1 /*
2 * Copyright 1999-2009 Sun Microsystems, Inc. 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. Sun designates this
8 * particular file as subject to the "Classpath" exception as provided
9 * by Sun 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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
22 * CA 95054 USA or visit www.sun.com if you need additional information or
23 * have any 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.Map;
33 import java.util.ArrayList;
34 import java.util.HashMap;
35 import java.util.Arrays;
36
37
38 /**
39 * A compiled representation of a regular expression.
40 *
41 * <p> A regular expression, specified as a string, must first be compiled into
42 * an instance of this class. The resulting pattern can then be used to create
43 * a {@link Matcher} object that can match arbitrary {@link
44 * java.lang.CharSequence </code>character sequences<code>} against the regular
45 * expression. All of the state involved in performing a match resides in the
46 * matcher, so many matchers can share the same pattern.
47 *
48 * <p> A typical invocation sequence is thus
49 *
50 * <blockquote><pre>
51 * Pattern p = Pattern.{@link #compile compile}("a*b");
52 * Matcher m = p.{@link #matcher matcher}("aaaaab");
53 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote>
54 *
55 * <p> A {@link #matches matches} method is defined by this class as a
56 * convenience for when a regular expression is used just once. This method
57 * compiles an expression and matches an input sequence against it in a single
58 * invocation. The statement
59 *
60 * <blockquote><pre>
61 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote>
62 *
63 * is equivalent to the three statements above, though for repeated matches it
64 * is less efficient since it does not allow the compiled pattern to be reused.
65 *
66 * <p> Instances of this class are immutable and are safe for use by multiple
67 * concurrent threads. Instances of the {@link Matcher} class are not safe for
68 * such use.
69 *
70 *
71 * <a name="sum">
72 * <h4> Summary of regular-expression constructs </h4>
73 *
74 * <table border="0" cellpadding="1" cellspacing="0"
75 * summary="Regular expression constructs, and what they match">
76 *
77 * <tr align="left">
78 * <th bgcolor="#CCCCFF" align="left" id="construct">Construct</th>
79 * <th bgcolor="#CCCCFF" align="left" id="matches">Matches</th>
80 * </tr>
81 *
82 * <tr><th> </th></tr>
83 * <tr align="left"><th colspan="2" id="characters">Characters</th></tr>
84 *
85 * <tr><td valign="top" headers="construct characters"><i>x</i></td>
86 * <td headers="matches">The character <i>x</i></td></tr>
87 * <tr><td valign="top" headers="construct characters"><tt>\\</tt></td>
88 * <td headers="matches">The backslash character</td></tr>
89 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>n</i></td>
90 * <td headers="matches">The character with octal value <tt>0</tt><i>n</i>
91 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
92 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>nn</i></td>
93 * <td headers="matches">The character with octal value <tt>0</tt><i>nn</i>
94 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
95 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>mnn</i></td>
96 * <td headers="matches">The character with octal value <tt>0</tt><i>mnn</i>
97 * (0 <tt><=</tt> <i>m</i> <tt><=</tt> 3,
98 * 0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
99 * <tr><td valign="top" headers="construct characters"><tt>\x</tt><i>hh</i></td>
100 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hh</i></td></tr>
101 * <tr><td valign="top" headers="construct characters"><tt>\u</tt><i>hhhh</i></td>
102 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hhhh</i></td></tr>
103 * <tr><td valign="top" headers="matches"><tt>\t</tt></td>
104 * <td headers="matches">The tab character (<tt>'\u0009'</tt>)</td></tr>
105 * <tr><td valign="top" headers="construct characters"><tt>\n</tt></td>
106 * <td headers="matches">The newline (line feed) character (<tt>'\u000A'</tt>)</td></tr>
107 * <tr><td valign="top" headers="construct characters"><tt>\r</tt></td>
108 * <td headers="matches">The carriage-return character (<tt>'\u000D'</tt>)</td></tr>
109 * <tr><td valign="top" headers="construct characters"><tt>\f</tt></td>
110 * <td headers="matches">The form-feed character (<tt>'\u000C'</tt>)</td></tr>
111 * <tr><td valign="top" headers="construct characters"><tt>\a</tt></td>
112 * <td headers="matches">The alert (bell) character (<tt>'\u0007'</tt>)</td></tr>
113 * <tr><td valign="top" headers="construct characters"><tt>\e</tt></td>
114 * <td headers="matches">The escape character (<tt>'\u001B'</tt>)</td></tr>
115 * <tr><td valign="top" headers="construct characters"><tt>\c</tt><i>x</i></td>
116 * <td headers="matches">The control character corresponding to <i>x</i></td></tr>
117 *
118 * <tr><th> </th></tr>
119 * <tr align="left"><th colspan="2" id="classes">Character classes</th></tr>
120 *
121 * <tr><td valign="top" headers="construct classes"><tt>[abc]</tt></td>
122 * <td headers="matches"><tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (simple class)</td></tr>
123 * <tr><td valign="top" headers="construct classes"><tt>[^abc]</tt></td>
124 * <td headers="matches">Any character except <tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (negation)</td></tr>
125 * <tr><td valign="top" headers="construct classes"><tt>[a-zA-Z]</tt></td>
126 * <td headers="matches"><tt>a</tt> through <tt>z</tt>
127 * or <tt>A</tt> through <tt>Z</tt>, inclusive (range)</td></tr>
128 * <tr><td valign="top" headers="construct classes"><tt>[a-d[m-p]]</tt></td>
129 * <td headers="matches"><tt>a</tt> through <tt>d</tt>,
130 * or <tt>m</tt> through <tt>p</tt>: <tt>[a-dm-p]</tt> (union)</td></tr>
131 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[def]]</tt></td>
132 * <td headers="matches"><tt>d</tt>, <tt>e</tt>, or <tt>f</tt> (intersection)</tr>
133 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^bc]]</tt></td>
134 * <td headers="matches"><tt>a</tt> through <tt>z</tt>,
135 * except for <tt>b</tt> and <tt>c</tt>: <tt>[ad-z]</tt> (subtraction)</td></tr>
136 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^m-p]]</tt></td>
137 * <td headers="matches"><tt>a</tt> through <tt>z</tt>,
138 * and not <tt>m</tt> through <tt>p</tt>: <tt>[a-lq-z]</tt>(subtraction)</td></tr>
139 * <tr><th> </th></tr>
140 *
141 * <tr align="left"><th colspan="2" id="predef">Predefined character classes</th></tr>
142 *
143 * <tr><td valign="top" headers="construct predef"><tt>.</tt></td>
144 * <td headers="matches">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr>
145 * <tr><td valign="top" headers="construct predef"><tt>\d</tt></td>
146 * <td headers="matches">A digit: <tt>[0-9]</tt></td></tr>
147 * <tr><td valign="top" headers="construct predef"><tt>\D</tt></td>
148 * <td headers="matches">A non-digit: <tt>[^0-9]</tt></td></tr>
149 * <tr><td valign="top" headers="construct predef"><tt>\s</tt></td>
150 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr>
151 * <tr><td valign="top" headers="construct predef"><tt>\S</tt></td>
152 * <td headers="matches">A non-whitespace character: <tt>[^\s]</tt></td></tr>
153 * <tr><td valign="top" headers="construct predef"><tt>\w</tt></td>
154 * <td headers="matches">A word character: <tt>[a-zA-Z_0-9]</tt></td></tr>
155 * <tr><td valign="top" headers="construct predef"><tt>\W</tt></td>
156 * <td headers="matches">A non-word character: <tt>[^\w]</tt></td></tr>
157 *
158 * <tr><th> </th></tr>
159 * <tr align="left"><th colspan="2" id="posix">POSIX character classes</b> (US-ASCII only)<b></th></tr>
160 *
161 * <tr><td valign="top" headers="construct posix"><tt>\p{Lower}</tt></td>
162 * <td headers="matches">A lower-case alphabetic character: <tt>[a-z]</tt></td></tr>
163 * <tr><td valign="top" headers="construct posix"><tt>\p{Upper}</tt></td>
164 * <td headers="matches">An upper-case alphabetic character:<tt>[A-Z]</tt></td></tr>
165 * <tr><td valign="top" headers="construct posix"><tt>\p{ASCII}</tt></td>
166 * <td headers="matches">All ASCII:<tt>[\x00-\x7F]</tt></td></tr>
167 * <tr><td valign="top" headers="construct posix"><tt>\p{Alpha}</tt></td>
168 * <td headers="matches">An alphabetic character:<tt>[\p{Lower}\p{Upper}]</tt></td></tr>
169 * <tr><td valign="top" headers="construct posix"><tt>\p{Digit}</tt></td>
170 * <td headers="matches">A decimal digit: <tt>[0-9]</tt></td></tr>
171 * <tr><td valign="top" headers="construct posix"><tt>\p{Alnum}</tt></td>
172 * <td headers="matches">An alphanumeric character:<tt>[\p{Alpha}\p{Digit}]</tt></td></tr>
173 * <tr><td valign="top" headers="construct posix"><tt>\p{Punct}</tt></td>
174 * <td headers="matches">Punctuation: One of <tt>!"#$%&'()*+,-./:;<=>?@[\]^_`{|}~</tt></td></tr>
175 * <!-- <tt>[\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]</tt>
176 * <tt>[\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]</tt> -->
177 * <tr><td valign="top" headers="construct posix"><tt>\p{Graph}</tt></td>
178 * <td headers="matches">A visible character: <tt>[\p{Alnum}\p{Punct}]</tt></td></tr>
179 * <tr><td valign="top" headers="construct posix"><tt>\p{Print}</tt></td>
180 * <td headers="matches">A printable character: <tt>[\p{Graph}\x20]</tt></td></tr>
181 * <tr><td valign="top" headers="construct posix"><tt>\p{Blank}</tt></td>
182 * <td headers="matches">A space or a tab: <tt>[ \t]</tt></td></tr>
183 * <tr><td valign="top" headers="construct posix"><tt>\p{Cntrl}</tt></td>
184 * <td headers="matches">A control character: <tt>[\x00-\x1F\x7F]</tt></td></tr>
185 * <tr><td valign="top" headers="construct posix"><tt>\p{XDigit}</tt></td>
186 * <td headers="matches">A hexadecimal digit: <tt>[0-9a-fA-F]</tt></td></tr>
187 * <tr><td valign="top" headers="construct posix"><tt>\p{Space}</tt></td>
188 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr>
189 *
190 * <tr><th> </th></tr>
191 * <tr align="left"><th colspan="2">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr>
192 *
193 * <tr><td valign="top"><tt>\p{javaLowerCase}</tt></td>
194 * <td>Equivalent to java.lang.Character.isLowerCase()</td></tr>
195 * <tr><td valign="top"><tt>\p{javaUpperCase}</tt></td>
196 * <td>Equivalent to java.lang.Character.isUpperCase()</td></tr>
197 * <tr><td valign="top"><tt>\p{javaWhitespace}</tt></td>
198 * <td>Equivalent to java.lang.Character.isWhitespace()</td></tr>
199 * <tr><td valign="top"><tt>\p{javaMirrored}</tt></td>
200 * <td>Equivalent to java.lang.Character.isMirrored()</td></tr>
201 *
202 * <tr><th> </th></tr>
203 * <tr align="left"><th colspan="2" id="unicode">Classes for Unicode blocks and categories</th></tr>
204 *
205 * <tr><td valign="top" headers="construct unicode"><tt>\p{InGreek}</tt></td>
206 * <td headers="matches">A character in the Greek block (simple <a href="#ubc">block</a>)</td></tr>
207 * <tr><td valign="top" headers="construct unicode"><tt>\p{Lu}</tt></td>
208 * <td headers="matches">An uppercase letter (simple <a href="#ubc">category</a>)</td></tr>
209 * <tr><td valign="top" headers="construct unicode"><tt>\p{Sc}</tt></td>
210 * <td headers="matches">A currency symbol</td></tr>
211 * <tr><td valign="top" headers="construct unicode"><tt>\P{InGreek}</tt></td>
212 * <td headers="matches">Any character except one in the Greek block (negation)</td></tr>
213 * <tr><td valign="top" headers="construct unicode"><tt>[\p{L}&&[^\p{Lu}]] </tt></td>
214 * <td headers="matches">Any letter except an uppercase letter (subtraction)</td></tr>
215 *
216 * <tr><th> </th></tr>
217 * <tr align="left"><th colspan="2" id="bounds">Boundary matchers</th></tr>
218 *
219 * <tr><td valign="top" headers="construct bounds"><tt>^</tt></td>
220 * <td headers="matches">The beginning of a line</td></tr>
221 * <tr><td valign="top" headers="construct bounds"><tt>$</tt></td>
222 * <td headers="matches">The end of a line</td></tr>
223 * <tr><td valign="top" headers="construct bounds"><tt>\b</tt></td>
224 * <td headers="matches">A word boundary</td></tr>
225 * <tr><td valign="top" headers="construct bounds"><tt>\B</tt></td>
226 * <td headers="matches">A non-word boundary</td></tr>
227 * <tr><td valign="top" headers="construct bounds"><tt>\A</tt></td>
228 * <td headers="matches">The beginning of the input</td></tr>
229 * <tr><td valign="top" headers="construct bounds"><tt>\G</tt></td>
230 * <td headers="matches">The end of the previous match</td></tr>
231 * <tr><td valign="top" headers="construct bounds"><tt>\Z</tt></td>
232 * <td headers="matches">The end of the input but for the final
233 * <a href="#lt">terminator</a>, if any</td></tr>
234 * <tr><td valign="top" headers="construct bounds"><tt>\z</tt></td>
235 * <td headers="matches">The end of the input</td></tr>
236 *
237 * <tr><th> </th></tr>
238 * <tr align="left"><th colspan="2" id="greedy">Greedy quantifiers</th></tr>
239 *
240 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>?</tt></td>
241 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
242 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>*</tt></td>
243 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
244 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>+</tt></td>
245 * <td headers="matches"><i>X</i>, one or more times</td></tr>
246 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>}</tt></td>
247 * <td headers="matches"><i>X</i>, exactly <i>n</i> 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>, at least <i>n</i> times</td></tr>
250 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}</tt></td>
251 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
252 *
253 * <tr><th> </th></tr>
254 * <tr align="left"><th colspan="2" id="reluc">Reluctant quantifiers</th></tr>
255 *
256 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>??</tt></td>
257 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
258 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>*?</tt></td>
259 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
260 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>+?</tt></td>
261 * <td headers="matches"><i>X</i>, one or more times</td></tr>
262 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>}?</tt></td>
263 * <td headers="matches"><i>X</i>, exactly <i>n</i> 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>, at least <i>n</i> times</td></tr>
266 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}?</tt></td>
267 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
268 *
269 * <tr><th> </th></tr>
270 * <tr align="left"><th colspan="2" id="poss">Possessive quantifiers</th></tr>
271 *
272 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>?+</tt></td>
273 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
274 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>*+</tt></td>
275 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
276 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>++</tt></td>
277 * <td headers="matches"><i>X</i>, one or more times</td></tr>
278 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>}+</tt></td>
279 * <td headers="matches"><i>X</i>, exactly <i>n</i> 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>, at least <i>n</i> times</td></tr>
282 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}+</tt></td>
283 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
284 *
285 * <tr><th> </th></tr>
286 * <tr align="left"><th colspan="2" id="logical">Logical operators</th></tr>
287 *
288 * <tr><td valign="top" headers="construct logical"><i>XY</i></td>
289 * <td headers="matches"><i>X</i> followed by <i>Y</i></td></tr>
290 * <tr><td valign="top" headers="construct logical"><i>X</i><tt>|</tt><i>Y</i></td>
291 * <td headers="matches">Either <i>X</i> or <i>Y</i></td></tr>
292 * <tr><td valign="top" headers="construct logical"><tt>(</tt><i>X</i><tt>)</tt></td>
293 * <td headers="matches">X, as a <a href="#cg">capturing group</a></td></tr>
294 *
295 * <tr><th> </th></tr>
296 * <tr align="left"><th colspan="2" id="backref">Back references</th></tr>
297 *
298 * <tr><td valign="bottom" headers="construct backref"><tt>\</tt><i>n</i></td>
299 * <td valign="bottom" headers="matches">Whatever the <i>n</i><sup>th</sup>
300 * <a href="#cg">capturing group</a> matched</td></tr>
301 *
302 * <tr><td valign="bottom" headers="construct backref"><tt>\</tt><i>k</i><<i>name</i>></td>
303 * <td valign="bottom" headers="matches">Whatever the
304 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr>
305 *
306 * <tr><th> </th></tr>
307 * <tr align="left"><th colspan="2" id="quot">Quotation</th></tr>
308 *
309 * <tr><td valign="top" headers="construct quot"><tt>\</tt></td>
310 * <td headers="matches">Nothing, but quotes the following character</td></tr>
311 * <tr><td valign="top" headers="construct quot"><tt>\Q</tt></td>
312 * <td headers="matches">Nothing, but quotes all characters until <tt>\E</tt></td></tr>
313 * <tr><td valign="top" headers="construct quot"><tt>\E</tt></td>
314 * <td headers="matches">Nothing, but ends quoting started by <tt>\Q</tt></td></tr>
315 * <!-- Metachars: !$()*+.<>?[\]^{|} -->
316 *
317 * <tr><th> </th></tr>
318 * <tr align="left"><th colspan="2" id="special">Special constructs (named-capturing and non-capturing)</th></tr>
319 *
320 * <tr><td valign="top" headers="construct special"><tt>(?<<a href="#groupname">name</a>></tt><i>X</i><tt>)</tt></td>
321 * <td headers="matches"><i>X</i>, as a named-capturing group</td></tr>
322 * <tr><td valign="top" headers="construct special"><tt>(?:</tt><i>X</i><tt>)</tt></td>
323 * <td headers="matches"><i>X</i>, as a non-capturing group</td></tr>
324 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux) </tt></td>
325 * <td headers="matches">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a>
326 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a>
327 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> on - off</td></tr>
328 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux:</tt><i>X</i><tt>)</tt> </td>
329 * <td headers="matches"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the
330 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a>
331 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a >
332 * <a href="#COMMENTS">x</a> on - off</td></tr>
333 * <tr><td valign="top" headers="construct special"><tt>(?=</tt><i>X</i><tt>)</tt></td>
334 * <td headers="matches"><i>X</i>, via zero-width positive lookahead</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 negative 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 positive lookbehind</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 negative 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>, as an independent, non-capturing group</td></tr>
343 *
344 * </table>
345 *
346 * <hr>
347 *
348 *
349 * <a name="bs">
350 * <h4> Backslashes, escapes, and quoting </h4>
351 *
352 * <p> The backslash character (<tt>'\'</tt>) serves to introduce escaped
353 * constructs, as defined in the table above, as well as to quote characters
354 * that otherwise would be interpreted as unescaped constructs. Thus the
355 * expression <tt>\\</tt> matches a single backslash and <tt>\{</tt> matches a
356 * left brace.
357 *
358 * <p> It is an error to use a backslash prior to any alphabetic character that
359 * does not denote an escaped construct; these are reserved for future
360 * extensions to the regular-expression language. A backslash may be used
361 * prior to a non-alphabetic character regardless of whether that character is
362 * part of an unescaped construct.
363 *
364 * <p> Backslashes within string literals in Java source code are interpreted
365 * as required by the <a
366 * href="http://java.sun.com/docs/books/jls">Java Language
367 * Specification</a> as either <a
368 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">Unicode
369 * escapes</a> or other <a
370 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#101089">character
371 * escapes</a>. It is therefore necessary to double backslashes in string
372 * literals that represent regular expressions to protect them from
373 * interpretation by the Java bytecode compiler. The string literal
374 * <tt>"\b"</tt>, for example, matches a single backspace character when
375 * interpreted as a regular expression, while <tt>"\\b"</tt> matches a
376 * word boundary. The string literal <tt>"\(hello\)"</tt> is illegal
377 * and leads to a compile-time error; in order to match the string
378 * <tt>(hello)</tt> the string literal <tt>"\\(hello\\)"</tt>
379 * must be used.
380 *
381 * <a name="cc">
382 * <h4> Character Classes </h4>
383 *
384 * <p> Character classes may appear within other character classes, and
385 * may be composed by the union operator (implicit) and the intersection
386 * operator (<tt>&&</tt>).
387 * The union operator denotes a class that contains every character that is
388 * in at least one of its operand classes. The intersection operator
389 * denotes a class that contains every character that is in both of its
390 * operand classes.
391 *
392 * <p> The precedence of character-class operators is as follows, from
393 * highest to lowest:
394 *
395 * <blockquote><table border="0" cellpadding="1" cellspacing="0"
396 * summary="Precedence of character class operators.">
397 * <tr><th>1 </th>
398 * <td>Literal escape </td>
399 * <td><tt>\x</tt></td></tr>
400 * <tr><th>2 </th>
401 * <td>Grouping</td>
402 * <td><tt>[...]</tt></td></tr>
403 * <tr><th>3 </th>
404 * <td>Range</td>
405 * <td><tt>a-z</tt></td></tr>
406 * <tr><th>4 </th>
407 * <td>Union</td>
408 * <td><tt>[a-e][i-u]</tt></td></tr>
409 * <tr><th>5 </th>
410 * <td>Intersection</td>
411 * <td><tt>[a-z&&[aeiou]]</tt></td></tr>
412 * </table></blockquote>
413 *
414 * <p> Note that a different set of metacharacters are in effect inside
415 * a character class than outside a character class. For instance, the
416 * regular expression <tt>.</tt> loses its special meaning inside a
417 * character class, while the expression <tt>-</tt> becomes a range
418 * forming metacharacter.
419 *
420 * <a name="lt">
421 * <h4> Line terminators </h4>
422 *
423 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks
424 * the end of a line of the input character sequence. The following are
425 * recognized as line terminators:
426 *
427 * <ul>
428 *
429 * <li> A newline (line feed) character (<tt>'\n'</tt>),
430 *
431 * <li> A carriage-return character followed immediately by a newline
432 * character (<tt>"\r\n"</tt>),
433 *
434 * <li> A standalone carriage-return character (<tt>'\r'</tt>),
435 *
436 * <li> A next-line character (<tt>'\u0085'</tt>),
437 *
438 * <li> A line-separator character (<tt>'\u2028'</tt>), or
439 *
440 * <li> A paragraph-separator character (<tt>'\u2029</tt>).
441 *
442 * </ul>
443 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators
444 * recognized are newline characters.
445 *
446 * <p> The regular expression <tt>.</tt> matches any character except a line
447 * terminator unless the {@link #DOTALL} flag is specified.
448 *
449 * <p> By default, the regular expressions <tt>^</tt> and <tt>$</tt> ignore
450 * line terminators and only match at the beginning and the end, respectively,
451 * of the entire input sequence. If {@link #MULTILINE} mode is activated then
452 * <tt>^</tt> matches at the beginning of input and after any line terminator
453 * except at the end of input. When in {@link #MULTILINE} mode <tt>$</tt>
454 * matches just before a line terminator or the end of the input sequence.
455 *
456 * <a name="cg">
457 * <h4> Groups and capturing </h4>
458 *
459 * <a name="gnumber">
460 * <h5> Group number </h5>
461 * <p> Capturing groups are numbered by counting their opening parentheses from
462 * left to right. In the expression <tt>((A)(B(C)))</tt>, for example, there
463 * are four such groups: </p>
464 *
465 * <blockquote><table cellpadding=1 cellspacing=0 summary="Capturing group numberings">
466 * <tr><th>1 </th>
467 * <td><tt>((A)(B(C)))</tt></td></tr>
468 * <tr><th>2 </th>
469 * <td><tt>(A)</tt></td></tr>
470 * <tr><th>3 </th>
471 * <td><tt>(B(C))</tt></td></tr>
472 * <tr><th>4 </th>
473 * <td><tt>(C)</tt></td></tr>
474 * </table></blockquote>
475 *
476 * <p> Group zero always stands for the entire expression.
477 *
478 * <p> Capturing groups are so named because, during a match, each subsequence
479 * of the input sequence that matches such a group is saved. The captured
480 * subsequence may be used later in the expression, via a back reference, and
481 * may also be retrieved from the matcher once the match operation is complete.
482 *
483 * <a name="groupname">
484 * <h5> Group name </h5>
485 * <p>A capturing group can also be assigned a "name", a <tt>named-capturing group</tt>,
486 * and then be back-referenced later by the "name". Group names are composed of
487 * the following characters. The first character must be a <tt>letter</tt>.
488 *
489 * <ul>
490 * <li> The uppercase letters <tt>'A'</tt> through <tt>'Z'</tt>
491 * (<tt>'\u0041'</tt> through <tt>'\u005a'</tt>),
492 * <li> The lowercase letters <tt>'a'</tt> through <tt>'z'</tt>
493 * (<tt>'\u0061'</tt> through <tt>'\u007a'</tt>),
494 * <li> The digits <tt>'0'</tt> through <tt>'9'</tt>
495 * (<tt>'\u0030'</tt> through <tt>'\u0039'</tt>),
496 * </ul>
497 *
498 * <p> A <tt>named-capturing group</tt> is still numbered as described in
499 * <a href="#gnumber">Group number</a>.
500 *
501 * <p> The captured input associated with a group is always the subsequence
502 * that the group most recently matched. If a group is evaluated a second time
503 * because of quantification then its previously-captured value, if any, will
504 * be retained if the second evaluation fails. Matching the string
505 * <tt>"aba"</tt> against the expression <tt>(a(b)?)+</tt>, for example, leaves
506 * group two set to <tt>"b"</tt>. All captured input is discarded at the
507 * beginning of each match.
508 *
509 * <p> Groups beginning with <tt>(?</tt> are either pure, <i>non-capturing</i> groups
510 * that do not capture text and do not count towards the group total, or
511 * <i>named-capturing</i> group.
512 *
513 * <h4> Unicode support </h4>
514 *
515 * <p> This class is in conformance with Level 1 of <a
516 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical
517 * Standard #18: Unicode Regular Expression Guidelines</i></a>, plus RL2.1
518 * Canonical Equivalents.
519 *
520 * <p> Unicode escape sequences such as <tt>\u2014</tt> in Java source code
521 * are processed as described in <a
522 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">\u00A73.3</a>
523 * of the Java Language Specification. Such escape sequences are also
524 * implemented directly by the regular-expression parser so that Unicode
525 * escapes can be used in expressions that are read from files or from the
526 * keyboard. Thus the strings <tt>"\u2014"</tt> and <tt>"\\u2014"</tt>,
527 * while not equal, compile into the same pattern, which matches the character
528 * with hexadecimal value <tt>0x2014</tt>.
529 *
530 * <a name="ubc"> <p>Unicode blocks and categories are written with the
531 * <tt>\p</tt> and <tt>\P</tt> constructs as in
532 * Perl. <tt>\p{</tt><i>prop</i><tt>}</tt> matches if the input has the
533 * property <i>prop</i>, while <tt>\P{</tt><i>prop</i><tt>}</tt> does not match if
534 * the input has that property. Blocks are specified with the prefix
535 * <tt>In</tt>, as in <tt>InMongolian</tt>. Categories may be specified with
536 * the optional prefix <tt>Is</tt>: Both <tt>\p{L}</tt> and <tt>\p{IsL}</tt>
537 * denote the category of Unicode letters. Blocks and categories can be used
538 * both inside and outside of a character class.
539 *
540 * <p> The supported categories are those of
541 * <a href="http://www.unicode.org/unicode/standard/standard.html">
542 * <i>The Unicode Standard</i></a> in the version specified by the
543 * {@link java.lang.Character Character} class. The category names are those
544 * defined in the Standard, both normative and informative.
545 * The block names supported by <code>Pattern</code> are the valid block names
546 * accepted and defined by
547 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}.
548 *
549 * <a name="jcc"> <p>Categories that behave like the java.lang.Character
550 * boolean is<i>methodname</i> methods (except for the deprecated ones) are
551 * available through the same <tt>\p{</tt><i>prop</i><tt>}</tt> syntax where
552 * the specified property has the name <tt>java<i>methodname</i></tt>.
553 *
554 * <h4> Comparison to Perl 5 </h4>
555 *
556 * <p>The <code>Pattern</code> engine performs traditional NFA-based matching
557 * with ordered alternation as occurs in Perl 5.
558 *
559 * <p> Perl constructs not supported by this class: </p>
560 *
561 * <ul>
562 *
563 * <li><p> The conditional constructs <tt>(?{</tt><i>X</i><tt>})</tt> and
564 * <tt>(?(</tt><i>condition</i><tt>)</tt><i>X</i><tt>|</tt><i>Y</i><tt>)</tt>,
565 * </p></li>
566 *
567 * <li><p> The embedded code constructs <tt>(?{</tt><i>code</i><tt>})</tt>
568 * and <tt>(??{</tt><i>code</i><tt>})</tt>,</p></li>
569 *
570 * <li><p> The embedded comment syntax <tt>(?#comment)</tt>, and </p></li>
571 *
572 * <li><p> The preprocessing operations <tt>\l</tt> <tt>\u</tt>,
573 * <tt>\L</tt>, and <tt>\U</tt>. </p></li>
574 *
575 * </ul>
576 *
577 * <p> Constructs supported by this class but not by Perl: </p>
578 *
579 * <ul>
580 *
581 * <li><p> Possessive quantifiers, which greedily match as much as they can
582 * and do not back off, even when doing so would allow the overall match to
583 * succeed. </p></li>
584 *
585 * <li><p> Character-class union and intersection as described
586 * <a href="#cc">above</a>.</p></li>
587 *
588 * </ul>
589 *
590 * <p> Notable differences from Perl: </p>
591 *
592 * <ul>
593 *
594 * <li><p> In Perl, <tt>\1</tt> through <tt>\9</tt> are always interpreted
595 * as back references; a backslash-escaped number greater than <tt>9</tt> is
596 * treated as a back reference if at least that many subexpressions exist,
597 * otherwise it is interpreted, if possible, as an octal escape. In this
598 * class octal escapes must always begin with a zero. In this class,
599 * <tt>\1</tt> through <tt>\9</tt> are always interpreted as back
600 * references, and a larger number is accepted as a back reference if at
601 * least that many subexpressions exist at that point in the regular
602 * expression, otherwise the parser will drop digits until the number is
603 * smaller or equal to the existing number of groups or it is one digit.
604 * </p></li>
605 *
606 * <li><p> Perl uses the <tt>g</tt> flag to request a match that resumes
607 * where the last match left off. This functionality is provided implicitly
608 * by the {@link Matcher} class: Repeated invocations of the {@link
609 * Matcher#find find} method will resume where the last match left off,
610 * unless the matcher is reset. </p></li>
611 *
612 * <li><p> In Perl, embedded flags at the top level of an expression affect
613 * the whole expression. In this class, embedded flags always take effect
614 * at the point at which they appear, whether they are at the top level or
615 * within a group; in the latter case, flags are restored at the end of the
616 * group just as in Perl. </p></li>
617 *
618 * <li><p> Perl is forgiving about malformed matching constructs, as in the
619 * expression <tt>*a</tt>, as well as dangling brackets, as in the
620 * expression <tt>abc]</tt>, and treats them as literals. This
621 * class also accepts dangling brackets but is strict about dangling
622 * metacharacters like +, ? and *, and will throw a
623 * {@link PatternSyntaxException} if it encounters them. </p></li>
624 *
625 * </ul>
626 *
627 *
628 * <p> For a more precise description of the behavior of regular expression
629 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/">
630 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl,
631 * O'Reilly and Associates, 2006.</a>
632 * </p>
633 *
634 * @see java.lang.String#split(String, int)
635 * @see java.lang.String#split(String)
636 *
637 * @author Mike McCloskey
638 * @author Mark Reinhold
639 * @author JSR-51 Expert Group
640 * @since 1.4
641 * @spec JSR-51
642 */
643
644 public final class Pattern
645 implements java.io.Serializable
646 {
647
648 /**
649 * Regular expression modifier values. Instead of being passed as
650 * arguments, they can also be passed as inline modifiers.
651 * For example, the following statements have the same effect.
652 * <pre>
653 * RegExp r1 = RegExp.compile("abc", Pattern.I|Pattern.M);
654 * RegExp r2 = RegExp.compile("(?im)abc", 0);
655 * </pre>
656 *
657 * The flags are duplicated so that the familiar Perl match flag
658 * names are available.
659 */
660
661 /**
662 * Enables Unix lines mode.
663 *
664 * <p> In this mode, only the <tt>'\n'</tt> line terminator is recognized
665 * in the behavior of <tt>.</tt>, <tt>^</tt>, and <tt>$</tt>.
666 *
667 * <p> Unix lines mode can also be enabled via the embedded flag
668 * expression <tt>(?d)</tt>.
669 */
670 public static final int UNIX_LINES = 0x01;
671
672 /**
673 * Enables case-insensitive matching.
674 *
675 * <p> By default, case-insensitive matching assumes that only characters
676 * in the US-ASCII charset are being matched. Unicode-aware
677 * case-insensitive matching can be enabled by specifying the {@link
678 * #UNICODE_CASE} flag in conjunction with this flag.
679 *
680 * <p> Case-insensitive matching can also be enabled via the embedded flag
681 * expression <tt>(?i)</tt>.
682 *
683 * <p> Specifying this flag may impose a slight performance penalty. </p>
684 */
685 public static final int CASE_INSENSITIVE = 0x02;
686
687 /**
688 * Permits whitespace and comments in pattern.
689 *
690 * <p> In this mode, whitespace is ignored, and embedded comments starting
691 * with <tt>#</tt> are ignored until the end of a line.
692 *
693 * <p> Comments mode can also be enabled via the embedded flag
694 * expression <tt>(?x)</tt>.
695 */
696 public static final int COMMENTS = 0x04;
697
698 /**
699 * Enables multiline mode.
700 *
701 * <p> In multiline mode the expressions <tt>^</tt> and <tt>$</tt> match
702 * just after or just before, respectively, a line terminator or the end of
703 * the input sequence. By default these expressions only match at the
704 * beginning and the end of the entire input sequence.
705 *
706 * <p> Multiline mode can also be enabled via the embedded flag
707 * expression <tt>(?m)</tt>. </p>
708 */
709 public static final int MULTILINE = 0x08;
710
711 /**
712 * Enables literal parsing of the pattern.
713 *
714 * <p> When this flag is specified then the input string that specifies
715 * the pattern is treated as a sequence of literal characters.
716 * Metacharacters or escape sequences in the input sequence will be
717 * given no special meaning.
718 *
719 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on
720 * matching when used in conjunction with this flag. The other flags
721 * become superfluous.
722 *
723 * <p> There is no embedded flag character for enabling literal parsing.
724 * @since 1.5
725 */
726 public static final int LITERAL = 0x10;
727
728 /**
729 * Enables dotall mode.
730 *
731 * <p> In dotall mode, the expression <tt>.</tt> matches any character,
732 * including a line terminator. By default this expression does not match
733 * line terminators.
734 *
735 * <p> Dotall mode can also be enabled via the embedded flag
736 * expression <tt>(?s)</tt>. (The <tt>s</tt> is a mnemonic for
737 * "single-line" mode, which is what this is called in Perl.) </p>
738 */
739 public static final int DOTALL = 0x20;
740
741 /**
742 * Enables Unicode-aware case folding.
743 *
744 * <p> When this flag is specified then case-insensitive matching, when
745 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner
746 * consistent with the Unicode Standard. By default, case-insensitive
747 * matching assumes that only characters in the US-ASCII charset are being
748 * matched.
749 *
750 * <p> Unicode-aware case folding can also be enabled via the embedded flag
751 * expression <tt>(?u)</tt>.
752 *
753 * <p> Specifying this flag may impose a performance penalty. </p>
754 */
755 public static final int UNICODE_CASE = 0x40;
756
757 /**
758 * Enables canonical equivalence.
759 *
760 * <p> When this flag is specified then two characters will be considered
761 * to match if, and only if, their full canonical decompositions match.
762 * The expression <tt>"a\u030A"</tt>, for example, will match the
763 * string <tt>"\u00E5"</tt> when this flag is specified. By default,
764 * matching does not take canonical equivalence into account.
765 *
766 * <p> There is no embedded flag character for enabling canonical
767 * equivalence.
768 *
769 * <p> Specifying this flag may impose a performance penalty. </p>
770 */
771 public static final int CANON_EQ = 0x80;
772
773 /* Pattern has only two serialized components: The pattern string
774 * and the flags, which are all that is needed to recompile the pattern
775 * when it is deserialized.
776 */
777
778 /** use serialVersionUID from Merlin b59 for interoperability */
779 private static final long serialVersionUID = 5073258162644648461L;
780
781 /**
782 * The original regular-expression pattern string.
783 *
784 * @serial
785 */
786 private String pattern;
787
788 /**
789 * The original pattern flags.
790 *
791 * @serial
792 */
793 private int flags;
794
795 /**
796 * Boolean indicating this Pattern is compiled; this is necessary in order
797 * to lazily compile deserialized Patterns.
798 */
799 private transient volatile boolean compiled = false;
800
801 /**
802 * The normalized pattern string.
803 */
804 private transient String normalizedPattern;
805
806 /**
807 * The starting point of state machine for the find operation. This allows
808 * a match to start anywhere in the input.
809 */
810 transient Node root;
811
812 /**
813 * The root of object tree for a match operation. The pattern is matched
814 * at the beginning. This may include a find that uses BnM or a First
815 * node.
816 */
817 transient Node matchRoot;
818
819 /**
820 * Temporary storage used by parsing pattern slice.
821 */
822 transient int[] buffer;
823
824 /**
825 * Map the "name" of the "named capturing group" to its group id
826 * node.
827 */
828 transient volatile Map<String, Integer> namedGroups;
829
830 /**
831 * Temporary storage used while parsing group references.
832 */
833 transient GroupHead[] groupNodes;
834
835 /**
836 * Temporary null terminated code point array used by pattern compiling.
837 */
838 private transient int[] temp;
839
840 /**
841 * The number of capturing groups in this Pattern. Used by matchers to
842 * allocate storage needed to perform a match.
843 */
844 transient int capturingGroupCount;
845
846 /**
847 * The local variable count used by parsing tree. Used by matchers to
848 * allocate storage needed to perform a match.
849 */
850 transient int localCount;
851
852 /**
853 * Index into the pattern string that keeps track of how much has been
854 * parsed.
855 */
856 private transient int cursor;
857
858 /**
859 * Holds the length of the pattern string.
860 */
861 private transient int patternLength;
862
863 /**
864 * If the Start node might possibly match supplementary characters.
865 * It is set to true during compiling if
866 * (1) There is supplementary char in pattern, or
867 * (2) There is complement node of Category or Block
868 */
869 private transient boolean hasSupplementary;
870
871 /**
872 * Compiles the given regular expression into a pattern. </p>
873 *
874 * @param regex
875 * The expression to be compiled
876 *
877 * @throws PatternSyntaxException
878 * If the expression's syntax is invalid
879 */
880 public static Pattern compile(String regex) {
881 return new Pattern(regex, 0);
882 }
883
884 /**
885 * Compiles the given regular expression into a pattern with the given
886 * flags. </p>
887 *
888 * @param regex
889 * The expression to be compiled
890 *
891 * @param flags
892 * Match flags, a bit mask that may include
893 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL},
894 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES},
895 * {@link #LITERAL} and {@link #COMMENTS}
896 *
897 * @throws IllegalArgumentException
898 * If bit values other than those corresponding to the defined
899 * match flags are set in <tt>flags</tt>
900 *
901 * @throws PatternSyntaxException
902 * If the expression's syntax is invalid
903 */
904 public static Pattern compile(String regex, int flags) {
905 return new Pattern(regex, flags);
906 }
907
908 /**
909 * Returns the regular expression from which this pattern was compiled.
910 * </p>
911 *
912 * @return The source of this pattern
913 */
914 public String pattern() {
915 return pattern;
916 }
917
918 /**
919 * <p>Returns the string representation of this pattern. This
920 * is the regular expression from which this pattern was
921 * compiled.</p>
922 *
923 * @return The string representation of this pattern
924 * @since 1.5
925 */
926 public String toString() {
927 return pattern;
928 }
929
930 /**
931 * Creates a matcher that will match the given input against this pattern.
932 * </p>
933 *
934 * @param input
935 * The character sequence to be matched
936 *
937 * @return A new matcher for this pattern
938 */
939 public Matcher matcher(CharSequence input) {
940 if (!compiled) {
941 synchronized(this) {
942 if (!compiled)
943 compile();
944 }
945 }
946 Matcher m = new Matcher(this, input);
947 return m;
948 }
949
950 /**
951 * Returns this pattern's match flags. </p>
952 *
953 * @return The match flags specified when this pattern was compiled
954 */
955 public int flags() {
956 return flags;
957 }
958
959 /**
960 * Compiles the given regular expression and attempts to match the given
961 * input against it.
962 *
963 * <p> An invocation of this convenience method of the form
964 *
965 * <blockquote><pre>
966 * Pattern.matches(regex, input);</pre></blockquote>
967 *
968 * behaves in exactly the same way as the expression
969 *
970 * <blockquote><pre>
971 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote>
972 *
973 * <p> If a pattern is to be used multiple times, compiling it once and reusing
974 * it will be more efficient than invoking this method each time. </p>
975 *
976 * @param regex
977 * The expression to be compiled
978 *
979 * @param input
980 * The character sequence to be matched
981 *
982 * @throws PatternSyntaxException
983 * If the expression's syntax is invalid
984 */
985 public static boolean matches(String regex, CharSequence input) {
986 Pattern p = Pattern.compile(regex);
987 Matcher m = p.matcher(input);
988 return m.matches();
989 }
990
991 /**
992 * Splits the given input sequence around matches of this pattern.
993 *
994 * <p> The array returned by this method contains each substring of the
995 * input sequence that is terminated by another subsequence that matches
996 * this pattern or is terminated by the end of the input sequence. The
997 * substrings in the array are in the order in which they occur in the
998 * input. If this pattern does not match any subsequence of the input then
999 * the resulting array has just one element, namely the input sequence in
1000 * string form.
1001 *
1002 * <p> The <tt>limit</tt> parameter controls the number of times the
1003 * pattern is applied and therefore affects the length of the resulting
1004 * array. If the limit <i>n</i> is greater than zero then the pattern
1005 * will be applied at most <i>n</i> - 1 times, the array's
1006 * length will be no greater than <i>n</i>, and the array's last entry
1007 * will contain all input beyond the last matched delimiter. If <i>n</i>
1008 * is non-positive then the pattern will be applied as many times as
1009 * possible and the array can have any length. If <i>n</i> is zero then
1010 * the pattern will be applied as many times as possible, the array can
1011 * have any length, and trailing empty strings will be discarded.
1012 *
1013 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following
1014 * results with these parameters:
1015 *
1016 * <blockquote><table cellpadding=1 cellspacing=0
1017 * summary="Split examples showing regex, limit, and result">
1018 * <tr><th><P align="left"><i>Regex </i></th>
1019 * <th><P align="left"><i>Limit </i></th>
1020 * <th><P align="left"><i>Result </i></th></tr>
1021 * <tr><td align=center>:</td>
1022 * <td align=center>2</td>
1023 * <td><tt>{ "boo", "and:foo" }</tt></td></tr>
1024 * <tr><td align=center>:</td>
1025 * <td align=center>5</td>
1026 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
1027 * <tr><td align=center>:</td>
1028 * <td align=center>-2</td>
1029 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
1030 * <tr><td align=center>o</td>
1031 * <td align=center>5</td>
1032 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr>
1033 * <tr><td align=center>o</td>
1034 * <td align=center>-2</td>
1035 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr>
1036 * <tr><td align=center>o</td>
1037 * <td align=center>0</td>
1038 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr>
1039 * </table></blockquote>
1040 *
1041 *
1042 * @param input
1043 * The character sequence to be split
1044 *
1045 * @param limit
1046 * The result threshold, as described above
1047 *
1048 * @return The array of strings computed by splitting the input
1049 * around matches of this pattern
1050 */
1051 public String[] split(CharSequence input, int limit) {
1052 int index = 0;
1053 boolean matchLimited = limit > 0;
1054 ArrayList<String> matchList = new ArrayList<String>();
1055 Matcher m = matcher(input);
1056
1057 // Add segments before each match found
1058 while(m.find()) {
1059 if (!matchLimited || matchList.size() < limit - 1) {
1060 String match = input.subSequence(index, m.start()).toString();
1061 matchList.add(match);
1062 index = m.end();
1063 } else if (matchList.size() == limit - 1) { // last one
1064 String match = input.subSequence(index,
1065 input.length()).toString();
1066 matchList.add(match);
1067 index = m.end();
1068 }
1069 }
1070
1071 // If no match was found, return this
1072 if (index == 0)
1073 return new String[] {input.toString()};
1074
1075 // Add remaining segment
1076 if (!matchLimited || matchList.size() < limit)
1077 matchList.add(input.subSequence(index, input.length()).toString());
1078
1079 // Construct result
1080 int resultSize = matchList.size();
1081 if (limit == 0)
1082 while (resultSize > 0 && matchList.get(resultSize-1).equals(""))
1083 resultSize--;
1084 String[] result = new String[resultSize];
1085 return matchList.subList(0, resultSize).toArray(result);
1086 }
1087
1088 /**
1089 * Splits the given input sequence around matches of this pattern.
1090 *
1091 * <p> This method works as if by invoking the two-argument {@link
1092 * #split(java.lang.CharSequence, int) split} method with the given input
1093 * sequence and a limit argument of zero. Trailing empty strings are
1094 * therefore not included in the resulting array. </p>
1095 *
1096 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following
1097 * results with these expressions:
1098 *
1099 * <blockquote><table cellpadding=1 cellspacing=0
1100 * summary="Split examples showing regex and result">
1101 * <tr><th><P align="left"><i>Regex </i></th>
1102 * <th><P align="left"><i>Result</i></th></tr>
1103 * <tr><td align=center>:</td>
1104 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
1105 * <tr><td align=center>o</td>
1106 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr>
1107 * </table></blockquote>
1108 *
1109 *
1110 * @param input
1111 * The character sequence to be split
1112 *
1113 * @return The array of strings computed by splitting the input
1114 * around matches of this pattern
1115 */
1116 public String[] split(CharSequence input) {
1117 return split(input, 0);
1118 }
1119
1120 /**
1121 * Returns a literal pattern <code>String</code> for the specified
1122 * <code>String</code>.
1123 *
1124 * <p>This method produces a <code>String</code> that can be used to
1125 * create a <code>Pattern</code> that would match the string
1126 * <code>s</code> as if it were a literal pattern.</p> Metacharacters
1127 * or escape sequences in the input sequence will be given no special
1128 * meaning.
1129 *
1130 * @param s The string to be literalized
1131 * @return A literal string replacement
1132 * @since 1.5
1133 */
1134 public static String quote(String s) {
1135 int slashEIndex = s.indexOf("\\E");
1136 if (slashEIndex == -1)
1137 return "\\Q" + s + "\\E";
1138
1139 StringBuilder sb = new StringBuilder(s.length() * 2);
1140 sb.append("\\Q");
1141 slashEIndex = 0;
1142 int current = 0;
1143 while ((slashEIndex = s.indexOf("\\E", current)) != -1) {
1144 sb.append(s.substring(current, slashEIndex));
1145 current = slashEIndex + 2;
1146 sb.append("\\E\\\\E\\Q");
1147 }
1148 sb.append(s.substring(current, s.length()));
1149 sb.append("\\E");
1150 return sb.toString();
1151 }
1152
1153 /**
1154 * Recompile the Pattern instance from a stream. The original pattern
1155 * string is read in and the object tree is recompiled from it.
1156 */
1157 private void readObject(java.io.ObjectInputStream s)
1158 throws java.io.IOException, ClassNotFoundException {
1159
1160 // Read in all fields
1161 s.defaultReadObject();
1162
1163 // Initialize counts
1164 capturingGroupCount = 1;
1165 localCount = 0;
1166
1167 // if length > 0, the Pattern is lazily compiled
1168 compiled = false;
1169 if (pattern.length() == 0) {
1170 root = new Start(lastAccept);
1171 matchRoot = lastAccept;
1172 compiled = true;
1173 }
1174 }
1175
1176 /**
1177 * This private constructor is used to create all Patterns. The pattern
1178 * string and match flags are all that is needed to completely describe
1179 * a Pattern. An empty pattern string results in an object tree with
1180 * only a Start node and a LastNode node.
1181 */
1182 private Pattern(String p, int f) {
1183 pattern = p;
1184 flags = f;
1185
1186 // Reset group index count
1187 capturingGroupCount = 1;
1188 localCount = 0;
1189
1190 if (pattern.length() > 0) {
1191 compile();
1192 } else {
1193 root = new Start(lastAccept);
1194 matchRoot = lastAccept;
1195 }
1196 }
1197
1198 /**
1199 * The pattern is converted to normalizedD form and then a pure group
1200 * is constructed to match canonical equivalences of the characters.
1201 */
1202 private void normalize() {
1203 boolean inCharClass = false;
1204 int lastCodePoint = -1;
1205
1206 // Convert pattern into normalizedD form
1207 normalizedPattern = Normalizer.normalize(pattern, Normalizer.Form.NFD);
1208 patternLength = normalizedPattern.length();
1209
1210 // Modify pattern to match canonical equivalences
1211 StringBuilder newPattern = new StringBuilder(patternLength);
1212 for(int i=0; i<patternLength; ) {
1213 int c = normalizedPattern.codePointAt(i);
1214 StringBuilder sequenceBuffer;
1215 if ((Character.getType(c) == Character.NON_SPACING_MARK)
1216 && (lastCodePoint != -1)) {
1217 sequenceBuffer = new StringBuilder();
1218 sequenceBuffer.appendCodePoint(lastCodePoint);
1219 sequenceBuffer.appendCodePoint(c);
1220 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1221 i += Character.charCount(c);
1222 if (i >= patternLength)
1223 break;
1224 c = normalizedPattern.codePointAt(i);
1225 sequenceBuffer.appendCodePoint(c);
1226 }
1227 String ea = produceEquivalentAlternation(
1228 sequenceBuffer.toString());
1229 newPattern.setLength(newPattern.length()-Character.charCount(lastCodePoint));
1230 newPattern.append("(?:").append(ea).append(")");
1231 } else if (c == '[' && lastCodePoint != '\\') {
1232 i = normalizeCharClass(newPattern, i);
1233 } else {
1234 newPattern.appendCodePoint(c);
1235 }
1236 lastCodePoint = c;
1237 i += Character.charCount(c);
1238 }
1239 normalizedPattern = newPattern.toString();
1240 }
1241
1242 /**
1243 * Complete the character class being parsed and add a set
1244 * of alternations to it that will match the canonical equivalences
1245 * of the characters within the class.
1246 */
1247 private int normalizeCharClass(StringBuilder newPattern, int i) {
1248 StringBuilder charClass = new StringBuilder();
1249 StringBuilder eq = null;
1250 int lastCodePoint = -1;
1251 String result;
1252
1253 i++;
1254 charClass.append("[");
1255 while(true) {
1256 int c = normalizedPattern.codePointAt(i);
1257 StringBuilder sequenceBuffer;
1258
1259 if (c == ']' && lastCodePoint != '\\') {
1260 charClass.append((char)c);
1261 break;
1262 } else if (Character.getType(c) == Character.NON_SPACING_MARK) {
1263 sequenceBuffer = new StringBuilder();
1264 sequenceBuffer.appendCodePoint(lastCodePoint);
1265 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1266 sequenceBuffer.appendCodePoint(c);
1267 i += Character.charCount(c);
1268 if (i >= normalizedPattern.length())
1269 break;
1270 c = normalizedPattern.codePointAt(i);
1271 }
1272 String ea = produceEquivalentAlternation(
1273 sequenceBuffer.toString());
1274
1275 charClass.setLength(charClass.length()-Character.charCount(lastCodePoint));
1276 if (eq == null)
1277 eq = new StringBuilder();
1278 eq.append('|');
1279 eq.append(ea);
1280 } else {
1281 charClass.appendCodePoint(c);
1282 i++;
1283 }
1284 if (i == normalizedPattern.length())
1285 throw error("Unclosed character class");
1286 lastCodePoint = c;
1287 }
1288
1289 if (eq != null) {
1290 result = "(?:"+charClass.toString()+eq.toString()+")";
1291 } else {
1292 result = charClass.toString();
1293 }
1294
1295 newPattern.append(result);
1296 return i;
1297 }
1298
1299 /**
1300 * Given a specific sequence composed of a regular character and
1301 * combining marks that follow it, produce the alternation that will
1302 * match all canonical equivalences of that sequence.
1303 */
1304 private String produceEquivalentAlternation(String source) {
1305 int len = countChars(source, 0, 1);
1306 if (source.length() == len)
1307 // source has one character.
1308 return source;
1309
1310 String base = source.substring(0,len);
1311 String combiningMarks = source.substring(len);
1312
1313 String[] perms = producePermutations(combiningMarks);
1314 StringBuilder result = new StringBuilder(source);
1315
1316 // Add combined permutations
1317 for(int x=0; x<perms.length; x++) {
1318 String next = base + perms[x];
1319 if (x>0)
1320 result.append("|"+next);
1321 next = composeOneStep(next);
1322 if (next != null)
1323 result.append("|"+produceEquivalentAlternation(next));
1324 }
1325 return result.toString();
1326 }
1327
1328 /**
1329 * Returns an array of strings that have all the possible
1330 * permutations of the characters in the input string.
1331 * This is used to get a list of all possible orderings
1332 * of a set of combining marks. Note that some of the permutations
1333 * are invalid because of combining class collisions, and these
1334 * possibilities must be removed because they are not canonically
1335 * equivalent.
1336 */
1337 private String[] producePermutations(String input) {
1338 if (input.length() == countChars(input, 0, 1))
1339 return new String[] {input};
1340
1341 if (input.length() == countChars(input, 0, 2)) {
1342 int c0 = Character.codePointAt(input, 0);
1343 int c1 = Character.codePointAt(input, Character.charCount(c0));
1344 if (getClass(c1) == getClass(c0)) {
1345 return new String[] {input};
1346 }
1347 String[] result = new String[2];
1348 result[0] = input;
1349 StringBuilder sb = new StringBuilder(2);
1350 sb.appendCodePoint(c1);
1351 sb.appendCodePoint(c0);
1352 result[1] = sb.toString();
1353 return result;
1354 }
1355
1356 int length = 1;
1357 int nCodePoints = countCodePoints(input);
1358 for(int x=1; x<nCodePoints; x++)
1359 length = length * (x+1);
1360
1361 String[] temp = new String[length];
1362
1363 int combClass[] = new int[nCodePoints];
1364 for(int x=0, i=0; x<nCodePoints; x++) {
1365 int c = Character.codePointAt(input, i);
1366 combClass[x] = getClass(c);
1367 i += Character.charCount(c);
1368 }
1369
1370 // For each char, take it out and add the permutations
1371 // of the remaining chars
1372 int index = 0;
1373 int len;
1374 // offset maintains the index in code units.
1375 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) {
1376 len = countChars(input, offset, 1);
1377 boolean skip = false;
1378 for(int y=x-1; y>=0; y--) {
1379 if (combClass[y] == combClass[x]) {
1380 continue loop;
1381 }
1382 }
1383 StringBuilder sb = new StringBuilder(input);
1384 String otherChars = sb.delete(offset, offset+len).toString();
1385 String[] subResult = producePermutations(otherChars);
1386
1387 String prefix = input.substring(offset, offset+len);
1388 for(int y=0; y<subResult.length; y++)
1389 temp[index++] = prefix + subResult[y];
1390 }
1391 String[] result = new String[index];
1392 for (int x=0; x<index; x++)
1393 result[x] = temp[x];
1394 return result;
1395 }
1396
1397 private int getClass(int c) {
1398 return sun.text.Normalizer.getCombiningClass(c);
1399 }
1400
1401 /**
1402 * Attempts to compose input by combining the first character
1403 * with the first combining mark following it. Returns a String
1404 * that is the composition of the leading character with its first
1405 * combining mark followed by the remaining combining marks. Returns
1406 * null if the first two characters cannot be further composed.
1407 */
1408 private String composeOneStep(String input) {
1409 int len = countChars(input, 0, 2);
1410 String firstTwoCharacters = input.substring(0, len);
1411 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC);
1412
1413 if (result.equals(firstTwoCharacters))
1414 return null;
1415 else {
1416 String remainder = input.substring(len);
1417 return result + remainder;
1418 }
1419 }
1420
1421 /**
1422 * Preprocess any \Q...\E sequences in `temp', meta-quoting them.
1423 * See the description of `quotemeta' in perlfunc(1).
1424 */
1425 private void RemoveQEQuoting() {
1426 final int pLen = patternLength;
1427 int i = 0;
1428 while (i < pLen-1) {
1429 if (temp[i] != '\\')
1430 i += 1;
1431 else if (temp[i + 1] != 'Q')
1432 i += 2;
1433 else
1434 break;
1435 }
1436 if (i >= pLen - 1) // No \Q sequence found
1437 return;
1438 int j = i;
1439 i += 2;
1440 int[] newtemp = new int[j + 2*(pLen-i) + 2];
1441 System.arraycopy(temp, 0, newtemp, 0, j);
1442
1443 boolean inQuote = true;
1444 while (i < pLen) {
1445 int c = temp[i++];
1446 if (! ASCII.isAscii(c) || ASCII.isAlnum(c)) {
1447 newtemp[j++] = c;
1448 } else if (c != '\\') {
1449 if (inQuote) newtemp[j++] = '\\';
1450 newtemp[j++] = c;
1451 } else if (inQuote) {
1452 if (temp[i] == 'E') {
1453 i++;
1454 inQuote = false;
1455 } else {
1456 newtemp[j++] = '\\';
1457 newtemp[j++] = '\\';
1458 }
1459 } else {
1460 if (temp[i] == 'Q') {
1461 i++;
1462 inQuote = true;
1463 } else {
1464 newtemp[j++] = c;
1465 if (i != pLen)
1466 newtemp[j++] = temp[i++];
1467 }
1468 }
1469 }
1470
1471 patternLength = j;
1472 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination
1473 }
1474
1475 /**
1476 * Copies regular expression to an int array and invokes the parsing
1477 * of the expression which will create the object tree.
1478 */
1479 private void compile() {
1480 // Handle canonical equivalences
1481 if (has(CANON_EQ) && !has(LITERAL)) {
1482 normalize();
1483 } else {
1484 normalizedPattern = pattern;
1485 }
1486 patternLength = normalizedPattern.length();
1487
1488 // Copy pattern to int array for convenience
1489 // Use double zero to terminate pattern
1490 temp = new int[patternLength + 2];
1491
1492 hasSupplementary = false;
1493 int c, count = 0;
1494 // Convert all chars into code points
1495 for (int x = 0; x < patternLength; x += Character.charCount(c)) {
1496 c = normalizedPattern.codePointAt(x);
1497 if (isSupplementary(c)) {
1498 hasSupplementary = true;
1499 }
1500 temp[count++] = c;
1501 }
1502
1503 patternLength = count; // patternLength now in code points
1504
1505 if (! has(LITERAL))
1506 RemoveQEQuoting();
1507
1508 // Allocate all temporary objects here.
1509 buffer = new int[32];
1510 groupNodes = new GroupHead[10];
1511 namedGroups = null;
1512
1513 if (has(LITERAL)) {
1514 // Literal pattern handling
1515 matchRoot = newSlice(temp, patternLength, hasSupplementary);
1516 matchRoot.next = lastAccept;
1517 } else {
1518 // Start recursive descent parsing
1519 matchRoot = expr(lastAccept);
1520 // Check extra pattern characters
1521 if (patternLength != cursor) {
1522 if (peek() == ')') {
1523 throw error("Unmatched closing ')'");
1524 } else {
1525 throw error("Unexpected internal error");
1526 }
1527 }
1528 }
1529
1530 // Peephole optimization
1531 if (matchRoot instanceof Slice) {
1532 root = BnM.optimize(matchRoot);
1533 if (root == matchRoot) {
1534 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1535 }
1536 } else if (matchRoot instanceof Begin || matchRoot instanceof First) {
1537 root = matchRoot;
1538 } else {
1539 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1540 }
1541
1542 // Release temporary storage
1543 temp = null;
1544 buffer = null;
1545 groupNodes = null;
1546 patternLength = 0;
1547 compiled = true;
1548 }
1549
1550 Map<String, Integer> namedGroups() {
1551 if (namedGroups == null)
1552 namedGroups = new HashMap<String, Integer>(2);
1553 return namedGroups;
1554 }
1555
1556 /**
1557 * Used to print out a subtree of the Pattern to help with debugging.
1558 */
1559 private static void printObjectTree(Node node) {
1560 while(node != null) {
1561 if (node instanceof Prolog) {
1562 System.out.println(node);
1563 printObjectTree(((Prolog)node).loop);
1564 System.out.println("**** end contents prolog loop");
1565 } else if (node instanceof Loop) {
1566 System.out.println(node);
1567 printObjectTree(((Loop)node).body);
1568 System.out.println("**** end contents Loop body");
1569 } else if (node instanceof Curly) {
1570 System.out.println(node);
1571 printObjectTree(((Curly)node).atom);
1572 System.out.println("**** end contents Curly body");
1573 } else if (node instanceof GroupCurly) {
1574 System.out.println(node);
1575 printObjectTree(((GroupCurly)node).atom);
1576 System.out.println("**** end contents GroupCurly body");
1577 } else if (node instanceof GroupTail) {
1578 System.out.println(node);
1579 System.out.println("Tail next is "+node.next);
1580 return;
1581 } else {
1582 System.out.println(node);
1583 }
1584 node = node.next;
1585 if (node != null)
1586 System.out.println("->next:");
1587 if (node == Pattern.accept) {
1588 System.out.println("Accept Node");
1589 node = null;
1590 }
1591 }
1592 }
1593
1594 /**
1595 * Used to accumulate information about a subtree of the object graph
1596 * so that optimizations can be applied to the subtree.
1597 */
1598 static final class TreeInfo {
1599 int minLength;
1600 int maxLength;
1601 boolean maxValid;
1602 boolean deterministic;
1603
1604 TreeInfo() {
1605 reset();
1606 }
1607 void reset() {
1608 minLength = 0;
1609 maxLength = 0;
1610 maxValid = true;
1611 deterministic = true;
1612 }
1613 }
1614
1615 /*
1616 * The following private methods are mainly used to improve the
1617 * readability of the code. In order to let the Java compiler easily
1618 * inline them, we should not put many assertions or error checks in them.
1619 */
1620
1621 /**
1622 * Indicates whether a particular flag is set or not.
1623 */
1624 private boolean has(int f) {
1625 return (flags & f) != 0;
1626 }
1627
1628 /**
1629 * Match next character, signal error if failed.
1630 */
1631 private void accept(int ch, String s) {
1632 int testChar = temp[cursor++];
1633 if (has(COMMENTS))
1634 testChar = parsePastWhitespace(testChar);
1635 if (ch != testChar) {
1636 throw error(s);
1637 }
1638 }
1639
1640 /**
1641 * Mark the end of pattern with a specific character.
1642 */
1643 private void mark(int c) {
1644 temp[patternLength] = c;
1645 }
1646
1647 /**
1648 * Peek the next character, and do not advance the cursor.
1649 */
1650 private int peek() {
1651 int ch = temp[cursor];
1652 if (has(COMMENTS))
1653 ch = peekPastWhitespace(ch);
1654 return ch;
1655 }
1656
1657 /**
1658 * Read the next character, and advance the cursor by one.
1659 */
1660 private int read() {
1661 int ch = temp[cursor++];
1662 if (has(COMMENTS))
1663 ch = parsePastWhitespace(ch);
1664 return ch;
1665 }
1666
1667 /**
1668 * Read the next character, and advance the cursor by one,
1669 * ignoring the COMMENTS setting
1670 */
1671 private int readEscaped() {
1672 int ch = temp[cursor++];
1673 return ch;
1674 }
1675
1676 /**
1677 * Advance the cursor by one, and peek the next character.
1678 */
1679 private int next() {
1680 int ch = temp[++cursor];
1681 if (has(COMMENTS))
1682 ch = peekPastWhitespace(ch);
1683 return ch;
1684 }
1685
1686 /**
1687 * Advance the cursor by one, and peek the next character,
1688 * ignoring the COMMENTS setting
1689 */
1690 private int nextEscaped() {
1691 int ch = temp[++cursor];
1692 return ch;
1693 }
1694
1695 /**
1696 * If in xmode peek past whitespace and comments.
1697 */
1698 private int peekPastWhitespace(int ch) {
1699 while (ASCII.isSpace(ch) || ch == '#') {
1700 while (ASCII.isSpace(ch))
1701 ch = temp[++cursor];
1702 if (ch == '#') {
1703 ch = peekPastLine();
1704 }
1705 }
1706 return ch;
1707 }
1708
1709 /**
1710 * If in xmode parse past whitespace and comments.
1711 */
1712 private int parsePastWhitespace(int ch) {
1713 while (ASCII.isSpace(ch) || ch == '#') {
1714 while (ASCII.isSpace(ch))
1715 ch = temp[cursor++];
1716 if (ch == '#')
1717 ch = parsePastLine();
1718 }
1719 return ch;
1720 }
1721
1722 /**
1723 * xmode parse past comment to end of line.
1724 */
1725 private int parsePastLine() {
1726 int ch = temp[cursor++];
1727 while (ch != 0 && !isLineSeparator(ch))
1728 ch = temp[cursor++];
1729 return ch;
1730 }
1731
1732 /**
1733 * xmode peek past comment to end of line.
1734 */
1735 private int peekPastLine() {
1736 int ch = temp[++cursor];
1737 while (ch != 0 && !isLineSeparator(ch))
1738 ch = temp[++cursor];
1739 return ch;
1740 }
1741
1742 /**
1743 * Determines if character is a line separator in the current mode
1744 */
1745 private boolean isLineSeparator(int ch) {
1746 if (has(UNIX_LINES)) {
1747 return ch == '\n';
1748 } else {
1749 return (ch == '\n' ||
1750 ch == '\r' ||
1751 (ch|1) == '\u2029' ||
1752 ch == '\u0085');
1753 }
1754 }
1755
1756 /**
1757 * Read the character after the next one, and advance the cursor by two.
1758 */
1759 private int skip() {
1760 int i = cursor;
1761 int ch = temp[i+1];
1762 cursor = i + 2;
1763 return ch;
1764 }
1765
1766 /**
1767 * Unread one next character, and retreat cursor by one.
1768 */
1769 private void unread() {
1770 cursor--;
1771 }
1772
1773 /**
1774 * Internal method used for handling all syntax errors. The pattern is
1775 * displayed with a pointer to aid in locating the syntax error.
1776 */
1777 private PatternSyntaxException error(String s) {
1778 return new PatternSyntaxException(s, normalizedPattern, cursor - 1);
1779 }
1780
1781 /**
1782 * Determines if there is any supplementary character or unpaired
1783 * surrogate in the specified range.
1784 */
1785 private boolean findSupplementary(int start, int end) {
1786 for (int i = start; i < end; i++) {
1787 if (isSupplementary(temp[i]))
1788 return true;
1789 }
1790 return false;
1791 }
1792
1793 /**
1794 * Determines if the specified code point is a supplementary
1795 * character or unpaired surrogate.
1796 */
1797 private static final boolean isSupplementary(int ch) {
1798 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT ||
1799 Character.isSurrogate((char)ch);
1800 }
1801
1802 /**
1803 * The following methods handle the main parsing. They are sorted
1804 * according to their precedence order, the lowest one first.
1805 */
1806
1807 /**
1808 * The expression is parsed with branch nodes added for alternations.
1809 * This may be called recursively to parse sub expressions that may
1810 * contain alternations.
1811 */
1812 private Node expr(Node end) {
1813 Node prev = null;
1814 Node firstTail = null;
1815 Node branchConn = null;
1816
1817 for (;;) {
1818 Node node = sequence(end);
1819 Node nodeTail = root; //double return
1820 if (prev == null) {
1821 prev = node;
1822 firstTail = nodeTail;
1823 } else {
1824 // Branch
1825 if (branchConn == null) {
1826 branchConn = new BranchConn();
1827 branchConn.next = end;
1828 }
1829 if (node == end) {
1830 // if the node returned from sequence() is "end"
1831 // we have an empty expr, set a null atom into
1832 // the branch to indicate to go "next" directly.
1833 node = null;
1834 } else {
1835 // the "tail.next" of each atom goes to branchConn
1836 nodeTail.next = branchConn;
1837 }
1838 if (prev instanceof Branch) {
1839 ((Branch)prev).add(node);
1840 } else {
1841 if (prev == end) {
1842 prev = null;
1843 } else {
1844 // replace the "end" with "branchConn" at its tail.next
1845 // when put the "prev" into the branch as the first atom.
1846 firstTail.next = branchConn;
1847 }
1848 prev = new Branch(prev, node, branchConn);
1849 }
1850 }
1851 if (peek() != '|') {
1852 return prev;
1853 }
1854 next();
1855 }
1856 }
1857
1858 /**
1859 * Parsing of sequences between alternations.
1860 */
1861 private Node sequence(Node end) {
1862 Node head = null;
1863 Node tail = null;
1864 Node node = null;
1865 LOOP:
1866 for (;;) {
1867 int ch = peek();
1868 switch (ch) {
1869 case '(':
1870 // Because group handles its own closure,
1871 // we need to treat it differently
1872 node = group0();
1873 // Check for comment or flag group
1874 if (node == null)
1875 continue;
1876 if (head == null)
1877 head = node;
1878 else
1879 tail.next = node;
1880 // Double return: Tail was returned in root
1881 tail = root;
1882 continue;
1883 case '[':
1884 node = clazz(true);
1885 break;
1886 case '\\':
1887 ch = nextEscaped();
1888 if (ch == 'p' || ch == 'P') {
1889 boolean oneLetter = true;
1890 boolean comp = (ch == 'P');
1891 ch = next(); // Consume { if present
1892 if (ch != '{') {
1893 unread();
1894 } else {
1895 oneLetter = false;
1896 }
1897 node = family(oneLetter, comp);
1898 } else {
1899 unread();
1900 node = atom();
1901 }
1902 break;
1903 case '^':
1904 next();
1905 if (has(MULTILINE)) {
1906 if (has(UNIX_LINES))
1907 node = new UnixCaret();
1908 else
1909 node = new Caret();
1910 } else {
1911 node = new Begin();
1912 }
1913 break;
1914 case '$':
1915 next();
1916 if (has(UNIX_LINES))
1917 node = new UnixDollar(has(MULTILINE));
1918 else
1919 node = new Dollar(has(MULTILINE));
1920 break;
1921 case '.':
1922 next();
1923 if (has(DOTALL)) {
1924 node = new All();
1925 } else {
1926 if (has(UNIX_LINES))
1927 node = new UnixDot();
1928 else {
1929 node = new Dot();
1930 }
1931 }
1932 break;
1933 case '|':
1934 case ')':
1935 break LOOP;
1936 case ']': // Now interpreting dangling ] and } as literals
1937 case '}':
1938 node = atom();
1939 break;
1940 case '?':
1941 case '*':
1942 case '+':
1943 next();
1944 throw error("Dangling meta character '" + ((char)ch) + "'");
1945 case 0:
1946 if (cursor >= patternLength) {
1947 break LOOP;
1948 }
1949 // Fall through
1950 default:
1951 node = atom();
1952 break;
1953 }
1954
1955 node = closure(node);
1956
1957 if (head == null) {
1958 head = tail = node;
1959 } else {
1960 tail.next = node;
1961 tail = node;
1962 }
1963 }
1964 if (head == null) {
1965 return end;
1966 }
1967 tail.next = end;
1968 root = tail; //double return
1969 return head;
1970 }
1971
1972 /**
1973 * Parse and add a new Single or Slice.
1974 */
1975 private Node atom() {
1976 int first = 0;
1977 int prev = -1;
1978 boolean hasSupplementary = false;
1979 int ch = peek();
1980 for (;;) {
1981 switch (ch) {
1982 case '*':
1983 case '+':
1984 case '?':
1985 case '{':
1986 if (first > 1) {
1987 cursor = prev; // Unwind one character
1988 first--;
1989 }
1990 break;
1991 case '$':
1992 case '.':
1993 case '^':
1994 case '(':
1995 case '[':
1996 case '|':
1997 case ')':
1998 break;
1999 case '\\':
2000 ch = nextEscaped();
2001 if (ch == 'p' || ch == 'P') { // Property
2002 if (first > 0) { // Slice is waiting; handle it first
2003 unread();
2004 break;
2005 } else { // No slice; just return the family node
2006 boolean comp = (ch == 'P');
2007 boolean oneLetter = true;
2008 ch = next(); // Consume { if present
2009 if (ch != '{')
2010 unread();
2011 else
2012 oneLetter = false;
2013 return family(oneLetter, comp);
2014 }
2015 }
2016 unread();
2017 prev = cursor;
2018 ch = escape(false, first == 0);
2019 if (ch >= 0) {
2020 append(ch, first);
2021 first++;
2022 if (isSupplementary(ch)) {
2023 hasSupplementary = true;
2024 }
2025 ch = peek();
2026 continue;
2027 } else if (first == 0) {
2028 return root;
2029 }
2030 // Unwind meta escape sequence
2031 cursor = prev;
2032 break;
2033 case 0:
2034 if (cursor >= patternLength) {
2035 break;
2036 }
2037 // Fall through
2038 default:
2039 prev = cursor;
2040 append(ch, first);
2041 first++;
2042 if (isSupplementary(ch)) {
2043 hasSupplementary = true;
2044 }
2045 ch = next();
2046 continue;
2047 }
2048 break;
2049 }
2050 if (first == 1) {
2051 return newSingle(buffer[0]);
2052 } else {
2053 return newSlice(buffer, first, hasSupplementary);
2054 }
2055 }
2056
2057 private void append(int ch, int len) {
2058 if (len >= buffer.length) {
2059 int[] tmp = new int[len+len];
2060 System.arraycopy(buffer, 0, tmp, 0, len);
2061 buffer = tmp;
2062 }
2063 buffer[len] = ch;
2064 }
2065
2066 /**
2067 * Parses a backref greedily, taking as many numbers as it
2068 * can. The first digit is always treated as a backref, but
2069 * multi digit numbers are only treated as a backref if at
2070 * least that many backrefs exist at this point in the regex.
2071 */
2072 private Node ref(int refNum) {
2073 boolean done = false;
2074 while(!done) {
2075 int ch = peek();
2076 switch(ch) {
2077 case '0':
2078 case '1':
2079 case '2':
2080 case '3':
2081 case '4':
2082 case '5':
2083 case '6':
2084 case '7':
2085 case '8':
2086 case '9':
2087 int newRefNum = (refNum * 10) + (ch - '0');
2088 // Add another number if it doesn't make a group
2089 // that doesn't exist
2090 if (capturingGroupCount - 1 < newRefNum) {
2091 done = true;
2092 break;
2093 }
2094 refNum = newRefNum;
2095 read();
2096 break;
2097 default:
2098 done = true;
2099 break;
2100 }
2101 }
2102 if (has(CASE_INSENSITIVE))
2103 return new CIBackRef(refNum, has(UNICODE_CASE));
2104 else
2105 return new BackRef(refNum);
2106 }
2107
2108 /**
2109 * Parses an escape sequence to determine the actual value that needs
2110 * to be matched.
2111 * If -1 is returned and create was true a new object was added to the tree
2112 * to handle the escape sequence.
2113 * If the returned value is greater than zero, it is the value that
2114 * matches the escape sequence.
2115 */
2116 private int escape(boolean inclass, boolean create) {
2117 int ch = skip();
2118 switch (ch) {
2119 case '0':
2120 return o();
2121 case '1':
2122 case '2':
2123 case '3':
2124 case '4':
2125 case '5':
2126 case '6':
2127 case '7':
2128 case '8':
2129 case '9':
2130 if (inclass) break;
2131 if (create) {
2132 root = ref((ch - '0'));
2133 }
2134 return -1;
2135 case 'A':
2136 if (inclass) break;
2137 if (create) root = new Begin();
2138 return -1;
2139 case 'B':
2140 if (inclass) break;
2141 if (create) root = new Bound(Bound.NONE);
2142 return -1;
2143 case 'C':
2144 break;
2145 case 'D':
2146 if (create) root = new Ctype(ASCII.DIGIT).complement();
2147 return -1;
2148 case 'E':
2149 case 'F':
2150 break;
2151 case 'G':
2152 if (inclass) break;
2153 if (create) root = new LastMatch();
2154 return -1;
2155 case 'H':
2156 case 'I':
2157 case 'J':
2158 case 'K':
2159 case 'L':
2160 case 'M':
2161 case 'N':
2162 case 'O':
2163 case 'P':
2164 case 'Q':
2165 case 'R':
2166 break;
2167 case 'S':
2168 if (create) root = new Ctype(ASCII.SPACE).complement();
2169 return -1;
2170 case 'T':
2171 case 'U':
2172 case 'V':
2173 break;
2174 case 'W':
2175 if (create) root = new Ctype(ASCII.WORD).complement();
2176 return -1;
2177 case 'X':
2178 case 'Y':
2179 break;
2180 case 'Z':
2181 if (inclass) break;
2182 if (create) {
2183 if (has(UNIX_LINES))
2184 root = new UnixDollar(false);
2185 else
2186 root = new Dollar(false);
2187 }
2188 return -1;
2189 case 'a':
2190 return '\007';
2191 case 'b':
2192 if (inclass) break;
2193 if (create) root = new Bound(Bound.BOTH);
2194 return -1;
2195 case 'c':
2196 return c();
2197 case 'd':
2198 if (create) root = new Ctype(ASCII.DIGIT);
2199 return -1;
2200 case 'e':
2201 return '\033';
2202 case 'f':
2203 return '\f';
2204 case 'g':
2205 case 'h':
2206 case 'i':
2207 case 'j':
2208 break;
2209 case 'k':
2210 if (inclass)
2211 break;
2212 if (read() != '<')
2213 throw error("\\k is not followed by '<' for named capturing group");
2214 String name = groupname(read());
2215 if (!namedGroups().containsKey(name))
2216 throw error("(named capturing group <"+ name+"> does not exit");
2217 if (create) {
2218 if (has(CASE_INSENSITIVE))
2219 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE));
2220 else
2221 root = new BackRef(namedGroups().get(name));
2222 }
2223 return -1;
2224 case 'l':
2225 case 'm':
2226 break;
2227 case 'n':
2228 return '\n';
2229 case 'o':
2230 case 'p':
2231 case 'q':
2232 break;
2233 case 'r':
2234 return '\r';
2235 case 's':
2236 if (create) root = new Ctype(ASCII.SPACE);
2237 return -1;
2238 case 't':
2239 return '\t';
2240 case 'u':
2241 return u();
2242 case 'v':
2243 return '\013';
2244 case 'w':
2245 if (create) root = new Ctype(ASCII.WORD);
2246 return -1;
2247 case 'x':
2248 return x();
2249 case 'y':
2250 break;
2251 case 'z':
2252 if (inclass) break;
2253 if (create) root = new End();
2254 return -1;
2255 default:
2256 return ch;
2257 }
2258 throw error("Illegal/unsupported escape sequence");
2259 }
2260
2261 /**
2262 * Parse a character class, and return the node that matches it.
2263 *
2264 * Consumes a ] on the way out if consume is true. Usually consume
2265 * is true except for the case of [abc&&def] where def is a separate
2266 * right hand node with "understood" brackets.
2267 */
2268 private CharProperty clazz(boolean consume) {
2269 CharProperty prev = null;
2270 CharProperty node = null;
2271 BitClass bits = new BitClass();
2272 boolean include = true;
2273 boolean firstInClass = true;
2274 int ch = next();
2275 for (;;) {
2276 switch (ch) {
2277 case '^':
2278 // Negates if first char in a class, otherwise literal
2279 if (firstInClass) {
2280 if (temp[cursor-1] != '[')
2281 break;
2282 ch = next();
2283 include = !include;
2284 continue;
2285 } else {
2286 // ^ not first in class, treat as literal
2287 break;
2288 }
2289 case '[':
2290 firstInClass = false;
2291 node = clazz(true);
2292 if (prev == null)
2293 prev = node;
2294 else
2295 prev = union(prev, node);
2296 ch = peek();
2297 continue;
2298 case '&':
2299 firstInClass = false;
2300 ch = next();
2301 if (ch == '&') {
2302 ch = next();
2303 CharProperty rightNode = null;
2304 while (ch != ']' && ch != '&') {
2305 if (ch == '[') {
2306 if (rightNode == null)
2307 rightNode = clazz(true);
2308 else
2309 rightNode = union(rightNode, clazz(true));
2310 } else { // abc&&def
2311 unread();
2312 rightNode = clazz(false);
2313 }
2314 ch = peek();
2315 }
2316 if (rightNode != null)
2317 node = rightNode;
2318 if (prev == null) {
2319 if (rightNode == null)
2320 throw error("Bad class syntax");
2321 else
2322 prev = rightNode;
2323 } else {
2324 prev = intersection(prev, node);
2325 }
2326 } else {
2327 // treat as a literal &
2328 unread();
2329 break;
2330 }
2331 continue;
2332 case 0:
2333 firstInClass = false;
2334 if (cursor >= patternLength)
2335 throw error("Unclosed character class");
2336 break;
2337 case ']':
2338 firstInClass = false;
2339 if (prev != null) {
2340 if (consume)
2341 next();
2342 return prev;
2343 }
2344 break;
2345 default:
2346 firstInClass = false;
2347 break;
2348 }
2349 node = range(bits);
2350 if (include) {
2351 if (prev == null) {
2352 prev = node;
2353 } else {
2354 if (prev != node)
2355 prev = union(prev, node);
2356 }
2357 } else {
2358 if (prev == null) {
2359 prev = node.complement();
2360 } else {
2361 if (prev != node)
2362 prev = setDifference(prev, node);
2363 }
2364 }
2365 ch = peek();
2366 }
2367 }
2368
2369 private CharProperty bitsOrSingle(BitClass bits, int ch) {
2370 /* Bits can only handle codepoints in [u+0000-u+00ff] range.
2371 Use "single" node instead of bits when dealing with unicode
2372 case folding for codepoints listed below.
2373 (1)Uppercase out of range: u+00ff, u+00b5
2374 toUpperCase(u+00ff) -> u+0178
2375 toUpperCase(u+00b5) -> u+039c
2376 (2)LatinSmallLetterLongS u+17f
2377 toUpperCase(u+017f) -> u+0053
2378 (3)LatinSmallLetterDotlessI u+131
2379 toUpperCase(u+0131) -> u+0049
2380 (4)LatinCapitalLetterIWithDotAbove u+0130
2381 toLowerCase(u+0130) -> u+0069
2382 (5)KelvinSign u+212a
2383 toLowerCase(u+212a) ==> u+006B
2384 (6)AngstromSign u+212b
2385 toLowerCase(u+212b) ==> u+00e5
2386 */
2387 int d;
2388 if (ch < 256 &&
2389 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) &&
2390 (ch == 0xff || ch == 0xb5 ||
2391 ch == 0x49 || ch == 0x69 || //I and i
2392 ch == 0x53 || ch == 0x73 || //S and s
2393 ch == 0x4b || ch == 0x6b || //K and k
2394 ch == 0xc5 || ch == 0xe5))) //A+ring
2395 return bits.add(ch, flags());
2396 return newSingle(ch);
2397 }
2398
2399 /**
2400 * Parse a single character or a character range in a character class
2401 * and return its representative node.
2402 */
2403 private CharProperty range(BitClass bits) {
2404 int ch = peek();
2405 if (ch == '\\') {
2406 ch = nextEscaped();
2407 if (ch == 'p' || ch == 'P') { // A property
2408 boolean comp = (ch == 'P');
2409 boolean oneLetter = true;
2410 // Consume { if present
2411 ch = next();
2412 if (ch != '{')
2413 unread();
2414 else
2415 oneLetter = false;
2416 return family(oneLetter, comp);
2417 } else { // ordinary escape
2418 unread();
2419 ch = escape(true, true);
2420 if (ch == -1)
2421 return (CharProperty) root;
2422 }
2423 } else {
2424 ch = single();
2425 }
2426 if (ch >= 0) {
2427 if (peek() == '-') {
2428 int endRange = temp[cursor+1];
2429 if (endRange == '[') {
2430 return bitsOrSingle(bits, ch);
2431 }
2432 if (endRange != ']') {
2433 next();
2434 int m = single();
2435 if (m < ch)
2436 throw error("Illegal character range");
2437 if (has(CASE_INSENSITIVE))
2438 return caseInsensitiveRangeFor(ch, m);
2439 else
2440 return rangeFor(ch, m);
2441 }
2442 }
2443 return bitsOrSingle(bits, ch);
2444 }
2445 throw error("Unexpected character '"+((char)ch)+"'");
2446 }
2447
2448 private int single() {
2449 int ch = peek();
2450 switch (ch) {
2451 case '\\':
2452 return escape(true, false);
2453 default:
2454 next();
2455 return ch;
2456 }
2457 }
2458
2459 /**
2460 * Parses a Unicode character family and returns its representative node.
2461 */
2462 private CharProperty family(boolean singleLetter,
2463 boolean maybeComplement)
2464 {
2465 next();
2466 String name;
2467 CharProperty node;
2468
2469 if (singleLetter) {
2470 int c = temp[cursor];
2471 if (!Character.isSupplementaryCodePoint(c)) {
2472 name = String.valueOf((char)c);
2473 } else {
2474 name = new String(temp, cursor, 1);
2475 }
2476 read();
2477 } else {
2478 int i = cursor;
2479 mark('}');
2480 while(read() != '}') {
2481 }
2482 mark('\000');
2483 int j = cursor;
2484 if (j > patternLength)
2485 throw error("Unclosed character family");
2486 if (i + 1 >= j)
2487 throw error("Empty character family");
2488 name = new String(temp, i, j-i-1);
2489 }
2490
2491 if (name.startsWith("In")) {
2492 node = unicodeBlockPropertyFor(name.substring(2));
2493 } else {
2494 if (name.startsWith("Is"))
2495 name = name.substring(2);
2496 node = charPropertyNodeFor(name);
2497 }
2498 if (maybeComplement) {
2499 if (node instanceof Category || node instanceof Block)
2500 hasSupplementary = true;
2501 node = node.complement();
2502 }
2503 return node;
2504 }
2505
2506 /**
2507 * Returns a CharProperty matching all characters in a UnicodeBlock.
2508 */
2509 private CharProperty unicodeBlockPropertyFor(String name) {
2510 final Character.UnicodeBlock block;
2511 try {
2512 block = Character.UnicodeBlock.forName(name);
2513 } catch (IllegalArgumentException iae) {
2514 throw error("Unknown character block name {" + name + "}");
2515 }
2516 return new Block(block);
2517 }
2518
2519 /**
2520 * Returns a CharProperty matching all characters in a named property.
2521 */
2522 private CharProperty charPropertyNodeFor(String name) {
2523 CharProperty p = CharPropertyNames.charPropertyFor(name);
2524 if (p == null)
2525 throw error("Unknown character property name {" + name + "}");
2526 return p;
2527 }
2528
2529 /**
2530 * Parses and returns the name of a "named capturing group", the trailing
2531 * ">" is consumed after parsing.
2532 */
2533 private String groupname(int ch) {
2534 StringBuilder sb = new StringBuilder();
2535 sb.append(Character.toChars(ch));
2536 while (ASCII.isLower(ch=read()) || ASCII.isUpper(ch) ||
2537 ASCII.isDigit(ch)) {
2538 sb.append(Character.toChars(ch));
2539 }
2540 if (sb.length() == 0)
2541 throw error("named capturing group has 0 length name");
2542 if (ch != '>')
2543 throw error("named capturing group is missing trailing '>'");
2544 return sb.toString();
2545 }
2546
2547 /**
2548 * Parses a group and returns the head node of a set of nodes that process
2549 * the group. Sometimes a double return system is used where the tail is
2550 * returned in root.
2551 */
2552 private Node group0() {
2553 boolean capturingGroup = false;
2554 Node head = null;
2555 Node tail = null;
2556 int save = flags;
2557 root = null;
2558 int ch = next();
2559 if (ch == '?') {
2560 ch = skip();
2561 switch (ch) {
2562 case ':': // (?:xxx) pure group
2563 head = createGroup(true);
2564 tail = root;
2565 head.next = expr(tail);
2566 break;
2567 case '=': // (?=xxx) and (?!xxx) lookahead
2568 case '!':
2569 head = createGroup(true);
2570 tail = root;
2571 head.next = expr(tail);
2572 if (ch == '=') {
2573 head = tail = new Pos(head);
2574 } else {
2575 head = tail = new Neg(head);
2576 }
2577 break;
2578 case '>': // (?>xxx) independent group
2579 head = createGroup(true);
2580 tail = root;
2581 head.next = expr(tail);
2582 head = tail = new Ques(head, INDEPENDENT);
2583 break;
2584 case '<': // (?<xxx) look behind
2585 ch = read();
2586 if (ASCII.isLower(ch) || ASCII.isUpper(ch)) {
2587 // named captured group
2588 String name = groupname(ch);
2589 if (namedGroups().containsKey(name))
2590 throw error("Named capturing group <" + name
2591 + "> is already defined");
2592 capturingGroup = true;
2593 head = createGroup(false);
2594 tail = root;
2595 namedGroups().put(name, capturingGroupCount-1);
2596 head.next = expr(tail);
2597 break;
2598 }
2599 int start = cursor;
2600 head = createGroup(true);
2601 tail = root;
2602 head.next = expr(tail);
2603 tail.next = lookbehindEnd;
2604 TreeInfo info = new TreeInfo();
2605 head.study(info);
2606 if (info.maxValid == false) {
2607 throw error("Look-behind group does not have "
2608 + "an obvious maximum length");
2609 }
2610 boolean hasSupplementary = findSupplementary(start, patternLength);
2611 if (ch == '=') {
2612 head = tail = (hasSupplementary ?
2613 new BehindS(head, info.maxLength,
2614 info.minLength) :
2615 new Behind(head, info.maxLength,
2616 info.minLength));
2617 } else if (ch == '!') {
2618 head = tail = (hasSupplementary ?
2619 new NotBehindS(head, info.maxLength,
2620 info.minLength) :
2621 new NotBehind(head, info.maxLength,
2622 info.minLength));
2623 } else {
2624 throw error("Unknown look-behind group");
2625 }
2626 break;
2627 case '$':
2628 case '@':
2629 throw error("Unknown group type");
2630 default: // (?xxx:) inlined match flags
2631 unread();
2632 addFlag();
2633 ch = read();
2634 if (ch == ')') {
2635 return null; // Inline modifier only
2636 }
2637 if (ch != ':') {
2638 throw error("Unknown inline modifier");
2639 }
2640 head = createGroup(true);
2641 tail = root;
2642 head.next = expr(tail);
2643 break;
2644 }
2645 } else { // (xxx) a regular group
2646 capturingGroup = true;
2647 head = createGroup(false);
2648 tail = root;
2649 head.next = expr(tail);
2650 }
2651
2652 accept(')', "Unclosed group");
2653 flags = save;
2654
2655 // Check for quantifiers
2656 Node node = closure(head);
2657 if (node == head) { // No closure
2658 root = tail;
2659 return node; // Dual return
2660 }
2661 if (head == tail) { // Zero length assertion
2662 root = node;
2663 return node; // Dual return
2664 }
2665
2666 if (node instanceof Ques) {
2667 Ques ques = (Ques) node;
2668 if (ques.type == POSSESSIVE) {
2669 root = node;
2670 return node;
2671 }
2672 tail.next = new BranchConn();
2673 tail = tail.next;
2674 if (ques.type == GREEDY) {
2675 head = new Branch(head, null, tail);
2676 } else { // Reluctant quantifier
2677 head = new Branch(null, head, tail);
2678 }
2679 root = tail;
2680 return head;
2681 } else if (node instanceof Curly) {
2682 Curly curly = (Curly) node;
2683 if (curly.type == POSSESSIVE) {
2684 root = node;
2685 return node;
2686 }
2687 // Discover if the group is deterministic
2688 TreeInfo info = new TreeInfo();
2689 if (head.study(info)) { // Deterministic
2690 GroupTail temp = (GroupTail) tail;
2691 head = root = new GroupCurly(head.next, curly.cmin,
2692 curly.cmax, curly.type,
2693 ((GroupTail)tail).localIndex,
2694 ((GroupTail)tail).groupIndex,
2695 capturingGroup);
2696 return head;
2697 } else { // Non-deterministic
2698 int temp = ((GroupHead) head).localIndex;
2699 Loop loop;
2700 if (curly.type == GREEDY)
2701 loop = new Loop(this.localCount, temp);
2702 else // Reluctant Curly
2703 loop = new LazyLoop(this.localCount, temp);
2704 Prolog prolog = new Prolog(loop);
2705 this.localCount += 1;
2706 loop.cmin = curly.cmin;
2707 loop.cmax = curly.cmax;
2708 loop.body = head;
2709 tail.next = loop;
2710 root = loop;
2711 return prolog; // Dual return
2712 }
2713 }
2714 throw error("Internal logic error");
2715 }
2716
2717 /**
2718 * Create group head and tail nodes using double return. If the group is
2719 * created with anonymous true then it is a pure group and should not
2720 * affect group counting.
2721 */
2722 private Node createGroup(boolean anonymous) {
2723 int localIndex = localCount++;
2724 int groupIndex = 0;
2725 if (!anonymous)
2726 groupIndex = capturingGroupCount++;
2727 GroupHead head = new GroupHead(localIndex);
2728 root = new GroupTail(localIndex, groupIndex);
2729 if (!anonymous && groupIndex < 10)
2730 groupNodes[groupIndex] = head;
2731 return head;
2732 }
2733
2734 /**
2735 * Parses inlined match flags and set them appropriately.
2736 */
2737 private void addFlag() {
2738 int ch = peek();
2739 for (;;) {
2740 switch (ch) {
2741 case 'i':
2742 flags |= CASE_INSENSITIVE;
2743 break;
2744 case 'm':
2745 flags |= MULTILINE;
2746 break;
2747 case 's':
2748 flags |= DOTALL;
2749 break;
2750 case 'd':
2751 flags |= UNIX_LINES;
2752 break;
2753 case 'u':
2754 flags |= UNICODE_CASE;
2755 break;
2756 case 'c':
2757 flags |= CANON_EQ;
2758 break;
2759 case 'x':
2760 flags |= COMMENTS;
2761 break;
2762 case '-': // subFlag then fall through
2763 ch = next();
2764 subFlag();
2765 default:
2766 return;
2767 }
2768 ch = next();
2769 }
2770 }
2771
2772 /**
2773 * Parses the second part of inlined match flags and turns off
2774 * flags appropriately.
2775 */
2776 private void subFlag() {
2777 int ch = peek();
2778 for (;;) {
2779 switch (ch) {
2780 case 'i':
2781 flags &= ~CASE_INSENSITIVE;
2782 break;
2783 case 'm':
2784 flags &= ~MULTILINE;
2785 break;
2786 case 's':
2787 flags &= ~DOTALL;
2788 break;
2789 case 'd':
2790 flags &= ~UNIX_LINES;
2791 break;
2792 case 'u':
2793 flags &= ~UNICODE_CASE;
2794 break;
2795 case 'c':
2796 flags &= ~CANON_EQ;
2797 break;
2798 case 'x':
2799 flags &= ~COMMENTS;
2800 break;
2801 default:
2802 return;
2803 }
2804 ch = next();
2805 }
2806 }
2807
2808 static final int MAX_REPS = 0x7FFFFFFF;
2809
2810 static final int GREEDY = 0;
2811
2812 static final int LAZY = 1;
2813
2814 static final int POSSESSIVE = 2;
2815
2816 static final int INDEPENDENT = 3;
2817
2818 /**
2819 * Processes repetition. If the next character peeked is a quantifier
2820 * then new nodes must be appended to handle the repetition.
2821 * Prev could be a single or a group, so it could be a chain of nodes.
2822 */
2823 private Node closure(Node prev) {
2824 Node atom;
2825 int ch = peek();
2826 switch (ch) {
2827 case '?':
2828 ch = next();
2829 if (ch == '?') {
2830 next();
2831 return new Ques(prev, LAZY);
2832 } else if (ch == '+') {
2833 next();
2834 return new Ques(prev, POSSESSIVE);
2835 }
2836 return new Ques(prev, GREEDY);
2837 case '*':
2838 ch = next();
2839 if (ch == '?') {
2840 next();
2841 return new Curly(prev, 0, MAX_REPS, LAZY);
2842 } else if (ch == '+') {
2843 next();
2844 return new Curly(prev, 0, MAX_REPS, POSSESSIVE);
2845 }
2846 return new Curly(prev, 0, MAX_REPS, GREEDY);
2847 case '+':
2848 ch = next();
2849 if (ch == '?') {
2850 next();
2851 return new Curly(prev, 1, MAX_REPS, LAZY);
2852 } else if (ch == '+') {
2853 next();
2854 return new Curly(prev, 1, MAX_REPS, POSSESSIVE);
2855 }
2856 return new Curly(prev, 1, MAX_REPS, GREEDY);
2857 case '{':
2858 ch = temp[cursor+1];
2859 if (ASCII.isDigit(ch)) {
2860 skip();
2861 int cmin = 0;
2862 do {
2863 cmin = cmin * 10 + (ch - '0');
2864 } while (ASCII.isDigit(ch = read()));
2865 int cmax = cmin;
2866 if (ch == ',') {
2867 ch = read();
2868 cmax = MAX_REPS;
2869 if (ch != '}') {
2870 cmax = 0;
2871 while (ASCII.isDigit(ch)) {
2872 cmax = cmax * 10 + (ch - '0');
2873 ch = read();
2874 }
2875 }
2876 }
2877 if (ch != '}')
2878 throw error("Unclosed counted closure");
2879 if (((cmin) | (cmax) | (cmax - cmin)) < 0)
2880 throw error("Illegal repetition range");
2881 Curly curly;
2882 ch = peek();
2883 if (ch == '?') {
2884 next();
2885 curly = new Curly(prev, cmin, cmax, LAZY);
2886 } else if (ch == '+') {
2887 next();
2888 curly = new Curly(prev, cmin, cmax, POSSESSIVE);
2889 } else {
2890 curly = new Curly(prev, cmin, cmax, GREEDY);
2891 }
2892 return curly;
2893 } else {
2894 throw error("Illegal repetition");
2895 }
2896 default:
2897 return prev;
2898 }
2899 }
2900
2901 /**
2902 * Utility method for parsing control escape sequences.
2903 */
2904 private int c() {
2905 if (cursor < patternLength) {
2906 return read() ^ 64;
2907 }
2908 throw error("Illegal control escape sequence");
2909 }
2910
2911 /**
2912 * Utility method for parsing octal escape sequences.
2913 */
2914 private int o() {
2915 int n = read();
2916 if (((n-'0')|('7'-n)) >= 0) {
2917 int m = read();
2918 if (((m-'0')|('7'-m)) >= 0) {
2919 int o = read();
2920 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) {
2921 return (n - '0') * 64 + (m - '0') * 8 + (o - '0');
2922 }
2923 unread();
2924 return (n - '0') * 8 + (m - '0');
2925 }
2926 unread();
2927 return (n - '0');
2928 }
2929 throw error("Illegal octal escape sequence");
2930 }
2931
2932 /**
2933 * Utility method for parsing hexadecimal escape sequences.
2934 */
2935 private int x() {
2936 int n = read();
2937 if (ASCII.isHexDigit(n)) {
2938 int m = read();
2939 if (ASCII.isHexDigit(m)) {
2940 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m);
2941 }
2942 }
2943 throw error("Illegal hexadecimal escape sequence");
2944 }
2945
2946 /**
2947 * Utility method for parsing unicode escape sequences.
2948 */
2949 private int cursor() {
2950 return cursor;
2951 }
2952
2953 private void setcursor(int pos) {
2954 cursor = pos;
2955 }
2956
2957 private int uxxxx() {
2958 int n = 0;
2959 for (int i = 0; i < 4; i++) {
2960 int ch = read();
2961 if (!ASCII.isHexDigit(ch)) {
2962 throw error("Illegal Unicode escape sequence");
2963 }
2964 n = n * 16 + ASCII.toDigit(ch);
2965 }
2966 return n;
2967 }
2968
2969 private int u() {
2970 int n = uxxxx();
2971 if (Character.isHighSurrogate((char)n)) {
2972 int cur = cursor();
2973 if (read() == '\\' && read() == 'u') {
2974 int n2 = uxxxx();
2975 if (Character.isLowSurrogate((char)n2))
2976 return Character.toCodePoint((char)n, (char)n2);
2977 }
2978 setcursor(cur);
2979 }
2980 return n;
2981 }
2982
2983 //
2984 // Utility methods for code point support
2985 //
2986
2987 private static final int countChars(CharSequence seq, int index,
2988 int lengthInCodePoints) {
2989 // optimization
2990 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) {
2991 assert (index >= 0 && index < seq.length());
2992 return 1;
2993 }
2994 int length = seq.length();
2995 int x = index;
2996 if (lengthInCodePoints >= 0) {
2997 assert (index >= 0 && index < length);
2998 for (int i = 0; x < length && i < lengthInCodePoints; i++) {
2999 if (Character.isHighSurrogate(seq.charAt(x++))) {
3000 if (x < length && Character.isLowSurrogate(seq.charAt(x))) {
3001 x++;
3002 }
3003 }
3004 }
3005 return x - index;
3006 }
3007
3008 assert (index >= 0 && index <= length);
3009 if (index == 0) {
3010 return 0;
3011 }
3012 int len = -lengthInCodePoints;
3013 for (int i = 0; x > 0 && i < len; i++) {
3014 if (Character.isLowSurrogate(seq.charAt(--x))) {
3015 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) {
3016 x--;
3017 }
3018 }
3019 }
3020 return index - x;
3021 }
3022
3023 private static final int countCodePoints(CharSequence seq) {
3024 int length = seq.length();
3025 int n = 0;
3026 for (int i = 0; i < length; ) {
3027 n++;
3028 if (Character.isHighSurrogate(seq.charAt(i++))) {
3029 if (i < length && Character.isLowSurrogate(seq.charAt(i))) {
3030 i++;
3031 }
3032 }
3033 }
3034 return n;
3035 }
3036
3037 /**
3038 * Creates a bit vector for matching Latin-1 values. A normal BitClass
3039 * never matches values above Latin-1, and a complemented BitClass always
3040 * matches values above Latin-1.
3041 */
3042 private static final class BitClass extends BmpCharProperty {
3043 final boolean[] bits;
3044 BitClass() { bits = new boolean[256]; }
3045 private BitClass(boolean[] bits) { this.bits = bits; }
3046 BitClass add(int c, int flags) {
3047 assert c >= 0 && c <= 255;
3048 if ((flags & CASE_INSENSITIVE) != 0) {
3049 if (ASCII.isAscii(c)) {
3050 bits[ASCII.toUpper(c)] = true;
3051 bits[ASCII.toLower(c)] = true;
3052 } else if ((flags & UNICODE_CASE) != 0) {
3053 bits[Character.toLowerCase(c)] = true;
3054 bits[Character.toUpperCase(c)] = true;
3055 }
3056 }
3057 bits[c] = true;
3058 return this;
3059 }
3060 boolean isSatisfiedBy(int ch) {
3061 return ch < 256 && bits[ch];
3062 }
3063 }
3064
3065 /**
3066 * Returns a suitably optimized, single character matcher.
3067 */
3068 private CharProperty newSingle(final int ch) {
3069 if (has(CASE_INSENSITIVE)) {
3070 int lower, upper;
3071 if (has(UNICODE_CASE)) {
3072 upper = Character.toUpperCase(ch);
3073 lower = Character.toLowerCase(upper);
3074 if (upper != lower)
3075 return new SingleU(lower);
3076 } else if (ASCII.isAscii(ch)) {
3077 lower = ASCII.toLower(ch);
3078 upper = ASCII.toUpper(ch);
3079 if (lower != upper)
3080 return new SingleI(lower, upper);
3081 }
3082 }
3083 if (isSupplementary(ch))
3084 return new SingleS(ch); // Match a given Unicode character
3085 return new Single(ch); // Match a given BMP character
3086 }
3087
3088 /**
3089 * Utility method for creating a string slice matcher.
3090 */
3091 private Node newSlice(int[] buf, int count, boolean hasSupplementary) {
3092 int[] tmp = new int[count];
3093 if (has(CASE_INSENSITIVE)) {
3094 if (has(UNICODE_CASE)) {
3095 for (int i = 0; i < count; i++) {
3096 tmp[i] = Character.toLowerCase(
3097 Character.toUpperCase(buf[i]));
3098 }
3099 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp);
3100 }
3101 for (int i = 0; i < count; i++) {
3102 tmp[i] = ASCII.toLower(buf[i]);
3103 }
3104 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp);
3105 }
3106 for (int i = 0; i < count; i++) {
3107 tmp[i] = buf[i];
3108 }
3109 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp);
3110 }
3111
3112 /**
3113 * The following classes are the building components of the object
3114 * tree that represents a compiled regular expression. The object tree
3115 * is made of individual elements that handle constructs in the Pattern.
3116 * Each type of object knows how to match its equivalent construct with
3117 * the match() method.
3118 */
3119
3120 /**
3121 * Base class for all node classes. Subclasses should override the match()
3122 * method as appropriate. This class is an accepting node, so its match()
3123 * always returns true.
3124 */
3125 static class Node extends Object {
3126 Node next;
3127 Node() {
3128 next = Pattern.accept;
3129 }
3130 /**
3131 * This method implements the classic accept node.
3132 */
3133 boolean match(Matcher matcher, int i, CharSequence seq) {
3134 matcher.last = i;
3135 matcher.groups[0] = matcher.first;
3136 matcher.groups[1] = matcher.last;
3137 return true;
3138 }
3139 /**
3140 * This method is good for all zero length assertions.
3141 */
3142 boolean study(TreeInfo info) {
3143 if (next != null) {
3144 return next.study(info);
3145 } else {
3146 return info.deterministic;
3147 }
3148 }
3149 }
3150
3151 static class LastNode extends Node {
3152 /**
3153 * This method implements the classic accept node with
3154 * the addition of a check to see if the match occurred
3155 * using all of the input.
3156 */
3157 boolean match(Matcher matcher, int i, CharSequence seq) {
3158 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to)
3159 return false;
3160 matcher.last = i;
3161 matcher.groups[0] = matcher.first;
3162 matcher.groups[1] = matcher.last;
3163 return true;
3164 }
3165 }
3166
3167 /**
3168 * Used for REs that can start anywhere within the input string.
3169 * This basically tries to match repeatedly at each spot in the
3170 * input string, moving forward after each try. An anchored search
3171 * or a BnM will bypass this node completely.
3172 */
3173 static class Start extends Node {
3174 int minLength;
3175 Start(Node node) {
3176 this.next = node;
3177 TreeInfo info = new TreeInfo();
3178 next.study(info);
3179 minLength = info.minLength;
3180 }
3181 boolean match(Matcher matcher, int i, CharSequence seq) {
3182 if (i > matcher.to - minLength) {
3183 matcher.hitEnd = true;
3184 return false;
3185 }
3186 int guard = matcher.to - minLength;
3187 for (; i <= guard; i++) {
3188 if (next.match(matcher, i, seq)) {
3189 matcher.first = i;
3190 matcher.groups[0] = matcher.first;
3191 matcher.groups[1] = matcher.last;
3192 return true;
3193 }
3194 }
3195 matcher.hitEnd = true;
3196 return false;
3197 }
3198 boolean study(TreeInfo info) {
3199 next.study(info);
3200 info.maxValid = false;
3201 info.deterministic = false;
3202 return false;
3203 }
3204 }
3205
3206 /*
3207 * StartS supports supplementary characters, including unpaired surrogates.
3208 */
3209 static final class StartS extends Start {
3210 StartS(Node node) {
3211 super(node);
3212 }
3213 boolean match(Matcher matcher, int i, CharSequence seq) {
3214 if (i > matcher.to - minLength) {
3215 matcher.hitEnd = true;
3216 return false;
3217 }
3218 int guard = matcher.to - minLength;
3219 while (i <= guard) {
3220 //if ((ret = next.match(matcher, i, seq)) || i == guard)
3221 if (next.match(matcher, i, seq)) {
3222 matcher.first = i;
3223 matcher.groups[0] = matcher.first;
3224 matcher.groups[1] = matcher.last;
3225 return true;
3226 }
3227 if (i == guard)
3228 break;
3229 // Optimization to move to the next character. This is
3230 // faster than countChars(seq, i, 1).
3231 if (Character.isHighSurrogate(seq.charAt(i++))) {
3232 if (i < seq.length() &&
3233 Character.isLowSurrogate(seq.charAt(i))) {
3234 i++;
3235 }
3236 }
3237 }
3238 matcher.hitEnd = true;
3239 return false;
3240 }
3241 }
3242
3243 /**
3244 * Node to anchor at the beginning of input. This object implements the
3245 * match for a \A sequence, and the caret anchor will use this if not in
3246 * multiline mode.
3247 */
3248 static final class Begin extends Node {
3249 boolean match(Matcher matcher, int i, CharSequence seq) {
3250 int fromIndex = (matcher.anchoringBounds) ?
3251 matcher.from : 0;
3252 if (i == fromIndex && next.match(matcher, i, seq)) {
3253 matcher.first = i;
3254 matcher.groups[0] = i;
3255 matcher.groups[1] = matcher.last;
3256 return true;
3257 } else {
3258 return false;
3259 }
3260 }
3261 }
3262
3263 /**
3264 * Node to anchor at the end of input. This is the absolute end, so this
3265 * should not match at the last newline before the end as $ will.
3266 */
3267 static final class End extends Node {
3268 boolean match(Matcher matcher, int i, CharSequence seq) {
3269 int endIndex = (matcher.anchoringBounds) ?
3270 matcher.to : matcher.getTextLength();
3271 if (i == endIndex) {
3272 matcher.hitEnd = true;
3273 return next.match(matcher, i, seq);
3274 }
3275 return false;
3276 }
3277 }
3278
3279 /**
3280 * Node to anchor at the beginning of a line. This is essentially the
3281 * object to match for the multiline ^.
3282 */
3283 static final class Caret extends Node {
3284 boolean match(Matcher matcher, int i, CharSequence seq) {
3285 int startIndex = matcher.from;
3286 int endIndex = matcher.to;
3287 if (!matcher.anchoringBounds) {
3288 startIndex = 0;
3289 endIndex = matcher.getTextLength();
3290 }
3291 // Perl does not match ^ at end of input even after newline
3292 if (i == endIndex) {
3293 matcher.hitEnd = true;
3294 return false;
3295 }
3296 if (i > startIndex) {
3297 char ch = seq.charAt(i-1);
3298 if (ch != '\n' && ch != '\r'
3299 && (ch|1) != '\u2029'
3300 && ch != '\u0085' ) {
3301 return false;
3302 }
3303 // Should treat /r/n as one newline
3304 if (ch == '\r' && seq.charAt(i) == '\n')
3305 return false;
3306 }
3307 return next.match(matcher, i, seq);
3308 }
3309 }
3310
3311 /**
3312 * Node to anchor at the beginning of a line when in unixdot mode.
3313 */
3314 static final class UnixCaret extends Node {
3315 boolean match(Matcher matcher, int i, CharSequence seq) {
3316 int startIndex = matcher.from;
3317 int endIndex = matcher.to;
3318 if (!matcher.anchoringBounds) {
3319 startIndex = 0;
3320 endIndex = matcher.getTextLength();
3321 }
3322 // Perl does not match ^ at end of input even after newline
3323 if (i == endIndex) {
3324 matcher.hitEnd = true;
3325 return false;
3326 }
3327 if (i > startIndex) {
3328 char ch = seq.charAt(i-1);
3329 if (ch != '\n') {
3330 return false;
3331 }
3332 }
3333 return next.match(matcher, i, seq);
3334 }
3335 }
3336
3337 /**
3338 * Node to match the location where the last match ended.
3339 * This is used for the \G construct.
3340 */
3341 static final class LastMatch extends Node {
3342 boolean match(Matcher matcher, int i, CharSequence seq) {
3343 if (i != matcher.oldLast)
3344 return false;
3345 return next.match(matcher, i, seq);
3346 }
3347 }
3348
3349 /**
3350 * Node to anchor at the end of a line or the end of input based on the
3351 * multiline mode.
3352 *
3353 * When not in multiline mode, the $ can only match at the very end
3354 * of the input, unless the input ends in a line terminator in which
3355 * it matches right before the last line terminator.
3356 *
3357 * Note that \r\n is considered an atomic line terminator.
3358 *
3359 * Like ^ the $ operator matches at a position, it does not match the
3360 * line terminators themselves.
3361 */
3362 static final class Dollar extends Node {
3363 boolean multiline;
3364 Dollar(boolean mul) {
3365 multiline = mul;
3366 }
3367 boolean match(Matcher matcher, int i, CharSequence seq) {
3368 int endIndex = (matcher.anchoringBounds) ?
3369 matcher.to : matcher.getTextLength();
3370 if (!multiline) {
3371 if (i < endIndex - 2)
3372 return false;
3373 if (i == endIndex - 2) {
3374 char ch = seq.charAt(i);
3375 if (ch != '\r')
3376 return false;
3377 ch = seq.charAt(i + 1);
3378 if (ch != '\n')
3379 return false;
3380 }
3381 }
3382 // Matches before any line terminator; also matches at the
3383 // end of input
3384 // Before line terminator:
3385 // If multiline, we match here no matter what
3386 // If not multiline, fall through so that the end
3387 // is marked as hit; this must be a /r/n or a /n
3388 // at the very end so the end was hit; more input
3389 // could make this not match here
3390 if (i < endIndex) {
3391 char ch = seq.charAt(i);
3392 if (ch == '\n') {
3393 // No match between \r\n
3394 if (i > 0 && seq.charAt(i-1) == '\r')
3395 return false;
3396 if (multiline)
3397 return next.match(matcher, i, seq);
3398 } else if (ch == '\r' || ch == '\u0085' ||
3399 (ch|1) == '\u2029') {
3400 if (multiline)
3401 return next.match(matcher, i, seq);
3402 } else { // No line terminator, no match
3403 return false;
3404 }
3405 }
3406 // Matched at current end so hit end
3407 matcher.hitEnd = true;
3408 // If a $ matches because of end of input, then more input
3409 // could cause it to fail!
3410 matcher.requireEnd = true;
3411 return next.match(matcher, i, seq);
3412 }
3413 boolean study(TreeInfo info) {
3414 next.study(info);
3415 return info.deterministic;
3416 }
3417 }
3418
3419 /**
3420 * Node to anchor at the end of a line or the end of input based on the
3421 * multiline mode when in unix lines mode.
3422 */
3423 static final class UnixDollar extends Node {
3424 boolean multiline;
3425 UnixDollar(boolean mul) {
3426 multiline = mul;
3427 }
3428 boolean match(Matcher matcher, int i, CharSequence seq) {
3429 int endIndex = (matcher.anchoringBounds) ?
3430 matcher.to : matcher.getTextLength();
3431 if (i < endIndex) {
3432 char ch = seq.charAt(i);
3433 if (ch == '\n') {
3434 // If not multiline, then only possible to
3435 // match at very end or one before end
3436 if (multiline == false && i != endIndex - 1)
3437 return false;
3438 // If multiline return next.match without setting
3439 // matcher.hitEnd
3440 if (multiline)
3441 return next.match(matcher, i, seq);
3442 } else {
3443 return false;
3444 }
3445 }
3446 // Matching because at the end or 1 before the end;
3447 // more input could change this so set hitEnd
3448 matcher.hitEnd = true;
3449 // If a $ matches because of end of input, then more input
3450 // could cause it to fail!
3451 matcher.requireEnd = true;
3452 return next.match(matcher, i, seq);
3453 }
3454 boolean study(TreeInfo info) {
3455 next.study(info);
3456 return info.deterministic;
3457 }
3458 }
3459
3460 /**
3461 * Abstract node class to match one character satisfying some
3462 * boolean property.
3463 */
3464 private static abstract class CharProperty extends Node {
3465 abstract boolean isSatisfiedBy(int ch);
3466 CharProperty complement() {
3467 return new CharProperty() {
3468 boolean isSatisfiedBy(int ch) {
3469 return ! CharProperty.this.isSatisfiedBy(ch);}};
3470 }
3471 boolean match(Matcher matcher, int i, CharSequence seq) {
3472 if (i < matcher.to) {
3473 int ch = Character.codePointAt(seq, i);
3474 return isSatisfiedBy(ch)
3475 && next.match(matcher, i+Character.charCount(ch), seq);
3476 } else {
3477 matcher.hitEnd = true;
3478 return false;
3479 }
3480 }
3481 boolean study(TreeInfo info) {
3482 info.minLength++;
3483 info.maxLength++;
3484 return next.study(info);
3485 }
3486 }
3487
3488 /**
3489 * Optimized version of CharProperty that works only for
3490 * properties never satisfied by Supplementary characters.
3491 */
3492 private static abstract class BmpCharProperty extends CharProperty {
3493 boolean match(Matcher matcher, int i, CharSequence seq) {
3494 if (i < matcher.to) {
3495 return isSatisfiedBy(seq.charAt(i))
3496 && next.match(matcher, i+1, seq);
3497 } else {
3498 matcher.hitEnd = true;
3499 return false;
3500 }
3501 }
3502 }
3503
3504 /**
3505 * Node class that matches a Supplementary Unicode character
3506 */
3507 static final class SingleS extends CharProperty {
3508 final int c;
3509 SingleS(int c) { this.c = c; }
3510 boolean isSatisfiedBy(int ch) {
3511 return ch == c;
3512 }
3513 }
3514
3515 /**
3516 * Optimization -- matches a given BMP character
3517 */
3518 static final class Single extends BmpCharProperty {
3519 final int c;
3520 Single(int c) { this.c = c; }
3521 boolean isSatisfiedBy(int ch) {
3522 return ch == c;
3523 }
3524 }
3525
3526 /**
3527 * Case insensitive matches a given BMP character
3528 */
3529 static final class SingleI extends BmpCharProperty {
3530 final int lower;
3531 final int upper;
3532 SingleI(int lower, int upper) {
3533 this.lower = lower;
3534 this.upper = upper;
3535 }
3536 boolean isSatisfiedBy(int ch) {
3537 return ch == lower || ch == upper;
3538 }
3539 }
3540
3541 /**
3542 * Unicode case insensitive matches a given Unicode character
3543 */
3544 static final class SingleU extends CharProperty {
3545 final int lower;
3546 SingleU(int lower) {
3547 this.lower = lower;
3548 }
3549 boolean isSatisfiedBy(int ch) {
3550 return lower == ch ||
3551 lower == Character.toLowerCase(Character.toUpperCase(ch));
3552 }
3553 }
3554
3555
3556 /**
3557 * Node class that matches a Unicode block.
3558 */
3559 static final class Block extends CharProperty {
3560 final Character.UnicodeBlock block;
3561 Block(Character.UnicodeBlock block) {
3562 this.block = block;
3563 }
3564 boolean isSatisfiedBy(int ch) {
3565 return block == Character.UnicodeBlock.of(ch);
3566 }
3567 }
3568
3569 /**
3570 * Node class that matches a Unicode category.
3571 */
3572 static final class Category extends CharProperty {
3573 final int typeMask;
3574 Category(int typeMask) { this.typeMask = typeMask; }
3575 boolean isSatisfiedBy(int ch) {
3576 return (typeMask & (1 << Character.getType(ch))) != 0;
3577 }
3578 }
3579
3580 /**
3581 * Node class that matches a POSIX type.
3582 */
3583 static final class Ctype extends BmpCharProperty {
3584 final int ctype;
3585 Ctype(int ctype) { this.ctype = ctype; }
3586 boolean isSatisfiedBy(int ch) {
3587 return ch < 128 && ASCII.isType(ch, ctype);
3588 }
3589 }
3590
3591 /**
3592 * Base class for all Slice nodes
3593 */
3594 static class SliceNode extends Node {
3595 int[] buffer;
3596 SliceNode(int[] buf) {
3597 buffer = buf;
3598 }
3599 boolean study(TreeInfo info) {
3600 info.minLength += buffer.length;
3601 info.maxLength += buffer.length;
3602 return next.study(info);
3603 }
3604 }
3605
3606 /**
3607 * Node class for a case sensitive/BMP-only sequence of literal
3608 * characters.
3609 */
3610 static final class Slice extends SliceNode {
3611 Slice(int[] buf) {
3612 super(buf);
3613 }
3614 boolean match(Matcher matcher, int i, CharSequence seq) {
3615 int[] buf = buffer;
3616 int len = buf.length;
3617 for (int j=0; j<len; j++) {
3618 if ((i+j) >= matcher.to) {
3619 matcher.hitEnd = true;
3620 return false;
3621 }
3622 if (buf[j] != seq.charAt(i+j))
3623 return false;
3624 }
3625 return next.match(matcher, i+len, seq);
3626 }
3627 }
3628
3629 /**
3630 * Node class for a case_insensitive/BMP-only sequence of literal
3631 * characters.
3632 */
3633 static class SliceI extends SliceNode {
3634 SliceI(int[] buf) {
3635 super(buf);
3636 }
3637 boolean match(Matcher matcher, int i, CharSequence seq) {
3638 int[] buf = buffer;
3639 int len = buf.length;
3640 for (int j=0; j<len; j++) {
3641 if ((i+j) >= matcher.to) {
3642 matcher.hitEnd = true;
3643 return false;
3644 }
3645 int c = seq.charAt(i+j);
3646 if (buf[j] != c &&
3647 buf[j] != ASCII.toLower(c))
3648 return false;
3649 }
3650 return next.match(matcher, i+len, seq);
3651 }
3652 }
3653
3654 /**
3655 * Node class for a unicode_case_insensitive/BMP-only sequence of
3656 * literal characters. Uses unicode case folding.
3657 */
3658 static final class SliceU extends SliceNode {
3659 SliceU(int[] buf) {
3660 super(buf);
3661 }
3662 boolean match(Matcher matcher, int i, CharSequence seq) {
3663 int[] buf = buffer;
3664 int len = buf.length;
3665 for (int j=0; j<len; j++) {
3666 if ((i+j) >= matcher.to) {
3667 matcher.hitEnd = true;
3668 return false;
3669 }
3670 int c = seq.charAt(i+j);
3671 if (buf[j] != c &&
3672 buf[j] != Character.toLowerCase(Character.toUpperCase(c)))
3673 return false;
3674 }
3675 return next.match(matcher, i+len, seq);
3676 }
3677 }
3678
3679 /**
3680 * Node class for a case sensitive sequence of literal characters
3681 * including supplementary characters.
3682 */
3683 static final class SliceS extends SliceNode {
3684 SliceS(int[] buf) {
3685 super(buf);
3686 }
3687 boolean match(Matcher matcher, int i, CharSequence seq) {
3688 int[] buf = buffer;
3689 int x = i;
3690 for (int j = 0; j < buf.length; j++) {
3691 if (x >= matcher.to) {
3692 matcher.hitEnd = true;
3693 return false;
3694 }
3695 int c = Character.codePointAt(seq, x);
3696 if (buf[j] != c)
3697 return false;
3698 x += Character.charCount(c);
3699 if (x > matcher.to) {
3700 matcher.hitEnd = true;
3701 return false;
3702 }
3703 }
3704 return next.match(matcher, x, seq);
3705 }
3706 }
3707
3708 /**
3709 * Node class for a case insensitive sequence of literal characters
3710 * including supplementary characters.
3711 */
3712 static class SliceIS extends SliceNode {
3713 SliceIS(int[] buf) {
3714 super(buf);
3715 }
3716 int toLower(int c) {
3717 return ASCII.toLower(c);
3718 }
3719 boolean match(Matcher matcher, int i, CharSequence seq) {
3720 int[] buf = buffer;
3721 int x = i;
3722 for (int j = 0; j < buf.length; j++) {
3723 if (x >= matcher.to) {
3724 matcher.hitEnd = true;
3725 return false;
3726 }
3727 int c = Character.codePointAt(seq, x);
3728 if (buf[j] != c && buf[j] != toLower(c))
3729 return false;
3730 x += Character.charCount(c);
3731 if (x > matcher.to) {
3732 matcher.hitEnd = true;
3733 return false;
3734 }
3735 }
3736 return next.match(matcher, x, seq);
3737 }
3738 }
3739
3740 /**
3741 * Node class for a case insensitive sequence of literal characters.
3742 * Uses unicode case folding.
3743 */
3744 static final class SliceUS extends SliceIS {
3745 SliceUS(int[] buf) {
3746 super(buf);
3747 }
3748 int toLower(int c) {
3749 return Character.toLowerCase(Character.toUpperCase(c));
3750 }
3751 }
3752
3753 private static boolean inRange(int lower, int ch, int upper) {
3754 return lower <= ch && ch <= upper;
3755 }
3756
3757 /**
3758 * Returns node for matching characters within an explicit value range.
3759 */
3760 private static CharProperty rangeFor(final int lower,
3761 final int upper) {
3762 return new CharProperty() {
3763 boolean isSatisfiedBy(int ch) {
3764 return inRange(lower, ch, upper);}};
3765 }
3766
3767 /**
3768 * Returns node for matching characters within an explicit value
3769 * range in a case insensitive manner.
3770 */
3771 private CharProperty caseInsensitiveRangeFor(final int lower,
3772 final int upper) {
3773 if (has(UNICODE_CASE))
3774 return new CharProperty() {
3775 boolean isSatisfiedBy(int ch) {
3776 if (inRange(lower, ch, upper))
3777 return true;
3778 int up = Character.toUpperCase(ch);
3779 return inRange(lower, up, upper) ||
3780 inRange(lower, Character.toLowerCase(up), upper);}};
3781 return new CharProperty() {
3782 boolean isSatisfiedBy(int ch) {
3783 return inRange(lower, ch, upper) ||
3784 ASCII.isAscii(ch) &&
3785 (inRange(lower, ASCII.toUpper(ch), upper) ||
3786 inRange(lower, ASCII.toLower(ch), upper));
3787 }};
3788 }
3789
3790 /**
3791 * Implements the Unicode category ALL and the dot metacharacter when
3792 * in dotall mode.
3793 */
3794 static final class All extends CharProperty {
3795 boolean isSatisfiedBy(int ch) {
3796 return true;
3797 }
3798 }
3799
3800 /**
3801 * Node class for the dot metacharacter when dotall is not enabled.
3802 */
3803 static final class Dot extends CharProperty {
3804 boolean isSatisfiedBy(int ch) {
3805 return (ch != '\n' && ch != '\r'
3806 && (ch|1) != '\u2029'
3807 && ch != '\u0085');
3808 }
3809 }
3810
3811 /**
3812 * Node class for the dot metacharacter when dotall is not enabled
3813 * but UNIX_LINES is enabled.
3814 */
3815 static final class UnixDot extends CharProperty {
3816 boolean isSatisfiedBy(int ch) {
3817 return ch != '\n';
3818 }
3819 }
3820
3821 /**
3822 * The 0 or 1 quantifier. This one class implements all three types.
3823 */
3824 static final class Ques extends Node {
3825 Node atom;
3826 int type;
3827 Ques(Node node, int type) {
3828 this.atom = node;
3829 this.type = type;
3830 }
3831 boolean match(Matcher matcher, int i, CharSequence seq) {
3832 switch (type) {
3833 case GREEDY:
3834 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq))
3835 || next.match(matcher, i, seq);
3836 case LAZY:
3837 return next.match(matcher, i, seq)
3838 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq));
3839 case POSSESSIVE:
3840 if (atom.match(matcher, i, seq)) i = matcher.last;
3841 return next.match(matcher, i, seq);
3842 default:
3843 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq);
3844 }
3845 }
3846 boolean study(TreeInfo info) {
3847 if (type != INDEPENDENT) {
3848 int minL = info.minLength;
3849 atom.study(info);
3850 info.minLength = minL;
3851 info.deterministic = false;
3852 return next.study(info);
3853 } else {
3854 atom.study(info);
3855 return next.study(info);
3856 }
3857 }
3858 }
3859
3860 /**
3861 * Handles the curly-brace style repetition with a specified minimum and
3862 * maximum occurrences. The * quantifier is handled as a special case.
3863 * This class handles the three types.
3864 */
3865 static final class Curly extends Node {
3866 Node atom;
3867 int type;
3868 int cmin;
3869 int cmax;
3870
3871 Curly(Node node, int cmin, int cmax, int type) {
3872 this.atom = node;
3873 this.type = type;
3874 this.cmin = cmin;
3875 this.cmax = cmax;
3876 }
3877 boolean match(Matcher matcher, int i, CharSequence seq) {
3878 int j;
3879 for (j = 0; j < cmin; j++) {
3880 if (atom.match(matcher, i, seq)) {
3881 i = matcher.last;
3882 continue;
3883 }
3884 return false;
3885 }
3886 if (type == GREEDY)
3887 return match0(matcher, i, j, seq);
3888 else if (type == LAZY)
3889 return match1(matcher, i, j, seq);
3890 else
3891 return match2(matcher, i, j, seq);
3892 }
3893 // Greedy match.
3894 // i is the index to start matching at
3895 // j is the number of atoms that have matched
3896 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
3897 if (j >= cmax) {
3898 // We have matched the maximum... continue with the rest of
3899 // the regular expression
3900 return next.match(matcher, i, seq);
3901 }
3902 int backLimit = j;
3903 while (atom.match(matcher, i, seq)) {
3904 // k is the length of this match
3905 int k = matcher.last - i;
3906 if (k == 0) // Zero length match
3907 break;
3908 // Move up index and number matched
3909 i = matcher.last;
3910 j++;
3911 // We are greedy so match as many as we can
3912 while (j < cmax) {
3913 if (!atom.match(matcher, i, seq))
3914 break;
3915 if (i + k != matcher.last) {
3916 if (match0(matcher, matcher.last, j+1, seq))
3917 return true;
3918 break;
3919 }
3920 i += k;
3921 j++;
3922 }
3923 // Handle backing off if match fails
3924 while (j >= backLimit) {
3925 if (next.match(matcher, i, seq))
3926 return true;
3927 i -= k;
3928 j--;
3929 }
3930 return false;
3931 }
3932 return next.match(matcher, i, seq);
3933 }
3934 // Reluctant match. At this point, the minimum has been satisfied.
3935 // i is the index to start matching at
3936 // j is the number of atoms that have matched
3937 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
3938 for (;;) {
3939 // Try finishing match without consuming any more
3940 if (next.match(matcher, i, seq))
3941 return true;
3942 // At the maximum, no match found
3943 if (j >= cmax)
3944 return false;
3945 // Okay, must try one more atom
3946 if (!atom.match(matcher, i, seq))
3947 return false;
3948 // If we haven't moved forward then must break out
3949 if (i == matcher.last)
3950 return false;
3951 // Move up index and number matched
3952 i = matcher.last;
3953 j++;
3954 }
3955 }
3956 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
3957 for (; j < cmax; j++) {
3958 if (!atom.match(matcher, i, seq))
3959 break;
3960 if (i == matcher.last)
3961 break;
3962 i = matcher.last;
3963 }
3964 return next.match(matcher, i, seq);
3965 }
3966 boolean study(TreeInfo info) {
3967 // Save original info
3968 int minL = info.minLength;
3969 int maxL = info.maxLength;
3970 boolean maxV = info.maxValid;
3971 boolean detm = info.deterministic;
3972 info.reset();
3973
3974 atom.study(info);
3975
3976 int temp = info.minLength * cmin + minL;
3977 if (temp < minL) {
3978 temp = 0xFFFFFFF; // arbitrary large number
3979 }
3980 info.minLength = temp;
3981
3982 if (maxV & info.maxValid) {
3983 temp = info.maxLength * cmax + maxL;
3984 info.maxLength = temp;
3985 if (temp < maxL) {
3986 info.maxValid = false;
3987 }
3988 } else {
3989 info.maxValid = false;
3990 }
3991
3992 if (info.deterministic && cmin == cmax)
3993 info.deterministic = detm;
3994 else
3995 info.deterministic = false;
3996
3997 return next.study(info);
3998 }
3999 }
4000
4001 /**
4002 * Handles the curly-brace style repetition with a specified minimum and
4003 * maximum occurrences in deterministic cases. This is an iterative
4004 * optimization over the Prolog and Loop system which would handle this
4005 * in a recursive way. The * quantifier is handled as a special case.
4006 * If capture is true then this class saves group settings and ensures
4007 * that groups are unset when backing off of a group match.
4008 */
4009 static final class GroupCurly extends Node {
4010 Node atom;
4011 int type;
4012 int cmin;
4013 int cmax;
4014 int localIndex;
4015 int groupIndex;
4016 boolean capture;
4017
4018 GroupCurly(Node node, int cmin, int cmax, int type, int local,
4019 int group, boolean capture) {
4020 this.atom = node;
4021 this.type = type;
4022 this.cmin = cmin;
4023 this.cmax = cmax;
4024 this.localIndex = local;
4025 this.groupIndex = group;
4026 this.capture = capture;
4027 }
4028 boolean match(Matcher matcher, int i, CharSequence seq) {
4029 int[] groups = matcher.groups;
4030 int[] locals = matcher.locals;
4031 int save0 = locals[localIndex];
4032 int save1 = 0;
4033 int save2 = 0;
4034
4035 if (capture) {
4036 save1 = groups[groupIndex];
4037 save2 = groups[groupIndex+1];
4038 }
4039
4040 // Notify GroupTail there is no need to setup group info
4041 // because it will be set here
4042 locals[localIndex] = -1;
4043
4044 boolean ret = true;
4045 for (int j = 0; j < cmin; j++) {
4046 if (atom.match(matcher, i, seq)) {
4047 if (capture) {
4048 groups[groupIndex] = i;
4049 groups[groupIndex+1] = matcher.last;
4050 }
4051 i = matcher.last;
4052 } else {
4053 ret = false;
4054 break;
4055 }
4056 }
4057 if (ret) {
4058 if (type == GREEDY) {
4059 ret = match0(matcher, i, cmin, seq);
4060 } else if (type == LAZY) {
4061 ret = match1(matcher, i, cmin, seq);
4062 } else {
4063 ret = match2(matcher, i, cmin, seq);
4064 }
4065 }
4066 if (!ret) {
4067 locals[localIndex] = save0;
4068 if (capture) {
4069 groups[groupIndex] = save1;
4070 groups[groupIndex+1] = save2;
4071 }
4072 }
4073 return ret;
4074 }
4075 // Aggressive group match
4076 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4077 int[] groups = matcher.groups;
4078 int save0 = 0;
4079 int save1 = 0;
4080 if (capture) {
4081 save0 = groups[groupIndex];
4082 save1 = groups[groupIndex+1];
4083 }
4084 for (;;) {
4085 if (j >= cmax)
4086 break;
4087 if (!atom.match(matcher, i, seq))
4088 break;
4089 int k = matcher.last - i;
4090 if (k <= 0) {
4091 if (capture) {
4092 groups[groupIndex] = i;
4093 groups[groupIndex+1] = i + k;
4094 }
4095 i = i + k;
4096 break;
4097 }
4098 for (;;) {
4099 if (capture) {
4100 groups[groupIndex] = i;
4101 groups[groupIndex+1] = i + k;
4102 }
4103 i = i + k;
4104 if (++j >= cmax)
4105 break;
4106 if (!atom.match(matcher, i, seq))
4107 break;
4108 if (i + k != matcher.last) {
4109 if (match0(matcher, i, j, seq))
4110 return true;
4111 break;
4112 }
4113 }
4114 while (j > cmin) {
4115 if (next.match(matcher, i, seq)) {
4116 if (capture) {
4117 groups[groupIndex+1] = i;
4118 groups[groupIndex] = i - k;
4119 }
4120 i = i - k;
4121 return true;
4122 }
4123 // backing off
4124 if (capture) {
4125 groups[groupIndex+1] = i;
4126 groups[groupIndex] = i - k;
4127 }
4128 i = i - k;
4129 j--;
4130 }
4131 break;
4132 }
4133 if (capture) {
4134 groups[groupIndex] = save0;
4135 groups[groupIndex+1] = save1;
4136 }
4137 return next.match(matcher, i, seq);
4138 }
4139 // Reluctant matching
4140 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4141 for (;;) {
4142 if (next.match(matcher, i, seq))
4143 return true;
4144 if (j >= cmax)
4145 return false;
4146 if (!atom.match(matcher, i, seq))
4147 return false;
4148 if (i == matcher.last)
4149 return false;
4150 if (capture) {
4151 matcher.groups[groupIndex] = i;
4152 matcher.groups[groupIndex+1] = matcher.last;
4153 }
4154 i = matcher.last;
4155 j++;
4156 }
4157 }
4158 // Possessive matching
4159 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4160 for (; j < cmax; j++) {
4161 if (!atom.match(matcher, i, seq)) {
4162 break;
4163 }
4164 if (capture) {
4165 matcher.groups[groupIndex] = i;
4166 matcher.groups[groupIndex+1] = matcher.last;
4167 }
4168 if (i == matcher.last) {
4169 break;
4170 }
4171 i = matcher.last;
4172 }
4173 return next.match(matcher, i, seq);
4174 }
4175 boolean study(TreeInfo info) {
4176 // Save original info
4177 int minL = info.minLength;
4178 int maxL = info.maxLength;
4179 boolean maxV = info.maxValid;
4180 boolean detm = info.deterministic;
4181 info.reset();
4182
4183 atom.study(info);
4184
4185 int temp = info.minLength * cmin + minL;
4186 if (temp < minL) {
4187 temp = 0xFFFFFFF; // Arbitrary large number
4188 }
4189 info.minLength = temp;
4190
4191 if (maxV & info.maxValid) {
4192 temp = info.maxLength * cmax + maxL;
4193 info.maxLength = temp;
4194 if (temp < maxL) {
4195 info.maxValid = false;
4196 }
4197 } else {
4198 info.maxValid = false;
4199 }
4200
4201 if (info.deterministic && cmin == cmax) {
4202 info.deterministic = detm;
4203 } else {
4204 info.deterministic = false;
4205 }
4206
4207 return next.study(info);
4208 }
4209 }
4210
4211 /**
4212 * A Guard node at the end of each atom node in a Branch. It
4213 * serves the purpose of chaining the "match" operation to
4214 * "next" but not the "study", so we can collect the TreeInfo
4215 * of each atom node without including the TreeInfo of the
4216 * "next".
4217 */
4218 static final class BranchConn extends Node {
4219 BranchConn() {};
4220 boolean match(Matcher matcher, int i, CharSequence seq) {
4221 return next.match(matcher, i, seq);
4222 }
4223 boolean study(TreeInfo info) {
4224 return info.deterministic;
4225 }
4226 }
4227
4228 /**
4229 * Handles the branching of alternations. Note this is also used for
4230 * the ? quantifier to branch between the case where it matches once
4231 * and where it does not occur.
4232 */
4233 static final class Branch extends Node {
4234 Node[] atoms = new Node[2];
4235 int size = 2;
4236 Node conn;
4237 Branch(Node first, Node second, Node branchConn) {
4238 conn = branchConn;
4239 atoms[0] = first;
4240 atoms[1] = second;
4241 }
4242
4243 void add(Node node) {
4244 if (size >= atoms.length) {
4245 Node[] tmp = new Node[atoms.length*2];
4246 System.arraycopy(atoms, 0, tmp, 0, atoms.length);
4247 atoms = tmp;
4248 }
4249 atoms[size++] = node;
4250 }
4251
4252 boolean match(Matcher matcher, int i, CharSequence seq) {
4253 for (int n = 0; n < size; n++) {
4254 if (atoms[n] == null) {
4255 if (conn.next.match(matcher, i, seq))
4256 return true;
4257 } else if (atoms[n].match(matcher, i, seq)) {
4258 return true;
4259 }
4260 }
4261 return false;
4262 }
4263
4264 boolean study(TreeInfo info) {
4265 int minL = info.minLength;
4266 int maxL = info.maxLength;
4267 boolean maxV = info.maxValid;
4268
4269 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num
4270 int maxL2 = -1;
4271 for (int n = 0; n < size; n++) {
4272 info.reset();
4273 if (atoms[n] != null)
4274 atoms[n].study(info);
4275 minL2 = Math.min(minL2, info.minLength);
4276 maxL2 = Math.max(maxL2, info.maxLength);
4277 maxV = (maxV & info.maxValid);
4278 }
4279
4280 minL += minL2;
4281 maxL += maxL2;
4282
4283 info.reset();
4284 conn.next.study(info);
4285
4286 info.minLength += minL;
4287 info.maxLength += maxL;
4288 info.maxValid &= maxV;
4289 info.deterministic = false;
4290 return false;
4291 }
4292 }
4293
4294 /**
4295 * The GroupHead saves the location where the group begins in the locals
4296 * and restores them when the match is done.
4297 *
4298 * The matchRef is used when a reference to this group is accessed later
4299 * in the expression. The locals will have a negative value in them to
4300 * indicate that we do not want to unset the group if the reference
4301 * doesn't match.
4302 */
4303 static final class GroupHead extends Node {
4304 int localIndex;
4305 GroupHead(int localCount) {
4306 localIndex = localCount;
4307 }
4308 boolean match(Matcher matcher, int i, CharSequence seq) {
4309 int save = matcher.locals[localIndex];
4310 matcher.locals[localIndex] = i;
4311 boolean ret = next.match(matcher, i, seq);
4312 matcher.locals[localIndex] = save;
4313 return ret;
4314 }
4315 boolean matchRef(Matcher matcher, int i, CharSequence seq) {
4316 int save = matcher.locals[localIndex];
4317 matcher.locals[localIndex] = ~i; // HACK
4318 boolean ret = next.match(matcher, i, seq);
4319 matcher.locals[localIndex] = save;
4320 return ret;
4321 }
4322 }
4323
4324 /**
4325 * Recursive reference to a group in the regular expression. It calls
4326 * matchRef because if the reference fails to match we would not unset
4327 * the group.
4328 */
4329 static final class GroupRef extends Node {
4330 GroupHead head;
4331 GroupRef(GroupHead head) {
4332 this.head = head;
4333 }
4334 boolean match(Matcher matcher, int i, CharSequence seq) {
4335 return head.matchRef(matcher, i, seq)
4336 && next.match(matcher, matcher.last, seq);
4337 }
4338 boolean study(TreeInfo info) {
4339 info.maxValid = false;
4340 info.deterministic = false;
4341 return next.study(info);
4342 }
4343 }
4344
4345 /**
4346 * The GroupTail handles the setting of group beginning and ending
4347 * locations when groups are successfully matched. It must also be able to
4348 * unset groups that have to be backed off of.
4349 *
4350 * The GroupTail node is also used when a previous group is referenced,
4351 * and in that case no group information needs to be set.
4352 */
4353 static final class GroupTail extends Node {
4354 int localIndex;
4355 int groupIndex;
4356 GroupTail(int localCount, int groupCount) {
4357 localIndex = localCount;
4358 groupIndex = groupCount + groupCount;
4359 }
4360 boolean match(Matcher matcher, int i, CharSequence seq) {
4361 int tmp = matcher.locals[localIndex];
4362 if (tmp >= 0) { // This is the normal group case.
4363 // Save the group so we can unset it if it
4364 // backs off of a match.
4365 int groupStart = matcher.groups[groupIndex];
4366 int groupEnd = matcher.groups[groupIndex+1];
4367
4368 matcher.groups[groupIndex] = tmp;
4369 matcher.groups[groupIndex+1] = i;
4370 if (next.match(matcher, i, seq)) {
4371 return true;
4372 }
4373 matcher.groups[groupIndex] = groupStart;
4374 matcher.groups[groupIndex+1] = groupEnd;
4375 return false;
4376 } else {
4377 // This is a group reference case. We don't need to save any
4378 // group info because it isn't really a group.
4379 matcher.last = i;
4380 return true;
4381 }
4382 }
4383 }
4384
4385 /**
4386 * This sets up a loop to handle a recursive quantifier structure.
4387 */
4388 static final class Prolog extends Node {
4389 Loop loop;
4390 Prolog(Loop loop) {
4391 this.loop = loop;
4392 }
4393 boolean match(Matcher matcher, int i, CharSequence seq) {
4394 return loop.matchInit(matcher, i, seq);
4395 }
4396 boolean study(TreeInfo info) {
4397 return loop.study(info);
4398 }
4399 }
4400
4401 /**
4402 * Handles the repetition count for a greedy Curly. The matchInit
4403 * is called from the Prolog to save the index of where the group
4404 * beginning is stored. A zero length group check occurs in the
4405 * normal match but is skipped in the matchInit.
4406 */
4407 static class Loop extends Node {
4408 Node body;
4409 int countIndex; // local count index in matcher locals
4410 int beginIndex; // group beginning index
4411 int cmin, cmax;
4412 Loop(int countIndex, int beginIndex) {
4413 this.countIndex = countIndex;
4414 this.beginIndex = beginIndex;
4415 }
4416 boolean match(Matcher matcher, int i, CharSequence seq) {
4417 // Avoid infinite loop in zero-length case.
4418 if (i > matcher.locals[beginIndex]) {
4419 int count = matcher.locals[countIndex];
4420
4421 // This block is for before we reach the minimum
4422 // iterations required for the loop to match
4423 if (count < cmin) {
4424 matcher.locals[countIndex] = count + 1;
4425 boolean b = body.match(matcher, i, seq);
4426 // If match failed we must backtrack, so
4427 // the loop count should NOT be incremented
4428 if (!b)
4429 matcher.locals[countIndex] = count;
4430 // Return success or failure since we are under
4431 // minimum
4432 return b;
4433 }
4434 // This block is for after we have the minimum
4435 // iterations required for the loop to match
4436 if (count < cmax) {
4437 matcher.locals[countIndex] = count + 1;
4438 boolean b = body.match(matcher, i, seq);
4439 // If match failed we must backtrack, so
4440 // the loop count should NOT be incremented
4441 if (!b)
4442 matcher.locals[countIndex] = count;
4443 else
4444 return true;
4445 }
4446 }
4447 return next.match(matcher, i, seq);
4448 }
4449 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4450 int save = matcher.locals[countIndex];
4451 boolean ret = false;
4452 if (0 < cmin) {
4453 matcher.locals[countIndex] = 1;
4454 ret = body.match(matcher, i, seq);
4455 } else if (0 < cmax) {
4456 matcher.locals[countIndex] = 1;
4457 ret = body.match(matcher, i, seq);
4458 if (ret == false)
4459 ret = next.match(matcher, i, seq);
4460 } else {
4461 ret = next.match(matcher, i, seq);
4462 }
4463 matcher.locals[countIndex] = save;
4464 return ret;
4465 }
4466 boolean study(TreeInfo info) {
4467 info.maxValid = false;
4468 info.deterministic = false;
4469 return false;
4470 }
4471 }
4472
4473 /**
4474 * Handles the repetition count for a reluctant Curly. The matchInit
4475 * is called from the Prolog to save the index of where the group
4476 * beginning is stored. A zero length group check occurs in the
4477 * normal match but is skipped in the matchInit.
4478 */
4479 static final class LazyLoop extends Loop {
4480 LazyLoop(int countIndex, int beginIndex) {
4481 super(countIndex, beginIndex);
4482 }
4483 boolean match(Matcher matcher, int i, CharSequence seq) {
4484 // Check for zero length group
4485 if (i > matcher.locals[beginIndex]) {
4486 int count = matcher.locals[countIndex];
4487 if (count < cmin) {
4488 matcher.locals[countIndex] = count + 1;
4489 boolean result = body.match(matcher, i, seq);
4490 // If match failed we must backtrack, so
4491 // the loop count should NOT be incremented
4492 if (!result)
4493 matcher.locals[countIndex] = count;
4494 return result;
4495 }
4496 if (next.match(matcher, i, seq))
4497 return true;
4498 if (count < cmax) {
4499 matcher.locals[countIndex] = count + 1;
4500 boolean result = body.match(matcher, i, seq);
4501 // If match failed we must backtrack, so
4502 // the loop count should NOT be incremented
4503 if (!result)
4504 matcher.locals[countIndex] = count;
4505 return result;
4506 }
4507 return false;
4508 }
4509 return next.match(matcher, i, seq);
4510 }
4511 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4512 int save = matcher.locals[countIndex];
4513 boolean ret = false;
4514 if (0 < cmin) {
4515 matcher.locals[countIndex] = 1;
4516 ret = body.match(matcher, i, seq);
4517 } else if (next.match(matcher, i, seq)) {
4518 ret = true;
4519 } else if (0 < cmax) {
4520 matcher.locals[countIndex] = 1;
4521 ret = body.match(matcher, i, seq);
4522 }
4523 matcher.locals[countIndex] = save;
4524 return ret;
4525 }
4526 boolean study(TreeInfo info) {
4527 info.maxValid = false;
4528 info.deterministic = false;
4529 return false;
4530 }
4531 }
4532
4533 /**
4534 * Refers to a group in the regular expression. Attempts to match
4535 * whatever the group referred to last matched.
4536 */
4537 static class BackRef extends Node {
4538 int groupIndex;
4539 BackRef(int groupCount) {
4540 super();
4541 groupIndex = groupCount + groupCount;
4542 }
4543 boolean match(Matcher matcher, int i, CharSequence seq) {
4544 int j = matcher.groups[groupIndex];
4545 int k = matcher.groups[groupIndex+1];
4546
4547 int groupSize = k - j;
4548
4549 // If the referenced group didn't match, neither can this
4550 if (j < 0)
4551 return false;
4552
4553 // If there isn't enough input left no match
4554 if (i + groupSize > matcher.to) {
4555 matcher.hitEnd = true;
4556 return false;
4557 }
4558
4559 // Check each new char to make sure it matches what the group
4560 // referenced matched last time around
4561 for (int index=0; index<groupSize; index++)
4562 if (seq.charAt(i+index) != seq.charAt(j+index))
4563 return false;
4564
4565 return next.match(matcher, i+groupSize, seq);
4566 }
4567 boolean study(TreeInfo info) {
4568 info.maxValid = false;
4569 return next.study(info);
4570 }
4571 }
4572
4573 static class CIBackRef extends Node {
4574 int groupIndex;
4575 boolean doUnicodeCase;
4576 CIBackRef(int groupCount, boolean doUnicodeCase) {
4577 super();
4578 groupIndex = groupCount + groupCount;
4579 this.doUnicodeCase = doUnicodeCase;
4580 }
4581 boolean match(Matcher matcher, int i, CharSequence seq) {
4582 int j = matcher.groups[groupIndex];
4583 int k = matcher.groups[groupIndex+1];
4584
4585 int groupSize = k - j;
4586
4587 // If the referenced group didn't match, neither can this
4588 if (j < 0)
4589 return false;
4590
4591 // If there isn't enough input left no match
4592 if (i + groupSize > matcher.to) {
4593 matcher.hitEnd = true;
4594 return false;
4595 }
4596
4597 // Check each new char to make sure it matches what the group
4598 // referenced matched last time around
4599 int x = i;
4600 for (int index=0; index<groupSize; index++) {
4601 int c1 = Character.codePointAt(seq, x);
4602 int c2 = Character.codePointAt(seq, j);
4603 if (c1 != c2) {
4604 if (doUnicodeCase) {
4605 int cc1 = Character.toUpperCase(c1);
4606 int cc2 = Character.toUpperCase(c2);
4607 if (cc1 != cc2 &&
4608 Character.toLowerCase(cc1) !=
4609 Character.toLowerCase(cc2))
4610 return false;
4611 } else {
4612 if (ASCII.toLower(c1) != ASCII.toLower(c2))
4613 return false;
4614 }
4615 }
4616 x += Character.charCount(c1);
4617 j += Character.charCount(c2);
4618 }
4619
4620 return next.match(matcher, i+groupSize, seq);
4621 }
4622 boolean study(TreeInfo info) {
4623 info.maxValid = false;
4624 return next.study(info);
4625 }
4626 }
4627
4628 /**
4629 * Searches until the next instance of its atom. This is useful for
4630 * finding the atom efficiently without passing an instance of it
4631 * (greedy problem) and without a lot of wasted search time (reluctant
4632 * problem).
4633 */
4634 static final class First extends Node {
4635 Node atom;
4636 First(Node node) {
4637 this.atom = BnM.optimize(node);
4638 }
4639 boolean match(Matcher matcher, int i, CharSequence seq) {
4640 if (atom instanceof BnM) {
4641 return atom.match(matcher, i, seq)
4642 && next.match(matcher, matcher.last, seq);
4643 }
4644 for (;;) {
4645 if (i > matcher.to) {
4646 matcher.hitEnd = true;
4647 return false;
4648 }
4649 if (atom.match(matcher, i, seq)) {
4650 return next.match(matcher, matcher.last, seq);
4651 }
4652 i += countChars(seq, i, 1);
4653 matcher.first++;
4654 }
4655 }
4656 boolean study(TreeInfo info) {
4657 atom.study(info);
4658 info.maxValid = false;
4659 info.deterministic = false;
4660 return next.study(info);
4661 }
4662 }
4663
4664 static final class Conditional extends Node {
4665 Node cond, yes, not;
4666 Conditional(Node cond, Node yes, Node not) {
4667 this.cond = cond;
4668 this.yes = yes;
4669 this.not = not;
4670 }
4671 boolean match(Matcher matcher, int i, CharSequence seq) {
4672 if (cond.match(matcher, i, seq)) {
4673 return yes.match(matcher, i, seq);
4674 } else {
4675 return not.match(matcher, i, seq);
4676 }
4677 }
4678 boolean study(TreeInfo info) {
4679 int minL = info.minLength;
4680 int maxL = info.maxLength;
4681 boolean maxV = info.maxValid;
4682 info.reset();
4683 yes.study(info);
4684
4685 int minL2 = info.minLength;
4686 int maxL2 = info.maxLength;
4687 boolean maxV2 = info.maxValid;
4688 info.reset();
4689 not.study(info);
4690
4691 info.minLength = minL + Math.min(minL2, info.minLength);
4692 info.maxLength = maxL + Math.max(maxL2, info.maxLength);
4693 info.maxValid = (maxV & maxV2 & info.maxValid);
4694 info.deterministic = false;
4695 return next.study(info);
4696 }
4697 }
4698
4699 /**
4700 * Zero width positive lookahead.
4701 */
4702 static final class Pos extends Node {
4703 Node cond;
4704 Pos(Node cond) {
4705 this.cond = cond;
4706 }
4707 boolean match(Matcher matcher, int i, CharSequence seq) {
4708 int savedTo = matcher.to;
4709 boolean conditionMatched = false;
4710
4711 // Relax transparent region boundaries for lookahead
4712 if (matcher.transparentBounds)
4713 matcher.to = matcher.getTextLength();
4714 try {
4715 conditionMatched = cond.match(matcher, i, seq);
4716 } finally {
4717 // Reinstate region boundaries
4718 matcher.to = savedTo;
4719 }
4720 return conditionMatched && next.match(matcher, i, seq);
4721 }
4722 }
4723
4724 /**
4725 * Zero width negative lookahead.
4726 */
4727 static final class Neg extends Node {
4728 Node cond;
4729 Neg(Node cond) {
4730 this.cond = cond;
4731 }
4732 boolean match(Matcher matcher, int i, CharSequence seq) {
4733 int savedTo = matcher.to;
4734 boolean conditionMatched = false;
4735
4736 // Relax transparent region boundaries for lookahead
4737 if (matcher.transparentBounds)
4738 matcher.to = matcher.getTextLength();
4739 try {
4740 if (i < matcher.to) {
4741 conditionMatched = !cond.match(matcher, i, seq);
4742 } else {
4743 // If a negative lookahead succeeds then more input
4744 // could cause it to fail!
4745 matcher.requireEnd = true;
4746 conditionMatched = !cond.match(matcher, i, seq);
4747 }
4748 } finally {
4749 // Reinstate region boundaries
4750 matcher.to = savedTo;
4751 }
4752 return conditionMatched && next.match(matcher, i, seq);
4753 }
4754 }
4755
4756 /**
4757 * For use with lookbehinds; matches the position where the lookbehind
4758 * was encountered.
4759 */
4760 static Node lookbehindEnd = new Node() {
4761 boolean match(Matcher matcher, int i, CharSequence seq) {
4762 return i == matcher.lookbehindTo;
4763 }
4764 };
4765
4766 /**
4767 * Zero width positive lookbehind.
4768 */
4769 static class Behind extends Node {
4770 Node cond;
4771 int rmax, rmin;
4772 Behind(Node cond, int rmax, int rmin) {
4773 this.cond = cond;
4774 this.rmax = rmax;
4775 this.rmin = rmin;
4776 }
4777
4778 boolean match(Matcher matcher, int i, CharSequence seq) {
4779 int savedFrom = matcher.from;
4780 boolean conditionMatched = false;
4781 int startIndex = (!matcher.transparentBounds) ?
4782 matcher.from : 0;
4783 int from = Math.max(i - rmax, startIndex);
4784 // Set end boundary
4785 int savedLBT = matcher.lookbehindTo;
4786 matcher.lookbehindTo = i;
4787 // Relax transparent region boundaries for lookbehind
4788 if (matcher.transparentBounds)
4789 matcher.from = 0;
4790 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
4791 conditionMatched = cond.match(matcher, j, seq);
4792 }
4793 matcher.from = savedFrom;
4794 matcher.lookbehindTo = savedLBT;
4795 return conditionMatched && next.match(matcher, i, seq);
4796 }
4797 }
4798
4799 /**
4800 * Zero width positive lookbehind, including supplementary
4801 * characters or unpaired surrogates.
4802 */
4803 static final class BehindS extends Behind {
4804 BehindS(Node cond, int rmax, int rmin) {
4805 super(cond, rmax, rmin);
4806 }
4807 boolean match(Matcher matcher, int i, CharSequence seq) {
4808 int rmaxChars = countChars(seq, i, -rmax);
4809 int rminChars = countChars(seq, i, -rmin);
4810 int savedFrom = matcher.from;
4811 int startIndex = (!matcher.transparentBounds) ?
4812 matcher.from : 0;
4813 boolean conditionMatched = false;
4814 int from = Math.max(i - rmaxChars, startIndex);
4815 // Set end boundary
4816 int savedLBT = matcher.lookbehindTo;
4817 matcher.lookbehindTo = i;
4818 // Relax transparent region boundaries for lookbehind
4819 if (matcher.transparentBounds)
4820 matcher.from = 0;
4821
4822 for (int j = i - rminChars;
4823 !conditionMatched && j >= from;
4824 j -= j>from ? countChars(seq, j, -1) : 1) {
4825 conditionMatched = cond.match(matcher, j, seq);
4826 }
4827 matcher.from = savedFrom;
4828 matcher.lookbehindTo = savedLBT;
4829 return conditionMatched && next.match(matcher, i, seq);
4830 }
4831 }
4832
4833 /**
4834 * Zero width negative lookbehind.
4835 */
4836 static class NotBehind extends Node {
4837 Node cond;
4838 int rmax, rmin;
4839 NotBehind(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 savedLBT = matcher.lookbehindTo;
4847 int savedFrom = matcher.from;
4848 boolean conditionMatched = false;
4849 int startIndex = (!matcher.transparentBounds) ?
4850 matcher.from : 0;
4851 int from = Math.max(i - rmax, startIndex);
4852 matcher.lookbehindTo = i;
4853 // Relax transparent region boundaries for lookbehind
4854 if (matcher.transparentBounds)
4855 matcher.from = 0;
4856 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
4857 conditionMatched = cond.match(matcher, j, seq);
4858 }
4859 // Reinstate region boundaries
4860 matcher.from = savedFrom;
4861 matcher.lookbehindTo = savedLBT;
4862 return !conditionMatched && next.match(matcher, i, seq);
4863 }
4864 }
4865
4866 /**
4867 * Zero width negative lookbehind, including supplementary
4868 * characters or unpaired surrogates.
4869 */
4870 static final class NotBehindS extends NotBehind {
4871 NotBehindS(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 savedLBT = matcher.lookbehindTo;
4879 boolean conditionMatched = false;
4880 int startIndex = (!matcher.transparentBounds) ?
4881 matcher.from : 0;
4882 int from = Math.max(i - rmaxChars, startIndex);
4883 matcher.lookbehindTo = i;
4884 // Relax transparent region boundaries for lookbehind
4885 if (matcher.transparentBounds)
4886 matcher.from = 0;
4887 for (int j = i - rminChars;
4888 !conditionMatched && j >= from;
4889 j -= j>from ? countChars(seq, j, -1) : 1) {
4890 conditionMatched = cond.match(matcher, j, seq);
4891 }
4892 //Reinstate region boundaries
4893 matcher.from = savedFrom;
4894 matcher.lookbehindTo = savedLBT;
4895 return !conditionMatched && next.match(matcher, i, seq);
4896 }
4897 }
4898
4899 /**
4900 * Returns the set union of two CharProperty nodes.
4901 */
4902 private static CharProperty union(final CharProperty lhs,
4903 final CharProperty rhs) {
4904 return new CharProperty() {
4905 boolean isSatisfiedBy(int ch) {
4906 return lhs.isSatisfiedBy(ch) || rhs.isSatisfiedBy(ch);}};
4907 }
4908
4909 /**
4910 * Returns the set intersection of two CharProperty nodes.
4911 */
4912 private static CharProperty intersection(final CharProperty lhs,
4913 final CharProperty rhs) {
4914 return new CharProperty() {
4915 boolean isSatisfiedBy(int ch) {
4916 return lhs.isSatisfiedBy(ch) && rhs.isSatisfiedBy(ch);}};
4917 }
4918
4919 /**
4920 * Returns the set difference of two CharProperty nodes.
4921 */
4922 private static CharProperty setDifference(final CharProperty lhs,
4923 final CharProperty rhs) {
4924 return new CharProperty() {
4925 boolean isSatisfiedBy(int ch) {
4926 return ! rhs.isSatisfiedBy(ch) && lhs.isSatisfiedBy(ch);}};
4927 }
4928
4929 /**
4930 * Handles word boundaries. Includes a field to allow this one class to
4931 * deal with the different types of word boundaries we can match. The word
4932 * characters include underscores, letters, and digits. Non spacing marks
4933 * can are also part of a word if they have a base character, otherwise
4934 * they are ignored for purposes of finding word boundaries.
4935 */
4936 static final class Bound extends Node {
4937 static int LEFT = 0x1;
4938 static int RIGHT= 0x2;
4939 static int BOTH = 0x3;
4940 static int NONE = 0x4;
4941 int type;
4942 Bound(int n) {
4943 type = n;
4944 }
4945 int check(Matcher matcher, int i, CharSequence seq) {
4946 int ch;
4947 boolean left = false;
4948 int startIndex = matcher.from;
4949 int endIndex = matcher.to;
4950 if (matcher.transparentBounds) {
4951 startIndex = 0;
4952 endIndex = matcher.getTextLength();
4953 }
4954 if (i > startIndex) {
4955 ch = Character.codePointBefore(seq, i);
4956 left = (ch == '_' || Character.isLetterOrDigit(ch) ||
4957 ((Character.getType(ch) == Character.NON_SPACING_MARK)
4958 && hasBaseCharacter(matcher, i-1, seq)));
4959 }
4960 boolean right = false;
4961 if (i < endIndex) {
4962 ch = Character.codePointAt(seq, i);
4963 right = (ch == '_' || Character.isLetterOrDigit(ch) ||
4964 ((Character.getType(ch) == Character.NON_SPACING_MARK)
4965 && hasBaseCharacter(matcher, i, seq)));
4966 } else {
4967 // Tried to access char past the end
4968 matcher.hitEnd = true;
4969 // The addition of another char could wreck a boundary
4970 matcher.requireEnd = true;
4971 }
4972 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE);
4973 }
4974 boolean match(Matcher matcher, int i, CharSequence seq) {
4975 return (check(matcher, i, seq) & type) > 0
4976 && next.match(matcher, i, seq);
4977 }
4978 }
4979
4980 /**
4981 * Non spacing marks only count as word characters in bounds calculations
4982 * if they have a base character.
4983 */
4984 private static boolean hasBaseCharacter(Matcher matcher, int i,
4985 CharSequence seq)
4986 {
4987 int start = (!matcher.transparentBounds) ?
4988 matcher.from : 0;
4989 for (int x=i; x >= start; x--) {
4990 int ch = Character.codePointAt(seq, x);
4991 if (Character.isLetterOrDigit(ch))
4992 return true;
4993 if (Character.getType(ch) == Character.NON_SPACING_MARK)
4994 continue;
4995 return false;
4996 }
4997 return false;
4998 }
4999
5000 /**
5001 * Attempts to match a slice in the input using the Boyer-Moore string
5002 * matching algorithm. The algorithm is based on the idea that the
5003 * pattern can be shifted farther ahead in the search text if it is
5004 * matched right to left.
5005 * <p>
5006 * The pattern is compared to the input one character at a time, from
5007 * the rightmost character in the pattern to the left. If the characters
5008 * all match the pattern has been found. If a character does not match,
5009 * the pattern is shifted right a distance that is the maximum of two
5010 * functions, the bad character shift and the good suffix shift. This
5011 * shift moves the attempted match position through the input more
5012 * quickly than a naive one position at a time check.
5013 * <p>
5014 * The bad character shift is based on the character from the text that
5015 * did not match. If the character does not appear in the pattern, the
5016 * pattern can be shifted completely beyond the bad character. If the
5017 * character does occur in the pattern, the pattern can be shifted to
5018 * line the pattern up with the next occurrence of that character.
5019 * <p>
5020 * The good suffix shift is based on the idea that some subset on the right
5021 * side of the pattern has matched. When a bad character is found, the
5022 * pattern can be shifted right by the pattern length if the subset does
5023 * not occur again in pattern, or by the amount of distance to the
5024 * next occurrence of the subset in the pattern.
5025 *
5026 * Boyer-Moore search methods adapted from code by Amy Yu.
5027 */
5028 static class BnM extends Node {
5029 int[] buffer;
5030 int[] lastOcc;
5031 int[] optoSft;
5032
5033 /**
5034 * Pre calculates arrays needed to generate the bad character
5035 * shift and the good suffix shift. Only the last seven bits
5036 * are used to see if chars match; This keeps the tables small
5037 * and covers the heavily used ASCII range, but occasionally
5038 * results in an aliased match for the bad character shift.
5039 */
5040 static Node optimize(Node node) {
5041 if (!(node instanceof Slice)) {
5042 return node;
5043 }
5044
5045 int[] src = ((Slice) node).buffer;
5046 int patternLength = src.length;
5047 // The BM algorithm requires a bit of overhead;
5048 // If the pattern is short don't use it, since
5049 // a shift larger than the pattern length cannot
5050 // be used anyway.
5051 if (patternLength < 4) {
5052 return node;
5053 }
5054 int i, j, k;
5055 int[] lastOcc = new int[128];
5056 int[] optoSft = new int[patternLength];
5057 // Precalculate part of the bad character shift
5058 // It is a table for where in the pattern each
5059 // lower 7-bit value occurs
5060 for (i = 0; i < patternLength; i++) {
5061 lastOcc[src[i]&0x7F] = i + 1;
5062 }
5063 // Precalculate the good suffix shift
5064 // i is the shift amount being considered
5065 NEXT: for (i = patternLength; i > 0; i--) {
5066 // j is the beginning index of suffix being considered
5067 for (j = patternLength - 1; j >= i; j--) {
5068 // Testing for good suffix
5069 if (src[j] == src[j-i]) {
5070 // src[j..len] is a good suffix
5071 optoSft[j-1] = i;
5072 } else {
5073 // No match. The array has already been
5074 // filled up with correct values before.
5075 continue NEXT;
5076 }
5077 }
5078 // This fills up the remaining of optoSft
5079 // any suffix can not have larger shift amount
5080 // then its sub-suffix. Why???
5081 while (j > 0) {
5082 optoSft[--j] = i;
5083 }
5084 }
5085 // Set the guard value because of unicode compression
5086 optoSft[patternLength-1] = 1;
5087 if (node instanceof SliceS)
5088 return new BnMS(src, lastOcc, optoSft, node.next);
5089 return new BnM(src, lastOcc, optoSft, node.next);
5090 }
5091 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5092 this.buffer = src;
5093 this.lastOcc = lastOcc;
5094 this.optoSft = optoSft;
5095 this.next = next;
5096 }
5097 boolean match(Matcher matcher, int i, CharSequence seq) {
5098 int[] src = buffer;
5099 int patternLength = src.length;
5100 int last = matcher.to - patternLength;
5101
5102 // Loop over all possible match positions in text
5103 NEXT: while (i <= last) {
5104 // Loop over pattern from right to left
5105 for (int j = patternLength - 1; j >= 0; j--) {
5106 int ch = seq.charAt(i+j);
5107 if (ch != src[j]) {
5108 // Shift search to the right by the maximum of the
5109 // bad character shift and the good suffix shift
5110 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]);
5111 continue NEXT;
5112 }
5113 }
5114 // Entire pattern matched starting at i
5115 matcher.first = i;
5116 boolean ret = next.match(matcher, i + patternLength, seq);
5117 if (ret) {
5118 matcher.first = i;
5119 matcher.groups[0] = matcher.first;
5120 matcher.groups[1] = matcher.last;
5121 return true;
5122 }
5123 i++;
5124 }
5125 // BnM is only used as the leading node in the unanchored case,
5126 // and it replaced its Start() which always searches to the end
5127 // if it doesn't find what it's looking for, so hitEnd is true.
5128 matcher.hitEnd = true;
5129 return false;
5130 }
5131 boolean study(TreeInfo info) {
5132 info.minLength += buffer.length;
5133 info.maxValid = false;
5134 return next.study(info);
5135 }
5136 }
5137
5138 /**
5139 * Supplementary support version of BnM(). Unpaired surrogates are
5140 * also handled by this class.
5141 */
5142 static final class BnMS extends BnM {
5143 int lengthInChars;
5144
5145 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5146 super(src, lastOcc, optoSft, next);
5147 for (int x = 0; x < buffer.length; x++) {
5148 lengthInChars += Character.charCount(buffer[x]);
5149 }
5150 }
5151 boolean match(Matcher matcher, int i, CharSequence seq) {
5152 int[] src = buffer;
5153 int patternLength = src.length;
5154 int last = matcher.to - lengthInChars;
5155
5156 // Loop over all possible match positions in text
5157 NEXT: while (i <= last) {
5158 // Loop over pattern from right to left
5159 int ch;
5160 for (int j = countChars(seq, i, patternLength), x = patternLength - 1;
5161 j > 0; j -= Character.charCount(ch), x--) {
5162 ch = Character.codePointBefore(seq, i+j);
5163 if (ch != src[x]) {
5164 // Shift search to the right by the maximum of the
5165 // bad character shift and the good suffix shift
5166 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]);
5167 i += countChars(seq, i, n);
5168 continue NEXT;
5169 }
5170 }
5171 // Entire pattern matched starting at i
5172 matcher.first = i;
5173 boolean ret = next.match(matcher, i + lengthInChars, seq);
5174 if (ret) {
5175 matcher.first = i;
5176 matcher.groups[0] = matcher.first;
5177 matcher.groups[1] = matcher.last;
5178 return true;
5179 }
5180 i += countChars(seq, i, 1);
5181 }
5182 matcher.hitEnd = true;
5183 return false;
5184 }
5185 }
5186
5187 ///////////////////////////////////////////////////////////////////////////////
5188 ///////////////////////////////////////////////////////////////////////////////
5189
5190 /**
5191 * This must be the very first initializer.
5192 */
5193 static Node accept = new Node();
5194
5195 static Node lastAccept = new LastNode();
5196
5197 private static class CharPropertyNames {
5198
5199 static CharProperty charPropertyFor(String name) {
5200 CharPropertyFactory m = map.get(name);
5201 return m == null ? null : m.make();
5202 }
5203
5204 private static abstract class CharPropertyFactory {
5205 abstract CharProperty make();
5206 }
5207
5208 private static void defCategory(String name,
5209 final int typeMask) {
5210 map.put(name, new CharPropertyFactory() {
5211 CharProperty make() { return new Category(typeMask);}});
5212 }
5213
5214 private static void defRange(String name,
5215 final int lower, final int upper) {
5216 map.put(name, new CharPropertyFactory() {
5217 CharProperty make() { return rangeFor(lower, upper);}});
5218 }
5219
5220 private static void defCtype(String name,
5221 final int ctype) {
5222 map.put(name, new CharPropertyFactory() {
5223 CharProperty make() { return new Ctype(ctype);}});
5224 }
5225
5226 private static abstract class CloneableProperty
5227 extends CharProperty implements Cloneable
5228 {
5229 public CloneableProperty clone() {
5230 try {
5231 return (CloneableProperty) super.clone();
5232 } catch (CloneNotSupportedException e) {
5233 throw new AssertionError(e);
5234 }
5235 }
5236 }
5237
5238 private static void defClone(String name,
5239 final CloneableProperty p) {
5240 map.put(name, new CharPropertyFactory() {
5241 CharProperty make() { return p.clone();}});
5242 }
5243
5244 private static final HashMap<String, CharPropertyFactory> map
5245 = new HashMap<String, CharPropertyFactory>();
5246
5247 static {
5248 // Unicode character property aliases, defined in
5249 // http://www.unicode.org/Public/UNIDATA/PropertyValueAliases.txt
5250 defCategory("Cn", 1<<Character.UNASSIGNED);
5251 defCategory("Lu", 1<<Character.UPPERCASE_LETTER);
5252 defCategory("Ll", 1<<Character.LOWERCASE_LETTER);
5253 defCategory("Lt", 1<<Character.TITLECASE_LETTER);
5254 defCategory("Lm", 1<<Character.MODIFIER_LETTER);
5255 defCategory("Lo", 1<<Character.OTHER_LETTER);
5256 defCategory("Mn", 1<<Character.NON_SPACING_MARK);
5257 defCategory("Me", 1<<Character.ENCLOSING_MARK);
5258 defCategory("Mc", 1<<Character.COMBINING_SPACING_MARK);
5259 defCategory("Nd", 1<<Character.DECIMAL_DIGIT_NUMBER);
5260 defCategory("Nl", 1<<Character.LETTER_NUMBER);
5261 defCategory("No", 1<<Character.OTHER_NUMBER);
5262 defCategory("Zs", 1<<Character.SPACE_SEPARATOR);
5263 defCategory("Zl", 1<<Character.LINE_SEPARATOR);
5264 defCategory("Zp", 1<<Character.PARAGRAPH_SEPARATOR);
5265 defCategory("Cc", 1<<Character.CONTROL);
5266 defCategory("Cf", 1<<Character.FORMAT);
5267 defCategory("Co", 1<<Character.PRIVATE_USE);
5268 defCategory("Cs", 1<<Character.SURROGATE);
5269 defCategory("Pd", 1<<Character.DASH_PUNCTUATION);
5270 defCategory("Ps", 1<<Character.START_PUNCTUATION);
5271 defCategory("Pe", 1<<Character.END_PUNCTUATION);
5272 defCategory("Pc", 1<<Character.CONNECTOR_PUNCTUATION);
5273 defCategory("Po", 1<<Character.OTHER_PUNCTUATION);
5274 defCategory("Sm", 1<<Character.MATH_SYMBOL);
5275 defCategory("Sc", 1<<Character.CURRENCY_SYMBOL);
5276 defCategory("Sk", 1<<Character.MODIFIER_SYMBOL);
5277 defCategory("So", 1<<Character.OTHER_SYMBOL);
5278 defCategory("Pi", 1<<Character.INITIAL_QUOTE_PUNCTUATION);
5279 defCategory("Pf", 1<<Character.FINAL_QUOTE_PUNCTUATION);
5280 defCategory("L", ((1<<Character.UPPERCASE_LETTER) |
5281 (1<<Character.LOWERCASE_LETTER) |
5282 (1<<Character.TITLECASE_LETTER) |
5283 (1<<Character.MODIFIER_LETTER) |
5284 (1<<Character.OTHER_LETTER)));
5285 defCategory("M", ((1<<Character.NON_SPACING_MARK) |
5286 (1<<Character.ENCLOSING_MARK) |
5287 (1<<Character.COMBINING_SPACING_MARK)));
5288 defCategory("N", ((1<<Character.DECIMAL_DIGIT_NUMBER) |
5289 (1<<Character.LETTER_NUMBER) |
5290 (1<<Character.OTHER_NUMBER)));
5291 defCategory("Z", ((1<<Character.SPACE_SEPARATOR) |
5292 (1<<Character.LINE_SEPARATOR) |
5293 (1<<Character.PARAGRAPH_SEPARATOR)));
5294 defCategory("C", ((1<<Character.CONTROL) |
5295 (1<<Character.FORMAT) |
5296 (1<<Character.PRIVATE_USE) |
5297 (1<<Character.SURROGATE))); // Other
5298 defCategory("P", ((1<<Character.DASH_PUNCTUATION) |
5299 (1<<Character.START_PUNCTUATION) |
5300 (1<<Character.END_PUNCTUATION) |
5301 (1<<Character.CONNECTOR_PUNCTUATION) |
5302 (1<<Character.OTHER_PUNCTUATION) |
5303 (1<<Character.INITIAL_QUOTE_PUNCTUATION) |
5304 (1<<Character.FINAL_QUOTE_PUNCTUATION)));
5305 defCategory("S", ((1<<Character.MATH_SYMBOL) |
5306 (1<<Character.CURRENCY_SYMBOL) |
5307 (1<<Character.MODIFIER_SYMBOL) |
5308 (1<<Character.OTHER_SYMBOL)));
5309 defCategory("LC", ((1<<Character.UPPERCASE_LETTER) |
5310 (1<<Character.LOWERCASE_LETTER) |
5311 (1<<Character.TITLECASE_LETTER)));
5312 defCategory("LD", ((1<<Character.UPPERCASE_LETTER) |
5313 (1<<Character.LOWERCASE_LETTER) |
5314 (1<<Character.TITLECASE_LETTER) |
5315 (1<<Character.MODIFIER_LETTER) |
5316 (1<<Character.OTHER_LETTER) |
5317 (1<<Character.DECIMAL_DIGIT_NUMBER)));
5318 defRange("L1", 0x00, 0xFF); // Latin-1
5319 map.put("all", new CharPropertyFactory() {
5320 CharProperty make() { return new All(); }});
5321
5322 // Posix regular expression character classes, defined in
5323 // http://www.unix.org/onlinepubs/009695399/basedefs/xbd_chap09.html
5324 defRange("ASCII", 0x00, 0x7F); // ASCII
5325 defCtype("Alnum", ASCII.ALNUM); // Alphanumeric characters
5326 defCtype("Alpha", ASCII.ALPHA); // Alphabetic characters
5327 defCtype("Blank", ASCII.BLANK); // Space and tab characters
5328 defCtype("Cntrl", ASCII.CNTRL); // Control characters
5329 defRange("Digit", '0', '9'); // Numeric characters
5330 defCtype("Graph", ASCII.GRAPH); // printable and visible
5331 defRange("Lower", 'a', 'z'); // Lower-case alphabetic
5332 defRange("Print", 0x20, 0x7E); // Printable characters
5333 defCtype("Punct", ASCII.PUNCT); // Punctuation characters
5334 defCtype("Space", ASCII.SPACE); // Space characters
5335 defRange("Upper", 'A', 'Z'); // Upper-case alphabetic
5336 defCtype("XDigit",ASCII.XDIGIT); // hexadecimal digits
5337
5338 // Java character properties, defined by methods in Character.java
5339 defClone("javaLowerCase", new CloneableProperty() {
5340 boolean isSatisfiedBy(int ch) {
5341 return Character.isLowerCase(ch);}});
5342 defClone("javaUpperCase", new CloneableProperty() {
5343 boolean isSatisfiedBy(int ch) {
5344 return Character.isUpperCase(ch);}});
5345 defClone("javaTitleCase", new CloneableProperty() {
5346 boolean isSatisfiedBy(int ch) {
5347 return Character.isTitleCase(ch);}});
5348 defClone("javaDigit", new CloneableProperty() {
5349 boolean isSatisfiedBy(int ch) {
5350 return Character.isDigit(ch);}});
5351 defClone("javaDefined", new CloneableProperty() {
5352 boolean isSatisfiedBy(int ch) {
5353 return Character.isDefined(ch);}});
5354 defClone("javaLetter", new CloneableProperty() {
5355 boolean isSatisfiedBy(int ch) {
5356 return Character.isLetter(ch);}});
5357 defClone("javaLetterOrDigit", new CloneableProperty() {
5358 boolean isSatisfiedBy(int ch) {
5359 return Character.isLetterOrDigit(ch);}});
5360 defClone("javaJavaIdentifierStart", new CloneableProperty() {
5361 boolean isSatisfiedBy(int ch) {
5362 return Character.isJavaIdentifierStart(ch);}});
5363 defClone("javaJavaIdentifierPart", new CloneableProperty() {
5364 boolean isSatisfiedBy(int ch) {
5365 return Character.isJavaIdentifierPart(ch);}});
5366 defClone("javaUnicodeIdentifierStart", new CloneableProperty() {
5367 boolean isSatisfiedBy(int ch) {
5368 return Character.isUnicodeIdentifierStart(ch);}});
5369 defClone("javaUnicodeIdentifierPart", new CloneableProperty() {
5370 boolean isSatisfiedBy(int ch) {
5371 return Character.isUnicodeIdentifierPart(ch);}});
5372 defClone("javaIdentifierIgnorable", new CloneableProperty() {
5373 boolean isSatisfiedBy(int ch) {
5374 return Character.isIdentifierIgnorable(ch);}});
5375 defClone("javaSpaceChar", new CloneableProperty() {
5376 boolean isSatisfiedBy(int ch) {
5377 return Character.isSpaceChar(ch);}});
5378 defClone("javaWhitespace", new CloneableProperty() {
5379 boolean isSatisfiedBy(int ch) {
5380 return Character.isWhitespace(ch);}});
5381 defClone("javaISOControl", new CloneableProperty() {
5382 boolean isSatisfiedBy(int ch) {
5383 return Character.isISOControl(ch);}});
5384 defClone("javaMirrored", new CloneableProperty() {
5385 boolean isSatisfiedBy(int ch) {
5386 return Character.isMirrored(ch);}});
5387 }
5388 }
5389 }