero length assertions.
3195 */
3196 boolean study(TreeInfo info) {
3197 if (next != null) {
3198 return next.study(info);
3199 } else {
3200 return info.deterministic;
3201 }
3202 }
3203 }
3204
3205 static class LastNode extends Node {
3206 /**
3207 * This method implements the classic accept node with
3208 * the addition of a check to see if the match occurred
3209 * using all of the input.
3210 */
3211 boolean match(Matcher matcher, int i, CharSequence seq) {
3212 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to)
3213 return false;
3214 matcher.last = i;
3215 matcher.groups[0] = matcher.first;
3216 matcher.groups[1] = matcher.last;
3217 return true;
3218 }
3219 }
3220
3221 /**
3222 * Used for REs that can start anywhere within the input string.
3223 * This basically tries to match repeatedly at each spot in the
3224 * input string, moving forward after each try. An anchored search
3225 * or a BnM will bypass this node completely.
3226 */
3227 static class Start extends Node {
3228 int minLength;
3229 Start(Node node) {
3230 this.next = node;
3231 TreeInfo info = new TreeInfo();
3232 next.study(info);
3233 minLength = info.minLength;
3234 }
3235 boolean match(Matcher matcher, int i, CharSequence seq) {
3236 if (i > matcher.to - minLength) {
3237 matcher.hitEnd = true;
3238 return false;
3239 }
3240 int guard = matcher.to - minLength;
3241 for (; i <= guard; i++) {
3242 if (next.match(matcher, i, seq)) {
3243 matcher.first = i;
3244 matcher.groups[0] = matcher.first;
3245 matcher.groups[1] = matcher.last;
3246 return true;
3247 }
3248 }
3249 matcher.hitEnd = true;
3250 return false;
3251 }
3252 boolean study(TreeInfo info) {
3253 next.study(info);
3254 info.maxValid = false;
3255 info.deterministic = false;
3256 return false;
3257 }
3258 }
3259
3260 /*
3261 * StartS supports supplementary characters, including unpaired surrogates.
3262 */
3263 static final class StartS extends Start {
3264 StartS(Node node) {
3265 super(node);
3266 }
3267 boolean match(Matcher matcher, int i, CharSequence seq) {
3268 if (i > matcher.to - minLength) {
3269 matcher.hitEnd = true;
3270 return false;
3271 }
3272 int guard = matcher.to - minLength;
3273 while (i <= guard) {
3274 //if ((ret = next.match(matcher, i, seq)) || i == guard)
3275 if (next.match(matcher, i, seq)) {
3276 matcher.first = i;
3277 matcher.groups[0] = matcher.first;
3278 matcher.groups[1] = matcher.last;
3279 return true;
3280 }
3281 if (i == guard)
3282 break;
3283 // Optimization to move to the next character. This is
3284 // faster than countChars(seq, i, 1).
3285 if (Character.isHighSurrogate(seq.charAt(i++))) {
3286 if (i < seq.length() &&
3287 Character.isLowSurrogate(seq.charAt(i))) {
3288 i++;
3289 }
3290 }
3291 }
3292 matcher.hitEnd = true;
3293 return false;
3294 }
3295 }
3296
3297 /**
3298 * Node to anchor at the beginning of input. This object implements the
3299 * match for a \A sequence, and the caret anchor will use this if not in
3300 * multiline mode.
3301 */
3302 static final class Begin extends Node {
3303 boolean match(Matcher matcher, int i, CharSequence seq) {
3304 int fromIndex = (matcher.anchoringBounds) ?
3305 matcher.from : 0;
3306 if (i == fromIndex && next.match(matcher, i, seq)) {
3307 matcher.first = i;
3308 matcher.groups[0] = i;
3309 matcher.groups[1] = matcher.last;
3310 return true;
3311 } else {
3312 return false;
3313 }
3314 }
3315 }
3316
3317 /**
3318 * Node to anchor at the end of input. This is the absolute end, so this
3319 * should not match at the last newline before the end as $ will.
3320 */
3321 static final class End extends Node {
3322 boolean match(Matcher matcher, int i, CharSequence seq) {
3323 int endIndex = (matcher.anchoringBounds) ?
3324 matcher.to : matcher.getTextLength();
3325 if (i == endIndex) {
3326 matcher.hitEnd = true;
3327 return next.match(matcher, i, seq);
3328 }
3329 return false;
3330 }
3331 }
3332
3333 /**
3334 * Node to anchor at the beginning of a line. This is essentially the
3335 * object to match for the multiline ^.
3336 */
3337 static final class Caret extends Node {
3338 boolean match(Matcher matcher, int i, CharSequence seq) {
3339 int startIndex = matcher.from;
3340 int endIndex = matcher.to;
3341 if (!matcher.anchoringBounds) {
3342 startIndex = 0;
3343 endIndex = matcher.getTextLength();
3344 }
3345 // Perl does not match ^ at end of input even after newline
3346 if (i == endIndex) {
3347 matcher.hitEnd = true;
3348 return false;
3349 }
3350 if (i > startIndex) {
3351 char ch = seq.charAt(i-1);
3352 if (ch != '\n' && ch != '\r'
3353 && (ch|1) != '\u2029'
3354 && ch != '\u0085' ) {
3355 return false;
3356 }
3357 // Should treat /r/n as one newline
3358 if (ch == '\r' && seq.charAt(i) == '\n')
3359 return false;
3360 }
3361 return next.match(matcher, i, seq);
3362 }
3363 }
3364
3365 /**
3366 * Node to anchor at the beginning of a line when in unixdot mode.
3367 */
3368 static final class UnixCaret extends Node {
3369 boolean match(Matcher matcher, int i, CharSequence seq) {
3370 int startIndex = matcher.from;
3371 int endIndex = matcher.to;
3372 if (!matcher.anchoringBounds) {
3373 startIndex = 0;
3374 endIndex = matcher.getTextLength();
3375 }
3376 // Perl does not match ^ at end of input even after newline
3377 if (i == endIndex) {
3378 matcher.hitEnd = true;
3379 return false;
3380 }
3381 if (i > startIndex) {
3382 char ch = seq.charAt(i-1);
3383 if (ch != '\n') {
3384 return false;
3385 }
3386 }
3387 return next.match(matcher, i, seq);
3388 }
3389 }
3390
3391 /**
3392 * Node to match the location where the last match ended.
3393 * This is used for the \G construct.
3394 */
3395 static final class LastMatch extends Node {
3396 boolean match(Matcher matcher, int i, CharSequence seq) {
3397 if (i != matcher.oldLast)
3398 return false;
3399 return next.match(matcher, i, seq);
3400 }
3401 }
3402
3403 /**
3404 * Node to anchor at the end of a line or the end of input based on the
3405 * multiline mode.
3406 *
3407 * When not in multiline mode, the $ can only match at the very end
3408 * of the input, unless the input ends in a line terminator in which
3409 * it matches right before the last line terminator.
3410 *
3411 * Note that \r\n is considered an atomic line terminator.
3412 *
3413 * Like ^ the $ operator matches at a position, it does not match the
3414 * line terminators themselves.
3415 */
3416 static final class Dollar extends Node {
3417 boolean multiline;
3418 Dollar(boolean mul) {
3419 multiline = mul;
3420 }
3421 boolean match(Matcher matcher, int i, CharSequence seq) {
3422 int endIndex = (matcher.anchoringBounds) ?
3423 matcher.to : matcher.getTextLength();
3424 if (!multiline) {
3425 if (i < endIndex - 2)
3426 return false;
3427 if (i == endIndex - 2) {
3428 char ch = seq.charAt(i);
3429 if (ch != '\r')
3430 return false;
3431 ch = seq.charAt(i + 1);
3432 if (ch != '\n')
3433 return false;
3434 }
3435 }
3436 // Matches before any line terminator; also matches at the
3437 // end of input
3438 // Before line terminator:
3439 // If multiline, we match here no matter what
3440 // If not multiline, fall through so that the end
3441 // is marked as hit; this must be a /r/n or a /n
3442 // at the very end so the end was hit; more input
3443 // could make this not match here
3444 if (i < endIndex) {
3445 char ch = seq.charAt(i);
3446 if (ch == '\n') {
3447 // No match between \r\n
3448 if (i > 0 && seq.charAt(i-1) == '\r')
3449 return false;
3450 if (multiline)
3451 return next.match(matcher, i, seq);
3452 } else if (ch == '\r' || ch == '\u0085' ||
3453 (ch|1) == '\u2029') {
3454 if (multiline)
3455 return next.match(matcher, i, seq);
3456 } else { // No line terminator, no match
3457 return false;
3458 }
3459 }
3460 // Matched at current end so hit end
3461 matcher.hitEnd = true;
3462 // If a $ matches because of end of input, then more input
3463 // could cause it to fail!
3464 matcher.requireEnd = true;
3465 return next.match(matcher, i, seq);
3466 }
3467 boolean study(TreeInfo info) {
3468 next.study(info);
3469 return info.deterministic;
3470 }
3471 }
3472
3473 /**
3474 * Node to anchor at the end of a line or the end of input based on the
3475 * multiline mode when in unix lines mode.
3476 */
3477 static final class UnixDollar extends Node {
3478 boolean multiline;
3479 UnixDollar(boolean mul) {
3480 multiline = mul;
3481 }
3482 boolean match(Matcher matcher, int i, CharSequence seq) {
3483 int endIndex = (matcher.anchoringBounds) ?
3484 matcher.to : matcher.getTextLength();
3485 if (i < endIndex) {
3486 char ch = seq.charAt(i);
3487 if (ch == '\n') {
3488 // If not multiline, then only possible to
3489 // match at very end or one before end
3490 if (multiline == false && i != endIndex - 1)
3491 return false;
3492 // If multiline return next.match without setting
3493 // matcher.hitEnd
3494 if (multiline)
3495 return next.match(matcher, i, seq);
3496 } else {
3497 return false;
3498 }
3499 }
3500 // Matching because at the end or 1 before the end;
3501 // more input could change this so set hitEnd
3502 matcher.hitEnd = true;
3503 // If a $ matches because of end of input, then more input
3504 // could cause it to fail!
3505 matcher.requireEnd = true;
3506 return next.match(matcher, i, seq);
3507 }
3508 boolean study(TreeInfo info) {
3509 next.study(info);
3510 return info.deterministic;
3511 }
3512 }
3513
3514 /**
3515 * Abstract node class to match one character satisfying some
3516 * boolean property.
3517 */
3518 private static abstract class CharProperty extends Node {
3519 abstract boolean isSatisfiedBy(int ch);
3520 CharProperty complement() {
3521 return new CharProperty() {
3522 boolean isSatisfiedBy(int ch) {
3523 return ! CharProperty.this.isSatisfiedBy(ch);}};
3524 }
3525 boolean match(Matcher matcher, int i, CharSequence seq) {
3526 if (i < matcher.to) {
3527 int ch = Character.codePointAt(seq, i);
3528 return isSatisfiedBy(ch)
3529 && next.match(matcher, i+Character.charCount(ch), seq);
3530 } else {
3531 matcher.hitEnd = true;
3532 return false;
3533 }
3534 }
3535 boolean study(TreeInfo info) {
3536 info.minLength++;
3537 info.maxLength++;
3538 return next.study(info);
3539 }
3540 }
3541
3542 /**
3543 * Optimized version of CharProperty that works only for
3544 * properties never satisfied by Supplementary characters.
3545 */
3546 private static abstract class BmpCharProperty extends CharProperty {
3547 boolean match(Matcher matcher, int i, CharSequence seq) {
3548 if (i < matcher.to) {
3549 return isSatisfiedBy(seq.charAt(i))
3550 && next.match(matcher, i+1, seq);
3551 } else {
3552 matcher.hitEnd = true;
3553 return false;
3554 }
3555 }
3556 }
3557
3558 /**
3559 * Node class that matches a Supplementary Unicode character
3560 */
3561 static final class SingleS extends CharProperty {
3562 final int c;
3563 SingleS(int c) { this.c = c; }
3564 boolean isSatisfiedBy(int ch) {
3565 return ch == c;
3566 }
3567 }
3568
3569 /**
3570 * Optimization -- matches a given BMP character
3571 */
3572 static final class Single extends BmpCharProperty {
3573 final int c;
3574 Single(int c) { this.c = c; }
3575 boolean isSatisfiedBy(int ch) {
3576 return ch == c;
3577 }
3578 }
3579
3580 /**
3581 * Case insensitive matches a given BMP character
3582 */
3583 static final class SingleI extends BmpCharProperty {
3584 final int lower;
3585 final int upper;
3586 SingleI(int lower, int upper) {
3587 this.lower = lower;
3588 this.upper = upper;
3589 }
3590 boolean isSatisfiedBy(int ch) {
3591 return ch == lower || ch == upper;
3592 }
3593 }
3594
3595 /**
3596 * Unicode case insensitive matches a given Unicode character
3597 */
3598 static final class SingleU extends CharProperty {
3599 final int lower;
3600 SingleU(int lower) {
3601 this.lower = lower;
3602 }
3603 boolean isSatisfiedBy(int ch) {
3604 return lower == ch ||
3605 lower == Character.toLowerCase(Character.toUpperCase(ch));
3606 }
3607 }
3608
3609
3610 /**
3611 * Node class that matches a Unicode block.
3612 */
3613 static final class Block extends CharProperty {
3614 final Character.UnicodeBlock block;
3615 Block(Character.UnicodeBlock block) {
3616 this.block = block;
3617 }
3618 boolean isSatisfiedBy(int ch) {
3619 return block == Character.UnicodeBlock.of(ch);
3620 }
3621 }
3622
3623 /**
3624 * Node class that matches a Unicode script
3625 */
3626 static final class Script extends CharProperty {
3627 final Character.UnicodeScript script;
3628 Script(Character.UnicodeScript script) {
3629 this.script = script;
3630 }
3631 boolean isSatisfiedBy(int ch) {
3632 return script == Character.UnicodeScript.of(ch);
3633 }
3634 }
3635
3636 /**
3637 * Node class that matches a Unicode category.
3638 */
3639 static final class Category extends CharProperty {
3640 final int typeMask;
3641 Category(int typeMask) { this.typeMask = typeMask; }
3642 boolean isSatisfiedBy(int ch) {
3643 return (typeMask & (1 << Character.getType(ch))) != 0;
3644 }
3645 }
3646
3647 /**
3648 * Node class that matches a POSIX type.
3649 */
3650 static final class Ctype extends BmpCharProperty {
3651 final int ctype;
3652 Ctype(int ctype) { this.ctype = ctype; }
3653 boolean isSatisfiedBy(int ch) {
3654 return ch < 128 && ASCII.isType(ch, ctype);
3655 }
3656 }
3657
3658 /**
3659 * Base class for all Slice nodes
3660 */
3661 static class SliceNode extends Node {
3662 int[] buffer;
3663 SliceNode(int[] buf) {
3664 buffer = buf;
3665 }
3666 boolean study(TreeInfo info) {
3667 info.minLength += buffer.length;
3668 info.maxLength += buffer.length;
3669 return next.study(info);
3670 }
3671 }
3672
3673 /**
3674 * Node class for a case sensitive/BMP-only sequence of literal
3675 * characters.
3676 */
3677 static final class Slice extends SliceNode {
3678 Slice(int[] buf) {
3679 super(buf);
3680 }
3681 boolean match(Matcher matcher, int i, CharSequence seq) {
3682 int[] buf = buffer;
3683 int len = buf.length;
3684 for (int j=0; j<len; j++) {
3685 if ((i+j) >= matcher.to) {
3686 matcher.hitEnd = true;
3687 return false;
3688 }
3689 if (buf[j] != seq.charAt(i+j))
3690 return false;
3691 }
3692 return next.match(matcher, i+len, seq);
3693 }
3694 }
3695
3696 /**
3697 * Node class for a case_insensitive/BMP-only sequence of literal
3698 * characters.
3699 */
3700 static class SliceI extends SliceNode {
3701 SliceI(int[] buf) {
3702 super(buf);
3703 }
3704 boolean match(Matcher matcher, int i, CharSequence seq) {
3705 int[] buf = buffer;
3706 int len = buf.length;
3707 for (int j=0; j<len; j++) {
3708 if ((i+j) >= matcher.to) {
3709 matcher.hitEnd = true;
3710 return false;
3711 }
3712 int c = seq.charAt(i+j);
3713 if (buf[j] != c &&
3714 buf[j] != ASCII.toLower(c))
3715 return false;
3716 }
3717 return next.match(matcher, i+len, seq);
3718 }
3719 }
3720
3721 /**
3722 * Node class for a unicode_case_insensitive/BMP-only sequence of
3723 * literal characters. Uses unicode case folding.
3724 */
3725 static final class SliceU extends SliceNode {
3726 SliceU(int[] buf) {
3727 super(buf);
3728 }
3729 boolean match(Matcher matcher, int i, CharSequence seq) {
3730 int[] buf = buffer;
3731 int len = buf.length;
3732 for (int j=0; j<len; j++) {
3733 if ((i+j) >= matcher.to) {
3734 matcher.hitEnd = true;
3735 return false;
3736 }
3737 int c = seq.charAt(i+j);
3738 if (buf[j] != c &&
3739 buf[j] != Character.toLowerCase(Character.toUpperCase(c)))
3740 return false;
3741 }
3742 return next.match(matcher, i+len, seq);
3743 }
3744 }
3745
3746 /**
3747 * Node class for a case sensitive sequence of literal characters
3748 * including supplementary characters.
3749 */
3750 static final class SliceS extends SliceNode {
3751 SliceS(int[] buf) {
3752 super(buf);
3753 }
3754 boolean match(Matcher matcher, int i, CharSequence seq) {
3755 int[] buf = buffer;
3756 int x = i;
3757 for (int j = 0; j < buf.length; j++) {
3758 if (x >= matcher.to) {
3759 matcher.hitEnd = true;
3760 return false;
3761 }
3762 int c = Character.codePointAt(seq, x);
3763 if (buf[j] != c)
3764 return false;
3765 x += Character.charCount(c);
3766 if (x > matcher.to) {
3767 matcher.hitEnd = true;
3768 return false;
3769 }
3770 }
3771 return next.match(matcher, x, seq);
3772 }
3773 }
3774
3775 /**
3776 * Node class for a case insensitive sequence of literal characters
3777 * including supplementary characters.
3778 */
3779 static class SliceIS extends SliceNode {
3780 SliceIS(int[] buf) {
3781 super(buf);
3782 }
3783 int toLower(int c) {
3784 return ASCII.toLower(c);
3785 }
3786 boolean match(Matcher matcher, int i, CharSequence seq) {
3787 int[] buf = buffer;
3788 int x = i;
3789 for (int j = 0; j < buf.length; j++) {
3790 if (x >= matcher.to) {
3791 matcher.hitEnd = true;
3792 return false;
3793 }
3794 int c = Character.codePointAt(seq, x);
3795 if (buf[j] != c && buf[j] != toLower(c))
3796 return false;
3797 x += Character.charCount(c);
3798 if (x > matcher.to) {
3799 matcher.hitEnd = true;
3800 return false;
3801 }
3802 }
3803 return next.match(matcher, x, seq);
3804 }
3805 }
3806
3807 /**
3808 * Node class for a case insensitive sequence of literal characters.
3809 * Uses unicode case folding.
3810 */
3811 static final class SliceUS extends SliceIS {
3812 SliceUS(int[] buf) {
3813 super(buf);
3814 }
3815 int toLower(int c) {
3816 return Character.toLowerCase(Character.toUpperCase(c));
3817 }
3818 }
3819
3820 private static boolean inRange(int lower, int ch, int upper) {
3821 return lower <= ch && ch <= upper;
3822 }
3823
3824 /**
3825 * Returns node for matching characters within an explicit value range.
3826 */
3827 private static CharProperty rangeFor(final int lower,
3828 final int upper) {
3829 return new CharProperty() {
3830 boolean isSatisfiedBy(int ch) {
3831 return inRange(lower, ch, upper);}};
3832 }
3833
3834 /**
3835 * Returns node for matching characters within an explicit value
3836 * range in a case insensitive manner.
3837 */
3838 private CharProperty caseInsensitiveRangeFor(final int lower,
3839 final int upper) {
3840 if (has(UNICODE_CASE))
3841 return new CharProperty() {
3842 boolean isSatisfiedBy(int ch) {
3843 if (inRange(lower, ch, upper))
3844 return true;
3845 int up = Character.toUpperCase(ch);
3846 return inRange(lower, up, upper) ||
3847 inRange(lower, Character.toLowerCase(up), upper);}};
3848 return new CharProperty() {
3849 boolean isSatisfiedBy(int ch) {
3850 return inRange(lower, ch, upper) ||
3851 ASCII.isAscii(ch) &&
3852 (inRange(lower, ASCII.toUpper(ch), upper) ||
3853 inRange(lower, ASCII.toLower(ch), upper));
3854 }};
3855 }
3856
3857 /**
3858 * Implements the Unicode category ALL and the dot metacharacter when
3859 * in dotall mode.
3860 */
3861 static final class All extends CharProperty {
3862 boolean isSatisfiedBy(int ch) {
3863 return true;
3864 }
3865 }
3866
3867 /**
3868 * Node class for the dot metacharacter when dotall is not enabled.
3869 */
3870 static final class Dot extends CharProperty {
3871 boolean isSatisfiedBy(int ch) {
3872 return (ch != '\n' && ch != '\r'
3873 && (ch|1) != '\u2029'
3874 && ch != '\u0085');
3875 }
3876 }
3877
3878 /**
3879 * Node class for the dot metacharacter when dotall is not enabled
3880 * but UNIX_LINES is enabled.
3881 */
3882 static final class UnixDot extends CharProperty {
3883 boolean isSatisfiedBy(int ch) {
3884 return ch != '\n';
3885 }
3886 }
3887
3888 /**
3889 * The 0 or 1 quantifier. This one class implements all three types.
3890 */
3891 static final class Ques extends Node {
3892 Node atom;
3893 int type;
3894 Ques(Node node, int type) {
3895 this.atom = node;
3896 this.type = type;
3897 }
3898 boolean match(Matcher matcher, int i, CharSequence seq) {
3899 switch (type) {
3900 case GREEDY:
3901 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq))
3902 || next.match(matcher, i, seq);
3903 case LAZY:
3904 return next.match(matcher, i, seq)
3905 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq));
3906 case POSSESSIVE:
3907 if (atom.match(matcher, i, seq)) i = matcher.last;
3908 return next.match(matcher, i, seq);
3909 default:
3910 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq);
3911 }
3912 }
3913 boolean study(TreeInfo info) {
3914 if (type != INDEPENDENT) {
3915 int minL = info.minLength;
3916 atom.study(info);
3917 info.minLength = minL;
3918 info.deterministic = false;
3919 return next.study(info);
3920 } else {
3921 atom.study(info);
3922 return next.study(info);
3923 }
3924 }
3925 }
3926
3927 /**
3928 * Handles the curly-brace style repetition with a specified minimum and
3929 * maximum occurrences. The * quantifier is handled as a special case.
3930 * This class handles the three types.
3931 */
3932 static final class Curly extends Node {
3933 Node atom;
3934 int type;
3935 int cmin;
3936 int cmax;
3937
3938 Curly(Node node, int cmin, int cmax, int type) {
3939 this.atom = node;
3940 this.type = type;
3941 this.cmin = cmin;
3942 this.cmax = cmax;
3943 }
3944 boolean match(Matcher matcher, int i, CharSequence seq) {
3945 int j;
3946 for (j = 0; j < cmin; j++) {
3947 if (atom.match(matcher, i, seq)) {
3948 i = matcher.last;
3949 continue;
3950 }
3951 return false;
3952 }
3953 if (type == GREEDY)
3954 return match0(matcher, i, j, seq);
3955 else if (type == LAZY)
3956 return match1(matcher, i, j, seq);
3957 else
3958 return match2(matcher, i, j, seq);
3959 }
3960 // Greedy match.
3961 // i is the index to start matching at
3962 // j is the number of atoms that have matched
3963 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
3964 if (j >= cmax) {
3965 // We have matched the maximum... continue with the rest of
3966 // the regular expression
3967 return next.match(matcher, i, seq);
3968 }
3969 int backLimit = j;
3970 while (atom.match(matcher, i, seq)) {
3971 // k is the length of this match
3972 int k = matcher.last - i;
3973 if (k == 0) // Zero length match
3974 break;
3975 // Move up index and number matched
3976 i = matcher.last;
3977 j++;
3978 // We are greedy so match as many as we can
3979 while (j < cmax) {
3980 if (!atom.match(matcher, i, seq))
3981 break;
3982 if (i + k != matcher.last) {
3983 if (match0(matcher, matcher.last, j+1, seq))
3984 return true;
3985 break;
3986 }
3987 i += k;
3988 j++;
3989 }
3990 // Handle backing off if match fails
3991 while (j >= backLimit) {
3992 if (next.match(matcher, i, seq))
3993 return true;
3994 i -= k;
3995 j--;
3996 }
3997 return false;
3998 }
3999 return next.match(matcher, i, seq);
4000 }
4001 // Reluctant match. At this point, the minimum has been satisfied.
4002 // i is the index to start matching at
4003 // j is the number of atoms that have matched
4004 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4005 for (;;) {
4006 // Try finishing match without consuming any more
4007 if (next.match(matcher, i, seq))
4008 return true;
4009 // At the maximum, no match found
4010 if (j >= cmax)
4011 return false;
4012 // Okay, must try one more atom
4013 if (!atom.match(matcher, i, seq))
4014 return false;
4015 // If we haven't moved forward then must break out
4016 if (i == matcher.last)
4017 return false;
4018 // Move up index and number matched
4019 i = matcher.last;
4020 j++;
4021 }
4022 }
4023 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4024 for (; j < cmax; j++) {
4025 if (!atom.match(matcher, i, seq))
4026 break;
4027 if (i == matcher.last)
4028 break;
4029 i = matcher.last;
4030 }
4031 return next.match(matcher, i, seq);
4032 }
4033 boolean study(TreeInfo info) {
4034 // Save original info
4035 int minL = info.minLength;
4036 int maxL = info.maxLength;
4037 boolean maxV = info.maxValid;
4038 boolean detm = info.deterministic;
4039 info.reset();
4040
4041 atom.study(info);
4042
4043 int temp = info.minLength * cmin + minL;
4044 if (temp < minL) {
4045 temp = 0xFFFFFFF; // arbitrary large number
4046 }
4047 info.minLength = temp;
4048
4049 if (maxV & info.maxValid) {
4050 temp = info.maxLength * cmax + maxL;
4051 info.maxLength = temp;
4052 if (temp < maxL) {
4053 info.maxValid = false;
4054 }
4055 } else {
4056 info.maxValid = false;
4057 }
4058
4059 if (info.deterministic && cmin == cmax)
4060 info.deterministic = detm;
4061 else
4062 info.deterministic = false;
4063
4064 return next.study(info);
4065 }
4066 }
4067
4068 /**
4069 * Handles the curly-brace style repetition with a specified minimum and
4070 * maximum occurrences in deterministic cases. This is an iterative
4071 * optimization over the Prolog and Loop system which would handle this
4072 * in a recursive way. The * quantifier is handled as a special case.
4073 * If capture is true then this class saves group settings and ensures
4074 * that groups are unset when backing off of a group match.
4075 */
4076 static final class GroupCurly extends Node {
4077 Node atom;
4078 int type;
4079 int cmin;
4080 int cmax;
4081 int localIndex;
4082 int groupIndex;
4083 boolean capture;
4084
4085 GroupCurly(Node node, int cmin, int cmax, int type, int local,
4086 int group, boolean capture) {
4087 this.atom = node;
4088 this.type = type;
4089 this.cmin = cmin;
4090 this.cmax = cmax;
4091 this.localIndex = local;
4092 this.groupIndex = group;
4093 this.capture = capture;
4094 }
4095 boolean match(Matcher matcher, int i, CharSequence seq) {
4096 int[] groups = matcher.groups;
4097 int[] locals = matcher.locals;
4098 int save0 = locals[localIndex];
4099 int save1 = 0;
4100 int save2 = 0;
4101
4102 if (capture) {
4103 save1 = groups[groupIndex];
4104 save2 = groups[groupIndex+1];
4105 }
4106
4107 // Notify GroupTail there is no need to setup group info
4108 // because it will be set here
4109 locals[localIndex] = -1;
4110
4111 boolean ret = true;
4112 for (int j = 0; j < cmin; j++) {
4113 if (atom.match(matcher, i, seq)) {
4114 if (capture) {
4115 groups[groupIndex] = i;
4116 groups[groupIndex+1] = matcher.last;
4117 }
4118 i = matcher.last;
4119 } else {
4120 ret = false;
4121 break;
4122 }
4123 }
4124 if (ret) {
4125 if (type == GREEDY) {
4126 ret = match0(matcher, i, cmin, seq);
4127 } else if (type == LAZY) {
4128 ret = match1(matcher, i, cmin, seq);
4129 } else {
4130 ret = match2(matcher, i, cmin, seq);
4131 }
4132 }
4133 if (!ret) {
4134 locals[localIndex] = save0;
4135 if (capture) {
4136 groups[groupIndex] = save1;
4137 groups[groupIndex+1] = save2;
4138 }
4139 }
4140 return ret;
4141 }
4142 // Aggressive group match
4143 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4144 int[] groups = matcher.groups;
4145 int save0 = 0;
4146 int save1 = 0;
4147 if (capture) {
4148 save0 = groups[groupIndex];
4149 save1 = groups[groupIndex+1];
4150 }
4151 for (;;) {
4152 if (j >= cmax)
4153 break;
4154 if (!atom.match(matcher, i, seq))
4155 break;
4156 int k = matcher.last - i;
4157 if (k <= 0) {
4158 if (capture) {
4159 groups[groupIndex] = i;
4160 groups[groupIndex+1] = i + k;
4161 }
4162 i = i + k;
4163 break;
4164 }
4165 for (;;) {
4166 if (capture) {
4167 groups[groupIndex] = i;
4168 groups[groupIndex+1] = i + k;
4169 }
4170 i = i + k;
4171 if (++j >= cmax)
4172 break;
4173 if (!atom.match(matcher, i, seq))
4174 break;
4175 if (i + k != matcher.last) {
4176 if (match0(matcher, i, j, seq))
4177 return true;
4178 break;
4179 }
4180 }
4181 while (j > cmin) {
4182 if (next.match(matcher, i, seq)) {
4183 if (capture) {
4184 groups[groupIndex+1] = i;
4185 groups[groupIndex] = i - k;
4186 }
4187 i = i - k;
4188 return true;
4189 }
4190 // backing off
4191 if (capture) {
4192 groups[groupIndex+1] = i;
4193 groups[groupIndex] = i - k;
4194 }
4195 i = i - k;
4196 j--;
4197 }
4198 break;
4199 }
4200 if (capture) {
4201 groups[groupIndex] = save0;
4202 groups[groupIndex+1] = save1;
4203 }
4204 return next.match(matcher, i, seq);
4205 }
4206 // Reluctant matching
4207 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4208 for (;;) {
4209 if (next.match(matcher, i, seq))
4210 return true;
4211 if (j >= cmax)
4212 return false;
4213 if (!atom.match(matcher, i, seq))
4214 return false;
4215 if (i == matcher.last)
4216 return false;
4217 if (capture) {
4218 matcher.groups[groupIndex] = i;
4219 matcher.groups[groupIndex+1] = matcher.last;
4220 }
4221 i = matcher.last;
4222 j++;
4223 }
4224 }
4225 // Possessive matching
4226 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4227 for (; j < cmax; j++) {
4228 if (!atom.match(matcher, i, seq)) {
4229 break;
4230 }
4231 if (capture) {
4232 matcher.groups[groupIndex] = i;
4233 matcher.groups[groupIndex+1] = matcher.last;
4234 }
4235 if (i == matcher.last) {
4236 break;
4237 }
4238 i = matcher.last;
4239 }
4240 return next.match(matcher, i, seq);
4241 }
4242 boolean study(TreeInfo info) {
4243 // Save original info
4244 int minL = info.minLength;
4245 int maxL = info.maxLength;
4246 boolean maxV = info.maxValid;
4247 boolean detm = info.deterministic;
4248 info.reset();
4249
4250 atom.study(info);
4251
4252 int temp = info.minLength * cmin + minL;
4253 if (temp < minL) {
4254 temp = 0xFFFFFFF; // Arbitrary large number
4255 }
4256 info.minLength = temp;
4257
4258 if (maxV & info.maxValid) {
4259 temp = info.maxLength * cmax + maxL;
4260 info.maxLength = temp;
4261 if (temp < maxL) {
4262 info.maxValid = false;
4263 }
4264 } else {
4265 info.maxValid = false;
4266 }
4267
4268 if (info.deterministic && cmin == cmax) {
4269 info.deterministic = detm;
4270 } else {
4271 info.deterministic = false;
4272 }
4273
4274 return next.study(info);
4275 }
4276 }
4277
4278 /**
4279 * A Guard node at the end of each atom node in a Branch. It
4280 * serves the purpose of chaining the "match" operation to
4281 * "next" but not the "study", so we can collect the TreeInfo
4282 * of each atom node without including the TreeInfo of the
4283 * "next".
4284 */
4285 static final class BranchConn extends Node {
4286 BranchConn() {};
4287 boolean match(Matcher matcher, int i, CharSequence seq) {
4288 return next.match(matcher, i, seq);
4289 }
4290 boolean study(TreeInfo info) {
4291 return info.deterministic;
4292 }
4293 }
4294
4295 /**
4296 * Handles the branching of alternations. Note this is also used for
4297 * the ? quantifier to branch between the case where it matches once
4298 * and where it does not occur.
4299 */
4300 static final class Branch extends Node {
4301 Node[] atoms = new Node[2];
4302 int size = 2;
4303 Node conn;
4304 Branch(Node first, Node second, Node branchConn) {
4305 conn = branchConn;
4306 atoms[0] = first;
4307 atoms[1] = second;
4308 }
4309
4310 void add(Node node) {
4311 if (size >= atoms.length) {
4312 Node[] tmp = new Node[atoms.length*2];
4313 System.arraycopy(atoms, 0, tmp, 0, atoms.length);
4314 atoms = tmp;
4315 }
4316 atoms[size++] = node;
4317 }
4318
4319 boolean match(Matcher matcher, int i, CharSequence seq) {
4320 for (int n = 0; n < size; n++) {
4321 if (atoms[n] == null) {
4322 if (conn.next.match(matcher, i, seq))
4323 return true;
4324 } else if (atoms[n].match(matcher, i, seq)) {
4325 return true;
4326 }
4327 }
4328 return false;
4329 }
4330
4331 boolean study(TreeInfo info) {
4332 int minL = info.minLength;
4333 int maxL = info.maxLength;
4334 boolean maxV = info.maxValid;
4335
4336 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num
4337 int maxL2 = -1;
4338 for (int n = 0; n < size; n++) {
4339 info.reset();
4340 if (atoms[n] != null)
4341 atoms[n].study(info);
4342 minL2 = Math.min(minL2, info.minLength);
4343 maxL2 = Math.max(maxL2, info.maxLength);
4344 maxV = (maxV & info.maxValid);
4345 }
4346
4347 minL += minL2;
4348 maxL += maxL2;
4349
4350 info.reset();
4351 conn.next.study(info);
4352
4353 info.minLength += minL;
4354 info.maxLength += maxL;
4355 info.maxValid &= maxV;
4356 info.deterministic = false;
4357 return false;
4358 }
4359 }
4360
4361 /**
4362 * The GroupHead saves the location where the group begins in the locals
4363 * and restores them when the match is done.
4364 *
4365 * The matchRef is used when a reference to this group is accessed later
4366 * in the expression. The locals will have a negative value in them to
4367 * indicate that we do not want to unset the group if the reference
4368 * doesn't match.
4369 */
4370 static final class GroupHead extends Node {
4371 int localIndex;
4372 GroupHead(int localCount) {
4373 localIndex = localCount;
4374 }
4375 boolean match(Matcher matcher, int i, CharSequence seq) {
4376 int save = matcher.locals[localIndex];
4377 matcher.locals[localIndex] = i;
4378 boolean ret = next.match(matcher, i, seq);
4379 matcher.locals[localIndex] = save;
4380 return ret;
4381 }
4382 boolean matchRef(Matcher matcher, int i, CharSequence seq) {
4383 int save = matcher.locals[localIndex];
4384 matcher.locals[localIndex] = ~i; // HACK
4385 boolean ret = next.match(matcher, i, seq);
4386 matcher.locals[localIndex] = save;
4387 return ret;
4388 }
4389 }
4390
4391 /**
4392 * Recursive reference to a group in the regular expression. It calls
4393 * matchRef because if the reference fails to match we would not unset
4394 * the group.
4395 */
4396 static final class GroupRef extends Node {
4397 GroupHead head;
4398 GroupRef(GroupHead head) {
4399 this.head = head;
4400 }
4401 boolean match(Matcher matcher, int i, CharSequence seq) {
4402 return head.matchRef(matcher, i, seq)
4403 && next.match(matcher, matcher.last, seq);
4404 }
4405 boolean study(TreeInfo info) {
4406 info.maxValid = false;
4407 info.deterministic = false;
4408 return next.study(info);
4409 }
4410 }
4411
4412 /**
4413 * The GroupTail handles the setting of group beginning and ending
4414 * locations when groups are successfully matched. It must also be able to
4415 * unset groups that have to be backed off of.
4416 *
4417 * The GroupTail node is also used when a previous group is referenced,
4418 * and in that case no group information needs to be set.
4419 */
4420 static final class GroupTail extends Node {
4421 int localIndex;
4422 int groupIndex;
4423 GroupTail(int localCount, int groupCount) {
4424 localIndex = localCount;
4425 groupIndex = groupCount + groupCount;
4426 }
4427 boolean match(Matcher matcher, int i, CharSequence seq) {
4428 int tmp = matcher.locals[localIndex];
4429 if (tmp >= 0) { // This is the normal group case.
4430 // Save the group so we can unset it if it
4431 // backs off of a match.
4432 int groupStart = matcher.groups[groupIndex];
4433 int groupEnd = matcher.groups[groupIndex+1];
4434
4435 matcher.groups[groupIndex] = tmp;
4436 matcher.groups[groupIndex+1] = i;
4437 if (next.match(matcher, i, seq)) {
4438 return true;
4439 }
4440 matcher.groups[groupIndex] = groupStart;
4441 matcher.groups[groupIndex+1] = groupEnd;
4442 return false;
4443 } else {
4444 // This is a group reference case. We don't need to save any
4445 // group info because it isn't really a group.
4446 matcher.last = i;
4447 return true;
4448 }
4449 }
4450 }
4451
4452 /**
4453 * This sets up a loop to handle a recursive quantifier structure.
4454 */
4455 static final class Prolog extends Node {
4456 Loop loop;
4457 Prolog(Loop loop) {
4458 this.loop = loop;
4459 }
4460 boolean match(Matcher matcher, int i, CharSequence seq) {
4461 return loop.matchInit(matcher, i, seq);
4462 }
4463 boolean study(TreeInfo info) {
4464 return loop.study(info);
4465 }
4466 }
4467
4468 /**
4469 * Handles the repetition count for a greedy Curly. The matchInit
4470 * is called from the Prolog to save the index of where the group
4471 * beginning is stored. A zero length group check occurs in the
4472 * normal match but is skipped in the matchInit.
4473 */
4474 static class Loop extends Node {
4475 Node body;
4476 int countIndex; // local count index in matcher locals
4477 int beginIndex; // group beginning index
4478 int cmin, cmax;
4479 Loop(int countIndex, int beginIndex) {
4480 this.countIndex = countIndex;
4481 this.beginIndex = beginIndex;
4482 }
4483 boolean match(Matcher matcher, int i, CharSequence seq) {
4484 // Avoid infinite loop in zero-length case.
4485 if (i > matcher.locals[beginIndex]) {
4486 int count = matcher.locals[countIndex];
4487
4488 // This block is for before we reach the minimum
4489 // iterations required for the loop to match
4490 if (count < cmin) {
4491 matcher.locals[countIndex] = count + 1;
4492 boolean b = body.match(matcher, i, seq);
4493 // If match failed we must backtrack, so
4494 // the loop count should NOT be incremented
4495 if (!b)
4496 matcher.locals[countIndex] = count;
4497 // Return success or failure since we are under
4498 // minimum
4499 return b;
4500 }
4501 // This block is for after we have the minimum
4502 // iterations required for the loop to match
4503 if (count < cmax) {
4504 matcher.locals[countIndex] = count + 1;
4505 boolean b = body.match(matcher, i, seq);
4506 // If match failed we must backtrack, so
4507 // the loop count should NOT be incremented
4508 if (!b)
4509 matcher.locals[countIndex] = count;
4510 else
4511 return true;
4512 }
4513 }
4514 return next.match(matcher, i, seq);
4515 }
4516 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4517 int save = matcher.locals[countIndex];
4518 boolean ret = false;
4519 if (0 < cmin) {
4520 matcher.locals[countIndex] = 1;
4521 ret = body.match(matcher, i, seq);
4522 } else if (0 < cmax) {
4523 matcher.locals[countIndex] = 1;
4524 ret = body.match(matcher, i, seq);
4525 if (ret == false)
4526 ret = next.match(matcher, i, seq);
4527 } else {
4528 ret = next.match(matcher, i, seq);
4529 }
4530 matcher.locals[countIndex] = save;
4531 return ret;
4532 }
4533 boolean study(TreeInfo info) {
4534 info.maxValid = false;
4535 info.deterministic = false;
4536 return false;
4537 }
4538 }
4539
4540 /**
4541 * Handles the repetition count for a reluctant Curly. The matchInit
4542 * is called from the Prolog to save the index of where the group
4543 * beginning is stored. A zero length group check occurs in the
4544 * normal match but is skipped in the matchInit.
4545 */
4546 static final class LazyLoop extends Loop {
4547 LazyLoop(int countIndex, int beginIndex) {
4548 super(countIndex, beginIndex);
4549 }
4550 boolean match(Matcher matcher, int i, CharSequence seq) {
4551 // Check for zero length group
4552 if (i > matcher.locals[beginIndex]) {
4553 int count = matcher.locals[countIndex];
4554 if (count < cmin) {
4555 matcher.locals[countIndex] = count + 1;
4556 boolean result = body.match(matcher, i, seq);
4557 // If match failed we must backtrack, so
4558 // the loop count should NOT be incremented
4559 if (!result)
4560 matcher.locals[countIndex] = count;
4561 return result;
4562 }
4563 if (next.match(matcher, i, seq))
4564 return true;
4565 if (count < cmax) {
4566 matcher.locals[countIndex] = count + 1;
4567 boolean result = body.match(matcher, i, seq);
4568 // If match failed we must backtrack, so
4569 // the loop count should NOT be incremented
4570 if (!result)
4571 matcher.locals[countIndex] = count;
4572 return result;
4573 }
4574 return false;
4575 }
4576 return next.match(matcher, i, seq);
4577 }
4578 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4579 int save = matcher.locals[countIndex];
4580 boolean ret = false;
4581 if (0 < cmin) {
4582 matcher.locals[countIndex] = 1;
4583 ret = body.match(matcher, i, seq);
4584 } else if (next.match(matcher, i, seq)) {
4585 ret = true;
4586 } else if (0 < cmax) {
4587 matcher.locals[countIndex] = 1;
4588 ret = body.match(matcher, i, seq);
4589 }
4590 matcher.locals[countIndex] = save;
4591 return ret;
4592 }
4593 boolean study(TreeInfo info) {
4594 info.maxValid = false;
4595 info.deterministic = false;
4596 return false;
4597 }
4598 }
4599
4600 /**
4601 * Refers to a group in the regular expression. Attempts to match
4602 * whatever the group referred to last matched.
4603 */
4604 static class BackRef extends Node {
4605 int groupIndex;
4606 BackRef(int groupCount) {
4607 super();
4608 groupIndex = groupCount + groupCount;
4609 }
4610 boolean match(Matcher matcher, int i, CharSequence seq) {
4611 int j = matcher.groups[groupIndex];
4612 int k = matcher.groups[groupIndex+1];
4613
4614 int groupSize = k - j;
4615
4616 // If the referenced group didn't match, neither can this
4617 if (j < 0)
4618 return false;
4619
4620 // If there isn't enough input left no match
4621 if (i + groupSize > matcher.to) {
4622 matcher.hitEnd = true;
4623 return false;
4624 }
4625
4626 // Check each new char to make sure it matches what the group
4627 // referenced matched last time around
4628 for (int index=0; index<groupSize; index++)
4629 if (seq.charAt(i+index) != seq.charAt(j+index))
4630 return false;
4631
4632 return next.match(matcher, i+groupSize, seq);
4633 }
4634 boolean study(TreeInfo info) {
4635 info.maxValid = false;
4636 return next.study(info);
4637 }
4638 }
4639
4640 static class CIBackRef extends Node {
4641 int groupIndex;
4642 boolean doUnicodeCase;
4643 CIBackRef(int groupCount, boolean doUnicodeCase) {
4644 super();
4645 groupIndex = groupCount + groupCount;
4646 this.doUnicodeCase = doUnicodeCase;
4647 }
4648 boolean match(Matcher matcher, int i, CharSequence seq) {
4649 int j = matcher.groups[groupIndex];
4650 int k = matcher.groups[groupIndex+1];
4651
4652 int groupSize = k - j;
4653
4654 // If the referenced group didn't match, neither can this
4655 if (j < 0)
4656 return false;
4657
4658 // If there isn't enough input left no match
4659 if (i + groupSize > matcher.to) {
4660 matcher.hitEnd = true;
4661 return false;
4662 }
4663
4664 // Check each new char to make sure it matches what the group
4665 // referenced matched last time around
4666 int x = i;
4667 for (int index=0; index<groupSize; index++) {
4668 int c1 = Character.codePointAt(seq, x);
4669 int c2 = Character.codePointAt(seq, j);
4670 if (c1 != c2) {
4671 if (doUnicodeCase) {
4672 int cc1 = Character.toUpperCase(c1);
4673 int cc2 = Character.toUpperCase(c2);
4674 if (cc1 != cc2 &&
4675 Character.toLowerCase(cc1) !=
4676 Character.toLowerCase(cc2))
4677 return false;
4678 } else {
4679 if (ASCII.toLower(c1) != ASCII.toLower(c2))
4680 return false;
4681 }
4682 }
4683 x += Character.charCount(c1);
4684 j += Character.charCount(c2);
4685 }
4686
4687 return next.match(matcher, i+groupSize, seq);
4688 }
4689 boolean study(TreeInfo info) {
4690 info.maxValid = false;
4691 return next.study(info);
4692 }
4693 }
4694
4695 /**
4696 * Searches until the next instance of its atom. This is useful for
4697 * finding the atom efficiently without passing an instance of it
4698 * (greedy problem) and without a lot of wasted search time (reluctant
4699 * problem).
4700 */
4701 static final class First extends Node {
4702 Node atom;
4703 First(Node node) {
4704 this.atom = BnM.optimize(node);
4705 }
4706 boolean match(Matcher matcher, int i, CharSequence seq) {
4707 if (atom instanceof BnM) {
4708 return atom.match(matcher, i, seq)
4709 && next.match(matcher, matcher.last, seq);
4710 }
4711 for (;;) {
4712 if (i > matcher.to) {
4713 matcher.hitEnd = true;
4714 return false;
4715 }
4716 if (atom.match(matcher, i, seq)) {
4717 return next.match(matcher, matcher.last, seq);
4718 }
4719 i += countChars(seq, i, 1);
4720 matcher.first++;
4721 }
4722 }
4723 boolean study(TreeInfo info) {
4724 atom.study(info);
4725 info.maxValid = false;
4726 info.deterministic = false;
4727 return next.study(info);
4728 }
4729 }
4730
4731 static final class Conditional extends Node {
4732 Node cond, yes, not;
4733 Conditional(Node cond, Node yes, Node not) {
4734 this.cond = cond;
4735 this.yes = yes;
4736 this.not = not;
4737 }
4738 boolean match(Matcher matcher, int i, CharSequence seq) {
4739 if (cond.match(matcher, i, seq)) {
4740 return yes.match(matcher, i, seq);
4741 } else {
4742 return not.match(matcher, i, seq);
4743 }
4744 }
4745 boolean study(TreeInfo info) {
4746 int minL = info.minLength;
4747 int maxL = info.maxLength;
4748 boolean maxV = info.maxValid;
4749 info.reset();
4750 yes.study(info);
4751
4752 int minL2 = info.minLength;
4753 int maxL2 = info.maxLength;
4754 boolean maxV2 = info.maxValid;
4755 info.reset();
4756 not.study(info);
4757
4758 info.minLength = minL + Math.min(minL2, info.minLength);
4759 info.maxLength = maxL + Math.max(maxL2, info.maxLength);
4760 info.maxValid = (maxV & maxV2 & info.maxValid);
4761 info.deterministic = false;
4762 return next.study(info);
4763 }
4764 }
4765
4766 /**
4767 * Zero width positive lookahead.
4768 */
4769 static final class Pos extends Node {
4770 Node cond;
4771 Pos(Node cond) {
4772 this.cond = cond;
4773 }
4774 boolean match(Matcher matcher, int i, CharSequence seq) {
4775 int savedTo = matcher.to;
4776 boolean conditionMatched = false;
4777
4778 // Relax transparent region boundaries for lookahead
4779 if (matcher.transparentBounds)
4780 matcher.to = matcher.getTextLength();
4781 try {
4782 conditionMatched = cond.match(matcher, i, seq);
4783 } finally {
4784 // Reinstate region boundaries
4785 matcher.to = savedTo;
4786 }
4787 return conditionMatched && next.match(matcher, i, seq);
4788 }
4789 }
4790
4791 /**
4792 * Zero width negative lookahead.
4793 */
4794 static final class Neg extends Node {
4795 Node cond;
4796 Neg(Node cond) {
4797 this.cond = cond;
4798 }
4799 boolean match(Matcher matcher, int i, CharSequence seq) {
4800 int savedTo = matcher.to;
4801 boolean conditionMatched = false;
4802
4803 // Relax transparent region boundaries for lookahead
4804 if (matcher.transparentBounds)
4805 matcher.to = matcher.getTextLength();
4806 try {
4807 if (i < matcher.to) {
4808 conditionMatched = !cond.match(matcher, i, seq);
4809 } else {
4810 // If a negative lookahead succeeds then more input
4811 // could cause it to fail!
4812 matcher.requireEnd = true;
4813 conditionMatched = !cond.match(matcher, i, seq);
4814 }
4815 } finally {
4816 // Reinstate region boundaries
4817 matcher.to = savedTo;
4818 }
4819 return conditionMatched && next.match(matcher, i, seq);
4820 }
4821 }
4822
4823 /**
4824 * For use with lookbehinds; matches the position where the lookbehind
4825 * was encountered.
4826 */
4827 static Node lookbehindEnd = new Node() {
4828 boolean match(Matcher matcher, int i, CharSequence seq) {
4829 return i == matcher.lookbehindTo;
4830 }
4831 };
4832
4833 /**
4834 * Zero width positive lookbehind.
4835 */
4836 static class Behind extends Node {
4837 Node cond;
4838 int rmax, rmin;
4839 Behind(Node cond, int rmax, int rmin) {
4840 this.cond = cond;
4841 this.rmax = rmax;
4842 this.rmin = rmin;
4843 }
4844
4845 boolean match(Matcher matcher, int i, CharSequence seq) {
4846 int savedFrom = matcher.from;
4847 boolean conditionMatched = false;
4848 int startIndex = (!matcher.transparentBounds) ?
4849 matcher.from : 0;
4850 int from = Math.max(i - rmax, startIndex);
4851 // Set end boundary
4852 int savedLBT = matcher.lookbehindTo;
4853 matcher.lookbehindTo = i;
4854 // Relax transparent region boundaries for lookbehind
4855 if (matcher.transparentBounds)
4856 matcher.from = 0;
4857 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
4858 conditionMatched = cond.match(matcher, j, seq);
4859 }
4860 matcher.from = savedFrom;
4861 matcher.lookbehindTo = savedLBT;
4862 return conditionMatched && next.match(matcher, i, seq);
4863 }
4864 }
4865
4866 /**
4867 * Zero width positive lookbehind, including supplementary
4868 * characters or unpaired surrogates.
4869 */
4870 static final class BehindS extends Behind {
4871 BehindS(Node cond, int rmax, int rmin) {
4872 super(cond, rmax, rmin);
4873 }
4874 boolean match(Matcher matcher, int i, CharSequence seq) {
4875 int rmaxChars = countChars(seq, i, -rmax);
4876 int rminChars = countChars(seq, i, -rmin);
4877 int savedFrom = matcher.from;
4878 int startIndex = (!matcher.transparentBounds) ?
4879 matcher.from : 0;
4880 boolean conditionMatched = false;
4881 int from = Math.max(i - rmaxChars, startIndex);
4882 // Set end boundary
4883 int savedLBT = matcher.lookbehindTo;
4884 matcher.lookbehindTo = i;
4885 // Relax transparent region boundaries for lookbehind
4886 if (matcher.transparentBounds)
4887 matcher.from = 0;
4888
4889 for (int j = i - rminChars;
4890 !conditionMatched && j >= from;
4891 j -= j>from ? countChars(seq, j, -1) : 1) {
4892 conditionMatched = cond.match(matcher, j, seq);
4893 }
4894 matcher.from = savedFrom;
4895 matcher.lookbehindTo = savedLBT;
4896 return conditionMatched && next.match(matcher, i, seq);
4897 }
4898 }
4899
4900 /**
4901 * Zero width negative lookbehind.
4902 */
4903 static class NotBehind extends Node {
4904 Node cond;
4905 int rmax, rmin;
4906 NotBehind(Node cond, int rmax, int rmin) {
4907 this.cond = cond;
4908 this.rmax = rmax;
4909 this.rmin = rmin;
4910 }
4911
4912 boolean match(Matcher matcher, int i, CharSequence seq) {
4913 int savedLBT = matcher.lookbehindTo;
4914 int savedFrom = matcher.from;
4915 boolean conditionMatched = false;
4916 int startIndex = (!matcher.transparentBounds) ?
4917 matcher.from : 0;
4918 int from = Math.max(i - rmax, startIndex);
4919 matcher.lookbehindTo = i;
4920 // Relax transparent region boundaries for lookbehind
4921 if (matcher.transparentBounds)
4922 matcher.from = 0;
4923 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
4924 conditionMatched = cond.match(matcher, j, seq);
4925 }
4926 // Reinstate region boundaries
4927 matcher.from = savedFrom;
4928 matcher.lookbehindTo = savedLBT;
4929 return !conditionMatched && next.match(matcher, i, seq);
4930 }
4931 }
4932
4933 /**
4934 * Zero width negative lookbehind, including supplementary
4935 * characters or unpaired surrogates.
4936 */
4937 static final class NotBehindS extends NotBehind {
4938 NotBehindS(Node cond, int rmax, int rmin) {
4939 super(cond, rmax, rmin);
4940 }
4941 boolean match(Matcher matcher, int i, CharSequence seq) {
4942 int rmaxChars = countChars(seq, i, -rmax);
4943 int rminChars = countChars(seq, i, -rmin);
4944 int savedFrom = matcher.from;
4945 int savedLBT = matcher.lookbehindTo;
4946 boolean conditionMatched = false;
4947 int startIndex = (!matcher.transparentBounds) ?
4948 matcher.from : 0;
4949 int from = Math.max(i - rmaxChars, startIndex);
4950 matcher.lookbehindTo = i;
4951 // Relax transparent region boundaries for lookbehind
4952 if (matcher.transparentBounds)
4953 matcher.from = 0;
4954 for (int j = i - rminChars;
4955 !conditionMatched && j >= from;
4956 j -= j>from ? countChars(seq, j, -1) : 1) {
4957 conditionMatched = cond.match(matcher, j, seq);
4958 }
4959 //Reinstate region boundaries
4960 matcher.from = savedFrom;
4961 matcher.lookbehindTo = savedLBT;
4962 return !conditionMatched && next.match(matcher, i, seq);
4963 }
4964 }
4965
4966 /**
4967 * Returns the set union of two CharProperty nodes.
4968 */
4969 private static CharProperty union(final CharProperty lhs,
4970 final CharProperty rhs) {
4971 return new CharProperty() {
4972 boolean isSatisfiedBy(int ch) {
4973 return lhs.isSatisfiedBy(ch) || rhs.isSatisfiedBy(ch);}};
4974 }
4975
4976 /**
4977 * Returns the set intersection of two CharProperty nodes.
4978 */
4979 private static CharProperty intersection(final CharProperty lhs,
4980 final CharProperty rhs) {
4981 return new CharProperty() {
4982 boolean isSatisfiedBy(int ch) {
4983 return lhs.isSatisfiedBy(ch) && rhs.isSatisfiedBy(ch);}};
4984 }
4985
4986 /**
4987 * Returns the set difference of two CharProperty nodes.
4988 */
4989 private static CharProperty setDifference(final CharProperty lhs,
4990 final CharProperty rhs) {
4991 return new CharProperty() {
4992 boolean isSatisfiedBy(int ch) {
4993 return ! rhs.isSatisfiedBy(ch) && lhs.isSatisfiedBy(ch);}};
4994 }
4995
4996 /**
4997 * Handles word boundaries. Includes a field to allow this one class to
4998 * deal with the different types of word boundaries we can match. The word
4999 * characters include underscores, letters, and digits. Non spacing marks
5000 * can are also part of a word if they have a base character, otherwise
5001 * they are ignored for purposes of finding word boundaries.
5002 */
5003 static final class Bound extends Node {
5004 static int LEFT = 0x1;
5005 static int RIGHT= 0x2;
5006 static int BOTH = 0x3;
5007 static int NONE = 0x4;
5008 int type;
5009 Bound(int n) {
5010 type = n;
5011 }
5012 int check(Matcher matcher, int i, CharSequence seq) {
5013 int ch;
5014 boolean left = false;
5015 int startIndex = matcher.from;
5016 int endIndex = matcher.to;
5017 if (matcher.transparentBounds) {
5018 startIndex = 0;
5019 endIndex = matcher.getTextLength();
5020 }
5021 if (i > startIndex) {
5022 ch = Character.codePointBefore(seq, i);
5023 left = (ch == '_' || Character.isLetterOrDigit(ch) ||
5024 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5025 && hasBaseCharacter(matcher, i-1, seq)));
5026 }
5027 boolean right = false;
5028 if (i < endIndex) {
5029 ch = Character.codePointAt(seq, i);
5030 right = (ch == '_' || Character.isLetterOrDigit(ch) ||
5031 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5032 && hasBaseCharacter(matcher, i, seq)));
5033 } else {
5034 // Tried to access char past the end
5035 matcher.hitEnd = true;
5036 // The addition of another char could wreck a boundary
5037 matcher.requireEnd = true;
5038 }
5039 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE);
5040 }
5041 boolean match(Matcher matcher, int i, CharSequence seq) {
5042 return (check(matcher, i, seq) & type) > 0
5043 && next.match(matcher, i, seq);
5044 }
5045 }
5046
5047 /**
5048 * Non spacing marks only count as word characters in bounds calculations
5049 * if they have a base character.
5050 */
5051 private static boolean hasBaseCharacter(Matcher matcher, int i,
5052 CharSequence seq)
5053 {
5054 int start = (!matcher.transparentBounds) ?
5055 matcher.from : 0;
5056 for (int x=i; x >= start; x--) {
5057 int ch = Character.codePointAt(seq, x);
5058 if (Character.isLetterOrDigit(ch))
5059 return true;
5060 if (Character.getType(ch) == Character.NON_SPACING_MARK)
5061 continue;
5062 return false;
5063 }
5064 return false;
5065 }
5066
5067 /**
5068 * Attempts to match a slice in the input using the Boyer-Moore string
5069 * matching algorithm. The algorithm is based on the idea that the
5070 * pattern can be shifted farther ahead in the search text if it is
5071 * matched right to left.
5072 * <p>
5073 * The pattern is compared to the input one character at a time, from
5074 * the rightmost character in the pattern to the left. If the characters
5075 * all match the pattern has been found. If a character does not match,
5076 * the pattern is shifted right a distance that is the maximum of two
5077 * functions, the bad character shift and the good suffix shift. This
5078 * shift moves the attempted match position through the input more
5079 * quickly than a naive one position at a time check.
5080 * <p>
5081 * The bad character shift is based on the character from the text that
5082 * did not match. If the character does not appear in the pattern, the
5083 * pattern can be shifted completely beyond the bad character. If the
5084 * character does occur in the pattern, the pattern can be shifted to
5085 * line the pattern up with the next occurrence of that character.
5086 * <p>
5087 * The good suffix shift is based on the idea that some subset on the right
5088 * side of the pattern has matched. When a bad character is found, the
5089 * pattern can be shifted right by the pattern length if the subset does
5090 * not occur again in pattern, or by the amount of distance to the
5091 * next occurrence of the subset in the pattern.
5092 *
5093 * Boyer-Moore search methods adapted from code by Amy Yu.
5094 */
5095 static class BnM extends Node {
5096 int[] buffer;
5097 int[] lastOcc;
5098 int[] optoSft;
5099
5100 /**
5101 * Pre calculates arrays needed to generate the bad character
5102 * shift and the good suffix shift. Only the last seven bits
5103 * are used to see if chars match; This keeps the tables small
5104 * and covers the heavily used ASCII range, but occasionally
5105 * results in an aliased match for the bad character shift.
5106 */
5107 static Node optimize(Node node) {
5108 if (!(node instanceof Slice)) {
5109 return node;
5110 }
5111
5112 int[] src = ((Slice) node).buffer;
5113 int patternLength = src.length;
5114 // The BM algorithm requires a bit of overhead;
5115 // If the pattern is short don't use it, since
5116 // a shift larger than the pattern length cannot
5117 // be used anyway.
5118 if (patternLength < 4) {
5119 return node;
5120 }
5121 int i, j, k;
5122 int[] lastOcc = new int[128];
5123 int[] optoSft = new int[patternLength];
5124 // Precalculate part of the bad character shift
5125 // It is a table for where in the pattern each
5126 // lower 7-bit value occurs
5127 for (i = 0; i < patternLength; i++) {
5128 lastOcc[src[i]&0x7F] = i + 1;
5129 }
5130 // Precalculate the good suffix shift
5131 // i is the shift amount being considered
5132 NEXT: for (i = patternLength; i > 0; i--) {
5133 // j is the beginning index of suffix being considered
5134 for (j = patternLength - 1; j >= i; j--) {
5135 // Testing for good suffix
5136 if (src[j] == src[j-i]) {
5137 // src[j..len] is a good suffix
5138 optoSft[j-1] = i;
5139 } else {
5140 // No match. The array has already been
5141 // filled up with correct values before.
5142 continue NEXT;
5143 }
5144 }
5145 // This fills up the remaining of optoSft
5146 // any suffix can not have larger shift amount
5147 // then its sub-suffix. Why???
5148 while (j > 0) {
5149 optoSft[--j] = i;
5150 }
5151 }
5152 // Set the guard value because of unicode compression
5153 optoSft[patternLength-1] = 1;
5154 if (node instanceof SliceS)
5155 return new BnMS(src, lastOcc, optoSft, node.next);
5156 return new BnM(src, lastOcc, optoSft, node.next);
5157 }
5158 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5159 this.buffer = src;
5160 this.lastOcc = lastOcc;
5161 this.optoSft = optoSft;
5162 this.next = next;
5163 }
5164 boolean match(Matcher matcher, int i, CharSequence seq) {
5165 int[] src = buffer;
5166 int patternLength = src.length;
5167 int last = matcher.to - patternLength;
5168
5169 // Loop over all possible match positions in text
5170 NEXT: while (i <= last) {
5171 // Loop over pattern from right to left
5172 for (int j = patternLength - 1; j >= 0; j--) {
5173 int ch = seq.charAt(i+j);
5174 if (ch != src[j]) {
5175 // Shift search to the right by the maximum of the
5176 // bad character shift and the good suffix shift
5177 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]);
5178 continue NEXT;
5179 }
5180 }
5181 // Entire pattern matched starting at i
5182 matcher.first = i;
5183 boolean ret = next.match(matcher, i + patternLength, seq);
5184 if (ret) {
5185 matcher.first = i;
5186 matcher.groups[0] = matcher.first;
5187 matcher.groups[1] = matcher.last;
5188 return true;
5189 }
5190 i++;
5191 }
5192 // BnM is only used as the leading node in the unanchored case,
5193 // and it replaced its Start() which always searches to the end
5194 // if it doesn't find what it's looking for, so hitEnd is true.
5195 matcher.hitEnd = true;
5196 return false;
5197 }
5198 boolean study(TreeInfo info) {
5199 info.minLength += buffer.length;
5200 info.maxValid = false;
5201 return next.study(info);
5202 }
5203 }
5204
5205 /**
5206 * Supplementary support version of BnM(). Unpaired surrogates are
5207 * also handled by this class.
5208 */
5209 static final class BnMS extends BnM {
5210 int lengthInChars;
5211
5212 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5213 super(src, lastOcc, optoSft, next);
5214 for (int x = 0; x < buffer.length; x++) {
5215 lengthInChars += Character.charCount(buffer[x]);
5216 }
5217 }
5218 boolean match(Matcher matcher, int i, CharSequence seq) {
5219 int[] src = buffer;
5220 int patternLength = src.length;
5221 int last = matcher.to - lengthInChars;
5222
5223 // Loop over all possible match positions in text
5224 NEXT: while (i <= last) {
5225 // Loop over pattern from right to left
5226 int ch;
5227 for (int j = countChars(seq, i, patternLength), x = patternLength - 1;
5228 j > 0; j -= Character.charCount(ch), x--) {
5229 ch = Character.codePointBefore(seq, i+j);
5230 if (ch != src[x]) {
5231 // Shift search to the right by the maximum of the
5232 // bad character shift and the good suffix shift
5233 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]);
5234 i += countChars(seq, i, n);
5235 continue NEXT;
5236 }
5237 }
5238 // Entire pattern matched starting at i
5239 matcher.first = i;
5240 boolean ret = next.match(matcher, i + lengthInChars, seq);
5241 if (ret) {
5242 matcher.first = i;
5243 matcher.groups[0] = matcher.first;
5244 matcher.groups[1] = matcher.last;
5245 return true;
5246 }
5247 i += countChars(seq, i, 1);
5248 }
5249 matcher.hitEnd = true;
5250 return false;
5251 }
5252 }
5253
5254 ///////////////////////////////////////////////////////////////////////////////
5255 ///////////////////////////////////////////////////////////////////////////////
5256
5257 /**
5258 * This must be the very first initializer.
5259 */
5260 static Node accept = new Node();
5261
5262 static Node lastAccept = new LastNode();
5263
5264 private static class CharPropertyNames {
5265
5266 static CharProperty charPropertyFor(String name) {
5267 CharPropertyFactory m = map.get(name);
5268 return m == null ? null : m.make();
5269 }
5270
5271 private static abstract class CharPropertyFactory {
5272 abstract CharProperty make();
5273 }
5274
5275 private static void defCategory(String name,
5276 final int typeMask) {
5277 map.put(name, new CharPropertyFactory() {
5278 CharProperty make() { return new Category(typeMask);}});
5279 }
5280
5281 private static void defRange(String name,
5282 final int lower, final int upper) {
5283 map.put(name, new CharPropertyFactory() {
5284 CharProperty make() { return rangeFor(lower, upper);}});
5285 }
5286
5287 private static void defCtype(String name,
5288 final int ctype) {
5289 map.put(name, new CharPropertyFactory() {
5290 CharProperty make() { return new Ctype(ctype);}});
5291 }
5292
5293 private static abstract class CloneableProperty
5294 extends CharProperty implements Cloneable
5295 {
5296 public CloneableProperty clone() {
5297 try {
5298 return (CloneableProperty) super.clone();
5299 } catch (CloneNotSupportedException e) {
5300 throw new AssertionError(e);
5301 }
5302 }
5303 }
5304
5305 private static void defClone(String name,
5306 final CloneableProperty p) {
5307 map.put(name, new CharPropertyFactory() {
5308 CharProperty make() { return p.clone();}});
5309 }
5310
5311 private static final HashMap<String, CharPropertyFactory> map
5312 = new HashMap<String, CharPropertyFactory>();
5313
5314 static {
5315 // Unicode character property aliases, defined in
5316 // http://www.unicode.org/Public/UNIDATA/PropertyValueAliases.txt
5317 defCategory("Cn", 1<<Character.UNASSIGNED);
5318 defCategory("Lu", 1<<Character.UPPERCASE_LETTER);
5319 defCategory("Ll", 1<<Character.LOWERCASE_LETTER);
5320 defCategory("Lt", 1<<Character.TITLECASE_LETTER);
5321 defCategory("Lm", 1<<Character.MODIFIER_LETTER);
5322 defCategory("Lo", 1<<Character.OTHER_LETTER);
5323 defCategory("Mn", 1<<Character.NON_SPACING_MARK);
5324 defCategory("Me", 1<<Character.ENCLOSING_MARK);
5325 defCategory("Mc", 1<<Character.COMBINING_SPACING_MARK);
5326 defCategory("Nd", 1<<Character.DECIMAL_DIGIT_NUMBER);
5327 defCategory("Nl", 1<<Character.LETTER_NUMBER);
5328 defCategory("No", 1<<Character.OTHER_NUMBER);
5329 defCategory("Zs", 1<<Character.SPACE_SEPARATOR);
5330 defCategory("Zl", 1<<Character.LINE_SEPARATOR);
5331 defCategory("Zp", 1<<Character.PARAGRAPH_SEPARATOR);
5332 defCategory("Cc", 1<<Character.CONTROL);
5333 defCategory("Cf", 1<<Character.FORMAT);
5334 defCategory("Co", 1<<Character.PRIVATE_USE);
5335 defCategory("Cs", 1<<Character.SURROGATE);
5336 defCategory("Pd", 1<<Character.DASH_PUNCTUATION);
5337 defCategory("Ps", 1<<Character.START_PUNCTUATION);
5338 defCategory("Pe", 1<<Character.END_PUNCTUATION);
5339 defCategory("Pc", 1<<Character.CONNECTOR_PUNCTUATION);
5340 defCategory("Po", 1<<Character.OTHER_PUNCTUATION);
5341 defCategory("Sm", 1<<Character.MATH_SYMBOL);
5342 defCategory("Sc", 1<<Character.CURRENCY_SYMBOL);
5343 defCategory("Sk", 1<<Character.MODIFIER_SYMBOL);
5344 defCategory("So", 1<<Character.OTHER_SYMBOL);
5345 defCategory("Pi", 1<<Character.INITIAL_QUOTE_PUNCTUATION);
5346 defCategory("Pf", 1<<Character.FINAL_QUOTE_PUNCTUATION);
5347 defCategory("L", ((1<<Character.UPPERCASE_LETTER) |
5348 (1<<Character.LOWERCASE_LETTER) |
5349 (1<<Character.TITLECASE_LETTER) |
5350 (1<<Character.MODIFIER_LETTER) |
5351 (1<<Character.OTHER_LETTER)));
5352 defCategory("M", ((1<<Character.NON_SPACING_MARK) |
5353 (1<<Character.ENCLOSING_MARK) |
5354 (1<<Character.COMBINING_SPACING_MARK)));
5355 defCategory("N", ((1<<Character.DECIMAL_DIGIT_NUMBER) |
5356 (1<<Character.LETTER_NUMBER) |
5357 (1<<Character.OTHER_NUMBER)));
5358 defCategory("Z", ((1<<Character.SPACE_SEPARATOR) |
5359 (1<<Character.LINE_SEPARATOR) |
5360 (1<<Character.PARAGRAPH_SEPARATOR)));
5361 defCategory("C", ((1<<Character.CONTROL) |
5362 (1<<Character.FORMAT) |
5363 (1<<Character.PRIVATE_USE) |
5364 (1<<Character.SURROGATE))); // Other
5365 defCategory("P", ((1<<Character.DASH_PUNCTUATION) |
5366 (1<<Character.START_PUNCTUATION) |
5367 (1<<Character.END_PUNCTUATION) |
5368 (1<<Character.CONNECTOR_PUNCTUATION) |
5369 (1<<Character.OTHER_PUNCTUATION) |
5370 (1<<Character.INITIAL_QUOTE_PUNCTUATION) |
5371 (1<<Character.FINAL_QUOTE_PUNCTUATION)));
5372 defCategory("S", ((1<<Character.MATH_SYMBOL) |
5373 (1<<Character.CURRENCY_SYMBOL) |
5374 (1<<Character.MODIFIER_SYMBOL) |
5375 (1<<Character.OTHER_SYMBOL)));
5376 defCategory("LC", ((1<<Character.UPPERCASE_LETTER) |
5377 (1<<Character.LOWERCASE_LETTER) |
5378 (1<<Character.TITLECASE_LETTER)));
5379 defCategory("LD", ((1<<Character.UPPERCASE_LETTER) |
5380 (1<<Character.LOWERCASE_LETTER) |
5381 (1<<Character.TITLECASE_LETTER) |
5382 (1<<Character.MODIFIER_LETTER) |
5383 (1<<Character.OTHER_LETTER) |
5384 (1<<Character.DECIMAL_DIGIT_NUMBER)));
5385 defRange("L1", 0x00, 0xFF); // Latin-1
5386 map.put("all", new CharPropertyFactory() {
5387 CharProperty make() { return new All(); }});
5388
5389 // Posix regular expression character classes, defined in
5390 // http://www.unix.org/onlinepubs/009695399/basedefs/xbd_chap09.html
5391 defRange("ASCII", 0x00, 0x7F); // ASCII
5392 defCtype("Alnum", ASCII.ALNUM); // Alphanumeric characters
5393 defCtype("Alpha", ASCII.ALPHA); // Alphabetic characters
5394 defCtype("Blank", ASCII.BLANK); // Space and tab characters
5395 defCtype("Cntrl", ASCII.CNTRL); // Control characters
5396 defRange("Digit", '0', '9'); // Numeric characters
5397 defCtype("Graph", ASCII.GRAPH); // printable and visible
5398 defRange("Lower", 'a', 'z'); // Lower-case alphabetic
5399 defRange("Print", 0x20, 0x7E); // Printable characters
5400 defCtype("Punct", ASCII.PUNCT); // Punctuation characters
5401 defCtype("Space", ASCII.SPACE); // Space characters
5402 defRange("Upper", 'A', 'Z'); // Upper-case alphabetic
5403 defCtype("XDigit",ASCII.XDIGIT); // hexadecimal digits
5404
5405 // Java character properties, defined by methods in Character.java
5406 defClone("javaLowerCase", new CloneableProperty() {
5407 boolean isSatisfiedBy(int ch) {
5408 return Character.isLowerCase(ch);}});
5409 defClone("javaUpperCase", new CloneableProperty() {
5410 boolean isSatisfiedBy(int ch) {
5411 return Character.isUpperCase(ch);}});
5412 defClone("javaTitleCase", new CloneableProperty() {
5413 boolean isSatisfiedBy(int ch) {
5414 return Character.isTitleCase(ch);}});
5415 defClone("javaDigit", new CloneableProperty() {
5416 boolean isSatisfiedBy(int ch) {
5417 return Character.isDigit(ch);}});
5418 defClone("javaDefined", new CloneableProperty() {
5419 boolean isSatisfiedBy(int ch) {
5420 return Character.isDefined(ch);}});
5421 defClone("javaLetter", new CloneableProperty() {
5422 boolean isSatisfiedBy(int ch) {
5423 return Character.isLetter(ch);}});
5424 defClone("javaLetterOrDigit", new CloneableProperty() {
5425 boolean isSatisfiedBy(int ch) {
5426 return Character.isLetterOrDigit(ch);}});
5427 defClone("javaJavaIdentifierStart", new CloneableProperty() {
5428 boolean isSatisfiedBy(int ch) {
5429 return Character.isJavaIdentifierStart(ch);}});
5430 defClone("javaJavaIdentifierPart", new CloneableProperty() {
5431 boolean isSatisfiedBy(int ch) {
5432 return Character.isJavaIdentifierPart(ch);}});
5433 defClone("javaUnicodeIdentifierStart", new CloneableProperty() {
5434 boolean isSatisfiedBy(int ch) {
5435 return Character.isUnicodeIdentifierStart(ch);}});
5436 defClone("javaUnicodeIdentifierPart", new CloneableProperty() {
5437 boolean isSatisfiedBy(int ch) {
5438 return Character.isUnicodeIdentifierPart(ch);}});
5439 defClone("javaIdentifierIgnorable", new CloneableProperty() {
5440 boolean isSatisfiedBy(int ch) {
5441 return Character.isIdentifierIgnorable(ch);}});
5442 defClone("javaSpaceChar", new CloneableProperty() {
5443 boolean isSatisfiedBy(int ch) {
5444 return Character.isSpaceChar(ch);}});
5445 defClone("javaWhitespace", new CloneableProperty() {
5446 boolean isSatisfiedBy(int ch) {
5447 return Character.isWhitespace(ch);}});
5448 defClone("javaISOControl", new CloneableProperty() {
5449 boolean isSatisfiedBy(int ch) {
5450 return Character.isISOControl(ch);}});
5451 defClone("javaMirrored", new CloneableProperty() {
5452 boolean isSatisfiedBy(int ch) {
5453 return Character.isMirrored(ch);}});
5454 }
5455 }
5456 }