| Pattern.java |
1 /*
2 * %W% %E%
3 *
4 * Copyright (c) 2006, Oracle and/or its affiliates. All rights reserved.
5 * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
6 */
7
8 package java.util.regex;
9
10 import java.security.AccessController;
11 import java.security.PrivilegedAction;
12 import java.text.CharacterIterator;
13 import java.text.Normalizer;
14 import java.util.ArrayList;
15 import java.util.HashMap;
16 import java.util.Arrays;
17
18
19 /**
20 * A compiled representation of a regular expression.
21 *
22 * <p> A regular expression, specified as a string, must first be compiled into
23 * an instance of this class. The resulting pattern can then be used to create
24 * a {@link Matcher} object that can match arbitrary {@link
25 * java.lang.CharSequence </code>character sequences<code>} against the regular
26 * expression. All of the state involved in performing a match resides in the
27 * matcher, so many matchers can share the same pattern.
28 *
29 * <p> A typical invocation sequence is thus
30 *
31 * <blockquote><pre>
32 * Pattern p = Pattern.{@link #compile compile}("a*b");
33 * Matcher m = p.{@link #matcher matcher}("aaaaab");
34 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote>
35 *
36 * <p> A {@link #matches matches} method is defined by this class as a
37 * convenience for when a regular expression is used just once. This method
38 * compiles an expression and matches an input sequence against it in a single
39 * invocation. The statement
40 *
41 * <blockquote><pre>
42 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote>
43 *
44 * is equivalent to the three statements above, though for repeated matches it
45 * is less efficient since it does not allow the compiled pattern to be reused.
46 *
47 * <p> Instances of this class are immutable and are safe for use by multiple
48 * concurrent threads. Instances of the {@link Matcher} class are not safe for
49 * such use.
50 *
51 *
52 * <a name="sum">
53 * <h4> Summary of regular-expression constructs </h4>
54 *
55 * <table border="0" cellpadding="1" cellspacing="0"
56 * summary="Regular expression constructs, and what they match">
57 *
58 * <tr align="left">
59 * <th bgcolor="#CCCCFF" align="left" id="construct">Construct</th>
60 * <th bgcolor="#CCCCFF" align="left" id="matches">Matches</th>
61 * </tr>
62 *
63 * <tr><th> </th></tr>
64 * <tr align="left"><th colspan="2" id="characters">Characters</th></tr>
65 *
66 * <tr><td valign="top" headers="construct characters"><i>x</i></td>
67 * <td headers="matches">The character <i>x</i></td></tr>
68 * <tr><td valign="top" headers="construct characters"><tt>\\</tt></td>
69 * <td headers="matches">The backslash character</td></tr>
70 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>n</i></td>
71 * <td headers="matches">The character with octal value <tt>0</tt><i>n</i>
72 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
73 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>nn</i></td>
74 * <td headers="matches">The character with octal value <tt>0</tt><i>nn</i>
75 * (0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
76 * <tr><td valign="top" headers="construct characters"><tt>\0</tt><i>mnn</i></td>
77 * <td headers="matches">The character with octal value <tt>0</tt><i>mnn</i>
78 * (0 <tt><=</tt> <i>m</i> <tt><=</tt> 3,
79 * 0 <tt><=</tt> <i>n</i> <tt><=</tt> 7)</td></tr>
80 * <tr><td valign="top" headers="construct characters"><tt>\x</tt><i>hh</i></td>
81 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hh</i></td></tr>
82 * <tr><td valign="top" headers="construct characters"><tt>\u</tt><i>hhhh</i></td>
83 * <td headers="matches">The character with hexadecimal value <tt>0x</tt><i>hhhh</i></td></tr>
84 * <tr><td valign="top" headers="matches"><tt>\t</tt></td>
85 * <td headers="matches">The tab character (<tt>'\u0009'</tt>)</td></tr>
86 * <tr><td valign="top" headers="construct characters"><tt>\n</tt></td>
87 * <td headers="matches">The newline (line feed) character (<tt>'\u000A'</tt>)</td></tr>
88 * <tr><td valign="top" headers="construct characters"><tt>\r</tt></td>
89 * <td headers="matches">The carriage-return character (<tt>'\u000D'</tt>)</td></tr>
90 * <tr><td valign="top" headers="construct characters"><tt>\f</tt></td>
91 * <td headers="matches">The form-feed character (<tt>'\u000C'</tt>)</td></tr>
92 * <tr><td valign="top" headers="construct characters"><tt>\a</tt></td>
93 * <td headers="matches">The alert (bell) character (<tt>'\u0007'</tt>)</td></tr>
94 * <tr><td valign="top" headers="construct characters"><tt>\e</tt></td>
95 * <td headers="matches">The escape character (<tt>'\u001B'</tt>)</td></tr>
96 * <tr><td valign="top" headers="construct characters"><tt>\c</tt><i>x</i></td>
97 * <td headers="matches">The control character corresponding to <i>x</i></td></tr>
98 *
99 * <tr><th> </th></tr>
100 * <tr align="left"><th colspan="2" id="classes">Character classes</th></tr>
101 *
102 * <tr><td valign="top" headers="construct classes"><tt>[abc]</tt></td>
103 * <td headers="matches"><tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (simple class)</td></tr>
104 * <tr><td valign="top" headers="construct classes"><tt>[^abc]</tt></td>
105 * <td headers="matches">Any character except <tt>a</tt>, <tt>b</tt>, or <tt>c</tt> (negation)</td></tr>
106 * <tr><td valign="top" headers="construct classes"><tt>[a-zA-Z]</tt></td>
107 * <td headers="matches"><tt>a</tt> through <tt>z</tt>
108 * or <tt>A</tt> through <tt>Z</tt>, inclusive (range)</td></tr>
109 * <tr><td valign="top" headers="construct classes"><tt>[a-d[m-p]]</tt></td>
110 * <td headers="matches"><tt>a</tt> through <tt>d</tt>,
111 * or <tt>m</tt> through <tt>p</tt>: <tt>[a-dm-p]</tt> (union)</td></tr>
112 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[def]]</tt></td>
113 * <td headers="matches"><tt>d</tt>, <tt>e</tt>, or <tt>f</tt> (intersection)</tr>
114 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^bc]]</tt></td>
115 * <td headers="matches"><tt>a</tt> through <tt>z</tt>,
116 * except for <tt>b</tt> and <tt>c</tt>: <tt>[ad-z]</tt> (subtraction)</td></tr>
117 * <tr><td valign="top" headers="construct classes"><tt>[a-z&&[^m-p]]</tt></td>
118 * <td headers="matches"><tt>a</tt> through <tt>z</tt>,
119 * and not <tt>m</tt> through <tt>p</tt>: <tt>[a-lq-z]</tt>(subtraction)</td></tr>
120 * <tr><th> </th></tr>
121 *
122 * <tr align="left"><th colspan="2" id="predef">Predefined character classes</th></tr>
123 *
124 * <tr><td valign="top" headers="construct predef"><tt>.</tt></td>
125 * <td headers="matches">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr>
126 * <tr><td valign="top" headers="construct predef"><tt>\d</tt></td>
127 * <td headers="matches">A digit: <tt>[0-9]</tt></td></tr>
128 * <tr><td valign="top" headers="construct predef"><tt>\D</tt></td>
129 * <td headers="matches">A non-digit: <tt>[^0-9]</tt></td></tr>
130 * <tr><td valign="top" headers="construct predef"><tt>\s</tt></td>
131 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr>
132 * <tr><td valign="top" headers="construct predef"><tt>\S</tt></td>
133 * <td headers="matches">A non-whitespace character: <tt>[^\s]</tt></td></tr>
134 * <tr><td valign="top" headers="construct predef"><tt>\w</tt></td>
135 * <td headers="matches">A word character: <tt>[a-zA-Z_0-9]</tt></td></tr>
136 * <tr><td valign="top" headers="construct predef"><tt>\W</tt></td>
137 * <td headers="matches">A non-word character: <tt>[^\w]</tt></td></tr>
138 *
139 * <tr><th> </th></tr>
140 * <tr align="left"><th colspan="2" id="posix">POSIX character classes</b> (US-ASCII only)<b></th></tr>
141 *
142 * <tr><td valign="top" headers="construct posix"><tt>\p{Lower}</tt></td>
143 * <td headers="matches">A lower-case alphabetic character: <tt>[a-z]</tt></td></tr>
144 * <tr><td valign="top" headers="construct posix"><tt>\p{Upper}</tt></td>
145 * <td headers="matches">An upper-case alphabetic character:<tt>[A-Z]</tt></td></tr>
146 * <tr><td valign="top" headers="construct posix"><tt>\p{ASCII}</tt></td>
147 * <td headers="matches">All ASCII:<tt>[\x00-\x7F]</tt></td></tr>
148 * <tr><td valign="top" headers="construct posix"><tt>\p{Alpha}</tt></td>
149 * <td headers="matches">An alphabetic character:<tt>[\p{Lower}\p{Upper}]</tt></td></tr>
150 * <tr><td valign="top" headers="construct posix"><tt>\p{Digit}</tt></td>
151 * <td headers="matches">A decimal digit: <tt>[0-9]</tt></td></tr>
152 * <tr><td valign="top" headers="construct posix"><tt>\p{Alnum}</tt></td>
153 * <td headers="matches">An alphanumeric character:<tt>[\p{Alpha}\p{Digit}]</tt></td></tr>
154 * <tr><td valign="top" headers="construct posix"><tt>\p{Punct}</tt></td>
155 * <td headers="matches">Punctuation: One of <tt>!"#$%&'()*+,-./:;<=>?@[\]^_`{|}~</tt></td></tr>
156 * <!-- <tt>[\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]</tt>
157 * <tt>[\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]</tt> -->
158 * <tr><td valign="top" headers="construct posix"><tt>\p{Graph}</tt></td>
159 * <td headers="matches">A visible character: <tt>[\p{Alnum}\p{Punct}]</tt></td></tr>
160 * <tr><td valign="top" headers="construct posix"><tt>\p{Print}</tt></td>
161 * <td headers="matches">A printable character: <tt>[\p{Graph}\x20]</tt></td></tr>
162 * <tr><td valign="top" headers="construct posix"><tt>\p{Blank}</tt></td>
163 * <td headers="matches">A space or a tab: <tt>[ \t]</tt></td></tr>
164 * <tr><td valign="top" headers="construct posix"><tt>\p{Cntrl}</tt></td>
165 * <td headers="matches">A control character: <tt>[\x00-\x1F\x7F]</tt></td></tr>
166 * <tr><td valign="top" headers="construct posix"><tt>\p{XDigit}</tt></td>
167 * <td headers="matches">A hexadecimal digit: <tt>[0-9a-fA-F]</tt></td></tr>
168 * <tr><td valign="top" headers="construct posix"><tt>\p{Space}</tt></td>
169 * <td headers="matches">A whitespace character: <tt>[ \t\n\x0B\f\r]</tt></td></tr>
170 *
171 * <tr><th> </th></tr>
172 * <tr align="left"><th colspan="2">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr>
173 *
174 * <tr><td valign="top"><tt>\p{javaLowerCase}</tt></td>
175 * <td>Equivalent to java.lang.Character.isLowerCase()</td></tr>
176 * <tr><td valign="top"><tt>\p{javaUpperCase}</tt></td>
177 * <td>Equivalent to java.lang.Character.isUpperCase()</td></tr>
178 * <tr><td valign="top"><tt>\p{javaWhitespace}</tt></td>
179 * <td>Equivalent to java.lang.Character.isWhitespace()</td></tr>
180 * <tr><td valign="top"><tt>\p{javaMirrored}</tt></td>
181 * <td>Equivalent to java.lang.Character.isMirrored()</td></tr>
182 *
183 * <tr><th> </th></tr>
184 * <tr align="left"><th colspan="2" id="unicode">Classes for Unicode blocks and categories</th></tr>
185 *
186 * <tr><td valign="top" headers="construct unicode"><tt>\p{InGreek}</tt></td>
187 * <td headers="matches">A character in the Greek block (simple <a href="#ubc">block</a>)</td></tr>
188 * <tr><td valign="top" headers="construct unicode"><tt>\p{Lu}</tt></td>
189 * <td headers="matches">An uppercase letter (simple <a href="#ubc">category</a>)</td></tr>
190 * <tr><td valign="top" headers="construct unicode"><tt>\p{Sc}</tt></td>
191 * <td headers="matches">A currency symbol</td></tr>
192 * <tr><td valign="top" headers="construct unicode"><tt>\P{InGreek}</tt></td>
193 * <td headers="matches">Any character except one in the Greek block (negation)</td></tr>
194 * <tr><td valign="top" headers="construct unicode"><tt>[\p{L}&&[^\p{Lu}]] </tt></td>
195 * <td headers="matches">Any letter except an uppercase letter (subtraction)</td></tr>
196 *
197 * <tr><th> </th></tr>
198 * <tr align="left"><th colspan="2" id="bounds">Boundary matchers</th></tr>
199 *
200 * <tr><td valign="top" headers="construct bounds"><tt>^</tt></td>
201 * <td headers="matches">The beginning of a line</td></tr>
202 * <tr><td valign="top" headers="construct bounds"><tt>$</tt></td>
203 * <td headers="matches">The end of a line</td></tr>
204 * <tr><td valign="top" headers="construct bounds"><tt>\b</tt></td>
205 * <td headers="matches">A word boundary</td></tr>
206 * <tr><td valign="top" headers="construct bounds"><tt>\B</tt></td>
207 * <td headers="matches">A non-word boundary</td></tr>
208 * <tr><td valign="top" headers="construct bounds"><tt>\A</tt></td>
209 * <td headers="matches">The beginning of the input</td></tr>
210 * <tr><td valign="top" headers="construct bounds"><tt>\G</tt></td>
211 * <td headers="matches">The end of the previous match</td></tr>
212 * <tr><td valign="top" headers="construct bounds"><tt>\Z</tt></td>
213 * <td headers="matches">The end of the input but for the final
214 * <a href="#lt">terminator</a>, if any</td></tr>
215 * <tr><td valign="top" headers="construct bounds"><tt>\z</tt></td>
216 * <td headers="matches">The end of the input</td></tr>
217 *
218 * <tr><th> </th></tr>
219 * <tr align="left"><th colspan="2" id="greedy">Greedy quantifiers</th></tr>
220 *
221 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>?</tt></td>
222 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
223 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>*</tt></td>
224 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
225 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>+</tt></td>
226 * <td headers="matches"><i>X</i>, one or more times</td></tr>
227 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>}</tt></td>
228 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
229 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>,}</tt></td>
230 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
231 * <tr><td valign="top" headers="construct greedy"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}</tt></td>
232 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
233 *
234 * <tr><th> </th></tr>
235 * <tr align="left"><th colspan="2" id="reluc">Reluctant quantifiers</th></tr>
236 *
237 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>??</tt></td>
238 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
239 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>*?</tt></td>
240 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
241 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>+?</tt></td>
242 * <td headers="matches"><i>X</i>, one or more times</td></tr>
243 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>}?</tt></td>
244 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
245 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>,}?</tt></td>
246 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
247 * <tr><td valign="top" headers="construct reluc"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}?</tt></td>
248 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
249 *
250 * <tr><th> </th></tr>
251 * <tr align="left"><th colspan="2" id="poss">Possessive quantifiers</th></tr>
252 *
253 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>?+</tt></td>
254 * <td headers="matches"><i>X</i>, once or not at all</td></tr>
255 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>*+</tt></td>
256 * <td headers="matches"><i>X</i>, zero or more times</td></tr>
257 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>++</tt></td>
258 * <td headers="matches"><i>X</i>, one or more times</td></tr>
259 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>}+</tt></td>
260 * <td headers="matches"><i>X</i>, exactly <i>n</i> times</td></tr>
261 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>,}+</tt></td>
262 * <td headers="matches"><i>X</i>, at least <i>n</i> times</td></tr>
263 * <tr><td valign="top" headers="construct poss"><i>X</i><tt>{</tt><i>n</i><tt>,</tt><i>m</i><tt>}+</tt></td>
264 * <td headers="matches"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
265 *
266 * <tr><th> </th></tr>
267 * <tr align="left"><th colspan="2" id="logical">Logical operators</th></tr>
268 *
269 * <tr><td valign="top" headers="construct logical"><i>XY</i></td>
270 * <td headers="matches"><i>X</i> followed by <i>Y</i></td></tr>
271 * <tr><td valign="top" headers="construct logical"><i>X</i><tt>|</tt><i>Y</i></td>
272 * <td headers="matches">Either <i>X</i> or <i>Y</i></td></tr>
273 * <tr><td valign="top" headers="construct logical"><tt>(</tt><i>X</i><tt>)</tt></td>
274 * <td headers="matches">X, as a <a href="#cg">capturing group</a></td></tr>
275 *
276 * <tr><th> </th></tr>
277 * <tr align="left"><th colspan="2" id="backref">Back references</th></tr>
278 *
279 * <tr><td valign="bottom" headers="construct backref"><tt>\</tt><i>n</i></td>
280 * <td valign="bottom" headers="matches">Whatever the <i>n</i><sup>th</sup>
281 * <a href="#cg">capturing group</a> matched</td></tr>
282 *
283 * <tr><th> </th></tr>
284 * <tr align="left"><th colspan="2" id="quot">Quotation</th></tr>
285 *
286 * <tr><td valign="top" headers="construct quot"><tt>\</tt></td>
287 * <td headers="matches">Nothing, but quotes the following character</td></tr>
288 * <tr><td valign="top" headers="construct quot"><tt>\Q</tt></td>
289 * <td headers="matches">Nothing, but quotes all characters until <tt>\E</tt></td></tr>
290 * <tr><td valign="top" headers="construct quot"><tt>\E</tt></td>
291 * <td headers="matches">Nothing, but ends quoting started by <tt>\Q</tt></td></tr>
292 * <!-- Metachars: !$()*+.<>?[\]^{|} -->
293 *
294 * <tr><th> </th></tr>
295 * <tr align="left"><th colspan="2" id="special">Special constructs (non-capturing)</th></tr>
296 *
297 * <tr><td valign="top" headers="construct special"><tt>(?:</tt><i>X</i><tt>)</tt></td>
298 * <td headers="matches"><i>X</i>, as a non-capturing group</td></tr>
299 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux) </tt></td>
300 * <td headers="matches">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a>
301 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a>
302 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> on - off</td></tr>
303 * <tr><td valign="top" headers="construct special"><tt>(?idmsux-idmsux:</tt><i>X</i><tt>)</tt> </td>
304 * <td headers="matches"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the
305 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a>
306 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a >
307 * <a href="#COMMENTS">x</a> on - off</td></tr>
308 * <tr><td valign="top" headers="construct special"><tt>(?=</tt><i>X</i><tt>)</tt></td>
309 * <td headers="matches"><i>X</i>, via zero-width positive lookahead</td></tr>
310 * <tr><td valign="top" headers="construct special"><tt>(?!</tt><i>X</i><tt>)</tt></td>
311 * <td headers="matches"><i>X</i>, via zero-width negative lookahead</td></tr>
312 * <tr><td valign="top" headers="construct special"><tt>(?<=</tt><i>X</i><tt>)</tt></td>
313 * <td headers="matches"><i>X</i>, via zero-width positive lookbehind</td></tr>
314 * <tr><td valign="top" headers="construct special"><tt>(?<!</tt><i>X</i><tt>)</tt></td>
315 * <td headers="matches"><i>X</i>, via zero-width negative lookbehind</td></tr>
316 * <tr><td valign="top" headers="construct special"><tt>(?></tt><i>X</i><tt>)</tt></td>
317 * <td headers="matches"><i>X</i>, as an independent, non-capturing group</td></tr>
318 *
319 * </table>
320 *
321 * <hr>
322 *
323 *
324 * <a name="bs">
325 * <h4> Backslashes, escapes, and quoting </h4>
326 *
327 * <p> The backslash character (<tt>'\'</tt>) serves to introduce escaped
328 * constructs, as defined in the table above, as well as to quote characters
329 * that otherwise would be interpreted as unescaped constructs. Thus the
330 * expression <tt>\\</tt> matches a single backslash and <tt>\{</tt> matches a
331 * left brace.
332 *
333 * <p> It is an error to use a backslash prior to any alphabetic character that
334 * does not denote an escaped construct; these are reserved for future
335 * extensions to the regular-expression language. A backslash may be used
336 * prior to a non-alphabetic character regardless of whether that character is
337 * part of an unescaped construct.
338 *
339 * <p> Backslashes within string literals in Java source code are interpreted
340 * as required by the <a
341 * href="http://java.sun.com/docs/books/jls">Java Language
342 * Specification</a> as either <a
343 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">Unicode
344 * escapes</a> or other <a
345 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#101089">character
346 * escapes</a>. It is therefore necessary to double backslashes in string
347 * literals that represent regular expressions to protect them from
348 * interpretation by the Java bytecode compiler. The string literal
349 * <tt>"\b"</tt>, for example, matches a single backspace character when
350 * interpreted as a regular expression, while <tt>"\\b"</tt> matches a
351 * word boundary. The string literal <tt>"\(hello\)"</tt> is illegal
352 * and leads to a compile-time error; in order to match the string
353 * <tt>(hello)</tt> the string literal <tt>"\\(hello\\)"</tt>
354 * must be used.
355 *
356 * <a name="cc">
357 * <h4> Character Classes </h4>
358 *
359 * <p> Character classes may appear within other character classes, and
360 * may be composed by the union operator (implicit) and the intersection
361 * operator (<tt>&&</tt>).
362 * The union operator denotes a class that contains every character that is
363 * in at least one of its operand classes. The intersection operator
364 * denotes a class that contains every character that is in both of its
365 * operand classes.
366 *
367 * <p> The precedence of character-class operators is as follows, from
368 * highest to lowest:
369 *
370 * <blockquote><table border="0" cellpadding="1" cellspacing="0"
371 * summary="Precedence of character class operators.">
372 * <tr><th>1 </th>
373 * <td>Literal escape </td>
374 * <td><tt>\x</tt></td></tr>
375 * <tr><th>2 </th>
376 * <td>Grouping</td>
377 * <td><tt>[...]</tt></td></tr>
378 * <tr><th>3 </th>
379 * <td>Range</td>
380 * <td><tt>a-z</tt></td></tr>
381 * <tr><th>4 </th>
382 * <td>Union</td>
383 * <td><tt>[a-e][i-u]</tt></td></tr>
384 * <tr><th>5 </th>
385 * <td>Intersection</td>
386 * <td><tt>[a-z&&[aeiou]]</tt></td></tr>
387 * </table></blockquote>
388 *
389 * <p> Note that a different set of metacharacters are in effect inside
390 * a character class than outside a character class. For instance, the
391 * regular expression <tt>.</tt> loses its special meaning inside a
392 * character class, while the expression <tt>-</tt> becomes a range
393 * forming metacharacter.
394 *
395 * <a name="lt">
396 * <h4> Line terminators </h4>
397 *
398 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks
399 * the end of a line of the input character sequence. The following are
400 * recognized as line terminators:
401 *
402 * <ul>
403 *
404 * <li> A newline (line feed) character (<tt>'\n'</tt>),
405 *
406 * <li> A carriage-return character followed immediately by a newline
407 * character (<tt>"\r\n"</tt>),
408 *
409 * <li> A standalone carriage-return character (<tt>'\r'</tt>),
410 *
411 * <li> A next-line character (<tt>'\u0085'</tt>),
412 *
413 * <li> A line-separator character (<tt>'\u2028'</tt>), or
414 *
415 * <li> A paragraph-separator character (<tt>'\u2029</tt>).
416 *
417 * </ul>
418 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators
419 * recognized are newline characters.
420 *
421 * <p> The regular expression <tt>.</tt> matches any character except a line
422 * terminator unless the {@link #DOTALL} flag is specified.
423 *
424 * <p> By default, the regular expressions <tt>^</tt> and <tt>$</tt> ignore
425 * line terminators and only match at the beginning and the end, respectively,
426 * of the entire input sequence. If {@link #MULTILINE} mode is activated then
427 * <tt>^</tt> matches at the beginning of input and after any line terminator
428 * except at the end of input. When in {@link #MULTILINE} mode <tt>$</tt>
429 * matches just before a line terminator or the end of the input sequence.
430 *
431 * <a name="cg">
432 * <h4> Groups and capturing </h4>
433 *
434 * <p> Capturing groups are numbered by counting their opening parentheses from
435 * left to right. In the expression <tt>((A)(B(C)))</tt>, for example, there
436 * are four such groups: </p>
437 *
438 * <blockquote><table cellpadding=1 cellspacing=0 summary="Capturing group numberings">
439 * <tr><th>1 </th>
440 * <td><tt>((A)(B(C)))</tt></td></tr>
441 * <tr><th>2 </th>
442 * <td><tt>(A)</tt></td></tr>
443 * <tr><th>3 </th>
444 * <td><tt>(B(C))</tt></td></tr>
445 * <tr><th>4 </th>
446 * <td><tt>(C)</tt></td></tr>
447 * </table></blockquote>
448 *
449 * <p> Group zero always stands for the entire expression.
450 *
451 * <p> Capturing groups are so named because, during a match, each subsequence
452 * of the input sequence that matches such a group is saved. The captured
453 * subsequence may be used later in the expression, via a back reference, and
454 * may also be retrieved from the matcher once the match operation is complete.
455 *
456 * <p> The captured input associated with a group is always the subsequence
457 * that the group most recently matched. If a group is evaluated a second time
458 * because of quantification then its previously-captured value, if any, will
459 * be retained if the second evaluation fails. Matching the string
460 * <tt>"aba"</tt> against the expression <tt>(a(b)?)+</tt>, for example, leaves
461 * group two set to <tt>"b"</tt>. All captured input is discarded at the
462 * beginning of each match.
463 *
464 * <p> Groups beginning with <tt>(?</tt> are pure, <i>non-capturing</i> groups
465 * that do not capture text and do not count towards the group total.
466 *
467 *
468 * <h4> Unicode support </h4>
469 *
470 * <p> This class is in conformance with Level 1 of <a
471 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical
472 * Standard #18: Unicode Regular Expression Guidelines</i></a>, plus RL2.1
473 * Canonical Equivalents.
474 *
475 * <p> Unicode escape sequences such as <tt>\u2014</tt> in Java source code
476 * are processed as described in <a
477 * href="http://java.sun.com/docs/books/jls/third_edition/html/lexical.html#100850">ยง3.3</a>
478 * of the Java Language Specification. Such escape sequences are also
479 * implemented directly by the regular-expression parser so that Unicode
480 * escapes can be used in expressions that are read from files or from the
481 * keyboard. Thus the strings <tt>"\u2014"</tt> and <tt>"\\u2014"</tt>,
482 * while not equal, compile into the same pattern, which matches the character
483 * with hexadecimal value <tt>0x2014</tt>.
484 *
485 * <a name="ubc"> <p>Unicode blocks and categories are written with the
486 * <tt>\p</tt> and <tt>\P</tt> constructs as in
487 * Perl. <tt>\p{</tt><i>prop</i><tt>}</tt> matches if the input has the
488 * property <i>prop</i>, while <tt>\P{</tt><i>prop</i><tt>}</tt> does not match if
489 * the input has that property. Blocks are specified with the prefix
490 * <tt>In</tt>, as in <tt>InMongolian</tt>. Categories may be specified with
491 * the optional prefix <tt>Is</tt>: Both <tt>\p{L}</tt> and <tt>\p{IsL}</tt>
492 * denote the category of Unicode letters. Blocks and categories can be used
493 * both inside and outside of a character class.
494 *
495 * <p> The supported categories are those of
496 * <a href="http://www.unicode.org/unicode/standard/standard.html">
497 * <i>The Unicode Standard</i></a> in the version specified by the
498 * {@link java.lang.Character Character} class. The category names are those
499 * defined in the Standard, both normative and informative.
500 * The block names supported by <code>Pattern</code> are the valid block names
501 * accepted and defined by
502 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}.
503 *
504 * <a name="jcc"> <p>Categories that behave like the java.lang.Character
505 * boolean is<i>methodname</i> methods (except for the deprecated ones) are
506 * available through the same <tt>\p{</tt><i>prop</i><tt>}</tt> syntax where
507 * the specified property has the name <tt>java<i>methodname</i></tt>.
508 *
509 * <h4> Comparison to Perl 5 </h4>
510 *
511 * <p>The <code>Pattern</code> engine performs traditional NFA-based matching
512 * with ordered alternation as occurs in Perl 5.
513 *
514 * <p> Perl constructs not supported by this class: </p>
515 *
516 * <ul>
517 *
518 * <li><p> The conditional constructs <tt>(?{</tt><i>X</i><tt>})</tt> and
519 * <tt>(?(</tt><i>condition</i><tt>)</tt><i>X</i><tt>|</tt><i>Y</i><tt>)</tt>,
520 * </p></li>
521 *
522 * <li><p> The embedded code constructs <tt>(?{</tt><i>code</i><tt>})</tt>
523 * and <tt>(??{</tt><i>code</i><tt>})</tt>,</p></li>
524 *
525 * <li><p> The embedded comment syntax <tt>(?#comment)</tt>, and </p></li>
526 *
527 * <li><p> The preprocessing operations <tt>\l</tt> <tt>\u</tt>,
528 * <tt>\L</tt>, and <tt>\U</tt>. </p></li>
529 *
530 * </ul>
531 *
532 * <p> Constructs supported by this class but not by Perl: </p>
533 *
534 * <ul>
535 *
536 * <li><p> Possessive quantifiers, which greedily match as much as they can
537 * and do not back off, even when doing so would allow the overall match to
538 * succeed. </p></li>
539 *
540 * <li><p> Character-class union and intersection as described
541 * <a href="#cc">above</a>.</p></li>
542 *
543 * </ul>
544 *
545 * <p> Notable differences from Perl: </p>
546 *
547 * <ul>
548 *
549 * <li><p> In Perl, <tt>\1</tt> through <tt>\9</tt> are always interpreted
550 * as back references; a backslash-escaped number greater than <tt>9</tt> is
551 * treated as a back reference if at least that many subexpressions exist,
552 * otherwise it is interpreted, if possible, as an octal escape. In this
553 * class octal escapes must always begin with a zero. In this class,
554 * <tt>\1</tt> through <tt>\9</tt> are always interpreted as back
555 * references, and a larger number is accepted as a back reference if at
556 * least that many subexpressions exist at that point in the regular
557 * expression, otherwise the parser will drop digits until the number is
558 * smaller or equal to the existing number of groups or it is one digit.
559 * </p></li>
560 *
561 * <li><p> Perl uses the <tt>g</tt> flag to request a match that resumes
562 * where the last match left off. This functionality is provided implicitly
563 * by the {@link Matcher} class: Repeated invocations of the {@link
564 * Matcher#find find} method will resume where the last match left off,
565 * unless the matcher is reset. </p></li>
566 *
567 * <li><p> In Perl, embedded flags at the top level of an expression affect
568 * the whole expression. In this class, embedded flags always take effect
569 * at the point at which they appear, whether they are at the top level or
570 * within a group; in the latter case, flags are restored at the end of the
571 * group just as in Perl. </p></li>
572 *
573 * <li><p> Perl is forgiving about malformed matching constructs, as in the
574 * expression <tt>*a</tt>, as well as dangling brackets, as in the
575 * expression <tt>abc]</tt>, and treats them as literals. This
576 * class also accepts dangling brackets but is strict about dangling
577 * metacharacters like +, ? and *, and will throw a
578 * {@link PatternSyntaxException} if it encounters them. </p></li>
579 *
580 * </ul>
581 *
582 *
583 * <p> For a more precise description of the behavior of regular expression
584 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/">
585 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl,
586 * O'Reilly and Associates, 2006.</a>
587 * </p>
588 *
589 * @see java.lang.String#split(String, int)
590 * @see java.lang.String#split(String)
591 *
592 * @author Mike McCloskey
593 * @author Mark Reinhold
594 * @author JSR-51 Expert Group
595 * @version %I%, %E%
596 * @since 1.4
597 * @spec JSR-51
598 */
599
600 public final class Pattern
601 implements java.io.Serializable
602 {
603
604 /**
605 * Regular expression modifier values. Instead of being passed as
606 * arguments, they can also be passed as inline modifiers.
607 * For example, the following statements have the same effect.
608 * <pre>
609 * RegExp r1 = RegExp.compile("abc", Pattern.I|Pattern.M);
610 * RegExp r2 = RegExp.compile("(?im)abc", 0);
611 * </pre>
612 *
613 * The flags are duplicated so that the familiar Perl match flag
614 * names are available.
615 */
616
617 /**
618 * Enables Unix lines mode.
619 *
620 * <p> In this mode, only the <tt>'\n'</tt> line terminator is recognized
621 * in the behavior of <tt>.</tt>, <tt>^</tt>, and <tt>$</tt>.
622 *
623 * <p> Unix lines mode can also be enabled via the embedded flag
624 * expression <tt>(?d)</tt>.
625 */
626 public static final int UNIX_LINES = 0x01;
627
628 /**
629 * Enables case-insensitive matching.
630 *
631 * <p> By default, case-insensitive matching assumes that only characters
632 * in the US-ASCII charset are being matched. Unicode-aware
633 * case-insensitive matching can be enabled by specifying the {@link
634 * #UNICODE_CASE} flag in conjunction with this flag.
635 *
636 * <p> Case-insensitive matching can also be enabled via the embedded flag
637 * expression <tt>(?i)</tt>.
638 *
639 * <p> Specifying this flag may impose a slight performance penalty. </p>
640 */
641 public static final int CASE_INSENSITIVE = 0x02;
642
643 /**
644 * Permits whitespace and comments in pattern.
645 *
646 * <p> In this mode, whitespace is ignored, and embedded comments starting
647 * with <tt>#</tt> are ignored until the end of a line.
648 *
649 * <p> Comments mode can also be enabled via the embedded flag
650 * expression <tt>(?x)</tt>.
651 */
652 public static final int COMMENTS = 0x04;
653
654 /**
655 * Enables multiline mode.
656 *
657 * <p> In multiline mode the expressions <tt>^</tt> and <tt>$</tt> match
658 * just after or just before, respectively, a line terminator or the end of
659 * the input sequence. By default these expressions only match at the
660 * beginning and the end of the entire input sequence.
661 *
662 * <p> Multiline mode can also be enabled via the embedded flag
663 * expression <tt>(?m)</tt>. </p>
664 */
665 public static final int MULTILINE = 0x08;
666
667 /**
668 * Enables literal parsing of the pattern.
669 *
670 * <p> When this flag is specified then the input string that specifies
671 * the pattern is treated as a sequence of literal characters.
672 * Metacharacters or escape sequences in the input sequence will be
673 * given no special meaning.
674 *
675 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on
676 * matching when used in conjunction with this flag. The other flags
677 * become superfluous.
678 *
679 * <p> There is no embedded flag character for enabling literal parsing.
680 * @since 1.5
681 */
682 public static final int LITERAL = 0x10;
683
684 /**
685 * Enables dotall mode.
686 *
687 * <p> In dotall mode, the expression <tt>.</tt> matches any character,
688 * including a line terminator. By default this expression does not match
689 * line terminators.
690 *
691 * <p> Dotall mode can also be enabled via the embedded flag
692 * expression <tt>(?s)</tt>. (The <tt>s</tt> is a mnemonic for
693 * "single-line" mode, which is what this is called in Perl.) </p>
694 */
695 public static final int DOTALL = 0x20;
696
697 /**
698 * Enables Unicode-aware case folding.
699 *
700 * <p> When this flag is specified then case-insensitive matching, when
701 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner
702 * consistent with the Unicode Standard. By default, case-insensitive
703 * matching assumes that only characters in the US-ASCII charset are being
704 * matched.
705 *
706 * <p> Unicode-aware case folding can also be enabled via the embedded flag
707 * expression <tt>(?u)</tt>.
708 *
709 * <p> Specifying this flag may impose a performance penalty. </p>
710 */
711 public static final int UNICODE_CASE = 0x40;
712
713 /**
714 * Enables canonical equivalence.
715 *
716 * <p> When this flag is specified then two characters will be considered
717 * to match if, and only if, their full canonical decompositions match.
718 * The expression <tt>"a\u030A"</tt>, for example, will match the
719 * string <tt>"\u00E5"</tt> when this flag is specified. By default,
720 * matching does not take canonical equivalence into account.
721 *
722 * <p> There is no embedded flag character for enabling canonical
723 * equivalence.
724 *
725 * <p> Specifying this flag may impose a performance penalty. </p>
726 */
727 public static final int CANON_EQ = 0x80;
728
729 /* Pattern has only two serialized components: The pattern string
730 * and the flags, which are all that is needed to recompile the pattern
731 * when it is deserialized.
732 */
733
734 /** use serialVersionUID from Merlin b59 for interoperability */
735 private static final long serialVersionUID = 5073258162644648461L;
736
737 /**
738 * The original regular-expression pattern string.
739 *
740 * @serial
741 */
742 private String pattern;
743
744 /**
745 * The original pattern flags.
746 *
747 * @serial
748 */
749 private int flags;
750
751 /**
752 * Boolean indicating this Pattern is compiled; this is necessary in order
753 * to lazily compile deserialized Patterns.
754 */
755 private transient volatile boolean compiled = false;
756
757 /**
758 * The normalized pattern string.
759 */
760 private transient String normalizedPattern;
761
762 /**
763 * The starting point of state machine for the find operation. This allows
764 * a match to start anywhere in the input.
765 */
766 transient Node root;
767
768 /**
769 * The root of object tree for a match operation. The pattern is matched
770 * at the beginning. This may include a find that uses BnM or a First
771 * node.
772 */
773 transient Node matchRoot;
774
775 /**
776 * Temporary storage used by parsing pattern slice.
777 */
778 transient int[] buffer;
779
780 /**
781 * Temporary storage used while parsing group references.
782 */
783 transient GroupHead[] groupNodes;
784
785 /**
786 * Temporary null terminated code point array used by pattern compiling.
787 */
788 private transient int[] temp;
789
790 /**
791 * The number of capturing groups in this Pattern. Used by matchers to
792 * allocate storage needed to perform a match.
793 */
794 transient int capturingGroupCount;
795
796 /**
797 * The local variable count used by parsing tree. Used by matchers to
798 * allocate storage needed to perform a match.
799 */
800 transient int localCount;
801
802 /**
803 * Index into the pattern string that keeps track of how much has been
804 * parsed.
805 */
806 private transient int cursor;
807
808 /**
809 * Holds the length of the pattern string.
810 */
811 private transient int patternLength;
812
813 /**
814 * Compiles the given regular expression into a pattern. </p>
815 *
816 * @param regex
817 * The expression to be compiled
818 *
819 * @throws PatternSyntaxException
820 * If the expression's syntax is invalid
821 */
822 public static Pattern compile(String regex) {
823 return new Pattern(regex, 0);
824 }
825
826 /**
827 * Compiles the given regular expression into a pattern with the given
828 * flags. </p>
829 *
830 * @param regex
831 * The expression to be compiled
832 *
833 * @param flags
834 * Match flags, a bit mask that may include
835 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL},
836 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES},
837 * {@link #LITERAL} and {@link #COMMENTS}
838 *
839 * @throws IllegalArgumentException
840 * If bit values other than those corresponding to the defined
841 * match flags are set in <tt>flags</tt>
842 *
843 * @throws PatternSyntaxException
844 * If the expression's syntax is invalid
845 */
846 public static Pattern compile(String regex, int flags) {
847 return new Pattern(regex, flags);
848 }
849
850 /**
851 * Returns the regular expression from which this pattern was compiled.
852 * </p>
853 *
854 * @return The source of this pattern
855 */
856 public String pattern() {
857 return pattern;
858 }
859
860 /**
861 * <p>Returns the string representation of this pattern. This
862 * is the regular expression from which this pattern was
863 * compiled.</p>
864 *
865 * @return The string representation of this pattern
866 * @since 1.5
867 */
868 public String toString() {
869 return pattern;
870 }
871
872 /**
873 * Creates a matcher that will match the given input against this pattern.
874 * </p>
875 *
876 * @param input
877 * The character sequence to be matched
878 *
879 * @return A new matcher for this pattern
880 */
881 public Matcher matcher(CharSequence input) {
882 if (!compiled) {
883 synchronized(this) {
884 if (!compiled)
885 compile();
886 }
887 }
888 Matcher m = new Matcher(this, input);
889 return m;
890 }
891
892 /**
893 * Returns this pattern's match flags. </p>
894 *
895 * @return The match flags specified when this pattern was compiled
896 */
897 public int flags() {
898 return flags;
899 }
900
901 /**
902 * Compiles the given regular expression and attempts to match the given
903 * input against it.
904 *
905 * <p> An invocation of this convenience method of the form
906 *
907 * <blockquote><pre>
908 * Pattern.matches(regex, input);</pre></blockquote>
909 *
910 * behaves in exactly the same way as the expression
911 *
912 * <blockquote><pre>
913 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote>
914 *
915 * <p> If a pattern is to be used multiple times, compiling it once and reusing
916 * it will be more efficient than invoking this method each time. </p>
917 *
918 * @param regex
919 * The expression to be compiled
920 *
921 * @param input
922 * The character sequence to be matched
923 *
924 * @throws PatternSyntaxException
925 * If the expression's syntax is invalid
926 */
927 public static boolean matches(String regex, CharSequence input) {
928 Pattern p = Pattern.compile(regex);
929 Matcher m = p.matcher(input);
930 return m.matches();
931 }
932
933 /**
934 * Splits the given input sequence around matches of this pattern.
935 *
936 * <p> The array returned by this method contains each substring of the
937 * input sequence that is terminated by another subsequence that matches
938 * this pattern or is terminated by the end of the input sequence. The
939 * substrings in the array are in the order in which they occur in the
940 * input. If this pattern does not match any subsequence of the input then
941 * the resulting array has just one element, namely the input sequence in
942 * string form.
943 *
944 * <p> The <tt>limit</tt> parameter controls the number of times the
945 * pattern is applied and therefore affects the length of the resulting
946 * array. If the limit <i>n</i> is greater than zero then the pattern
947 * will be applied at most <i>n</i> - 1 times, the array's
948 * length will be no greater than <i>n</i>, and the array's last entry
949 * will contain all input beyond the last matched delimiter. If <i>n</i>
950 * is non-positive then the pattern will be applied as many times as
951 * possible and the array can have any length. If <i>n</i> is zero then
952 * the pattern will be applied as many times as possible, the array can
953 * have any length, and trailing empty strings will be discarded.
954 *
955 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following
956 * results with these parameters:
957 *
958 * <blockquote><table cellpadding=1 cellspacing=0
959 * summary="Split examples showing regex, limit, and result">
960 * <tr><th><P align="left"><i>Regex </i></th>
961 * <th><P align="left"><i>Limit </i></th>
962 * <th><P align="left"><i>Result </i></th></tr>
963 * <tr><td align=center>:</td>
964 * <td align=center>2</td>
965 * <td><tt>{ "boo", "and:foo" }</tt></td></tr>
966 * <tr><td align=center>:</td>
967 * <td align=center>5</td>
968 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
969 * <tr><td align=center>:</td>
970 * <td align=center>-2</td>
971 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
972 * <tr><td align=center>o</td>
973 * <td align=center>5</td>
974 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr>
975 * <tr><td align=center>o</td>
976 * <td align=center>-2</td>
977 * <td><tt>{ "b", "", ":and:f", "", "" }</tt></td></tr>
978 * <tr><td align=center>o</td>
979 * <td align=center>0</td>
980 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr>
981 * </table></blockquote>
982 *
983 *
984 * @param input
985 * The character sequence to be split
986 *
987 * @param limit
988 * The result threshold, as described above
989 *
990 * @return The array of strings computed by splitting the input
991 * around matches of this pattern
992 */
993 public String[] split(CharSequence input, int limit) {
994 int index = 0;
995 boolean matchLimited = limit > 0;
996 ArrayList<String> matchList = new ArrayList<String>();
997 Matcher m = matcher(input);
998
999 // Add segments before each match found
1000 while(m.find()) {
1001 if (!matchLimited || matchList.size() < limit - 1) {
1002 String match = input.subSequence(index, m.start()).toString();
1003 matchList.add(match);
1004 index = m.end();
1005 } else if (matchList.size() == limit - 1) { // last one
1006 String match = input.subSequence(index,
1007 input.length()).toString();
1008 matchList.add(match);
1009 index = m.end();
1010 }
1011 }
1012
1013 // If no match was found, return this
1014 if (index == 0)
1015 return new String[] {input.toString()};
1016
1017 // Add remaining segment
1018 if (!matchLimited || matchList.size() < limit)
1019 matchList.add(input.subSequence(index, input.length()).toString());
1020
1021 // Construct result
1022 int resultSize = matchList.size();
1023 if (limit == 0)
1024 while (resultSize > 0 && matchList.get(resultSize-1).equals(""))
1025 resultSize--;
1026 String[] result = new String[resultSize];
1027 return matchList.subList(0, resultSize).toArray(result);
1028 }
1029
1030 /**
1031 * Splits the given input sequence around matches of this pattern.
1032 *
1033 * <p> This method works as if by invoking the two-argument {@link
1034 * #split(java.lang.CharSequence, int) split} method with the given input
1035 * sequence and a limit argument of zero. Trailing empty strings are
1036 * therefore not included in the resulting array. </p>
1037 *
1038 * <p> The input <tt>"boo:and:foo"</tt>, for example, yields the following
1039 * results with these expressions:
1040 *
1041 * <blockquote><table cellpadding=1 cellspacing=0
1042 * summary="Split examples showing regex and result">
1043 * <tr><th><P align="left"><i>Regex </i></th>
1044 * <th><P align="left"><i>Result</i></th></tr>
1045 * <tr><td align=center>:</td>
1046 * <td><tt>{ "boo", "and", "foo" }</tt></td></tr>
1047 * <tr><td align=center>o</td>
1048 * <td><tt>{ "b", "", ":and:f" }</tt></td></tr>
1049 * </table></blockquote>
1050 *
1051 *
1052 * @param input
1053 * The character sequence to be split
1054 *
1055 * @return The array of strings computed by splitting the input
1056 * around matches of this pattern
1057 */
1058 public String[] split(CharSequence input) {
1059 return split(input, 0);
1060 }
1061
1062 /**
1063 * Returns a literal pattern <code>String</code> for the specified
1064 * <code>String</code>.
1065 *
1066 * <p>This method produces a <code>String</code> that can be used to
1067 * create a <code>Pattern</code> that would match the string
1068 * <code>s</code> as if it were a literal pattern.</p> Metacharacters
1069 * or escape sequences in the input sequence will be given no special
1070 * meaning.
1071 *
1072 * @param s The string to be literalized
1073 * @return A literal string replacement
1074 * @since 1.5
1075 */
1076 public static String quote(String s) {
1077 int slashEIndex = s.indexOf("\\E");
1078 if (slashEIndex == -1)
1079 return "\\Q" + s + "\\E";
1080
1081 StringBuilder sb = new StringBuilder(s.length() * 2);
1082 sb.append("\\Q");
1083 slashEIndex = 0;
1084 int current = 0;
1085 while ((slashEIndex = s.indexOf("\\E", current)) != -1) {
1086 sb.append(s.substring(current, slashEIndex));
1087 current = slashEIndex + 2;
1088 sb.append("\\E\\\\E\\Q");
1089 }
1090 sb.append(s.substring(current, s.length()));
1091 sb.append("\\E");
1092 return sb.toString();
1093 }
1094
1095 /**
1096 * Recompile the Pattern instance from a stream. The original pattern
1097 * string is read in and the object tree is recompiled from it.
1098 */
1099 private void readObject(java.io.ObjectInputStream s)
1100 throws java.io.IOException, ClassNotFoundException {
1101
1102 // Read in all fields
1103 s.defaultReadObject();
1104
1105 // Initialize counts
1106 capturingGroupCount = 1;
1107 localCount = 0;
1108
1109 // if length > 0, the Pattern is lazily compiled
1110 compiled = false;
1111 if (pattern.length() == 0) {
1112 root = new Start(lastAccept);
1113 matchRoot = lastAccept;
1114 compiled = true;
1115 }
1116 }
1117
1118 /**
1119 * This private constructor is used to create all Patterns. The pattern
1120 * string and match flags are all that is needed to completely describe
1121 * a Pattern. An empty pattern string results in an object tree with
1122 * only a Start node and a LastNode node.
1123 */
1124 private Pattern(String p, int f) {
1125 pattern = p;
1126 flags = f;
1127
1128 // Reset group index count
1129 capturingGroupCount = 1;
1130 localCount = 0;
1131
1132 if (pattern.length() > 0) {
1133 compile();
1134 } else {
1135 root = new Start(lastAccept);
1136 matchRoot = lastAccept;
1137 }
1138 }
1139
1140 /**
1141 * The pattern is converted to normalizedD form and then a pure group
1142 * is constructed to match canonical equivalences of the characters.
1143 */
1144 private void normalize() {
1145 boolean inCharClass = false;
1146 int lastCodePoint = -1;
1147
1148 // Convert pattern into normalizedD form
1149 normalizedPattern = Normalizer.normalize(pattern, Normalizer.Form.NFD);
1150 patternLength = normalizedPattern.length();
1151
1152 // Modify pattern to match canonical equivalences
1153 StringBuilder newPattern = new StringBuilder(patternLength);
1154 for(int i=0; i<patternLength; ) {
1155 int c = normalizedPattern.codePointAt(i);
1156 StringBuilder sequenceBuffer;
1157 if ((Character.getType(c) == Character.NON_SPACING_MARK)
1158 && (lastCodePoint != -1)) {
1159 sequenceBuffer = new StringBuilder();
1160 sequenceBuffer.appendCodePoint(lastCodePoint);
1161 sequenceBuffer.appendCodePoint(c);
1162 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1163 i += Character.charCount(c);
1164 if (i >= patternLength)
1165 break;
1166 c = normalizedPattern.codePointAt(i);
1167 sequenceBuffer.appendCodePoint(c);
1168 }
1169 String ea = produceEquivalentAlternation(
1170 sequenceBuffer.toString());
1171 newPattern.setLength(newPattern.length()-Character.charCount(lastCodePoint));
1172 newPattern.append("(?:").append(ea).append(")");
1173 } else if (c == '[' && lastCodePoint != '\\') {
1174 i = normalizeCharClass(newPattern, i);
1175 } else {
1176 newPattern.appendCodePoint(c);
1177 }
1178 lastCodePoint = c;
1179 i += Character.charCount(c);
1180 }
1181 normalizedPattern = newPattern.toString();
1182 }
1183
1184 /**
1185 * Complete the character class being parsed and add a set
1186 * of alternations to it that will match the canonical equivalences
1187 * of the characters within the class.
1188 */
1189 private int normalizeCharClass(StringBuilder newPattern, int i) {
1190 StringBuilder charClass = new StringBuilder();
1191 StringBuilder eq = null;
1192 int lastCodePoint = -1;
1193 String result;
1194
1195 i++;
1196 charClass.append("[");
1197 while(true) {
1198 int c = normalizedPattern.codePointAt(i);
1199 StringBuilder sequenceBuffer;
1200
1201 if (c == ']' && lastCodePoint != '\\') {
1202 charClass.append((char)c);
1203 break;
1204 } else if (Character.getType(c) == Character.NON_SPACING_MARK) {
1205 sequenceBuffer = new StringBuilder();
1206 sequenceBuffer.appendCodePoint(lastCodePoint);
1207 while(Character.getType(c) == Character.NON_SPACING_MARK) {
1208 sequenceBuffer.appendCodePoint(c);
1209 i += Character.charCount(c);
1210 if (i >= normalizedPattern.length())
1211 break;
1212 c = normalizedPattern.codePointAt(i);
1213 }
1214 String ea = produceEquivalentAlternation(
1215 sequenceBuffer.toString());
1216
1217 charClass.setLength(charClass.length()-Character.charCount(lastCodePoint));
1218 if (eq == null)
1219 eq = new StringBuilder();
1220 eq.append('|');
1221 eq.append(ea);
1222 } else {
1223 charClass.appendCodePoint(c);
1224 i++;
1225 }
1226 if (i == normalizedPattern.length())
1227 throw error("Unclosed character class");
1228 lastCodePoint = c;
1229 }
1230
1231 if (eq != null) {
1232 result = "(?:"+charClass.toString()+eq.toString()+")";
1233 } else {
1234 result = charClass.toString();
1235 }
1236
1237 newPattern.append(result);
1238 return i;
1239 }
1240
1241 /**
1242 * Given a specific sequence composed of a regular character and
1243 * combining marks that follow it, produce the alternation that will
1244 * match all canonical equivalences of that sequence.
1245 */
1246 private String produceEquivalentAlternation(String source) {
1247 int len = countChars(source, 0, 1);
1248 if (source.length() == len)
1249 // source has one character.
1250 return source;
1251
1252 String base = source.substring(0,len);
1253 String combiningMarks = source.substring(len);
1254
1255 String[] perms = producePermutations(combiningMarks);
1256 StringBuilder result = new StringBuilder(source);
1257
1258 // Add combined permutations
1259 for(int x=0; x<perms.length; x++) {
1260 String next = base + perms[x];
1261 if (x>0)
1262 result.append("|"+next);
1263 next = composeOneStep(next);
1264 if (next != null)
1265 result.append("|"+produceEquivalentAlternation(next));
1266 }
1267 return result.toString();
1268 }
1269
1270 /**
1271 * Returns an array of strings that have all the possible
1272 * permutations of the characters in the input string.
1273 * This is used to get a list of all possible orderings
1274 * of a set of combining marks. Note that some of the permutations
1275 * are invalid because of combining class collisions, and these
1276 * possibilities must be removed because they are not canonically
1277 * equivalent.
1278 */
1279 private String[] producePermutations(String input) {
1280 if (input.length() == countChars(input, 0, 1))
1281 return new String[] {input};
1282
1283 if (input.length() == countChars(input, 0, 2)) {
1284 int c0 = Character.codePointAt(input, 0);
1285 int c1 = Character.codePointAt(input, Character.charCount(c0));
1286 if (getClass(c1) == getClass(c0)) {
1287 return new String[] {input};
1288 }
1289 String[] result = new String[2];
1290 result[0] = input;
1291 StringBuilder sb = new StringBuilder(2);
1292 sb.appendCodePoint(c1);
1293 sb.appendCodePoint(c0);
1294 result[1] = sb.toString();
1295 return result;
1296 }
1297
1298 int length = 1;
1299 int nCodePoints = countCodePoints(input);
1300 for(int x=1; x<nCodePoints; x++)
1301 length = length * (x+1);
1302
1303 String[] temp = new String[length];
1304
1305 int combClass[] = new int[nCodePoints];
1306 for(int x=0, i=0; x<nCodePoints; x++) {
1307 int c = Character.codePointAt(input, i);
1308 combClass[x] = getClass(c);
1309 i += Character.charCount(c);
1310 }
1311
1312 // For each char, take it out and add the permutations
1313 // of the remaining chars
1314 int index = 0;
1315 int len;
1316 // offset maintains the index in code units.
1317loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) {
1318 len = countChars(input, offset, 1);
1319 boolean skip = false;
1320 for(int y=x-1; y>=0; y--) {
1321 if (combClass[y] == combClass[x]) {
1322 continue loop;
1323 }
1324 }
1325 StringBuilder sb = new StringBuilder(input);
1326 String otherChars = sb.delete(offset, offset+len).toString();
1327 String[] subResult = producePermutations(otherChars);
1328
1329 String prefix = input.substring(offset, offset+len);
1330 for(int y=0; y<subResult.length; y++)
1331 temp[index++] = prefix + subResult[y];
1332 }
1333 String[] result = new String[index];
1334 for (int x=0; x<index; x++)
1335 result[x] = temp[x];
1336 return result;
1337 }
1338
1339 private int getClass(int c) {
1340 return sun.text.Normalizer.getCombiningClass(c);
1341 }
1342
1343 /**
1344 * Attempts to compose input by combining the first character
1345 * with the first combining mark following it. Returns a String
1346 * that is the composition of the leading character with its first
1347 * combining mark followed by the remaining combining marks. Returns
1348 * null if the first two characters cannot be further composed.
1349 */
1350 private String composeOneStep(String input) {
1351 int len = countChars(input, 0, 2);
1352 String firstTwoCharacters = input.substring(0, len);
1353 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC);
1354
1355 if (result.equals(firstTwoCharacters))
1356 return null;
1357 else {
1358 String remainder = input.substring(len);
1359 return result + remainder;
1360 }
1361 }
1362
1363 /**
1364 * Preprocess any \Q...\E sequences in `temp', meta-quoting them.
1365 * See the description of `quotemeta' in perlfunc(1).
1366 */
1367 private void RemoveQEQuoting() {
1368 final int pLen = patternLength;
1369 int i = 0;
1370 while (i < pLen-1) {
1371 if (temp[i] != '\\')
1372 i += 1;
1373 else if (temp[i + 1] != 'Q')
1374 i += 2;
1375 else
1376 break;
1377 }
1378 if (i >= pLen - 1) // No \Q sequence found
1379 return;
1380 int j = i;
1381 i += 2;
1382 int[] newtemp = new int[j + 2*(pLen-i) + 2];
1383 System.arraycopy(temp, 0, newtemp, 0, j);
1384
1385 boolean inQuote = true;
1386 while (i < pLen) {
1387 int c = temp[i++];
1388 if (! ASCII.isAscii(c) || ASCII.isAlnum(c)) {
1389 newtemp[j++] = c;
1390 } else if (c != '\\') {
1391 if (inQuote) newtemp[j++] = '\\';
1392 newtemp[j++] = c;
1393 } else if (inQuote) {
1394 if (temp[i] == 'E') {
1395 i++;
1396 inQuote = false;
1397 } else {
1398 newtemp[j++] = '\\';
1399 newtemp[j++] = '\\';
1400 }
1401 } else {
1402 if (temp[i] == 'Q') {
1403 i++;
1404 inQuote = true;
1405 } else {
1406 newtemp[j++] = c;
1407 if (i != pLen)
1408 newtemp[j++] = temp[i++];
1409 }
1410 }
1411 }
1412
1413 patternLength = j;
1414 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination
1415 }
1416
1417 /**
1418 * Copies regular expression to an int array and invokes the parsing
1419 * of the expression which will create the object tree.
1420 */
1421 private void compile() {
1422 // Handle canonical equivalences
1423 if (has(CANON_EQ) && !has(LITERAL)) {
1424 normalize();
1425 } else {
1426 normalizedPattern = pattern;
1427 }
1428 patternLength = normalizedPattern.length();
1429
1430 // Copy pattern to int array for convenience
1431 // Use double zero to terminate pattern
1432 temp = new int[patternLength + 2];
1433
1434 boolean hasSupplementary = false;
1435 int c, count = 0;
1436 // Convert all chars into code points
1437 for (int x = 0; x < patternLength; x += Character.charCount(c)) {
1438 c = normalizedPattern.codePointAt(x);
1439 if (isSupplementary(c)) {
1440 hasSupplementary = true;
1441 }
1442 temp[count++] = c;
1443 }
1444
1445 patternLength = count; // patternLength now in code points
1446
1447 if (! has(LITERAL))
1448 RemoveQEQuoting();
1449
1450 // Allocate all temporary objects here.
1451 buffer = new int[32];
1452 groupNodes = new GroupHead[10];
1453
1454 if (has(LITERAL)) {
1455 // Literal pattern handling
1456 matchRoot = newSlice(temp, patternLength, hasSupplementary);
1457 matchRoot.next = lastAccept;
1458 } else {
1459 // Start recursive descent parsing
1460 matchRoot = expr(lastAccept);
1461 // Check extra pattern characters
1462 if (patternLength != cursor) {
1463 if (peek() == ')') {
1464 throw error("Unmatched closing ')'");
1465 } else {
1466 throw error("Unexpected internal error");
1467 }
1468 }
1469 }
1470
1471 // Peephole optimization
1472 if (matchRoot instanceof Slice) {
1473 root = BnM.optimize(matchRoot);
1474 if (root == matchRoot) {
1475 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1476 }
1477 } else if (matchRoot instanceof Begin || matchRoot instanceof First) {
1478 root = matchRoot;
1479 } else {
1480 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1481 }
1482
1483 // Release temporary storage
1484 temp = null;
1485 buffer = null;
1486 groupNodes = null;
1487 patternLength = 0;
1488 compiled = true;
1489 }
1490
1491 /**
1492 * Used to print out a subtree of the Pattern to help with debugging.
1493 */
1494 private static void printObjectTree(Node node) {
1495 while(node != null) {
1496 if (node instanceof Prolog) {
1497 System.out.println(node);
1498 printObjectTree(((Prolog)node).loop);
1499 System.out.println("**** end contents prolog loop");
1500 } else if (node instanceof Loop) {
1501 System.out.println(node);
1502 printObjectTree(((Loop)node).body);
1503 System.out.println("**** end contents Loop body");
1504 } else if (node instanceof Curly) {
1505 System.out.println(node);
1506 printObjectTree(((Curly)node).atom);
1507 System.out.println("**** end contents Curly body");
1508 } else if (node instanceof GroupCurly) {
1509 System.out.println(node);
1510 printObjectTree(((GroupCurly)node).atom);
1511 System.out.println("**** end contents GroupCurly body");
1512 } else if (node instanceof GroupTail) {
1513 System.out.println(node);
1514 System.out.println("Tail next is "+node.next);
1515 return;
1516 } else {
1517 System.out.println(node);
1518 }
1519 node = node.next;
1520 if (node != null)
1521 System.out.println("->next:");
1522 if (node == Pattern.accept) {
1523 System.out.println("Accept Node");
1524 node = null;
1525 }
1526 }
1527 }
1528
1529 /**
1530 * Used to accumulate information about a subtree of the object graph
1531 * so that optimizations can be applied to the subtree.
1532 */
1533 static final class TreeInfo {
1534 int minLength;
1535 int maxLength;
1536 boolean maxValid;
1537 boolean deterministic;
1538
1539 TreeInfo() {
1540 reset();
1541 }
1542 void reset() {
1543 minLength = 0;
1544 maxLength = 0;
1545 maxValid = true;
1546 deterministic = true;
1547 }
1548 }
1549
1550 /*
1551 * The following private methods are mainly used to improve the
1552 * readability of the code. In order to let the Java compiler easily
1553 * inline them, we should not put many assertions or error checks in them.
1554 */
1555
1556 /**
1557 * Indicates whether a particular flag is set or not.
1558 */
1559 private boolean has(int f) {
1560 return (flags & f) != 0;
1561 }
1562
1563 /**
1564 * Match next character, signal error if failed.
1565 */
1566 private void accept(int ch, String s) {
1567 int testChar = temp[cursor++];
1568 if (has(COMMENTS))
1569 testChar = parsePastWhitespace(testChar);
1570 if (ch != testChar) {
1571 throw error(s);
1572 }
1573 }
1574
1575 /**
1576 * Mark the end of pattern with a specific character.
1577 */
1578 private void mark(int c) {
1579 temp[patternLength] = c;
1580 }
1581
1582 /**
1583 * Peek the next character, and do not advance the cursor.
1584 */
1585 private int peek() {
1586 int ch = temp[cursor];
1587 if (has(COMMENTS))
1588 ch = peekPastWhitespace(ch);
1589 return ch;
1590 }
1591
1592 /**
1593 * Read the next character, and advance the cursor by one.
1594 */
1595 private int read() {
1596 int ch = temp[cursor++];
1597 if (has(COMMENTS))
1598 ch = parsePastWhitespace(ch);
1599 return ch;
1600 }
1601
1602 /**
1603 * Read the next character, and advance the cursor by one,
1604 * ignoring the COMMENTS setting
1605 */
1606 private int readEscaped() {
1607 int ch = temp[cursor++];
1608 return ch;
1609 }
1610
1611 /**
1612 * Advance the cursor by one, and peek the next character.
1613 */
1614 private int next() {
1615 int ch = temp[++cursor];
1616 if (has(COMMENTS))
1617 ch = peekPastWhitespace(ch);
1618 return ch;
1619 }
1620
1621 /**
1622 * Advance the cursor by one, and peek the next character,
1623 * ignoring the COMMENTS setting
1624 */
1625 private int nextEscaped() {
1626 int ch = temp[++cursor];
1627 return ch;
1628 }
1629
1630 /**
1631 * If in xmode peek past whitespace and comments.
1632 */
1633 private int peekPastWhitespace(int ch) {
1634 while (ASCII.isSpace(ch) || ch == '#') {
1635 while (ASCII.isSpace(ch))
1636 ch = temp[++cursor];
1637 if (ch == '#') {
1638 ch = peekPastLine();
1639 }
1640 }
1641 return ch;
1642 }
1643
1644 /**
1645 * If in xmode parse past whitespace and comments.
1646 */
1647 private int parsePastWhitespace(int ch) {
1648 while (ASCII.isSpace(ch) || ch == '#') {
1649 while (ASCII.isSpace(ch))
1650 ch = temp[cursor++];
1651 if (ch == '#')
1652 ch = parsePastLine();
1653 }
1654 return ch;
1655 }
1656
1657 /**
1658 * xmode parse past comment to end of line.
1659 */
1660 private int parsePastLine() {
1661 int ch = temp[cursor++];
1662 while (ch != 0 && !isLineSeparator(ch))
1663 ch = temp[cursor++];
1664 return ch;
1665 }
1666
1667 /**
1668 * xmode peek past comment to end of line.
1669 */
1670 private int peekPastLine() {
1671 int ch = temp[++cursor];
1672 while (ch != 0 && !isLineSeparator(ch))
1673 ch = temp[++cursor];
1674 return ch;
1675 }
1676
1677 /**
1678 * Determines if character is a line separator in the current mode
1679 */
1680 private boolean isLineSeparator(int ch) {
1681 if (has(UNIX_LINES)) {
1682 return ch == '\n';
1683 } else {
1684 return (ch == '\n' ||
1685 ch == '\r' ||
1686 (ch|1) \u2029= 'โฉ' ||
1687 ch \u0085= 'ย
');
1688 }
1689 }
1690
1691 /**
1692 * Read the character after the next one, and advance the cursor by two.
1693 */
1694 private int skip() {
1695 int i = cursor;
1696 int ch = temp[i+1];
1697 cursor = i + 2;
1698 return ch;
1699 }
1700
1701 /**
1702 * Unread one next character, and retreat cursor by one.
1703 */
1704 private void unread() {
1705 cursor--;
1706 }
1707
1708 /**
1709 * Internal method used for handling all syntax errors. The pattern is
1710 * displayed with a pointer to aid in locating the syntax error.
1711 */
1712 private PatternSyntaxException error(String s) {
1713 return new PatternSyntaxException(s, normalizedPattern, cursor - 1);
1714 }
1715
1716 /**
1717 * Determines if there is any supplementary character or unpaired
1718 * surrogate in the specified range.
1719 */
1720 private boolean findSupplementary(int start, int end) {
1721 for (int i = start; i < end; i++) {
1722 if (isSupplementary(temp[i]))
1723 return true;
1724 }
1725 return false;
1726 }
1727
1728 /**
1729 * Determines if the specified code point is a supplementary
1730 * character or unpaired surrogate.
1731 */
1732 private static final boolean isSupplementary(int ch) {
1733 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT || isSurrogate(ch);
1734 }
1735
1736 /**
1737 * The following methods handle the main parsing. They are sorted
1738 * according to their precedence order, the lowest one first.
1739 */
1740
1741 /**
1742 * The expression is parsed with branch nodes added for alternations.
1743 * This may be called recursively to parse sub expressions that may
1744 * contain alternations.
1745 */
1746 private Node expr(Node end) {
1747 Node prev = null;
1748 Node firstTail = null;
1749 Node branchConn = null;
1750
1751 for (;;) {
1752 Node node = sequence(end);
1753 Node nodeTail = root; //double return
1754 if (prev == null) {
1755 prev = node;
1756 firstTail = nodeTail;
1757 } else {
1758 // Branch
1759 if (branchConn == null) {
1760 branchConn = new BranchConn();
1761 branchConn.next = end;
1762 }
1763 if (node == end) {
1764 // if the node returned from sequence() is "end"
1765 // we have an empty expr, set a null atom into
1766 // the branch to indicate to go "next" directly.
1767 node = null;
1768 } else {
1769 // the "tail.next" of each atom goes to branchConn
1770 nodeTail.next = branchConn;
1771 }
1772 if (prev instanceof Branch) {
1773 ((Branch)prev).add(node);
1774 } else {
1775 if (prev == end) {
1776 prev = null;
1777 } else {
1778 // replace the "end" with "branchConn" at its tail.next
1779 // when put the "prev" into the branch as the first atom.
1780 firstTail.next = branchConn;
1781 }
1782 prev = new Branch(prev, node, branchConn);
1783 }
1784 }
1785 if (peek() != '|') {
1786 return prev;
1787 }
1788 next();
1789 }
1790 }
1791
1792 /**
1793 * Parsing of sequences between alternations.
1794 */
1795 private Node sequence(Node end) {
1796 Node head = null;
1797 Node tail = null;
1798 Node node = null;
1799 LOOP:
1800 for (;;) {
1801 int ch = peek();
1802 switch (ch) {
1803 case '(':
1804 // Because group handles its own closure,
1805 // we need to treat it differently
1806 node = group0();
1807 // Check for comment or flag group
1808 if (node == null)
1809 continue;
1810 if (head == null)
1811 head = node;
1812 else
1813 tail.next = node;
1814 // Double return: Tail was returned in root
1815 tail = root;
1816 continue;
1817 case '[':
1818 node = clazz(true);
1819 break;
1820 case '\\':
1821 ch = nextEscaped();
1822 if (ch == 'p' || ch == 'P') {
1823 boolean oneLetter = true;
1824 boolean comp = (ch == 'P');
1825 ch = next(); // Consume { if present
1826 if (ch != '{') {
1827 unread();
1828 } else {
1829 oneLetter = false;
1830 }
1831 node = family(oneLetter).maybeComplement(comp);
1832 } else {
1833 unread();
1834 node = atom();
1835 }
1836 break;
1837 case '^':
1838 next();
1839 if (has(MULTILINE)) {
1840 if (has(UNIX_LINES))
1841 node = new UnixCaret();
1842 else
1843 node = new Caret();
1844 } else {
1845 node = new Begin();
1846 }
1847 break;
1848 case '$':
1849 next();
1850 if (has(UNIX_LINES))
1851 node = new UnixDollar(has(MULTILINE));
1852 else
1853 node = new Dollar(has(MULTILINE));
1854 break;
1855 case '.':
1856 next();
1857 if (has(DOTALL)) {
1858 node = new All();
1859 } else {
1860 if (has(UNIX_LINES))
1861 node = new UnixDot();
1862 else {
1863 node = new Dot();
1864 }
1865 }
1866 break;
1867 case '|':
1868 case ')':
1869 break LOOP;
1870 case ']': // Now interpreting dangling ] and } as literals
1871 case '}':
1872 node = atom();
1873 break;
1874 case '?':
1875 case '*':
1876 case '+':
1877 next();
1878 throw error("Dangling meta character '" + ((char)ch) + "'");
1879 case 0:
1880 if (cursor >= patternLength) {
1881 break LOOP;
1882 }
1883 // Fall through
1884 default:
1885 node = atom();
1886 break;
1887 }
1888
1889 node = closure(node);
1890
1891 if (head == null) {
1892 head = tail = node;
1893 } else {
1894 tail.next = node;
1895 tail = node;
1896 }
1897 }
1898 if (head == null) {
1899 return end;
1900 }
1901 tail.next = end;
1902 root = tail; //double return
1903 return head;
1904 }
1905
1906 /**
1907 * Parse and add a new Single or Slice.
1908 */
1909 private Node atom() {
1910 int first = 0;
1911 int prev = -1;
1912 boolean hasSupplementary = false;
1913 int ch = peek();
1914 for (;;) {
1915 switch (ch) {
1916 case '*':
1917 case '+':
1918 case '?':
1919 case '{':
1920 if (first > 1) {
1921 cursor = prev; // Unwind one character
1922 first--;
1923 }
1924 break;
1925 case '$':
1926 case '.':
1927 case '^':
1928 case '(':
1929 case '[':
1930 case '|':
1931 case ')':
1932 break;
1933 case '\\':
1934 ch = nextEscaped();
1935 if (ch == 'p' || ch == 'P') { // Property
1936 if (first > 0) { // Slice is waiting; handle it first
1937 unread();
1938 break;
1939 } else { // No slice; just return the family node
1940 boolean comp = (ch == 'P');
1941 boolean oneLetter = true;
1942 ch = next(); // Consume { if present
1943 if (ch != '{')
1944 unread();
1945 else
1946 oneLetter = false;
1947 return family(oneLetter).maybeComplement(comp);
1948 }
1949 }
1950 unread();
1951 prev = cursor;
1952 ch = escape(false, first == 0);
1953 if (ch >= 0) {
1954 append(ch, first);
1955 first++;
1956 if (isSupplementary(ch)) {
1957 hasSupplementary = true;
1958 }
1959 ch = peek();
1960 continue;
1961 } else if (first == 0) {
1962 return root;
1963 }
1964 // Unwind meta escape sequence
1965 cursor = prev;
1966 break;
1967 case 0:
1968 if (cursor >= patternLength) {
1969 break;
1970 }
1971 // Fall through
1972 default:
1973 prev = cursor;
1974 append(ch, first);
1975 first++;
1976 if (isSupplementary(ch)) {
1977 hasSupplementary = true;
1978 }
1979 ch = next();
1980 continue;
1981 }
1982 break;
1983 }
1984 if (first == 1) {
1985 return newSingle(buffer[0]);
1986 } else {
1987 return newSlice(buffer, first, hasSupplementary);
1988 }
1989 }
1990
1991 private void append(int ch, int len) {
1992 if (len >= buffer.length) {
1993 int[] tmp = new int[len+len];
1994 System.arraycopy(buffer, 0, tmp, 0, len);
1995 buffer = tmp;
1996 }
1997 buffer[len] = ch;
1998 }
1999
2000 /**
2001 * Parses a backref greedily, taking as many numbers as it
2002 * can. The first digit is always treated as a backref, but
2003 * multi digit numbers are only treated as a backref if at
2004 * least that many backrefs exist at this point in the regex.
2005 */
2006 private Node ref(int refNum) {
2007 boolean done = false;
2008 while(!done) {
2009 int ch = peek();
2010 switch(ch) {
2011 case '0':
2012 case '1':
2013 case '2':
2014 case '3':
2015 case '4':
2016 case '5':
2017 case '6':
2018 case '7':
2019 case '8':
2020 case '9':
2021 int newRefNum = (refNum * 10) + (ch - '0');
2022 // Add another number if it doesn't make a group
2023 // that doesn't exist
2024 if (capturingGroupCount - 1 < newRefNum) {
2025 done = true;
2026 break;
2027 }
2028 refNum = newRefNum;
2029 read();
2030 break;
2031 default:
2032 done = true;
2033 break;
2034 }
2035 }
2036 if (has(CASE_INSENSITIVE))
2037 return new CIBackRef(refNum, has(UNICODE_CASE));
2038 else
2039 return new BackRef(refNum);
2040 }
2041
2042 /**
2043 * Parses an escape sequence to determine the actual value that needs
2044 * to be matched.
2045 * If -1 is returned and create was true a new object was added to the tree
2046 * to handle the escape sequence.
2047 * If the returned value is greater than zero, it is the value that
2048 * matches the escape sequence.
2049 */
2050 private int escape(boolean inclass, boolean create) {
2051 int ch = skip();
2052 switch (ch) {
2053 case '0':
2054 return o();
2055 case '1':
2056 case '2':
2057 case '3':
2058 case '4':
2059 case '5':
2060 case '6':
2061 case '7':
2062 case '8':
2063 case '9':
2064 if (inclass) break;
2065 if (create) {
2066 root = ref((ch - '0'));
2067 }
2068 return -1;
2069 case 'A':
2070 if (inclass) break;
2071 if (create) root = new Begin();
2072 return -1;
2073 case 'B':
2074 if (inclass) break;
2075 if (create) root = new Bound(Bound.NONE);
2076 return -1;
2077 case 'C':
2078 break;
2079 case 'D':
2080 if (create) root = new Ctype(ASCII.DIGIT).complement();
2081 return -1;
2082 case 'E':
2083 case 'F':
2084 break;
2085 case 'G':
2086 if (inclass) break;
2087 if (create) root = new LastMatch();
2088 return -1;
2089 case 'H':
2090 case 'I':
2091 case 'J':
2092 case 'K':
2093 case 'L':
2094 case 'M':
2095 case 'N':
2096 case 'O':
2097 case 'P':
2098 case 'Q':
2099 case 'R':
2100 break;
2101 case 'S':
2102 if (create) root = new Ctype(ASCII.SPACE).complement();
2103 return -1;
2104 case 'T':
2105 case 'U':
2106 case 'V':
2107 break;
2108 case 'W':
2109 if (create) root = new Ctype(ASCII.WORD).complement();
2110 return -1;
2111 case 'X':
2112 case 'Y':
2113 break;
2114 case 'Z':
2115 if (inclass) break;
2116 if (create) {
2117 if (has(UNIX_LINES))
2118 root = new UnixDollar(false);
2119 else
2120 root = new Dollar(false);
2121 }
2122 return -1;
2123 case 'a':
2124 return '\007';
2125 case 'b':
2126 if (inclass) break;
2127 if (create) root = new Bound(Bound.BOTH);
2128 return -1;
2129 case 'c':
2130 return c();
2131 case 'd':
2132 if (create) root = new Ctype(ASCII.DIGIT);
2133 return -1;
2134 case 'e':
2135 return '\033';
2136 case 'f':
2137 return '\f';
2138 case 'g':
2139 case 'h':
2140 case 'i':
2141 case 'j':
2142 case 'k':
2143 case 'l':
2144 case 'm':
2145 break;
2146 case 'n':
2147 return '\n';
2148 case 'o':
2149 case 'p':
2150 case 'q':
2151 break;
2152 case 'r':
2153 return '\r';
2154 case 's':
2155 if (create) root = new Ctype(ASCII.SPACE);
2156 return -1;
2157 case 't':
2158 return '\t';
2159 case 'u':
2160 return u();
2161 case 'v':
2162 return '\013';
2163 case 'w':
2164 if (create) root = new Ctype(ASCII.WORD);
2165 return -1;
2166 case 'x':
2167 return x();
2168 case 'y':
2169 break;
2170 case 'z':
2171 if (inclass) break;
2172 if (create) root = new End();
2173 return -1;
2174 default:
2175 return ch;
2176 }
2177 throw error("Illegal/unsupported escape sequence");
2178 }
2179
2180 /**
2181 * Parse a character class, and return the node that matches it.
2182 *
2183 * Consumes a ] on the way out if consume is true. Usually consume
2184 * is true except for the case of [abc&&def] where def is a separate
2185 * right hand node with "understood" brackets.
2186 */
2187 private CharProperty clazz(boolean consume) {
2188 CharProperty prev = null;
2189 CharProperty node = null;
2190 BitClass bits = new BitClass();
2191 boolean include = true;
2192 boolean firstInClass = true;
2193 int ch = next();
2194 for (;;) {
2195 switch (ch) {
2196 case '^':
2197 // Negates if first char in a class, otherwise literal
2198 if (firstInClass) {
2199 if (temp[cursor-1] != '[')
2200 break;
2201 ch = next();
2202 include = !include;
2203 continue;
2204 } else {
2205 // ^ not first in class, treat as literal
2206 break;
2207 }
2208 case '[':
2209 firstInClass = false;
2210 node = clazz(true);
2211 if (prev == null)
2212 prev = node;
2213 else
2214 prev = union(prev, node);
2215 ch = peek();
2216 continue;
2217 case '&':
2218 firstInClass = false;
2219 ch = next();
2220 if (ch == '&') {
2221 ch = next();
2222 CharProperty rightNode = null;
2223 while (ch != ']' && ch != '&') {
2224 if (ch == '[') {
2225 if (rightNode == null)
2226 rightNode = clazz(true);
2227 else
2228 rightNode = union(rightNode, clazz(true));
2229 } else { // abc&&def
2230 unread();
2231 rightNode = clazz(false);
2232 }
2233 ch = peek();
2234 }
2235 if (rightNode != null)
2236 node = rightNode;
2237 if (prev == null) {
2238 if (rightNode == null)
2239 throw error("Bad class syntax");
2240 else
2241 prev = rightNode;
2242 } else {
2243 prev = intersection(prev, node);
2244 }
2245 } else {
2246 // treat as a literal &
2247 unread();
2248 break;
2249 }
2250 continue;
2251 case 0:
2252 firstInClass = false;
2253 if (cursor >= patternLength)
2254 throw error("Unclosed character class");
2255 break;
2256 case ']':
2257 firstInClass = false;
2258 if (prev != null) {
2259 if (consume)
2260 next();
2261 return prev;
2262 }
2263 break;
2264 default:
2265 firstInClass = false;
2266 break;
2267 }
2268 node = range(bits);
2269 if (include) {
2270 if (prev == null) {
2271 prev = node;
2272 } else {
2273 if (prev != node)
2274 prev = union(prev, node);
2275 }
2276 } else {
2277 if (prev == null) {
2278 prev = node.complement();
2279 } else {
2280 if (prev != node)
2281 prev = setDifference(prev, node);
2282 }
2283 }
2284 ch = peek();
2285 }
2286 }
2287
2288 private CharProperty bitsOrSingle(BitClass bits, int ch) {
2289 /* Bits can only handle codepoints in [u+0000-u+00ff] range.
2290 Use "single" node instead of bits when dealing with unicode
2291 case folding for codepoints listed below.
2292 (1)Uppercase out of range: u+00ff, u+00b5
2293 toUpperCase(u+00ff) -> u+0178
2294 toUpperCase(u+00b5) -> u+039c
2295 (2)LatinSmallLetterLongS u+17f
2296 toUpperCase(u+017f) -> u+0053
2297 (3)LatinSmallLetterDotlessI u+131
2298 toUpperCase(u+0131) -> u+0049
2299 (4)LatinCapitalLetterIWithDotAbove u+0130
2300 toLowerCase(u+0130) -> u+0069
2301 (5)KelvinSign u+212a
2302 toLowerCase(u+212a) ==> u+006B
2303 (6)AngstromSign u+212b
2304 toLowerCase(u+212b) ==> u+00e5
2305 */
2306 int d;
2307 if (ch < 256 &&
2308 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) &&
2309 (ch == 0xff || ch == 0xb5 ||
2310 ch == 0x49 || ch == 0x69 || //I and i
2311 ch == 0x53 || ch == 0x73 || //S and s
2312 ch == 0x4b || ch == 0x6b || //K and k
2313 ch == 0xc5 || ch == 0xe5))) //A+ring
2314 return bits.add(ch, flags());
2315 return newSingle(ch);
2316 }
2317
2318 /**
2319 * Parse a single character or a character range in a character class
2320 * and return its representative node.
2321 */
2322 private CharProperty range(BitClass bits) {
2323 int ch = peek();
2324 if (ch == '\\') {
2325 ch = nextEscaped();
2326 if (ch == 'p' || ch == 'P') { // A property
2327 boolean comp = (ch == 'P');
2328 boolean oneLetter = true;
2329 // Consume { if present
2330 ch = next();
2331 if (ch != '{')
2332 unread();
2333 else
2334 oneLetter = false;
2335 return family(oneLetter).maybeComplement(comp);
2336 } else { // ordinary escape
2337 unread();
2338 ch = escape(true, true);
2339 if (ch == -1)
2340 return (CharProperty) root;
2341 }
2342 } else {
2343 ch = single();
2344 }
2345 if (ch >= 0) {
2346 if (peek() == '-') {
2347 int endRange = temp[cursor+1];
2348 if (endRange == '[') {
2349 return bitsOrSingle(bits, ch);
2350 }
2351 if (endRange != ']') {
2352 next();
2353 int m = single();
2354 if (m < ch)
2355 throw error("Illegal character range");
2356 if (has(CASE_INSENSITIVE))
2357 return caseInsensitiveRangeFor(ch, m);
2358 else
2359 return rangeFor(ch, m);
2360 }
2361 }
2362 return bitsOrSingle(bits, ch);
2363 }
2364 throw error("Unexpected character '"+((char)ch)+"'");
2365 }
2366
2367 private int single() {
2368 int ch = peek();
2369 switch (ch) {
2370 case '\\':
2371 return escape(true, false);
2372 default:
2373 next();
2374 return ch;
2375 }
2376 }
2377
2378 /**
2379 * Parses a Unicode character family and returns its representative node.
2380 */
2381 private CharProperty family(boolean singleLetter) {
2382 next();
2383 String name;
2384
2385 if (singleLetter) {
2386 int c = temp[cursor];
2387 if (!Character.isSupplementaryCodePoint(c)) {
2388 name = String.valueOf((char)c);
2389 } else {
2390 name = new String(temp, cursor, 1);
2391 }
2392 read();
2393 } else {
2394 int i = cursor;
2395 mark('}');
2396 while(read() != '}') {
2397 }
2398 mark('\000');
2399 int j = cursor;
2400 if (j > patternLength)
2401 throw error("Unclosed character family");
2402 if (i + 1 >= j)
2403 throw error("Empty character family");
2404 name = new String(temp, i, j-i-1);
2405 }
2406
2407 if (name.startsWith("In")) {
2408 return unicodeBlockPropertyFor(name.substring(2));
2409 } else {
2410 if (name.startsWith("Is"))
2411 name = name.substring(2);
2412 return charPropertyNodeFor(name);
2413 }
2414 }
2415
2416 /**
2417 * Returns a CharProperty matching all characters in a UnicodeBlock.
2418 */
2419 private CharProperty unicodeBlockPropertyFor(String name) {
2420 final Character.UnicodeBlock block;
2421 try {
2422 block = Character.UnicodeBlock.forName(name);
2423 } catch (IllegalArgumentException iae) {
2424 throw error("Unknown character block name {" + name + "}");
2425 }
2426 return new CharProperty() {
2427 boolean isSatisfiedBy(int ch) {
2428 return block == Character.UnicodeBlock.of(ch);}};
2429 }
2430
2431 /**
2432 * Returns a CharProperty matching all characters in a named property.
2433 */
2434 private CharProperty charPropertyNodeFor(String name) {
2435 CharProperty p = CharPropertyNames.charPropertyFor(name);
2436 if (p == null)
2437 throw error("Unknown character property name {" + name + "}");
2438 return p;
2439 }
2440
2441 /**
2442 * Parses a group and returns the head node of a set of nodes that process
2443 * the group. Sometimes a double return system is used where the tail is
2444 * returned in root.
2445 */
2446 private Node group0() {
2447 boolean capturingGroup = false;
2448 Node head = null;
2449 Node tail = null;
2450 int save = flags;
2451 root = null;
2452 int ch = next();
2453 if (ch == '?') {
2454 ch = skip();
2455 switch (ch) {
2456 case ':': // (?:xxx) pure group
2457 head = createGroup(true);
2458 tail = root;
2459 head.next = expr(tail);
2460 break;
2461 case '=': // (?=xxx) and (?!xxx) lookahead
2462 case '!':
2463 head = createGroup(true);
2464 tail = root;
2465 head.next = expr(tail);
2466 if (ch == '=') {
2467 head = tail = new Pos(head);
2468 } else {
2469 head = tail = new Neg(head);
2470 }
2471 break;
2472 case '>': // (?>xxx) independent group
2473 head = createGroup(true);
2474 tail = root;
2475 head.next = expr(tail);
2476 head = tail = new Ques(head, INDEPENDENT);
2477 break;
2478 case '<': // (?<xxx) look behind
2479 ch = read();
2480 int start = cursor;
2481 head = createGroup(true);
2482 tail = root;
2483 head.next = expr(tail);
2484 tail.next = lookbehindEnd;
2485 TreeInfo info = new TreeInfo();
2486 head.study(info);
2487 if (info.maxValid == false) {
2488 throw error("Look-behind group does not have "
2489 + "an obvious maximum length");
2490 }
2491 boolean hasSupplementary = findSupplementary(start, patternLength);
2492 if (ch == '=') {
2493 head = tail = (hasSupplementary ?
2494 new BehindS(head, info.maxLength,
2495 info.minLength) :
2496 new Behind(head, info.maxLength,
2497 info.minLength));
2498 } else if (ch == '!') {
2499 head = tail = (hasSupplementary ?
2500 new NotBehindS(head, info.maxLength,
2501 info.minLength) :
2502 new NotBehind(head, info.maxLength,
2503 info.minLength));
2504 } else {
2505 throw error("Unknown look-behind group");
2506 }
2507 break;
2508 case '$':
2509 case '@':
2510 throw error("Unknown group type");
2511 default: // (?xxx:) inlined match flags
2512 unread();
2513 addFlag();
2514 ch = read();
2515 if (ch == ')') {
2516 return null; // Inline modifier only
2517 }
2518 if (ch != ':') {
2519 throw error("Unknown inline modifier");
2520 }
2521 head = createGroup(true);
2522 tail = root;
2523 head.next = expr(tail);
2524 break;
2525 }
2526 } else { // (xxx) a regular group
2527 capturingGroup = true;
2528 head = createGroup(false);
2529 tail = root;
2530 head.next = expr(tail);
2531 }
2532
2533 accept(')', "Unclosed group");
2534 flags = save;
2535
2536 // Check for quantifiers
2537 Node node = closure(head);
2538 if (node == head) { // No closure
2539 root = tail;
2540 return node; // Dual return
2541 }
2542 if (head == tail) { // Zero length assertion
2543 root = node;
2544 return node; // Dual return
2545 }
2546
2547 if (node instanceof Ques) {
2548 Ques ques = (Ques) node;
2549 if (ques.type == POSSESSIVE) {
2550 root = node;
2551 return node;
2552 }
2553 tail.next = new BranchConn();
2554 tail = tail.next;
2555 if (ques.type == GREEDY) {
2556 head = new Branch(head, null, tail);
2557 } else { // Reluctant quantifier
2558 head = new Branch(null, head, tail);
2559 }
2560 root = tail;
2561 return head;
2562 } else if (node instanceof Curly) {
2563 Curly curly = (Curly) node;
2564 if (curly.type == POSSESSIVE) {
2565 root = node;
2566 return node;
2567 }
2568 // Discover if the group is deterministic
2569 TreeInfo info = new TreeInfo();
2570 if (head.study(info)) { // Deterministic
2571 GroupTail temp = (GroupTail) tail;
2572 head = root = new GroupCurly(head.next, curly.cmin,
2573 curly.cmax, curly.type,
2574 ((GroupTail)tail).localIndex,
2575 ((GroupTail)tail).groupIndex,
2576 capturingGroup);
2577 return head;
2578 } else { // Non-deterministic
2579 int temp = ((GroupHead) head).localIndex;
2580 Loop loop;
2581 if (curly.type == GREEDY)
2582 loop = new Loop(this.localCount, temp);
2583 else // Reluctant Curly
2584 loop = new LazyLoop(this.localCount, temp);
2585 Prolog prolog = new Prolog(loop);
2586 this.localCount += 1;
2587 loop.cmin = curly.cmin;
2588 loop.cmax = curly.cmax;
2589 loop.body = head;
2590 tail.next = loop;
2591 root = loop;
2592 return prolog; // Dual return
2593 }
2594 }
2595 throw error("Internal logic error");
2596 }
2597
2598 /**
2599 * Create group head and tail nodes using double return. If the group is
2600 * created with anonymous true then it is a pure group and should not
2601 * affect group counting.
2602 */
2603 private Node createGroup(boolean anonymous) {
2604 int localIndex = localCount++;
2605 int groupIndex = 0;
2606 if (!anonymous)
2607 groupIndex = capturingGroupCount++;
2608 GroupHead head = new GroupHead(localIndex);
2609 root = new GroupTail(localIndex, groupIndex);
2610 if (!anonymous && groupIndex < 10)
2611 groupNodes[groupIndex] = head;
2612 return head;
2613 }
2614
2615 /**
2616 * Parses inlined match flags and set them appropriately.
2617 */
2618 private void addFlag() {
2619 int ch = peek();
2620 for (;;) {
2621 switch (ch) {
2622 case 'i':
2623 flags |= CASE_INSENSITIVE;
2624 break;
2625 case 'm':
2626 flags |= MULTILINE;
2627 break;
2628 case 's':
2629 flags |= DOTALL;
2630 break;
2631 case 'd':
2632 flags |= UNIX_LINES;
2633 break;
2634 case 'u':
2635 flags |= UNICODE_CASE;
2636 break;
2637 case 'c':
2638 flags |= CANON_EQ;
2639 break;
2640 case 'x':
2641 flags |= COMMENTS;
2642 break;
2643 case '-': // subFlag then fall through
2644 ch = next();
2645 subFlag();
2646 default:
2647 return;
2648 }
2649 ch = next();
2650 }
2651 }
2652
2653 /**
2654 * Parses the second part of inlined match flags and turns off
2655 * flags appropriately.
2656 */
2657 private void subFlag() {
2658 int ch = peek();
2659 for (;;) {
2660 switch (ch) {
2661 case 'i':
2662 flags &= ~CASE_INSENSITIVE;
2663 break;
2664 case 'm':
2665 flags &= ~MULTILINE;
2666 break;
2667 case 's':
2668 flags &= ~DOTALL;
2669 break;
2670 case 'd':
2671 flags &= ~UNIX_LINES;
2672 break;
2673 case 'u':
2674 flags &= ~UNICODE_CASE;
2675 break;
2676 case 'c':
2677 flags &= ~CANON_EQ;
2678 break;
2679 case 'x':
2680 flags &= ~COMMENTS;
2681 break;
2682 default:
2683 return;
2684 }
2685 ch = next();
2686 }
2687 }
2688
2689 static final int MAX_REPS = 0x7FFFFFFF;
2690
2691 static final int GREEDY = 0;
2692
2693 static final int LAZY = 1;
2694
2695 static final int POSSESSIVE = 2;
2696
2697 static final int INDEPENDENT = 3;
2698
2699 /**
2700 * Processes repetition. If the next character peeked is a quantifier
2701 * then new nodes must be appended to handle the repetition.
2702 * Prev could be a single or a group, so it could be a chain of nodes.
2703 */
2704 private Node closure(Node prev) {
2705 Node atom;
2706 int ch = peek();
2707 switch (ch) {
2708 case '?':
2709 ch = next();
2710 if (ch == '?') {
2711 next();
2712 return new Ques(prev, LAZY);
2713 } else if (ch == '+') {
2714 next();
2715 return new Ques(prev, POSSESSIVE);
2716 }
2717 return new Ques(prev, GREEDY);
2718 case '*':
2719 ch = next();
2720 if (ch == '?') {
2721 next();
2722 return new Curly(prev, 0, MAX_REPS, LAZY);
2723 } else if (ch == '+') {
2724 next();
2725 return new Curly(prev, 0, MAX_REPS, POSSESSIVE);
2726 }
2727 return new Curly(prev, 0, MAX_REPS, GREEDY);
2728 case '+':
2729 ch = next();
2730 if (ch == '?') {
2731 next();
2732 return new Curly(prev, 1, MAX_REPS, LAZY);
2733 } else if (ch == '+') {
2734 next();
2735 return new Curly(prev, 1, MAX_REPS, POSSESSIVE);
2736 }
2737 return new Curly(prev, 1, MAX_REPS, GREEDY);
2738 case '{':
2739 ch = temp[cursor+1];
2740 if (ASCII.isDigit(ch)) {
2741 skip();
2742 int cmin = 0;
2743 do {
2744 cmin = cmin * 10 + (ch - '0');
2745 } while (ASCII.isDigit(ch = read()));
2746 int cmax = cmin;
2747 if (ch == ',') {
2748 ch = read();
2749 cmax = MAX_REPS;
2750 if (ch != '}') {
2751 cmax = 0;
2752 while (ASCII.isDigit(ch)) {
2753 cmax = cmax * 10 + (ch - '0');
2754 ch = read();
2755 }
2756 }
2757 }
2758 if (ch != '}')
2759 throw error("Unclosed counted closure");
2760 if (((cmin) | (cmax) | (cmax - cmin)) < 0)
2761 throw error("Illegal repetition range");
2762 Curly curly;
2763 ch = peek();
2764 if (ch == '?') {
2765 next();
2766 curly = new Curly(prev, cmin, cmax, LAZY);
2767 } else if (ch == '+') {
2768 next();
2769 curly = new Curly(prev, cmin, cmax, POSSESSIVE);
2770 } else {
2771 curly = new Curly(prev, cmin, cmax, GREEDY);
2772 }
2773 return curly;
2774 } else {
2775 throw error("Illegal repetition");
2776 }
2777 default:
2778 return prev;
2779 }
2780 }
2781
2782 /**
2783 * Utility method for parsing control escape sequences.
2784 */
2785 private int c() {
2786 if (cursor < patternLength) {
2787 return read() ^ 64;
2788 }
2789 throw error("Illegal control escape sequence");
2790 }
2791
2792 /**
2793 * Utility method for parsing octal escape sequences.
2794 */
2795 private int o() {
2796 int n = read();
2797 if (((n-'0')|('7'-n)) >= 0) {
2798 int m = read();
2799 if (((m-'0')|('7'-m)) >= 0) {
2800 int o = read();
2801 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) {
2802 return (n - '0') * 64 + (m - '0') * 8 + (o - '0');
2803 }
2804 unread();
2805 return (n - '0') * 8 + (m - '0');
2806 }
2807 unread();
2808 return (n - '0');
2809 }
2810 throw error("Illegal octal escape sequence");
2811 }
2812
2813 /**
2814 * Utility method for parsing hexadecimal escape sequences.
2815 */
2816 private int x() {
2817 int n = read();
2818 if (ASCII.isHexDigit(n)) {
2819 int m = read();
2820 if (ASCII.isHexDigit(m)) {
2821 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m);
2822 }
2823 }
2824 throw error("Illegal hexadecimal escape sequence");
2825 }
2826
2827 /**
2828 * Utility method for parsing unicode escape sequences.
2829 */
2830 private int u() {
2831 int n = 0;
2832 for (int i = 0; i < 4; i++) {
2833 int ch = read();
2834 if (!ASCII.isHexDigit(ch)) {
2835 throw error("Illegal Unicode escape sequence");
2836 }
2837 n = n * 16 + ASCII.toDigit(ch);
2838 }
2839 return n;
2840 }
2841
2842 //
2843 // Utility methods for code point support
2844 //
2845
2846 /**
2847 * Tests a surrogate value.
2848 */
2849 private static final boolean isSurrogate(int c) {
2850 return c >= Character.MIN_HIGH_SURROGATE && c <= Character.MAX_LOW_SURROGATE;
2851 }
2852
2853 private static final int countChars(CharSequence seq, int index,
2854 int lengthInCodePoints) {
2855 // optimization
2856 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) {
2857 assert (index >= 0 && index < seq.length());
2858 return 1;
2859 }
2860 int length = seq.length();
2861 int x = index;
2862 if (lengthInCodePoints >= 0) {
2863 assert (index >= 0 && index < length);
2864 for (int i = 0; x < length && i < lengthInCodePoints; i++) {
2865 if (Character.isHighSurrogate(seq.charAt(x++))) {
2866 if (x < length && Character.isLowSurrogate(seq.charAt(x))) {
2867 x++;
2868 }
2869 }
2870 }
2871 return x - index;
2872 }
2873
2874 assert (index >= 0 && index <= length);
2875 if (index == 0) {
2876 return 0;
2877 }
2878 int len = -lengthInCodePoints;
2879 for (int i = 0; x > 0 && i < len; i++) {
2880 if (Character.isLowSurrogate(seq.charAt(--x))) {
2881 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) {
2882 x--;
2883 }
2884 }
2885 }
2886 return index - x;
2887 }
2888
2889 private static final int countCodePoints(CharSequence seq) {
2890 int length = seq.length();
2891 int n = 0;
2892 for (int i = 0; i < length; ) {
2893 n++;
2894 if (Character.isHighSurrogate(seq.charAt(i++))) {
2895 if (i < length && Character.isLowSurrogate(seq.charAt(i))) {
2896 i++;
2897 }
2898 }
2899 }
2900 return n;
2901 }
2902
2903 /**
2904 * Creates a bit vector for matching Latin-1 values. A normal BitClass
2905 * never matches values above Latin-1, and a complemented BitClass always
2906 * matches values above Latin-1.
2907 */
2908 private static final class BitClass extends BmpCharProperty {
2909 final boolean[] bits;
2910 BitClass() { bits = new boolean[256]; }
2911 private BitClass(boolean[] bits) { this.bits = bits; }
2912 BitClass add(int c, int flags) {
2913 assert c >= 0 && c <= 255;
2914 if ((flags & CASE_INSENSITIVE) != 0) {
2915 if (ASCII.isAscii(c)) {
2916 bits[ASCII.toUpper(c)] = true;
2917 bits[ASCII.toLower(c)] = true;
2918 } else if ((flags & UNICODE_CASE) != 0) {
2919 bits[Character.toLowerCase(c)] = true;
2920 bits[Character.toUpperCase(c)] = true;
2921 }
2922 }
2923 bits[c] = true;
2924 return this;
2925 }
2926 boolean isSatisfiedBy(int ch) {
2927 return ch < 256 && bits[ch];
2928 }
2929 }
2930
2931 /**
2932 * Returns a suitably optimized, single character matcher.
2933 */
2934 private CharProperty newSingle(final int ch) {
2935 if (has(CASE_INSENSITIVE)) {
2936 int lower, upper;
2937 if (has(UNICODE_CASE)) {
2938 upper = Character.toUpperCase(ch);
2939 lower = Character.toLowerCase(upper);
2940 if (upper != lower)
2941 return new SingleU(lower);
2942 } else if (ASCII.isAscii(ch)) {
2943 lower = ASCII.toLower(ch);
2944 upper = ASCII.toUpper(ch);
2945 if (lower != upper)
2946 return new SingleI(lower, upper);
2947 }
2948 }
2949 if (isSupplementary(ch))
2950 return new SingleS(ch); // Match a given Unicode character
2951 return new Single(ch); // Match a given BMP character
2952 }
2953
2954 /**
2955 * Utility method for creating a string slice matcher.
2956 */
2957 private Node newSlice(int[] buf, int count, boolean hasSupplementary) {
2958 int[] tmp = new int[count];
2959 if (has(CASE_INSENSITIVE)) {
2960 if (has(UNICODE_CASE)) {
2961 for (int i = 0; i < count; i++) {
2962 tmp[i] = Character.toLowerCase(
2963 Character.toUpperCase(buf[i]));
2964 }
2965 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp);
2966 }
2967 for (int i = 0; i < count; i++) {
2968 tmp[i] = ASCII.toLower(buf[i]);
2969 }
2970 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp);
2971 }
2972 for (int i = 0; i < count; i++) {
2973 tmp[i] = buf[i];
2974 }
2975 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp);
2976 }
2977
2978 /**
2979 * The following classes are the building components of the object
2980 * tree that represents a compiled regular expression. The object tree
2981 * is made of individual elements that handle constructs in the Pattern.
2982 * Each type of object knows how to match its equivalent construct with
2983 * the match() method.
2984 */
2985
2986 /**
2987 * Base class for all node classes. Subclasses should override the match()
2988 * method as appropriate. This class is an accepting node, so its match()
2989 * always returns true.
2990 */
2991 static class Node extends Object {
2992 Node next;
2993 Node() {
2994 next = Pattern.accept;
2995 }
2996 /**
2997 * This method implements the classic accept node.
2998 */
2999 boolean match(Matcher matcher, int i, CharSequence seq) {
3000 matcher.last = i;
3001 matcher.groups[0] = matcher.first;
3002 matcher.groups[1] = matcher.last;
3003 return true;
3004 }
3005 /**
3006 * This method is good for all zero length assertions.
3007 */
3008 boolean study(TreeInfo info) {
3009 if (next != null) {
3010 return next.study(info);
3011 } else {
3012 return info.deterministic;
3013 }
3014 }
3015 }
3016
3017 static class LastNode extends Node {
3018 /**
3019 * This method implements the classic accept node with
3020 * the addition of a check to see if the match occurred
3021 * using all of the input.
3022 */
3023 boolean match(Matcher matcher, int i, CharSequence seq) {
3024 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to)
3025 return false;
3026 matcher.last = i;
3027 matcher.groups[0] = matcher.first;
3028 matcher.groups[1] = matcher.last;
3029 return true;
3030 }
3031 }
3032
3033 /**
3034 * Used for REs that can start anywhere within the input string.
3035 * This basically tries to match repeatedly at each spot in the
3036 * input string, moving forward after each try. An anchored search
3037 * or a BnM will bypass this node completely.
3038 */
3039 static class Start extends Node {
3040 int minLength;
3041 Start(Node node) {
3042 this.next = node;
3043 TreeInfo info = new TreeInfo();
3044 next.study(info);
3045 minLength = info.minLength;
3046 }
3047 boolean match(Matcher matcher, int i, CharSequence seq) {
3048 if (i > matcher.to - minLength) {
3049 matcher.hitEnd = true;
3050 return false;
3051 }
3052 boolean ret = false;
3053 int guard = matcher.to - minLength;
3054 for (; i <= guard; i++) {
3055 if (ret = next.match(matcher, i, seq))
3056 break;
3057 if (i == guard)
3058 matcher.hitEnd = true;
3059 }
3060 if (ret) {
3061 matcher.first = i;
3062 matcher.groups[0] = matcher.first;
3063 matcher.groups[1] = matcher.last;
3064 }
3065 return ret;
3066 }
3067 boolean study(TreeInfo info) {
3068 next.study(info);
3069 info.maxValid = false;
3070 info.deterministic = false;
3071 return false;
3072 }
3073 }
3074
3075 /*
3076 * StartS supports supplementary characters, including unpaired surrogates.
3077 */
3078 static final class StartS extends Start {
3079 StartS(Node node) {
3080 super(node);
3081 }
3082 boolean match(Matcher matcher, int i, CharSequence seq) {
3083 if (i > matcher.to - minLength) {
3084 matcher.hitEnd = true;
3085 return false;
3086 }
3087 boolean ret = false;
3088 int guard = matcher.to - minLength;
3089 while (i <= guard) {
3090 if ((ret = next.match(matcher, i, seq)) || i == guard)
3091 break;
3092 // Optimization to move to the next character. This is
3093 // faster than countChars(seq, i, 1).
3094 if (Character.isHighSurrogate(seq.charAt(i++))) {
3095 if (i < seq.length() && Character.isLowSurrogate(seq.charAt(i))) {
3096 i++;
3097 }
3098 }
3099 if (i == guard)
3100 matcher.hitEnd = true;
3101 }
3102 if (ret) {
3103 matcher.first = i;
3104 matcher.groups[0] = matcher.first;
3105 matcher.groups[1] = matcher.last;
3106 }
3107 return ret;
3108 }
3109 }
3110
3111 /**
3112 * Node to anchor at the beginning of input. This object implements the
3113 * match for a \A sequence, and the caret anchor will use this if not in
3114 * multiline mode.
3115 */
3116 static final class Begin extends Node {
3117 boolean match(Matcher matcher, int i, CharSequence seq) {
3118 int fromIndex = (matcher.anchoringBounds) ?
3119 matcher.from : 0;
3120 if (i == fromIndex && next.match(matcher, i, seq)) {
3121 matcher.first = i;
3122 matcher.groups[0] = i;
3123 matcher.groups[1] = matcher.last;
3124 return true;
3125 } else {
3126 return false;
3127 }
3128 }
3129 }
3130
3131 /**
3132 * Node to anchor at the end of input. This is the absolute end, so this
3133 * should not match at the last newline before the end as $ will.
3134 */
3135 static final class End extends Node {
3136 boolean match(Matcher matcher, int i, CharSequence seq) {
3137 int endIndex = (matcher.anchoringBounds) ?
3138 matcher.to : matcher.getTextLength();
3139 if (i == endIndex) {
3140 matcher.hitEnd = true;
3141 return next.match(matcher, i, seq);
3142 }
3143 return false;
3144 }
3145 }
3146
3147 /**
3148 * Node to anchor at the beginning of a line. This is essentially the
3149 * object to match for the multiline ^.
3150 */
3151 static final class Caret extends Node {
3152 boolean match(Matcher matcher, int i, CharSequence seq) {
3153 int startIndex = matcher.from;
3154 int endIndex = matcher.to;
3155 if (!matcher.anchoringBounds) {
3156 startIndex = 0;
3157 endIndex = matcher.getTextLength();
3158 }
3159 // Perl does not match ^ at end of input even after newline
3160 if (i == endIndex) {
3161 matcher.hitEnd = true;
3162 return false;
3163 }
3164 if (i > startIndex) {
3165 char ch = seq.charAt(i-1);
3166 if (ch != '\n' && ch != '\r'
3167 && (ch|1) \u2029= 'โฉ'
3168 && ch \u0085= 'ย
' ) {
3169 return false;
3170 }
3171 // Should treat /r/n as one newline
3172 if (ch == '\r' && seq.charAt(i) == '\n')
3173 return false;
3174 }
3175 return next.match(matcher, i, seq);
3176 }
3177 }
3178
3179 /**
3180 * Node to anchor at the beginning of a line when in unixdot mode.
3181 */
3182 static final class UnixCaret extends Node {
3183 boolean match(Matcher matcher, int i, CharSequence seq) {
3184 int startIndex = matcher.from;
3185 int endIndex = matcher.to;
3186 if (!matcher.anchoringBounds) {
3187 startIndex = 0;
3188 endIndex = matcher.getTextLength();
3189 }
3190 // Perl does not match ^ at end of input even after newline
3191 if (i == endIndex) {
3192 matcher.hitEnd = true;
3193 return false;
3194 }
3195 if (i > startIndex) {
3196 char ch = seq.charAt(i-1);
3197 if (ch != '\n') {
3198 return false;
3199 }
3200 }
3201 return next.match(matcher, i, seq);
3202 }
3203 }
3204
3205 /**
3206 * Node to match the location where the last match ended.
3207 * This is used for the \G construct.
3208 */
3209 static final class LastMatch extends Node {
3210 boolean match(Matcher matcher, int i, CharSequence seq) {
3211 if (i != matcher.oldLast)
3212 return false;
3213 return next.match(matcher, i, seq);
3214 }
3215 }
3216
3217 /**
3218 * Node to anchor at the end of a line or the end of input based on the
3219 * multiline mode.
3220 *
3221 * When not in multiline mode, the $ can only match at the very end
3222 * of the input, unless the input ends in a line terminator in which
3223 * it matches right before the last line terminator.
3224 *
3225 * Note that \r\n is considered an atomic line terminator.
3226 *
3227 * Like ^ the $ operator matches at a position, it does not match the
3228 * line terminators themselves.
3229 */
3230 static final class Dollar extends Node {
3231 boolean multiline;
3232 Dollar(boolean mul) {
3233 multiline = mul;
3234 }
3235 boolean match(Matcher matcher, int i, CharSequence seq) {
3236 int endIndex = (matcher.anchoringBounds) ?
3237 matcher.to : matcher.getTextLength();
3238 if (!multiline) {
3239 if (i < endIndex - 2)
3240 return false;
3241 if (i == endIndex - 2) {
3242 char ch = seq.charAt(i);
3243 if (ch != '\r')
3244 return false;
3245 ch = seq.charAt(i + 1);
3246 if (ch != '\n')
3247 return false;
3248 }
3249 }
3250 // Matches before any line terminator; also matches at the
3251 // end of input
3252 // Before line terminator:
3253 // If multiline, we match here no matter what
3254 // If not multiline, fall through so that the end
3255 // is marked as hit; this must be a /r/n or a /n
3256 // at the very end so the end was hit; more input
3257 // could make this not match here
3258 if (i < endIndex) {
3259 char ch = seq.charAt(i);
3260 if (ch == '\n') {
3261 // No match between \r\n
3262 if (i > 0 && seq.charAt(i-1) == '\r')
3263 return false;
3264 if (multiline)
3265 return next.match(matcher, i, seq);
3266 } else if (ch == '\r' || ch \u0085= 'ย
' ||
3267 (ch|1) \u2029= 'โฉ') {
3268 if (multiline)
3269 return next.match(matcher, i, seq);
3270 } else { // No line terminator, no match
3271 return false;
3272 }
3273 }
3274 // Matched at current end so hit end
3275 matcher.hitEnd = true;
3276 // If a $ matches because of end of input, then more input
3277 // could cause it to fail!
3278 matcher.requireEnd = true;
3279 return next.match(matcher, i, seq);
3280 }
3281 boolean study(TreeInfo info) {
3282 next.study(info);
3283 return info.deterministic;
3284 }
3285 }
3286
3287 /**
3288 * Node to anchor at the end of a line or the end of input based on the
3289 * multiline mode when in unix lines mode.
3290 */
3291 static final class UnixDollar extends Node {
3292 boolean multiline;
3293 UnixDollar(boolean mul) {
3294 multiline = mul;
3295 }
3296 boolean match(Matcher matcher, int i, CharSequence seq) {
3297 int endIndex = (matcher.anchoringBounds) ?
3298 matcher.to : matcher.getTextLength();
3299 if (i < endIndex) {
3300 char ch = seq.charAt(i);
3301 if (ch == '\n') {
3302 // If not multiline, then only possible to
3303 // match at very end or one before end
3304 if (multiline == false && i != endIndex - 1)
3305 return false;
3306 // If multiline return next.match without setting
3307 // matcher.hitEnd
3308 if (multiline)
3309 return next.match(matcher, i, seq);
3310 } else {
3311 return false;
3312 }
3313 }
3314 // Matching because at the end or 1 before the end;
3315 // more input could change this so set hitEnd
3316 matcher.hitEnd = true;
3317 // If a $ matches because of end of input, then more input
3318 // could cause it to fail!
3319 matcher.requireEnd = true;
3320 return next.match(matcher, i, seq);
3321 }
3322 boolean study(TreeInfo info) {
3323 next.study(info);
3324 return info.deterministic;
3325 }
3326 }
3327
3328 /**
3329 * Abstract node class to match one character satisfying some
3330 * boolean property.
3331 */
3332 private static abstract class CharProperty extends Node {
3333 abstract boolean isSatisfiedBy(int ch);
3334 CharProperty complement() {
3335 return new CharProperty() {
3336 boolean isSatisfiedBy(int ch) {
3337 return ! CharProperty.this.isSatisfiedBy(ch);}};
3338 }
3339 CharProperty maybeComplement(boolean complement) {
3340 return complement ? complement() : this;
3341 }
3342 boolean match(Matcher matcher, int i, CharSequence seq) {
3343 if (i < matcher.to) {
3344 int ch = Character.codePointAt(seq, i);
3345 return isSatisfiedBy(ch)
3346 && next.match(matcher, i+Character.charCount(ch), seq);
3347 } else {
3348 matcher.hitEnd = true;
3349 return false;
3350 }
3351 }
3352 boolean study(TreeInfo info) {
3353 info.minLength++;
3354 info.maxLength++;
3355 return next.study(info);
3356 }
3357 }
3358
3359 /**
3360 * Optimized version of CharProperty that works only for
3361 * properties never satisfied by Supplementary characters.
3362 */
3363 private static abstract class BmpCharProperty extends CharProperty {
3364 boolean match(Matcher matcher, int i, CharSequence seq) {
3365 if (i < matcher.to) {
3366 return isSatisfiedBy(seq.charAt(i))
3367 && next.match(matcher, i+1, seq);
3368 } else {
3369 matcher.hitEnd = true;
3370 return false;
3371 }
3372 }
3373 }
3374
3375 /**
3376 * Node class that matches a Supplementary Unicode character
3377 */
3378 static final class SingleS extends CharProperty {
3379 final int c;
3380 SingleS(int c) { this.c = c; }
3381 boolean isSatisfiedBy(int ch) {
3382 return ch == c;
3383 }
3384 }
3385
3386 /**
3387 * Optimization -- matches a given BMP character
3388 */
3389 static final class Single extends BmpCharProperty {
3390 final int c;
3391 Single(int c) { this.c = c; }
3392 boolean isSatisfiedBy(int ch) {
3393 return ch == c;
3394 }
3395 }
3396
3397 /**
3398 * Case insensitive matches a given BMP character
3399 */
3400 static final class SingleI extends BmpCharProperty {
3401 final int lower;
3402 final int upper;
3403 SingleI(int lower, int upper) {
3404 this.lower = lower;
3405 this.upper = upper;
3406 }
3407 boolean isSatisfiedBy(int ch) {
3408 return ch == lower || ch == upper;
3409 }
3410 }
3411
3412 /**
3413 * Unicode case insensitive matches a given Unicode character
3414 */
3415 static final class SingleU extends CharProperty {
3416 final int lower;
3417 SingleU(int lower) {
3418 this.lower = lower;
3419 }
3420 boolean isSatisfiedBy(int ch) {
3421 return lower == ch ||
3422 lower == Character.toLowerCase(Character.toUpperCase(ch));
3423 }
3424 }
3425
3426 /**
3427 * Node class that matches a Unicode category.
3428 */
3429 static final class Category extends CharProperty {
3430 final int typeMask;
3431 Category(int typeMask) { this.typeMask = typeMask; }
3432 boolean isSatisfiedBy(int ch) {
3433 return (typeMask & (1 << Character.getType(ch))) != 0;
3434 }
3435 }
3436
3437 /**
3438 * Node class that matches a POSIX type.
3439 */
3440 static final class Ctype extends BmpCharProperty {
3441 final int ctype;
3442 Ctype(int ctype) { this.ctype = ctype; }
3443 boolean isSatisfiedBy(int ch) {
3444 return ch < 128 && ASCII.isType(ch, ctype);
3445 }
3446 }
3447
3448 /**
3449 * Base class for all Slice nodes
3450 */
3451 static class SliceNode extends Node {
3452 int[] buffer;
3453 SliceNode(int[] buf) {
3454 buffer = buf;
3455 }
3456 boolean study(TreeInfo info) {
3457 info.minLength += buffer.length;
3458 info.maxLength += buffer.length;
3459 return next.study(info);
3460 }
3461 }
3462
3463 /**
3464 * Node class for a case sensitive/BMP-only sequence of literal
3465 * characters.
3466 */
3467 static final class Slice extends SliceNode {
3468 Slice(int[] buf) {
3469 super(buf);
3470 }
3471 boolean match(Matcher matcher, int i, CharSequence seq) {
3472 int[] buf = buffer;
3473 int len = buf.length;
3474 for (int j=0; j<len; j++) {
3475 if ((i+j) >= matcher.to) {
3476 matcher.hitEnd = true;
3477 return false;
3478 }
3479 if (buf[j] != seq.charAt(i+j))
3480 return false;
3481 }
3482 return next.match(matcher, i+len, seq);
3483 }
3484 }
3485
3486 /**
3487 * Node class for a case_insensitive/BMP-only sequence of literal
3488 * characters.
3489 */
3490 static class SliceI extends SliceNode {
3491 SliceI(int[] buf) {
3492 super(buf);
3493 }
3494 boolean match(Matcher matcher, int i, CharSequence seq) {
3495 int[] buf = buffer;
3496 int len = buf.length;
3497 for (int j=0; j<len; j++) {
3498 if ((i+j) >= matcher.to) {
3499 matcher.hitEnd = true;
3500 return false;
3501 }
3502 int c = seq.charAt(i+j);
3503 if (buf[j] != c &&
3504 buf[j] != ASCII.toLower(c))
3505 return false;
3506 }
3507 return next.match(matcher, i+len, seq);
3508 }
3509 }
3510
3511 /**
3512 * Node class for a unicode_case_insensitive/BMP-only sequence of
3513 * literal characters. Uses unicode case folding.
3514 */
3515 static final class SliceU extends SliceNode {
3516 SliceU(int[] buf) {
3517 super(buf);
3518 }
3519 boolean match(Matcher matcher, int i, CharSequence seq) {
3520 int[] buf = buffer;
3521 int len = buf.length;
3522 for (int j=0; j<len; j++) {
3523 if ((i+j) >= matcher.to) {
3524 matcher.hitEnd = true;
3525 return false;
3526 }
3527 int c = seq.charAt(i+j);
3528 if (buf[j] != c &&
3529 buf[j] != Character.toLowerCase(Character.toUpperCase(c)))
3530 return false;
3531 }
3532 return next.match(matcher, i+len, seq);
3533 }
3534 }
3535
3536 /**
3537 * Node class for a case sensitive sequence of literal characters
3538 * including supplementary characters.
3539 */
3540 static final class SliceS extends SliceNode {
3541 SliceS(int[] buf) {
3542 super(buf);
3543 }
3544 boolean match(Matcher matcher, int i, CharSequence seq) {
3545 int[] buf = buffer;
3546 int x = i;
3547 for (int j = 0; j < buf.length; j++) {
3548 if (x >= matcher.to) {
3549 matcher.hitEnd = true;
3550 return false;
3551 }
3552 int c = Character.codePointAt(seq, x);
3553 if (buf[j] != c)
3554 return false;
3555 x += Character.charCount(c);
3556 if (x > matcher.to) {
3557 matcher.hitEnd = true;
3558 return false;
3559 }
3560 }
3561 return next.match(matcher, x, seq);
3562 }
3563 }
3564
3565 /**
3566 * Node class for a case insensitive sequence of literal characters
3567 * including supplementary characters.
3568 */
3569 static class SliceIS extends SliceNode {
3570 SliceIS(int[] buf) {
3571 super(buf);
3572 }
3573 int toLower(int c) {
3574 return ASCII.toLower(c);
3575 }
3576 boolean match(Matcher matcher, int i, CharSequence seq) {
3577 int[] buf = buffer;
3578 int x = i;
3579 for (int j = 0; j < buf.length; j++) {
3580 if (x >= matcher.to) {
3581 matcher.hitEnd = true;
3582 return false;
3583 }
3584 int c = Character.codePointAt(seq, x);
3585 if (buf[j] != c && buf[j] != toLower(c))
3586 return false;
3587 x += Character.charCount(c);
3588 if (x > matcher.to) {
3589 matcher.hitEnd = true;
3590 return false;
3591 }
3592 }
3593 return next.match(matcher, x, seq);
3594 }
3595 }
3596
3597 /**
3598 * Node class for a case insensitive sequence of literal characters.
3599 * Uses unicode case folding.
3600 */
3601 static final class SliceUS extends SliceIS {
3602 SliceUS(int[] buf) {
3603 super(buf);
3604 }
3605 int toLower(int c) {
3606 return Character.toLowerCase(Character.toUpperCase(c));
3607 }
3608 }
3609
3610 private static boolean inRange(int lower, int ch, int upper) {
3611 return lower <= ch && ch <= upper;
3612 }
3613
3614 /**
3615 * Returns node for matching characters within an explicit value range.
3616 */
3617 private static CharProperty rangeFor(final int lower,
3618 final int upper) {
3619 return new CharProperty() {
3620 boolean isSatisfiedBy(int ch) {
3621 return inRange(lower, ch, upper);}};
3622 }
3623
3624 /**
3625 * Returns node for matching characters within an explicit value
3626 * range in a case insensitive manner.
3627 */
3628 private CharProperty caseInsensitiveRangeFor(final int lower,
3629 final int upper) {
3630 if (has(UNICODE_CASE))
3631 return new CharProperty() {
3632 boolean isSatisfiedBy(int ch) {
3633 if (inRange(lower, ch, upper))
3634 return true;
3635 int up = Character.toUpperCase(ch);
3636 return inRange(lower, up, upper) ||
3637 inRange(lower, Character.toLowerCase(up), upper);}};
3638 return new CharProperty() {
3639 boolean isSatisfiedBy(int ch) {
3640 return inRange(lower, ch, upper) ||
3641 ASCII.isAscii(ch) &&
3642 (inRange(lower, ASCII.toUpper(ch), upper) ||
3643 inRange(lower, ASCII.toLower(ch), upper));
3644 }};
3645 }
3646
3647 /**
3648 * Implements the Unicode category ALL and the dot metacharacter when
3649 * in dotall mode.
3650 */
3651 static final class All extends CharProperty {
3652 boolean isSatisfiedBy(int ch) {
3653 return true;
3654 }
3655 }
3656
3657 /**
3658 * Node class for the dot metacharacter when dotall is not enabled.
3659 */
3660 static final class Dot extends CharProperty {
3661 boolean isSatisfiedBy(int ch) {
3662 return (ch != '\n' && ch != '\r'
3663 && (ch|1) \u2029= 'โฉ'
3664 && ch \u0085= 'ย
');
3665 }
3666 }
3667
3668 /**
3669 * Node class for the dot metacharacter when dotall is not enabled
3670 * but UNIX_LINES is enabled.
3671 */
3672 static final class UnixDot extends CharProperty {
3673 boolean isSatisfiedBy(int ch) {
3674 return ch != '\n';
3675 }
3676 }
3677
3678 /**
3679 * The 0 or 1 quantifier. This one class implements all three types.
3680 */
3681 static final class Ques extends Node {
3682 Node atom;
3683 int type;
3684 Ques(Node node, int type) {
3685 this.atom = node;
3686 this.type = type;
3687 }
3688 boolean match(Matcher matcher, int i, CharSequence seq) {
3689 switch (type) {
3690 case GREEDY:
3691 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq))
3692 || next.match(matcher, i, seq);
3693 case LAZY:
3694 return next.match(matcher, i, seq)
3695 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq));
3696 case POSSESSIVE:
3697 if (atom.match(matcher, i, seq)) i = matcher.last;
3698 return next.match(matcher, i, seq);
3699 default:
3700 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq);
3701 }
3702 }
3703 boolean study(TreeInfo info) {
3704 if (type != INDEPENDENT) {
3705 int minL = info.minLength;
3706 atom.study(info);
3707 info.minLength = minL;
3708 info.deterministic = false;
3709 return next.study(info);
3710 } else {
3711 atom.study(info);
3712 return next.study(info);
3713 }
3714 }
3715 }
3716
3717 /**
3718 * Handles the curly-brace style repetition with a specified minimum and
3719 * maximum occurrences. The * quantifier is handled as a special case.
3720 * This class handles the three types.
3721 */
3722 static final class Curly extends Node {
3723 Node atom;
3724 int type;
3725 int cmin;
3726 int cmax;
3727
3728 Curly(Node node, int cmin, int cmax, int type) {
3729 this.atom = node;
3730 this.type = type;
3731 this.cmin = cmin;
3732 this.cmax = cmax;
3733 }
3734 boolean match(Matcher matcher, int i, CharSequence seq) {
3735 int j;
3736 for (j = 0; j < cmin; j++) {
3737 if (atom.match(matcher, i, seq)) {
3738 i = matcher.last;
3739 continue;
3740 }
3741 return false;
3742 }
3743 if (type == GREEDY)
3744 return match0(matcher, i, j, seq);
3745 else if (type == LAZY)
3746 return match1(matcher, i, j, seq);
3747 else
3748 return match2(matcher, i, j, seq);
3749 }
3750 // Greedy match.
3751 // i is the index to start matching at
3752 // j is the number of atoms that have matched
3753 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
3754 if (j >= cmax) {
3755 // We have matched the maximum... continue with the rest of
3756 // the regular expression
3757 return next.match(matcher, i, seq);
3758 }
3759 int backLimit = j;
3760 while (atom.match(matcher, i, seq)) {
3761 // k is the length of this match
3762 int k = matcher.last - i;
3763 if (k == 0) // Zero length match
3764 break;
3765 // Move up index and number matched
3766 i = matcher.last;
3767 j++;
3768 // We are greedy so match as many as we can
3769 while (j < cmax) {
3770 if (!atom.match(matcher, i, seq))
3771 break;
3772 if (i + k != matcher.last) {
3773 if (match0(matcher, matcher.last, j+1, seq))
3774 return true;
3775 break;
3776 }
3777 i += k;
3778 j++;
3779 }
3780 // Handle backing off if match fails
3781 while (j >= backLimit) {
3782 if (next.match(matcher, i, seq))
3783 return true;
3784 i -= k;
3785 j--;
3786 }
3787 return false;
3788 }
3789 return next.match(matcher, i, seq);
3790 }
3791 // Reluctant match. At this point, the minimum has been satisfied.
3792 // i is the index to start matching at
3793 // j is the number of atoms that have matched
3794 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
3795 for (;;) {
3796 // Try finishing match without consuming any more
3797 if (next.match(matcher, i, seq))
3798 return true;
3799 // At the maximum, no match found
3800 if (j >= cmax)
3801 return false;
3802 // Okay, must try one more atom
3803 if (!atom.match(matcher, i, seq))
3804 return false;
3805 // If we haven't moved forward then must break out
3806 if (i == matcher.last)
3807 return false;
3808 // Move up index and number matched
3809 i = matcher.last;
3810 j++;
3811 }
3812 }
3813 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
3814 for (; j < cmax; j++) {
3815 if (!atom.match(matcher, i, seq))
3816 break;
3817 if (i == matcher.last)
3818 break;
3819 i = matcher.last;
3820 }
3821 return next.match(matcher, i, seq);
3822 }
3823 boolean study(TreeInfo info) {
3824 // Save original info
3825 int minL = info.minLength;
3826 int maxL = info.maxLength;
3827 boolean maxV = info.maxValid;
3828 boolean detm = info.deterministic;
3829 info.reset();
3830
3831 atom.study(info);
3832
3833 int temp = info.minLength * cmin + minL;
3834 if (temp < minL) {
3835 temp = 0xFFFFFFF; // arbitrary large number
3836 }
3837 info.minLength = temp;
3838
3839 if (maxV & info.maxValid) {
3840 temp = info.maxLength * cmax + maxL;
3841 info.maxLength = temp;
3842 if (temp < maxL) {
3843 info.maxValid = false;
3844 }
3845 } else {
3846 info.maxValid = false;
3847 }
3848
3849 if (info.deterministic && cmin == cmax)
3850 info.deterministic = detm;
3851 else
3852 info.deterministic = false;
3853
3854 return next.study(info);
3855 }
3856 }
3857
3858 /**
3859 * Handles the curly-brace style repetition with a specified minimum and
3860 * maximum occurrences in deterministic cases. This is an iterative
3861 * optimization over the Prolog and Loop system which would handle this
3862 * in a recursive way. The * quantifier is handled as a special case.
3863 * If capture is true then this class saves group settings and ensures
3864 * that groups are unset when backing off of a group match.
3865 */
3866 static final class GroupCurly extends Node {
3867 Node atom;
3868 int type;
3869 int cmin;
3870 int cmax;
3871 int localIndex;
3872 int groupIndex;
3873 boolean capture;
3874
3875 GroupCurly(Node node, int cmin, int cmax, int type, int local,
3876 int group, boolean capture) {
3877 this.atom = node;
3878 this.type = type;
3879 this.cmin = cmin;
3880 this.cmax = cmax;
3881 this.localIndex = local;
3882 this.groupIndex = group;
3883 this.capture = capture;
3884 }
3885 boolean match(Matcher matcher, int i, CharSequence seq) {
3886 int[] groups = matcher.groups;
3887 int[] locals = matcher.locals;
3888 int save0 = locals[localIndex];
3889 int save1 = 0;
3890 int save2 = 0;
3891
3892 if (capture) {
3893 save1 = groups[groupIndex];
3894 save2 = groups[groupIndex+1];
3895 }
3896
3897 // Notify GroupTail there is no need to setup group info
3898 // because it will be set here
3899 locals[localIndex] = -1;
3900
3901 boolean ret = true;
3902 for (int j = 0; j < cmin; j++) {
3903 if (atom.match(matcher, i, seq)) {
3904 if (capture) {
3905 groups[groupIndex] = i;
3906 groups[groupIndex+1] = matcher.last;
3907 }
3908 i = matcher.last;
3909 } else {
3910 ret = false;
3911 break;
3912 }
3913 }
3914 if (ret) {
3915 if (type == GREEDY) {
3916 ret = match0(matcher, i, cmin, seq);
3917 } else if (type == LAZY) {
3918 ret = match1(matcher, i, cmin, seq);
3919 } else {
3920 ret = match2(matcher, i, cmin, seq);
3921 }
3922 }
3923 if (!ret) {
3924 locals[localIndex] = save0;
3925 if (capture) {
3926 groups[groupIndex] = save1;
3927 groups[groupIndex+1] = save2;
3928 }
3929 }
3930 return ret;
3931 }
3932 // Aggressive group match
3933 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
3934 int[] groups = matcher.groups;
3935 int save0 = 0;
3936 int save1 = 0;
3937 if (capture) {
3938 save0 = groups[groupIndex];
3939 save1 = groups[groupIndex+1];
3940 }
3941 for (;;) {
3942 if (j >= cmax)
3943 break;
3944 if (!atom.match(matcher, i, seq))
3945 break;
3946 int k = matcher.last - i;
3947 if (k <= 0) {
3948 if (capture) {
3949 groups[groupIndex] = i;
3950 groups[groupIndex+1] = i + k;
3951 }
3952 i = i + k;
3953 break;
3954 }
3955 for (;;) {
3956 if (capture) {
3957 groups[groupIndex] = i;
3958 groups[groupIndex+1] = i + k;
3959 }
3960 i = i + k;
3961 if (++j >= cmax)
3962 break;
3963 if (!atom.match(matcher, i, seq))
3964 break;
3965 if (i + k != matcher.last) {
3966 if (match0(matcher, i, j, seq))
3967 return true;
3968 break;
3969 }
3970 }
3971 while (j > cmin) {
3972 if (next.match(matcher, i, seq)) {
3973 if (capture) {
3974 groups[groupIndex+1] = i;
3975 groups[groupIndex] = i - k;
3976 }
3977 i = i - k;
3978 return true;
3979 }
3980 // backing off
3981 if (capture) {
3982 groups[groupIndex+1] = i;
3983 groups[groupIndex] = i - k;
3984 }
3985 i = i - k;
3986 j--;
3987 }
3988 break;
3989 }
3990 if (capture) {
3991 groups[groupIndex] = save0;
3992 groups[groupIndex+1] = save1;
3993 }
3994 return next.match(matcher, i, seq);
3995 }
3996 // Reluctant matching
3997 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
3998 for (;;) {
3999 if (next.match(matcher, i, seq))
4000 return true;
4001 if (j >= cmax)
4002 return false;
4003 if (!atom.match(matcher, i, seq))
4004 return false;
4005 if (i == matcher.last)
4006 return false;
4007 if (capture) {
4008 matcher.groups[groupIndex] = i;
4009 matcher.groups[groupIndex+1] = matcher.last;
4010 }
4011 i = matcher.last;
4012 j++;
4013 }
4014 }
4015 // Possessive matching
4016 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4017 for (; j < cmax; j++) {
4018 if (!atom.match(matcher, i, seq)) {
4019 break;
4020 }
4021 if (capture) {
4022 matcher.groups[groupIndex] = i;
4023 matcher.groups[groupIndex+1] = matcher.last;
4024 }
4025 if (i == matcher.last) {
4026 break;
4027 }
4028 i = matcher.last;
4029 }
4030 return next.match(matcher, i, seq);
4031 }
4032 boolean study(TreeInfo info) {
4033 // Save original info
4034 int minL = info.minLength;
4035 int maxL = info.maxLength;
4036 boolean maxV = info.maxValid;
4037 boolean detm = info.deterministic;
4038 info.reset();
4039
4040 atom.study(info);
4041
4042 int temp = info.minLength * cmin + minL;
4043 if (temp < minL) {
4044 temp = 0xFFFFFFF; // Arbitrary large number
4045 }
4046 info.minLength = temp;
4047
4048 if (maxV & info.maxValid) {
4049 temp = info.maxLength * cmax + maxL;
4050 info.maxLength = temp;
4051 if (temp < maxL) {
4052 info.maxValid = false;
4053 }
4054 } else {
4055 info.maxValid = false;
4056 }
4057
4058 if (info.deterministic && cmin == cmax) {
4059 info.deterministic = detm;
4060 } else {
4061 info.deterministic = false;
4062 }
4063
4064 return next.study(info);
4065 }
4066 }
4067
4068 /**
4069 * A Guard node at the end of each atom node in a Branch. It
4070 * serves the purpose of chaining the "match" operation to
4071 * "next" but not the "study", so we can collect the TreeInfo
4072 * of each atom node without including the TreeInfo of the
4073 * "next".
4074 */
4075 static final class BranchConn extends Node {
4076 BranchConn() {};
4077 boolean match(Matcher matcher, int i, CharSequence seq) {
4078 return next.match(matcher, i, seq);
4079 }
4080 boolean study(TreeInfo info) {
4081 return info.deterministic;
4082 }
4083 }
4084
4085 /**
4086 * Handles the branching of alternations. Note this is also used for
4087 * the ? quantifier to branch between the case where it matches once
4088 * and where it does not occur.
4089 */
4090 static final class Branch extends Node {
4091 Node[] atoms = new Node[2];
4092 int size = 2;
4093 Node conn;
4094 Branch(Node first, Node second, Node branchConn) {
4095 conn = branchConn;
4096 atoms[0] = first;
4097 atoms[1] = second;
4098 }
4099
4100 void add(Node node) {
4101 if (size >= atoms.length) {
4102 Node[] tmp = new Node[atoms.length*2];
4103 System.arraycopy(atoms, 0, tmp, 0, atoms.length);
4104 atoms = tmp;
4105 }
4106 atoms[size++] = node;
4107 }
4108
4109 boolean match(Matcher matcher, int i, CharSequence seq) {
4110 for (int n = 0; n < size; n++) {
4111 if (atoms[n] == null) {
4112 if (conn.next.match(matcher, i, seq))
4113 return true;
4114 } else if (atoms[n].match(matcher, i, seq)) {
4115 return true;
4116 }
4117 }
4118 return false;
4119 }
4120
4121 boolean study(TreeInfo info) {
4122 int minL = info.minLength;
4123 int maxL = info.maxLength;
4124 boolean maxV = info.maxValid;
4125
4126 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num
4127 int maxL2 = -1;
4128 for (int n = 0; n < size; n++) {
4129 info.reset();
4130 if (atoms[n] != null)
4131 atoms[n].study(info);
4132 minL2 = Math.min(minL2, info.minLength);
4133 maxL2 = Math.max(maxL2, info.maxLength);
4134 maxV = (maxV & info.maxValid);
4135 }
4136
4137 minL += minL2;
4138 maxL += maxL2;
4139
4140 info.reset();
4141 conn.next.study(info);
4142
4143 info.minLength += minL;
4144 info.maxLength += maxL;
4145 info.maxValid &= maxV;
4146 info.deterministic = false;
4147 return false;
4148 }
4149 }
4150
4151 /**
4152 * The GroupHead saves the location where the group begins in the locals
4153 * and restores them when the match is done.
4154 *
4155 * The matchRef is used when a reference to this group is accessed later
4156 * in the expression. The locals will have a negative value in them to
4157 * indicate that we do not want to unset the group if the reference
4158 * doesn't match.
4159 */
4160 static final class GroupHead extends Node {
4161 int localIndex;
4162 GroupHead(int localCount) {
4163 localIndex = localCount;
4164 }
4165 boolean match(Matcher matcher, int i, CharSequence seq) {
4166 int save = matcher.locals[localIndex];
4167 matcher.locals[localIndex] = i;
4168 boolean ret = next.match(matcher, i, seq);
4169 matcher.locals[localIndex] = save;
4170 return ret;
4171 }
4172 boolean matchRef(Matcher matcher, int i, CharSequence seq) {
4173 int save = matcher.locals[localIndex];
4174 matcher.locals[localIndex] = ~i; // HACK
4175 boolean ret = next.match(matcher, i, seq);
4176 matcher.locals[localIndex] = save;
4177 return ret;
4178 }
4179 }
4180
4181 /**
4182 * Recursive reference to a group in the regular expression. It calls
4183 * matchRef because if the reference fails to match we would not unset
4184 * the group.
4185 */
4186 static final class GroupRef extends Node {
4187 GroupHead head;
4188 GroupRef(GroupHead head) {
4189 this.head = head;
4190 }
4191 boolean match(Matcher matcher, int i, CharSequence seq) {
4192 return head.matchRef(matcher, i, seq)
4193 && next.match(matcher, matcher.last, seq);
4194 }
4195 boolean study(TreeInfo info) {
4196 info.maxValid = false;
4197 info.deterministic = false;
4198 return next.study(info);
4199 }
4200 }
4201
4202 /**
4203 * The GroupTail handles the setting of group beginning and ending
4204 * locations when groups are successfully matched. It must also be able to
4205 * unset groups that have to be backed off of.
4206 *
4207 * The GroupTail node is also used when a previous group is referenced,
4208 * and in that case no group information needs to be set.
4209 */
4210 static final class GroupTail extends Node {
4211 int localIndex;
4212 int groupIndex;
4213 GroupTail(int localCount, int groupCount) {
4214 localIndex = localCount;
4215 groupIndex = groupCount + groupCount;
4216 }
4217 boolean match(Matcher matcher, int i, CharSequence seq) {
4218 int tmp = matcher.locals[localIndex];
4219 if (tmp >= 0) { // This is the normal group case.
4220 // Save the group so we can unset it if it
4221 // backs off of a match.
4222 int groupStart = matcher.groups[groupIndex];
4223 int groupEnd = matcher.groups[groupIndex+1];
4224
4225 matcher.groups[groupIndex] = tmp;
4226 matcher.groups[groupIndex+1] = i;
4227 if (next.match(matcher, i, seq)) {
4228 return true;
4229 }
4230 matcher.groups[groupIndex] = groupStart;
4231 matcher.groups[groupIndex+1] = groupEnd;
4232 return false;
4233 } else {
4234 // This is a group reference case. We don't need to save any
4235 // group info because it isn't really a group.
4236 matcher.last = i;
4237 return true;
4238 }
4239 }
4240 }
4241
4242 /**
4243 * This sets up a loop to handle a recursive quantifier structure.
4244 */
4245 static final class Prolog extends Node {
4246 Loop loop;
4247 Prolog(Loop loop) {
4248 this.loop = loop;
4249 }
4250 boolean match(Matcher matcher, int i, CharSequence seq) {
4251 return loop.matchInit(matcher, i, seq);
4252 }
4253 boolean study(TreeInfo info) {
4254 return loop.study(info);
4255 }
4256 }
4257
4258 /**
4259 * Handles the repetition count for a greedy Curly. The matchInit
4260 * is called from the Prolog to save the index of where the group
4261 * beginning is stored. A zero length group check occurs in the
4262 * normal match but is skipped in the matchInit.
4263 */
4264 static class Loop extends Node {
4265 Node body;
4266 int countIndex; // local count index in matcher locals
4267 int beginIndex; // group beginning index
4268 int cmin, cmax;
4269 Loop(int countIndex, int beginIndex) {
4270 this.countIndex = countIndex;
4271 this.beginIndex = beginIndex;
4272 }
4273 boolean match(Matcher matcher, int i, CharSequence seq) {
4274 // Avoid infinite loop in zero-length case.
4275 if (i > matcher.locals[beginIndex]) {
4276 int count = matcher.locals[countIndex];
4277
4278 // This block is for before we reach the minimum
4279 // iterations required for the loop to match
4280 if (count < cmin) {
4281 matcher.locals[countIndex] = count + 1;
4282 boolean b = body.match(matcher, i, seq);
4283 // If match failed we must backtrack, so
4284 // the loop count should NOT be incremented
4285 if (!b)
4286 matcher.locals[countIndex] = count;
4287 // Return success or failure since we are under
4288 // minimum
4289 return b;
4290 }
4291 // This block is for after we have the minimum
4292 // iterations required for the loop to match
4293 if (count < cmax) {
4294 matcher.locals[countIndex] = count + 1;
4295 boolean b = body.match(matcher, i, seq);
4296 // If match failed we must backtrack, so
4297 // the loop count should NOT be incremented
4298 if (!b)
4299 matcher.locals[countIndex] = count;
4300 else
4301 return true;
4302 }
4303 }
4304 return next.match(matcher, i, seq);
4305 }
4306 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4307 int save = matcher.locals[countIndex];
4308 boolean ret = false;
4309 if (0 < cmin) {
4310 matcher.locals[countIndex] = 1;
4311 ret = body.match(matcher, i, seq);
4312 } else if (0 < cmax) {
4313 matcher.locals[countIndex] = 1;
4314 ret = body.match(matcher, i, seq);
4315 if (ret == false)
4316 ret = next.match(matcher, i, seq);
4317 } else {
4318 ret = next.match(matcher, i, seq);
4319 }
4320 matcher.locals[countIndex] = save;
4321 return ret;
4322 }
4323 boolean study(TreeInfo info) {
4324 info.maxValid = false;
4325 info.deterministic = false;
4326 return false;
4327 }
4328 }
4329
4330 /**
4331 * Handles the repetition count for a reluctant Curly. The matchInit
4332 * is called from the Prolog to save the index of where the group
4333 * beginning is stored. A zero length group check occurs in the
4334 * normal match but is skipped in the matchInit.
4335 */
4336 static final class LazyLoop extends Loop {
4337 LazyLoop(int countIndex, int beginIndex) {
4338 super(countIndex, beginIndex);
4339 }
4340 boolean match(Matcher matcher, int i, CharSequence seq) {
4341 // Check for zero length group
4342 if (i > matcher.locals[beginIndex]) {
4343 int count = matcher.locals[countIndex];
4344 if (count < cmin) {
4345 matcher.locals[countIndex] = count + 1;
4346 boolean result = body.match(matcher, i, seq);
4347 // If match failed we must backtrack, so
4348 // the loop count should NOT be incremented
4349 if (!result)
4350 matcher.locals[countIndex] = count;
4351 return result;
4352 }
4353 if (next.match(matcher, i, seq))
4354 return true;
4355 if (count < cmax) {
4356 matcher.locals[countIndex] = count + 1;
4357 boolean result = body.match(matcher, i, seq);
4358 // If match failed we must backtrack, so
4359 // the loop count should NOT be incremented
4360 if (!result)
4361 matcher.locals[countIndex] = count;
4362 return result;
4363 }
4364 return false;
4365 }
4366 return next.match(matcher, i, seq);
4367 }
4368 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4369 int save = matcher.locals[countIndex];
4370 boolean ret = false;
4371 if (0 < cmin) {
4372 matcher.locals[countIndex] = 1;
4373 ret = body.match(matcher, i, seq);
4374 } else if (next.match(matcher, i, seq)) {
4375 ret = true;
4376 } else if (0 < cmax) {
4377 matcher.locals[countIndex] = 1;
4378 ret = body.match(matcher, i, seq);
4379 }
4380 matcher.locals[countIndex] = save;
4381 return ret;
4382 }
4383 boolean study(TreeInfo info) {
4384 info.maxValid = false;
4385 info.deterministic = false;
4386 return false;
4387 }
4388 }
4389
4390 /**
4391 * Refers to a group in the regular expression. Attempts to match
4392 * whatever the group referred to last matched.
4393 */
4394 static class BackRef extends Node {
4395 int groupIndex;
4396 BackRef(int groupCount) {
4397 super();
4398 groupIndex = groupCount + groupCount;
4399 }
4400 boolean match(Matcher matcher, int i, CharSequence seq) {
4401 int j = matcher.groups[groupIndex];
4402 int k = matcher.groups[groupIndex+1];
4403
4404 int groupSize = k - j;
4405
4406 // If the referenced group didn't match, neither can this
4407 if (j < 0)
4408 return false;
4409
4410 // If there isn't enough input left no match
4411 if (i + groupSize > matcher.to) {
4412 matcher.hitEnd = true;
4413 return false;
4414 }
4415
4416 // Check each new char to make sure it matches what the group
4417 // referenced matched last time around
4418 for (int index=0; index<groupSize; index++)
4419 if (seq.charAt(i+index) != seq.charAt(j+index))
4420 return false;
4421
4422 return next.match(matcher, i+groupSize, seq);
4423 }
4424 boolean study(TreeInfo info) {
4425 info.maxValid = false;
4426 return next.study(info);
4427 }
4428 }
4429
4430 static class CIBackRef extends Node {
4431 int groupIndex;
4432 boolean doUnicodeCase;
4433 CIBackRef(int groupCount, boolean doUnicodeCase) {
4434 super();
4435 groupIndex = groupCount + groupCount;
4436 this.doUnicodeCase = doUnicodeCase;
4437 }
4438 boolean match(Matcher matcher, int i, CharSequence seq) {
4439 int j = matcher.groups[groupIndex];
4440 int k = matcher.groups[groupIndex+1];
4441
4442 int groupSize = k - j;
4443
4444 // If the referenced group didn't match, neither can this
4445 if (j < 0)
4446 return false;
4447
4448 // If there isn't enough input left no match
4449 if (i + groupSize > matcher.to) {
4450 matcher.hitEnd = true;
4451 return false;
4452 }
4453
4454 // Check each new char to make sure it matches what the group
4455 // referenced matched last time around
4456 int x = i;
4457 for (int index=0; index<groupSize; index++) {
4458 int c1 = Character.codePointAt(seq, x);
4459 int c2 = Character.codePointAt(seq, j);
4460 if (c1 != c2) {
4461 if (doUnicodeCase) {
4462 int cc1 = Character.toUpperCase(c1);
4463 int cc2 = Character.toUpperCase(c2);
4464 if (cc1 != cc2 &&
4465 Character.toLowerCase(cc1) !=
4466 Character.toLowerCase(cc2))
4467 return false;
4468 } else {
4469 if (ASCII.toLower(c1) != ASCII.toLower(c2))
4470 return false;
4471 }
4472 }
4473 x += Character.charCount(c1);
4474 j += Character.charCount(c2);
4475 }
4476
4477 return next.match(matcher, i+groupSize, seq);
4478 }
4479 boolean study(TreeInfo info) {
4480 info.maxValid = false;
4481 return next.study(info);
4482 }
4483 }
4484
4485 /**
4486 * Searches until the next instance of its atom. This is useful for
4487 * finding the atom efficiently without passing an instance of it
4488 * (greedy problem) and without a lot of wasted search time (reluctant
4489 * problem).
4490 */
4491 static final class First extends Node {
4492 Node atom;
4493 First(Node node) {
4494 this.atom = BnM.optimize(node);
4495 }
4496 boolean match(Matcher matcher, int i, CharSequence seq) {
4497 if (atom instanceof BnM) {
4498 return atom.match(matcher, i, seq)
4499 && next.match(matcher, matcher.last, seq);
4500 }
4501 for (;;) {
4502 if (i > matcher.to) {
4503 matcher.hitEnd = true;
4504 return false;
4505 }
4506 if (atom.match(matcher, i, seq)) {
4507 return next.match(matcher, matcher.last, seq);
4508 }
4509 i += countChars(seq, i, 1);
4510 matcher.first++;
4511 }
4512 }
4513 boolean study(TreeInfo info) {
4514 atom.study(info);
4515 info.maxValid = false;
4516 info.deterministic = false;
4517 return next.study(info);
4518 }
4519 }
4520
4521 static final class Conditional extends Node {
4522 Node cond, yes, not;
4523 Conditional(Node cond, Node yes, Node not) {
4524 this.cond = cond;
4525 this.yes = yes;
4526 this.not = not;
4527 }
4528 boolean match(Matcher matcher, int i, CharSequence seq) {
4529 if (cond.match(matcher, i, seq)) {
4530 return yes.match(matcher, i, seq);
4531 } else {
4532 return not.match(matcher, i, seq);
4533 }
4534 }
4535 boolean study(TreeInfo info) {
4536 int minL = info.minLength;
4537 int maxL = info.maxLength;
4538 boolean maxV = info.maxValid;
4539 info.reset();
4540 yes.study(info);
4541
4542 int minL2 = info.minLength;
4543 int maxL2 = info.maxLength;
4544 boolean maxV2 = info.maxValid;
4545 info.reset();
4546 not.study(info);
4547
4548 info.minLength = minL + Math.min(minL2, info.minLength);
4549 info.maxLength = maxL + Math.max(maxL2, info.maxLength);
4550 info.maxValid = (maxV & maxV2 & info.maxValid);
4551 info.deterministic = false;
4552 return next.study(info);
4553 }
4554 }
4555
4556 /**
4557 * Zero width positive lookahead.
4558 */
4559 static final class Pos extends Node {
4560 Node cond;
4561 Pos(Node cond) {
4562 this.cond = cond;
4563 }
4564 boolean match(Matcher matcher, int i, CharSequence seq) {
4565 int savedTo = matcher.to;
4566 boolean conditionMatched = false;
4567
4568 // Relax transparent region boundaries for lookahead
4569 if (matcher.transparentBounds)
4570 matcher.to = matcher.getTextLength();
4571 try {
4572 conditionMatched = cond.match(matcher, i, seq);
4573 } finally {
4574 // Reinstate region boundaries
4575 matcher.to = savedTo;
4576 }
4577 return conditionMatched && next.match(matcher, i, seq);
4578 }
4579 }
4580
4581 /**
4582 * Zero width negative lookahead.
4583 */
4584 static final class Neg extends Node {
4585 Node cond;
4586 Neg(Node cond) {
4587 this.cond = cond;
4588 }
4589 boolean match(Matcher matcher, int i, CharSequence seq) {
4590 int savedTo = matcher.to;
4591 boolean conditionMatched = false;
4592
4593 // Relax transparent region boundaries for lookahead
4594 if (matcher.transparentBounds)
4595 matcher.to = matcher.getTextLength();
4596 try {
4597 if (i < matcher.to) {
4598 conditionMatched = !cond.match(matcher, i, seq);
4599 } else {
4600 // If a negative lookahead succeeds then more input
4601 // could cause it to fail!
4602 matcher.requireEnd = true;
4603 conditionMatched = !cond.match(matcher, i, seq);
4604 }
4605 } finally {
4606 // Reinstate region boundaries
4607 matcher.to = savedTo;
4608 }
4609 return conditionMatched && next.match(matcher, i, seq);
4610 }
4611 }
4612
4613 /**
4614 * For use with lookbehinds; matches the position where the lookbehind
4615 * was encountered.
4616 */
4617 static Node lookbehindEnd = new Node() {
4618 boolean match(Matcher matcher, int i, CharSequence seq) {
4619 return i == matcher.lookbehindTo;
4620 }
4621 };
4622
4623 /**
4624 * Zero width positive lookbehind.
4625 */
4626 static class Behind extends Node {
4627 Node cond;
4628 int rmax, rmin;
4629 Behind(Node cond, int rmax, int rmin) {
4630 this.cond = cond;
4631 this.rmax = rmax;
4632 this.rmin = rmin;
4633 }
4634
4635 boolean match(Matcher matcher, int i, CharSequence seq) {
4636 int savedFrom = matcher.from;
4637 boolean conditionMatched = false;
4638 int startIndex = (!matcher.transparentBounds) ?
4639 matcher.from : 0;
4640 int from = Math.max(i - rmax, startIndex);
4641 // Set end boundary
4642 int savedLBT = matcher.lookbehindTo;
4643 matcher.lookbehindTo = i;
4644 // Relax transparent region boundaries for lookbehind
4645 if (matcher.transparentBounds)
4646 matcher.from = 0;
4647 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
4648 conditionMatched = cond.match(matcher, j, seq);
4649 }
4650 matcher.from = savedFrom;
4651 matcher.lookbehindTo = savedLBT;
4652 return conditionMatched && next.match(matcher, i, seq);
4653 }
4654 }
4655
4656 /**
4657 * Zero width positive lookbehind, including supplementary
4658 * characters or unpaired surrogates.
4659 */
4660 static final class BehindS extends Behind {
4661 BehindS(Node cond, int rmax, int rmin) {
4662 super(cond, rmax, rmin);
4663 }
4664 boolean match(Matcher matcher, int i, CharSequence seq) {
4665 int rmaxChars = countChars(seq, i, -rmax);
4666 int rminChars = countChars(seq, i, -rmin);
4667 int savedFrom = matcher.from;
4668 int startIndex = (!matcher.transparentBounds) ?
4669 matcher.from : 0;
4670 boolean conditionMatched = false;
4671 int from = Math.max(i - rmaxChars, startIndex);
4672 // Set end boundary
4673 int savedLBT = matcher.lookbehindTo;
4674 matcher.lookbehindTo = i;
4675 // Relax transparent region boundaries for lookbehind
4676 if (matcher.transparentBounds)
4677 matcher.from = 0;
4678
4679 for (int j = i - rminChars;
4680 !conditionMatched && j >= from;
4681 j -= j>from ? countChars(seq, j, -1) : 1) {
4682 conditionMatched = cond.match(matcher, j, seq);
4683 }
4684 matcher.from = savedFrom;
4685 matcher.lookbehindTo = savedLBT;
4686 return conditionMatched && next.match(matcher, i, seq);
4687 }
4688 }
4689
4690 /**
4691 * Zero width negative lookbehind.
4692 */
4693 static class NotBehind extends Node {
4694 Node cond;
4695 int rmax, rmin;
4696 NotBehind(Node cond, int rmax, int rmin) {
4697 this.cond = cond;
4698 this.rmax = rmax;
4699 this.rmin = rmin;
4700 }
4701
4702 boolean match(Matcher matcher, int i, CharSequence seq) {
4703 int savedLBT = matcher.lookbehindTo;
4704 int savedFrom = matcher.from;
4705 boolean conditionMatched = false;
4706 int startIndex = (!matcher.transparentBounds) ?
4707 matcher.from : 0;
4708 int from = Math.max(i - rmax, startIndex);
4709 matcher.lookbehindTo = i;
4710 // Relax transparent region boundaries for lookbehind
4711 if (matcher.transparentBounds)
4712 matcher.from = 0;
4713 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
4714 conditionMatched = cond.match(matcher, j, seq);
4715 }
4716 // Reinstate region boundaries
4717 matcher.from = savedFrom;
4718 matcher.lookbehindTo = savedLBT;
4719 return !conditionMatched && next.match(matcher, i, seq);
4720 }
4721 }
4722
4723 /**
4724 * Zero width negative lookbehind, including supplementary
4725 * characters or unpaired surrogates.
4726 */
4727 static final class NotBehindS extends NotBehind {
4728 NotBehindS(Node cond, int rmax, int rmin) {
4729 super(cond, rmax, rmin);
4730 }
4731 boolean match(Matcher matcher, int i, CharSequence seq) {
4732 int rmaxChars = countChars(seq, i, -rmax);
4733 int rminChars = countChars(seq, i, -rmin);
4734 int savedFrom = matcher.from;
4735 int savedLBT = matcher.lookbehindTo;
4736 boolean conditionMatched = false;
4737 int startIndex = (!matcher.transparentBounds) ?
4738 matcher.from : 0;
4739 int from = Math.max(i - rmaxChars, startIndex);
4740 matcher.lookbehindTo = i;
4741 // Relax transparent region boundaries for lookbehind
4742 if (matcher.transparentBounds)
4743 matcher.from = 0;
4744 for (int j = i - rminChars;
4745 !conditionMatched && j >= from;
4746 j -= j>from ? countChars(seq, j, -1) : 1) {
4747 conditionMatched = cond.match(matcher, j, seq);
4748 }
4749 //Reinstate region boundaries
4750 matcher.from = savedFrom;
4751 matcher.lookbehindTo = savedLBT;
4752 return !conditionMatched && next.match(matcher, i, seq);
4753 }
4754 }
4755
4756 /**
4757 * Returns the set union of two CharProperty nodes.
4758 */
4759 private static CharProperty union(final CharProperty lhs,
4760 final CharProperty rhs) {
4761 return new CharProperty() {
4762 boolean isSatisfiedBy(int ch) {
4763 return lhs.isSatisfiedBy(ch) || rhs.isSatisfiedBy(ch);}};
4764 }
4765
4766 /**
4767 * Returns the set intersection of two CharProperty nodes.
4768 */
4769 private static CharProperty intersection(final CharProperty lhs,
4770 final CharProperty rhs) {
4771 return new CharProperty() {
4772 boolean isSatisfiedBy(int ch) {
4773 return lhs.isSatisfiedBy(ch) && rhs.isSatisfiedBy(ch);}};
4774 }
4775
4776 /**
4777 * Returns the set difference of two CharProperty nodes.
4778 */
4779 private static CharProperty setDifference(final CharProperty lhs,
4780 final CharProperty rhs) {
4781 return new CharProperty() {
4782 boolean isSatisfiedBy(int ch) {
4783 return ! rhs.isSatisfiedBy(ch) && lhs.isSatisfiedBy(ch);}};
4784 }
4785
4786 /**
4787 * Handles word boundaries. Includes a field to allow this one class to
4788 * deal with the different types of word boundaries we can match. The word
4789 * characters include underscores, letters, and digits. Non spacing marks
4790 * can are also part of a word if they have a base character, otherwise
4791 * they are ignored for purposes of finding word boundaries.
4792 */
4793 static final class Bound extends Node {
4794 static int LEFT = 0x1;
4795 static int RIGHT= 0x2;
4796 static int BOTH = 0x3;
4797 static int NONE = 0x4;
4798 int type;
4799 Bound(int n) {
4800 type = n;
4801 }
4802 int check(Matcher matcher, int i, CharSequence seq) {
4803 int ch;
4804 boolean left = false;
4805 int startIndex = matcher.from;
4806 int endIndex = matcher.to;
4807 if (matcher.transparentBounds) {
4808 startIndex = 0;
4809 endIndex = matcher.getTextLength();
4810 }
4811 if (i > startIndex) {
4812 ch = Character.codePointBefore(seq, i);
4813 left = (ch == '_' || Character.isLetterOrDigit(ch) ||
4814 ((Character.getType(ch) == Character.NON_SPACING_MARK)
4815 && hasBaseCharacter(matcher, i-1, seq)));
4816 }
4817 boolean right = false;
4818 if (i < endIndex) {
4819 ch = Character.codePointAt(seq, i);
4820 right = (ch == '_' || Character.isLetterOrDigit(ch) ||
4821 ((Character.getType(ch) == Character.NON_SPACING_MARK)
4822 && hasBaseCharacter(matcher, i, seq)));
4823 } else {
4824 // Tried to access char past the end
4825 matcher.hitEnd = true;
4826 // The addition of another char could wreck a boundary
4827 matcher.requireEnd = true;
4828 }
4829 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE);
4830 }
4831 boolean match(Matcher matcher, int i, CharSequence seq) {
4832 return (check(matcher, i, seq) & type) > 0
4833 && next.match(matcher, i, seq);
4834 }
4835 }
4836
4837 /**
4838 * Non spacing marks only count as word characters in bounds calculations
4839 * if they have a base character.
4840 */
4841 private static boolean hasBaseCharacter(Matcher matcher, int i,
4842 CharSequence seq)
4843 {
4844 int start = (!matcher.transparentBounds) ?
4845 matcher.from : 0;
4846 for (int x=i; x >= start; x--) {
4847 int ch = Character.codePointAt(seq, x);
4848 if (Character.isLetterOrDigit(ch))
4849 return true;
4850 if (Character.getType(ch) == Character.NON_SPACING_MARK)
4851 continue;
4852 return false;
4853 }
4854 return false;
4855 }
4856
4857 /**
4858 * Attempts to match a slice in the input using the Boyer-Moore string
4859 * matching algorithm. The algorithm is based on the idea that the
4860 * pattern can be shifted farther ahead in the search text if it is
4861 * matched right to left.
4862 * <p>
4863 * The pattern is compared to the input one character at a time, from
4864 * the rightmost character in the pattern to the left. If the characters
4865 * all match the pattern has been found. If a character does not match,
4866 * the pattern is shifted right a distance that is the maximum of two
4867 * functions, the bad character shift and the good suffix shift. This
4868 * shift moves the attempted match position through the input more
4869 * quickly than a naive one position at a time check.
4870 * <p>
4871 * The bad character shift is based on the character from the text that
4872 * did not match. If the character does not appear in the pattern, the
4873 * pattern can be shifted completely beyond the bad character. If the
4874 * character does occur in the pattern, the pattern can be shifted to
4875 * line the pattern up with the next occurrence of that character.
4876 * <p>
4877 * The good suffix shift is based on the idea that some subset on the right
4878 * side of the pattern has matched. When a bad character is found, the
4879 * pattern can be shifted right by the pattern length if the subset does
4880 * not occur again in pattern, or by the amount of distance to the
4881 * next occurrence of the subset in the pattern.
4882 *
4883 * Boyer-Moore search methods adapted from code by Amy Yu.
4884 */
4885 static class BnM extends Node {
4886 int[] buffer;
4887 int[] lastOcc;
4888 int[] optoSft;
4889
4890 /**
4891 * Pre calculates arrays needed to generate the bad character
4892 * shift and the good suffix shift. Only the last seven bits
4893 * are used to see if chars match; This keeps the tables small
4894 * and covers the heavily used ASCII range, but occasionally
4895 * results in an aliased match for the bad character shift.
4896 */
4897 static Node optimize(Node node) {
4898 if (!(node instanceof Slice)) {
4899 return node;
4900 }
4901
4902 int[] src = ((Slice) node).buffer;
4903 int patternLength = src.length;
4904 // The BM algorithm requires a bit of overhead;
4905 // If the pattern is short don't use it, since
4906 // a shift larger than the pattern length cannot
4907 // be used anyway.
4908 if (patternLength < 4) {
4909 return node;
4910 }
4911 int i, j, k;
4912 int[] lastOcc = new int[128];
4913 int[] optoSft = new int[patternLength];
4914 // Precalculate part of the bad character shift
4915 // It is a table for where in the pattern each
4916 // lower 7-bit value occurs
4917 for (i = 0; i < patternLength; i++) {
4918 lastOcc[src[i]&0x7F] = i + 1;
4919 }
4920 // Precalculate the good suffix shift
4921 // i is the shift amount being considered
4922NEXT: for (i = patternLength; i > 0; i--) {
4923 // j is the beginning index of suffix being considered
4924 for (j = patternLength - 1; j >= i; j--) {
4925 // Testing for good suffix
4926 if (src[j] == src[j-i]) {
4927 // src[j..len] is a good suffix
4928 optoSft[j-1] = i;
4929 } else {
4930 // No match. The array has already been
4931 // filled up with correct values before.
4932 continue NEXT;
4933 }
4934 }
4935 // This fills up the remaining of optoSft
4936 // any suffix can not have larger shift amount
4937 // then its sub-suffix. Why???
4938 while (j > 0) {
4939 optoSft[--j] = i;
4940 }
4941 }
4942 // Set the guard value because of unicode compression
4943 optoSft[patternLength-1] = 1;
4944 if (node instanceof SliceS)
4945 return new BnMS(src, lastOcc, optoSft, node.next);
4946 return new BnM(src, lastOcc, optoSft, node.next);
4947 }
4948 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) {
4949 this.buffer = src;
4950 this.lastOcc = lastOcc;
4951 this.optoSft = optoSft;
4952 this.next = next;
4953 }
4954 boolean match(Matcher matcher, int i, CharSequence seq) {
4955 int[] src = buffer;
4956 int patternLength = src.length;
4957 int last = matcher.to - patternLength;
4958
4959 // Loop over all possible match positions in text
4960NEXT: while (i <= last) {
4961 // Loop over pattern from right to left
4962 for (int j = patternLength - 1; j >= 0; j--) {
4963 int ch = seq.charAt(i+j);
4964 if (ch != src[j]) {
4965 // Shift search to the right by the maximum of the
4966 // bad character shift and the good suffix shift
4967 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]);
4968 continue NEXT;
4969 }
4970 }
4971 // Entire pattern matched starting at i
4972 matcher.first = i;
4973 boolean ret = next.match(matcher, i + patternLength, seq);
4974 if (ret) {
4975 matcher.first = i;
4976 matcher.groups[0] = matcher.first;
4977 matcher.groups[1] = matcher.last;
4978 return true;
4979 }
4980 i++;
4981 }
4982 // BnM is only used as the leading node in the unanchored case,
4983 // and it replaced its Start() which always searches to the end
4984 // if it doesn't find what it's looking for, so hitEnd is true.
4985 matcher.hitEnd = true;
4986 return false;
4987 }
4988 boolean study(TreeInfo info) {
4989 info.minLength += buffer.length;
4990 info.maxValid = false;
4991 return next.study(info);
4992 }
4993 }
4994
4995 /**
4996 * Supplementary support version of BnM(). Unpaired surrogates are
4997 * also handled by this class.
4998 */
4999 static final class BnMS extends BnM {
5000 int lengthInChars;
5001
5002 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5003 super(src, lastOcc, optoSft, next);
5004 for (int x = 0; x < buffer.length; x++) {
5005 lengthInChars += Character.charCount(buffer[x]);
5006 }
5007 }
5008 boolean match(Matcher matcher, int i, CharSequence seq) {
5009 int[] src = buffer;
5010 int patternLength = src.length;
5011 int last = matcher.to - lengthInChars;
5012
5013 // Loop over all possible match positions in text
5014NEXT: while (i <= last) {
5015 // Loop over pattern from right to left
5016 int ch;
5017 for (int j = countChars(seq, i, patternLength), x = patternLength - 1;
5018 j > 0; j -= Character.charCount(ch), x--) {
5019 ch = Character.codePointBefore(seq, i+j);
5020 if (ch != src[x]) {
5021 // Shift search to the right by the maximum of the
5022 // bad character shift and the good suffix shift
5023 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]);
5024 i += countChars(seq, i, n);
5025 continue NEXT;
5026 }
5027 }
5028 // Entire pattern matched starting at i
5029 matcher.first = i;
5030 boolean ret = next.match(matcher, i + lengthInChars, seq);
5031 if (ret) {
5032 matcher.first = i;
5033 matcher.groups[0] = matcher.first;
5034 matcher.groups[1] = matcher.last;
5035 return true;
5036 }
5037 i += countChars(seq, i, 1);
5038 }
5039 matcher.hitEnd = true;
5040 return false;
5041 }
5042 }
5043
5044///////////////////////////////////////////////////////////////////////////////
5045///////////////////////////////////////////////////////////////////////////////
5046
5047 /**
5048 * This must be the very first initializer.
5049 */
5050 static Node accept = new Node();
5051
5052 static Node lastAccept = new LastNode();
5053
5054 private static class CharPropertyNames {
5055
5056 static CharProperty charPropertyFor(String name) {
5057 CharPropertyFactory m = map.get(name);
5058 return m == null ? null : m.make();
5059 }
5060
5061 private static abstract class CharPropertyFactory {
5062 abstract CharProperty make();
5063 }
5064
5065 private static void defCategory(String name,
5066 final int typeMask) {
5067 map.put(name, new CharPropertyFactory() {
5068 CharProperty make() { return new Category(typeMask);}});
5069 }
5070
5071 private static void defRange(String name,
5072 final int lower, final int upper) {
5073 map.put(name, new CharPropertyFactory() {
5074 CharProperty make() { return rangeFor(lower, upper);}});
5075 }
5076
5077 private static void defCtype(String name,
5078 final int ctype) {
5079 map.put(name, new CharPropertyFactory() {
5080 CharProperty make() { return new Ctype(ctype);}});
5081 }
5082
5083 private static abstract class CloneableProperty
5084 extends CharProperty implements Cloneable
5085 {
5086 public CloneableProperty clone() {
5087 try {
5088 return (CloneableProperty) super.clone();
5089 } catch (CloneNotSupportedException e) {
5090 throw new AssertionError(e);
5091 }
5092 }
5093 }
5094
5095 private static void defClone(String name,
5096 final CloneableProperty p) {
5097 map.put(name, new CharPropertyFactory() {
5098 CharProperty make() { return p.clone();}});
5099 }
5100
5101 private static final HashMap<String, CharPropertyFactory> map
5102 = new HashMap<String, CharPropertyFactory>();
5103
5104 static {
5105 // Unicode character property aliases, defined in
5106 // http://www.unicode.org/Public/UNIDATA/PropertyValueAliases.txt
5107 defCategory("Cn", 1<<Character.UNASSIGNED);
5108 defCategory("Lu", 1<<Character.UPPERCASE_LETTER);
5109 defCategory("Ll", 1<<Character.LOWERCASE_LETTER);
5110 defCategory("Lt", 1<<Character.TITLECASE_LETTER);
5111 defCategory("Lm", 1<<Character.MODIFIER_LETTER);
5112 defCategory("Lo", 1<<Character.OTHER_LETTER);
5113 defCategory("Mn", 1<<Character.NON_SPACING_MARK);
5114 defCategory("Me", 1<<Character.ENCLOSING_MARK);
5115 defCategory("Mc", 1<<Character.COMBINING_SPACING_MARK);
5116 defCategory("Nd", 1<<Character.DECIMAL_DIGIT_NUMBER);
5117 defCategory("Nl", 1<<Character.LETTER_NUMBER);
5118 defCategory("No", 1<<Character.OTHER_NUMBER);
5119 defCategory("Zs", 1<<Character.SPACE_SEPARATOR);
5120 defCategory("Zl", 1<<Character.LINE_SEPARATOR);
5121 defCategory("Zp", 1<<Character.PARAGRAPH_SEPARATOR);
5122 defCategory("Cc", 1<<Character.CONTROL);
5123 defCategory("Cf", 1<<Character.FORMAT);
5124 defCategory("Co", 1<<Character.PRIVATE_USE);
5125 defCategory("Cs", 1<<Character.SURROGATE);
5126 defCategory("Pd", 1<<Character.DASH_PUNCTUATION);
5127 defCategory("Ps", 1<<Character.START_PUNCTUATION);
5128 defCategory("Pe", 1<<Character.END_PUNCTUATION);
5129 defCategory("Pc", 1<<Character.CONNECTOR_PUNCTUATION);
5130 defCategory("Po", 1<<Character.OTHER_PUNCTUATION);
5131 defCategory("Sm", 1<<Character.MATH_SYMBOL);
5132 defCategory("Sc", 1<<Character.CURRENCY_SYMBOL);
5133 defCategory("Sk", 1<<Character.MODIFIER_SYMBOL);
5134 defCategory("So", 1<<Character.OTHER_SYMBOL);
5135 defCategory("Pi", 1<<Character.INITIAL_QUOTE_PUNCTUATION);
5136 defCategory("Pf", 1<<Character.FINAL_QUOTE_PUNCTUATION);
5137 defCategory("L", ((1<<Character.UPPERCASE_LETTER) |
5138 (1<<Character.LOWERCASE_LETTER) |
5139 (1<<Character.TITLECASE_LETTER) |
5140 (1<<Character.MODIFIER_LETTER) |
5141 (1<<Character.OTHER_LETTER)));
5142 defCategory("M", ((1<<Character.NON_SPACING_MARK) |
5143 (1<<Character.ENCLOSING_MARK) |
5144 (1<<Character.COMBINING_SPACING_MARK)));
5145 defCategory("N", ((1<<Character.DECIMAL_DIGIT_NUMBER) |
5146 (1<<Character.LETTER_NUMBER) |
5147 (1<<Character.OTHER_NUMBER)));
5148 defCategory("Z", ((1<<Character.SPACE_SEPARATOR) |
5149 (1<<Character.LINE_SEPARATOR) |
5150 (1<<Character.PARAGRAPH_SEPARATOR)));
5151 defCategory("C", ((1<<Character.CONTROL) |
5152 (1<<Character.FORMAT) |
5153 (1<<Character.PRIVATE_USE) |
5154 (1<<Character.SURROGATE))); // Other
5155 defCategory("P", ((1<<Character.DASH_PUNCTUATION) |
5156 (1<<Character.START_PUNCTUATION) |
5157 (1<<Character.END_PUNCTUATION) |
5158 (1<<Character.CONNECTOR_PUNCTUATION) |
5159 (1<<Character.OTHER_PUNCTUATION) |
5160 (1<<Character.INITIAL_QUOTE_PUNCTUATION) |
5161 (1<<Character.FINAL_QUOTE_PUNCTUATION)));
5162 defCategory("S", ((1<<Character.MATH_SYMBOL) |
5163 (1<<Character.CURRENCY_SYMBOL) |
5164 (1<<Character.MODIFIER_SYMBOL) |
5165 (1<<Character.OTHER_SYMBOL)));
5166 defCategory("LC", ((1<<Character.UPPERCASE_LETTER) |
5167 (1<<Character.LOWERCASE_LETTER) |
5168 (1<<Character.TITLECASE_LETTER)));
5169 defCategory("LD", ((1<<Character.UPPERCASE_LETTER) |
5170 (1<<Character.LOWERCASE_LETTER) |
5171 (1<<Character.TITLECASE_LETTER) |
5172 (1<<Character.MODIFIER_LETTER) |
5173 (1<<Character.OTHER_LETTER) |
5174 (1<<Character.DECIMAL_DIGIT_NUMBER)));
5175 defRange("L1", 0x00, 0xFF); // Latin-1
5176 map.put("all", new CharPropertyFactory() {
5177 CharProperty make() { return new All(); }});
5178
5179 // Posix regular expression character classes, defined in
5180 // http://www.unix.org/onlinepubs/009695399/basedefs/xbd_chap09.html
5181 defRange("ASCII", 0x00, 0x7F); // ASCII
5182 defCtype("Alnum", ASCII.ALNUM); // Alphanumeric characters
5183 defCtype("Alpha", ASCII.ALPHA); // Alphabetic characters
5184 defCtype("Blank", ASCII.BLANK); // Space and tab characters
5185 defCtype("Cntrl", ASCII.CNTRL); // Control characters
5186 defRange("Digit", '0', '9'); // Numeric characters
5187 defCtype("Graph", ASCII.GRAPH); // printable and visible
5188 defRange("Lower", 'a', 'z'); // Lower-case alphabetic
5189 defRange("Print", 0x20, 0x7E); // Printable characters
5190 defCtype("Punct", ASCII.PUNCT); // Punctuation characters
5191 defCtype("Space", ASCII.SPACE); // Space characters
5192 defRange("Upper", 'A', 'Z'); // Upper-case alphabetic
5193 defCtype("XDigit",ASCII.XDIGIT); // hexadecimal digits
5194
5195 // Java character properties, defined by methods in Character.java
5196 defClone("javaLowerCase", new CloneableProperty() {
5197 boolean isSatisfiedBy(int ch) {
5198 return Character.isLowerCase(ch);}});
5199 defClone("javaUpperCase", new CloneableProperty() {
5200 boolean isSatisfiedBy(int ch) {
5201 return Character.isUpperCase(ch);}});
5202 defClone("javaTitleCase", new CloneableProperty() {
5203 boolean isSatisfiedBy(int ch) {
5204 return Character.isTitleCase(ch);}});
5205 defClone("javaDigit", new CloneableProperty() {
5206 boolean isSatisfiedBy(int ch) {
5207 return Character.isDigit(ch);}});
5208 defClone("javaDefined", new CloneableProperty() {
5209 boolean isSatisfiedBy(int ch) {
5210 return Character.isDefined(ch);}});
5211 defClone("javaLetter", new CloneableProperty() {
5212 boolean isSatisfiedBy(int ch) {
5213 return Character.isLetter(ch);}});
5214 defClone("javaLetterOrDigit", new CloneableProperty() {
5215 boolean isSatisfiedBy(int ch) {
5216 return Character.isLetterOrDigit(ch);}});
5217 defClone("javaJavaIdentifierStart", new CloneableProperty() {
5218 boolean isSatisfiedBy(int ch) {
5219 return Character.isJavaIdentifierStart(ch);}});
5220 defClone("javaJavaIdentifierPart", new CloneableProperty() {
5221 boolean isSatisfiedBy(int ch) {
5222 return Character.isJavaIdentifierPart(ch);}});
5223 defClone("javaUnicodeIdentifierStart", new CloneableProperty() {
5224 boolean isSatisfiedBy(int ch) {
5225 return Character.isUnicodeIdentifierStart(ch);}});
5226 defClone("javaUnicodeIdentifierPart", new CloneableProperty() {
5227 boolean isSatisfiedBy(int ch) {
5228 return Character.isUnicodeIdentifierPart(ch);}});
5229 defClone("javaIdentifierIgnorable", new CloneableProperty() {
5230 boolean isSatisfiedBy(int ch) {
5231 return Character.isIdentifierIgnorable(ch);}});
5232 defClone("javaSpaceChar", new CloneableProperty() {
5233 boolean isSatisfiedBy(int ch) {
5234 return Character.isSpaceChar(ch);}});
5235 defClone("javaWhitespace", new CloneableProperty() {
5236 boolean isSatisfiedBy(int ch) {
5237 return Character.isWhitespace(ch);}});
5238 defClone("javaISOControl", new CloneableProperty() {
5239 boolean isSatisfiedBy(int ch) {
5240 return Character.isISOControl(ch);}});
5241 defClone("javaMirrored", new CloneableProperty() {
5242 boolean isSatisfiedBy(int ch) {
5243 return Character.isMirrored(ch);}});
5244 }
5245 }
5246}
5247