1 /*
2 * Copyright (c) 1999, 2019, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation. Oracle designates this
8 * particular file as subject to the "Classpath" exception as provided
9 * by Oracle in the LICENSE file that accompanied this code.
10 *
11 * This code is distributed in the hope that it will be useful, but WITHOUT
12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
14 * version 2 for more details (a copy is included in the LICENSE file that
15 * accompanied this code).
16 *
17 * You should have received a copy of the GNU General Public License version
18 * 2 along with this work; if not, write to the Free Software Foundation,
19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20 *
21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
22 * or visit www.oracle.com if you need additional information or have any
23 * questions.
24 */
25
26 package java.util.regex;
27
28 import java.text.Normalizer;
29 import java.text.Normalizer.Form;
30 import java.util.Locale;
31 import java.util.Iterator;
32 import java.util.Map;
33 import java.util.ArrayList;
34 import java.util.HashMap;
35 import java.util.LinkedHashSet;
36 import java.util.List;
37 import java.util.Set;
38 import java.util.Arrays;
39 import java.util.NoSuchElementException;
40 import java.util.Spliterator;
41 import java.util.Spliterators;
42 import java.util.function.Predicate;
43 import java.util.stream.Stream;
44 import java.util.stream.StreamSupport;
45
46 import jdk.internal.util.ArraysSupport;
47
48 /**
49 * A compiled representation of a regular expression.
50 *
51 * <p> A regular expression, specified as a string, must first be compiled into
52 * an instance of this class. The resulting pattern can then be used to create
53 * a {@link Matcher} object that can match arbitrary {@linkplain
54 * java.lang.CharSequence character sequences} against the regular
55 * expression. All of the state involved in performing a match resides in the
56 * matcher, so many matchers can share the same pattern.
57 *
58 * <p> A typical invocation sequence is thus
59 *
60 * <blockquote><pre>
61 * Pattern p = Pattern.{@link #compile compile}("a*b");
62 * Matcher m = p.{@link #matcher matcher}("aaaaab");
63 * boolean b = m.{@link Matcher#matches matches}();</pre></blockquote>
64 *
65 * <p> A {@link #matches matches} method is defined by this class as a
66 * convenience for when a regular expression is used just once. This method
67 * compiles an expression and matches an input sequence against it in a single
68 * invocation. The statement
69 *
70 * <blockquote><pre>
71 * boolean b = Pattern.matches("a*b", "aaaaab");</pre></blockquote>
72 *
73 * is equivalent to the three statements above, though for repeated matches it
74 * is less efficient since it does not allow the compiled pattern to be reused.
75 *
76 * <p> Instances of this class are immutable and are safe for use by multiple
77 * concurrent threads. Instances of the {@link Matcher} class are not safe for
78 * such use.
79 *
80 *
81 * <h2><a id="sum">Summary of regular-expression constructs</a></h2>
82 *
83 * <table class="borderless">
84 * <caption style="display:none">Regular expression constructs, and what they match</caption>
85 * <thead style="text-align:left">
86 * <tr>
87 * <th id="construct">Construct</th>
88 * <th id="matches">Matches</th>
89 * </tr>
90 * </thead>
91 * <tbody style="text-align:left">
92 *
93 * <tr><th colspan="2" style="padding-top:20px" id="characters">Characters</th></tr>
94 *
95 * <tr><th style="vertical-align:top; font-weight: normal" id="x"><i>x</i></th>
96 * <td headers="matches characters x">The character <i>x</i></td></tr>
97 * <tr><th style="vertical-align:top; font-weight: normal" id="backslash">{@code \\}</th>
98 * <td headers="matches characters backslash">The backslash character</td></tr>
99 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_n">{@code \0}<i>n</i></th>
100 * <td headers="matches characters octal_n">The character with octal value {@code 0}<i>n</i>
101 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr>
102 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nn">{@code \0}<i>nn</i></th>
103 * <td headers="matches characters octal_nn">The character with octal value {@code 0}<i>nn</i>
104 * (0 {@code <=} <i>n</i> {@code <=} 7)</td></tr>
105 * <tr><th style="vertical-align:top; font-weight: normal" id="octal_nnn">{@code \0}<i>mnn</i></th>
106 * <td headers="matches characters octal_nnn">The character with octal value {@code 0}<i>mnn</i>
107 * (0 {@code <=} <i>m</i> {@code <=} 3,
108 * 0 {@code <=} <i>n</i> {@code <=} 7)</td></tr>
109 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hh">{@code \x}<i>hh</i></th>
110 * <td headers="matches characters hex_hh">The character with hexadecimal value {@code 0x}<i>hh</i></td></tr>
111 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_hhhh"><code>\u</code><i>hhhh</i></th>
112 * <td headers="matches characters hex_hhhh">The character with hexadecimal value {@code 0x}<i>hhhh</i></td></tr>
113 * <tr><th style="vertical-align:top; font-weight: normal" id="hex_h_h"><code>\x</code><i>{h...h}</i></th>
114 * <td headers="matches characters hex_h_h">The character with hexadecimal value {@code 0x}<i>h...h</i>
115 * ({@link java.lang.Character#MIN_CODE_POINT Character.MIN_CODE_POINT}
116 * <= {@code 0x}<i>h...h</i> <=
117 * {@link java.lang.Character#MAX_CODE_POINT Character.MAX_CODE_POINT})</td></tr>
118 * <tr><th style="vertical-align:top; font-weight: normal" id="unicode_name"><code>\N{</code><i>name</i><code>}</code></th>
119 * <td headers="matches characters unicode_name">The character with Unicode character name <i>'name'</i></td></tr>
120 * <tr><th style="vertical-align:top; font-weight:normal" id="tab">{@code \t}</th>
121 * <td headers="matches characters tab">The tab character (<code>'\u0009'</code>)</td></tr>
122 * <tr><th style="vertical-align:top; font-weight:normal" id="newline">{@code \n}</th>
123 * <td headers="matches characters newline">The newline (line feed) character (<code>'\u000A'</code>)</td></tr>
124 * <tr><th style="vertical-align:top; font-weight:normal" id="return">{@code \r}</th>
125 * <td headers="matches characters return">The carriage-return character (<code>'\u000D'</code>)</td></tr>
126 * <tr><th style="vertical-align:top; font-weight:normal" id="form_feed">{@code \f}</th>
127 * <td headers="matches characters form_feed">The form-feed character (<code>'\u000C'</code>)</td></tr>
128 * <tr><th style="vertical-align:top; font-weight:normal" id="bell">{@code \a}</th>
129 * <td headers="matches characters bell">The alert (bell) character (<code>'\u0007'</code>)</td></tr>
130 * <tr><th style="vertical-align:top; font-weight:normal" id="escape">{@code \e}</th>
131 * <td headers="matches characters escape">The escape character (<code>'\u001B'</code>)</td></tr>
132 * <tr><th style="vertical-align:top; font-weight:normal" id="ctrl_x">{@code \c}<i>x</i></th>
133 * <td headers="matches characters ctrl_x">The control character corresponding to <i>x</i></td></tr>
134 *
135 * <tr><th colspan="2" style="padding-top:20px" id="classes">Character classes</th></tr>
136 *
137 * <tr><th style="vertical-align:top; font-weight:normal" id="simple">{@code [abc]}</th>
138 * <td headers="matches classes simple">{@code a}, {@code b}, or {@code c} (simple class)</td></tr>
139 * <tr><th style="vertical-align:top; font-weight:normal" id="negation">{@code [^abc]}</th>
140 * <td headers="matches classes negation">Any character except {@code a}, {@code b}, or {@code c} (negation)</td></tr>
141 * <tr><th style="vertical-align:top; font-weight:normal" id="range">{@code [a-zA-Z]}</th>
142 * <td headers="matches classes range">{@code a} through {@code z}
143 * or {@code A} through {@code Z}, inclusive (range)</td></tr>
144 * <tr><th style="vertical-align:top; font-weight:normal" id="union">{@code [a-d[m-p]]}</th>
145 * <td headers="matches classes union">{@code a} through {@code d},
146 * or {@code m} through {@code p}: {@code [a-dm-p]} (union)</td></tr>
147 * <tr><th style="vertical-align:top; font-weight:normal" id="intersection">{@code [a-z&&[def]]}</th>
148 * <td headers="matches classes intersection">{@code d}, {@code e}, or {@code f} (intersection)</tr>
149 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction1">{@code [a-z&&[^bc]]}</th>
150 * <td headers="matches classes subtraction1">{@code a} through {@code z},
151 * except for {@code b} and {@code c}: {@code [ad-z]} (subtraction)</td></tr>
152 * <tr><th style="vertical-align:top; font-weight:normal" id="subtraction2">{@code [a-z&&[^m-p]]}</th>
153 * <td headers="matches classes subtraction2">{@code a} through {@code z},
154 * and not {@code m} through {@code p}: {@code [a-lq-z]}(subtraction)</td></tr>
155 *
156 * <tr><th colspan="2" style="padding-top:20px" id="predef">Predefined character classes</th></tr>
157 *
158 * <tr><th style="vertical-align:top; font-weight:normal" id="any">{@code .}</th>
159 * <td headers="matches predef any">Any character (may or may not match <a href="#lt">line terminators</a>)</td></tr>
160 * <tr><th style="vertical-align:top; font-weight:normal" id="digit">{@code \d}</th>
161 * <td headers="matches predef digit">A digit: {@code [0-9]}</td></tr>
162 * <tr><th style="vertical-align:top; font-weight:normal" id="non_digit">{@code \D}</th>
163 * <td headers="matches predef non_digit">A non-digit: {@code [^0-9]}</td></tr>
164 * <tr><th style="vertical-align:top; font-weight:normal" id="horiz_white">{@code \h}</th>
165 * <td headers="matches predef horiz_white">A horizontal whitespace character:
166 * <code>[ \t\xA0\u1680\u180e\u2000-\u200a\u202f\u205f\u3000]</code></td></tr>
167 * <tr><th style="vertical-align:top; font-weight:normal" id="non_horiz_white">{@code \H}</th>
168 * <td headers="matches predef non_horiz_white">A non-horizontal whitespace character: {@code [^\h]}</td></tr>
169 * <tr><th style="vertical-align:top; font-weight:normal" id="white">{@code \s}</th>
170 * <td headers="matches predef white">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr>
171 * <tr><th style="vertical-align:top; font-weight:normal" id="non_white">{@code \S}</th>
172 * <td headers="matches predef non_white">A non-whitespace character: {@code [^\s]}</td></tr>
173 * <tr><th style="vertical-align:top; font-weight:normal" id="vert_white">{@code \v}</th>
174 * <td headers="matches predef vert_white">A vertical whitespace character: <code>[\n\x0B\f\r\x85\u2028\u2029]</code>
175 * </td></tr>
176 * <tr><th style="vertical-align:top; font-weight:normal" id="non_vert_white">{@code \V}</th>
177 * <td headers="matches predef non_vert_white">A non-vertical whitespace character: {@code [^\v]}</td></tr>
178 * <tr><th style="vertical-align:top; font-weight:normal" id="word">{@code \w}</th>
179 * <td headers="matches predef word">A word character: {@code [a-zA-Z_0-9]}</td></tr>
180 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word">{@code \W}</th>
181 * <td headers="matches predef non_word">A non-word character: {@code [^\w]}</td></tr>
182 *
183 * <tr><th colspan="2" style="padding-top:20px" id="posix"><b>POSIX character classes (US-ASCII only)</b></th></tr>
184 *
185 * <tr><th style="vertical-align:top; font-weight:normal" id="Lower">{@code \p{Lower}}</th>
186 * <td headers="matches posix Lower">A lower-case alphabetic character: {@code [a-z]}</td></tr>
187 * <tr><th style="vertical-align:top; font-weight:normal" id="Upper">{@code \p{Upper}}</th>
188 * <td headers="matches posix Upper">An upper-case alphabetic character:{@code [A-Z]}</td></tr>
189 * <tr><th style="vertical-align:top; font-weight:normal" id="ASCII">{@code \p{ASCII}}</th>
190 * <td headers="matches posix ASCII">All ASCII:{@code [\x00-\x7F]}</td></tr>
191 * <tr><th style="vertical-align:top; font-weight:normal" id="Alpha">{@code \p{Alpha}}</th>
192 * <td headers="matches posix Alpha">An alphabetic character:{@code [\p{Lower}\p{Upper}]}</td></tr>
193 * <tr><th style="vertical-align:top; font-weight:normal" id="Digit">{@code \p{Digit}}</th>
194 * <td headers="matches posix Digit">A decimal digit: {@code [0-9]}</td></tr>
195 * <tr><th style="vertical-align:top; font-weight:normal" id="Alnum">{@code \p{Alnum}}</th>
196 * <td headers="matches posix Alnum">An alphanumeric character:{@code [\p{Alpha}\p{Digit}]}</td></tr>
197 * <tr><th style="vertical-align:top; font-weight:normal" id="Punct">{@code \p{Punct}}</th>
198 * <td headers="matches posix Punct">Punctuation: One of {@code !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~}</td></tr>
199 * <!-- {@code [\!"#\$%&'\(\)\*\+,\-\./:;\<=\>\?@\[\\\]\^_`\{\|\}~]}
200 * {@code [\X21-\X2F\X31-\X40\X5B-\X60\X7B-\X7E]} -->
201 * <tr><th style="vertical-align:top; font-weight:normal" id="Graph">{@code \p{Graph}}</th>
202 * <td headers="matches posix Graph">A visible character: {@code [\p{Alnum}\p{Punct}]}</td></tr>
203 * <tr><th style="vertical-align:top; font-weight:normal" id="Print">{@code \p{Print}}</th>
204 * <td headers="matches posix Print">A printable character: {@code [\p{Graph}\x20]}</td></tr>
205 * <tr><th style="vertical-align:top; font-weight:normal" id="Blank">{@code \p{Blank}}</th>
206 * <td headers="matches posix Blank">A space or a tab: {@code [ \t]}</td></tr>
207 * <tr><th style="vertical-align:top; font-weight:normal" id="Cntrl">{@code \p{Cntrl}}</th>
208 * <td headers="matches posix Cntrl">A control character: {@code [\x00-\x1F\x7F]}</td></tr>
209 * <tr><th style="vertical-align:top; font-weight:normal" id="XDigit">{@code \p{XDigit}}</th>
210 * <td headers="matches posix XDigit">A hexadecimal digit: {@code [0-9a-fA-F]}</td></tr>
211 * <tr><th style="vertical-align:top; font-weight:normal" id="Space">{@code \p{Space}}</th>
212 * <td headers="matches posix Space">A whitespace character: {@code [ \t\n\x0B\f\r]}</td></tr>
213 *
214 * <tr><th colspan="2" style="padding-top:20px" id="java">java.lang.Character classes (simple <a href="#jcc">java character type</a>)</th></tr>
215 *
216 * <tr><th style="vertical-align:top; font-weight:normal" id="javaLowerCase">{@code \p{javaLowerCase}}</th>
217 * <td headers="matches java javaLowerCase">Equivalent to java.lang.Character.isLowerCase()</td></tr>
218 * <tr><th style="vertical-align:top; font-weight:normal" id="javaUpperCase">{@code \p{javaUpperCase}}</th>
219 * <td headers="matches java javaUpperCase">Equivalent to java.lang.Character.isUpperCase()</td></tr>
220 * <tr><th style="vertical-align:top; font-weight:normal" id="javaWhitespace">{@code \p{javaWhitespace}}</th>
221 * <td headers="matches java javaWhitespace">Equivalent to java.lang.Character.isWhitespace()</td></tr>
222 * <tr><th style="vertical-align:top; font-weight:normal" id="javaMirrored">{@code \p{javaMirrored}}</th>
223 * <td headers="matches java javaMirrored">Equivalent to java.lang.Character.isMirrored()</td></tr>
224 *
225 * <tr><th colspan="2" style="padding-top:20px" id="unicode">Classes for Unicode scripts, blocks, categories and binary properties</th></tr>
226 *
227 * <tr><th style="vertical-align:top; font-weight:normal" id="IsLatin">{@code \p{IsLatin}}</th>
228 * <td headers="matches unicode IsLatin">A Latin script character (<a href="#usc">script</a>)</td></tr>
229 * <tr><th style="vertical-align:top; font-weight:normal" id="InGreek">{@code \p{InGreek}}</th>
230 * <td headers="matches unicode InGreek">A character in the Greek block (<a href="#ubc">block</a>)</td></tr>
231 * <tr><th style="vertical-align:top; font-weight:normal" id="Lu">{@code \p{Lu}}</th>
232 * <td headers="matches unicode Lu">An uppercase letter (<a href="#ucc">category</a>)</td></tr>
233 * <tr><th style="vertical-align:top; font-weight:normal" id="IsAlphabetic">{@code \p{IsAlphabetic}}</th>
234 * <td headers="matches unicode IsAlphabetic">An alphabetic character (<a href="#ubpc">binary property</a>)</td></tr>
235 * <tr><th style="vertical-align:top; font-weight:normal" id="Sc">{@code \p{Sc}}</th>
236 * <td headers="matches unicode Sc">A currency symbol</td></tr>
237 * <tr><th style="vertical-align:top; font-weight:normal" id="not_InGreek">{@code \P{InGreek}}</th>
238 * <td headers="matches unicode not_InGreek">Any character except one in the Greek block (negation)</td></tr>
239 * <tr><th style="vertical-align:top; font-weight:normal" id="not_uppercase">{@code [\p{L}&&[^\p{Lu}]]}</th>
240 * <td headers="matches unicode not_uppercase">Any letter except an uppercase letter (subtraction)</td></tr>
241 *
242 * <tr><th colspan="2" style="padding-top:20px" id="bounds">Boundary matchers</th></tr>
243 *
244 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_line">{@code ^}</th>
245 * <td headers="matches bounds begin_line">The beginning of a line</td></tr>
246 * <tr><th style="vertical-align:top; font-weight:normal" id="end_line">{@code $}</th>
247 * <td headers="matches bounds end_line">The end of a line</td></tr>
248 * <tr><th style="vertical-align:top; font-weight:normal" id="word_boundary">{@code \b}</th>
249 * <td headers="matches bounds word_boundary">A word boundary</td></tr>
250 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_cluster_boundary">{@code \b{g}}</th>
251 * <td headers="matches bounds grapheme_cluster_boundary">A Unicode extended grapheme cluster boundary</td></tr>
252 * <tr><th style="vertical-align:top; font-weight:normal" id="non_word_boundary">{@code \B}</th>
253 * <td headers="matches bounds non_word_boundary">A non-word boundary</td></tr>
254 * <tr><th style="vertical-align:top; font-weight:normal" id="begin_input">{@code \A}</th>
255 * <td headers="matches bounds begin_input">The beginning of the input</td></tr>
256 * <tr><th style="vertical-align:top; font-weight:normal" id="end_prev_match">{@code \G}</th>
257 * <td headers="matches bounds end_prev_match">The end of the previous match</td></tr>
258 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input_except_term">{@code \Z}</th>
259 * <td headers="matches bounds end_input_except_term">The end of the input but for the final
260 * <a href="#lt">terminator</a>, if any</td></tr>
261 * <tr><th style="vertical-align:top; font-weight:normal" id="end_input">{@code \z}</th>
262 * <td headers="matches bounds end_input">The end of the input</td></tr>
263 *
264 * <tr><th colspan="2" style="padding-top:20px" id="linebreak">Linebreak matcher</th></tr>
265 *
266 * <tr><th style="vertical-align:top; font-weight:normal" id="any_unicode_linebreak">{@code \R}</th>
267 * <td headers="matches linebreak any_unicode_linebreak">Any Unicode linebreak sequence, is equivalent to
268 * <code>\u000D\u000A|[\u000A\u000B\u000C\u000D\u0085\u2028\u2029]
269 * </code></td></tr>
270 *
271 * <tr><th colspan="2" style="padding-top:20px" id="grapheme">Unicode Extended Grapheme matcher</th></tr>
272 *
273 * <tr><th style="vertical-align:top; font-weight:normal" id="grapheme_any">{@code \X}</th>
274 * <td headers="matches grapheme grapheme_any">Any Unicode extended grapheme cluster</td></tr>
275 *
276 * <tr><th colspan="2" style="padding-top:20px" id="greedy">Greedy quantifiers</th></tr>
277 *
278 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_once_or_not"><i>X</i>{@code ?}</th>
279 * <td headers="matches greedy greedy_once_or_not"><i>X</i>, once or not at all</td></tr>
280 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_zero_or_more"><i>X</i>{@code *}</th>
281 * <td headers="matches greedy greedy_zero_or_more"><i>X</i>, zero or more times</td></tr>
282 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_one_or_more"><i>X</i>{@code +}</th>
283 * <td headers="matches greedy greedy_one_or_more"><i>X</i>, one or more times</td></tr>
284 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_exactly"><i>X</i><code>{</code><i>n</i><code>}</code></th>
285 * <td headers="matches greedy greedy_exactly"><i>X</i>, exactly <i>n</i> times</td></tr>
286 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least"><i>X</i><code>{</code><i>n</i>{@code ,}}</th>
287 * <td headers="matches greedy greedy_at_least"><i>X</i>, at least <i>n</i> times</td></tr>
288 * <tr><th style="vertical-align:top; font-weight:normal" id="greedy_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}</code></th>
289 * <td headers="matches greedy greedy_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
290 *
291 * <tr><th colspan="2" style="padding-top:20px" id="reluc">Reluctant quantifiers</th></tr>
292 *
293 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_once_or_not"><i>X</i>{@code ??}</th>
294 * <td headers="matches reluc reluc_once_or_not"><i>X</i>, once or not at all</td></tr>
295 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_zero_or_more"><i>X</i>{@code *?}</th>
296 * <td headers="matches reluc reluc_zero_or_more"><i>X</i>, zero or more times</td></tr>
297 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_one_or_more"><i>X</i>{@code +?}</th>
298 * <td headers="matches reluc reluc_one_or_more"><i>X</i>, one or more times</td></tr>
299 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_exactly"><i>X</i><code>{</code><i>n</i><code>}?</code></th>
300 * <td headers="matches reluc reluc_exactly"><i>X</i>, exactly <i>n</i> times</td></tr>
301 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least"><i>X</i><code>{</code><i>n</i><code>,}?</code></th>
302 * <td headers="matches reluc reluc_at_least"><i>X</i>, at least <i>n</i> times</td></tr>
303 * <tr><th style="vertical-align:top; font-weight:normal" id="reluc_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}?</code></th>
304 * <td headers="matches reluc reluc_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
305 *
306 * <tr><th colspan="2" style="padding-top:20px" id="poss">Possessive quantifiers</th></tr>
307 *
308 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_once_or_not"><i>X</i>{@code ?+}</th>
309 * <td headers="matches poss poss_once_or_not"><i>X</i>, once or not at all</td></tr>
310 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_zero_or_more"><i>X</i>{@code *+}</th>
311 * <td headers="matches poss poss_zero_or_more"><i>X</i>, zero or more times</td></tr>
312 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_one_or_more"><i>X</i>{@code ++}</th>
313 * <td headers="matches poss poss_one_or_more"><i>X</i>, one or more times</td></tr>
314 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_exactly"><i>X</i><code>{</code><i>n</i><code>}+</code></th>
315 * <td headers="matches poss poss_exactly"><i>X</i>, exactly <i>n</i> times</td></tr>
316 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least"><i>X</i><code>{</code><i>n</i><code>,}+</code></th>
317 * <td headers="matches poss poss_at_least"><i>X</i>, at least <i>n</i> times</td></tr>
318 * <tr><th style="vertical-align:top; font-weight:normal" id="poss_at_least_up_to"><i>X</i><code>{</code><i>n</i>{@code ,}<i>m</i><code>}+</code></th>
319 * <td headers="matches poss poss_at_least_up_to"><i>X</i>, at least <i>n</i> but not more than <i>m</i> times</td></tr>
320 *
321 * <tr><th colspan="2" style="padding-top:20px" id="logical">Logical operators</th></tr>
322 *
323 * <tr><th style="vertical-align:top; font-weight:normal" id="concat"><i>XY</i></th>
324 * <td headers="matches logical concat"><i>X</i> followed by <i>Y</i></td></tr>
325 * <tr><th style="vertical-align:top; font-weight:normal" id="alternate"><i>X</i>{@code |}<i>Y</i></th>
326 * <td headers="matches logical alternate">Either <i>X</i> or <i>Y</i></td></tr>
327 * <tr><th style="vertical-align:top; font-weight:normal" id="group">{@code (}<i>X</i>{@code )}</th>
328 * <td headers="matches logical group">X, as a <a href="#cg">capturing group</a></td></tr>
329 *
330 * <tr><th colspan="2" style="padding-top:20px" id="backref">Back references</th></tr>
331 *
332 * <tr><th style="vertical-align:top; font-weight:normal" id="back_nth">{@code \}<i>n</i></th>
333 * <td headers="matches backref back_nth">Whatever the <i>n</i><sup>th</sup>
334 * <a href="#cg">capturing group</a> matched</td></tr>
335 * <tr><th style="vertical-align:top; font-weight:normal" id="back_named">{@code \}<i>k</i><<i>name</i>></th>
336 * <td headers="matches backref back_named">Whatever the
337 * <a href="#groupname">named-capturing group</a> "name" matched</td></tr>
338 *
339 * <tr><th colspan="2" style="padding-top:20px" id="quote">Quotation</th></tr>
340 *
341 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_follow">{@code \}</th>
342 * <td headers="matches quote quote_follow">Nothing, but quotes the following character</td></tr>
343 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_begin">{@code \Q}</th>
344 * <td headers="matches quote quote_begin">Nothing, but quotes all characters until {@code \E}</td></tr>
345 * <tr><th style="vertical-align:top; font-weight:normal" id="quote_end">{@code \E}</th>
346 * <td headers="matches quote quote_end">Nothing, but ends quoting started by {@code \Q}</td></tr>
347 * <!-- Metachars: !$()*+.<>?[\]^{|} -->
348 *
349 * <tr><th colspan="2" style="padding-top:20px" id="special">Special constructs (named-capturing and non-capturing)</th></tr>
350 *
351 * <tr><th style="vertical-align:top; font-weight:normal" id="named_group"><code>(?<<a href="#groupname">name</a>></code><i>X</i>{@code )}</th>
352 * <td headers="matches special named_group"><i>X</i>, as a named-capturing group</td></tr>
353 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group">{@code (?:}<i>X</i>{@code )}</th>
354 * <td headers="matches special non_capture_group"><i>X</i>, as a non-capturing group</td></tr>
355 * <tr><th style="vertical-align:top; font-weight:normal" id="flags"><code>(?idmsuxU-idmsuxU) </code></th>
356 * <td headers="matches special flags">Nothing, but turns match flags <a href="#CASE_INSENSITIVE">i</a>
357 * <a href="#UNIX_LINES">d</a> <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a>
358 * <a href="#UNICODE_CASE">u</a> <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a>
359 * on - off</td></tr>
360 * <tr><th style="vertical-align:top; font-weight:normal" id="non_capture_group_flags"><code>(?idmsuxU-idmsuxU:</code><i>X</i>{@code )} </th>
361 * <td headers="matches special non_capture_group_flags"><i>X</i>, as a <a href="#cg">non-capturing group</a> with the
362 * given flags <a href="#CASE_INSENSITIVE">i</a> <a href="#UNIX_LINES">d</a>
363 * <a href="#MULTILINE">m</a> <a href="#DOTALL">s</a> <a href="#UNICODE_CASE">u</a >
364 * <a href="#COMMENTS">x</a> <a href="#UNICODE_CHARACTER_CLASS">U</a> on - off</td></tr>
365 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookahead">{@code (?=}<i>X</i>{@code )}</th>
366 * <td headers="matches special pos_lookahead"><i>X</i>, via zero-width positive lookahead</td></tr>
367 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookahead">{@code (?!}<i>X</i>{@code )}</th>
368 * <td headers="matches special neg_lookahead"><i>X</i>, via zero-width negative lookahead</td></tr>
369 * <tr><th style="vertical-align:top; font-weight:normal" id="pos_lookbehind">{@code (?<=}<i>X</i>{@code )}</th>
370 * <td headers="matches special pos_lookbehind"><i>X</i>, via zero-width positive lookbehind</td></tr>
371 * <tr><th style="vertical-align:top; font-weight:normal" id="neg_lookbehind">{@code (?<!}<i>X</i>{@code )}</th>
372 * <td headers="matches special neg_lookbehind"><i>X</i>, via zero-width negative lookbehind</td></tr>
373 * <tr><th style="vertical-align:top; font-weight:normal" id="indep_non_capture_group">{@code (?>}<i>X</i>{@code )}</th>
374 * <td headers="matches special indep_non_capture_group"><i>X</i>, as an independent, non-capturing group</td></tr>
375 *
376 * </tbody>
377 * </table>
378 *
379 * <hr>
380 *
381 *
382 * <h2><a id="bs">Backslashes, escapes, and quoting</a></h2>
383 *
384 * <p> The backslash character ({@code '\'}) serves to introduce escaped
385 * constructs, as defined in the table above, as well as to quote characters
386 * that otherwise would be interpreted as unescaped constructs. Thus the
387 * expression {@code \\} matches a single backslash and <code>\{</code> matches a
388 * left brace.
389 *
390 * <p> It is an error to use a backslash prior to any alphabetic character that
391 * does not denote an escaped construct; these are reserved for future
392 * extensions to the regular-expression language. A backslash may be used
393 * prior to a non-alphabetic character regardless of whether that character is
394 * part of an unescaped construct.
395 *
396 * <p> Backslashes within string literals in Java source code are interpreted
397 * as required by
398 * <cite>The Java™ Language Specification</cite>
399 * as either Unicode escapes (section 3.3) or other character escapes (section 3.10.6)
400 * It is therefore necessary to double backslashes in string
401 * literals that represent regular expressions to protect them from
402 * interpretation by the Java bytecode compiler. The string literal
403 * <code>"\b"</code>, for example, matches a single backspace character when
404 * interpreted as a regular expression, while {@code "\\b"} matches a
405 * word boundary. The string literal {@code "\(hello\)"} is illegal
406 * and leads to a compile-time error; in order to match the string
407 * {@code (hello)} the string literal {@code "\\(hello\\)"}
408 * must be used.
409 *
410 * <h2><a id="cc">Character Classes</a></h2>
411 *
412 * <p> Character classes may appear within other character classes, and
413 * may be composed by the union operator (implicit) and the intersection
414 * operator ({@code &&}).
415 * The union operator denotes a class that contains every character that is
416 * in at least one of its operand classes. The intersection operator
417 * denotes a class that contains every character that is in both of its
418 * operand classes.
419 *
420 * <p> The precedence of character-class operators is as follows, from
421 * highest to lowest:
422 *
423 * <table class="striped" style="margin-left: 2em;">
424 * <caption style="display:none">Precedence of character class operators.</caption>
425 * <thead>
426 * <tr><th scope="col">Precedence<th scope="col">Name<th scope="col">Example
427 * </thead>
428 * <tbody>
429 * <tr><th scope="row">1</th>
430 * <td>Literal escape </td>
431 * <td>{@code \x}</td></tr>
432 * <tr><th scope="row">2</th>
433 * <td>Grouping</td>
434 * <td>{@code [...]}</td></tr>
435 * <tr><th scope="row">3</th>
436 * <td>Range</td>
437 * <td>{@code a-z}</td></tr>
438 * <tr><th scope="row">4</th>
439 * <td>Union</td>
440 * <td>{@code [a-e][i-u]}</td></tr>
441 * <tr><th scope="row">5</th>
442 * <td>Intersection</td>
443 * <td>{@code [a-z&&[aeiou]]}</td></tr>
444 * </tbody>
445 * </table>
446 *
447 * <p> Note that a different set of metacharacters are in effect inside
448 * a character class than outside a character class. For instance, the
449 * regular expression {@code .} loses its special meaning inside a
450 * character class, while the expression {@code -} becomes a range
451 * forming metacharacter.
452 *
453 * <h2><a id="lt">Line terminators</a></h2>
454 *
455 * <p> A <i>line terminator</i> is a one- or two-character sequence that marks
456 * the end of a line of the input character sequence. The following are
457 * recognized as line terminators:
458 *
459 * <ul>
460 *
461 * <li> A newline (line feed) character ({@code '\n'}),
462 *
463 * <li> A carriage-return character followed immediately by a newline
464 * character ({@code "\r\n"}),
465 *
466 * <li> A standalone carriage-return character ({@code '\r'}),
467 *
468 * <li> A next-line character (<code>'\u0085'</code>),
469 *
470 * <li> A line-separator character (<code>'\u2028'</code>), or
471 *
472 * <li> A paragraph-separator character (<code>'\u2029'</code>).
473 *
474 * </ul>
475 * <p>If {@link #UNIX_LINES} mode is activated, then the only line terminators
476 * recognized are newline characters.
477 *
478 * <p> The regular expression {@code .} matches any character except a line
479 * terminator unless the {@link #DOTALL} flag is specified.
480 *
481 * <p> By default, the regular expressions {@code ^} and {@code $} ignore
482 * line terminators and only match at the beginning and the end, respectively,
483 * of the entire input sequence. If {@link #MULTILINE} mode is activated then
484 * {@code ^} matches at the beginning of input and after any line terminator
485 * except at the end of input. When in {@link #MULTILINE} mode {@code $}
486 * matches just before a line terminator or the end of the input sequence.
487 *
488 * <h2><a id="cg">Groups and capturing</a></h2>
489 *
490 * <h3><a id="gnumber">Group number</a></h3>
491 * <p> Capturing groups are numbered by counting their opening parentheses from
492 * left to right. In the expression {@code ((A)(B(C)))}, for example, there
493 * are four such groups: </p>
494 *
495 * <ol style="margin-left:2em;">
496 * <li> {@code ((A)(B(C)))}
497 * <li> {@code (A)}
498 * <li> {@code (B(C))}
499 * <li> {@code (C)}
500 * </ol>
501 *
502 * <p> Group zero always stands for the entire expression.
503 *
504 * <p> Capturing groups are so named because, during a match, each subsequence
505 * of the input sequence that matches such a group is saved. The captured
506 * subsequence may be used later in the expression, via a back reference, and
507 * may also be retrieved from the matcher once the match operation is complete.
508 *
509 * <h3><a id="groupname">Group name</a></h3>
510 * <p>A capturing group can also be assigned a "name", a {@code named-capturing group},
511 * and then be back-referenced later by the "name". Group names are composed of
512 * the following characters. The first character must be a {@code letter}.
513 *
514 * <ul>
515 * <li> The uppercase letters {@code 'A'} through {@code 'Z'}
516 * (<code>'\u0041'</code> through <code>'\u005a'</code>),
517 * <li> The lowercase letters {@code 'a'} through {@code 'z'}
518 * (<code>'\u0061'</code> through <code>'\u007a'</code>),
519 * <li> The digits {@code '0'} through {@code '9'}
520 * (<code>'\u0030'</code> through <code>'\u0039'</code>),
521 * </ul>
522 *
523 * <p> A {@code named-capturing group} is still numbered as described in
524 * <a href="#gnumber">Group number</a>.
525 *
526 * <p> The captured input associated with a group is always the subsequence
527 * that the group most recently matched. If a group is evaluated a second time
528 * because of quantification then its previously-captured value, if any, will
529 * be retained if the second evaluation fails. Matching the string
530 * {@code "aba"} against the expression {@code (a(b)?)+}, for example, leaves
531 * group two set to {@code "b"}. All captured input is discarded at the
532 * beginning of each match.
533 *
534 * <p> Groups beginning with {@code (?} are either pure, <i>non-capturing</i> groups
535 * that do not capture text and do not count towards the group total, or
536 * <i>named-capturing</i> group.
537 *
538 * <h2> Unicode support </h2>
539 *
540 * <p> This class is in conformance with Level 1 of <a
541 * href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical
542 * Standard #18: Unicode Regular Expression</i></a>, plus RL2.1
543 * Canonical Equivalents and RL2.2 Extended Grapheme Clusters.
544 * <p>
545 * <b>Unicode escape sequences</b> such as <code>\u2014</code> in Java source code
546 * are processed as described in section 3.3 of
547 * <cite>The Java™ Language Specification</cite>.
548 * Such escape sequences are also implemented directly by the regular-expression
549 * parser so that Unicode escapes can be used in expressions that are read from
550 * files or from the keyboard. Thus the strings <code>"\u2014"</code> and
551 * {@code "\\u2014"}, while not equal, compile into the same pattern, which
552 * matches the character with hexadecimal value {@code 0x2014}.
553 * <p>
554 * A Unicode character can also be represented by using its <b>Hex notation</b>
555 * (hexadecimal code point value) directly as described in construct
556 * <code>\x{...}</code>, for example a supplementary character U+2011F can be
557 * specified as <code>\x{2011F}</code>, instead of two consecutive Unicode escape
558 * sequences of the surrogate pair <code>\uD840</code><code>\uDD1F</code>.
559 * <p>
560 * <b>Unicode character names</b> are supported by the named character construct
561 * <code>\N{</code>...<code>}</code>, for example, <code>\N{WHITE SMILING FACE}</code>
562 * specifies character <code>\u263A</code>. The character names supported
563 * by this class are the valid Unicode character names matched by
564 * {@link java.lang.Character#codePointOf(String) Character.codePointOf(name)}.
565 * <p>
566 * <a href="http://www.unicode.org/reports/tr18/#Default_Grapheme_Clusters">
567 * <b>Unicode extended grapheme clusters</b></a> are supported by the grapheme
568 * cluster matcher {@code \X} and the corresponding boundary matcher {@code \b{g}}.
569 * <p>
570 * Unicode scripts, blocks, categories and binary properties are written with
571 * the {@code \p} and {@code \P} constructs as in Perl.
572 * <code>\p{</code><i>prop</i><code>}</code> matches if
573 * the input has the property <i>prop</i>, while <code>\P{</code><i>prop</i><code>}</code>
574 * does not match if the input has that property.
575 * <p>
576 * Scripts, blocks, categories and binary properties can be used both inside
577 * and outside of a character class.
578 *
579 * <p>
580 * <b><a id="usc">Scripts</a></b> are specified either with the prefix {@code Is}, as in
581 * {@code IsHiragana}, or by using the {@code script} keyword (or its short
582 * form {@code sc}) as in {@code script=Hiragana} or {@code sc=Hiragana}.
583 * <p>
584 * The script names supported by {@code Pattern} are the valid script names
585 * accepted and defined by
586 * {@link java.lang.Character.UnicodeScript#forName(String) UnicodeScript.forName}.
587 *
588 * <p>
589 * <b><a id="ubc">Blocks</a></b> are specified with the prefix {@code In}, as in
590 * {@code InMongolian}, or by using the keyword {@code block} (or its short
591 * form {@code blk}) as in {@code block=Mongolian} or {@code blk=Mongolian}.
592 * <p>
593 * The block names supported by {@code Pattern} are the valid block names
594 * accepted and defined by
595 * {@link java.lang.Character.UnicodeBlock#forName(String) UnicodeBlock.forName}.
596 * <p>
597 *
598 * <b><a id="ucc">Categories</a></b> may be specified with the optional prefix {@code Is}:
599 * Both {@code \p{L}} and {@code \p{IsL}} denote the category of Unicode
600 * letters. Same as scripts and blocks, categories can also be specified
601 * by using the keyword {@code general_category} (or its short form
602 * {@code gc}) as in {@code general_category=Lu} or {@code gc=Lu}.
603 * <p>
604 * The supported categories are those of
605 * <a href="http://www.unicode.org/unicode/standard/standard.html">
606 * <i>The Unicode Standard</i></a> in the version specified by the
607 * {@link java.lang.Character Character} class. The category names are those
608 * defined in the Standard, both normative and informative.
609 * <p>
610 *
611 * <b><a id="ubpc">Binary properties</a></b> are specified with the prefix {@code Is}, as in
612 * {@code IsAlphabetic}. The supported binary properties by {@code Pattern}
613 * are
614 * <ul>
615 * <li> Alphabetic
616 * <li> Ideographic
617 * <li> Letter
618 * <li> Lowercase
619 * <li> Uppercase
620 * <li> Titlecase
621 * <li> Punctuation
622 * <Li> Control
623 * <li> White_Space
624 * <li> Digit
625 * <li> Hex_Digit
626 * <li> Join_Control
627 * <li> Noncharacter_Code_Point
628 * <li> Assigned
629 * </ul>
630 * <p>
631 * The following <b>Predefined Character classes</b> and <b>POSIX character classes</b>
632 * are in conformance with the recommendation of <i>Annex C: Compatibility Properties</i>
633 * of <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Regular Expression
634 * </i></a>, when {@link #UNICODE_CHARACTER_CLASS} flag is specified.
635 *
636 * <table class="striped">
637 * <caption style="display:none">predefined and posix character classes in Unicode mode</caption>
638 * <thead>
639 * <tr>
640 * <th scope="col" id="predef_classes">Classes</th>
641 * <th scope="col" id="predef_matches">Matches</th>
642 * </tr>
643 * </thead>
644 * <tbody>
645 * <tr><th scope="row">{@code \p{Lower}}</th>
646 * <td>A lowercase character:{@code \p{IsLowercase}}</td></tr>
647 * <tr><th scope="row">{@code \p{Upper}}</th>
648 * <td>An uppercase character:{@code \p{IsUppercase}}</td></tr>
649 * <tr><th scope="row">{@code \p{ASCII}}</th>
650 * <td>All ASCII:{@code [\x00-\x7F]}</td></tr>
651 * <tr><th scope="row">{@code \p{Alpha}}</th>
652 * <td>An alphabetic character:{@code \p{IsAlphabetic}}</td></tr>
653 * <tr><th scope="row">{@code \p{Digit}}</th>
654 * <td>A decimal digit character:{@code \p{IsDigit}}</td></tr>
655 * <tr><th scope="row">{@code \p{Alnum}}</th>
656 * <td>An alphanumeric character:{@code [\p{IsAlphabetic}\p{IsDigit}]}</td></tr>
657 * <tr><th scope="row">{@code \p{Punct}}</th>
658 * <td>A punctuation character:{@code \p{IsPunctuation}}</td></tr>
659 * <tr><th scope="row">{@code \p{Graph}}</th>
660 * <td>A visible character: {@code [^\p{IsWhite_Space}\p{gc=Cc}\p{gc=Cs}\p{gc=Cn}]}</td></tr>
661 * <tr><th scope="row">{@code \p{Print}}</th>
662 * <td>A printable character: {@code [\p{Graph}\p{Blank}&&[^\p{Cntrl}]]}</td></tr>
663 * <tr><th scope="row">{@code \p{Blank}}</th>
664 * <td>A space or a tab: {@code [\p{IsWhite_Space}&&[^\p{gc=Zl}\p{gc=Zp}\x0a\x0b\x0c\x0d\x85]]}</td></tr>
665 * <tr><th scope="row">{@code \p{Cntrl}}</th>
666 * <td>A control character: {@code \p{gc=Cc}}</td></tr>
667 * <tr><th scope="row">{@code \p{XDigit}}</th>
668 * <td>A hexadecimal digit: {@code [\p{gc=Nd}\p{IsHex_Digit}]}</td></tr>
669 * <tr><th scope="row">{@code \p{Space}}</th>
670 * <td>A whitespace character:{@code \p{IsWhite_Space}}</td></tr>
671 * <tr><th scope="row">{@code \d}</th>
672 * <td>A digit: {@code \p{IsDigit}}</td></tr>
673 * <tr><th scope="row">{@code \D}</th>
674 * <td>A non-digit: {@code [^\d]}</td></tr>
675 * <tr><th scope="row">{@code \s}</th>
676 * <td>A whitespace character: {@code \p{IsWhite_Space}}</td></tr>
677 * <tr><th scope="row">{@code \S}</th>
678 * <td>A non-whitespace character: {@code [^\s]}</td></tr>
679 * <tr><th scope="row">{@code \w}</th>
680 * <td>A word character: {@code [\p{Alpha}\p{gc=Mn}\p{gc=Me}\p{gc=Mc}\p{Digit}\p{gc=Pc}\p{IsJoin_Control}]}</td></tr>
681 * <tr><th scope="row">{@code \W}</th>
682 * <td>A non-word character: {@code [^\w]}</td></tr>
683 * </tbody>
684 * </table>
685 * <p>
686 * <a id="jcc">
687 * Categories that behave like the java.lang.Character
688 * boolean is<i>methodname</i> methods (except for the deprecated ones) are
689 * available through the same <code>\p{</code><i>prop</i><code>}</code> syntax where
690 * the specified property has the name <code>java<i>methodname</i></code></a>.
691 *
692 * <h2> Comparison to Perl 5 </h2>
693 *
694 * <p>The {@code Pattern} engine performs traditional NFA-based matching
695 * with ordered alternation as occurs in Perl 5.
696 *
697 * <p> Perl constructs not supported by this class: </p>
698 *
699 * <ul>
700 * <li><p> The backreference constructs, <code>\g{</code><i>n</i><code>}</code> for
701 * the <i>n</i><sup>th</sup><a href="#cg">capturing group</a> and
702 * <code>\g{</code><i>name</i><code>}</code> for
703 * <a href="#groupname">named-capturing group</a>.
704 * </p></li>
705 *
706 * <li><p> The conditional constructs
707 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code )} and
708 * {@code (?(}<i>condition</i>{@code )}<i>X</i>{@code |}<i>Y</i>{@code )},
709 * </p></li>
710 *
711 * <li><p> The embedded code constructs <code>(?{</code><i>code</i><code>})</code>
712 * and <code>(??{</code><i>code</i><code>})</code>,</p></li>
713 *
714 * <li><p> The embedded comment syntax {@code (?#comment)}, and </p></li>
715 *
716 * <li><p> The preprocessing operations {@code \l} <code>\u</code>,
717 * {@code \L}, and {@code \U}. </p></li>
718 *
719 * </ul>
720 *
721 * <p> Constructs supported by this class but not by Perl: </p>
722 *
723 * <ul>
724 *
725 * <li><p> Character-class union and intersection as described
726 * <a href="#cc">above</a>.</p></li>
727 *
728 * </ul>
729 *
730 * <p> Notable differences from Perl: </p>
731 *
732 * <ul>
733 *
734 * <li><p> In Perl, {@code \1} through {@code \9} are always interpreted
735 * as back references; a backslash-escaped number greater than {@code 9} is
736 * treated as a back reference if at least that many subexpressions exist,
737 * otherwise it is interpreted, if possible, as an octal escape. In this
738 * class octal escapes must always begin with a zero. In this class,
739 * {@code \1} through {@code \9} are always interpreted as back
740 * references, and a larger number is accepted as a back reference if at
741 * least that many subexpressions exist at that point in the regular
742 * expression, otherwise the parser will drop digits until the number is
743 * smaller or equal to the existing number of groups or it is one digit.
744 * </p></li>
745 *
746 * <li><p> Perl uses the {@code g} flag to request a match that resumes
747 * where the last match left off. This functionality is provided implicitly
748 * by the {@link Matcher} class: Repeated invocations of the {@link
749 * Matcher#find find} method will resume where the last match left off,
750 * unless the matcher is reset. </p></li>
751 *
752 * <li><p> In Perl, embedded flags at the top level of an expression affect
753 * the whole expression. In this class, embedded flags always take effect
754 * at the point at which they appear, whether they are at the top level or
755 * within a group; in the latter case, flags are restored at the end of the
756 * group just as in Perl. </p></li>
757 *
758 * </ul>
759 *
760 *
761 * <p> For a more precise description of the behavior of regular expression
762 * constructs, please see <a href="http://www.oreilly.com/catalog/regex3/">
763 * <i>Mastering Regular Expressions, 3nd Edition</i>, Jeffrey E. F. Friedl,
764 * O'Reilly and Associates, 2006.</a>
765 * </p>
766 *
767 * @see java.lang.String#split(String, int)
768 * @see java.lang.String#split(String)
769 *
770 * @author Mike McCloskey
771 * @author Mark Reinhold
772 * @author JSR-51 Expert Group
773 * @since 1.4
774 * @spec JSR-51
775 */
776
777 public final class Pattern
778 implements java.io.Serializable
779 {
780
781 /**
782 * Regular expression modifier values. Instead of being passed as
783 * arguments, they can also be passed as inline modifiers.
784 * For example, the following statements have the same effect.
785 * <pre>
786 * Pattern p1 = Pattern.compile("abc", Pattern.CASE_INSENSITIVE|Pattern.MULTILINE);
787 * Pattern p2 = Pattern.compile("(?im)abc", 0);
788 * </pre>
789 */
790
791 /**
792 * Enables Unix lines mode.
793 *
794 * <p> In this mode, only the {@code '\n'} line terminator is recognized
795 * in the behavior of {@code .}, {@code ^}, and {@code $}.
796 *
797 * <p> Unix lines mode can also be enabled via the embedded flag
798 * expression {@code (?d)}.
799 */
800 public static final int UNIX_LINES = 0x01;
801
802 /**
803 * Enables case-insensitive matching.
804 *
805 * <p> By default, case-insensitive matching assumes that only characters
806 * in the US-ASCII charset are being matched. Unicode-aware
807 * case-insensitive matching can be enabled by specifying the {@link
808 * #UNICODE_CASE} flag in conjunction with this flag.
809 *
810 * <p> Case-insensitive matching can also be enabled via the embedded flag
811 * expression {@code (?i)}.
812 *
813 * <p> Specifying this flag may impose a slight performance penalty. </p>
814 */
815 public static final int CASE_INSENSITIVE = 0x02;
816
817 /**
818 * Permits whitespace and comments in pattern.
819 *
820 * <p> In this mode, whitespace is ignored, and embedded comments starting
821 * with {@code #} are ignored until the end of a line.
822 *
823 * <p> Comments mode can also be enabled via the embedded flag
824 * expression {@code (?x)}.
825 */
826 public static final int COMMENTS = 0x04;
827
828 /**
829 * Enables multiline mode.
830 *
831 * <p> In multiline mode the expressions {@code ^} and {@code $} match
832 * just after or just before, respectively, a line terminator or the end of
833 * the input sequence. By default these expressions only match at the
834 * beginning and the end of the entire input sequence.
835 *
836 * <p> Multiline mode can also be enabled via the embedded flag
837 * expression {@code (?m)}. </p>
838 */
839 public static final int MULTILINE = 0x08;
840
841 /**
842 * Enables literal parsing of the pattern.
843 *
844 * <p> When this flag is specified then the input string that specifies
845 * the pattern is treated as a sequence of literal characters.
846 * Metacharacters or escape sequences in the input sequence will be
847 * given no special meaning.
848 *
849 * <p>The flags CASE_INSENSITIVE and UNICODE_CASE retain their impact on
850 * matching when used in conjunction with this flag. The other flags
851 * become superfluous.
852 *
853 * <p> There is no embedded flag character for enabling literal parsing.
854 * @since 1.5
855 */
856 public static final int LITERAL = 0x10;
857
858 /**
859 * Enables dotall mode.
860 *
861 * <p> In dotall mode, the expression {@code .} matches any character,
862 * including a line terminator. By default this expression does not match
863 * line terminators.
864 *
865 * <p> Dotall mode can also be enabled via the embedded flag
866 * expression {@code (?s)}. (The {@code s} is a mnemonic for
867 * "single-line" mode, which is what this is called in Perl.) </p>
868 */
869 public static final int DOTALL = 0x20;
870
871 /**
872 * Enables Unicode-aware case folding.
873 *
874 * <p> When this flag is specified then case-insensitive matching, when
875 * enabled by the {@link #CASE_INSENSITIVE} flag, is done in a manner
876 * consistent with the Unicode Standard. By default, case-insensitive
877 * matching assumes that only characters in the US-ASCII charset are being
878 * matched.
879 *
880 * <p> Unicode-aware case folding can also be enabled via the embedded flag
881 * expression {@code (?u)}.
882 *
883 * <p> Specifying this flag may impose a performance penalty. </p>
884 */
885 public static final int UNICODE_CASE = 0x40;
886
887 /**
888 * Enables canonical equivalence.
889 *
890 * <p> When this flag is specified then two characters will be considered
891 * to match if, and only if, their full canonical decompositions match.
892 * The expression <code>"a\u030A"</code>, for example, will match the
893 * string <code>"\u00E5"</code> when this flag is specified. By default,
894 * matching does not take canonical equivalence into account.
895 *
896 * <p> There is no embedded flag character for enabling canonical
897 * equivalence.
898 *
899 * <p> Specifying this flag may impose a performance penalty. </p>
900 */
901 public static final int CANON_EQ = 0x80;
902
903 /**
904 * Enables the Unicode version of <i>Predefined character classes</i> and
905 * <i>POSIX character classes</i>.
906 *
907 * <p> When this flag is specified then the (US-ASCII only)
908 * <i>Predefined character classes</i> and <i>POSIX character classes</i>
909 * are in conformance with
910 * <a href="http://www.unicode.org/reports/tr18/"><i>Unicode Technical
911 * Standard #18: Unicode Regular Expression</i></a>
912 * <i>Annex C: Compatibility Properties</i>.
913 * <p>
914 * The UNICODE_CHARACTER_CLASS mode can also be enabled via the embedded
915 * flag expression {@code (?U)}.
916 * <p>
917 * The flag implies UNICODE_CASE, that is, it enables Unicode-aware case
918 * folding.
919 * <p>
920 * Specifying this flag may impose a performance penalty. </p>
921 * @since 1.7
922 */
923 public static final int UNICODE_CHARACTER_CLASS = 0x100;
924
925 /**
926 * Contains all possible flags for compile(regex, flags).
927 */
928 private static final int ALL_FLAGS = CASE_INSENSITIVE | MULTILINE |
929 DOTALL | UNICODE_CASE | CANON_EQ | UNIX_LINES | LITERAL |
930 UNICODE_CHARACTER_CLASS | COMMENTS;
931
932 /* Pattern has only two serialized components: The pattern string
933 * and the flags, which are all that is needed to recompile the pattern
934 * when it is deserialized.
935 */
936
937 /** use serialVersionUID from Merlin b59 for interoperability */
938 @java.io.Serial
939 private static final long serialVersionUID = 5073258162644648461L;
940
941 /**
942 * The original regular-expression pattern string.
943 *
944 * @serial
945 */
946 private String pattern;
947
948 /**
949 * The original pattern flags.
950 *
951 * @serial
952 */
953 private int flags;
954
955 /**
956 * The temporary pattern flags used during compiling. The flags might be turn
957 * on and off by embedded flag.
958 */
959 private transient int flags0;
960
961 /**
962 * Boolean indicating this Pattern is compiled; this is necessary in order
963 * to lazily compile deserialized Patterns.
964 */
965 private transient volatile boolean compiled;
966
967 /**
968 * The normalized pattern string.
969 */
970 private transient String normalizedPattern;
971
972 /**
973 * The starting point of state machine for the find operation. This allows
974 * a match to start anywhere in the input.
975 */
976 transient Node root;
977
978 /**
979 * The root of object tree for a match operation. The pattern is matched
980 * at the beginning. This may include a find that uses BnM or a First
981 * node.
982 */
983 transient Node matchRoot;
984
985 /**
986 * Temporary storage used by parsing pattern slice.
987 */
988 transient int[] buffer;
989
990 /**
991 * A temporary storage used for predicate for double return.
992 */
993 transient CharPredicate predicate;
994
995 /**
996 * Map the "name" of the "named capturing group" to its group id
997 * node.
998 */
999 transient volatile Map<String, Integer> namedGroups;
1000
1001 /**
1002 * Temporary storage used while parsing group references.
1003 */
1004 transient GroupHead[] groupNodes;
1005
1006 /**
1007 * Temporary storage used to store the top level closure nodes.
1008 */
1009 transient List<Node> topClosureNodes;
1010
1011 /**
1012 * The number of top greedy closure nodes in this Pattern. Used by
1013 * matchers to allocate storage needed for a IntHashSet to keep the
1014 * beginning pos {@code i} of all failed match.
1015 */
1016 transient int localTCNCount;
1017
1018 /*
1019 * Turn off the stop-exponential-backtracking optimization if there
1020 * is a group ref in the pattern.
1021 */
1022 transient boolean hasGroupRef;
1023
1024 /**
1025 * Temporary null terminated code point array used by pattern compiling.
1026 */
1027 private transient int[] temp;
1028
1029 /**
1030 * The number of capturing groups in this Pattern. Used by matchers to
1031 * allocate storage needed to perform a match.
1032 */
1033 transient int capturingGroupCount;
1034
1035 /**
1036 * The local variable count used by parsing tree. Used by matchers to
1037 * allocate storage needed to perform a match.
1038 */
1039 transient int localCount;
1040
1041 /**
1042 * Index into the pattern string that keeps track of how much has been
1043 * parsed.
1044 */
1045 private transient int cursor;
1046
1047 /**
1048 * Holds the length of the pattern string.
1049 */
1050 private transient int patternLength;
1051
1052 /**
1053 * If the Start node might possibly match supplementary characters.
1054 * It is set to true during compiling if
1055 * (1) There is supplementary char in pattern, or
1056 * (2) There is complement node of a "family" CharProperty
1057 */
1058 private transient boolean hasSupplementary;
1059
1060 /**
1061 * Compiles the given regular expression into a pattern.
1062 *
1063 * @param regex
1064 * The expression to be compiled
1065 * @return the given regular expression compiled into a pattern
1066 * @throws PatternSyntaxException
1067 * If the expression's syntax is invalid
1068 */
1069 public static Pattern compile(String regex) {
1070 return new Pattern(regex, 0);
1071 }
1072
1073 /**
1074 * Compiles the given regular expression into a pattern with the given
1075 * flags.
1076 *
1077 * @param regex
1078 * The expression to be compiled
1079 *
1080 * @param flags
1081 * Match flags, a bit mask that may include
1082 * {@link #CASE_INSENSITIVE}, {@link #MULTILINE}, {@link #DOTALL},
1083 * {@link #UNICODE_CASE}, {@link #CANON_EQ}, {@link #UNIX_LINES},
1084 * {@link #LITERAL}, {@link #UNICODE_CHARACTER_CLASS}
1085 * and {@link #COMMENTS}
1086 *
1087 * @return the given regular expression compiled into a pattern with the given flags
1088 * @throws IllegalArgumentException
1089 * If bit values other than those corresponding to the defined
1090 * match flags are set in {@code flags}
1091 *
1092 * @throws PatternSyntaxException
1093 * If the expression's syntax is invalid
1094 */
1095 public static Pattern compile(String regex, int flags) {
1096 return new Pattern(regex, flags);
1097 }
1098
1099 /**
1100 * Returns the regular expression from which this pattern was compiled.
1101 *
1102 * @return The source of this pattern
1103 */
1104 public String pattern() {
1105 return pattern;
1106 }
1107
1108 /**
1109 * <p>Returns the string representation of this pattern. This
1110 * is the regular expression from which this pattern was
1111 * compiled.</p>
1112 *
1113 * @return The string representation of this pattern
1114 * @since 1.5
1115 */
1116 public String toString() {
1117 return pattern;
1118 }
1119
1120 /**
1121 * Creates a matcher that will match the given input against this pattern.
1122 *
1123 * @param input
1124 * The character sequence to be matched
1125 *
1126 * @return A new matcher for this pattern
1127 */
1128 public Matcher matcher(CharSequence input) {
1129 if (!compiled) {
1130 synchronized(this) {
1131 if (!compiled)
1132 compile();
1133 }
1134 }
1135 Matcher m = new Matcher(this, input);
1136 return m;
1137 }
1138
1139 /**
1140 * Returns this pattern's match flags.
1141 *
1142 * @return The match flags specified when this pattern was compiled
1143 */
1144 public int flags() {
1145 return flags0;
1146 }
1147
1148 /**
1149 * Compiles the given regular expression and attempts to match the given
1150 * input against it.
1151 *
1152 * <p> An invocation of this convenience method of the form
1153 *
1154 * <blockquote><pre>
1155 * Pattern.matches(regex, input);</pre></blockquote>
1156 *
1157 * behaves in exactly the same way as the expression
1158 *
1159 * <blockquote><pre>
1160 * Pattern.compile(regex).matcher(input).matches()</pre></blockquote>
1161 *
1162 * <p> If a pattern is to be used multiple times, compiling it once and reusing
1163 * it will be more efficient than invoking this method each time. </p>
1164 *
1165 * @param regex
1166 * The expression to be compiled
1167 *
1168 * @param input
1169 * The character sequence to be matched
1170 * @return whether or not the regular expression matches on the input
1171 * @throws PatternSyntaxException
1172 * If the expression's syntax is invalid
1173 */
1174 public static boolean matches(String regex, CharSequence input) {
1175 Pattern p = Pattern.compile(regex);
1176 Matcher m = p.matcher(input);
1177 return m.matches();
1178 }
1179
1180 /**
1181 * Splits the given input sequence around matches of this pattern.
1182 *
1183 * <p> The array returned by this method contains each substring of the
1184 * input sequence that is terminated by another subsequence that matches
1185 * this pattern or is terminated by the end of the input sequence. The
1186 * substrings in the array are in the order in which they occur in the
1187 * input. If this pattern does not match any subsequence of the input then
1188 * the resulting array has just one element, namely the input sequence in
1189 * string form.
1190 *
1191 * <p> When there is a positive-width match at the beginning of the input
1192 * sequence then an empty leading substring is included at the beginning
1193 * of the resulting array. A zero-width match at the beginning however
1194 * never produces such empty leading substring.
1195 *
1196 * <p> The {@code limit} parameter controls the number of times the
1197 * pattern is applied and therefore affects the length of the resulting
1198 * array.
1199 * <ul>
1200 * <li><p>
1201 * If the <i>limit</i> is positive then the pattern will be applied
1202 * at most <i>limit</i> - 1 times, the array's length will be
1203 * no greater than <i>limit</i>, and the array's last entry will contain
1204 * all input beyond the last matched delimiter.</p></li>
1205 *
1206 * <li><p>
1207 * If the <i>limit</i> is zero then the pattern will be applied as
1208 * many times as possible, the array can have any length, and trailing
1209 * empty strings will be discarded.</p></li>
1210 *
1211 * <li><p>
1212 * If the <i>limit</i> is negative then the pattern will be applied
1213 * as many times as possible and the array can have any length.</p></li>
1214 * </ul>
1215 *
1216 * <p> The input {@code "boo:and:foo"}, for example, yields the following
1217 * results with these parameters:
1218 *
1219 * <table class="plain" style="margin-left:2em;">
1220 * <caption style="display:none">Split example showing regex, limit, and result</caption>
1221 * <thead>
1222 * <tr>
1223 * <th scope="col">Regex</th>
1224 * <th scope="col">Limit</th>
1225 * <th scope="col">Result</th>
1226 * </tr>
1227 * </thead>
1228 * <tbody>
1229 * <tr><th scope="row" rowspan="3" style="font-weight:normal">:</th>
1230 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">2</th>
1231 * <td>{@code { "boo", "and:foo" }}</td></tr>
1232 * <tr><!-- : -->
1233 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th>
1234 * <td>{@code { "boo", "and", "foo" }}</td></tr>
1235 * <tr><!-- : -->
1236 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th>
1237 * <td>{@code { "boo", "and", "foo" }}</td></tr>
1238 * <tr><th scope="row" rowspan="3" style="font-weight:normal">o</th>
1239 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">5</th>
1240 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr>
1241 * <tr><!-- o -->
1242 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">-2</th>
1243 * <td>{@code { "b", "", ":and:f", "", "" }}</td></tr>
1244 * <tr><!-- o -->
1245 * <th scope="row" style="font-weight:normal; text-align:right; padding-right:1em">0</th>
1246 * <td>{@code { "b", "", ":and:f" }}</td></tr>
1247 * </tbody>
1248 * </table>
1249 *
1250 * @param input
1251 * The character sequence to be split
1252 *
1253 * @param limit
1254 * The result threshold, as described above
1255 *
1256 * @return The array of strings computed by splitting the input
1257 * around matches of this pattern
1258 */
1259 public String[] split(CharSequence input, int limit) {
1260 int index = 0;
1261 boolean matchLimited = limit > 0;
1262 ArrayList<String> matchList = new ArrayList<>();
1263 Matcher m = matcher(input);
1264
1265 // Add segments before each match found
1266 while(m.find()) {
1267 if (!matchLimited || matchList.size() < limit - 1) {
1268 if (index == 0 && index == m.start() && m.start() == m.end()) {
1269 // no empty leading substring included for zero-width match
1270 // at the beginning of the input char sequence.
1271 continue;
1272 }
1273 String match = input.subSequence(index, m.start()).toString();
1274 matchList.add(match);
1275 index = m.end();
1276 } else if (matchList.size() == limit - 1) { // last one
1277 String match = input.subSequence(index,
1278 input.length()).toString();
1279 matchList.add(match);
1280 index = m.end();
1281 }
1282 }
1283
1284 // If no match was found, return this
1285 if (index == 0)
1286 return new String[] {input.toString()};
1287
1288 // Add remaining segment
1289 if (!matchLimited || matchList.size() < limit)
1290 matchList.add(input.subSequence(index, input.length()).toString());
1291
1292 // Construct result
1293 int resultSize = matchList.size();
1294 if (limit == 0)
1295 while (resultSize > 0 && matchList.get(resultSize-1).isEmpty())
1296 resultSize--;
1297 String[] result = new String[resultSize];
1298 return matchList.subList(0, resultSize).toArray(result);
1299 }
1300
1301 /**
1302 * Splits the given input sequence around matches of this pattern.
1303 *
1304 * <p> This method works as if by invoking the two-argument {@link
1305 * #split(java.lang.CharSequence, int) split} method with the given input
1306 * sequence and a limit argument of zero. Trailing empty strings are
1307 * therefore not included in the resulting array. </p>
1308 *
1309 * <p> The input {@code "boo:and:foo"}, for example, yields the following
1310 * results with these expressions:
1311 *
1312 * <table class="plain" style="margin-left:2em">
1313 * <caption style="display:none">Split examples showing regex and result</caption>
1314 * <thead>
1315 * <tr>
1316 * <th scope="col">Regex</th>
1317 * <th scope="col">Result</th>
1318 * </tr>
1319 * </thead>
1320 * <tbody>
1321 * <tr><th scope="row" style="text-weight:normal">:</th>
1322 * <td>{@code { "boo", "and", "foo" }}</td></tr>
1323 * <tr><th scope="row" style="text-weight:normal">o</th>
1324 * <td>{@code { "b", "", ":and:f" }}</td></tr>
1325 * </tbody>
1326 * </table>
1327 *
1328 *
1329 * @param input
1330 * The character sequence to be split
1331 *
1332 * @return The array of strings computed by splitting the input
1333 * around matches of this pattern
1334 */
1335 public String[] split(CharSequence input) {
1336 return split(input, 0);
1337 }
1338
1339 /**
1340 * Returns a literal pattern {@code String} for the specified
1341 * {@code String}.
1342 *
1343 * <p>This method produces a {@code String} that can be used to
1344 * create a {@code Pattern} that would match the string
1345 * {@code s} as if it were a literal pattern.</p> Metacharacters
1346 * or escape sequences in the input sequence will be given no special
1347 * meaning.
1348 *
1349 * @param s The string to be literalized
1350 * @return A literal string replacement
1351 * @since 1.5
1352 */
1353 public static String quote(String s) {
1354 int slashEIndex = s.indexOf("\\E");
1355 if (slashEIndex == -1)
1356 return "\\Q" + s + "\\E";
1357
1358 int lenHint = s.length();
1359 lenHint = (lenHint < Integer.MAX_VALUE - 8 - lenHint) ?
1360 (lenHint << 1) : (Integer.MAX_VALUE - 8);
1361
1362 StringBuilder sb = new StringBuilder(lenHint);
1363 sb.append("\\Q");
1364 int current = 0;
1365 do {
1366 sb.append(s, current, slashEIndex)
1367 .append("\\E\\\\E\\Q");
1368 current = slashEIndex + 2;
1369 } while ((slashEIndex = s.indexOf("\\E", current)) != -1);
1370
1371 return sb.append(s, current, s.length())
1372 .append("\\E")
1373 .toString();
1374 }
1375
1376 /**
1377 * Recompile the Pattern instance from a stream. The original pattern
1378 * string is read in and the object tree is recompiled from it.
1379 */
1380 @java.io.Serial
1381 private void readObject(java.io.ObjectInputStream s)
1382 throws java.io.IOException, ClassNotFoundException {
1383
1384 // Read in all fields
1385 s.defaultReadObject();
1386
1387 // reset the flags
1388 flags0 = flags;
1389
1390 // Initialize counts
1391 capturingGroupCount = 1;
1392 localCount = 0;
1393 localTCNCount = 0;
1394
1395 // if length > 0, the Pattern is lazily compiled
1396 if (pattern.isEmpty()) {
1397 root = new Start(lastAccept);
1398 matchRoot = lastAccept;
1399 compiled = true;
1400 }
1401 }
1402
1403 /**
1404 * This private constructor is used to create all Patterns. The pattern
1405 * string and match flags are all that is needed to completely describe
1406 * a Pattern. An empty pattern string results in an object tree with
1407 * only a Start node and a LastNode node.
1408 */
1409 private Pattern(String p, int f) {
1410 if ((f & ~ALL_FLAGS) != 0) {
1411 throw new IllegalArgumentException("Unknown flag 0x"
1412 + Integer.toHexString(f));
1413 }
1414 pattern = p;
1415 flags = f;
1416
1417 // to use UNICODE_CASE if UNICODE_CHARACTER_CLASS present
1418 if ((flags & UNICODE_CHARACTER_CLASS) != 0)
1419 flags |= UNICODE_CASE;
1420
1421 // 'flags' for compiling
1422 flags0 = flags;
1423
1424 // Reset group index count
1425 capturingGroupCount = 1;
1426 localCount = 0;
1427 localTCNCount = 0;
1428
1429 if (!pattern.isEmpty()) {
1430 compile();
1431 } else {
1432 root = new Start(lastAccept);
1433 matchRoot = lastAccept;
1434 }
1435 }
1436
1437 /**
1438 * The pattern is converted to normalized form ({@link
1439 * java.text.Normalizer.Form#NFC NFC}, canonical decomposition,
1440 * followed by canonical composition for the character class
1441 * part, and {@link java.text.Normalizer.Form#NFD NFD},
1442 * canonical decomposition for the rest), and then a pure
1443 * group is constructed to match canonical equivalences of the
1444 * characters.
1445 */
1446 private static String normalize(String pattern) {
1447 int plen = pattern.length();
1448 StringBuilder pbuf = new StringBuilder(plen);
1449 char last = 0;
1450 int lastStart = 0;
1451 char cc = 0;
1452 for (int i = 0; i < plen;) {
1453 char c = pattern.charAt(i);
1454 if (cc == 0 && // top level
1455 c == '\\' && i + 1 < plen && pattern.charAt(i + 1) == '\\') {
1456 i += 2; last = 0;
1457 continue;
1458 }
1459 if (c == '[' && last != '\\') {
1460 if (cc == 0) {
1461 if (lastStart < i)
1462 normalizeSlice(pattern, lastStart, i, pbuf);
1463 lastStart = i;
1464 }
1465 cc++;
1466 } else if (c == ']' && last != '\\') {
1467 cc--;
1468 if (cc == 0) {
1469 normalizeClazz(pattern, lastStart, i + 1, pbuf);
1470 lastStart = i + 1;
1471 }
1472 }
1473 last = c;
1474 i++;
1475 }
1476 assert (cc == 0);
1477 if (lastStart < plen)
1478 normalizeSlice(pattern, lastStart, plen, pbuf);
1479 return pbuf.toString();
1480 }
1481
1482 private static void normalizeSlice(String src, int off, int limit,
1483 StringBuilder dst)
1484 {
1485 int len = src.length();
1486 int off0 = off;
1487 while (off < limit && ASCII.isAscii(src.charAt(off))) {
1488 off++;
1489 }
1490 if (off == limit) {
1491 dst.append(src, off0, limit);
1492 return;
1493 }
1494 off--;
1495 if (off < off0)
1496 off = off0;
1497 else
1498 dst.append(src, off0, off);
1499 while (off < limit) {
1500 int ch0 = src.codePointAt(off);
1501 if (".$|()[]{}^?*+\\".indexOf(ch0) != -1) {
1502 dst.append((char)ch0);
1503 off++;
1504 continue;
1505 }
1506 int j = Grapheme.nextBoundary(src, off, limit);
1507 int ch1;
1508 String seq = src.substring(off, j);
1509 String nfd = Normalizer.normalize(seq, Normalizer.Form.NFD);
1510 off = j;
1511 if (nfd.codePointCount(0, nfd.length()) > 1) {
1512 ch0 = nfd.codePointAt(0);
1513 ch1 = nfd.codePointAt(Character.charCount(ch0));
1514 if (Character.getType(ch1) == Character.NON_SPACING_MARK) {
1515 Set<String> altns = new LinkedHashSet<>();
1516 altns.add(seq);
1517 produceEquivalentAlternation(nfd, altns);
1518 dst.append("(?:");
1519 altns.forEach( s -> dst.append(s).append('|'));
1520 dst.delete(dst.length() - 1, dst.length());
1521 dst.append(")");
1522 continue;
1523 }
1524 }
1525 String nfc = Normalizer.normalize(seq, Normalizer.Form.NFC);
1526 if (!seq.equals(nfc) && !nfd.equals(nfc))
1527 dst.append("(?:" + seq + "|" + nfd + "|" + nfc + ")");
1528 else if (!seq.equals(nfd))
1529 dst.append("(?:" + seq + "|" + nfd + ")");
1530 else
1531 dst.append(seq);
1532 }
1533 }
1534
1535 private static void normalizeClazz(String src, int off, int limit,
1536 StringBuilder dst)
1537 {
1538 dst.append(Normalizer.normalize(src.substring(off, limit), Form.NFC));
1539 }
1540
1541 /**
1542 * Given a specific sequence composed of a regular character and
1543 * combining marks that follow it, produce the alternation that will
1544 * match all canonical equivalences of that sequence.
1545 */
1546 private static void produceEquivalentAlternation(String src,
1547 Set<String> dst)
1548 {
1549 int len = countChars(src, 0, 1);
1550 if (src.length() == len) {
1551 dst.add(src); // source has one character.
1552 return;
1553 }
1554 String base = src.substring(0,len);
1555 String combiningMarks = src.substring(len);
1556 String[] perms = producePermutations(combiningMarks);
1557 // Add combined permutations
1558 for(int x = 0; x < perms.length; x++) {
1559 String next = base + perms[x];
1560 dst.add(next);
1561 next = composeOneStep(next);
1562 if (next != null) {
1563 produceEquivalentAlternation(next, dst);
1564 }
1565 }
1566 }
1567
1568 /**
1569 * Returns an array of strings that have all the possible
1570 * permutations of the characters in the input string.
1571 * This is used to get a list of all possible orderings
1572 * of a set of combining marks. Note that some of the permutations
1573 * are invalid because of combining class collisions, and these
1574 * possibilities must be removed because they are not canonically
1575 * equivalent.
1576 */
1577 private static String[] producePermutations(String input) {
1578 if (input.length() == countChars(input, 0, 1))
1579 return new String[] {input};
1580
1581 if (input.length() == countChars(input, 0, 2)) {
1582 int c0 = Character.codePointAt(input, 0);
1583 int c1 = Character.codePointAt(input, Character.charCount(c0));
1584 if (getClass(c1) == getClass(c0)) {
1585 return new String[] {input};
1586 }
1587 String[] result = new String[2];
1588 result[0] = input;
1589 StringBuilder sb = new StringBuilder(2);
1590 sb.appendCodePoint(c1);
1591 sb.appendCodePoint(c0);
1592 result[1] = sb.toString();
1593 return result;
1594 }
1595
1596 int length = 1;
1597 int nCodePoints = countCodePoints(input);
1598 for(int x=1; x<nCodePoints; x++)
1599 length = length * (x+1);
1600
1601 String[] temp = new String[length];
1602
1603 int combClass[] = new int[nCodePoints];
1604 for(int x=0, i=0; x<nCodePoints; x++) {
1605 int c = Character.codePointAt(input, i);
1606 combClass[x] = getClass(c);
1607 i += Character.charCount(c);
1608 }
1609
1610 // For each char, take it out and add the permutations
1611 // of the remaining chars
1612 int index = 0;
1613 int len;
1614 // offset maintains the index in code units.
1615 loop: for(int x=0, offset=0; x<nCodePoints; x++, offset+=len) {
1616 len = countChars(input, offset, 1);
1617 for(int y=x-1; y>=0; y--) {
1618 if (combClass[y] == combClass[x]) {
1619 continue loop;
1620 }
1621 }
1622 StringBuilder sb = new StringBuilder(input);
1623 String otherChars = sb.delete(offset, offset+len).toString();
1624 String[] subResult = producePermutations(otherChars);
1625
1626 String prefix = input.substring(offset, offset+len);
1627 for (String sre : subResult)
1628 temp[index++] = prefix + sre;
1629 }
1630 String[] result = new String[index];
1631 System.arraycopy(temp, 0, result, 0, index);
1632 return result;
1633 }
1634
1635 private static int getClass(int c) {
1636 return sun.text.Normalizer.getCombiningClass(c);
1637 }
1638
1639 /**
1640 * Attempts to compose input by combining the first character
1641 * with the first combining mark following it. Returns a String
1642 * that is the composition of the leading character with its first
1643 * combining mark followed by the remaining combining marks. Returns
1644 * null if the first two characters cannot be further composed.
1645 */
1646 private static String composeOneStep(String input) {
1647 int len = countChars(input, 0, 2);
1648 String firstTwoCharacters = input.substring(0, len);
1649 String result = Normalizer.normalize(firstTwoCharacters, Normalizer.Form.NFC);
1650 if (result.equals(firstTwoCharacters))
1651 return null;
1652 else {
1653 String remainder = input.substring(len);
1654 return result + remainder;
1655 }
1656 }
1657
1658 /**
1659 * Preprocess any \Q...\E sequences in `temp', meta-quoting them.
1660 * See the description of `quotemeta' in perlfunc(1).
1661 */
1662 private void RemoveQEQuoting() {
1663 final int pLen = patternLength;
1664 int i = 0;
1665 while (i < pLen-1) {
1666 if (temp[i] != '\\')
1667 i += 1;
1668 else if (temp[i + 1] != 'Q')
1669 i += 2;
1670 else
1671 break;
1672 }
1673 if (i >= pLen - 1) // No \Q sequence found
1674 return;
1675 int j = i;
1676 i += 2;
1677 int newTempLen;
1678 try {
1679 newTempLen = Math.addExact(j + 2, Math.multiplyExact(3, pLen - i));
1680 } catch (ArithmeticException ae) {
1681 throw new OutOfMemoryError();
1682 }
1683 int[] newtemp = new int[newTempLen];
1684 System.arraycopy(temp, 0, newtemp, 0, j);
1685
1686 boolean inQuote = true;
1687 boolean beginQuote = true;
1688 while (i < pLen) {
1689 int c = temp[i++];
1690 if (!ASCII.isAscii(c) || ASCII.isAlpha(c)) {
1691 newtemp[j++] = c;
1692 } else if (ASCII.isDigit(c)) {
1693 if (beginQuote) {
1694 /*
1695 * A unicode escape \[0xu] could be before this quote,
1696 * and we don't want this numeric char to processed as
1697 * part of the escape.
1698 */
1699 newtemp[j++] = '\\';
1700 newtemp[j++] = 'x';
1701 newtemp[j++] = '3';
1702 }
1703 newtemp[j++] = c;
1704 } else if (c != '\\') {
1705 if (inQuote) newtemp[j++] = '\\';
1706 newtemp[j++] = c;
1707 } else if (inQuote) {
1708 if (temp[i] == 'E') {
1709 i++;
1710 inQuote = false;
1711 } else {
1712 newtemp[j++] = '\\';
1713 newtemp[j++] = '\\';
1714 }
1715 } else {
1716 if (temp[i] == 'Q') {
1717 i++;
1718 inQuote = true;
1719 beginQuote = true;
1720 continue;
1721 } else {
1722 newtemp[j++] = c;
1723 if (i != pLen)
1724 newtemp[j++] = temp[i++];
1725 }
1726 }
1727
1728 beginQuote = false;
1729 }
1730
1731 patternLength = j;
1732 temp = Arrays.copyOf(newtemp, j + 2); // double zero termination
1733 }
1734
1735 /**
1736 * Copies regular expression to an int array and invokes the parsing
1737 * of the expression which will create the object tree.
1738 */
1739 private void compile() {
1740 // Handle canonical equivalences
1741 if (has(CANON_EQ) && !has(LITERAL)) {
1742 normalizedPattern = normalize(pattern);
1743 } else {
1744 normalizedPattern = pattern;
1745 }
1746 patternLength = normalizedPattern.length();
1747
1748 // Copy pattern to int array for convenience
1749 // Use double zero to terminate pattern
1750 temp = new int[patternLength + 2];
1751
1752 hasSupplementary = false;
1753 int c, count = 0;
1754 // Convert all chars into code points
1755 for (int x = 0; x < patternLength; x += Character.charCount(c)) {
1756 c = normalizedPattern.codePointAt(x);
1757 if (isSupplementary(c)) {
1758 hasSupplementary = true;
1759 }
1760 temp[count++] = c;
1761 }
1762
1763 patternLength = count; // patternLength now in code points
1764
1765 if (! has(LITERAL))
1766 RemoveQEQuoting();
1767
1768 // Allocate all temporary objects here.
1769 buffer = new int[32];
1770 groupNodes = new GroupHead[10];
1771 namedGroups = null;
1772 topClosureNodes = new ArrayList<>(10);
1773
1774 if (has(LITERAL)) {
1775 // Literal pattern handling
1776 matchRoot = newSlice(temp, patternLength, hasSupplementary);
1777 matchRoot.next = lastAccept;
1778 } else {
1779 // Start recursive descent parsing
1780 matchRoot = expr(lastAccept);
1781 // Check extra pattern characters
1782 if (patternLength != cursor) {
1783 if (peek() == ')') {
1784 throw error("Unmatched closing ')'");
1785 } else {
1786 throw error("Unexpected internal error");
1787 }
1788 }
1789 }
1790
1791 // Peephole optimization
1792 if (matchRoot instanceof Slice) {
1793 root = BnM.optimize(matchRoot);
1794 if (root == matchRoot) {
1795 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1796 }
1797 } else if (matchRoot instanceof Begin || matchRoot instanceof First) {
1798 root = matchRoot;
1799 } else {
1800 root = hasSupplementary ? new StartS(matchRoot) : new Start(matchRoot);
1801 }
1802
1803 // Optimize the greedy Loop to prevent exponential backtracking, IF there
1804 // is no group ref in this pattern. With a non-negative localTCNCount value,
1805 // the greedy type Loop, Curly will skip the backtracking for any starting
1806 // position "i" that failed in the past.
1807 if (!hasGroupRef) {
1808 for (Node node : topClosureNodes) {
1809 if (node instanceof Loop) {
1810 // non-deterministic-greedy-group
1811 ((Loop)node).posIndex = localTCNCount++;
1812 }
1813 }
1814 }
1815
1816 // Release temporary storage
1817 temp = null;
1818 buffer = null;
1819 groupNodes = null;
1820 patternLength = 0;
1821 compiled = true;
1822 topClosureNodes = null;
1823 }
1824
1825 Map<String, Integer> namedGroups() {
1826 Map<String, Integer> groups = namedGroups;
1827 if (groups == null) {
1828 namedGroups = groups = new HashMap<>(2);
1829 }
1830 return groups;
1831 }
1832
1833 /**
1834 * Used to accumulate information about a subtree of the object graph
1835 * so that optimizations can be applied to the subtree.
1836 */
1837 static final class TreeInfo {
1838 int minLength;
1839 int maxLength;
1840 boolean maxValid;
1841 boolean deterministic;
1842
1843 TreeInfo() {
1844 reset();
1845 }
1846 void reset() {
1847 minLength = 0;
1848 maxLength = 0;
1849 maxValid = true;
1850 deterministic = true;
1851 }
1852 }
1853
1854 /*
1855 * The following private methods are mainly used to improve the
1856 * readability of the code. In order to let the Java compiler easily
1857 * inline them, we should not put many assertions or error checks in them.
1858 */
1859
1860 /**
1861 * Indicates whether a particular flag is set or not.
1862 */
1863 private boolean has(int f) {
1864 return (flags0 & f) != 0;
1865 }
1866
1867 /**
1868 * Match next character, signal error if failed.
1869 */
1870 private void accept(int ch, String s) {
1871 int testChar = temp[cursor++];
1872 if (has(COMMENTS))
1873 testChar = parsePastWhitespace(testChar);
1874 if (ch != testChar) {
1875 throw error(s);
1876 }
1877 }
1878
1879 /**
1880 * Mark the end of pattern with a specific character.
1881 */
1882 private void mark(int c) {
1883 temp[patternLength] = c;
1884 }
1885
1886 /**
1887 * Peek the next character, and do not advance the cursor.
1888 */
1889 private int peek() {
1890 int ch = temp[cursor];
1891 if (has(COMMENTS))
1892 ch = peekPastWhitespace(ch);
1893 return ch;
1894 }
1895
1896 /**
1897 * Read the next character, and advance the cursor by one.
1898 */
1899 private int read() {
1900 int ch = temp[cursor++];
1901 if (has(COMMENTS))
1902 ch = parsePastWhitespace(ch);
1903 return ch;
1904 }
1905
1906 /**
1907 * Read the next character, and advance the cursor by one,
1908 * ignoring the COMMENTS setting
1909 */
1910 private int readEscaped() {
1911 int ch = temp[cursor++];
1912 return ch;
1913 }
1914
1915 /**
1916 * Advance the cursor by one, and peek the next character.
1917 */
1918 private int next() {
1919 int ch = temp[++cursor];
1920 if (has(COMMENTS))
1921 ch = peekPastWhitespace(ch);
1922 return ch;
1923 }
1924
1925 /**
1926 * Advance the cursor by one, and peek the next character,
1927 * ignoring the COMMENTS setting
1928 */
1929 private int nextEscaped() {
1930 int ch = temp[++cursor];
1931 return ch;
1932 }
1933
1934 /**
1935 * If in xmode peek past whitespace and comments.
1936 */
1937 private int peekPastWhitespace(int ch) {
1938 while (ASCII.isSpace(ch) || ch == '#') {
1939 while (ASCII.isSpace(ch))
1940 ch = temp[++cursor];
1941 if (ch == '#') {
1942 ch = peekPastLine();
1943 }
1944 }
1945 return ch;
1946 }
1947
1948 /**
1949 * If in xmode parse past whitespace and comments.
1950 */
1951 private int parsePastWhitespace(int ch) {
1952 while (ASCII.isSpace(ch) || ch == '#') {
1953 while (ASCII.isSpace(ch))
1954 ch = temp[cursor++];
1955 if (ch == '#')
1956 ch = parsePastLine();
1957 }
1958 return ch;
1959 }
1960
1961 /**
1962 * xmode parse past comment to end of line.
1963 */
1964 private int parsePastLine() {
1965 int ch = temp[cursor++];
1966 while (ch != 0 && !isLineSeparator(ch))
1967 ch = temp[cursor++];
1968 return ch;
1969 }
1970
1971 /**
1972 * xmode peek past comment to end of line.
1973 */
1974 private int peekPastLine() {
1975 int ch = temp[++cursor];
1976 while (ch != 0 && !isLineSeparator(ch))
1977 ch = temp[++cursor];
1978 return ch;
1979 }
1980
1981 /**
1982 * Determines if character is a line separator in the current mode
1983 */
1984 private boolean isLineSeparator(int ch) {
1985 if (has(UNIX_LINES)) {
1986 return ch == '\n';
1987 } else {
1988 return (ch == '\n' ||
1989 ch == '\r' ||
1990 (ch|1) == '\u2029' ||
1991 ch == '\u0085');
1992 }
1993 }
1994
1995 /**
1996 * Read the character after the next one, and advance the cursor by two.
1997 */
1998 private int skip() {
1999 int i = cursor;
2000 int ch = temp[i+1];
2001 cursor = i + 2;
2002 return ch;
2003 }
2004
2005 /**
2006 * Unread one next character, and retreat cursor by one.
2007 */
2008 private void unread() {
2009 cursor--;
2010 }
2011
2012 /**
2013 * Internal method used for handling all syntax errors. The pattern is
2014 * displayed with a pointer to aid in locating the syntax error.
2015 */
2016 private PatternSyntaxException error(String s) {
2017 return new PatternSyntaxException(s, normalizedPattern, cursor - 1);
2018 }
2019
2020 /**
2021 * Determines if there is any supplementary character or unpaired
2022 * surrogate in the specified range.
2023 */
2024 private boolean findSupplementary(int start, int end) {
2025 for (int i = start; i < end; i++) {
2026 if (isSupplementary(temp[i]))
2027 return true;
2028 }
2029 return false;
2030 }
2031
2032 /**
2033 * Determines if the specified code point is a supplementary
2034 * character or unpaired surrogate.
2035 */
2036 private static final boolean isSupplementary(int ch) {
2037 return ch >= Character.MIN_SUPPLEMENTARY_CODE_POINT ||
2038 Character.isSurrogate((char)ch);
2039 }
2040
2041 /**
2042 * The following methods handle the main parsing. They are sorted
2043 * according to their precedence order, the lowest one first.
2044 */
2045
2046 /**
2047 * The expression is parsed with branch nodes added for alternations.
2048 * This may be called recursively to parse sub expressions that may
2049 * contain alternations.
2050 */
2051 private Node expr(Node end) {
2052 Node prev = null;
2053 Node firstTail = null;
2054 Branch branch = null;
2055 Node branchConn = null;
2056
2057 for (;;) {
2058 Node node = sequence(end);
2059 Node nodeTail = root; //double return
2060 if (prev == null) {
2061 prev = node;
2062 firstTail = nodeTail;
2063 } else {
2064 // Branch
2065 if (branchConn == null) {
2066 branchConn = new BranchConn();
2067 branchConn.next = end;
2068 }
2069 if (node == end) {
2070 // if the node returned from sequence() is "end"
2071 // we have an empty expr, set a null atom into
2072 // the branch to indicate to go "next" directly.
2073 node = null;
2074 } else {
2075 // the "tail.next" of each atom goes to branchConn
2076 nodeTail.next = branchConn;
2077 }
2078 if (prev == branch) {
2079 branch.add(node);
2080 } else {
2081 if (prev == end) {
2082 prev = null;
2083 } else {
2084 // replace the "end" with "branchConn" at its tail.next
2085 // when put the "prev" into the branch as the first atom.
2086 firstTail.next = branchConn;
2087 }
2088 prev = branch = new Branch(prev, node, branchConn);
2089 }
2090 }
2091 if (peek() != '|') {
2092 return prev;
2093 }
2094 next();
2095 }
2096 }
2097
2098 @SuppressWarnings("fallthrough")
2099 /**
2100 * Parsing of sequences between alternations.
2101 */
2102 private Node sequence(Node end) {
2103 Node head = null;
2104 Node tail = null;
2105 Node node;
2106 LOOP:
2107 for (;;) {
2108 int ch = peek();
2109 switch (ch) {
2110 case '(':
2111 // Because group handles its own closure,
2112 // we need to treat it differently
2113 node = group0();
2114 // Check for comment or flag group
2115 if (node == null)
2116 continue;
2117 if (head == null)
2118 head = node;
2119 else
2120 tail.next = node;
2121 // Double return: Tail was returned in root
2122 tail = root;
2123 continue;
2124 case '[':
2125 if (has(CANON_EQ) && !has(LITERAL))
2126 node = new NFCCharProperty(clazz(true));
2127 else
2128 node = newCharProperty(clazz(true));
2129 break;
2130 case '\\':
2131 ch = nextEscaped();
2132 if (ch == 'p' || ch == 'P') {
2133 boolean oneLetter = true;
2134 boolean comp = (ch == 'P');
2135 ch = next(); // Consume { if present
2136 if (ch != '{') {
2137 unread();
2138 } else {
2139 oneLetter = false;
2140 }
2141 // node = newCharProperty(family(oneLetter, comp));
2142 if (has(CANON_EQ) && !has(LITERAL))
2143 node = new NFCCharProperty(family(oneLetter, comp));
2144 else
2145 node = newCharProperty(family(oneLetter, comp));
2146 } else {
2147 unread();
2148 node = atom();
2149 }
2150 break;
2151 case '^':
2152 next();
2153 if (has(MULTILINE)) {
2154 if (has(UNIX_LINES))
2155 node = new UnixCaret();
2156 else
2157 node = new Caret();
2158 } else {
2159 node = new Begin();
2160 }
2161 break;
2162 case '$':
2163 next();
2164 if (has(UNIX_LINES))
2165 node = new UnixDollar(has(MULTILINE));
2166 else
2167 node = new Dollar(has(MULTILINE));
2168 break;
2169 case '.':
2170 next();
2171 if (has(DOTALL)) {
2172 node = new CharProperty(ALL());
2173 } else {
2174 if (has(UNIX_LINES)) {
2175 node = new CharProperty(UNIXDOT());
2176 } else {
2177 node = new CharProperty(DOT());
2178 }
2179 }
2180 break;
2181 case '|':
2182 case ')':
2183 break LOOP;
2184 case ']': // Now interpreting dangling ] and } as literals
2185 case '}':
2186 node = atom();
2187 break;
2188 case '?':
2189 case '*':
2190 case '+':
2191 next();
2192 throw error("Dangling meta character '" + ((char)ch) + "'");
2193 case 0:
2194 if (cursor >= patternLength) {
2195 break LOOP;
2196 }
2197 // Fall through
2198 default:
2199 node = atom();
2200 break;
2201 }
2202
2203 node = closure(node);
2204 /* save the top dot-greedy nodes (.*, .+) as well
2205 if (node instanceof GreedyCharProperty &&
2206 ((GreedyCharProperty)node).cp instanceof Dot) {
2207 topClosureNodes.add(node);
2208 }
2209 */
2210 if (head == null) {
2211 head = tail = node;
2212 } else {
2213 tail.next = node;
2214 tail = node;
2215 }
2216 }
2217 if (head == null) {
2218 return end;
2219 }
2220 tail.next = end;
2221 root = tail; //double return
2222 return head;
2223 }
2224
2225 @SuppressWarnings("fallthrough")
2226 /**
2227 * Parse and add a new Single or Slice.
2228 */
2229 private Node atom() {
2230 int first = 0;
2231 int prev = -1;
2232 boolean hasSupplementary = false;
2233 int ch = peek();
2234 for (;;) {
2235 switch (ch) {
2236 case '*':
2237 case '+':
2238 case '?':
2239 case '{':
2240 if (first > 1) {
2241 cursor = prev; // Unwind one character
2242 first--;
2243 }
2244 break;
2245 case '$':
2246 case '.':
2247 case '^':
2248 case '(':
2249 case '[':
2250 case '|':
2251 case ')':
2252 break;
2253 case '\\':
2254 ch = nextEscaped();
2255 if (ch == 'p' || ch == 'P') { // Property
2256 if (first > 0) { // Slice is waiting; handle it first
2257 unread();
2258 break;
2259 } else { // No slice; just return the family node
2260 boolean comp = (ch == 'P');
2261 boolean oneLetter = true;
2262 ch = next(); // Consume { if present
2263 if (ch != '{')
2264 unread();
2265 else
2266 oneLetter = false;
2267 if (has(CANON_EQ) && !has(LITERAL))
2268 return new NFCCharProperty(family(oneLetter, comp));
2269 else
2270 return newCharProperty(family(oneLetter, comp));
2271 }
2272 }
2273 unread();
2274 prev = cursor;
2275 ch = escape(false, first == 0, false);
2276 if (ch >= 0) {
2277 append(ch, first);
2278 first++;
2279 if (isSupplementary(ch)) {
2280 hasSupplementary = true;
2281 }
2282 ch = peek();
2283 continue;
2284 } else if (first == 0) {
2285 return root;
2286 }
2287 // Unwind meta escape sequence
2288 cursor = prev;
2289 break;
2290 case 0:
2291 if (cursor >= patternLength) {
2292 break;
2293 }
2294 // Fall through
2295 default:
2296 prev = cursor;
2297 append(ch, first);
2298 first++;
2299 if (isSupplementary(ch)) {
2300 hasSupplementary = true;
2301 }
2302 ch = next();
2303 continue;
2304 }
2305 break;
2306 }
2307 if (first == 1) {
2308 return newCharProperty(single(buffer[0]));
2309 } else {
2310 return newSlice(buffer, first, hasSupplementary);
2311 }
2312 }
2313
2314 private void append(int ch, int index) {
2315 int len = buffer.length;
2316 if (index - len >= 0) {
2317 len = ArraysSupport.newLength(len,
2318 1 + index - len, /* minimum growth */
2319 len /* preferred growth */);
2320 buffer = Arrays.copyOf(buffer, len);
2321 }
2322 buffer[index] = ch;
2323 }
2324
2325 /**
2326 * Parses a backref greedily, taking as many numbers as it
2327 * can. The first digit is always treated as a backref, but
2328 * multi digit numbers are only treated as a backref if at
2329 * least that many backrefs exist at this point in the regex.
2330 */
2331 private Node ref(int refNum) {
2332 boolean done = false;
2333 while(!done) {
2334 int ch = peek();
2335 switch(ch) {
2336 case '0':
2337 case '1':
2338 case '2':
2339 case '3':
2340 case '4':
2341 case '5':
2342 case '6':
2343 case '7':
2344 case '8':
2345 case '9':
2346 int newRefNum = (refNum * 10) + (ch - '0');
2347 // Add another number if it doesn't make a group
2348 // that doesn't exist
2349 if (capturingGroupCount - 1 < newRefNum) {
2350 done = true;
2351 break;
2352 }
2353 refNum = newRefNum;
2354 read();
2355 break;
2356 default:
2357 done = true;
2358 break;
2359 }
2360 }
2361 hasGroupRef = true;
2362 if (has(CASE_INSENSITIVE))
2363 return new CIBackRef(refNum, has(UNICODE_CASE));
2364 else
2365 return new BackRef(refNum);
2366 }
2367
2368 /**
2369 * Parses an escape sequence to determine the actual value that needs
2370 * to be matched.
2371 * If -1 is returned and create was true a new object was added to the tree
2372 * to handle the escape sequence.
2373 * If the returned value is greater than zero, it is the value that
2374 * matches the escape sequence.
2375 */
2376 private int escape(boolean inclass, boolean create, boolean isrange) {
2377 int ch = skip();
2378 switch (ch) {
2379 case '0':
2380 return o();
2381 case '1':
2382 case '2':
2383 case '3':
2384 case '4':
2385 case '5':
2386 case '6':
2387 case '7':
2388 case '8':
2389 case '9':
2390 if (inclass) break;
2391 if (create) {
2392 root = ref((ch - '0'));
2393 }
2394 return -1;
2395 case 'A':
2396 if (inclass) break;
2397 if (create) root = new Begin();
2398 return -1;
2399 case 'B':
2400 if (inclass) break;
2401 if (create) root = new Bound(Bound.NONE, has(UNICODE_CHARACTER_CLASS));
2402 return -1;
2403 case 'C':
2404 break;
2405 case 'D':
2406 if (create) {
2407 predicate = has(UNICODE_CHARACTER_CLASS) ?
2408 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT();
2409 predicate = predicate.negate();
2410 if (!inclass)
2411 root = newCharProperty(predicate);
2412 }
2413 return -1;
2414 case 'E':
2415 case 'F':
2416 break;
2417 case 'G':
2418 if (inclass) break;
2419 if (create) root = new LastMatch();
2420 return -1;
2421 case 'H':
2422 if (create) {
2423 predicate = HorizWS().negate();
2424 if (!inclass)
2425 root = newCharProperty(predicate);
2426 }
2427 return -1;
2428 case 'I':
2429 case 'J':
2430 case 'K':
2431 case 'L':
2432 case 'M':
2433 break;
2434 case 'N':
2435 return N();
2436 case 'O':
2437 case 'P':
2438 case 'Q':
2439 break;
2440 case 'R':
2441 if (inclass) break;
2442 if (create) root = new LineEnding();
2443 return -1;
2444 case 'S':
2445 if (create) {
2446 predicate = has(UNICODE_CHARACTER_CLASS) ?
2447 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE();
2448 predicate = predicate.negate();
2449 if (!inclass)
2450 root = newCharProperty(predicate);
2451 }
2452 return -1;
2453 case 'T':
2454 case 'U':
2455 break;
2456 case 'V':
2457 if (create) {
2458 predicate = VertWS().negate();
2459 if (!inclass)
2460 root = newCharProperty(predicate);
2461 }
2462 return -1;
2463 case 'W':
2464 if (create) {
2465 predicate = has(UNICODE_CHARACTER_CLASS) ?
2466 CharPredicates.WORD() : CharPredicates.ASCII_WORD();
2467 predicate = predicate.negate();
2468 if (!inclass)
2469 root = newCharProperty(predicate);
2470 }
2471 return -1;
2472 case 'X':
2473 if (inclass) break;
2474 if (create) {
2475 root = new XGrapheme();
2476 }
2477 return -1;
2478 case 'Y':
2479 break;
2480 case 'Z':
2481 if (inclass) break;
2482 if (create) {
2483 if (has(UNIX_LINES))
2484 root = new UnixDollar(false);
2485 else
2486 root = new Dollar(false);
2487 }
2488 return -1;
2489 case 'a':
2490 return '\007';
2491 case 'b':
2492 if (inclass) break;
2493 if (create) {
2494 if (peek() == '{') {
2495 if (skip() == 'g') {
2496 if (read() == '}') {
2497 root = new GraphemeBound();
2498 return -1;
2499 }
2500 break; // error missing trailing }
2501 }
2502 unread(); unread();
2503 }
2504 root = new Bound(Bound.BOTH, has(UNICODE_CHARACTER_CLASS));
2505 }
2506 return -1;
2507 case 'c':
2508 return c();
2509 case 'd':
2510 if (create) {
2511 predicate = has(UNICODE_CHARACTER_CLASS) ?
2512 CharPredicates.DIGIT() : CharPredicates.ASCII_DIGIT();
2513 if (!inclass)
2514 root = newCharProperty(predicate);
2515 }
2516 return -1;
2517 case 'e':
2518 return '\033';
2519 case 'f':
2520 return '\f';
2521 case 'g':
2522 break;
2523 case 'h':
2524 if (create) {
2525 predicate = HorizWS();
2526 if (!inclass)
2527 root = newCharProperty(predicate);
2528 }
2529 return -1;
2530 case 'i':
2531 case 'j':
2532 break;
2533 case 'k':
2534 if (inclass)
2535 break;
2536 if (read() != '<')
2537 throw error("\\k is not followed by '<' for named capturing group");
2538 String name = groupname(read());
2539 if (!namedGroups().containsKey(name))
2540 throw error("named capturing group <" + name + "> does not exist");
2541 if (create) {
2542 hasGroupRef = true;
2543 if (has(CASE_INSENSITIVE))
2544 root = new CIBackRef(namedGroups().get(name), has(UNICODE_CASE));
2545 else
2546 root = new BackRef(namedGroups().get(name));
2547 }
2548 return -1;
2549 case 'l':
2550 case 'm':
2551 break;
2552 case 'n':
2553 return '\n';
2554 case 'o':
2555 case 'p':
2556 case 'q':
2557 break;
2558 case 'r':
2559 return '\r';
2560 case 's':
2561 if (create) {
2562 predicate = has(UNICODE_CHARACTER_CLASS) ?
2563 CharPredicates.WHITE_SPACE() : CharPredicates.ASCII_SPACE();
2564 if (!inclass)
2565 root = newCharProperty(predicate);
2566 }
2567 return -1;
2568 case 't':
2569 return '\t';
2570 case 'u':
2571 return u();
2572 case 'v':
2573 // '\v' was implemented as VT/0x0B in releases < 1.8 (though
2574 // undocumented). In JDK8 '\v' is specified as a predefined
2575 // character class for all vertical whitespace characters.
2576 // So [-1, root=VertWS node] pair is returned (instead of a
2577 // single 0x0B). This breaks the range if '\v' is used as
2578 // the start or end value, such as [\v-...] or [...-\v], in
2579 // which a single definite value (0x0B) is expected. For
2580 // compatibility concern '\013'/0x0B is returned if isrange.
2581 if (isrange)
2582 return '\013';
2583 if (create) {
2584 predicate = VertWS();
2585 if (!inclass)
2586 root = newCharProperty(predicate);
2587 }
2588 return -1;
2589 case 'w':
2590 if (create) {
2591 predicate = has(UNICODE_CHARACTER_CLASS) ?
2592 CharPredicates.WORD() : CharPredicates.ASCII_WORD();
2593 if (!inclass)
2594 root = newCharProperty(predicate);
2595 }
2596 return -1;
2597 case 'x':
2598 return x();
2599 case 'y':
2600 break;
2601 case 'z':
2602 if (inclass) break;
2603 if (create) root = new End();
2604 return -1;
2605 default:
2606 return ch;
2607 }
2608 throw error("Illegal/unsupported escape sequence");
2609 }
2610
2611 /**
2612 * Parse a character class, and return the node that matches it.
2613 *
2614 * Consumes a ] on the way out if consume is true. Usually consume
2615 * is true except for the case of [abc&&def] where def is a separate
2616 * right hand node with "understood" brackets.
2617 */
2618 private CharPredicate clazz(boolean consume) {
2619 CharPredicate prev = null;
2620 CharPredicate curr = null;
2621 BitClass bits = new BitClass();
2622
2623 boolean isNeg = false;
2624 boolean hasBits = false;
2625 int ch = next();
2626
2627 // Negates if first char in a class, otherwise literal
2628 if (ch == '^' && temp[cursor-1] == '[') {
2629 ch = next();
2630 isNeg = true;
2631 }
2632 for (;;) {
2633 switch (ch) {
2634 case '[':
2635 curr = clazz(true);
2636 if (prev == null)
2637 prev = curr;
2638 else
2639 prev = prev.union(curr);
2640 ch = peek();
2641 continue;
2642 case '&':
2643 ch = next();
2644 if (ch == '&') {
2645 ch = next();
2646 CharPredicate right = null;
2647 while (ch != ']' && ch != '&') {
2648 if (ch == '[') {
2649 if (right == null)
2650 right = clazz(true);
2651 else
2652 right = right.union(clazz(true));
2653 } else { // abc&&def
2654 unread();
2655 right = clazz(false);
2656 }
2657 ch = peek();
2658 }
2659 if (hasBits) {
2660 // bits used, union has high precedence
2661 if (prev == null) {
2662 prev = curr = bits;
2663 } else {
2664 prev = prev.union(bits);
2665 }
2666 hasBits = false;
2667 }
2668 if (right != null)
2669 curr = right;
2670 if (prev == null) {
2671 if (right == null)
2672 throw error("Bad class syntax");
2673 else
2674 prev = right;
2675 } else {
2676 prev = prev.and(curr);
2677 }
2678 } else {
2679 // treat as a literal &
2680 unread();
2681 break;
2682 }
2683 continue;
2684 case 0:
2685 if (cursor >= patternLength)
2686 throw error("Unclosed character class");
2687 break;
2688 case ']':
2689 if (prev != null || hasBits) {
2690 if (consume)
2691 next();
2692 if (prev == null)
2693 prev = bits;
2694 else if (hasBits)
2695 prev = prev.union(bits);
2696 if (isNeg)
2697 return prev.negate();
2698 return prev;
2699 }
2700 break;
2701 default:
2702 break;
2703 }
2704 curr = range(bits);
2705 if (curr == null) { // the bits used
2706 hasBits = true;
2707 } else {
2708 if (prev == null)
2709 prev = curr;
2710 else if (prev != curr)
2711 prev = prev.union(curr);
2712 }
2713 ch = peek();
2714 }
2715 }
2716
2717 private CharPredicate bitsOrSingle(BitClass bits, int ch) {
2718 /* Bits can only handle codepoints in [u+0000-u+00ff] range.
2719 Use "single" node instead of bits when dealing with unicode
2720 case folding for codepoints listed below.
2721 (1)Uppercase out of range: u+00ff, u+00b5
2722 toUpperCase(u+00ff) -> u+0178
2723 toUpperCase(u+00b5) -> u+039c
2724 (2)LatinSmallLetterLongS u+17f
2725 toUpperCase(u+017f) -> u+0053
2726 (3)LatinSmallLetterDotlessI u+131
2727 toUpperCase(u+0131) -> u+0049
2728 (4)LatinCapitalLetterIWithDotAbove u+0130
2729 toLowerCase(u+0130) -> u+0069
2730 (5)KelvinSign u+212a
2731 toLowerCase(u+212a) ==> u+006B
2732 (6)AngstromSign u+212b
2733 toLowerCase(u+212b) ==> u+00e5
2734 */
2735 if (ch < 256 &&
2736 !(has(CASE_INSENSITIVE) && has(UNICODE_CASE) &&
2737 (ch == 0xff || ch == 0xb5 ||
2738 ch == 0x49 || ch == 0x69 || //I and i
2739 ch == 0x53 || ch == 0x73 || //S and s
2740 ch == 0x4b || ch == 0x6b || //K and k
2741 ch == 0xc5 || ch == 0xe5))) { //A+ring
2742 bits.add(ch, flags0);
2743 return null;
2744 }
2745 return single(ch);
2746 }
2747
2748 /**
2749 * Returns a suitably optimized, single character predicate
2750 */
2751 private CharPredicate single(final int ch) {
2752 if (has(CASE_INSENSITIVE)) {
2753 int lower, upper;
2754 if (has(UNICODE_CASE)) {
2755 upper = Character.toUpperCase(ch);
2756 lower = Character.toLowerCase(upper);
2757 // Unicode case insensitive matches
2758 if (upper != lower)
2759 return SingleU(lower);
2760 } else if (ASCII.isAscii(ch)) {
2761 lower = ASCII.toLower(ch);
2762 upper = ASCII.toUpper(ch);
2763 // Case insensitive matches a given BMP character
2764 if (lower != upper)
2765 return SingleI(lower, upper);
2766 }
2767 }
2768 if (isSupplementary(ch))
2769 return SingleS(ch);
2770 return Single(ch); // Match a given BMP character
2771 }
2772
2773 /**
2774 * Parse a single character or a character range in a character class
2775 * and return its representative node.
2776 */
2777 private CharPredicate range(BitClass bits) {
2778 int ch = peek();
2779 if (ch == '\\') {
2780 ch = nextEscaped();
2781 if (ch == 'p' || ch == 'P') { // A property
2782 boolean comp = (ch == 'P');
2783 boolean oneLetter = true;
2784 // Consume { if present
2785 ch = next();
2786 if (ch != '{')
2787 unread();
2788 else
2789 oneLetter = false;
2790 return family(oneLetter, comp);
2791 } else { // ordinary escape
2792 boolean isrange = temp[cursor+1] == '-';
2793 unread();
2794 ch = escape(true, true, isrange);
2795 if (ch == -1)
2796 return predicate;
2797 }
2798 } else {
2799 next();
2800 }
2801 if (ch >= 0) {
2802 if (peek() == '-') {
2803 int endRange = temp[cursor+1];
2804 if (endRange == '[') {
2805 return bitsOrSingle(bits, ch);
2806 }
2807 if (endRange != ']') {
2808 next();
2809 int m = peek();
2810 if (m == '\\') {
2811 m = escape(true, false, true);
2812 } else {
2813 next();
2814 }
2815 if (m < ch) {
2816 throw error("Illegal character range");
2817 }
2818 if (has(CASE_INSENSITIVE)) {
2819 if (has(UNICODE_CASE))
2820 return CIRangeU(ch, m);
2821 return CIRange(ch, m);
2822 } else {
2823 return Range(ch, m);
2824 }
2825 }
2826 }
2827 return bitsOrSingle(bits, ch);
2828 }
2829 throw error("Unexpected character '"+((char)ch)+"'");
2830 }
2831
2832 /**
2833 * Parses a Unicode character family and returns its representative node.
2834 */
2835 private CharPredicate family(boolean singleLetter, boolean isComplement) {
2836 next();
2837 String name;
2838 CharPredicate p = null;
2839
2840 if (singleLetter) {
2841 int c = temp[cursor];
2842 if (!Character.isSupplementaryCodePoint(c)) {
2843 name = String.valueOf((char)c);
2844 } else {
2845 name = new String(temp, cursor, 1);
2846 }
2847 read();
2848 } else {
2849 int i = cursor;
2850 mark('}');
2851 while(read() != '}') {
2852 }
2853 mark('\000');
2854 int j = cursor;
2855 if (j > patternLength)
2856 throw error("Unclosed character family");
2857 if (i + 1 >= j)
2858 throw error("Empty character family");
2859 name = new String(temp, i, j-i-1);
2860 }
2861
2862 int i = name.indexOf('=');
2863 if (i != -1) {
2864 // property construct \p{name=value}
2865 String value = name.substring(i + 1);
2866 name = name.substring(0, i).toLowerCase(Locale.ENGLISH);
2867 switch (name) {
2868 case "sc":
2869 case "script":
2870 p = CharPredicates.forUnicodeScript(value);
2871 break;
2872 case "blk":
2873 case "block":
2874 p = CharPredicates.forUnicodeBlock(value);
2875 break;
2876 case "gc":
2877 case "general_category":
2878 p = CharPredicates.forProperty(value);
2879 break;
2880 default:
2881 break;
2882 }
2883 if (p == null)
2884 throw error("Unknown Unicode property {name=<" + name + ">, "
2885 + "value=<" + value + ">}");
2886
2887 } else {
2888 if (name.startsWith("In")) {
2889 // \p{InBlockName}
2890 p = CharPredicates.forUnicodeBlock(name.substring(2));
2891 } else if (name.startsWith("Is")) {
2892 // \p{IsGeneralCategory} and \p{IsScriptName}
2893 name = name.substring(2);
2894 p = CharPredicates.forUnicodeProperty(name);
2895 if (p == null)
2896 p = CharPredicates.forProperty(name);
2897 if (p == null)
2898 p = CharPredicates.forUnicodeScript(name);
2899 } else {
2900 if (has(UNICODE_CHARACTER_CLASS)) {
2901 p = CharPredicates.forPOSIXName(name);
2902 }
2903 if (p == null)
2904 p = CharPredicates.forProperty(name);
2905 }
2906 if (p == null)
2907 throw error("Unknown character property name {In/Is" + name + "}");
2908 }
2909 if (isComplement) {
2910 // it might be too expensive to detect if a complement of
2911 // CharProperty can match "certain" supplementary. So just
2912 // go with StartS.
2913 hasSupplementary = true;
2914 p = p.negate();
2915 }
2916 return p;
2917 }
2918
2919 private CharProperty newCharProperty(CharPredicate p) {
2920 if (p == null)
2921 return null;
2922 if (p instanceof BmpCharPredicate)
2923 return new BmpCharProperty((BmpCharPredicate)p);
2924 else
2925 return new CharProperty(p);
2926 }
2927
2928 /**
2929 * Parses and returns the name of a "named capturing group", the trailing
2930 * ">" is consumed after parsing.
2931 */
2932 private String groupname(int ch) {
2933 StringBuilder sb = new StringBuilder();
2934 if (!ASCII.isAlpha(ch))
2935 throw error("capturing group name does not start with a Latin letter");
2936 do {
2937 sb.append((char) ch);
2938 } while (ASCII.isAlnum(ch=read()));
2939 if (ch != '>')
2940 throw error("named capturing group is missing trailing '>'");
2941 return sb.toString();
2942 }
2943
2944 /**
2945 * Parses a group and returns the head node of a set of nodes that process
2946 * the group. Sometimes a double return system is used where the tail is
2947 * returned in root.
2948 */
2949 private Node group0() {
2950 boolean capturingGroup = false;
2951 Node head;
2952 Node tail;
2953 int save = flags0;
2954 int saveTCNCount = topClosureNodes.size();
2955 root = null;
2956 int ch = next();
2957 if (ch == '?') {
2958 ch = skip();
2959 switch (ch) {
2960 case ':': // (?:xxx) pure group
2961 head = createGroup(true);
2962 tail = root;
2963 head.next = expr(tail);
2964 break;
2965 case '=': // (?=xxx) and (?!xxx) lookahead
2966 case '!':
2967 head = createGroup(true);
2968 tail = root;
2969 head.next = expr(tail);
2970 if (ch == '=') {
2971 head = tail = new Pos(head);
2972 } else {
2973 head = tail = new Neg(head);
2974 }
2975 break;
2976 case '>': // (?>xxx) independent group
2977 head = createGroup(true);
2978 tail = root;
2979 head.next = expr(tail);
2980 head = tail = new Ques(head, Qtype.INDEPENDENT);
2981 break;
2982 case '<': // (?<xxx) look behind
2983 ch = read();
2984 if (ch != '=' && ch != '!') {
2985 // named captured group
2986 String name = groupname(ch);
2987 if (namedGroups().containsKey(name))
2988 throw error("Named capturing group <" + name
2989 + "> is already defined");
2990 capturingGroup = true;
2991 head = createGroup(false);
2992 tail = root;
2993 namedGroups().put(name, capturingGroupCount-1);
2994 head.next = expr(tail);
2995 break;
2996 }
2997 int start = cursor;
2998 head = createGroup(true);
2999 tail = root;
3000 head.next = expr(tail);
3001 tail.next = LookBehindEndNode.INSTANCE;
3002 TreeInfo info = new TreeInfo();
3003 head.study(info);
3004 if (info.maxValid == false) {
3005 throw error("Look-behind group does not have "
3006 + "an obvious maximum length");
3007 }
3008 boolean hasSupplementary = findSupplementary(start, patternLength);
3009 if (ch == '=') {
3010 head = tail = (hasSupplementary ?
3011 new BehindS(head, info.maxLength,
3012 info.minLength) :
3013 new Behind(head, info.maxLength,
3014 info.minLength));
3015 } else { // if (ch == '!')
3016 head = tail = (hasSupplementary ?
3017 new NotBehindS(head, info.maxLength,
3018 info.minLength) :
3019 new NotBehind(head, info.maxLength,
3020 info.minLength));
3021 }
3022 // clear all top-closure-nodes inside lookbehind
3023 if (saveTCNCount < topClosureNodes.size())
3024 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear();
3025 break;
3026 case '$':
3027 case '@':
3028 throw error("Unknown group type");
3029 default: // (?xxx:) inlined match flags
3030 unread();
3031 addFlag();
3032 ch = read();
3033 if (ch == ')') {
3034 return null; // Inline modifier only
3035 }
3036 if (ch != ':') {
3037 throw error("Unknown inline modifier");
3038 }
3039 head = createGroup(true);
3040 tail = root;
3041 head.next = expr(tail);
3042 break;
3043 }
3044 } else { // (xxx) a regular group
3045 capturingGroup = true;
3046 head = createGroup(false);
3047 tail = root;
3048 head.next = expr(tail);
3049 }
3050
3051 accept(')', "Unclosed group");
3052 flags0 = save;
3053
3054 // Check for quantifiers
3055 Node node = closure(head);
3056 if (node == head) { // No closure
3057 root = tail;
3058 return node; // Dual return
3059 }
3060 if (head == tail) { // Zero length assertion
3061 root = node;
3062 return node; // Dual return
3063 }
3064
3065 // have group closure, clear all inner closure nodes from the
3066 // top list (no backtracking stopper optimization for inner
3067 if (saveTCNCount < topClosureNodes.size())
3068 topClosureNodes.subList(saveTCNCount, topClosureNodes.size()).clear();
3069
3070 if (node instanceof Ques) {
3071 Ques ques = (Ques) node;
3072 if (ques.type == Qtype.POSSESSIVE) {
3073 root = node;
3074 return node;
3075 }
3076 tail.next = new BranchConn();
3077 tail = tail.next;
3078 if (ques.type == Qtype.GREEDY) {
3079 head = new Branch(head, null, tail);
3080 } else { // Reluctant quantifier
3081 head = new Branch(null, head, tail);
3082 }
3083 root = tail;
3084 return head;
3085 } else if (node instanceof Curly) {
3086 Curly curly = (Curly) node;
3087 if (curly.type == Qtype.POSSESSIVE) {
3088 root = node;
3089 return node;
3090 }
3091 // Discover if the group is deterministic
3092 TreeInfo info = new TreeInfo();
3093 if (head.study(info)) { // Deterministic
3094 GroupTail temp = (GroupTail) tail;
3095 head = root = new GroupCurly(head.next, curly.cmin,
3096 curly.cmax, curly.type,
3097 ((GroupTail)tail).localIndex,
3098 ((GroupTail)tail).groupIndex,
3099 capturingGroup);
3100 return head;
3101 } else { // Non-deterministic
3102 int temp = ((GroupHead) head).localIndex;
3103 Loop loop;
3104 if (curly.type == Qtype.GREEDY) {
3105 loop = new Loop(this.localCount, temp);
3106 // add the max_reps greedy to the top-closure-node list
3107 if (curly.cmax == MAX_REPS)
3108 topClosureNodes.add(loop);
3109 } else { // Reluctant Curly
3110 loop = new LazyLoop(this.localCount, temp);
3111 }
3112 Prolog prolog = new Prolog(loop);
3113 this.localCount += 1;
3114 loop.cmin = curly.cmin;
3115 loop.cmax = curly.cmax;
3116 loop.body = head;
3117 tail.next = loop;
3118 root = loop;
3119 return prolog; // Dual return
3120 }
3121 }
3122 throw error("Internal logic error");
3123 }
3124
3125 /**
3126 * Create group head and tail nodes using double return. If the group is
3127 * created with anonymous true then it is a pure group and should not
3128 * affect group counting.
3129 */
3130 private Node createGroup(boolean anonymous) {
3131 int localIndex = localCount++;
3132 int groupIndex = 0;
3133 if (!anonymous)
3134 groupIndex = capturingGroupCount++;
3135 GroupHead head = new GroupHead(localIndex);
3136 root = new GroupTail(localIndex, groupIndex);
3137
3138 // for debug/print only, head.match does NOT need the "tail" info
3139 head.tail = (GroupTail)root;
3140
3141 if (!anonymous && groupIndex < 10)
3142 groupNodes[groupIndex] = head;
3143 return head;
3144 }
3145
3146 @SuppressWarnings("fallthrough")
3147 /**
3148 * Parses inlined match flags and set them appropriately.
3149 */
3150 private void addFlag() {
3151 int ch = peek();
3152 for (;;) {
3153 switch (ch) {
3154 case 'i':
3155 flags0 |= CASE_INSENSITIVE;
3156 break;
3157 case 'm':
3158 flags0 |= MULTILINE;
3159 break;
3160 case 's':
3161 flags0 |= DOTALL;
3162 break;
3163 case 'd':
3164 flags0 |= UNIX_LINES;
3165 break;
3166 case 'u':
3167 flags0 |= UNICODE_CASE;
3168 break;
3169 case 'c':
3170 flags0 |= CANON_EQ;
3171 break;
3172 case 'x':
3173 flags0 |= COMMENTS;
3174 break;
3175 case 'U':
3176 flags0 |= (UNICODE_CHARACTER_CLASS | UNICODE_CASE);
3177 break;
3178 case '-': // subFlag then fall through
3179 ch = next();
3180 subFlag();
3181 default:
3182 return;
3183 }
3184 ch = next();
3185 }
3186 }
3187
3188 @SuppressWarnings("fallthrough")
3189 /**
3190 * Parses the second part of inlined match flags and turns off
3191 * flags appropriately.
3192 */
3193 private void subFlag() {
3194 int ch = peek();
3195 for (;;) {
3196 switch (ch) {
3197 case 'i':
3198 flags0 &= ~CASE_INSENSITIVE;
3199 break;
3200 case 'm':
3201 flags0 &= ~MULTILINE;
3202 break;
3203 case 's':
3204 flags0 &= ~DOTALL;
3205 break;
3206 case 'd':
3207 flags0 &= ~UNIX_LINES;
3208 break;
3209 case 'u':
3210 flags0 &= ~UNICODE_CASE;
3211 break;
3212 case 'c':
3213 flags0 &= ~CANON_EQ;
3214 break;
3215 case 'x':
3216 flags0 &= ~COMMENTS;
3217 break;
3218 case 'U':
3219 flags0 &= ~(UNICODE_CHARACTER_CLASS | UNICODE_CASE);
3220 break;
3221 default:
3222 return;
3223 }
3224 ch = next();
3225 }
3226 }
3227
3228 static final int MAX_REPS = 0x7FFFFFFF;
3229
3230 static enum Qtype {
3231 GREEDY, LAZY, POSSESSIVE, INDEPENDENT
3232 }
3233
3234 private Node curly(Node prev, int cmin) {
3235 int ch = next();
3236 if (ch == '?') {
3237 next();
3238 return new Curly(prev, cmin, MAX_REPS, Qtype.LAZY);
3239 } else if (ch == '+') {
3240 next();
3241 return new Curly(prev, cmin, MAX_REPS, Qtype.POSSESSIVE);
3242 }
3243 if (prev instanceof BmpCharProperty) {
3244 return new BmpCharPropertyGreedy((BmpCharProperty)prev, cmin);
3245 } else if (prev instanceof CharProperty) {
3246 return new CharPropertyGreedy((CharProperty)prev, cmin);
3247 }
3248 return new Curly(prev, cmin, MAX_REPS, Qtype.GREEDY);
3249 }
3250
3251 /**
3252 * Processes repetition. If the next character peeked is a quantifier
3253 * then new nodes must be appended to handle the repetition.
3254 * Prev could be a single or a group, so it could be a chain of nodes.
3255 */
3256 private Node closure(Node prev) {
3257 int ch = peek();
3258 switch (ch) {
3259 case '?':
3260 ch = next();
3261 if (ch == '?') {
3262 next();
3263 return new Ques(prev, Qtype.LAZY);
3264 } else if (ch == '+') {
3265 next();
3266 return new Ques(prev, Qtype.POSSESSIVE);
3267 }
3268 return new Ques(prev, Qtype.GREEDY);
3269 case '*':
3270 return curly(prev, 0);
3271 case '+':
3272 return curly(prev, 1);
3273 case '{':
3274 ch = skip();
3275 if (ASCII.isDigit(ch)) {
3276 int cmin = 0, cmax;
3277 try {
3278 do {
3279 cmin = Math.addExact(Math.multiplyExact(cmin, 10),
3280 ch - '0');
3281 } while (ASCII.isDigit(ch = read()));
3282 if (ch == ',') {
3283 ch = read();
3284 if (ch == '}') {
3285 unread();
3286 return curly(prev, cmin);
3287 } else {
3288 cmax = 0;
3289 while (ASCII.isDigit(ch)) {
3290 cmax = Math.addExact(Math.multiplyExact(cmax, 10),
3291 ch - '0');
3292 ch = read();
3293 }
3294 }
3295 } else {
3296 cmax = cmin;
3297 }
3298 } catch (ArithmeticException ae) {
3299 throw error("Illegal repetition range");
3300 }
3301 if (ch != '}')
3302 throw error("Unclosed counted closure");
3303 if (cmax < cmin)
3304 throw error("Illegal repetition range");
3305 ch = peek();
3306 if (ch == '?') {
3307 next();
3308 return new Curly(prev, cmin, cmax, Qtype.LAZY);
3309 } else if (ch == '+') {
3310 next();
3311 return new Curly(prev, cmin, cmax, Qtype.POSSESSIVE);
3312 } else {
3313 return new Curly(prev, cmin, cmax, Qtype.GREEDY);
3314 }
3315 } else {
3316 throw error("Illegal repetition");
3317 }
3318 default:
3319 return prev;
3320 }
3321 }
3322
3323 /**
3324 * Utility method for parsing control escape sequences.
3325 */
3326 private int c() {
3327 if (cursor < patternLength) {
3328 return read() ^ 64;
3329 }
3330 throw error("Illegal control escape sequence");
3331 }
3332
3333 /**
3334 * Utility method for parsing octal escape sequences.
3335 */
3336 private int o() {
3337 int n = read();
3338 if (((n-'0')|('7'-n)) >= 0) {
3339 int m = read();
3340 if (((m-'0')|('7'-m)) >= 0) {
3341 int o = read();
3342 if ((((o-'0')|('7'-o)) >= 0) && (((n-'0')|('3'-n)) >= 0)) {
3343 return (n - '0') * 64 + (m - '0') * 8 + (o - '0');
3344 }
3345 unread();
3346 return (n - '0') * 8 + (m - '0');
3347 }
3348 unread();
3349 return (n - '0');
3350 }
3351 throw error("Illegal octal escape sequence");
3352 }
3353
3354 /**
3355 * Utility method for parsing hexadecimal escape sequences.
3356 */
3357 private int x() {
3358 int n = read();
3359 if (ASCII.isHexDigit(n)) {
3360 int m = read();
3361 if (ASCII.isHexDigit(m)) {
3362 return ASCII.toDigit(n) * 16 + ASCII.toDigit(m);
3363 }
3364 } else if (n == '{' && ASCII.isHexDigit(peek())) {
3365 int ch = 0;
3366 while (ASCII.isHexDigit(n = read())) {
3367 ch = (ch << 4) + ASCII.toDigit(n);
3368 if (ch > Character.MAX_CODE_POINT)
3369 throw error("Hexadecimal codepoint is too big");
3370 }
3371 if (n != '}')
3372 throw error("Unclosed hexadecimal escape sequence");
3373 return ch;
3374 }
3375 throw error("Illegal hexadecimal escape sequence");
3376 }
3377
3378 /**
3379 * Utility method for parsing unicode escape sequences.
3380 */
3381 private int cursor() {
3382 return cursor;
3383 }
3384
3385 private void setcursor(int pos) {
3386 cursor = pos;
3387 }
3388
3389 private int uxxxx() {
3390 int n = 0;
3391 for (int i = 0; i < 4; i++) {
3392 int ch = read();
3393 if (!ASCII.isHexDigit(ch)) {
3394 throw error("Illegal Unicode escape sequence");
3395 }
3396 n = n * 16 + ASCII.toDigit(ch);
3397 }
3398 return n;
3399 }
3400
3401 private int u() {
3402 int n = uxxxx();
3403 if (Character.isHighSurrogate((char)n)) {
3404 int cur = cursor();
3405 if (read() == '\\' && read() == 'u') {
3406 int n2 = uxxxx();
3407 if (Character.isLowSurrogate((char)n2))
3408 return Character.toCodePoint((char)n, (char)n2);
3409 }
3410 setcursor(cur);
3411 }
3412 return n;
3413 }
3414
3415 private int N() {
3416 if (read() == '{') {
3417 int i = cursor;
3418 while (cursor < patternLength && read() != '}') {}
3419 if (cursor > patternLength)
3420 throw error("Unclosed character name escape sequence");
3421 String name = new String(temp, i, cursor - i - 1);
3422 try {
3423 return Character.codePointOf(name);
3424 } catch (IllegalArgumentException x) {
3425 throw error("Unknown character name [" + name + "]");
3426 }
3427 }
3428 throw error("Illegal character name escape sequence");
3429 }
3430
3431 //
3432 // Utility methods for code point support
3433 //
3434 private static final int countChars(CharSequence seq, int index,
3435 int lengthInCodePoints) {
3436 // optimization
3437 if (lengthInCodePoints == 1 && !Character.isHighSurrogate(seq.charAt(index))) {
3438 assert (index >= 0 && index < seq.length());
3439 return 1;
3440 }
3441 int length = seq.length();
3442 int x = index;
3443 if (lengthInCodePoints >= 0) {
3444 assert (index >= 0 && index < length);
3445 for (int i = 0; x < length && i < lengthInCodePoints; i++) {
3446 if (Character.isHighSurrogate(seq.charAt(x++))) {
3447 if (x < length && Character.isLowSurrogate(seq.charAt(x))) {
3448 x++;
3449 }
3450 }
3451 }
3452 return x - index;
3453 }
3454
3455 assert (index >= 0 && index <= length);
3456 if (index == 0) {
3457 return 0;
3458 }
3459 int len = -lengthInCodePoints;
3460 for (int i = 0; x > 0 && i < len; i++) {
3461 if (Character.isLowSurrogate(seq.charAt(--x))) {
3462 if (x > 0 && Character.isHighSurrogate(seq.charAt(x-1))) {
3463 x--;
3464 }
3465 }
3466 }
3467 return index - x;
3468 }
3469
3470 private static final int countCodePoints(CharSequence seq) {
3471 int length = seq.length();
3472 int n = 0;
3473 for (int i = 0; i < length; ) {
3474 n++;
3475 if (Character.isHighSurrogate(seq.charAt(i++))) {
3476 if (i < length && Character.isLowSurrogate(seq.charAt(i))) {
3477 i++;
3478 }
3479 }
3480 }
3481 return n;
3482 }
3483
3484 /**
3485 * Creates a bit vector for matching Latin-1 values. A normal BitClass
3486 * never matches values above Latin-1, and a complemented BitClass always
3487 * matches values above Latin-1.
3488 */
3489 static final class BitClass implements BmpCharPredicate {
3490 final boolean[] bits;
3491 BitClass() {
3492 bits = new boolean[256];
3493 }
3494 BitClass add(int c, int flags) {
3495 assert c >= 0 && c <= 255;
3496 if ((flags & CASE_INSENSITIVE) != 0) {
3497 if (ASCII.isAscii(c)) {
3498 bits[ASCII.toUpper(c)] = true;
3499 bits[ASCII.toLower(c)] = true;
3500 } else if ((flags & UNICODE_CASE) != 0) {
3501 bits[Character.toLowerCase(c)] = true;
3502 bits[Character.toUpperCase(c)] = true;
3503 }
3504 }
3505 bits[c] = true;
3506 return this;
3507 }
3508 public boolean is(int ch) {
3509 return ch < 256 && bits[ch];
3510 }
3511 }
3512
3513
3514 /**
3515 * Utility method for creating a string slice matcher.
3516 */
3517 private Node newSlice(int[] buf, int count, boolean hasSupplementary) {
3518 int[] tmp = new int[count];
3519 if (has(CASE_INSENSITIVE)) {
3520 if (has(UNICODE_CASE)) {
3521 for (int i = 0; i < count; i++) {
3522 tmp[i] = Character.toLowerCase(
3523 Character.toUpperCase(buf[i]));
3524 }
3525 return hasSupplementary? new SliceUS(tmp) : new SliceU(tmp);
3526 }
3527 for (int i = 0; i < count; i++) {
3528 tmp[i] = ASCII.toLower(buf[i]);
3529 }
3530 return hasSupplementary? new SliceIS(tmp) : new SliceI(tmp);
3531 }
3532 for (int i = 0; i < count; i++) {
3533 tmp[i] = buf[i];
3534 }
3535 return hasSupplementary ? new SliceS(tmp) : new Slice(tmp);
3536 }
3537
3538 /**
3539 * The following classes are the building components of the object
3540 * tree that represents a compiled regular expression. The object tree
3541 * is made of individual elements that handle constructs in the Pattern.
3542 * Each type of object knows how to match its equivalent construct with
3543 * the match() method.
3544 */
3545
3546 /**
3547 * Base class for all node classes. Subclasses should override the match()
3548 * method as appropriate. This class is an accepting node, so its match()
3549 * always returns true.
3550 */
3551 static class Node extends Object {
3552 Node next;
3553 Node() {
3554 next = Pattern.accept;
3555 }
3556 /**
3557 * This method implements the classic accept node.
3558 */
3559 boolean match(Matcher matcher, int i, CharSequence seq) {
3560 matcher.last = i;
3561 matcher.groups[0] = matcher.first;
3562 matcher.groups[1] = matcher.last;
3563 return true;
3564 }
3565 /**
3566 * This method is good for all zero length assertions.
3567 */
3568 boolean study(TreeInfo info) {
3569 if (next != null) {
3570 return next.study(info);
3571 } else {
3572 return info.deterministic;
3573 }
3574 }
3575 }
3576
3577 static class LastNode extends Node {
3578 /**
3579 * This method implements the classic accept node with
3580 * the addition of a check to see if the match occurred
3581 * using all of the input.
3582 */
3583 boolean match(Matcher matcher, int i, CharSequence seq) {
3584 if (matcher.acceptMode == Matcher.ENDANCHOR && i != matcher.to)
3585 return false;
3586 matcher.last = i;
3587 matcher.groups[0] = matcher.first;
3588 matcher.groups[1] = matcher.last;
3589 return true;
3590 }
3591 }
3592
3593 /**
3594 * Used for REs that can start anywhere within the input string.
3595 * This basically tries to match repeatedly at each spot in the
3596 * input string, moving forward after each try. An anchored search
3597 * or a BnM will bypass this node completely.
3598 */
3599 static class Start extends Node {
3600 int minLength;
3601 Start(Node node) {
3602 this.next = node;
3603 TreeInfo info = new TreeInfo();
3604 next.study(info);
3605 minLength = info.minLength;
3606 }
3607 boolean match(Matcher matcher, int i, CharSequence seq) {
3608 if (i > matcher.to - minLength) {
3609 matcher.hitEnd = true;
3610 return false;
3611 }
3612 int guard = matcher.to - minLength;
3613 for (; i <= guard; i++) {
3614 if (next.match(matcher, i, seq)) {
3615 matcher.first = i;
3616 matcher.groups[0] = matcher.first;
3617 matcher.groups[1] = matcher.last;
3618 return true;
3619 }
3620 }
3621 matcher.hitEnd = true;
3622 return false;
3623 }
3624 boolean study(TreeInfo info) {
3625 next.study(info);
3626 info.maxValid = false;
3627 info.deterministic = false;
3628 return false;
3629 }
3630 }
3631
3632 /*
3633 * StartS supports supplementary characters, including unpaired surrogates.
3634 */
3635 static final class StartS extends Start {
3636 StartS(Node node) {
3637 super(node);
3638 }
3639 boolean match(Matcher matcher, int i, CharSequence seq) {
3640 if (i > matcher.to - minLength) {
3641 matcher.hitEnd = true;
3642 return false;
3643 }
3644 int guard = matcher.to - minLength;
3645 while (i <= guard) {
3646 //if ((ret = next.match(matcher, i, seq)) || i == guard)
3647 if (next.match(matcher, i, seq)) {
3648 matcher.first = i;
3649 matcher.groups[0] = matcher.first;
3650 matcher.groups[1] = matcher.last;
3651 return true;
3652 }
3653 if (i == guard)
3654 break;
3655 // Optimization to move to the next character. This is
3656 // faster than countChars(seq, i, 1).
3657 if (Character.isHighSurrogate(seq.charAt(i++))) {
3658 if (i < seq.length() &&
3659 Character.isLowSurrogate(seq.charAt(i))) {
3660 i++;
3661 }
3662 }
3663 }
3664 matcher.hitEnd = true;
3665 return false;
3666 }
3667 }
3668
3669 /**
3670 * Node to anchor at the beginning of input. This object implements the
3671 * match for a \A sequence, and the caret anchor will use this if not in
3672 * multiline mode.
3673 */
3674 static final class Begin extends Node {
3675 boolean match(Matcher matcher, int i, CharSequence seq) {
3676 int fromIndex = (matcher.anchoringBounds) ?
3677 matcher.from : 0;
3678 if (i == fromIndex && next.match(matcher, i, seq)) {
3679 matcher.first = i;
3680 matcher.groups[0] = i;
3681 matcher.groups[1] = matcher.last;
3682 return true;
3683 } else {
3684 return false;
3685 }
3686 }
3687 }
3688
3689 /**
3690 * Node to anchor at the end of input. This is the absolute end, so this
3691 * should not match at the last newline before the end as $ will.
3692 */
3693 static final class End extends Node {
3694 boolean match(Matcher matcher, int i, CharSequence seq) {
3695 int endIndex = (matcher.anchoringBounds) ?
3696 matcher.to : matcher.getTextLength();
3697 if (i == endIndex) {
3698 matcher.hitEnd = true;
3699 return next.match(matcher, i, seq);
3700 }
3701 return false;
3702 }
3703 }
3704
3705 /**
3706 * Node to anchor at the beginning of a line. This is essentially the
3707 * object to match for the multiline ^.
3708 */
3709 static final class Caret extends Node {
3710 boolean match(Matcher matcher, int i, CharSequence seq) {
3711 int startIndex = matcher.from;
3712 int endIndex = matcher.to;
3713 if (!matcher.anchoringBounds) {
3714 startIndex = 0;
3715 endIndex = matcher.getTextLength();
3716 }
3717 // Perl does not match ^ at end of input even after newline
3718 if (i == endIndex) {
3719 matcher.hitEnd = true;
3720 return false;
3721 }
3722 if (i > startIndex) {
3723 char ch = seq.charAt(i-1);
3724 if (ch != '\n' && ch != '\r'
3725 && (ch|1) != '\u2029'
3726 && ch != '\u0085' ) {
3727 return false;
3728 }
3729 // Should treat /r/n as one newline
3730 if (ch == '\r' && seq.charAt(i) == '\n')
3731 return false;
3732 }
3733 return next.match(matcher, i, seq);
3734 }
3735 }
3736
3737 /**
3738 * Node to anchor at the beginning of a line when in unixdot mode.
3739 */
3740 static final class UnixCaret extends Node {
3741 boolean match(Matcher matcher, int i, CharSequence seq) {
3742 int startIndex = matcher.from;
3743 int endIndex = matcher.to;
3744 if (!matcher.anchoringBounds) {
3745 startIndex = 0;
3746 endIndex = matcher.getTextLength();
3747 }
3748 // Perl does not match ^ at end of input even after newline
3749 if (i == endIndex) {
3750 matcher.hitEnd = true;
3751 return false;
3752 }
3753 if (i > startIndex) {
3754 char ch = seq.charAt(i-1);
3755 if (ch != '\n') {
3756 return false;
3757 }
3758 }
3759 return next.match(matcher, i, seq);
3760 }
3761 }
3762
3763 /**
3764 * Node to match the location where the last match ended.
3765 * This is used for the \G construct.
3766 */
3767 static final class LastMatch extends Node {
3768 boolean match(Matcher matcher, int i, CharSequence seq) {
3769 if (i != matcher.oldLast)
3770 return false;
3771 return next.match(matcher, i, seq);
3772 }
3773 }
3774
3775 /**
3776 * Node to anchor at the end of a line or the end of input based on the
3777 * multiline mode.
3778 *
3779 * When not in multiline mode, the $ can only match at the very end
3780 * of the input, unless the input ends in a line terminator in which
3781 * it matches right before the last line terminator.
3782 *
3783 * Note that \r\n is considered an atomic line terminator.
3784 *
3785 * Like ^ the $ operator matches at a position, it does not match the
3786 * line terminators themselves.
3787 */
3788 static final class Dollar extends Node {
3789 boolean multiline;
3790 Dollar(boolean mul) {
3791 multiline = mul;
3792 }
3793 boolean match(Matcher matcher, int i, CharSequence seq) {
3794 int endIndex = (matcher.anchoringBounds) ?
3795 matcher.to : matcher.getTextLength();
3796 if (!multiline) {
3797 if (i < endIndex - 2)
3798 return false;
3799 if (i == endIndex - 2) {
3800 char ch = seq.charAt(i);
3801 if (ch != '\r')
3802 return false;
3803 ch = seq.charAt(i + 1);
3804 if (ch != '\n')
3805 return false;
3806 }
3807 }
3808 // Matches before any line terminator; also matches at the
3809 // end of input
3810 // Before line terminator:
3811 // If multiline, we match here no matter what
3812 // If not multiline, fall through so that the end
3813 // is marked as hit; this must be a /r/n or a /n
3814 // at the very end so the end was hit; more input
3815 // could make this not match here
3816 if (i < endIndex) {
3817 char ch = seq.charAt(i);
3818 if (ch == '\n') {
3819 // No match between \r\n
3820 if (i > 0 && seq.charAt(i-1) == '\r')
3821 return false;
3822 if (multiline)
3823 return next.match(matcher, i, seq);
3824 } else if (ch == '\r' || ch == '\u0085' ||
3825 (ch|1) == '\u2029') {
3826 if (multiline)
3827 return next.match(matcher, i, seq);
3828 } else { // No line terminator, no match
3829 return false;
3830 }
3831 }
3832 // Matched at current end so hit end
3833 matcher.hitEnd = true;
3834 // If a $ matches because of end of input, then more input
3835 // could cause it to fail!
3836 matcher.requireEnd = true;
3837 return next.match(matcher, i, seq);
3838 }
3839 boolean study(TreeInfo info) {
3840 next.study(info);
3841 return info.deterministic;
3842 }
3843 }
3844
3845 /**
3846 * Node to anchor at the end of a line or the end of input based on the
3847 * multiline mode when in unix lines mode.
3848 */
3849 static final class UnixDollar extends Node {
3850 boolean multiline;
3851 UnixDollar(boolean mul) {
3852 multiline = mul;
3853 }
3854 boolean match(Matcher matcher, int i, CharSequence seq) {
3855 int endIndex = (matcher.anchoringBounds) ?
3856 matcher.to : matcher.getTextLength();
3857 if (i < endIndex) {
3858 char ch = seq.charAt(i);
3859 if (ch == '\n') {
3860 // If not multiline, then only possible to
3861 // match at very end or one before end
3862 if (multiline == false && i != endIndex - 1)
3863 return false;
3864 // If multiline return next.match without setting
3865 // matcher.hitEnd
3866 if (multiline)
3867 return next.match(matcher, i, seq);
3868 } else {
3869 return false;
3870 }
3871 }
3872 // Matching because at the end or 1 before the end;
3873 // more input could change this so set hitEnd
3874 matcher.hitEnd = true;
3875 // If a $ matches because of end of input, then more input
3876 // could cause it to fail!
3877 matcher.requireEnd = true;
3878 return next.match(matcher, i, seq);
3879 }
3880 boolean study(TreeInfo info) {
3881 next.study(info);
3882 return info.deterministic;
3883 }
3884 }
3885
3886 /**
3887 * Node class that matches a Unicode line ending '\R'
3888 */
3889 static final class LineEnding extends Node {
3890 boolean match(Matcher matcher, int i, CharSequence seq) {
3891 // (u+000Du+000A|[u+000Au+000Bu+000Cu+000Du+0085u+2028u+2029])
3892 if (i < matcher.to) {
3893 int ch = seq.charAt(i);
3894 if (ch == 0x0A || ch == 0x0B || ch == 0x0C ||
3895 ch == 0x85 || ch == 0x2028 || ch == 0x2029)
3896 return next.match(matcher, i + 1, seq);
3897 if (ch == 0x0D) {
3898 i++;
3899 if (i < matcher.to) {
3900 if (seq.charAt(i) == 0x0A &&
3901 next.match(matcher, i + 1, seq)) {
3902 return true;
3903 }
3904 } else {
3905 matcher.hitEnd = true;
3906 }
3907 return next.match(matcher, i, seq);
3908 }
3909 } else {
3910 matcher.hitEnd = true;
3911 }
3912 return false;
3913 }
3914 boolean study(TreeInfo info) {
3915 info.minLength++;
3916 info.maxLength += 2;
3917 return next.study(info);
3918 }
3919 }
3920
3921 /**
3922 * Abstract node class to match one character satisfying some
3923 * boolean property.
3924 */
3925 static class CharProperty extends Node {
3926 final CharPredicate predicate;
3927
3928 CharProperty (CharPredicate predicate) {
3929 this.predicate = predicate;
3930 }
3931 boolean match(Matcher matcher, int i, CharSequence seq) {
3932 if (i < matcher.to) {
3933 int ch = Character.codePointAt(seq, i);
3934 return predicate.is(ch) &&
3935 next.match(matcher, i + Character.charCount(ch), seq);
3936 } else {
3937 matcher.hitEnd = true;
3938 return false;
3939 }
3940 }
3941 boolean study(TreeInfo info) {
3942 info.minLength++;
3943 info.maxLength++;
3944 return next.study(info);
3945 }
3946 }
3947
3948 /**
3949 * Optimized version of CharProperty that works only for
3950 * properties never satisfied by Supplementary characters.
3951 */
3952 private static class BmpCharProperty extends CharProperty {
3953 BmpCharProperty (BmpCharPredicate predicate) {
3954 super(predicate);
3955 }
3956 boolean match(Matcher matcher, int i, CharSequence seq) {
3957 if (i < matcher.to) {
3958 return predicate.is(seq.charAt(i)) &&
3959 next.match(matcher, i + 1, seq);
3960 } else {
3961 matcher.hitEnd = true;
3962 return false;
3963 }
3964 }
3965 }
3966
3967 private static class NFCCharProperty extends Node {
3968 CharPredicate predicate;
3969 NFCCharProperty (CharPredicate predicate) {
3970 this.predicate = predicate;
3971 }
3972
3973 boolean match(Matcher matcher, int i, CharSequence seq) {
3974 if (i < matcher.to) {
3975 int ch0 = Character.codePointAt(seq, i);
3976 int n = Character.charCount(ch0);
3977 int j = i + n;
3978 // Fast check if it's necessary to call Normalizer;
3979 // testing Grapheme.isBoundary is enough for this case
3980 while (j < matcher.to) {
3981 int ch1 = Character.codePointAt(seq, j);
3982 if (Grapheme.isBoundary(ch0, ch1))
3983 break;
3984 ch0 = ch1;
3985 j += Character.charCount(ch1);
3986 }
3987 if (i + n == j) { // single, assume nfc cp
3988 if (predicate.is(ch0))
3989 return next.match(matcher, j, seq);
3990 } else {
3991 while (i + n < j) {
3992 String nfc = Normalizer.normalize(
3993 seq.toString().substring(i, j), Normalizer.Form.NFC);
3994 if (nfc.codePointCount(0, nfc.length()) == 1) {
3995 if (predicate.is(nfc.codePointAt(0)) &&
3996 next.match(matcher, j, seq)) {
3997 return true;
3998 }
3999 }
4000
4001 ch0 = Character.codePointBefore(seq, j);
4002 j -= Character.charCount(ch0);
4003 }
4004 }
4005 if (j < matcher.to)
4006 return false;
4007 }
4008 matcher.hitEnd = true;
4009 return false;
4010 }
4011
4012 boolean study(TreeInfo info) {
4013 info.minLength++;
4014 info.deterministic = false;
4015 return next.study(info);
4016 }
4017 }
4018
4019 /**
4020 * Node class that matches an unicode extended grapheme cluster
4021 */
4022 static class XGrapheme extends Node {
4023 boolean match(Matcher matcher, int i, CharSequence seq) {
4024 if (i < matcher.to) {
4025 i = Grapheme.nextBoundary(seq, i, matcher.to);
4026 return next.match(matcher, i, seq);
4027 }
4028 matcher.hitEnd = true;
4029 return false;
4030 }
4031
4032 boolean study(TreeInfo info) {
4033 info.minLength++;
4034 info.deterministic = false;
4035 return next.study(info);
4036 }
4037 }
4038
4039 /**
4040 * Node class that handles grapheme boundaries
4041 */
4042 static class GraphemeBound extends Node {
4043 boolean match(Matcher matcher, int i, CharSequence seq) {
4044 int startIndex = matcher.from;
4045 int endIndex = matcher.to;
4046 if (matcher.transparentBounds) {
4047 startIndex = 0;
4048 endIndex = matcher.getTextLength();
4049 }
4050 if (i == startIndex) {
4051 return next.match(matcher, i, seq);
4052 }
4053 if (i < endIndex) {
4054 if (Character.isSurrogatePair(seq.charAt(i-1), seq.charAt(i)) ||
4055 Grapheme.nextBoundary(seq,
4056 i - Character.charCount(Character.codePointBefore(seq, i)),
4057 i + Character.charCount(Character.codePointAt(seq, i))) > i) {
4058 return false;
4059 }
4060 } else {
4061 matcher.hitEnd = true;
4062 matcher.requireEnd = true;
4063 }
4064 return next.match(matcher, i, seq);
4065 }
4066 }
4067
4068 /**
4069 * Base class for all Slice nodes
4070 */
4071 static class SliceNode extends Node {
4072 int[] buffer;
4073 SliceNode(int[] buf) {
4074 buffer = buf;
4075 }
4076 boolean study(TreeInfo info) {
4077 info.minLength += buffer.length;
4078 info.maxLength += buffer.length;
4079 return next.study(info);
4080 }
4081 }
4082
4083 /**
4084 * Node class for a case sensitive/BMP-only sequence of literal
4085 * characters.
4086 */
4087 static class Slice extends SliceNode {
4088 Slice(int[] buf) {
4089 super(buf);
4090 }
4091 boolean match(Matcher matcher, int i, CharSequence seq) {
4092 int[] buf = buffer;
4093 int len = buf.length;
4094 for (int j=0; j<len; j++) {
4095 if ((i+j) >= matcher.to) {
4096 matcher.hitEnd = true;
4097 return false;
4098 }
4099 if (buf[j] != seq.charAt(i+j))
4100 return false;
4101 }
4102 return next.match(matcher, i+len, seq);
4103 }
4104 }
4105
4106 /**
4107 * Node class for a case_insensitive/BMP-only sequence of literal
4108 * characters.
4109 */
4110 static class SliceI extends SliceNode {
4111 SliceI(int[] buf) {
4112 super(buf);
4113 }
4114 boolean match(Matcher matcher, int i, CharSequence seq) {
4115 int[] buf = buffer;
4116 int len = buf.length;
4117 for (int j=0; j<len; j++) {
4118 if ((i+j) >= matcher.to) {
4119 matcher.hitEnd = true;
4120 return false;
4121 }
4122 int c = seq.charAt(i+j);
4123 if (buf[j] != c &&
4124 buf[j] != ASCII.toLower(c))
4125 return false;
4126 }
4127 return next.match(matcher, i+len, seq);
4128 }
4129 }
4130
4131 /**
4132 * Node class for a unicode_case_insensitive/BMP-only sequence of
4133 * literal characters. Uses unicode case folding.
4134 */
4135 static final class SliceU extends SliceNode {
4136 SliceU(int[] buf) {
4137 super(buf);
4138 }
4139 boolean match(Matcher matcher, int i, CharSequence seq) {
4140 int[] buf = buffer;
4141 int len = buf.length;
4142 for (int j=0; j<len; j++) {
4143 if ((i+j) >= matcher.to) {
4144 matcher.hitEnd = true;
4145 return false;
4146 }
4147 int c = seq.charAt(i+j);
4148 if (buf[j] != c &&
4149 buf[j] != Character.toLowerCase(Character.toUpperCase(c)))
4150 return false;
4151 }
4152 return next.match(matcher, i+len, seq);
4153 }
4154 }
4155
4156 /**
4157 * Node class for a case sensitive sequence of literal characters
4158 * including supplementary characters.
4159 */
4160 static final class SliceS extends Slice {
4161 SliceS(int[] buf) {
4162 super(buf);
4163 }
4164 boolean match(Matcher matcher, int i, CharSequence seq) {
4165 int[] buf = buffer;
4166 int x = i;
4167 for (int j = 0; j < buf.length; j++) {
4168 if (x >= matcher.to) {
4169 matcher.hitEnd = true;
4170 return false;
4171 }
4172 int c = Character.codePointAt(seq, x);
4173 if (buf[j] != c)
4174 return false;
4175 x += Character.charCount(c);
4176 if (x > matcher.to) {
4177 matcher.hitEnd = true;
4178 return false;
4179 }
4180 }
4181 return next.match(matcher, x, seq);
4182 }
4183 }
4184
4185 /**
4186 * Node class for a case insensitive sequence of literal characters
4187 * including supplementary characters.
4188 */
4189 static class SliceIS extends SliceNode {
4190 SliceIS(int[] buf) {
4191 super(buf);
4192 }
4193 int toLower(int c) {
4194 return ASCII.toLower(c);
4195 }
4196 boolean match(Matcher matcher, int i, CharSequence seq) {
4197 int[] buf = buffer;
4198 int x = i;
4199 for (int j = 0; j < buf.length; j++) {
4200 if (x >= matcher.to) {
4201 matcher.hitEnd = true;
4202 return false;
4203 }
4204 int c = Character.codePointAt(seq, x);
4205 if (buf[j] != c && buf[j] != toLower(c))
4206 return false;
4207 x += Character.charCount(c);
4208 if (x > matcher.to) {
4209 matcher.hitEnd = true;
4210 return false;
4211 }
4212 }
4213 return next.match(matcher, x, seq);
4214 }
4215 }
4216
4217 /**
4218 * Node class for a case insensitive sequence of literal characters.
4219 * Uses unicode case folding.
4220 */
4221 static final class SliceUS extends SliceIS {
4222 SliceUS(int[] buf) {
4223 super(buf);
4224 }
4225 int toLower(int c) {
4226 return Character.toLowerCase(Character.toUpperCase(c));
4227 }
4228 }
4229
4230 /**
4231 * The 0 or 1 quantifier. This one class implements all three types.
4232 */
4233 static final class Ques extends Node {
4234 Node atom;
4235 Qtype type;
4236 Ques(Node node, Qtype type) {
4237 this.atom = node;
4238 this.type = type;
4239 }
4240 boolean match(Matcher matcher, int i, CharSequence seq) {
4241 switch (type) {
4242 case GREEDY:
4243 return (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq))
4244 || next.match(matcher, i, seq);
4245 case LAZY:
4246 return next.match(matcher, i, seq)
4247 || (atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq));
4248 case POSSESSIVE:
4249 if (atom.match(matcher, i, seq)) i = matcher.last;
4250 return next.match(matcher, i, seq);
4251 default:
4252 return atom.match(matcher, i, seq) && next.match(matcher, matcher.last, seq);
4253 }
4254 }
4255 boolean study(TreeInfo info) {
4256 if (type != Qtype.INDEPENDENT) {
4257 int minL = info.minLength;
4258 atom.study(info);
4259 info.minLength = minL;
4260 info.deterministic = false;
4261 return next.study(info);
4262 } else {
4263 atom.study(info);
4264 return next.study(info);
4265 }
4266 }
4267 }
4268
4269 /**
4270 * Handles the greedy style repetition with the specified minimum
4271 * and the maximum equal to MAX_REPS, for *, + and {N,} quantifiers.
4272 */
4273 static class CharPropertyGreedy extends Node {
4274 final CharPredicate predicate;
4275 final int cmin;
4276
4277 CharPropertyGreedy(CharProperty cp, int cmin) {
4278 this.predicate = cp.predicate;
4279 this.cmin = cmin;
4280 }
4281 boolean match(Matcher matcher, int i, CharSequence seq) {
4282 int n = 0;
4283 int to = matcher.to;
4284 // greedy, all the way down
4285 while (i < to) {
4286 int ch = Character.codePointAt(seq, i);
4287 if (!predicate.is(ch))
4288 break;
4289 i += Character.charCount(ch);
4290 n++;
4291 }
4292 if (i >= to) {
4293 matcher.hitEnd = true;
4294 }
4295 while (n >= cmin) {
4296 if (next.match(matcher, i, seq))
4297 return true;
4298 if (n == cmin)
4299 return false;
4300 // backing off if match fails
4301 int ch = Character.codePointBefore(seq, i);
4302 i -= Character.charCount(ch);
4303 n--;
4304 }
4305 return false;
4306 }
4307
4308 boolean study(TreeInfo info) {
4309 info.minLength += cmin;
4310 if (info.maxValid) {
4311 info.maxLength += MAX_REPS;
4312 }
4313 info.deterministic = false;
4314 return next.study(info);
4315 }
4316 }
4317
4318 static final class BmpCharPropertyGreedy extends CharPropertyGreedy {
4319
4320 BmpCharPropertyGreedy(BmpCharProperty bcp, int cmin) {
4321 super(bcp, cmin);
4322 }
4323
4324 boolean match(Matcher matcher, int i, CharSequence seq) {
4325 int n = 0;
4326 int to = matcher.to;
4327 while (i < to && predicate.is(seq.charAt(i))) {
4328 i++; n++;
4329 }
4330 if (i >= to) {
4331 matcher.hitEnd = true;
4332 }
4333 while (n >= cmin) {
4334 if (next.match(matcher, i, seq))
4335 return true;
4336 i--; n--; // backing off if match fails
4337 }
4338 return false;
4339 }
4340 }
4341
4342 /**
4343 * Handles the curly-brace style repetition with a specified minimum and
4344 * maximum occurrences. The * quantifier is handled as a special case.
4345 * This class handles the three types.
4346 */
4347 static final class Curly extends Node {
4348 Node atom;
4349 Qtype type;
4350 int cmin;
4351 int cmax;
4352
4353 Curly(Node node, int cmin, int cmax, Qtype type) {
4354 this.atom = node;
4355 this.type = type;
4356 this.cmin = cmin;
4357 this.cmax = cmax;
4358 }
4359 boolean match(Matcher matcher, int i, CharSequence seq) {
4360 int j;
4361 for (j = 0; j < cmin; j++) {
4362 if (atom.match(matcher, i, seq)) {
4363 i = matcher.last;
4364 continue;
4365 }
4366 return false;
4367 }
4368 if (type == Qtype.GREEDY)
4369 return match0(matcher, i, j, seq);
4370 else if (type == Qtype.LAZY)
4371 return match1(matcher, i, j, seq);
4372 else
4373 return match2(matcher, i, j, seq);
4374 }
4375 // Greedy match.
4376 // i is the index to start matching at
4377 // j is the number of atoms that have matched
4378 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4379 if (j >= cmax) {
4380 // We have matched the maximum... continue with the rest of
4381 // the regular expression
4382 return next.match(matcher, i, seq);
4383 }
4384 int backLimit = j;
4385 while (atom.match(matcher, i, seq)) {
4386 // k is the length of this match
4387 int k = matcher.last - i;
4388 if (k == 0) // Zero length match
4389 break;
4390 // Move up index and number matched
4391 i = matcher.last;
4392 j++;
4393 // We are greedy so match as many as we can
4394 while (j < cmax) {
4395 if (!atom.match(matcher, i, seq))
4396 break;
4397 if (i + k != matcher.last) {
4398 if (match0(matcher, matcher.last, j+1, seq))
4399 return true;
4400 break;
4401 }
4402 i += k;
4403 j++;
4404 }
4405 // Handle backing off if match fails
4406 while (j >= backLimit) {
4407 if (next.match(matcher, i, seq))
4408 return true;
4409 i -= k;
4410 j--;
4411 }
4412 return false;
4413 }
4414 return next.match(matcher, i, seq);
4415 }
4416 // Reluctant match. At this point, the minimum has been satisfied.
4417 // i is the index to start matching at
4418 // j is the number of atoms that have matched
4419 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4420 for (;;) {
4421 // Try finishing match without consuming any more
4422 if (next.match(matcher, i, seq))
4423 return true;
4424 // At the maximum, no match found
4425 if (j >= cmax)
4426 return false;
4427 // Okay, must try one more atom
4428 if (!atom.match(matcher, i, seq))
4429 return false;
4430 // If we haven't moved forward then must break out
4431 if (i == matcher.last)
4432 return false;
4433 // Move up index and number matched
4434 i = matcher.last;
4435 j++;
4436 }
4437 }
4438 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4439 for (; j < cmax; j++) {
4440 if (!atom.match(matcher, i, seq))
4441 break;
4442 if (i == matcher.last)
4443 break;
4444 i = matcher.last;
4445 }
4446 return next.match(matcher, i, seq);
4447 }
4448 boolean study(TreeInfo info) {
4449 // Save original info
4450 int minL = info.minLength;
4451 int maxL = info.maxLength;
4452 boolean maxV = info.maxValid;
4453 boolean detm = info.deterministic;
4454 info.reset();
4455
4456 atom.study(info);
4457
4458 int temp = info.minLength * cmin + minL;
4459 if (temp < minL) {
4460 temp = 0xFFFFFFF; // arbitrary large number
4461 }
4462 info.minLength = temp;
4463
4464 if (maxV & info.maxValid) {
4465 temp = info.maxLength * cmax + maxL;
4466 info.maxLength = temp;
4467 if (temp < maxL) {
4468 info.maxValid = false;
4469 }
4470 } else {
4471 info.maxValid = false;
4472 }
4473
4474 if (info.deterministic && cmin == cmax)
4475 info.deterministic = detm;
4476 else
4477 info.deterministic = false;
4478 return next.study(info);
4479 }
4480 }
4481
4482 /**
4483 * Handles the curly-brace style repetition with a specified minimum and
4484 * maximum occurrences in deterministic cases. This is an iterative
4485 * optimization over the Prolog and Loop system which would handle this
4486 * in a recursive way. The * quantifier is handled as a special case.
4487 * If capture is true then this class saves group settings and ensures
4488 * that groups are unset when backing off of a group match.
4489 */
4490 static final class GroupCurly extends Node {
4491 Node atom;
4492 Qtype type;
4493 int cmin;
4494 int cmax;
4495 int localIndex;
4496 int groupIndex;
4497 boolean capture;
4498
4499 GroupCurly(Node node, int cmin, int cmax, Qtype type, int local,
4500 int group, boolean capture) {
4501 this.atom = node;
4502 this.type = type;
4503 this.cmin = cmin;
4504 this.cmax = cmax;
4505 this.localIndex = local;
4506 this.groupIndex = group;
4507 this.capture = capture;
4508 }
4509 boolean match(Matcher matcher, int i, CharSequence seq) {
4510 int[] groups = matcher.groups;
4511 int[] locals = matcher.locals;
4512 int save0 = locals[localIndex];
4513 int save1 = 0;
4514 int save2 = 0;
4515
4516 if (capture) {
4517 save1 = groups[groupIndex];
4518 save2 = groups[groupIndex+1];
4519 }
4520
4521 // Notify GroupTail there is no need to setup group info
4522 // because it will be set here
4523 locals[localIndex] = -1;
4524
4525 boolean ret = true;
4526 for (int j = 0; j < cmin; j++) {
4527 if (atom.match(matcher, i, seq)) {
4528 if (capture) {
4529 groups[groupIndex] = i;
4530 groups[groupIndex+1] = matcher.last;
4531 }
4532 i = matcher.last;
4533 } else {
4534 ret = false;
4535 break;
4536 }
4537 }
4538 if (ret) {
4539 if (type == Qtype.GREEDY) {
4540 ret = match0(matcher, i, cmin, seq);
4541 } else if (type == Qtype.LAZY) {
4542 ret = match1(matcher, i, cmin, seq);
4543 } else {
4544 ret = match2(matcher, i, cmin, seq);
4545 }
4546 }
4547 if (!ret) {
4548 locals[localIndex] = save0;
4549 if (capture) {
4550 groups[groupIndex] = save1;
4551 groups[groupIndex+1] = save2;
4552 }
4553 }
4554 return ret;
4555 }
4556 // Aggressive group match
4557 boolean match0(Matcher matcher, int i, int j, CharSequence seq) {
4558 // don't back off passing the starting "j"
4559 int min = j;
4560 int[] groups = matcher.groups;
4561 int save0 = 0;
4562 int save1 = 0;
4563 if (capture) {
4564 save0 = groups[groupIndex];
4565 save1 = groups[groupIndex+1];
4566 }
4567 for (;;) {
4568 if (j >= cmax)
4569 break;
4570 if (!atom.match(matcher, i, seq))
4571 break;
4572 int k = matcher.last - i;
4573 if (k <= 0) {
4574 if (capture) {
4575 groups[groupIndex] = i;
4576 groups[groupIndex+1] = i + k;
4577 }
4578 i = i + k;
4579 break;
4580 }
4581 for (;;) {
4582 if (capture) {
4583 groups[groupIndex] = i;
4584 groups[groupIndex+1] = i + k;
4585 }
4586 i = i + k;
4587 if (++j >= cmax)
4588 break;
4589 if (!atom.match(matcher, i, seq))
4590 break;
4591 if (i + k != matcher.last) {
4592 if (match0(matcher, i, j, seq))
4593 return true;
4594 break;
4595 }
4596 }
4597 while (j > min) {
4598 if (next.match(matcher, i, seq)) {
4599 if (capture) {
4600 groups[groupIndex+1] = i;
4601 groups[groupIndex] = i - k;
4602 }
4603 return true;
4604 }
4605 // backing off
4606 i = i - k;
4607 if (capture) {
4608 groups[groupIndex+1] = i;
4609 groups[groupIndex] = i - k;
4610 }
4611 j--;
4612
4613 }
4614 break;
4615 }
4616 if (capture) {
4617 groups[groupIndex] = save0;
4618 groups[groupIndex+1] = save1;
4619 }
4620 return next.match(matcher, i, seq);
4621 }
4622 // Reluctant matching
4623 boolean match1(Matcher matcher, int i, int j, CharSequence seq) {
4624 for (;;) {
4625 if (next.match(matcher, i, seq))
4626 return true;
4627 if (j >= cmax)
4628 return false;
4629 if (!atom.match(matcher, i, seq))
4630 return false;
4631 if (i == matcher.last)
4632 return false;
4633 if (capture) {
4634 matcher.groups[groupIndex] = i;
4635 matcher.groups[groupIndex+1] = matcher.last;
4636 }
4637 i = matcher.last;
4638 j++;
4639 }
4640 }
4641 // Possessive matching
4642 boolean match2(Matcher matcher, int i, int j, CharSequence seq) {
4643 for (; j < cmax; j++) {
4644 if (!atom.match(matcher, i, seq)) {
4645 break;
4646 }
4647 if (capture) {
4648 matcher.groups[groupIndex] = i;
4649 matcher.groups[groupIndex+1] = matcher.last;
4650 }
4651 if (i == matcher.last) {
4652 break;
4653 }
4654 i = matcher.last;
4655 }
4656 return next.match(matcher, i, seq);
4657 }
4658 boolean study(TreeInfo info) {
4659 // Save original info
4660 int minL = info.minLength;
4661 int maxL = info.maxLength;
4662 boolean maxV = info.maxValid;
4663 boolean detm = info.deterministic;
4664 info.reset();
4665
4666 atom.study(info);
4667
4668 int temp = info.minLength * cmin + minL;
4669 if (temp < minL) {
4670 temp = 0xFFFFFFF; // Arbitrary large number
4671 }
4672 info.minLength = temp;
4673
4674 if (maxV & info.maxValid) {
4675 temp = info.maxLength * cmax + maxL;
4676 info.maxLength = temp;
4677 if (temp < maxL) {
4678 info.maxValid = false;
4679 }
4680 } else {
4681 info.maxValid = false;
4682 }
4683
4684 if (info.deterministic && cmin == cmax) {
4685 info.deterministic = detm;
4686 } else {
4687 info.deterministic = false;
4688 }
4689 return next.study(info);
4690 }
4691 }
4692
4693 /**
4694 * A Guard node at the end of each atom node in a Branch. It
4695 * serves the purpose of chaining the "match" operation to
4696 * "next" but not the "study", so we can collect the TreeInfo
4697 * of each atom node without including the TreeInfo of the
4698 * "next".
4699 */
4700 static final class BranchConn extends Node {
4701 BranchConn() {}
4702 boolean match(Matcher matcher, int i, CharSequence seq) {
4703 return next.match(matcher, i, seq);
4704 }
4705 boolean study(TreeInfo info) {
4706 return info.deterministic;
4707 }
4708 }
4709
4710 /**
4711 * Handles the branching of alternations. Note this is also used for
4712 * the ? quantifier to branch between the case where it matches once
4713 * and where it does not occur.
4714 */
4715 static final class Branch extends Node {
4716 Node[] atoms = new Node[2];
4717 int size = 2;
4718 Node conn;
4719 Branch(Node first, Node second, Node branchConn) {
4720 conn = branchConn;
4721 atoms[0] = first;
4722 atoms[1] = second;
4723 }
4724
4725 void add(Node node) {
4726 if (size >= atoms.length) {
4727 Node[] tmp = new Node[atoms.length*2];
4728 System.arraycopy(atoms, 0, tmp, 0, atoms.length);
4729 atoms = tmp;
4730 }
4731 atoms[size++] = node;
4732 }
4733
4734 boolean match(Matcher matcher, int i, CharSequence seq) {
4735 for (int n = 0; n < size; n++) {
4736 if (atoms[n] == null) {
4737 if (conn.next.match(matcher, i, seq))
4738 return true;
4739 } else if (atoms[n].match(matcher, i, seq)) {
4740 return true;
4741 }
4742 }
4743 return false;
4744 }
4745
4746 boolean study(TreeInfo info) {
4747 int minL = info.minLength;
4748 int maxL = info.maxLength;
4749 boolean maxV = info.maxValid;
4750
4751 int minL2 = Integer.MAX_VALUE; //arbitrary large enough num
4752 int maxL2 = -1;
4753 for (int n = 0; n < size; n++) {
4754 info.reset();
4755 if (atoms[n] != null)
4756 atoms[n].study(info);
4757 minL2 = Math.min(minL2, info.minLength);
4758 maxL2 = Math.max(maxL2, info.maxLength);
4759 maxV = (maxV & info.maxValid);
4760 }
4761
4762 minL += minL2;
4763 maxL += maxL2;
4764
4765 info.reset();
4766 conn.next.study(info);
4767
4768 info.minLength += minL;
4769 info.maxLength += maxL;
4770 info.maxValid &= maxV;
4771 info.deterministic = false;
4772 return false;
4773 }
4774 }
4775
4776 /**
4777 * The GroupHead saves the location where the group begins in the locals
4778 * and restores them when the match is done.
4779 *
4780 * The matchRef is used when a reference to this group is accessed later
4781 * in the expression. The locals will have a negative value in them to
4782 * indicate that we do not want to unset the group if the reference
4783 * doesn't match.
4784 */
4785 static final class GroupHead extends Node {
4786 int localIndex;
4787 GroupTail tail; // for debug/print only, match does not need to know
4788 GroupHead(int localCount) {
4789 localIndex = localCount;
4790 }
4791 boolean match(Matcher matcher, int i, CharSequence seq) {
4792 int save = matcher.locals[localIndex];
4793 matcher.locals[localIndex] = i;
4794 boolean ret = next.match(matcher, i, seq);
4795 matcher.locals[localIndex] = save;
4796 return ret;
4797 }
4798 }
4799
4800 /**
4801 * The GroupTail handles the setting of group beginning and ending
4802 * locations when groups are successfully matched. It must also be able to
4803 * unset groups that have to be backed off of.
4804 *
4805 * The GroupTail node is also used when a previous group is referenced,
4806 * and in that case no group information needs to be set.
4807 */
4808 static final class GroupTail extends Node {
4809 int localIndex;
4810 int groupIndex;
4811 GroupTail(int localCount, int groupCount) {
4812 localIndex = localCount;
4813 groupIndex = groupCount + groupCount;
4814 }
4815 boolean match(Matcher matcher, int i, CharSequence seq) {
4816 int tmp = matcher.locals[localIndex];
4817 if (tmp >= 0) { // This is the normal group case.
4818 // Save the group so we can unset it if it
4819 // backs off of a match.
4820 int groupStart = matcher.groups[groupIndex];
4821 int groupEnd = matcher.groups[groupIndex+1];
4822
4823 matcher.groups[groupIndex] = tmp;
4824 matcher.groups[groupIndex+1] = i;
4825 if (next.match(matcher, i, seq)) {
4826 return true;
4827 }
4828 matcher.groups[groupIndex] = groupStart;
4829 matcher.groups[groupIndex+1] = groupEnd;
4830 return false;
4831 } else {
4832 // This is a group reference case. We don't need to save any
4833 // group info because it isn't really a group.
4834 matcher.last = i;
4835 return true;
4836 }
4837 }
4838 }
4839
4840 /**
4841 * This sets up a loop to handle a recursive quantifier structure.
4842 */
4843 static final class Prolog extends Node {
4844 Loop loop;
4845 Prolog(Loop loop) {
4846 this.loop = loop;
4847 }
4848 boolean match(Matcher matcher, int i, CharSequence seq) {
4849 return loop.matchInit(matcher, i, seq);
4850 }
4851 boolean study(TreeInfo info) {
4852 return loop.study(info);
4853 }
4854 }
4855
4856 /**
4857 * Handles the repetition count for a greedy Curly. The matchInit
4858 * is called from the Prolog to save the index of where the group
4859 * beginning is stored. A zero length group check occurs in the
4860 * normal match but is skipped in the matchInit.
4861 */
4862 static class Loop extends Node {
4863 Node body;
4864 int countIndex; // local count index in matcher locals
4865 int beginIndex; // group beginning index
4866 int cmin, cmax;
4867 int posIndex;
4868 Loop(int countIndex, int beginIndex) {
4869 this.countIndex = countIndex;
4870 this.beginIndex = beginIndex;
4871 this.posIndex = -1;
4872 }
4873 boolean match(Matcher matcher, int i, CharSequence seq) {
4874 // Avoid infinite loop in zero-length case.
4875 if (i > matcher.locals[beginIndex]) {
4876 int count = matcher.locals[countIndex];
4877
4878 // This block is for before we reach the minimum
4879 // iterations required for the loop to match
4880 if (count < cmin) {
4881 matcher.locals[countIndex] = count + 1;
4882 boolean b = body.match(matcher, i, seq);
4883 // If match failed we must backtrack, so
4884 // the loop count should NOT be incremented
4885 if (!b)
4886 matcher.locals[countIndex] = count;
4887 // Return success or failure since we are under
4888 // minimum
4889 return b;
4890 }
4891 // This block is for after we have the minimum
4892 // iterations required for the loop to match
4893 if (count < cmax) {
4894 // Let's check if we have already tried and failed
4895 // at this starting position "i" in the past.
4896 // If yes, then just return false wihtout trying
4897 // again, to stop the exponential backtracking.
4898 if (posIndex != -1 &&
4899 matcher.localsPos[posIndex].contains(i)) {
4900 return next.match(matcher, i, seq);
4901 }
4902 matcher.locals[countIndex] = count + 1;
4903 boolean b = body.match(matcher, i, seq);
4904 // If match failed we must backtrack, so
4905 // the loop count should NOT be incremented
4906 if (b)
4907 return true;
4908 matcher.locals[countIndex] = count;
4909 // save the failed position
4910 if (posIndex != -1) {
4911 matcher.localsPos[posIndex].add(i);
4912 }
4913 }
4914 }
4915 return next.match(matcher, i, seq);
4916 }
4917 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4918 int save = matcher.locals[countIndex];
4919 boolean ret;
4920 if (posIndex != -1 && matcher.localsPos[posIndex] == null) {
4921 matcher.localsPos[posIndex] = new IntHashSet();
4922 }
4923 if (0 < cmin) {
4924 matcher.locals[countIndex] = 1;
4925 ret = body.match(matcher, i, seq);
4926 } else if (0 < cmax) {
4927 matcher.locals[countIndex] = 1;
4928 ret = body.match(matcher, i, seq);
4929 if (ret == false)
4930 ret = next.match(matcher, i, seq);
4931 } else {
4932 ret = next.match(matcher, i, seq);
4933 }
4934 matcher.locals[countIndex] = save;
4935 return ret;
4936 }
4937 boolean study(TreeInfo info) {
4938 info.maxValid = false;
4939 info.deterministic = false;
4940 return false;
4941 }
4942 }
4943
4944 /**
4945 * Handles the repetition count for a reluctant Curly. The matchInit
4946 * is called from the Prolog to save the index of where the group
4947 * beginning is stored. A zero length group check occurs in the
4948 * normal match but is skipped in the matchInit.
4949 */
4950 static final class LazyLoop extends Loop {
4951 LazyLoop(int countIndex, int beginIndex) {
4952 super(countIndex, beginIndex);
4953 }
4954 boolean match(Matcher matcher, int i, CharSequence seq) {
4955 // Check for zero length group
4956 if (i > matcher.locals[beginIndex]) {
4957 int count = matcher.locals[countIndex];
4958 if (count < cmin) {
4959 matcher.locals[countIndex] = count + 1;
4960 boolean result = body.match(matcher, i, seq);
4961 // If match failed we must backtrack, so
4962 // the loop count should NOT be incremented
4963 if (!result)
4964 matcher.locals[countIndex] = count;
4965 return result;
4966 }
4967 if (next.match(matcher, i, seq))
4968 return true;
4969 if (count < cmax) {
4970 matcher.locals[countIndex] = count + 1;
4971 boolean result = body.match(matcher, i, seq);
4972 // If match failed we must backtrack, so
4973 // the loop count should NOT be incremented
4974 if (!result)
4975 matcher.locals[countIndex] = count;
4976 return result;
4977 }
4978 return false;
4979 }
4980 return next.match(matcher, i, seq);
4981 }
4982 boolean matchInit(Matcher matcher, int i, CharSequence seq) {
4983 int save = matcher.locals[countIndex];
4984 boolean ret = false;
4985 if (0 < cmin) {
4986 matcher.locals[countIndex] = 1;
4987 ret = body.match(matcher, i, seq);
4988 } else if (next.match(matcher, i, seq)) {
4989 ret = true;
4990 } else if (0 < cmax) {
4991 matcher.locals[countIndex] = 1;
4992 ret = body.match(matcher, i, seq);
4993 }
4994 matcher.locals[countIndex] = save;
4995 return ret;
4996 }
4997 boolean study(TreeInfo info) {
4998 info.maxValid = false;
4999 info.deterministic = false;
5000 return false;
5001 }
5002 }
5003
5004 /**
5005 * Refers to a group in the regular expression. Attempts to match
5006 * whatever the group referred to last matched.
5007 */
5008 static class BackRef extends Node {
5009 int groupIndex;
5010 BackRef(int groupCount) {
5011 super();
5012 groupIndex = groupCount + groupCount;
5013 }
5014 boolean match(Matcher matcher, int i, CharSequence seq) {
5015 int j = matcher.groups[groupIndex];
5016 int k = matcher.groups[groupIndex+1];
5017
5018 int groupSize = k - j;
5019 // If the referenced group didn't match, neither can this
5020 if (j < 0)
5021 return false;
5022
5023 // If there isn't enough input left no match
5024 if (i + groupSize > matcher.to) {
5025 matcher.hitEnd = true;
5026 return false;
5027 }
5028 // Check each new char to make sure it matches what the group
5029 // referenced matched last time around
5030 for (int index=0; index<groupSize; index++)
5031 if (seq.charAt(i+index) != seq.charAt(j+index))
5032 return false;
5033
5034 return next.match(matcher, i+groupSize, seq);
5035 }
5036 boolean study(TreeInfo info) {
5037 info.maxValid = false;
5038 return next.study(info);
5039 }
5040 }
5041
5042 static class CIBackRef extends Node {
5043 int groupIndex;
5044 boolean doUnicodeCase;
5045 CIBackRef(int groupCount, boolean doUnicodeCase) {
5046 super();
5047 groupIndex = groupCount + groupCount;
5048 this.doUnicodeCase = doUnicodeCase;
5049 }
5050 boolean match(Matcher matcher, int i, CharSequence seq) {
5051 int j = matcher.groups[groupIndex];
5052 int k = matcher.groups[groupIndex+1];
5053
5054 int groupSize = k - j;
5055
5056 // If the referenced group didn't match, neither can this
5057 if (j < 0)
5058 return false;
5059
5060 // If there isn't enough input left no match
5061 if (i + groupSize > matcher.to) {
5062 matcher.hitEnd = true;
5063 return false;
5064 }
5065
5066 // Check each new char to make sure it matches what the group
5067 // referenced matched last time around
5068 int x = i;
5069 for (int index=0; index<groupSize; index++) {
5070 int c1 = Character.codePointAt(seq, x);
5071 int c2 = Character.codePointAt(seq, j);
5072 if (c1 != c2) {
5073 if (doUnicodeCase) {
5074 int cc1 = Character.toUpperCase(c1);
5075 int cc2 = Character.toUpperCase(c2);
5076 if (cc1 != cc2 &&
5077 Character.toLowerCase(cc1) !=
5078 Character.toLowerCase(cc2))
5079 return false;
5080 } else {
5081 if (ASCII.toLower(c1) != ASCII.toLower(c2))
5082 return false;
5083 }
5084 }
5085 x += Character.charCount(c1);
5086 j += Character.charCount(c2);
5087 }
5088
5089 return next.match(matcher, i+groupSize, seq);
5090 }
5091 boolean study(TreeInfo info) {
5092 info.maxValid = false;
5093 return next.study(info);
5094 }
5095 }
5096
5097 /**
5098 * Searches until the next instance of its atom. This is useful for
5099 * finding the atom efficiently without passing an instance of it
5100 * (greedy problem) and without a lot of wasted search time (reluctant
5101 * problem).
5102 */
5103 static final class First extends Node {
5104 Node atom;
5105 First(Node node) {
5106 this.atom = BnM.optimize(node);
5107 }
5108 boolean match(Matcher matcher, int i, CharSequence seq) {
5109 if (atom instanceof BnM) {
5110 return atom.match(matcher, i, seq)
5111 && next.match(matcher, matcher.last, seq);
5112 }
5113 for (;;) {
5114 if (i > matcher.to) {
5115 matcher.hitEnd = true;
5116 return false;
5117 }
5118 if (atom.match(matcher, i, seq)) {
5119 return next.match(matcher, matcher.last, seq);
5120 }
5121 i += countChars(seq, i, 1);
5122 matcher.first++;
5123 }
5124 }
5125 boolean study(TreeInfo info) {
5126 atom.study(info);
5127 info.maxValid = false;
5128 info.deterministic = false;
5129 return next.study(info);
5130 }
5131 }
5132
5133 /**
5134 * Zero width positive lookahead.
5135 */
5136 static final class Pos extends Node {
5137 Node cond;
5138 Pos(Node cond) {
5139 this.cond = cond;
5140 }
5141 boolean match(Matcher matcher, int i, CharSequence seq) {
5142 int savedTo = matcher.to;
5143 boolean conditionMatched;
5144
5145 // Relax transparent region boundaries for lookahead
5146 if (matcher.transparentBounds)
5147 matcher.to = matcher.getTextLength();
5148 try {
5149 conditionMatched = cond.match(matcher, i, seq);
5150 } finally {
5151 // Reinstate region boundaries
5152 matcher.to = savedTo;
5153 }
5154 return conditionMatched && next.match(matcher, i, seq);
5155 }
5156 }
5157
5158 /**
5159 * Zero width negative lookahead.
5160 */
5161 static final class Neg extends Node {
5162 Node cond;
5163 Neg(Node cond) {
5164 this.cond = cond;
5165 }
5166 boolean match(Matcher matcher, int i, CharSequence seq) {
5167 int savedTo = matcher.to;
5168 boolean conditionMatched;
5169
5170 // Relax transparent region boundaries for lookahead
5171 if (matcher.transparentBounds)
5172 matcher.to = matcher.getTextLength();
5173 try {
5174 if (i < matcher.to) {
5175 conditionMatched = !cond.match(matcher, i, seq);
5176 } else {
5177 // If a negative lookahead succeeds then more input
5178 // could cause it to fail!
5179 matcher.requireEnd = true;
5180 conditionMatched = !cond.match(matcher, i, seq);
5181 }
5182 } finally {
5183 // Reinstate region boundaries
5184 matcher.to = savedTo;
5185 }
5186 return conditionMatched && next.match(matcher, i, seq);
5187 }
5188 }
5189
5190 /**
5191 * For use with lookbehinds; matches the position where the lookbehind
5192 * was encountered.
5193 */
5194 static class LookBehindEndNode extends Node {
5195 private LookBehindEndNode() {} // Singleton
5196
5197 static LookBehindEndNode INSTANCE = new LookBehindEndNode();
5198
5199 boolean match(Matcher matcher, int i, CharSequence seq) {
5200 return i == matcher.lookbehindTo;
5201 }
5202 }
5203
5204 /**
5205 * Zero width positive lookbehind.
5206 */
5207 static class Behind extends Node {
5208 Node cond;
5209 int rmax, rmin;
5210 Behind(Node cond, int rmax, int rmin) {
5211 this.cond = cond;
5212 this.rmax = rmax;
5213 this.rmin = rmin;
5214 }
5215
5216 boolean match(Matcher matcher, int i, CharSequence seq) {
5217 int savedFrom = matcher.from;
5218 boolean conditionMatched = false;
5219 int startIndex = (!matcher.transparentBounds) ?
5220 matcher.from : 0;
5221 int from = Math.max(i - rmax, startIndex);
5222 // Set end boundary
5223 int savedLBT = matcher.lookbehindTo;
5224 matcher.lookbehindTo = i;
5225 // Relax transparent region boundaries for lookbehind
5226 if (matcher.transparentBounds)
5227 matcher.from = 0;
5228 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
5229 conditionMatched = cond.match(matcher, j, seq);
5230 }
5231 matcher.from = savedFrom;
5232 matcher.lookbehindTo = savedLBT;
5233 return conditionMatched && next.match(matcher, i, seq);
5234 }
5235 }
5236
5237 /**
5238 * Zero width positive lookbehind, including supplementary
5239 * characters or unpaired surrogates.
5240 */
5241 static final class BehindS extends Behind {
5242 BehindS(Node cond, int rmax, int rmin) {
5243 super(cond, rmax, rmin);
5244 }
5245 boolean match(Matcher matcher, int i, CharSequence seq) {
5246 int rmaxChars = countChars(seq, i, -rmax);
5247 int rminChars = countChars(seq, i, -rmin);
5248 int savedFrom = matcher.from;
5249 int startIndex = (!matcher.transparentBounds) ?
5250 matcher.from : 0;
5251 boolean conditionMatched = false;
5252 int from = Math.max(i - rmaxChars, startIndex);
5253 // Set end boundary
5254 int savedLBT = matcher.lookbehindTo;
5255 matcher.lookbehindTo = i;
5256 // Relax transparent region boundaries for lookbehind
5257 if (matcher.transparentBounds)
5258 matcher.from = 0;
5259
5260 for (int j = i - rminChars;
5261 !conditionMatched && j >= from;
5262 j -= j>from ? countChars(seq, j, -1) : 1) {
5263 conditionMatched = cond.match(matcher, j, seq);
5264 }
5265 matcher.from = savedFrom;
5266 matcher.lookbehindTo = savedLBT;
5267 return conditionMatched && next.match(matcher, i, seq);
5268 }
5269 }
5270
5271 /**
5272 * Zero width negative lookbehind.
5273 */
5274 static class NotBehind extends Node {
5275 Node cond;
5276 int rmax, rmin;
5277 NotBehind(Node cond, int rmax, int rmin) {
5278 this.cond = cond;
5279 this.rmax = rmax;
5280 this.rmin = rmin;
5281 }
5282
5283 boolean match(Matcher matcher, int i, CharSequence seq) {
5284 int savedLBT = matcher.lookbehindTo;
5285 int savedFrom = matcher.from;
5286 boolean conditionMatched = false;
5287 int startIndex = (!matcher.transparentBounds) ?
5288 matcher.from : 0;
5289 int from = Math.max(i - rmax, startIndex);
5290 matcher.lookbehindTo = i;
5291 // Relax transparent region boundaries for lookbehind
5292 if (matcher.transparentBounds)
5293 matcher.from = 0;
5294 for (int j = i - rmin; !conditionMatched && j >= from; j--) {
5295 conditionMatched = cond.match(matcher, j, seq);
5296 }
5297 // Reinstate region boundaries
5298 matcher.from = savedFrom;
5299 matcher.lookbehindTo = savedLBT;
5300 return !conditionMatched && next.match(matcher, i, seq);
5301 }
5302 }
5303
5304 /**
5305 * Zero width negative lookbehind, including supplementary
5306 * characters or unpaired surrogates.
5307 */
5308 static final class NotBehindS extends NotBehind {
5309 NotBehindS(Node cond, int rmax, int rmin) {
5310 super(cond, rmax, rmin);
5311 }
5312 boolean match(Matcher matcher, int i, CharSequence seq) {
5313 int rmaxChars = countChars(seq, i, -rmax);
5314 int rminChars = countChars(seq, i, -rmin);
5315 int savedFrom = matcher.from;
5316 int savedLBT = matcher.lookbehindTo;
5317 boolean conditionMatched = false;
5318 int startIndex = (!matcher.transparentBounds) ?
5319 matcher.from : 0;
5320 int from = Math.max(i - rmaxChars, startIndex);
5321 matcher.lookbehindTo = i;
5322 // Relax transparent region boundaries for lookbehind
5323 if (matcher.transparentBounds)
5324 matcher.from = 0;
5325 for (int j = i - rminChars;
5326 !conditionMatched && j >= from;
5327 j -= j>from ? countChars(seq, j, -1) : 1) {
5328 conditionMatched = cond.match(matcher, j, seq);
5329 }
5330 //Reinstate region boundaries
5331 matcher.from = savedFrom;
5332 matcher.lookbehindTo = savedLBT;
5333 return !conditionMatched && next.match(matcher, i, seq);
5334 }
5335 }
5336
5337 /**
5338 * Handles word boundaries. Includes a field to allow this one class to
5339 * deal with the different types of word boundaries we can match. The word
5340 * characters include underscores, letters, and digits. Non spacing marks
5341 * can are also part of a word if they have a base character, otherwise
5342 * they are ignored for purposes of finding word boundaries.
5343 */
5344 static final class Bound extends Node {
5345 static int LEFT = 0x1;
5346 static int RIGHT= 0x2;
5347 static int BOTH = 0x3;
5348 static int NONE = 0x4;
5349 int type;
5350 boolean useUWORD;
5351 Bound(int n, boolean useUWORD) {
5352 type = n;
5353 this.useUWORD = useUWORD;
5354 }
5355
5356 boolean isWord(int ch) {
5357 return useUWORD ? CharPredicates.WORD().is(ch)
5358 : (ch == '_' || Character.isLetterOrDigit(ch));
5359 }
5360
5361 int check(Matcher matcher, int i, CharSequence seq) {
5362 int ch;
5363 boolean left = false;
5364 int startIndex = matcher.from;
5365 int endIndex = matcher.to;
5366 if (matcher.transparentBounds) {
5367 startIndex = 0;
5368 endIndex = matcher.getTextLength();
5369 }
5370 if (i > startIndex) {
5371 ch = Character.codePointBefore(seq, i);
5372 left = (isWord(ch) ||
5373 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5374 && hasBaseCharacter(matcher, i-1, seq)));
5375 }
5376 boolean right = false;
5377 if (i < endIndex) {
5378 ch = Character.codePointAt(seq, i);
5379 right = (isWord(ch) ||
5380 ((Character.getType(ch) == Character.NON_SPACING_MARK)
5381 && hasBaseCharacter(matcher, i, seq)));
5382 } else {
5383 // Tried to access char past the end
5384 matcher.hitEnd = true;
5385 // The addition of another char could wreck a boundary
5386 matcher.requireEnd = true;
5387 }
5388 return ((left ^ right) ? (right ? LEFT : RIGHT) : NONE);
5389 }
5390 boolean match(Matcher matcher, int i, CharSequence seq) {
5391 return (check(matcher, i, seq) & type) > 0
5392 && next.match(matcher, i, seq);
5393 }
5394 }
5395
5396 /**
5397 * Non spacing marks only count as word characters in bounds calculations
5398 * if they have a base character.
5399 */
5400 private static boolean hasBaseCharacter(Matcher matcher, int i,
5401 CharSequence seq)
5402 {
5403 int start = (!matcher.transparentBounds) ?
5404 matcher.from : 0;
5405 for (int x=i; x >= start; x--) {
5406 int ch = Character.codePointAt(seq, x);
5407 if (Character.isLetterOrDigit(ch))
5408 return true;
5409 if (Character.getType(ch) == Character.NON_SPACING_MARK)
5410 continue;
5411 return false;
5412 }
5413 return false;
5414 }
5415
5416 /**
5417 * Attempts to match a slice in the input using the Boyer-Moore string
5418 * matching algorithm. The algorithm is based on the idea that the
5419 * pattern can be shifted farther ahead in the search text if it is
5420 * matched right to left.
5421 * <p>
5422 * The pattern is compared to the input one character at a time, from
5423 * the rightmost character in the pattern to the left. If the characters
5424 * all match the pattern has been found. If a character does not match,
5425 * the pattern is shifted right a distance that is the maximum of two
5426 * functions, the bad character shift and the good suffix shift. This
5427 * shift moves the attempted match position through the input more
5428 * quickly than a naive one position at a time check.
5429 * <p>
5430 * The bad character shift is based on the character from the text that
5431 * did not match. If the character does not appear in the pattern, the
5432 * pattern can be shifted completely beyond the bad character. If the
5433 * character does occur in the pattern, the pattern can be shifted to
5434 * line the pattern up with the next occurrence of that character.
5435 * <p>
5436 * The good suffix shift is based on the idea that some subset on the right
5437 * side of the pattern has matched. When a bad character is found, the
5438 * pattern can be shifted right by the pattern length if the subset does
5439 * not occur again in pattern, or by the amount of distance to the
5440 * next occurrence of the subset in the pattern.
5441 *
5442 * Boyer-Moore search methods adapted from code by Amy Yu.
5443 */
5444 static class BnM extends Node {
5445 int[] buffer;
5446 int[] lastOcc;
5447 int[] optoSft;
5448
5449 /**
5450 * Pre calculates arrays needed to generate the bad character
5451 * shift and the good suffix shift. Only the last seven bits
5452 * are used to see if chars match; This keeps the tables small
5453 * and covers the heavily used ASCII range, but occasionally
5454 * results in an aliased match for the bad character shift.
5455 */
5456 static Node optimize(Node node) {
5457 if (!(node instanceof Slice)) {
5458 return node;
5459 }
5460
5461 int[] src = ((Slice) node).buffer;
5462 int patternLength = src.length;
5463 // The BM algorithm requires a bit of overhead;
5464 // If the pattern is short don't use it, since
5465 // a shift larger than the pattern length cannot
5466 // be used anyway.
5467 if (patternLength < 4) {
5468 return node;
5469 }
5470 int i, j;
5471 int[] lastOcc = new int[128];
5472 int[] optoSft = new int[patternLength];
5473 // Precalculate part of the bad character shift
5474 // It is a table for where in the pattern each
5475 // lower 7-bit value occurs
5476 for (i = 0; i < patternLength; i++) {
5477 lastOcc[src[i]&0x7F] = i + 1;
5478 }
5479 // Precalculate the good suffix shift
5480 // i is the shift amount being considered
5481 NEXT: for (i = patternLength; i > 0; i--) {
5482 // j is the beginning index of suffix being considered
5483 for (j = patternLength - 1; j >= i; j--) {
5484 // Testing for good suffix
5485 if (src[j] == src[j-i]) {
5486 // src[j..len] is a good suffix
5487 optoSft[j-1] = i;
5488 } else {
5489 // No match. The array has already been
5490 // filled up with correct values before.
5491 continue NEXT;
5492 }
5493 }
5494 // This fills up the remaining of optoSft
5495 // any suffix can not have larger shift amount
5496 // then its sub-suffix. Why???
5497 while (j > 0) {
5498 optoSft[--j] = i;
5499 }
5500 }
5501 // Set the guard value because of unicode compression
5502 optoSft[patternLength-1] = 1;
5503 if (node instanceof SliceS)
5504 return new BnMS(src, lastOcc, optoSft, node.next);
5505 return new BnM(src, lastOcc, optoSft, node.next);
5506 }
5507 BnM(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5508 this.buffer = src;
5509 this.lastOcc = lastOcc;
5510 this.optoSft = optoSft;
5511 this.next = next;
5512 }
5513 boolean match(Matcher matcher, int i, CharSequence seq) {
5514 int[] src = buffer;
5515 int patternLength = src.length;
5516 int last = matcher.to - patternLength;
5517
5518 // Loop over all possible match positions in text
5519 NEXT: while (i <= last) {
5520 // Loop over pattern from right to left
5521 for (int j = patternLength - 1; j >= 0; j--) {
5522 int ch = seq.charAt(i+j);
5523 if (ch != src[j]) {
5524 // Shift search to the right by the maximum of the
5525 // bad character shift and the good suffix shift
5526 i += Math.max(j + 1 - lastOcc[ch&0x7F], optoSft[j]);
5527 continue NEXT;
5528 }
5529 }
5530 // Entire pattern matched starting at i
5531 matcher.first = i;
5532 boolean ret = next.match(matcher, i + patternLength, seq);
5533 if (ret) {
5534 matcher.first = i;
5535 matcher.groups[0] = matcher.first;
5536 matcher.groups[1] = matcher.last;
5537 return true;
5538 }
5539 i++;
5540 }
5541 // BnM is only used as the leading node in the unanchored case,
5542 // and it replaced its Start() which always searches to the end
5543 // if it doesn't find what it's looking for, so hitEnd is true.
5544 matcher.hitEnd = true;
5545 return false;
5546 }
5547 boolean study(TreeInfo info) {
5548 info.minLength += buffer.length;
5549 info.maxValid = false;
5550 return next.study(info);
5551 }
5552 }
5553
5554 /**
5555 * Supplementary support version of BnM(). Unpaired surrogates are
5556 * also handled by this class.
5557 */
5558 static final class BnMS extends BnM {
5559 int lengthInChars;
5560
5561 BnMS(int[] src, int[] lastOcc, int[] optoSft, Node next) {
5562 super(src, lastOcc, optoSft, next);
5563 for (int cp : buffer) {
5564 lengthInChars += Character.charCount(cp);
5565 }
5566 }
5567 boolean match(Matcher matcher, int i, CharSequence seq) {
5568 int[] src = buffer;
5569 int patternLength = src.length;
5570 int last = matcher.to - lengthInChars;
5571
5572 // Loop over all possible match positions in text
5573 NEXT: while (i <= last) {
5574 // Loop over pattern from right to left
5575 int ch;
5576 for (int j = countChars(seq, i, patternLength), x = patternLength - 1;
5577 j > 0; j -= Character.charCount(ch), x--) {
5578 ch = Character.codePointBefore(seq, i+j);
5579 if (ch != src[x]) {
5580 // Shift search to the right by the maximum of the
5581 // bad character shift and the good suffix shift
5582 int n = Math.max(x + 1 - lastOcc[ch&0x7F], optoSft[x]);
5583 i += countChars(seq, i, n);
5584 continue NEXT;
5585 }
5586 }
5587 // Entire pattern matched starting at i
5588 matcher.first = i;
5589 boolean ret = next.match(matcher, i + lengthInChars, seq);
5590 if (ret) {
5591 matcher.first = i;
5592 matcher.groups[0] = matcher.first;
5593 matcher.groups[1] = matcher.last;
5594 return true;
5595 }
5596 i += countChars(seq, i, 1);
5597 }
5598 matcher.hitEnd = true;
5599 return false;
5600 }
5601 }
5602
5603 @FunctionalInterface
5604 static interface CharPredicate {
5605 boolean is(int ch);
5606
5607 default CharPredicate and(CharPredicate p) {
5608 return ch -> is(ch) && p.is(ch);
5609 }
5610 default CharPredicate union(CharPredicate p) {
5611 return ch -> is(ch) || p.is(ch);
5612 }
5613 default CharPredicate union(CharPredicate p1,
5614 CharPredicate p2 ) {
5615 return ch -> is(ch) || p1.is(ch) || p2.is(ch);
5616 }
5617 default CharPredicate negate() {
5618 return ch -> !is(ch);
5619 }
5620 }
5621
5622 static interface BmpCharPredicate extends CharPredicate {
5623
5624 default CharPredicate and(CharPredicate p) {
5625 if (p instanceof BmpCharPredicate)
5626 return (BmpCharPredicate)(ch -> is(ch) && p.is(ch));
5627 return ch -> is(ch) && p.is(ch);
5628 }
5629 default CharPredicate union(CharPredicate p) {
5630 if (p instanceof BmpCharPredicate)
5631 return (BmpCharPredicate)(ch -> is(ch) || p.is(ch));
5632 return ch -> is(ch) || p.is(ch);
5633 }
5634 static CharPredicate union(CharPredicate... predicates) {
5635 CharPredicate cp = ch -> {
5636 for (CharPredicate p : predicates) {
5637 if (!p.is(ch))
5638 return false;
5639 }
5640 return true;
5641 };
5642 for (CharPredicate p : predicates) {
5643 if (! (p instanceof BmpCharPredicate))
5644 return cp;
5645 }
5646 return (BmpCharPredicate)cp;
5647 }
5648 }
5649
5650 /**
5651 * matches a Perl vertical whitespace
5652 */
5653 static BmpCharPredicate VertWS() {
5654 return cp -> (cp >= 0x0A && cp <= 0x0D) ||
5655 cp == 0x85 || cp == 0x2028 || cp == 0x2029;
5656 }
5657
5658 /**
5659 * matches a Perl horizontal whitespace
5660 */
5661 static BmpCharPredicate HorizWS() {
5662 return cp ->
5663 cp == 0x09 || cp == 0x20 || cp == 0xa0 || cp == 0x1680 ||
5664 cp == 0x180e || cp >= 0x2000 && cp <= 0x200a || cp == 0x202f ||
5665 cp == 0x205f || cp == 0x3000;
5666 }
5667
5668 /**
5669 * for the Unicode category ALL and the dot metacharacter when
5670 * in dotall mode.
5671 */
5672 static CharPredicate ALL() {
5673 return ch -> true;
5674 }
5675
5676 /**
5677 * for the dot metacharacter when dotall is not enabled.
5678 */
5679 static CharPredicate DOT() {
5680 return ch ->
5681 (ch != '\n' && ch != '\r'
5682 && (ch|1) != '\u2029'
5683 && ch != '\u0085');
5684 }
5685
5686 /**
5687 * the dot metacharacter when dotall is not enabled but UNIX_LINES is enabled.
5688 */
5689 static CharPredicate UNIXDOT() {
5690 return ch -> ch != '\n';
5691 }
5692
5693 /**
5694 * Indicate that matches a Supplementary Unicode character
5695 */
5696 static CharPredicate SingleS(int c) {
5697 return ch -> ch == c;
5698 }
5699
5700 /**
5701 * A bmp/optimized predicate of single
5702 */
5703 static BmpCharPredicate Single(int c) {
5704 return ch -> ch == c;
5705 }
5706
5707 /**
5708 * Case insensitive matches a given BMP character
5709 */
5710 static BmpCharPredicate SingleI(int lower, int upper) {
5711 return ch -> ch == lower || ch == upper;
5712 }
5713
5714 /**
5715 * Unicode case insensitive matches a given Unicode character
5716 */
5717 static CharPredicate SingleU(int lower) {
5718 return ch -> lower == ch ||
5719 lower == Character.toLowerCase(Character.toUpperCase(ch));
5720 }
5721
5722 private static boolean inRange(int lower, int ch, int upper) {
5723 return lower <= ch && ch <= upper;
5724 }
5725
5726 /**
5727 * Charactrs within a explicit value range
5728 */
5729 static CharPredicate Range(int lower, int upper) {
5730 if (upper < Character.MIN_HIGH_SURROGATE ||
5731 lower > Character.MAX_HIGH_SURROGATE &&
5732 upper < Character.MIN_SUPPLEMENTARY_CODE_POINT)
5733 return (BmpCharPredicate)(ch -> inRange(lower, ch, upper));
5734 return ch -> inRange(lower, ch, upper);
5735 }
5736
5737 /**
5738 * Charactrs within a explicit value range in a case insensitive manner.
5739 */
5740 static CharPredicate CIRange(int lower, int upper) {
5741 return ch -> inRange(lower, ch, upper) ||
5742 ASCII.isAscii(ch) &&
5743 (inRange(lower, ASCII.toUpper(ch), upper) ||
5744 inRange(lower, ASCII.toLower(ch), upper));
5745 }
5746
5747 static CharPredicate CIRangeU(int lower, int upper) {
5748 return ch -> {
5749 if (inRange(lower, ch, upper))
5750 return true;
5751 int up = Character.toUpperCase(ch);
5752 return inRange(lower, up, upper) ||
5753 inRange(lower, Character.toLowerCase(up), upper);
5754 };
5755 }
5756
5757 /**
5758 * This must be the very first initializer.
5759 */
5760 static final Node accept = new Node();
5761
5762 static final Node lastAccept = new LastNode();
5763
5764 /**
5765 * Creates a predicate that tests if this pattern is found in a given input
5766 * string.
5767 *
5768 * @apiNote
5769 * This method creates a predicate that behaves as if it creates a matcher
5770 * from the input sequence and then calls {@code find}, for example a
5771 * predicate of the form:
5772 * <pre>{@code
5773 * s -> matcher(s).find();
5774 * }</pre>
5775 *
5776 * @return The predicate which can be used for finding a match on a
5777 * subsequence of a string
5778 * @since 1.8
5779 * @see Matcher#find
5780 */
5781 public Predicate<String> asPredicate() {
5782 return s -> matcher(s).find();
5783 }
5784
5785 /**
5786 * Creates a predicate that tests if this pattern matches a given input string.
5787 *
5788 * @apiNote
5789 * This method creates a predicate that behaves as if it creates a matcher
5790 * from the input sequence and then calls {@code matches}, for example a
5791 * predicate of the form:
5792 * <pre>{@code
5793 * s -> matcher(s).matches();
5794 * }</pre>
5795 *
5796 * @return The predicate which can be used for matching an input string
5797 * against this pattern.
5798 * @since 11
5799 * @see Matcher#matches
5800 */
5801 public Predicate<String> asMatchPredicate() {
5802 return s -> matcher(s).matches();
5803 }
5804
5805 /**
5806 * Creates a stream from the given input sequence around matches of this
5807 * pattern.
5808 *
5809 * <p> The stream returned by this method contains each substring of the
5810 * input sequence that is terminated by another subsequence that matches
5811 * this pattern or is terminated by the end of the input sequence. The
5812 * substrings in the stream are in the order in which they occur in the
5813 * input. Trailing empty strings will be discarded and not encountered in
5814 * the stream.
5815 *
5816 * <p> If this pattern does not match any subsequence of the input then
5817 * the resulting stream has just one element, namely the input sequence in
5818 * string form.
5819 *
5820 * <p> When there is a positive-width match at the beginning of the input
5821 * sequence then an empty leading substring is included at the beginning
5822 * of the stream. A zero-width match at the beginning however never produces
5823 * such empty leading substring.
5824 *
5825 * <p> If the input sequence is mutable, it must remain constant during the
5826 * execution of the terminal stream operation. Otherwise, the result of the
5827 * terminal stream operation is undefined.
5828 *
5829 * @param input
5830 * The character sequence to be split
5831 *
5832 * @return The stream of strings computed by splitting the input
5833 * around matches of this pattern
5834 * @see #split(CharSequence)
5835 * @since 1.8
5836 */
5837 public Stream<String> splitAsStream(final CharSequence input) {
5838 class MatcherIterator implements Iterator<String> {
5839 private Matcher matcher;
5840 // The start position of the next sub-sequence of input
5841 // when current == input.length there are no more elements
5842 private int current;
5843 // null if the next element, if any, needs to obtained
5844 private String nextElement;
5845 // > 0 if there are N next empty elements
5846 private int emptyElementCount;
5847
5848 public String next() {
5849 if (!hasNext())
5850 throw new NoSuchElementException();
5851
5852 if (emptyElementCount == 0) {
5853 String n = nextElement;
5854 nextElement = null;
5855 return n;
5856 } else {
5857 emptyElementCount--;
5858 return "";
5859 }
5860 }
5861
5862 public boolean hasNext() {
5863 if (matcher == null) {
5864 matcher = matcher(input);
5865 // If the input is an empty string then the result can only be a
5866 // stream of the input. Induce that by setting the empty
5867 // element count to 1
5868 emptyElementCount = input.length() == 0 ? 1 : 0;
5869 }
5870 if (nextElement != null || emptyElementCount > 0)
5871 return true;
5872
5873 if (current == input.length())
5874 return false;
5875
5876 // Consume the next matching element
5877 // Count sequence of matching empty elements
5878 while (matcher.find()) {
5879 nextElement = input.subSequence(current, matcher.start()).toString();
5880 current = matcher.end();
5881 if (!nextElement.isEmpty()) {
5882 return true;
5883 } else if (current > 0) { // no empty leading substring for zero-width
5884 // match at the beginning of the input
5885 emptyElementCount++;
5886 }
5887 }
5888
5889 // Consume last matching element
5890 nextElement = input.subSequence(current, input.length()).toString();
5891 current = input.length();
5892 if (!nextElement.isEmpty()) {
5893 return true;
5894 } else {
5895 // Ignore a terminal sequence of matching empty elements
5896 emptyElementCount = 0;
5897 nextElement = null;
5898 return false;
5899 }
5900 }
5901 }
5902 return StreamSupport.stream(Spliterators.spliteratorUnknownSize(
5903 new MatcherIterator(), Spliterator.ORDERED | Spliterator.NONNULL), false);
5904 }
5905 }