1 /*
2  * Copyright (C) 2014 The Android Open Source Project
3  * Copyright (c) 2000, 2011, Oracle and/or its affiliates. All rights reserved.
4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5  *
6  * This code is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License version 2 only, as
8  * published by the Free Software Foundation.  Oracle designates this
9  * particular file as subject to the "Classpath" exception as provided
10  * by Oracle in the LICENSE file that accompanied this code.
11  *
12  * This code is distributed in the hope that it will be useful, but WITHOUT
13  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
14  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
15  * version 2 for more details (a copy is included in the LICENSE file that
16  * accompanied this code).
17  *
18  * You should have received a copy of the GNU General Public License version
19  * 2 along with this work; if not, write to the Free Software Foundation,
20  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
21  *
22  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
23  * or visit www.oracle.com if you need additional information or have any
24  * questions.
25  */
26 
27 package java.net;
28 
29 import java.io.IOException;
30 import java.io.InvalidObjectException;
31 import java.io.ObjectInputStream;
32 import java.io.ObjectOutputStream;
33 import java.io.Serializable;
34 import java.nio.ByteBuffer;
35 import java.nio.CharBuffer;
36 import java.nio.charset.CharsetDecoder;
37 import java.nio.charset.CoderResult;
38 import java.nio.charset.CodingErrorAction;
39 import java.nio.charset.CharacterCodingException;
40 import java.text.Normalizer;
41 import sun.nio.cs.ThreadLocalCoders;
42 
43 import java.lang.Character;             // for javadoc
44 import java.lang.NullPointerException;  // for javadoc
45 
46 
47 /**
48  * Represents a Uniform Resource Identifier (URI) reference.
49  *
50  * <p> Aside from some minor deviations noted below, an instance of this
51  * class represents a URI reference as defined by
52  * <a href="http://www.ietf.org/rfc/rfc2396.txt"><i>RFC&nbsp;2396: Uniform
53  * Resource Identifiers (URI): Generic Syntax</i></a>, amended by <a
54  * href="http://www.ietf.org/rfc/rfc2732.txt"><i>RFC&nbsp;2732: Format for
55  * Literal IPv6 Addresses in URLs</i></a>. The Literal IPv6 address format
56  * also supports scope_ids. The syntax and usage of scope_ids is described
57  * <a href="Inet6Address.html#scoped">here</a>.
58  * This class provides constructors for creating URI instances from
59  * their components or by parsing their string forms, methods for accessing the
60  * various components of an instance, and methods for normalizing, resolving,
61  * and relativizing URI instances.  Instances of this class are immutable.
62  *
63  *
64  * <h4> URI syntax and components </h4>
65  *
66  * At the highest level a URI reference (hereinafter simply "URI") in string
67  * form has the syntax
68  *
69  * <blockquote>
70  * [<i>scheme</i><tt><b>:</b></tt><i></i>]<i>scheme-specific-part</i>[<tt><b>#</b></tt><i>fragment</i>]
71  * </blockquote>
72  *
73  * where square brackets [...] delineate optional components and the characters
74  * <tt><b>:</b></tt> and <tt><b>#</b></tt> stand for themselves.
75  *
76  * <p> An <i>absolute</i> URI specifies a scheme; a URI that is not absolute is
77  * said to be <i>relative</i>.  URIs are also classified according to whether
78  * they are <i>opaque</i> or <i>hierarchical</i>.
79  *
80  * <p> An <i>opaque</i> URI is an absolute URI whose scheme-specific part does
81  * not begin with a slash character (<tt>'/'</tt>).  Opaque URIs are not
82  * subject to further parsing.  Some examples of opaque URIs are:
83  *
84  * <blockquote><table cellpadding=0 cellspacing=0 summary="layout">
85  * <tr><td><tt>mailto:java-net@java.sun.com</tt><td></tr>
86  * <tr><td><tt>news:comp.lang.java</tt><td></tr>
87  * <tr><td><tt>urn:isbn:096139210x</tt></td></tr>
88  * </table></blockquote>
89  *
90  * <p> A <i>hierarchical</i> URI is either an absolute URI whose
91  * scheme-specific part begins with a slash character, or a relative URI, that
92  * is, a URI that does not specify a scheme.  Some examples of hierarchical
93  * URIs are:
94  *
95  * <blockquote>
96  * <tt>http://java.sun.com/j2se/1.3/</tt><br>
97  * <tt>docs/guide/collections/designfaq.html#28</tt><br>
98  * <tt>../../../demo/jfc/SwingSet2/src/SwingSet2.java</tt><br>
99  * <tt>file:///~/calendar</tt>
100  * </blockquote>
101  *
102  * <p> A hierarchical URI is subject to further parsing according to the syntax
103  *
104  * <blockquote>
105  * [<i>scheme</i><tt><b>:</b></tt>][<tt><b>//</b></tt><i>authority</i>][<i>path</i>][<tt><b>?</b></tt><i>query</i>][<tt><b>#</b></tt><i>fragment</i>]
106  * </blockquote>
107  *
108  * where the characters <tt><b>:</b></tt>, <tt><b>/</b></tt>,
109  * <tt><b>?</b></tt>, and <tt><b>#</b></tt> stand for themselves.  The
110  * scheme-specific part of a hierarchical URI consists of the characters
111  * between the scheme and fragment components.
112  *
113  * <p> The authority component of a hierarchical URI is, if specified, either
114  * <i>server-based</i> or <i>registry-based</i>.  A server-based authority
115  * parses according to the familiar syntax
116  *
117  * <blockquote>
118  * [<i>user-info</i><tt><b>@</b></tt>]<i>host</i>[<tt><b>:</b></tt><i>port</i>]
119  * </blockquote>
120  *
121  * where the characters <tt><b>@</b></tt> and <tt><b>:</b></tt> stand for
122  * themselves.  Nearly all URI schemes currently in use are server-based.  An
123  * authority component that does not parse in this way is considered to be
124  * registry-based.
125  *
126  * <p> The path component of a hierarchical URI is itself said to be absolute
127  * if it begins with a slash character (<tt>'/'</tt>); otherwise it is
128  * relative.  The path of a hierarchical URI that is either absolute or
129  * specifies an authority is always absolute.
130  *
131  * <p> All told, then, a URI instance has the following nine components:
132  *
133  * <blockquote><table summary="Describes the components of a URI:scheme,scheme-specific-part,authority,user-info,host,port,path,query,fragment">
134  * <tr><th><i>Component</i></th><th><i>Type</i></th></tr>
135  * <tr><td>scheme</td><td><tt>String</tt></td></tr>
136  * <tr><td>scheme-specific-part&nbsp;&nbsp;&nbsp;&nbsp;</td><td><tt>String</tt></td></tr>
137  * <tr><td>authority</td><td><tt>String</tt></td></tr>
138  * <tr><td>user-info</td><td><tt>String</tt></td></tr>
139  * <tr><td>host</td><td><tt>String</tt></td></tr>
140  * <tr><td>port</td><td><tt>int</tt></td></tr>
141  * <tr><td>path</td><td><tt>String</tt></td></tr>
142  * <tr><td>query</td><td><tt>String</tt></td></tr>
143  * <tr><td>fragment</td><td><tt>String</tt></td></tr>
144  * </table></blockquote>
145  *
146  * In a given instance any particular component is either <i>undefined</i> or
147  * <i>defined</i> with a distinct value.  Undefined string components are
148  * represented by <tt>null</tt>, while undefined integer components are
149  * represented by <tt>-1</tt>.  A string component may be defined to have the
150  * empty string as its value; this is not equivalent to that component being
151  * undefined.
152  *
153  * <p> Whether a particular component is or is not defined in an instance
154  * depends upon the type of the URI being represented.  An absolute URI has a
155  * scheme component.  An opaque URI has a scheme, a scheme-specific part, and
156  * possibly a fragment, but has no other components.  A hierarchical URI always
157  * has a path (though it may be empty) and a scheme-specific-part (which at
158  * least contains the path), and may have any of the other components.  If the
159  * authority component is present and is server-based then the host component
160  * will be defined and the user-information and port components may be defined.
161  *
162  *
163  * <h4> Operations on URI instances </h4>
164  *
165  * The key operations supported by this class are those of
166  * <i>normalization</i>, <i>resolution</i>, and <i>relativization</i>.
167  *
168  * <p> <i>Normalization</i> is the process of removing unnecessary <tt>"."</tt>
169  * and <tt>".."</tt> segments from the path component of a hierarchical URI.
170  * Each <tt>"."</tt> segment is simply removed.  A <tt>".."</tt> segment is
171  * removed only if it is preceded by a non-<tt>".."</tt> segment.
172  * Normalization has no effect upon opaque URIs.
173  *
174  * <p> <i>Resolution</i> is the process of resolving one URI against another,
175  * <i>base</i> URI.  The resulting URI is constructed from components of both
176  * URIs in the manner specified by RFC&nbsp;2396, taking components from the
177  * base URI for those not specified in the original.  For hierarchical URIs,
178  * the path of the original is resolved against the path of the base and then
179  * normalized.  The result, for example, of resolving
180  *
181  * <blockquote>
182  * <tt>docs/guide/collections/designfaq.html#28&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt>(1)
183  * </blockquote>
184  *
185  * against the base URI <tt>http://java.sun.com/j2se/1.3/</tt> is the result
186  * URI
187  *
188  * <blockquote>
189  * <tt>http://java.sun.com/j2se/1.3/docs/guide/collections/designfaq.html#28</tt>
190  * </blockquote>
191  *
192  * Resolving the relative URI
193  *
194  * <blockquote>
195  * <tt>../../../demo/jfc/SwingSet2/src/SwingSet2.java&nbsp;&nbsp;&nbsp;&nbsp;</tt>(2)
196  * </blockquote>
197  *
198  * against this result yields, in turn,
199  *
200  * <blockquote>
201  * <tt>http://java.sun.com/j2se/1.3/demo/jfc/SwingSet2/src/SwingSet2.java</tt>
202  * </blockquote>
203  *
204  * Resolution of both absolute and relative URIs, and of both absolute and
205  * relative paths in the case of hierarchical URIs, is supported.  Resolving
206  * the URI <tt>file:///~calendar</tt> against any other URI simply yields the
207  * original URI, since it is absolute.  Resolving the relative URI (2) above
208  * against the relative base URI (1) yields the normalized, but still relative,
209  * URI
210  *
211  * <blockquote>
212  * <tt>demo/jfc/SwingSet2/src/SwingSet2.java</tt>
213  * </blockquote>
214  *
215  * <p> <i>Relativization</i>, finally, is the inverse of resolution: For any
216  * two normalized URIs <i>u</i> and&nbsp;<i>v</i>,
217  *
218  * <blockquote>
219  *   <i>u</i><tt>.relativize(</tt><i>u</i><tt>.resolve(</tt><i>v</i><tt>)).equals(</tt><i>v</i><tt>)</tt>&nbsp;&nbsp;and<br>
220  *   <i>u</i><tt>.resolve(</tt><i>u</i><tt>.relativize(</tt><i>v</i><tt>)).equals(</tt><i>v</i><tt>)</tt>&nbsp;&nbsp;.<br>
221  * </blockquote>
222  *
223  * This operation is often useful when constructing a document containing URIs
224  * that must be made relative to the base URI of the document wherever
225  * possible.  For example, relativizing the URI
226  *
227  * <blockquote>
228  * <tt>http://java.sun.com/j2se/1.3/docs/guide/index.html</tt>
229  * </blockquote>
230  *
231  * against the base URI
232  *
233  * <blockquote>
234  * <tt>http://java.sun.com/j2se/1.3</tt>
235  * </blockquote>
236  *
237  * yields the relative URI <tt>docs/guide/index.html</tt>.
238  *
239  *
240  * <h4> Character categories </h4>
241  *
242  * RFC&nbsp;2396 specifies precisely which characters are permitted in the
243  * various components of a URI reference.  The following categories, most of
244  * which are taken from that specification, are used below to describe these
245  * constraints:
246  *
247  * <blockquote><table cellspacing=2 summary="Describes categories alpha,digit,alphanum,unreserved,punct,reserved,escaped,and other">
248  *   <tr><th valign=top><i>alpha</i></th>
249  *       <td>The US-ASCII alphabetic characters,
250  *        <tt>'A'</tt>&nbsp;through&nbsp;<tt>'Z'</tt>
251  *        and <tt>'a'</tt>&nbsp;through&nbsp;<tt>'z'</tt></td></tr>
252  *   <tr><th valign=top><i>digit</i></th>
253  *       <td>The US-ASCII decimal digit characters,
254  *       <tt>'0'</tt>&nbsp;through&nbsp;<tt>'9'</tt></td></tr>
255  *   <tr><th valign=top><i>alphanum</i></th>
256  *       <td>All <i>alpha</i> and <i>digit</i> characters</td></tr>
257  *   <tr><th valign=top><i>unreserved</i>&nbsp;&nbsp;&nbsp;&nbsp;</th>
258  *       <td>All <i>alphanum</i> characters together with those in the string
259  *        <tt>"_-!.~'()*"</tt></td></tr>
260  *   <tr><th valign=top><i>punct</i></th>
261  *       <td>The characters in the string <tt>",;:$&+="</tt></td></tr>
262  *   <tr><th valign=top><i>reserved</i></th>
263  *       <td>All <i>punct</i> characters together with those in the string
264  *        <tt>"?/[]@"</tt></td></tr>
265  *   <tr><th valign=top><i>escaped</i></th>
266  *       <td>Escaped octets, that is, triplets consisting of the percent
267  *           character (<tt>'%'</tt>) followed by two hexadecimal digits
268  *           (<tt>'0'</tt>-<tt>'9'</tt>, <tt>'A'</tt>-<tt>'F'</tt>, and
269  *           <tt>'a'</tt>-<tt>'f'</tt>)</td></tr>
270  *   <tr><th valign=top><i>other</i></th>
271  *       <td>The Unicode characters that are not in the US-ASCII character set,
272  *           are not control characters (according to the {@link
273  *           java.lang.Character#isISOControl(char) Character.isISOControl}
274  *           method), and are not space characters (according to the {@link
275  *           java.lang.Character#isSpaceChar(char) Character.isSpaceChar}
276  *           method)&nbsp;&nbsp;<i>(<b>Deviation from RFC 2396</b>, which is
277  *           limited to US-ASCII)</i></td></tr>
278  * </table></blockquote>
279  *
280  * <p><a name="legal-chars"></a> The set of all legal URI characters consists of
281  * the <i>unreserved</i>, <i>reserved</i>, <i>escaped</i>, and <i>other</i>
282  * characters.
283  *
284  *
285  * <h4> Escaped octets, quotation, encoding, and decoding </h4>
286  *
287  * RFC 2396 allows escaped octets to appear in the user-info, path, query, and
288  * fragment components.  Escaping serves two purposes in URIs:
289  *
290  * <ul>
291  *
292  *   <li><p> To <i>encode</i> non-US-ASCII characters when a URI is required to
293  *   conform strictly to RFC&nbsp;2396 by not containing any <i>other</i>
294  *   characters.  </p></li>
295  *
296  *   <li><p> To <i>quote</i> characters that are otherwise illegal in a
297  *   component.  The user-info, path, query, and fragment components differ
298  *   slightly in terms of which characters are considered legal and illegal.
299  *   </p></li>
300  *
301  * </ul>
302  *
303  * These purposes are served in this class by three related operations:
304  *
305  * <ul>
306  *
307  *   <li><p><a name="encode"></a> A character is <i>encoded</i> by replacing it
308  *   with the sequence of escaped octets that represent that character in the
309  *   UTF-8 character set.  The Euro currency symbol (<tt>'&#92;u20AC'</tt>),
310  *   for example, is encoded as <tt>"%E2%82%AC"</tt>.  <i>(<b>Deviation from
311  *   RFC&nbsp;2396</b>, which does not specify any particular character
312  *   set.)</i> </p></li>
313  *
314  *   <li><p><a name="quote"></a> An illegal character is <i>quoted</i> simply by
315  *   encoding it.  The space character, for example, is quoted by replacing it
316  *   with <tt>"%20"</tt>.  UTF-8 contains US-ASCII, hence for US-ASCII
317  *   characters this transformation has exactly the effect required by
318  *   RFC&nbsp;2396. </p></li>
319  *
320  *   <li><p><a name="decode"></a>
321  *   A sequence of escaped octets is <i>decoded</i> by
322  *   replacing it with the sequence of characters that it represents in the
323  *   UTF-8 character set.  UTF-8 contains US-ASCII, hence decoding has the
324  *   effect of de-quoting any quoted US-ASCII characters as well as that of
325  *   decoding any encoded non-US-ASCII characters.  If a <a
326  *   href="../nio/charset/CharsetDecoder.html#ce">decoding error</a> occurs
327  *   when decoding the escaped octets then the erroneous octets are replaced by
328  *   <tt>'&#92;uFFFD'</tt>, the Unicode replacement character.  </p></li>
329  *
330  * </ul>
331  *
332  * These operations are exposed in the constructors and methods of this class
333  * as follows:
334  *
335  * <ul>
336  *
337  *   <li><p> The {@link #URI(java.lang.String) <code>single-argument
338  *   constructor</code>} requires any illegal characters in its argument to be
339  *   quoted and preserves any escaped octets and <i>other</i> characters that
340  *   are present.  </p></li>
341  *
342  *   <li><p> The {@link
343  *   #URI(java.lang.String,java.lang.String,java.lang.String,int,java.lang.String,java.lang.String,java.lang.String)
344  *   <code>multi-argument constructors</code>} quote illegal characters as
345  *   required by the components in which they appear.  The percent character
346  *   (<tt>'%'</tt>) is always quoted by these constructors.  Any <i>other</i>
347  *   characters are preserved.  </p></li>
348  *
349  *   <li><p> The {@link #getRawUserInfo() getRawUserInfo}, {@link #getRawPath()
350  *   getRawPath}, {@link #getRawQuery() getRawQuery}, {@link #getRawFragment()
351  *   getRawFragment}, {@link #getRawAuthority() getRawAuthority}, and {@link
352  *   #getRawSchemeSpecificPart() getRawSchemeSpecificPart} methods return the
353  *   values of their corresponding components in raw form, without interpreting
354  *   any escaped octets.  The strings returned by these methods may contain
355  *   both escaped octets and <i>other</i> characters, and will not contain any
356  *   illegal characters.  </p></li>
357  *
358  *   <li><p> The {@link #getUserInfo() getUserInfo}, {@link #getPath()
359  *   getPath}, {@link #getQuery() getQuery}, {@link #getFragment()
360  *   getFragment}, {@link #getAuthority() getAuthority}, and {@link
361  *   #getSchemeSpecificPart() getSchemeSpecificPart} methods decode any escaped
362  *   octets in their corresponding components.  The strings returned by these
363  *   methods may contain both <i>other</i> characters and illegal characters,
364  *   and will not contain any escaped octets.  </p></li>
365  *
366  *   <li><p> The {@link #toString() toString} method returns a URI string with
367  *   all necessary quotation but which may contain <i>other</i> characters.
368  *   </p></li>
369  *
370  *   <li><p> The {@link #toASCIIString() toASCIIString} method returns a fully
371  *   quoted and encoded URI string that does not contain any <i>other</i>
372  *   characters.  </p></li>
373  *
374  * </ul>
375  *
376  *
377  * <h4> Identities </h4>
378  *
379  * For any URI <i>u</i>, it is always the case that
380  *
381  * <blockquote>
382  * <tt>new URI(</tt><i>u</i><tt>.toString()).equals(</tt><i>u</i><tt>)</tt>&nbsp;.
383  * </blockquote>
384  *
385  * For any URI <i>u</i> that does not contain redundant syntax such as two
386  * slashes before an empty authority (as in <tt>file:///tmp/</tt>&nbsp;) or a
387  * colon following a host name but no port (as in
388  * <tt>http://java.sun.com:</tt>&nbsp;), and that does not encode characters
389  * except those that must be quoted, the following identities also hold:
390  *
391  * <blockquote>
392  * <tt>new URI(</tt><i>u</i><tt>.getScheme(),<br>
393  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getSchemeSpecificPart(),<br>
394  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getFragment())<br>
395  * .equals(</tt><i>u</i><tt>)</tt>
396  * </blockquote>
397  *
398  * in all cases,
399  *
400  * <blockquote>
401  * <tt>new URI(</tt><i>u</i><tt>.getScheme(),<br>
402  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getUserInfo(),&nbsp;</tt><i>u</i><tt>.getAuthority(),<br>
403  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getPath(),&nbsp;</tt><i>u</i><tt>.getQuery(),<br>
404  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getFragment())<br>
405  * .equals(</tt><i>u</i><tt>)</tt>
406  * </blockquote>
407  *
408  * if <i>u</i> is hierarchical, and
409  *
410  * <blockquote>
411  * <tt>new URI(</tt><i>u</i><tt>.getScheme(),<br>
412  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getUserInfo(),&nbsp;</tt><i>u</i><tt>.getHost(),&nbsp;</tt><i>u</i><tt>.getPort(),<br>
413  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getPath(),&nbsp;</tt><i>u</i><tt>.getQuery(),<br>
414  * &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</tt><i>u</i><tt>.getFragment())<br>
415  * .equals(</tt><i>u</i><tt>)</tt>
416  * </blockquote>
417  *
418  * if <i>u</i> is hierarchical and has either no authority or a server-based
419  * authority.
420  *
421  *
422  * <h4> URIs, URLs, and URNs </h4>
423  *
424  * A URI is a uniform resource <i>identifier</i> while a URL is a uniform
425  * resource <i>locator</i>.  Hence every URL is a URI, abstractly speaking, but
426  * not every URI is a URL.  This is because there is another subcategory of
427  * URIs, uniform resource <i>names</i> (URNs), which name resources but do not
428  * specify how to locate them.  The <tt>mailto</tt>, <tt>news</tt>, and
429  * <tt>isbn</tt> URIs shown above are examples of URNs.
430  *
431  * <p> The conceptual distinction between URIs and URLs is reflected in the
432  * differences between this class and the {@link URL} class.
433  *
434  * <p> An instance of this class represents a URI reference in the syntactic
435  * sense defined by RFC&nbsp;2396.  A URI may be either absolute or relative.
436  * A URI string is parsed according to the generic syntax without regard to the
437  * scheme, if any, that it specifies.  No lookup of the host, if any, is
438  * performed, and no scheme-dependent stream handler is constructed.  Equality,
439  * hashing, and comparison are defined strictly in terms of the character
440  * content of the instance.  In other words, a URI instance is little more than
441  * a structured string that supports the syntactic, scheme-independent
442  * operations of comparison, normalization, resolution, and relativization.
443  *
444  * <p> An instance of the {@link URL} class, by contrast, represents the
445  * syntactic components of a URL together with some of the information required
446  * to access the resource that it describes.  A URL must be absolute, that is,
447  * it must always specify a scheme.  A URL string is parsed according to its
448  * scheme.  A stream handler is always established for a URL, and in fact it is
449  * impossible to create a URL instance for a scheme for which no handler is
450  * available.  Equality and hashing depend upon both the scheme and the
451  * Internet address of the host, if any; comparison is not defined.  In other
452  * words, a URL is a structured string that supports the syntactic operation of
453  * resolution as well as the network I/O operations of looking up the host and
454  * opening a connection to the specified resource.
455  *
456  *
457  * @author Mark Reinhold
458  * @since 1.4
459  *
460  * @see <a href="http://www.ietf.org/rfc/rfc2279.txt">RFC&nbsp;2279: UTF-8, a transformation format of ISO 10646</a>
461  * @see <a href="http://www.ietf.org/rfc/rfc2373.txt">RFC&nbsp;2373: IPv6 Addressing Architecture</a>
462  * @see <a href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396: Uniform Resource Identifiers (URI): Generic Syntax</a>
463  * @see <a href="http://www.ietf.org/rfc/rfc2732.txt">RFC&nbsp;2732: Format for Literal IPv6 Addresses in URLs</a>
464  */
465 // Android changed: Reformat @see links.
466 public final class URI
467     implements Comparable<URI>, Serializable
468 {
469     // Note: Comments containing the word "ASSERT" indicate places where a
470     // throw of an InternalError should be replaced by an appropriate assertion
471     // statement once asserts are enabled in the build.
472 
473     static final long serialVersionUID = -6052424284110960213L;
474 
475 
476     // -- Properties and components of this instance --
477 
478     // Components of all URIs: [<scheme>:]<scheme-specific-part>[#<fragment>]
479     private transient String scheme;            // null ==> relative URI
480     private transient String fragment;
481 
482     // Hierarchical URI components: [//<authority>]<path>[?<query>]
483     private transient String authority;         // Registry or server
484 
485     // Server-based authority: [<userInfo>@]<host>[:<port>]
486     private transient String userInfo;
487     private transient String host;              // null ==> registry-based
488     private transient int port = -1;            // -1 ==> undefined
489 
490     // Remaining components of hierarchical URIs
491     private transient String path;              // null ==> opaque
492     private transient String query;
493 
494     // The remaining fields may be computed on demand
495 
496     private volatile transient String schemeSpecificPart;
497     private volatile transient int hash;        // Zero ==> undefined
498 
499     private volatile transient String decodedUserInfo = null;
500     private volatile transient String decodedAuthority = null;
501     private volatile transient String decodedPath = null;
502     private volatile transient String decodedQuery = null;
503     private volatile transient String decodedFragment = null;
504     private volatile transient String decodedSchemeSpecificPart = null;
505 
506     /**
507      * The string form of this URI.
508      *
509      * @serial
510      */
511     private volatile String string;             // The only serializable field
512 
513 
514 
515     // -- Constructors and factories --
516 
URI()517     private URI() { }                           // Used internally
518 
519     /**
520      * Constructs a URI by parsing the given string.
521      *
522      * <p> This constructor parses the given string exactly as specified by the
523      * grammar in <a
524      * href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>,
525      * Appendix&nbsp;A, <b><i>except for the following deviations:</i></b> </p>
526      *
527      * <ul type=disc>
528      *
529      *   <li><p> An empty authority component is permitted as long as it is
530      *   followed by a non-empty path, a query component, or a fragment
531      *   component.  This allows the parsing of URIs such as
532      *   <tt>"file:///foo/bar"</tt>, which seems to be the intent of
533      *   RFC&nbsp;2396 although the grammar does not permit it.  If the
534      *   authority component is empty then the user-information, host, and port
535      *   components are undefined. </p></li>
536      *
537      *   <li><p> Empty relative paths are permitted; this seems to be the
538      *   intent of RFC&nbsp;2396 although the grammar does not permit it.  The
539      *   primary consequence of this deviation is that a standalone fragment
540      *   such as <tt>"#foo"</tt> parses as a relative URI with an empty path
541      *   and the given fragment, and can be usefully <a
542      *   href="#resolve-frag">resolved</a> against a base URI.
543      *
544      *   <li><p> IPv4 addresses in host components are parsed rigorously, as
545      *   specified by <a
546      *   href="http://www.ietf.org/rfc/rfc2732.txt">RFC&nbsp;2732</a>: Each
547      *   element of a dotted-quad address must contain no more than three
548      *   decimal digits.  Each element is further constrained to have a value
549      *   no greater than 255. </p></li>
550      *
551      *   <li> <p> Hostnames in host components that comprise only a single
552      *   domain label are permitted to start with an <i>alphanum</i>
553      *   character. This seems to be the intent of <a
554      *   href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>
555      *   section&nbsp;3.2.2 although the grammar does not permit it. The
556      *   consequence of this deviation is that the authority component of a
557      *   hierarchical URI such as <tt>s://123</tt>, will parse as a server-based
558      *   authority. </p></li>
559      *
560      *   <li><p> IPv6 addresses are permitted for the host component.  An IPv6
561      *   address must be enclosed in square brackets (<tt>'['</tt> and
562      *   <tt>']'</tt>) as specified by <a
563      *   href="http://www.ietf.org/rfc/rfc2732.txt">RFC&nbsp;2732</a>.  The
564      *   IPv6 address itself must parse according to <a
565      *   href="http://www.ietf.org/rfc/rfc2373.txt">RFC&nbsp;2373</a>.  IPv6
566      *   addresses are further constrained to describe no more than sixteen
567      *   bytes of address information, a constraint implicit in RFC&nbsp;2373
568      *   but not expressible in the grammar. </p></li>
569      *
570      *   <li><p> Characters in the <i>other</i> category are permitted wherever
571      *   RFC&nbsp;2396 permits <i>escaped</i> octets, that is, in the
572      *   user-information, path, query, and fragment components, as well as in
573      *   the authority component if the authority is registry-based.  This
574      *   allows URIs to contain Unicode characters beyond those in the US-ASCII
575      *   character set. </p></li>
576      *
577      * </ul>
578      *
579      * @param  str   The string to be parsed into a URI
580      *
581      * @throws  NullPointerException
582      *          If <tt>str</tt> is <tt>null</tt>
583      *
584      * @throws  URISyntaxException
585      *          If the given string violates RFC&nbsp;2396, as augmented
586      *          by the above deviations
587      */
URI(String str)588     public URI(String str) throws URISyntaxException {
589         new Parser(str).parse(false);
590     }
591 
592     /**
593      * Constructs a hierarchical URI from the given components.
594      *
595      * <p> If a scheme is given then the path, if also given, must either be
596      * empty or begin with a slash character (<tt>'/'</tt>).  Otherwise a
597      * component of the new URI may be left undefined by passing <tt>null</tt>
598      * for the corresponding parameter or, in the case of the <tt>port</tt>
599      * parameter, by passing <tt>-1</tt>.
600      *
601      * <p> This constructor first builds a URI string from the given components
602      * according to the rules specified in <a
603      * href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>,
604      * section&nbsp;5.2, step&nbsp;7: </p>
605      *
606      * <ol>
607      *
608      *   <li><p> Initially, the result string is empty. </p></li>
609      *
610      *   <li><p> If a scheme is given then it is appended to the result,
611      *   followed by a colon character (<tt>':'</tt>).  </p></li>
612      *
613      *   <li><p> If user information, a host, or a port are given then the
614      *   string <tt>"//"</tt> is appended.  </p></li>
615      *
616      *   <li><p> If user information is given then it is appended, followed by
617      *   a commercial-at character (<tt>'@'</tt>).  Any character not in the
618      *   <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
619      *   categories is <a href="#quote">quoted</a>.  </p></li>
620      *
621      *   <li><p> If a host is given then it is appended.  If the host is a
622      *   literal IPv6 address but is not enclosed in square brackets
623      *   (<tt>'['</tt> and <tt>']'</tt>) then the square brackets are added.
624      *   </p></li>
625      *
626      *   <li><p> If a port number is given then a colon character
627      *   (<tt>':'</tt>) is appended, followed by the port number in decimal.
628      *   </p></li>
629      *
630      *   <li><p> If a path is given then it is appended.  Any character not in
631      *   the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
632      *   categories, and not equal to the slash character (<tt>'/'</tt>) or the
633      *   commercial-at character (<tt>'@'</tt>), is quoted.  </p></li>
634      *
635      *   <li><p> If a query is given then a question-mark character
636      *   (<tt>'?'</tt>) is appended, followed by the query.  Any character that
637      *   is not a <a href="#legal-chars">legal URI character</a> is quoted.
638      *   </p></li>
639      *
640      *   <li><p> Finally, if a fragment is given then a hash character
641      *   (<tt>'#'</tt>) is appended, followed by the fragment.  Any character
642      *   that is not a legal URI character is quoted.  </p></li>
643      *
644      * </ol>
645      *
646      * <p> The resulting URI string is then parsed as if by invoking the {@link
647      * #URI(String)} constructor and then invoking the {@link
648      * #parseServerAuthority()} method upon the result; this may cause a {@link
649      * URISyntaxException} to be thrown.  </p>
650      *
651      * @param   scheme    Scheme name
652      * @param   userInfo  User name and authorization information
653      * @param   host      Host name
654      * @param   port      Port number
655      * @param   path      Path
656      * @param   query     Query
657      * @param   fragment  Fragment
658      *
659      * @throws URISyntaxException
660      *         If both a scheme and a path are given but the path is relative,
661      *         if the URI string constructed from the given components violates
662      *         RFC&nbsp;2396, or if the authority component of the string is
663      *         present but cannot be parsed as a server-based authority
664      */
URI(String scheme, String userInfo, String host, int port, String path, String query, String fragment)665     public URI(String scheme,
666                String userInfo, String host, int port,
667                String path, String query, String fragment)
668         throws URISyntaxException
669     {
670         String s = toString(scheme, null,
671                             null, userInfo, host, port,
672                             path, query, fragment);
673         checkPath(s, scheme, path);
674         new Parser(s).parse(true);
675     }
676 
677     /**
678      * Constructs a hierarchical URI from the given components.
679      *
680      * <p> If a scheme is given then the path, if also given, must either be
681      * empty or begin with a slash character (<tt>'/'</tt>).  Otherwise a
682      * component of the new URI may be left undefined by passing <tt>null</tt>
683      * for the corresponding parameter.
684      *
685      * <p> This constructor first builds a URI string from the given components
686      * according to the rules specified in <a
687      * href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>,
688      * section&nbsp;5.2, step&nbsp;7: </p>
689      *
690      * <ol>
691      *
692      *   <li><p> Initially, the result string is empty.  </p></li>
693      *
694      *   <li><p> If a scheme is given then it is appended to the result,
695      *   followed by a colon character (<tt>':'</tt>).  </p></li>
696      *
697      *   <li><p> If an authority is given then the string <tt>"//"</tt> is
698      *   appended, followed by the authority.  If the authority contains a
699      *   literal IPv6 address then the address must be enclosed in square
700      *   brackets (<tt>'['</tt> and <tt>']'</tt>).  Any character not in the
701      *   <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
702      *   categories, and not equal to the commercial-at character
703      *   (<tt>'@'</tt>), is <a href="#quote">quoted</a>.  </p></li>
704      *
705      *   <li><p> If a path is given then it is appended.  Any character not in
706      *   the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
707      *   categories, and not equal to the slash character (<tt>'/'</tt>) or the
708      *   commercial-at character (<tt>'@'</tt>), is quoted.  </p></li>
709      *
710      *   <li><p> If a query is given then a question-mark character
711      *   (<tt>'?'</tt>) is appended, followed by the query.  Any character that
712      *   is not a <a href="#legal-chars">legal URI character</a> is quoted.
713      *   </p></li>
714      *
715      *   <li><p> Finally, if a fragment is given then a hash character
716      *   (<tt>'#'</tt>) is appended, followed by the fragment.  Any character
717      *   that is not a legal URI character is quoted.  </p></li>
718      *
719      * </ol>
720      *
721      * <p> The resulting URI string is then parsed as if by invoking the {@link
722      * #URI(String)} constructor and then invoking the {@link
723      * #parseServerAuthority()} method upon the result; this may cause a {@link
724      * URISyntaxException} to be thrown.  </p>
725      *
726      * @param   scheme     Scheme name
727      * @param   authority  Authority
728      * @param   path       Path
729      * @param   query      Query
730      * @param   fragment   Fragment
731      *
732      * @throws URISyntaxException
733      *         If both a scheme and a path are given but the path is relative,
734      *         if the URI string constructed from the given components violates
735      *         RFC&nbsp;2396, or if the authority component of the string is
736      *         present but cannot be parsed as a server-based authority
737      */
URI(String scheme, String authority, String path, String query, String fragment)738     public URI(String scheme,
739                String authority,
740                String path, String query, String fragment)
741         throws URISyntaxException
742     {
743         String s = toString(scheme, null,
744                             authority, null, null, -1,
745                             path, query, fragment);
746         checkPath(s, scheme, path);
747         new Parser(s).parse(false);
748     }
749 
750     /**
751      * Constructs a hierarchical URI from the given components.
752      *
753      * <p> A component may be left undefined by passing <tt>null</tt>.
754      *
755      * <p> This convenience constructor works as if by invoking the
756      * seven-argument constructor as follows:
757      *
758      * <blockquote><tt>
759      * new&nbsp;{@link #URI(String, String, String, int, String, String, String)
760      * URI}(scheme,&nbsp;null,&nbsp;host,&nbsp;-1,&nbsp;path,&nbsp;null,&nbsp;fragment);
761      * </tt></blockquote>
762      *
763      * @param   scheme    Scheme name
764      * @param   host      Host name
765      * @param   path      Path
766      * @param   fragment  Fragment
767      *
768      * @throws  URISyntaxException
769      *          If the URI string constructed from the given components
770      *          violates RFC&nbsp;2396
771      */
URI(String scheme, String host, String path, String fragment)772     public URI(String scheme, String host, String path, String fragment)
773         throws URISyntaxException
774     {
775         this(scheme, null, host, -1, path, null, fragment);
776     }
777 
778     /**
779      * Constructs a URI from the given components.
780      *
781      * <p> A component may be left undefined by passing <tt>null</tt>.
782      *
783      * <p> This constructor first builds a URI in string form using the given
784      * components as follows:  </p>
785      *
786      * <ol>
787      *
788      *   <li><p> Initially, the result string is empty.  </p></li>
789      *
790      *   <li><p> If a scheme is given then it is appended to the result,
791      *   followed by a colon character (<tt>':'</tt>).  </p></li>
792      *
793      *   <li><p> If a scheme-specific part is given then it is appended.  Any
794      *   character that is not a <a href="#legal-chars">legal URI character</a>
795      *   is <a href="#quote">quoted</a>.  </p></li>
796      *
797      *   <li><p> Finally, if a fragment is given then a hash character
798      *   (<tt>'#'</tt>) is appended to the string, followed by the fragment.
799      *   Any character that is not a legal URI character is quoted.  </p></li>
800      *
801      * </ol>
802      *
803      * <p> The resulting URI string is then parsed in order to create the new
804      * URI instance as if by invoking the {@link #URI(String)} constructor;
805      * this may cause a {@link URISyntaxException} to be thrown.  </p>
806      *
807      * @param   scheme    Scheme name
808      * @param   ssp       Scheme-specific part
809      * @param   fragment  Fragment
810      *
811      * @throws  URISyntaxException
812      *          If the URI string constructed from the given components
813      *          violates RFC&nbsp;2396
814      */
URI(String scheme, String ssp, String fragment)815     public URI(String scheme, String ssp, String fragment)
816         throws URISyntaxException
817     {
818         new Parser(toString(scheme, ssp,
819                             null, null, null, -1,
820                             null, null, fragment))
821             .parse(false);
822     }
823 
824     /**
825      * Creates a URI by parsing the given string.
826      *
827      * <p> This convenience factory method works as if by invoking the {@link
828      * #URI(String)} constructor; any {@link URISyntaxException} thrown by the
829      * constructor is caught and wrapped in a new {@link
830      * IllegalArgumentException} object, which is then thrown.
831      *
832      * <p> This method is provided for use in situations where it is known that
833      * the given string is a legal URI, for example for URI constants declared
834      * within in a program, and so it would be considered a programming error
835      * for the string not to parse as such.  The constructors, which throw
836      * {@link URISyntaxException} directly, should be used situations where a
837      * URI is being constructed from user input or from some other source that
838      * may be prone to errors.  </p>
839      *
840      * @param  str   The string to be parsed into a URI
841      * @return The new URI
842      *
843      * @throws  NullPointerException
844      *          If <tt>str</tt> is <tt>null</tt>
845      *
846      * @throws  IllegalArgumentException
847      *          If the given string violates RFC&nbsp;2396
848      */
create(String str)849     public static URI create(String str) {
850         try {
851             return new URI(str);
852         } catch (URISyntaxException x) {
853             throw new IllegalArgumentException(x.getMessage(), x);
854         }
855     }
856 
857 
858     // -- Operations --
859 
860     /**
861      * Attempts to parse this URI's authority component, if defined, into
862      * user-information, host, and port components.
863      *
864      * <p> If this URI's authority component has already been recognized as
865      * being server-based then it will already have been parsed into
866      * user-information, host, and port components.  In this case, or if this
867      * URI has no authority component, this method simply returns this URI.
868      *
869      * <p> Otherwise this method attempts once more to parse the authority
870      * component into user-information, host, and port components, and throws
871      * an exception describing why the authority component could not be parsed
872      * in that way.
873      *
874      * <p> This method is provided because the generic URI syntax specified in
875      * <a href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>
876      * cannot always distinguish a malformed server-based authority from a
877      * legitimate registry-based authority.  It must therefore treat some
878      * instances of the former as instances of the latter.  The authority
879      * component in the URI string <tt>"//foo:bar"</tt>, for example, is not a
880      * legal server-based authority but it is legal as a registry-based
881      * authority.
882      *
883      * <p> In many common situations, for example when working URIs that are
884      * known to be either URNs or URLs, the hierarchical URIs being used will
885      * always be server-based.  They therefore must either be parsed as such or
886      * treated as an error.  In these cases a statement such as
887      *
888      * <blockquote>
889      * <tt>URI </tt><i>u</i><tt> = new URI(str).parseServerAuthority();</tt>
890      * </blockquote>
891      *
892      * <p> can be used to ensure that <i>u</i> always refers to a URI that, if
893      * it has an authority component, has a server-based authority with proper
894      * user-information, host, and port components.  Invoking this method also
895      * ensures that if the authority could not be parsed in that way then an
896      * appropriate diagnostic message can be issued based upon the exception
897      * that is thrown. </p>
898      *
899      * @return  A URI whose authority field has been parsed
900      *          as a server-based authority
901      *
902      * @throws  URISyntaxException
903      *          If the authority component of this URI is defined
904      *          but cannot be parsed as a server-based authority
905      *          according to RFC&nbsp;2396
906      */
parseServerAuthority()907     public URI parseServerAuthority()
908         throws URISyntaxException
909     {
910         // We could be clever and cache the error message and index from the
911         // exception thrown during the original parse, but that would require
912         // either more fields or a more-obscure representation.
913         if ((host != null) || (authority == null))
914             return this;
915         defineString();
916         new Parser(string).parse(true);
917         return this;
918     }
919 
920     /**
921      * Normalizes this URI's path.
922      *
923      * <p> If this URI is opaque, or if its path is already in normal form,
924      * then this URI is returned.  Otherwise a new URI is constructed that is
925      * identical to this URI except that its path is computed by normalizing
926      * this URI's path in a manner consistent with <a
927      * href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>,
928      * section&nbsp;5.2, step&nbsp;6, sub-steps&nbsp;c through&nbsp;f; that is:
929      * </p>
930      *
931      * <ol>
932      *
933      *   <li><p> All <tt>"."</tt> segments are removed. </p></li>
934      *
935      *   <li><p> If a <tt>".."</tt> segment is preceded by a non-<tt>".."</tt>
936      *   segment then both of these segments are removed.  This step is
937      *   repeated until it is no longer applicable. </p></li>
938      *
939      *   <li><p> If the path is relative, and if its first segment contains a
940      *   colon character (<tt>':'</tt>), then a <tt>"."</tt> segment is
941      *   prepended.  This prevents a relative URI with a path such as
942      *   <tt>"a:b/c/d"</tt> from later being re-parsed as an opaque URI with a
943      *   scheme of <tt>"a"</tt> and a scheme-specific part of <tt>"b/c/d"</tt>.
944      *   <b><i>(Deviation from RFC&nbsp;2396)</i></b> </p></li>
945      *
946      * </ol>
947      *
948      * <p> A normalized path will begin with one or more <tt>".."</tt> segments
949      * if there were insufficient non-<tt>".."</tt> segments preceding them to
950      * allow their removal.  A normalized path will begin with a <tt>"."</tt>
951      * segment if one was inserted by step 3 above.  Otherwise, a normalized
952      * path will not contain any <tt>"."</tt> or <tt>".."</tt> segments. </p>
953      *
954      * @return  A URI equivalent to this URI,
955      *          but whose path is in normal form
956      */
normalize()957     public URI normalize() {
958         return normalize(this);
959     }
960 
961     /**
962      * Resolves the given URI against this URI.
963      *
964      * <p> If the given URI is already absolute, or if this URI is opaque, then
965      * the given URI is returned.
966      *
967      * <p><a name="resolve-frag"></a> If the given URI's fragment component is
968      * defined, its path component is empty, and its scheme, authority, and
969      * query components are undefined, then a URI with the given fragment but
970      * with all other components equal to those of this URI is returned.  This
971      * allows a URI representing a standalone fragment reference, such as
972      * <tt>"#foo"</tt>, to be usefully resolved against a base URI.
973      *
974      * <p> Otherwise this method constructs a new hierarchical URI in a manner
975      * consistent with <a
976      * href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>,
977      * section&nbsp;5.2; that is: </p>
978      *
979      * <ol>
980      *
981      *   <li><p> A new URI is constructed with this URI's scheme and the given
982      *   URI's query and fragment components. </p></li>
983      *
984      *   <li><p> If the given URI has an authority component then the new URI's
985      *   authority and path are taken from the given URI. </p></li>
986      *
987      *   <li><p> Otherwise the new URI's authority component is copied from
988      *   this URI, and its path is computed as follows: </p>
989      *
990      *   <ol type=a>
991      *
992      *     <li><p> If the given URI's path is absolute then the new URI's path
993      *     is taken from the given URI. </p></li>
994      *
995      *     <li><p> Otherwise the given URI's path is relative, and so the new
996      *     URI's path is computed by resolving the path of the given URI
997      *     against the path of this URI.  This is done by concatenating all but
998      *     the last segment of this URI's path, if any, with the given URI's
999      *     path and then normalizing the result as if by invoking the {@link
1000      *     #normalize() normalize} method. </p></li>
1001      *
1002      *   </ol></li>
1003      *
1004      * </ol>
1005      *
1006      * <p> The result of this method is absolute if, and only if, either this
1007      * URI is absolute or the given URI is absolute.  </p>
1008      *
1009      * @param  uri  The URI to be resolved against this URI
1010      * @return The resulting URI
1011      *
1012      * @throws  NullPointerException
1013      *          If <tt>uri</tt> is <tt>null</tt>
1014      */
resolve(URI uri)1015     public URI resolve(URI uri) {
1016         return resolve(this, uri);
1017     }
1018 
1019     /**
1020      * Constructs a new URI by parsing the given string and then resolving it
1021      * against this URI.
1022      *
1023      * <p> This convenience method works as if invoking it were equivalent to
1024      * evaluating the expression <tt>{@link #resolve(java.net.URI)
1025      * resolve}(URI.{@link #create(String) create}(str))</tt>. </p>
1026      *
1027      * @param  str   The string to be parsed into a URI
1028      * @return The resulting URI
1029      *
1030      * @throws  NullPointerException
1031      *          If <tt>str</tt> is <tt>null</tt>
1032      *
1033      * @throws  IllegalArgumentException
1034      *          If the given string violates RFC&nbsp;2396
1035      */
resolve(String str)1036     public URI resolve(String str) {
1037         return resolve(URI.create(str));
1038     }
1039 
1040     /**
1041      * Relativizes the given URI against this URI.
1042      *
1043      * <p> The relativization of the given URI against this URI is computed as
1044      * follows: </p>
1045      *
1046      * <ol>
1047      *
1048      *   <li><p> If either this URI or the given URI are opaque, or if the
1049      *   scheme and authority components of the two URIs are not identical, or
1050      *   if the path of this URI is not a prefix of the path of the given URI,
1051      *   then the given URI is returned. </p></li>
1052      *
1053      *   <li><p> Otherwise a new relative hierarchical URI is constructed with
1054      *   query and fragment components taken from the given URI and with a path
1055      *   component computed by removing this URI's path from the beginning of
1056      *   the given URI's path. </p></li>
1057      *
1058      * </ol>
1059      *
1060      * @param  uri  The URI to be relativized against this URI
1061      * @return The resulting URI
1062      *
1063      * @throws  NullPointerException
1064      *          If <tt>uri</tt> is <tt>null</tt>
1065      */
relativize(URI uri)1066     public URI relativize(URI uri) {
1067         return relativize(this, uri);
1068     }
1069 
1070     /**
1071      * Constructs a URL from this URI.
1072      *
1073      * <p> This convenience method works as if invoking it were equivalent to
1074      * evaluating the expression <tt>new&nbsp;URL(this.toString())</tt> after
1075      * first checking that this URI is absolute. </p>
1076      *
1077      * @return  A URL constructed from this URI
1078      *
1079      * @throws  IllegalArgumentException
1080      *          If this URL is not absolute
1081      *
1082      * @throws  MalformedURLException
1083      *          If a protocol handler for the URL could not be found,
1084      *          or if some other error occurred while constructing the URL
1085      */
toURL()1086     public URL toURL()
1087         throws MalformedURLException {
1088         if (!isAbsolute())
1089             throw new IllegalArgumentException("URI is not absolute");
1090         return new URL(toString());
1091     }
1092 
1093     // -- Component access methods --
1094 
1095     /**
1096      * Returns the scheme component of this URI.
1097      *
1098      * <p> The scheme component of a URI, if defined, only contains characters
1099      * in the <i>alphanum</i> category and in the string <tt>"-.+"</tt>.  A
1100      * scheme always starts with an <i>alpha</i> character. <p>
1101      *
1102      * The scheme component of a URI cannot contain escaped octets, hence this
1103      * method does not perform any decoding.
1104      *
1105      * @return  The scheme component of this URI,
1106      *          or <tt>null</tt> if the scheme is undefined
1107      */
getScheme()1108     public String getScheme() {
1109         return scheme;
1110     }
1111 
1112     /**
1113      * Tells whether or not this URI is absolute.
1114      *
1115      * <p> A URI is absolute if, and only if, it has a scheme component. </p>
1116      *
1117      * @return  <tt>true</tt> if, and only if, this URI is absolute
1118      */
isAbsolute()1119     public boolean isAbsolute() {
1120         return scheme != null;
1121     }
1122 
1123     /**
1124      * Tells whether or not this URI is opaque.
1125      *
1126      * <p> A URI is opaque if, and only if, it is absolute and its
1127      * scheme-specific part does not begin with a slash character ('/').
1128      * An opaque URI has a scheme, a scheme-specific part, and possibly
1129      * a fragment; all other components are undefined. </p>
1130      *
1131      * @return  <tt>true</tt> if, and only if, this URI is opaque
1132      */
isOpaque()1133     public boolean isOpaque() {
1134         return path == null;
1135     }
1136 
1137     /**
1138      * Returns the raw scheme-specific part of this URI.  The scheme-specific
1139      * part is never undefined, though it may be empty.
1140      *
1141      * <p> The scheme-specific part of a URI only contains legal URI
1142      * characters. </p>
1143      *
1144      * @return  The raw scheme-specific part of this URI
1145      *          (never <tt>null</tt>)
1146      */
getRawSchemeSpecificPart()1147     public String getRawSchemeSpecificPart() {
1148         defineSchemeSpecificPart();
1149         return schemeSpecificPart;
1150     }
1151 
1152     /**
1153      * Returns the decoded scheme-specific part of this URI.
1154      *
1155      * <p> The string returned by this method is equal to that returned by the
1156      * {@link #getRawSchemeSpecificPart() getRawSchemeSpecificPart} method
1157      * except that all sequences of escaped octets are <a
1158      * href="#decode">decoded</a>.  </p>
1159      *
1160      * @return  The decoded scheme-specific part of this URI
1161      *          (never <tt>null</tt>)
1162      */
getSchemeSpecificPart()1163     public String getSchemeSpecificPart() {
1164         if (decodedSchemeSpecificPart == null)
1165             decodedSchemeSpecificPart = decode(getRawSchemeSpecificPart());
1166         return decodedSchemeSpecificPart;
1167     }
1168 
1169     /**
1170      * Returns the raw authority component of this URI.
1171      *
1172      * <p> The authority component of a URI, if defined, only contains the
1173      * commercial-at character (<tt>'@'</tt>) and characters in the
1174      * <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, and <i>other</i>
1175      * categories.  If the authority is server-based then it is further
1176      * constrained to have valid user-information, host, and port
1177      * components. </p>
1178      *
1179      * @return  The raw authority component of this URI,
1180      *          or <tt>null</tt> if the authority is undefined
1181      */
getRawAuthority()1182     public String getRawAuthority() {
1183         return authority;
1184     }
1185 
1186     /**
1187      * Returns the decoded authority component of this URI.
1188      *
1189      * <p> The string returned by this method is equal to that returned by the
1190      * {@link #getRawAuthority() getRawAuthority} method except that all
1191      * sequences of escaped octets are <a href="#decode">decoded</a>.  </p>
1192      *
1193      * @return  The decoded authority component of this URI,
1194      *          or <tt>null</tt> if the authority is undefined
1195      */
getAuthority()1196     public String getAuthority() {
1197         if (decodedAuthority == null)
1198             decodedAuthority = decode(authority);
1199         return decodedAuthority;
1200     }
1201 
1202     /**
1203      * Returns the raw user-information component of this URI.
1204      *
1205      * <p> The user-information component of a URI, if defined, only contains
1206      * characters in the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, and
1207      * <i>other</i> categories. </p>
1208      *
1209      * @return  The raw user-information component of this URI,
1210      *          or <tt>null</tt> if the user information is undefined
1211      */
getRawUserInfo()1212     public String getRawUserInfo() {
1213         return userInfo;
1214     }
1215 
1216     /**
1217      * Returns the decoded user-information component of this URI.
1218      *
1219      * <p> The string returned by this method is equal to that returned by the
1220      * {@link #getRawUserInfo() getRawUserInfo} method except that all
1221      * sequences of escaped octets are <a href="#decode">decoded</a>.  </p>
1222      *
1223      * @return  The decoded user-information component of this URI,
1224      *          or <tt>null</tt> if the user information is undefined
1225      */
getUserInfo()1226     public String getUserInfo() {
1227         if ((decodedUserInfo == null) && (userInfo != null))
1228             decodedUserInfo = decode(userInfo);
1229         return decodedUserInfo;
1230     }
1231 
1232     /**
1233      * Returns the host component of this URI.
1234      *
1235      * <p> The host component of a URI, if defined, will have one of the
1236      * following forms: </p>
1237      *
1238      * <ul type=disc>
1239      *
1240      *   <li><p> A domain name consisting of one or more <i>labels</i>
1241      *   separated by period characters (<tt>'.'</tt>), optionally followed by
1242      *   a period character.  Each label consists of <i>alphanum</i> characters
1243      *   as well as hyphen characters (<tt>'-'</tt>), though hyphens never
1244      *   occur as the first or last characters in a label. The rightmost
1245      *   label of a domain name consisting of two or more labels, begins
1246      *   with an <i>alpha</i> character. </li>
1247      *
1248      *   <li><p> A dotted-quad IPv4 address of the form
1249      *   <i>digit</i><tt>+.</tt><i>digit</i><tt>+.</tt><i>digit</i><tt>+.</tt><i>digit</i><tt>+</tt>,
1250      *   where no <i>digit</i> sequence is longer than three characters and no
1251      *   sequence has a value larger than 255. </p></li>
1252      *
1253      *   <li><p> An IPv6 address enclosed in square brackets (<tt>'['</tt> and
1254      *   <tt>']'</tt>) and consisting of hexadecimal digits, colon characters
1255      *   (<tt>':'</tt>), and possibly an embedded IPv4 address.  The full
1256      *   syntax of IPv6 addresses is specified in <a
1257      *   href="http://www.ietf.org/rfc/rfc2373.txt"><i>RFC&nbsp;2373: IPv6
1258      *   Addressing Architecture</i></a>.  </p></li>
1259      *
1260      * </ul>
1261      *
1262      * The host component of a URI cannot contain escaped octets, hence this
1263      * method does not perform any decoding.
1264      *
1265      * @return  The host component of this URI,
1266      *          or <tt>null</tt> if the host is undefined
1267      */
getHost()1268     public String getHost() {
1269         return host;
1270     }
1271 
1272     /**
1273      * Returns the port number of this URI.
1274      *
1275      * <p> The port component of a URI, if defined, is a non-negative
1276      * integer. </p>
1277      *
1278      * @return  The port component of this URI,
1279      *          or <tt>-1</tt> if the port is undefined
1280      */
getPort()1281     public int getPort() {
1282         return port;
1283     }
1284 
1285     /**
1286      * Returns the raw path component of this URI.
1287      *
1288      * <p> The path component of a URI, if defined, only contains the slash
1289      * character (<tt>'/'</tt>), the commercial-at character (<tt>'@'</tt>),
1290      * and characters in the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>,
1291      * and <i>other</i> categories. </p>
1292      *
1293      * @return  The path component of this URI,
1294      *          or <tt>null</tt> if the path is undefined
1295      */
getRawPath()1296     public String getRawPath() {
1297         return path;
1298     }
1299 
1300     /**
1301      * Returns the decoded path component of this URI.
1302      *
1303      * <p> The string returned by this method is equal to that returned by the
1304      * {@link #getRawPath() getRawPath} method except that all sequences of
1305      * escaped octets are <a href="#decode">decoded</a>.  </p>
1306      *
1307      * @return  The decoded path component of this URI,
1308      *          or <tt>null</tt> if the path is undefined
1309      */
getPath()1310     public String getPath() {
1311         if ((decodedPath == null) && (path != null))
1312             decodedPath = decode(path);
1313         return decodedPath;
1314     }
1315 
1316     /**
1317      * Returns the raw query component of this URI.
1318      *
1319      * <p> The query component of a URI, if defined, only contains legal URI
1320      * characters. </p>
1321      *
1322      * @return  The raw query component of this URI,
1323      *          or <tt>null</tt> if the query is undefined
1324      */
getRawQuery()1325     public String getRawQuery() {
1326         return query;
1327     }
1328 
1329     /**
1330      * Returns the decoded query component of this URI.
1331      *
1332      * <p> The string returned by this method is equal to that returned by the
1333      * {@link #getRawQuery() getRawQuery} method except that all sequences of
1334      * escaped octets are <a href="#decode">decoded</a>.  </p>
1335      *
1336      * @return  The decoded query component of this URI,
1337      *          or <tt>null</tt> if the query is undefined
1338      */
getQuery()1339     public String getQuery() {
1340         if ((decodedQuery == null) && (query != null))
1341             decodedQuery = decode(query);
1342         return decodedQuery;
1343     }
1344 
1345     /**
1346      * Returns the raw fragment component of this URI.
1347      *
1348      * <p> The fragment component of a URI, if defined, only contains legal URI
1349      * characters. </p>
1350      *
1351      * @return  The raw fragment component of this URI,
1352      *          or <tt>null</tt> if the fragment is undefined
1353      */
getRawFragment()1354     public String getRawFragment() {
1355         return fragment;
1356     }
1357 
1358     /**
1359      * Returns the decoded fragment component of this URI.
1360      *
1361      * <p> The string returned by this method is equal to that returned by the
1362      * {@link #getRawFragment() getRawFragment} method except that all
1363      * sequences of escaped octets are <a href="#decode">decoded</a>.  </p>
1364      *
1365      * @return  The decoded fragment component of this URI,
1366      *          or <tt>null</tt> if the fragment is undefined
1367      */
getFragment()1368     public String getFragment() {
1369         if ((decodedFragment == null) && (fragment != null))
1370             decodedFragment = decode(fragment);
1371         return decodedFragment;
1372     }
1373 
1374 
1375     // -- Equality, comparison, hash code, toString, and serialization --
1376 
1377     /**
1378      * Tests this URI for equality with another object.
1379      *
1380      * <p> If the given object is not a URI then this method immediately
1381      * returns <tt>false</tt>.
1382      *
1383      * <p> For two URIs to be considered equal requires that either both are
1384      * opaque or both are hierarchical.  Their schemes must either both be
1385      * undefined or else be equal without regard to case. Their fragments
1386      * must either both be undefined or else be equal.
1387      *
1388      * <p> For two opaque URIs to be considered equal, their scheme-specific
1389      * parts must be equal.
1390      *
1391      * <p> For two hierarchical URIs to be considered equal, their paths must
1392      * be equal and their queries must either both be undefined or else be
1393      * equal.  Their authorities must either both be undefined, or both be
1394      * registry-based, or both be server-based.  If their authorities are
1395      * defined and are registry-based, then they must be equal.  If their
1396      * authorities are defined and are server-based, then their hosts must be
1397      * equal without regard to case, their port numbers must be equal, and
1398      * their user-information components must be equal.
1399      *
1400      * <p> When testing the user-information, path, query, fragment, authority,
1401      * or scheme-specific parts of two URIs for equality, the raw forms rather
1402      * than the encoded forms of these components are compared and the
1403      * hexadecimal digits of escaped octets are compared without regard to
1404      * case.
1405      *
1406      * <p> This method satisfies the general contract of the {@link
1407      * java.lang.Object#equals(Object) Object.equals} method. </p>
1408      *
1409      * @param   ob   The object to which this object is to be compared
1410      *
1411      * @return  <tt>true</tt> if, and only if, the given object is a URI that
1412      *          is identical to this URI
1413      */
equals(Object ob)1414     public boolean equals(Object ob) {
1415         if (ob == this)
1416             return true;
1417         if (!(ob instanceof URI))
1418             return false;
1419         URI that = (URI)ob;
1420         if (this.isOpaque() != that.isOpaque()) return false;
1421         if (!equalIgnoringCase(this.scheme, that.scheme)) return false;
1422         if (!equal(this.fragment, that.fragment)) return false;
1423 
1424         // Opaque
1425         if (this.isOpaque())
1426             return equal(this.schemeSpecificPart, that.schemeSpecificPart);
1427 
1428         // Hierarchical
1429         if (!equal(this.path, that.path)) return false;
1430         if (!equal(this.query, that.query)) return false;
1431 
1432         // Authorities
1433         if (this.authority == that.authority) return true;
1434         if (this.host != null) {
1435             // Server-based
1436             if (!equal(this.userInfo, that.userInfo)) return false;
1437             if (!equalIgnoringCase(this.host, that.host)) return false;
1438             if (this.port != that.port) return false;
1439         } else if (this.authority != null) {
1440             // Registry-based
1441             if (!equal(this.authority, that.authority)) return false;
1442         } else if (this.authority != that.authority) {
1443             return false;
1444         }
1445 
1446         return true;
1447     }
1448 
1449     /**
1450      * Returns a hash-code value for this URI.  The hash code is based upon all
1451      * of the URI's components, and satisfies the general contract of the
1452      * {@link java.lang.Object#hashCode() Object.hashCode} method.
1453      *
1454      * @return  A hash-code value for this URI
1455      */
hashCode()1456     public int hashCode() {
1457         if (hash != 0)
1458             return hash;
1459         int h = hashIgnoringCase(0, scheme);
1460         h = hash(h, fragment);
1461         if (isOpaque()) {
1462             h = hash(h, schemeSpecificPart);
1463         } else {
1464             h = hash(h, path);
1465             h = hash(h, query);
1466             if (host != null) {
1467                 h = hash(h, userInfo);
1468                 h = hashIgnoringCase(h, host);
1469                 h += 1949 * port;
1470             } else {
1471                 h = hash(h, authority);
1472             }
1473         }
1474         hash = h;
1475         return h;
1476     }
1477 
1478     /**
1479      * Compares this URI to another object, which must be a URI.
1480      *
1481      * <p> When comparing corresponding components of two URIs, if one
1482      * component is undefined but the other is defined then the first is
1483      * considered to be less than the second.  Unless otherwise noted, string
1484      * components are ordered according to their natural, case-sensitive
1485      * ordering as defined by the {@link java.lang.String#compareTo(Object)
1486      * String.compareTo} method.  String components that are subject to
1487      * encoding are compared by comparing their raw forms rather than their
1488      * encoded forms.
1489      *
1490      * <p> The ordering of URIs is defined as follows: </p>
1491      *
1492      * <ul type=disc>
1493      *
1494      *   <li><p> Two URIs with different schemes are ordered according the
1495      *   ordering of their schemes, without regard to case. </p></li>
1496      *
1497      *   <li><p> A hierarchical URI is considered to be less than an opaque URI
1498      *   with an identical scheme. </p></li>
1499      *
1500      *   <li><p> Two opaque URIs with identical schemes are ordered according
1501      *   to the ordering of their scheme-specific parts. </p></li>
1502      *
1503      *   <li><p> Two opaque URIs with identical schemes and scheme-specific
1504      *   parts are ordered according to the ordering of their
1505      *   fragments. </p></li>
1506      *
1507      *   <li><p> Two hierarchical URIs with identical schemes are ordered
1508      *   according to the ordering of their authority components: </p>
1509      *
1510      *   <ul type=disc>
1511      *
1512      *     <li><p> If both authority components are server-based then the URIs
1513      *     are ordered according to their user-information components; if these
1514      *     components are identical then the URIs are ordered according to the
1515      *     ordering of their hosts, without regard to case; if the hosts are
1516      *     identical then the URIs are ordered according to the ordering of
1517      *     their ports. </p></li>
1518      *
1519      *     <li><p> If one or both authority components are registry-based then
1520      *     the URIs are ordered according to the ordering of their authority
1521      *     components. </p></li>
1522      *
1523      *   </ul></li>
1524      *
1525      *   <li><p> Finally, two hierarchical URIs with identical schemes and
1526      *   authority components are ordered according to the ordering of their
1527      *   paths; if their paths are identical then they are ordered according to
1528      *   the ordering of their queries; if the queries are identical then they
1529      *   are ordered according to the order of their fragments. </p></li>
1530      *
1531      * </ul>
1532      *
1533      * <p> This method satisfies the general contract of the {@link
1534      * java.lang.Comparable#compareTo(Object) Comparable.compareTo}
1535      * method. </p>
1536      *
1537      * @param   that
1538      *          The object to which this URI is to be compared
1539      *
1540      * @return  A negative integer, zero, or a positive integer as this URI is
1541      *          less than, equal to, or greater than the given URI
1542      *
1543      * @throws  ClassCastException
1544      *          If the given object is not a URI
1545      */
compareTo(URI that)1546     public int compareTo(URI that) {
1547         int c;
1548 
1549         if ((c = compareIgnoringCase(this.scheme, that.scheme)) != 0)
1550             return c;
1551 
1552         if (this.isOpaque()) {
1553             if (that.isOpaque()) {
1554                 // Both opaque
1555                 if ((c = compare(this.schemeSpecificPart,
1556                                  that.schemeSpecificPart)) != 0)
1557                     return c;
1558                 return compare(this.fragment, that.fragment);
1559             }
1560             return +1;                  // Opaque > hierarchical
1561         } else if (that.isOpaque()) {
1562             return -1;                  // Hierarchical < opaque
1563         }
1564 
1565         // Hierarchical
1566         if ((this.host != null) && (that.host != null)) {
1567             // Both server-based
1568             if ((c = compare(this.userInfo, that.userInfo)) != 0)
1569                 return c;
1570             if ((c = compareIgnoringCase(this.host, that.host)) != 0)
1571                 return c;
1572             if ((c = this.port - that.port) != 0)
1573                 return c;
1574         } else {
1575             // If one or both authorities are registry-based then we simply
1576             // compare them in the usual, case-sensitive way.  If one is
1577             // registry-based and one is server-based then the strings are
1578             // guaranteed to be unequal, hence the comparison will never return
1579             // zero and the compareTo and equals methods will remain
1580             // consistent.
1581             if ((c = compare(this.authority, that.authority)) != 0) return c;
1582         }
1583 
1584         if ((c = compare(this.path, that.path)) != 0) return c;
1585         if ((c = compare(this.query, that.query)) != 0) return c;
1586         return compare(this.fragment, that.fragment);
1587     }
1588 
1589     /**
1590      * Returns the content of this URI as a string.
1591      *
1592      * <p> If this URI was created by invoking one of the constructors in this
1593      * class then a string equivalent to the original input string, or to the
1594      * string computed from the originally-given components, as appropriate, is
1595      * returned.  Otherwise this URI was created by normalization, resolution,
1596      * or relativization, and so a string is constructed from this URI's
1597      * components according to the rules specified in <a
1598      * href="http://www.ietf.org/rfc/rfc2396.txt">RFC&nbsp;2396</a>,
1599      * section&nbsp;5.2, step&nbsp;7. </p>
1600      *
1601      * @return  The string form of this URI
1602      */
toString()1603     public String toString() {
1604         defineString();
1605         return string;
1606     }
1607 
1608     /**
1609      * Returns the content of this URI as a US-ASCII string.
1610      *
1611      * <p> If this URI does not contain any characters in the <i>other</i>
1612      * category then an invocation of this method will return the same value as
1613      * an invocation of the {@link #toString() toString} method.  Otherwise
1614      * this method works as if by invoking that method and then <a
1615      * href="#encode">encoding</a> the result.  </p>
1616      *
1617      * @return  The string form of this URI, encoded as needed
1618      *          so that it only contains characters in the US-ASCII
1619      *          charset
1620      */
toASCIIString()1621     public String toASCIIString() {
1622         defineString();
1623         return encode(string);
1624     }
1625 
1626 
1627     // -- Serialization support --
1628 
1629     /**
1630      * Saves the content of this URI to the given serial stream.
1631      *
1632      * <p> The only serializable field of a URI instance is its <tt>string</tt>
1633      * field.  That field is given a value, if it does not have one already,
1634      * and then the {@link java.io.ObjectOutputStream#defaultWriteObject()}
1635      * method of the given object-output stream is invoked. </p>
1636      *
1637      * @param  os  The object-output stream to which this object
1638      *             is to be written
1639      */
writeObject(ObjectOutputStream os)1640     private void writeObject(ObjectOutputStream os)
1641         throws IOException
1642     {
1643         defineString();
1644         os.defaultWriteObject();        // Writes the string field only
1645     }
1646 
1647     /**
1648      * Reconstitutes a URI from the given serial stream.
1649      *
1650      * <p> The {@link java.io.ObjectInputStream#defaultReadObject()} method is
1651      * invoked to read the value of the <tt>string</tt> field.  The result is
1652      * then parsed in the usual way.
1653      *
1654      * @param  is  The object-input stream from which this object
1655      *             is being read
1656      */
readObject(ObjectInputStream is)1657     private void readObject(ObjectInputStream is)
1658         throws ClassNotFoundException, IOException
1659     {
1660         port = -1;                      // Argh
1661         is.defaultReadObject();
1662         try {
1663             new Parser(string).parse(false);
1664         } catch (URISyntaxException x) {
1665             IOException y = new InvalidObjectException("Invalid URI");
1666             y.initCause(x);
1667             throw y;
1668         }
1669     }
1670 
1671 
1672     // -- End of public methods --
1673 
1674 
1675     // -- Utility methods for string-field comparison and hashing --
1676 
1677     // These methods return appropriate values for null string arguments,
1678     // thereby simplifying the equals, hashCode, and compareTo methods.
1679     //
1680     // The case-ignoring methods should only be applied to strings whose
1681     // characters are all known to be US-ASCII.  Because of this restriction,
1682     // these methods are faster than the similar methods in the String class.
1683 
1684     // US-ASCII only
toLower(char c)1685     private static int toLower(char c) {
1686         if ((c >= 'A') && (c <= 'Z'))
1687             return c + ('a' - 'A');
1688         return c;
1689     }
1690 
equal(String s, String t)1691     private static boolean equal(String s, String t) {
1692         if (s == t) return true;
1693         if ((s != null) && (t != null)) {
1694             if (s.length() != t.length())
1695                 return false;
1696             if (s.indexOf('%') < 0)
1697                 return s.equals(t);
1698             int n = s.length();
1699             for (int i = 0; i < n;) {
1700                 char c = s.charAt(i);
1701                 char d = t.charAt(i);
1702                 if (c != '%') {
1703                     if (c != d)
1704                         return false;
1705                     i++;
1706                     continue;
1707                 }
1708                 if (d != '%')
1709                     return false;
1710                 i++;
1711                 if (toLower(s.charAt(i)) != toLower(t.charAt(i)))
1712                     return false;
1713                 i++;
1714                 if (toLower(s.charAt(i)) != toLower(t.charAt(i)))
1715                     return false;
1716                 i++;
1717             }
1718             return true;
1719         }
1720         return false;
1721     }
1722 
1723     // US-ASCII only
equalIgnoringCase(String s, String t)1724     private static boolean equalIgnoringCase(String s, String t) {
1725         if (s == t) return true;
1726         if ((s != null) && (t != null)) {
1727             int n = s.length();
1728             if (t.length() != n)
1729                 return false;
1730             for (int i = 0; i < n; i++) {
1731                 if (toLower(s.charAt(i)) != toLower(t.charAt(i)))
1732                     return false;
1733             }
1734             return true;
1735         }
1736         return false;
1737     }
1738 
hash(int hash, String s)1739     private static int hash(int hash, String s) {
1740         if (s == null) return hash;
1741         return hash * 127 + s.hashCode();
1742     }
1743 
1744     // US-ASCII only
hashIgnoringCase(int hash, String s)1745     private static int hashIgnoringCase(int hash, String s) {
1746         if (s == null) return hash;
1747         int h = hash;
1748         int n = s.length();
1749         for (int i = 0; i < n; i++)
1750             h = 31 * h + toLower(s.charAt(i));
1751         return h;
1752     }
1753 
compare(String s, String t)1754     private static int compare(String s, String t) {
1755         if (s == t) return 0;
1756         if (s != null) {
1757             if (t != null)
1758                 return s.compareTo(t);
1759             else
1760                 return +1;
1761         } else {
1762             return -1;
1763         }
1764     }
1765 
1766     // US-ASCII only
compareIgnoringCase(String s, String t)1767     private static int compareIgnoringCase(String s, String t) {
1768         if (s == t) return 0;
1769         if (s != null) {
1770             if (t != null) {
1771                 int sn = s.length();
1772                 int tn = t.length();
1773                 int n = sn < tn ? sn : tn;
1774                 for (int i = 0; i < n; i++) {
1775                     int c = toLower(s.charAt(i)) - toLower(t.charAt(i));
1776                     if (c != 0)
1777                         return c;
1778                 }
1779                 return sn - tn;
1780             }
1781             return +1;
1782         } else {
1783             return -1;
1784         }
1785     }
1786 
1787 
1788     // -- String construction --
1789 
1790     // If a scheme is given then the path, if given, must be absolute
1791     //
1792     private static void checkPath(String s, String scheme, String path)
1793         throws URISyntaxException
1794     {
1795         if (scheme != null) {
1796             if ((path != null)
1797                 && ((path.length() > 0) && (path.charAt(0) != '/')))
1798                 throw new URISyntaxException(s,
1799                                              "Relative path in absolute URI");
1800         }
1801     }
1802 
1803     private void appendAuthority(StringBuffer sb,
1804                                  String authority,
1805                                  String userInfo,
1806                                  String host,
1807                                  int port)
1808     {
1809         if (host != null) {
1810             sb.append("//");
1811             if (userInfo != null) {
1812                 sb.append(quote(userInfo, L_USERINFO, H_USERINFO));
1813                 sb.append('@');
1814             }
1815             boolean needBrackets = ((host.indexOf(':') >= 0)
1816                                     && !host.startsWith("[")
1817                                     && !host.endsWith("]"));
1818             if (needBrackets) sb.append('[');
1819             sb.append(host);
1820             if (needBrackets) sb.append(']');
1821             if (port != -1) {
1822                 sb.append(':');
1823                 sb.append(port);
1824             }
1825         } else if (authority != null) {
1826             sb.append("//");
1827             if (authority.startsWith("[")) {
1828                 // authority should (but may not) contain an embedded IPv6 address
1829                 int end = authority.indexOf("]");
1830                 String doquote = authority, dontquote = "";
1831                 if (end != -1 && authority.indexOf(":") != -1) {
1832                     // the authority contains an IPv6 address
1833                     if (end == authority.length()) {
1834                         dontquote = authority;
1835                         doquote = "";
1836                     } else {
1837                         dontquote = authority.substring(0 , end + 1);
1838                         doquote = authority.substring(end + 1);
1839                     }
1840                 }
1841                 sb.append(dontquote);
1842                 sb.append(quote(doquote,
1843                             L_REG_NAME | L_SERVER,
1844                             H_REG_NAME | H_SERVER));
1845             } else {
1846                 sb.append(quote(authority,
1847                             L_REG_NAME | L_SERVER,
1848                             H_REG_NAME | H_SERVER));
1849             }
1850         }
1851     }
1852 
appendSchemeSpecificPart(StringBuffer sb, String opaquePart, String authority, String userInfo, String host, int port, String path, String query)1853     private void appendSchemeSpecificPart(StringBuffer sb,
1854                                           String opaquePart,
1855                                           String authority,
1856                                           String userInfo,
1857                                           String host,
1858                                           int port,
1859                                           String path,
1860                                           String query)
1861     {
1862         if (opaquePart != null) {
1863             /* check if SSP begins with an IPv6 address
1864              * because we must not quote a literal IPv6 address
1865              */
1866             if (opaquePart.startsWith("//[")) {
1867                 int end =  opaquePart.indexOf("]");
1868                 if (end != -1 && opaquePart.indexOf(":")!=-1) {
1869                     String doquote, dontquote;
1870                     if (end == opaquePart.length()) {
1871                         dontquote = opaquePart;
1872                         doquote = "";
1873                     } else {
1874                         dontquote = opaquePart.substring(0,end+1);
1875                         doquote = opaquePart.substring(end+1);
1876                     }
1877                     sb.append (dontquote);
1878                     sb.append(quote(doquote, L_URIC, H_URIC));
1879                 }
1880             } else {
1881                 sb.append(quote(opaquePart, L_URIC, H_URIC));
1882             }
1883         } else {
1884             appendAuthority(sb, authority, userInfo, host, port);
1885             if (path != null)
1886                 sb.append(quote(path, L_PATH, H_PATH));
1887             if (query != null) {
1888                 sb.append('?');
1889                 sb.append(quote(query, L_URIC, H_URIC));
1890             }
1891         }
1892     }
1893 
appendFragment(StringBuffer sb, String fragment)1894     private void appendFragment(StringBuffer sb, String fragment) {
1895         if (fragment != null) {
1896             sb.append('#');
1897             sb.append(quote(fragment, L_URIC, H_URIC));
1898         }
1899     }
1900 
toString(String scheme, String opaquePart, String authority, String userInfo, String host, int port, String path, String query, String fragment)1901     private String toString(String scheme,
1902                             String opaquePart,
1903                             String authority,
1904                             String userInfo,
1905                             String host,
1906                             int port,
1907                             String path,
1908                             String query,
1909                             String fragment)
1910     {
1911         StringBuffer sb = new StringBuffer();
1912         if (scheme != null) {
1913             sb.append(scheme);
1914             sb.append(':');
1915         }
1916         appendSchemeSpecificPart(sb, opaquePart,
1917                                  authority, userInfo, host, port,
1918                                  path, query);
1919         appendFragment(sb, fragment);
1920         return sb.toString();
1921     }
1922 
defineSchemeSpecificPart()1923     private void defineSchemeSpecificPart() {
1924         if (schemeSpecificPart != null) return;
1925         StringBuffer sb = new StringBuffer();
1926         appendSchemeSpecificPart(sb, null, getAuthority(), getUserInfo(),
1927                                  host, port, getPath(), getQuery());
1928         if (sb.length() == 0) return;
1929         schemeSpecificPart = sb.toString();
1930     }
1931 
defineString()1932     private void defineString() {
1933         if (string != null) return;
1934 
1935         StringBuffer sb = new StringBuffer();
1936         if (scheme != null) {
1937             sb.append(scheme);
1938             sb.append(':');
1939         }
1940         if (isOpaque()) {
1941             sb.append(schemeSpecificPart);
1942         } else {
1943             if (host != null) {
1944                 sb.append("//");
1945                 if (userInfo != null) {
1946                     sb.append(userInfo);
1947                     sb.append('@');
1948                 }
1949                 boolean needBrackets = ((host.indexOf(':') >= 0)
1950                                     && !host.startsWith("[")
1951                                     && !host.endsWith("]"));
1952                 if (needBrackets) sb.append('[');
1953                 sb.append(host);
1954                 if (needBrackets) sb.append(']');
1955                 if (port != -1) {
1956                     sb.append(':');
1957                     sb.append(port);
1958                 }
1959             } else if (authority != null) {
1960                 sb.append("//");
1961                 sb.append(authority);
1962             }
1963             if (path != null)
1964                 sb.append(path);
1965             if (query != null) {
1966                 sb.append('?');
1967                 sb.append(query);
1968             }
1969         }
1970         if (fragment != null) {
1971             sb.append('#');
1972             sb.append(fragment);
1973         }
1974         string = sb.toString();
1975     }
1976 
1977 
1978     // -- Normalization, resolution, and relativization --
1979 
1980     // RFC2396 5.2 (6)
resolvePath(String base, String child, boolean absolute)1981     private static String resolvePath(String base, String child,
1982                                       boolean absolute)
1983     {
1984         int i = base.lastIndexOf('/');
1985         int cn = child.length();
1986         String path = "";
1987 
1988         if (cn == 0) {
1989             // 5.2 (6a)
1990             if (i >= 0)
1991                 path = base.substring(0, i + 1);
1992         } else {
1993             StringBuffer sb = new StringBuffer(base.length() + cn);
1994             // 5.2 (6a)
1995             if (i >= 0)
1996                 sb.append(base.substring(0, i + 1));
1997             // 5.2 (6b)
1998             sb.append(child);
1999             path = sb.toString();
2000         }
2001 
2002         // 5.2 (6c-f)
2003         String np = normalize(path, true);
2004 
2005         // 5.2 (6g): If the result is absolute but the path begins with "../",
2006         // then we simply leave the path as-is
2007 
2008         return np;
2009     }
2010 
2011     // RFC2396 5.2
resolve(URI base, URI child)2012     private static URI resolve(URI base, URI child) {
2013         // check if child if opaque first so that NPE is thrown
2014         // if child is null.
2015         if (child.isOpaque() || base.isOpaque())
2016             return child;
2017 
2018         // 5.2 (2): Reference to current document (lone fragment)
2019         if ((child.scheme == null) && (child.authority == null)
2020             && child.path.equals("") && (child.fragment != null)
2021             && (child.query == null)) {
2022             if ((base.fragment != null)
2023                 && child.fragment.equals(base.fragment)) {
2024                 return base;
2025             }
2026             URI ru = new URI();
2027             ru.scheme = base.scheme;
2028             ru.authority = base.authority;
2029             ru.userInfo = base.userInfo;
2030             ru.host = base.host;
2031             ru.port = base.port;
2032             ru.path = base.path;
2033             ru.fragment = child.fragment;
2034             ru.query = base.query;
2035             return ru;
2036         }
2037 
2038         // 5.2 (3): Child is absolute
2039         if (child.scheme != null)
2040             return child;
2041 
2042         URI ru = new URI();             // Resolved URI
2043         ru.scheme = base.scheme;
2044         ru.query = child.query;
2045         ru.fragment = child.fragment;
2046 
2047         // 5.2 (4): Authority
2048         if (child.authority == null) {
2049             ru.authority = base.authority;
2050             ru.host = base.host;
2051             ru.userInfo = base.userInfo;
2052             ru.port = base.port;
2053 
2054             if (child.path == null || child.path.isEmpty()) {
2055                 // This is an addtional path from RFC 3986 RI, which fixes following RFC 2396
2056                 // "normal" examples:
2057                 // Base: http://a/b/c/d;p?q
2058                 //   "?y" = "http://a/b/c/d;p?y"
2059                 //   ""   = "http://a/b/c/d;p?q"
2060                 // http://b/25897693
2061                 ru.path = base.path;
2062                 ru.query = child.query != null ? child.query : base.query;
2063             } else if ((child.path.length() > 0) && (child.path.charAt(0) == '/')) {
2064                 // 5.2 (5): Child path is absolute
2065                 //
2066                 // There is an additional step from RFC 3986 RI, requiring to remove dots for
2067                 // absolute path as well.
2068                 // http://b/25897693
2069                 ru.path = normalize(child.path, true);
2070             } else {
2071                 // 5.2 (6): Resolve relative path
2072                 ru.path = resolvePath(base.path, child.path, base.isAbsolute());
2073             }
2074         } else {
2075             ru.authority = child.authority;
2076             ru.host = child.host;
2077             ru.userInfo = child.userInfo;
2078             ru.host = child.host;
2079             ru.port = child.port;
2080             ru.path = child.path;
2081         }
2082 
2083         // 5.2 (7): Recombine (nothing to do here)
2084         return ru;
2085     }
2086 
2087     // If the given URI's path is normal then return the URI;
2088     // o.w., return a new URI containing the normalized path.
2089     //
normalize(URI u)2090     private static URI normalize(URI u) {
2091         if (u.isOpaque() || (u.path == null) || (u.path.length() == 0))
2092             return u;
2093 
2094         String np = normalize(u.path);
2095         if (np == u.path)
2096             return u;
2097 
2098         URI v = new URI();
2099         v.scheme = u.scheme;
2100         v.fragment = u.fragment;
2101         v.authority = u.authority;
2102         v.userInfo = u.userInfo;
2103         v.host = u.host;
2104         v.port = u.port;
2105         v.path = np;
2106         v.query = u.query;
2107         return v;
2108     }
2109 
2110     // If both URIs are hierarchical, their scheme and authority components are
2111     // identical, and the base path is a prefix of the child's path, then
2112     // return a relative URI that, when resolved against the base, yields the
2113     // child; otherwise, return the child.
2114     //
relativize(URI base, URI child)2115     private static URI relativize(URI base, URI child) {
2116         // check if child if opaque first so that NPE is thrown
2117         // if child is null.
2118         if (child.isOpaque() || base.isOpaque())
2119             return child;
2120         if (!equalIgnoringCase(base.scheme, child.scheme)
2121             || !equal(base.authority, child.authority))
2122             return child;
2123 
2124         String bp = normalize(base.path);
2125         String cp = normalize(child.path);
2126         if (!bp.equals(cp)) {
2127             // Android-changed: The original OpenJdk implementation would append a trailing slash
2128             // to paths like "/a/b" before relativizing them. This would relativize /a/b/c to
2129             // "/c" against "/a/b" the android implementation did not do this. It would assume that
2130             // "b" wasn't a directory and relativize the path to "/b/c". The spec is pretty vague
2131             // about this but this change is being made because we have several tests that expect
2132             // this behaviour.
2133             if (bp.indexOf('/') != -1) {
2134                 bp = bp.substring(0, bp.lastIndexOf('/') + 1);
2135             }
2136 
2137             if (!cp.startsWith(bp))
2138                 return child;
2139         }
2140 
2141         URI v = new URI();
2142         v.path = cp.substring(bp.length());
2143         v.query = child.query;
2144         v.fragment = child.fragment;
2145         return v;
2146     }
2147 
2148 
2149 
2150     // -- Path normalization --
2151 
2152     // The following algorithm for path normalization avoids the creation of a
2153     // string object for each segment, as well as the use of a string buffer to
2154     // compute the final result, by using a single char array and editing it in
2155     // place.  The array is first split into segments, replacing each slash
2156     // with '\0' and creating a segment-index array, each element of which is
2157     // the index of the first char in the corresponding segment.  We then walk
2158     // through both arrays, removing ".", "..", and other segments as necessary
2159     // by setting their entries in the index array to -1.  Finally, the two
2160     // arrays are used to rejoin the segments and compute the final result.
2161     //
2162     // This code is based upon src/solaris/native/java/io/canonicalize_md.c
2163 
2164 
2165     // Check the given path to see if it might need normalization.  A path
2166     // might need normalization if it contains duplicate slashes, a "."
2167     // segment, or a ".." segment.  Return -1 if no further normalization is
2168     // possible, otherwise return the number of segments found.
2169     //
2170     // This method takes a string argument rather than a char array so that
2171     // this test can be performed without invoking path.toCharArray().
2172     //
needsNormalization(String path)2173     static private int needsNormalization(String path) {
2174         boolean normal = true;
2175         int ns = 0;                     // Number of segments
2176         int end = path.length() - 1;    // Index of last char in path
2177         int p = 0;                      // Index of next char in path
2178 
2179         // Skip initial slashes
2180         while (p <= end) {
2181             if (path.charAt(p) != '/') break;
2182             p++;
2183         }
2184         if (p > 1) normal = false;
2185 
2186         // Scan segments
2187         while (p <= end) {
2188 
2189             // Looking at "." or ".." ?
2190             if ((path.charAt(p) == '.')
2191                 && ((p == end)
2192                     || ((path.charAt(p + 1) == '/')
2193                         || ((path.charAt(p + 1) == '.')
2194                             && ((p + 1 == end)
2195                                 || (path.charAt(p + 2) == '/')))))) {
2196                 normal = false;
2197             }
2198             ns++;
2199 
2200             // Find beginning of next segment
2201             while (p <= end) {
2202                 if (path.charAt(p++) != '/')
2203                     continue;
2204 
2205                 // Skip redundant slashes
2206                 while (p <= end) {
2207                     if (path.charAt(p) != '/') break;
2208                     normal = false;
2209                     p++;
2210                 }
2211 
2212                 break;
2213             }
2214         }
2215 
2216         return normal ? -1 : ns;
2217     }
2218 
2219 
2220     // Split the given path into segments, replacing slashes with nulls and
2221     // filling in the given segment-index array.
2222     //
2223     // Preconditions:
2224     //   segs.length == Number of segments in path
2225     //
2226     // Postconditions:
2227     //   All slashes in path replaced by '\0'
2228     //   segs[i] == Index of first char in segment i (0 <= i < segs.length)
2229     //
split(char[] path, int[] segs)2230     static private void split(char[] path, int[] segs) {
2231         int end = path.length - 1;      // Index of last char in path
2232         int p = 0;                      // Index of next char in path
2233         int i = 0;                      // Index of current segment
2234 
2235         // Skip initial slashes
2236         while (p <= end) {
2237             if (path[p] != '/') break;
2238             path[p] = '\0';
2239             p++;
2240         }
2241 
2242         while (p <= end) {
2243 
2244             // Note start of segment
2245             segs[i++] = p++;
2246 
2247             // Find beginning of next segment
2248             while (p <= end) {
2249                 if (path[p++] != '/')
2250                     continue;
2251                 path[p - 1] = '\0';
2252 
2253                 // Skip redundant slashes
2254                 while (p <= end) {
2255                     if (path[p] != '/') break;
2256                     path[p++] = '\0';
2257                 }
2258                 break;
2259             }
2260         }
2261 
2262         if (i != segs.length)
2263             throw new InternalError();  // ASSERT
2264     }
2265 
2266 
2267     // Join the segments in the given path according to the given segment-index
2268     // array, ignoring those segments whose index entries have been set to -1,
2269     // and inserting slashes as needed.  Return the length of the resulting
2270     // path.
2271     //
2272     // Preconditions:
2273     //   segs[i] == -1 implies segment i is to be ignored
2274     //   path computed by split, as above, with '\0' having replaced '/'
2275     //
2276     // Postconditions:
2277     //   path[0] .. path[return value] == Resulting path
2278     //
join(char[] path, int[] segs)2279     static private int join(char[] path, int[] segs) {
2280         int ns = segs.length;           // Number of segments
2281         int end = path.length - 1;      // Index of last char in path
2282         int p = 0;                      // Index of next path char to write
2283 
2284         if (path[p] == '\0') {
2285             // Restore initial slash for absolute paths
2286             path[p++] = '/';
2287         }
2288 
2289         for (int i = 0; i < ns; i++) {
2290             int q = segs[i];            // Current segment
2291             if (q == -1)
2292                 // Ignore this segment
2293                 continue;
2294 
2295             if (p == q) {
2296                 // We're already at this segment, so just skip to its end
2297                 while ((p <= end) && (path[p] != '\0'))
2298                     p++;
2299                 if (p <= end) {
2300                     // Preserve trailing slash
2301                     path[p++] = '/';
2302                 }
2303             } else if (p < q) {
2304                 // Copy q down to p
2305                 while ((q <= end) && (path[q] != '\0'))
2306                     path[p++] = path[q++];
2307                 if (q <= end) {
2308                     // Preserve trailing slash
2309                     path[p++] = '/';
2310                 }
2311             } else
2312                 throw new InternalError(); // ASSERT false
2313         }
2314 
2315         return p;
2316     }
2317 
2318 
2319     // Remove "." segments from the given path, and remove segment pairs
2320     // consisting of a non-".." segment followed by a ".." segment.
2321     //
removeDots(char[] path, int[] segs, boolean removeLeading)2322     private static void removeDots(char[] path, int[] segs, boolean removeLeading) {
2323         int ns = segs.length;
2324         int end = path.length - 1;
2325 
2326         for (int i = 0; i < ns; i++) {
2327             int dots = 0;               // Number of dots found (0, 1, or 2)
2328 
2329             // Find next occurrence of "." or ".."
2330             do {
2331                 int p = segs[i];
2332                 if (path[p] == '.') {
2333                     if (p == end) {
2334                         dots = 1;
2335                         break;
2336                     } else if (path[p + 1] == '\0') {
2337                         dots = 1;
2338                         break;
2339                     } else if ((path[p + 1] == '.')
2340                                && ((p + 1 == end)
2341                                    || (path[p + 2] == '\0'))) {
2342                         dots = 2;
2343                         break;
2344                     }
2345                 }
2346                 i++;
2347             } while (i < ns);
2348             if ((i > ns) || (dots == 0))
2349                 break;
2350 
2351             if (dots == 1) {
2352                 // Remove this occurrence of "."
2353                 segs[i] = -1;
2354             } else {
2355                 // If there is a preceding non-".." segment, remove both that
2356                 // segment and this occurrence of ".."
2357                 int j;
2358                 for (j = i - 1; j >= 0; j--) {
2359                     if (segs[j] != -1) break;
2360                 }
2361                 if (j >= 0) {
2362                     int q = segs[j];
2363                     if (!((path[q] == '.')
2364                           && (path[q + 1] == '.')
2365                           && (path[q + 2] == '\0'))) {
2366                         segs[i] = -1;
2367                         segs[j] = -1;
2368                     }
2369                 } else if (removeLeading) {
2370                     // This is a leading ".." segment. Per RFC 3986 RI, this should be removed as
2371                     // well. This fixes RFC 2396 "abnormal" examples.
2372                     // http://b/25897693
2373                     segs[i] = -1;
2374                 }
2375             }
2376         }
2377     }
2378 
2379 
2380     // DEVIATION: If the normalized path is relative, and if the first
2381     // segment could be parsed as a scheme name, then prepend a "." segment
2382     //
maybeAddLeadingDot(char[] path, int[] segs)2383     private static void maybeAddLeadingDot(char[] path, int[] segs) {
2384 
2385         if (path[0] == '\0')
2386             // The path is absolute
2387             return;
2388 
2389         int ns = segs.length;
2390         int f = 0;                      // Index of first segment
2391         while (f < ns) {
2392             if (segs[f] >= 0)
2393                 break;
2394             f++;
2395         }
2396         if ((f >= ns) || (f == 0))
2397             // The path is empty, or else the original first segment survived,
2398             // in which case we already know that no leading "." is needed
2399             return;
2400 
2401         int p = segs[f];
2402         while ((p < path.length) && (path[p] != ':') && (path[p] != '\0')) p++;
2403         if (p >= path.length || path[p] == '\0')
2404             // No colon in first segment, so no "." needed
2405             return;
2406 
2407         // At this point we know that the first segment is unused,
2408         // hence we can insert a "." segment at that position
2409         path[0] = '.';
2410         path[1] = '\0';
2411         segs[0] = 0;
2412     }
2413 
2414 
2415     // Normalize the given path string.  A normal path string has no empty
2416     // segments (i.e., occurrences of "//"), no segments equal to ".", and no
2417     // segments equal to ".." that are preceded by a segment not equal to "..".
2418     // In contrast to Unix-style pathname normalization, for URI paths we
2419     // always retain trailing slashes.
2420     //
normalize(String ps)2421     private static String normalize(String ps) {
2422         return normalize(ps, false);
2423     }
2424 
normalize(String ps, boolean removeLeading)2425     private static String normalize(String ps, boolean removeLeading) {
2426 
2427         // Does this path need normalization?
2428         int ns = needsNormalization(ps);        // Number of segments
2429         if (ns < 0)
2430             // Nope -- just return it
2431             return ps;
2432 
2433         char[] path = ps.toCharArray();         // Path in char-array form
2434 
2435         // Split path into segments
2436         int[] segs = new int[ns];               // Segment-index array
2437         split(path, segs);
2438 
2439         // Remove dots
2440         removeDots(path, segs, removeLeading);
2441 
2442         // Prevent scheme-name confusion
2443         maybeAddLeadingDot(path, segs);
2444 
2445         // Join the remaining segments and return the result
2446         String s = new String(path, 0, join(path, segs));
2447         if (s.equals(ps)) {
2448             // string was already normalized
2449             return ps;
2450         }
2451         return s;
2452     }
2453 
2454 
2455 
2456     // -- Character classes for parsing --
2457 
2458     // RFC2396 precisely specifies which characters in the US-ASCII charset are
2459     // permissible in the various components of a URI reference.  We here
2460     // define a set of mask pairs to aid in enforcing these restrictions.  Each
2461     // mask pair consists of two longs, a low mask and a high mask.  Taken
2462     // together they represent a 128-bit mask, where bit i is set iff the
2463     // character with value i is permitted.
2464     //
2465     // This approach is more efficient than sequentially searching arrays of
2466     // permitted characters.  It could be made still more efficient by
2467     // precompiling the mask information so that a character's presence in a
2468     // given mask could be determined by a single table lookup.
2469 
2470     // Compute the low-order mask for the characters in the given string
lowMask(String chars)2471     private static long lowMask(String chars) {
2472         int n = chars.length();
2473         long m = 0;
2474         for (int i = 0; i < n; i++) {
2475             char c = chars.charAt(i);
2476             if (c < 64)
2477                 m |= (1L << c);
2478         }
2479         return m;
2480     }
2481 
2482     // Compute the high-order mask for the characters in the given string
highMask(String chars)2483     private static long highMask(String chars) {
2484         int n = chars.length();
2485         long m = 0;
2486         for (int i = 0; i < n; i++) {
2487             char c = chars.charAt(i);
2488             if ((c >= 64) && (c < 128))
2489                 m |= (1L << (c - 64));
2490         }
2491         return m;
2492     }
2493 
2494     // Compute a low-order mask for the characters
2495     // between first and last, inclusive
lowMask(char first, char last)2496     private static long lowMask(char first, char last) {
2497         long m = 0;
2498         int f = Math.max(Math.min(first, 63), 0);
2499         int l = Math.max(Math.min(last, 63), 0);
2500         for (int i = f; i <= l; i++)
2501             m |= 1L << i;
2502         return m;
2503     }
2504 
2505     // Compute a high-order mask for the characters
2506     // between first and last, inclusive
highMask(char first, char last)2507     private static long highMask(char first, char last) {
2508         long m = 0;
2509         int f = Math.max(Math.min(first, 127), 64) - 64;
2510         int l = Math.max(Math.min(last, 127), 64) - 64;
2511         for (int i = f; i <= l; i++)
2512             m |= 1L << i;
2513         return m;
2514     }
2515 
2516     // Tell whether the given character is permitted by the given mask pair
match(char c, long lowMask, long highMask)2517     private static boolean match(char c, long lowMask, long highMask) {
2518         if (c == 0) // 0 doesn't have a slot in the mask. So, it never matches.
2519             return false;
2520         if (c < 64)
2521             return ((1L << c) & lowMask) != 0;
2522         if (c < 128)
2523             return ((1L << (c - 64)) & highMask) != 0;
2524         return false;
2525     }
2526 
2527     // Character-class masks, in reverse order from RFC2396 because
2528     // initializers for static fields cannot make forward references.
2529 
2530     // digit    = "0" | "1" | "2" | "3" | "4" | "5" | "6" | "7" |
2531     //            "8" | "9"
2532     private static final long L_DIGIT = lowMask('0', '9');
2533     private static final long H_DIGIT = 0L;
2534 
2535     // upalpha  = "A" | "B" | "C" | "D" | "E" | "F" | "G" | "H" | "I" |
2536     //            "J" | "K" | "L" | "M" | "N" | "O" | "P" | "Q" | "R" |
2537     //            "S" | "T" | "U" | "V" | "W" | "X" | "Y" | "Z"
2538     private static final long L_UPALPHA = 0L;
2539     private static final long H_UPALPHA = highMask('A', 'Z');
2540 
2541     // lowalpha = "a" | "b" | "c" | "d" | "e" | "f" | "g" | "h" | "i" |
2542     //            "j" | "k" | "l" | "m" | "n" | "o" | "p" | "q" | "r" |
2543     //            "s" | "t" | "u" | "v" | "w" | "x" | "y" | "z"
2544     private static final long L_LOWALPHA = 0L;
2545     private static final long H_LOWALPHA = highMask('a', 'z');
2546 
2547     // alpha         = lowalpha | upalpha
2548     private static final long L_ALPHA = L_LOWALPHA | L_UPALPHA;
2549     private static final long H_ALPHA = H_LOWALPHA | H_UPALPHA;
2550 
2551     // alphanum      = alpha | digit
2552     private static final long L_ALPHANUM = L_DIGIT | L_ALPHA;
2553     private static final long H_ALPHANUM = H_DIGIT | H_ALPHA;
2554 
2555     // hex           = digit | "A" | "B" | "C" | "D" | "E" | "F" |
2556     //                         "a" | "b" | "c" | "d" | "e" | "f"
2557     private static final long L_HEX = L_DIGIT;
2558     private static final long H_HEX = highMask('A', 'F') | highMask('a', 'f');
2559 
2560     // mark          = "-" | "_" | "." | "!" | "~" | "*" | "'" |
2561     //                 "(" | ")"
2562     private static final long L_MARK = lowMask("-_.!~*'()");
2563     private static final long H_MARK = highMask("-_.!~*'()");
2564 
2565     // unreserved    = alphanum | mark
2566     private static final long L_UNRESERVED = L_ALPHANUM | L_MARK;
2567     private static final long H_UNRESERVED = H_ALPHANUM | H_MARK;
2568 
2569     // reserved      = ";" | "/" | "?" | ":" | "@" | "&" | "=" | "+" |
2570     //                 "$" | "," | "[" | "]"
2571     // Added per RFC2732: "[", "]"
2572     private static final long L_RESERVED = lowMask(";/?:@&=+$,[]");
2573     private static final long H_RESERVED = highMask(";/?:@&=+$,[]");
2574 
2575     // The zero'th bit is used to indicate that escape pairs and non-US-ASCII
2576     // characters are allowed; this is handled by the scanEscape method below.
2577     private static final long L_ESCAPED = 1L;
2578     private static final long H_ESCAPED = 0L;
2579 
2580     // uric          = reserved | unreserved | escaped
2581     private static final long L_URIC = L_RESERVED | L_UNRESERVED | L_ESCAPED;
2582     private static final long H_URIC = H_RESERVED | H_UNRESERVED | H_ESCAPED;
2583 
2584     // pchar         = unreserved | escaped |
2585     //                 ":" | "@" | "&" | "=" | "+" | "$" | ","
2586     private static final long L_PCHAR
2587         = L_UNRESERVED | L_ESCAPED | lowMask(":@&=+$,");
2588     private static final long H_PCHAR
2589         = H_UNRESERVED | H_ESCAPED | highMask(":@&=+$,");
2590 
2591     // All valid path characters
2592     private static final long L_PATH = L_PCHAR | lowMask(";/");
2593     private static final long H_PATH = H_PCHAR | highMask(";/");
2594 
2595     // Dash, for use in domainlabel and toplabel
2596     private static final long L_DASH = lowMask("-");
2597     private static final long H_DASH = highMask("-");
2598 
2599     // UNDERSCORE, for use in domainlabel and toplabel
2600     private static final long L_UNDERSCORE = lowMask("_");
2601     private static final long H_UNDERSCORE = highMask("_");
2602 
2603     // Dot, for use in hostnames
2604     private static final long L_DOT = lowMask(".");
2605     private static final long H_DOT = highMask(".");
2606 
2607     // userinfo      = *( unreserved | escaped |
2608     //                    ";" | ":" | "&" | "=" | "+" | "$" | "," )
2609     private static final long L_USERINFO
2610         = L_UNRESERVED | L_ESCAPED | lowMask(";:&=+$,");
2611     private static final long H_USERINFO
2612         = H_UNRESERVED | H_ESCAPED | highMask(";:&=+$,");
2613 
2614     // reg_name      = 1*( unreserved | escaped | "$" | "," |
2615     //                     ";" | ":" | "@" | "&" | "=" | "+" )
2616     private static final long L_REG_NAME
2617         = L_UNRESERVED | L_ESCAPED | lowMask("$,;:@&=+");
2618     private static final long H_REG_NAME
2619         = H_UNRESERVED | H_ESCAPED | highMask("$,;:@&=+");
2620 
2621     // All valid characters for server-based authorities
2622     private static final long L_SERVER
2623         = L_USERINFO | L_ALPHANUM | L_DASH | lowMask(".:@[]");
2624     private static final long H_SERVER
2625         = H_USERINFO | H_ALPHANUM | H_DASH | highMask(".:@[]");
2626 
2627     // Special case of server authority that represents an IPv6 address
2628     // In this case, a % does not signify an escape sequence
2629     private static final long L_SERVER_PERCENT
2630         = L_SERVER | lowMask("%");
2631     private static final long H_SERVER_PERCENT
2632         = H_SERVER | highMask("%");
2633     private static final long L_LEFT_BRACKET = lowMask("[");
2634     private static final long H_LEFT_BRACKET = highMask("[");
2635 
2636     // scheme        = alpha *( alpha | digit | "+" | "-" | "." )
2637     private static final long L_SCHEME = L_ALPHA | L_DIGIT | lowMask("+-.");
2638     private static final long H_SCHEME = H_ALPHA | H_DIGIT | highMask("+-.");
2639 
2640     // uric_no_slash = unreserved | escaped | ";" | "?" | ":" | "@" |
2641     //                 "&" | "=" | "+" | "$" | ","
2642     private static final long L_URIC_NO_SLASH
2643         = L_UNRESERVED | L_ESCAPED | lowMask(";?:@&=+$,");
2644     private static final long H_URIC_NO_SLASH
2645         = H_UNRESERVED | H_ESCAPED | highMask(";?:@&=+$,");
2646 
2647 
2648     // -- Escaping and encoding --
2649 
2650     private final static char[] hexDigits = {
2651         '0', '1', '2', '3', '4', '5', '6', '7',
2652         '8', '9', 'A', 'B', 'C', 'D', 'E', 'F'
2653     };
2654 
appendEscape(StringBuffer sb, byte b)2655     private static void appendEscape(StringBuffer sb, byte b) {
2656         sb.append('%');
2657         sb.append(hexDigits[(b >> 4) & 0x0f]);
2658         sb.append(hexDigits[(b >> 0) & 0x0f]);
2659     }
2660 
appendEncoded(StringBuffer sb, char c)2661     private static void appendEncoded(StringBuffer sb, char c) {
2662         ByteBuffer bb = null;
2663         try {
2664             bb = ThreadLocalCoders.encoderFor("UTF-8")
2665                 .encode(CharBuffer.wrap("" + c));
2666         } catch (CharacterCodingException x) {
2667             assert false;
2668         }
2669         while (bb.hasRemaining()) {
2670             int b = bb.get() & 0xff;
2671             if (b >= 0x80)
2672                 appendEscape(sb, (byte)b);
2673             else
2674                 sb.append((char)b);
2675         }
2676     }
2677 
2678     // Quote any characters in s that are not permitted
2679     // by the given mask pair
2680     //
quote(String s, long lowMask, long highMask)2681     private static String quote(String s, long lowMask, long highMask) {
2682         int n = s.length();
2683         StringBuffer sb = null;
2684         boolean allowNonASCII = ((lowMask & L_ESCAPED) != 0);
2685         for (int i = 0; i < s.length(); i++) {
2686             char c = s.charAt(i);
2687             if (c < '\u0080') {
2688                 if (!match(c, lowMask, highMask)) {
2689                     if (sb == null) {
2690                         sb = new StringBuffer();
2691                         sb.append(s.substring(0, i));
2692                     }
2693                     appendEscape(sb, (byte)c);
2694                 } else {
2695                     if (sb != null)
2696                         sb.append(c);
2697                 }
2698             } else if (allowNonASCII
2699                        && (Character.isSpaceChar(c)
2700                            || Character.isISOControl(c))) {
2701                 if (sb == null) {
2702                     sb = new StringBuffer();
2703                     sb.append(s.substring(0, i));
2704                 }
2705                 appendEncoded(sb, c);
2706             } else {
2707                 if (sb != null)
2708                     sb.append(c);
2709             }
2710         }
2711         return (sb == null) ? s : sb.toString();
2712     }
2713 
2714     // Encodes all characters >= \u0080 into escaped, normalized UTF-8 octets,
2715     // assuming that s is otherwise legal
2716     //
encode(String s)2717     private static String encode(String s) {
2718         int n = s.length();
2719         if (n == 0)
2720             return s;
2721 
2722         // First check whether we actually need to encode
2723         for (int i = 0;;) {
2724             if (s.charAt(i) >= '\u0080')
2725                 break;
2726             if (++i >= n)
2727                 return s;
2728         }
2729 
2730         String ns = Normalizer.normalize(s, Normalizer.Form.NFC);
2731         ByteBuffer bb = null;
2732         try {
2733             bb = ThreadLocalCoders.encoderFor("UTF-8")
2734                 .encode(CharBuffer.wrap(ns));
2735         } catch (CharacterCodingException x) {
2736             assert false;
2737         }
2738 
2739         StringBuffer sb = new StringBuffer();
2740         while (bb.hasRemaining()) {
2741             int b = bb.get() & 0xff;
2742             if (b >= 0x80)
2743                 appendEscape(sb, (byte)b);
2744             else
2745                 sb.append((char)b);
2746         }
2747         return sb.toString();
2748     }
2749 
decode(char c)2750     private static int decode(char c) {
2751         if ((c >= '0') && (c <= '9'))
2752             return c - '0';
2753         if ((c >= 'a') && (c <= 'f'))
2754             return c - 'a' + 10;
2755         if ((c >= 'A') && (c <= 'F'))
2756             return c - 'A' + 10;
2757         assert false;
2758         return -1;
2759     }
2760 
decode(char c1, char c2)2761     private static byte decode(char c1, char c2) {
2762         return (byte)(  ((decode(c1) & 0xf) << 4)
2763                       | ((decode(c2) & 0xf) << 0));
2764     }
2765 
2766     // Evaluates all escapes in s, applying UTF-8 decoding if needed.  Assumes
2767     // that escapes are well-formed syntactically, i.e., of the form %XX.  If a
2768     // sequence of escaped octets is not valid UTF-8 then the erroneous octets
2769     // are replaced with '\uFFFD'.
2770     // Exception: any "%" found between "[]" is left alone. It is an IPv6 literal
2771     //            with a scope_id
2772     //
decode(String s)2773     private static String decode(String s) {
2774         if (s == null)
2775             return s;
2776         int n = s.length();
2777         if (n == 0)
2778             return s;
2779         if (s.indexOf('%') < 0)
2780             return s;
2781 
2782         StringBuffer sb = new StringBuffer(n);
2783         ByteBuffer bb = ByteBuffer.allocate(n);
2784         CharBuffer cb = CharBuffer.allocate(n);
2785         CharsetDecoder dec = ThreadLocalCoders.decoderFor("UTF-8")
2786             .onMalformedInput(CodingErrorAction.REPLACE)
2787             .onUnmappableCharacter(CodingErrorAction.REPLACE);
2788 
2789         // This is not horribly efficient, but it will do for now
2790         char c = s.charAt(0);
2791         boolean betweenBrackets = false;
2792 
2793         for (int i = 0; i < n;) {
2794             assert c == s.charAt(i);    // Loop invariant
2795             if (c == '[') {
2796                 betweenBrackets = true;
2797             } else if (betweenBrackets && c == ']') {
2798                 betweenBrackets = false;
2799             }
2800             if (c != '%' || betweenBrackets) {
2801                 sb.append(c);
2802                 if (++i >= n)
2803                     break;
2804                 c = s.charAt(i);
2805                 continue;
2806             }
2807             bb.clear();
2808             int ui = i;
2809             for (;;) {
2810                 assert (n - i >= 2);
2811                 bb.put(decode(s.charAt(++i), s.charAt(++i)));
2812                 if (++i >= n)
2813                     break;
2814                 c = s.charAt(i);
2815                 if (c != '%')
2816                     break;
2817             }
2818             bb.flip();
2819             cb.clear();
2820             dec.reset();
2821             CoderResult cr = dec.decode(bb, cb, true);
2822             assert cr.isUnderflow();
2823             cr = dec.flush(cb);
2824             assert cr.isUnderflow();
2825             sb.append(cb.flip().toString());
2826         }
2827 
2828         return sb.toString();
2829     }
2830 
2831 
2832     // -- Parsing --
2833 
2834     // For convenience we wrap the input URI string in a new instance of the
2835     // following internal class.  This saves always having to pass the input
2836     // string as an argument to each internal scan/parse method.
2837 
2838     private class Parser {
2839 
2840         private String input;           // URI input string
2841         private boolean requireServerAuthority = false;
2842 
Parser(String s)2843         Parser(String s) {
2844             input = s;
2845             string = s;
2846         }
2847 
2848         // -- Methods for throwing URISyntaxException in various ways --
2849 
fail(String reason)2850         private void fail(String reason) throws URISyntaxException {
2851             throw new URISyntaxException(input, reason);
2852         }
2853 
fail(String reason, int p)2854         private void fail(String reason, int p) throws URISyntaxException {
2855             throw new URISyntaxException(input, reason, p);
2856         }
2857 
failExpecting(String expected, int p)2858         private void failExpecting(String expected, int p)
2859             throws URISyntaxException
2860         {
2861             fail("Expected " + expected, p);
2862         }
2863 
failExpecting(String expected, String prior, int p)2864         private void failExpecting(String expected, String prior, int p)
2865             throws URISyntaxException
2866         {
2867             fail("Expected " + expected + " following " + prior, p);
2868         }
2869 
2870 
2871         // -- Simple access to the input string --
2872 
2873         // Return a substring of the input string
2874         //
substring(int start, int end)2875         private String substring(int start, int end) {
2876             return input.substring(start, end);
2877         }
2878 
2879         // Return the char at position p,
2880         // assuming that p < input.length()
2881         //
charAt(int p)2882         private char charAt(int p) {
2883             return input.charAt(p);
2884         }
2885 
2886         // Tells whether start < end and, if so, whether charAt(start) == c
2887         //
at(int start, int end, char c)2888         private boolean at(int start, int end, char c) {
2889             return (start < end) && (charAt(start) == c);
2890         }
2891 
2892         // Tells whether start + s.length() < end and, if so,
2893         // whether the chars at the start position match s exactly
2894         //
at(int start, int end, String s)2895         private boolean at(int start, int end, String s) {
2896             int p = start;
2897             int sn = s.length();
2898             if (sn > end - p)
2899                 return false;
2900             int i = 0;
2901             while (i < sn) {
2902                 if (charAt(p++) != s.charAt(i)) {
2903                     break;
2904                 }
2905                 i++;
2906             }
2907             return (i == sn);
2908         }
2909 
2910 
2911         // -- Scanning --
2912 
2913         // The various scan and parse methods that follow use a uniform
2914         // convention of taking the current start position and end index as
2915         // their first two arguments.  The start is inclusive while the end is
2916         // exclusive, just as in the String class, i.e., a start/end pair
2917         // denotes the left-open interval [start, end) of the input string.
2918         //
2919         // These methods never proceed past the end position.  They may return
2920         // -1 to indicate outright failure, but more often they simply return
2921         // the position of the first char after the last char scanned.  Thus
2922         // a typical idiom is
2923         //
2924         //     int p = start;
2925         //     int q = scan(p, end, ...);
2926         //     if (q > p)
2927         //         // We scanned something
2928         //         ...;
2929         //     else if (q == p)
2930         //         // We scanned nothing
2931         //         ...;
2932         //     else if (q == -1)
2933         //         // Something went wrong
2934         //         ...;
2935 
2936 
2937         // Scan a specific char: If the char at the given start position is
2938         // equal to c, return the index of the next char; otherwise, return the
2939         // start position.
2940         //
scan(int start, int end, char c)2941         private int scan(int start, int end, char c) {
2942             if ((start < end) && (charAt(start) == c))
2943                 return start + 1;
2944             return start;
2945         }
2946 
2947         // Scan forward from the given start position.  Stop at the first char
2948         // in the err string (in which case -1 is returned), or the first char
2949         // in the stop string (in which case the index of the preceding char is
2950         // returned), or the end of the input string (in which case the length
2951         // of the input string is returned).  May return the start position if
2952         // nothing matches.
2953         //
scan(int start, int end, String err, String stop)2954         private int scan(int start, int end, String err, String stop) {
2955             int p = start;
2956             while (p < end) {
2957                 char c = charAt(p);
2958                 if (err.indexOf(c) >= 0)
2959                     return -1;
2960                 if (stop.indexOf(c) >= 0)
2961                     break;
2962                 p++;
2963             }
2964             return p;
2965         }
2966 
2967         // Scan a potential escape sequence, starting at the given position,
2968         // with the given first char (i.e., charAt(start) == c).
2969         //
2970         // This method assumes that if escapes are allowed then visible
2971         // non-US-ASCII chars are also allowed.
2972         //
scanEscape(int start, int n, char first)2973         private int scanEscape(int start, int n, char first)
2974             throws URISyntaxException
2975         {
2976             int p = start;
2977             char c = first;
2978             if (c == '%') {
2979                 // Process escape pair
2980                 if ((p + 3 <= n)
2981                     && match(charAt(p + 1), L_HEX, H_HEX)
2982                     && match(charAt(p + 2), L_HEX, H_HEX)) {
2983                     return p + 3;
2984                 }
2985                 fail("Malformed escape pair", p);
2986             } else if ((c > 128)
2987                        && !Character.isSpaceChar(c)
2988                        && !Character.isISOControl(c)) {
2989                 // Allow unescaped but visible non-US-ASCII chars
2990                 return p + 1;
2991             }
2992             return p;
2993         }
2994 
2995         // Scan chars that match the given mask pair
2996         //
scan(int start, int n, long lowMask, long highMask)2997         private int scan(int start, int n, long lowMask, long highMask)
2998             throws URISyntaxException
2999         {
3000             int p = start;
3001             while (p < n) {
3002                 char c = charAt(p);
3003                 if (match(c, lowMask, highMask)) {
3004                     p++;
3005                     continue;
3006                 }
3007                 if ((lowMask & L_ESCAPED) != 0) {
3008                     int q = scanEscape(p, n, c);
3009                     if (q > p) {
3010                         p = q;
3011                         continue;
3012                     }
3013                 }
3014                 break;
3015             }
3016             return p;
3017         }
3018 
3019         // Check that each of the chars in [start, end) matches the given mask
3020         //
checkChars(int start, int end, long lowMask, long highMask, String what)3021         private void checkChars(int start, int end,
3022                                 long lowMask, long highMask,
3023                                 String what)
3024             throws URISyntaxException
3025         {
3026             int p = scan(start, end, lowMask, highMask);
3027             if (p < end)
3028                 fail("Illegal character in " + what, p);
3029         }
3030 
3031         // Check that the char at position p matches the given mask
3032         //
checkChar(int p, long lowMask, long highMask, String what)3033         private void checkChar(int p,
3034                                long lowMask, long highMask,
3035                                String what)
3036             throws URISyntaxException
3037         {
3038             checkChars(p, p + 1, lowMask, highMask, what);
3039         }
3040 
3041 
3042         // -- Parsing --
3043 
3044         // [<scheme>:]<scheme-specific-part>[#<fragment>]
3045         //
parse(boolean rsa)3046         void parse(boolean rsa) throws URISyntaxException {
3047             requireServerAuthority = rsa;
3048             int ssp;                    // Start of scheme-specific part
3049             int n = input.length();
3050             int p = scan(0, n, "/?#", ":");
3051             if ((p >= 0) && at(p, n, ':')) {
3052                 if (p == 0)
3053                     failExpecting("scheme name", 0);
3054                 checkChar(0, L_ALPHA, H_ALPHA, "scheme name");
3055                 checkChars(1, p, L_SCHEME, H_SCHEME, "scheme name");
3056                 scheme = substring(0, p);
3057                 p++;                    // Skip ':'
3058                 ssp = p;
3059                 if (at(p, n, '/')) {
3060                     p = parseHierarchical(p, n);
3061                 } else {
3062                     int q = scan(p, n, "", "#");
3063                     if (q <= p)
3064                         failExpecting("scheme-specific part", p);
3065                     checkChars(p, q, L_URIC, H_URIC, "opaque part");
3066                     p = q;
3067                 }
3068             } else {
3069                 ssp = 0;
3070                 p = parseHierarchical(0, n);
3071             }
3072             schemeSpecificPart = substring(ssp, p);
3073             if (at(p, n, '#')) {
3074                 checkChars(p + 1, n, L_URIC, H_URIC, "fragment");
3075                 fragment = substring(p + 1, n);
3076                 p = n;
3077             }
3078             if (p < n)
3079                 fail("end of URI", p);
3080         }
3081 
3082         // [//authority]<path>[?<query>]
3083         //
3084         // DEVIATION from RFC2396: We allow an empty authority component as
3085         // long as it's followed by a non-empty path, query component, or
3086         // fragment component.  This is so that URIs such as "file:///foo/bar"
3087         // will parse.  This seems to be the intent of RFC2396, though the
3088         // grammar does not permit it.  If the authority is empty then the
3089         // userInfo, host, and port components are undefined.
3090         //
3091         // DEVIATION from RFC2396: We allow empty relative paths.  This seems
3092         // to be the intent of RFC2396, but the grammar does not permit it.
3093         // The primary consequence of this deviation is that "#f" parses as a
3094         // relative URI with an empty path.
3095         //
parseHierarchical(int start, int n)3096         private int parseHierarchical(int start, int n)
3097             throws URISyntaxException
3098         {
3099             int p = start;
3100             if (at(p, n, '/') && at(p + 1, n, '/')) {
3101                 p += 2;
3102                 int q = scan(p, n, "", "/?#");
3103                 if (q > p) {
3104                     p = parseAuthority(p, q);
3105                 } else if (q < n) {
3106                     // DEVIATION: Allow empty authority prior to non-empty
3107                     // path, query component or fragment identifier
3108                 } else
3109                     failExpecting("authority", p);
3110             }
3111             int q = scan(p, n, "", "?#"); // DEVIATION: May be empty
3112             checkChars(p, q, L_PATH, H_PATH, "path");
3113             path = substring(p, q);
3114             p = q;
3115             if (at(p, n, '?')) {
3116                 p++;
3117                 q = scan(p, n, "", "#");
3118                 checkChars(p, q, L_URIC, H_URIC, "query");
3119                 query = substring(p, q);
3120                 p = q;
3121             }
3122             return p;
3123         }
3124 
3125         // authority     = server | reg_name
3126         //
3127         // Ambiguity: An authority that is a registry name rather than a server
3128         // might have a prefix that parses as a server.  We use the fact that
3129         // the authority component is always followed by '/' or the end of the
3130         // input string to resolve this: If the complete authority did not
3131         // parse as a server then we try to parse it as a registry name.
3132         //
parseAuthority(int start, int n)3133         private int parseAuthority(int start, int n)
3134             throws URISyntaxException
3135         {
3136             int p = start;
3137             int q = p;
3138             URISyntaxException ex = null;
3139 
3140             boolean serverChars;
3141             boolean regChars;
3142 
3143             if (scan(p, n, "", "]") > p) {
3144                 // contains a literal IPv6 address, therefore % is allowed
3145                 serverChars = (scan(p, n, L_SERVER_PERCENT, H_SERVER_PERCENT) == n);
3146             } else {
3147                 serverChars = (scan(p, n, L_SERVER, H_SERVER) == n);
3148             }
3149             regChars = (scan(p, n, L_REG_NAME, H_REG_NAME) == n);
3150 
3151             if (regChars && !serverChars) {
3152                 // Must be a registry-based authority
3153                 authority = substring(p, n);
3154                 return n;
3155             }
3156 
3157             if (serverChars) {
3158                 // Might be (probably is) a server-based authority, so attempt
3159                 // to parse it as such.  If the attempt fails, try to treat it
3160                 // as a registry-based authority.
3161                 try {
3162                     q = parseServer(p, n);
3163                     if (q < n)
3164                         failExpecting("end of authority", q);
3165                     authority = substring(p, n);
3166                 } catch (URISyntaxException x) {
3167                     // Undo results of failed parse
3168                     userInfo = null;
3169                     host = null;
3170                     port = -1;
3171                     if (requireServerAuthority) {
3172                         // If we're insisting upon a server-based authority,
3173                         // then just re-throw the exception
3174                         throw x;
3175                     } else {
3176                         // Save the exception in case it doesn't parse as a
3177                         // registry either
3178                         ex = x;
3179                         q = p;
3180                     }
3181                 }
3182             }
3183 
3184             if (q < n) {
3185                 if (regChars) {
3186                     // Registry-based authority
3187                     authority = substring(p, n);
3188                 } else if (ex != null) {
3189                     // Re-throw exception; it was probably due to
3190                     // a malformed IPv6 address
3191                     throw ex;
3192                 } else {
3193                     fail("Illegal character in authority", q);
3194                 }
3195             }
3196 
3197             return n;
3198         }
3199 
3200 
3201         // [<userinfo>@]<host>[:<port>]
3202         //
parseServer(int start, int n)3203         private int parseServer(int start, int n)
3204             throws URISyntaxException
3205         {
3206             int p = start;
3207             int q;
3208 
3209             // userinfo
3210             q = scan(p, n, "/?#", "@");
3211             if ((q >= p) && at(q, n, '@')) {
3212                 checkChars(p, q, L_USERINFO, H_USERINFO, "user info");
3213                 userInfo = substring(p, q);
3214                 p = q + 1;              // Skip '@'
3215             }
3216 
3217             // hostname, IPv4 address, or IPv6 address
3218             if (at(p, n, '[')) {
3219                 // DEVIATION from RFC2396: Support IPv6 addresses, per RFC2732
3220                 p++;
3221                 q = scan(p, n, "/?#", "]");
3222                 if ((q > p) && at(q, n, ']')) {
3223                     // look for a "%" scope id
3224                     int r = scan (p, q, "", "%");
3225                     if (r > p) {
3226                         parseIPv6Reference(p, r);
3227                         if (r+1 == q) {
3228                             fail ("scope id expected");
3229                         }
3230                         checkChars (r+1, q, L_ALPHANUM, H_ALPHANUM,
3231                                                 "scope id");
3232                     } else {
3233                         parseIPv6Reference(p, q);
3234                     }
3235                     host = substring(p-1, q+1);
3236                     p = q + 1;
3237                 } else {
3238                     failExpecting("closing bracket for IPv6 address", q);
3239                 }
3240             } else {
3241                 q = parseIPv4Address(p, n);
3242                 if (q <= p)
3243                     q = parseHostname(p, n);
3244                 p = q;
3245             }
3246 
3247             // port
3248             if (at(p, n, ':')) {
3249                 p++;
3250                 q = scan(p, n, "", "/");
3251                 if (q > p) {
3252                     checkChars(p, q, L_DIGIT, H_DIGIT, "port number");
3253                     try {
3254                         port = Integer.parseInt(substring(p, q));
3255                     } catch (NumberFormatException x) {
3256                         fail("Malformed port number", p);
3257                     }
3258                     p = q;
3259                 }
3260             }
3261             if (p < n)
3262                 failExpecting("port number", p);
3263 
3264             return p;
3265         }
3266 
3267         // Scan a string of decimal digits whose value fits in a byte
3268         //
scanByte(int start, int n)3269         private int scanByte(int start, int n)
3270             throws URISyntaxException
3271         {
3272             int p = start;
3273             int q = scan(p, n, L_DIGIT, H_DIGIT);
3274             if (q <= p) return q;
3275             if (Integer.parseInt(substring(p, q)) > 255) return p;
3276             return q;
3277         }
3278 
3279         // Scan an IPv4 address.
3280         //
3281         // If the strict argument is true then we require that the given
3282         // interval contain nothing besides an IPv4 address; if it is false
3283         // then we only require that it start with an IPv4 address.
3284         //
3285         // If the interval does not contain or start with (depending upon the
3286         // strict argument) a legal IPv4 address characters then we return -1
3287         // immediately; otherwise we insist that these characters parse as a
3288         // legal IPv4 address and throw an exception on failure.
3289         //
3290         // We assume that any string of decimal digits and dots must be an IPv4
3291         // address.  It won't parse as a hostname anyway, so making that
3292         // assumption here allows more meaningful exceptions to be thrown.
3293         //
scanIPv4Address(int start, int n, boolean strict)3294         private int scanIPv4Address(int start, int n, boolean strict)
3295             throws URISyntaxException
3296         {
3297             int p = start;
3298             int q;
3299             int m = scan(p, n, L_DIGIT | L_DOT, H_DIGIT | H_DOT);
3300             if ((m <= p) || (strict && (m != n)))
3301                 return -1;
3302             for (;;) {
3303                 // Per RFC2732: At most three digits per byte
3304                 // Further constraint: Each element fits in a byte
3305                 if ((q = scanByte(p, m)) <= p) break;   p = q;
3306                 if ((q = scan(p, m, '.')) <= p) break;  p = q;
3307                 if ((q = scanByte(p, m)) <= p) break;   p = q;
3308                 if ((q = scan(p, m, '.')) <= p) break;  p = q;
3309                 if ((q = scanByte(p, m)) <= p) break;   p = q;
3310                 if ((q = scan(p, m, '.')) <= p) break;  p = q;
3311                 if ((q = scanByte(p, m)) <= p) break;   p = q;
3312                 if (q < m) break;
3313                 return q;
3314             }
3315             fail("Malformed IPv4 address", q);
3316             return -1;
3317         }
3318 
3319         // Take an IPv4 address: Throw an exception if the given interval
3320         // contains anything except an IPv4 address
3321         //
takeIPv4Address(int start, int n, String expected)3322         private int takeIPv4Address(int start, int n, String expected)
3323             throws URISyntaxException
3324         {
3325             int p = scanIPv4Address(start, n, true);
3326             if (p <= start)
3327                 failExpecting(expected, start);
3328             return p;
3329         }
3330 
3331         // Attempt to parse an IPv4 address, returning -1 on failure but
3332         // allowing the given interval to contain [:<characters>] after
3333         // the IPv4 address.
3334         //
parseIPv4Address(int start, int n)3335         private int parseIPv4Address(int start, int n) {
3336             int p;
3337 
3338             try {
3339                 p = scanIPv4Address(start, n, false);
3340             } catch (URISyntaxException x) {
3341                 return -1;
3342             } catch (NumberFormatException nfe) {
3343                 return -1;
3344             }
3345 
3346             if (p > start && p < n) {
3347                 // IPv4 address is followed by something - check that
3348                 // it's a ":" as this is the only valid character to
3349                 // follow an address.
3350                 if (charAt(p) != ':') {
3351                     p = -1;
3352                 }
3353             }
3354 
3355             if (p > start)
3356                 host = substring(start, p);
3357 
3358             return p;
3359         }
3360 
3361         // hostname      = domainlabel [ "." ] | 1*( domainlabel "." ) toplabel [ "." ]
3362         // domainlabel   = alphanum | alphanum *( alphanum | "-" | "_" ) alphanum
3363         // toplabel      = alpha | alpha *( alphanum | "-" | "_" ) alphanum
3364         //
parseHostname(int start, int n)3365         private int parseHostname(int start, int n)
3366             throws URISyntaxException
3367         {
3368             int p = start;
3369             int q;
3370             int l = -1;                 // Start of last parsed label
3371 
3372             do {
3373                 // domainlabel = alphanum [ *( alphanum | "-" | "_" ) alphanum ]
3374 
3375                 // RFC1034#section-3.5 doesn't permit empty labels in hostnames, but we accepted
3376                 // this prior to N and the behavior is used by some apps. They're accepted for
3377                 // compatibility but we produce a warning in the log.
3378                 // http://b/25991669
3379                 if (p < n && charAt(p) == '.') {
3380                   java.lang.System.logE("URI " + substring(start, n) +  " has empty labels in " +
3381                                         "the hostname. This is malformed and will not be accepted" +
3382                                         "in future Android releases.");
3383                   q = ++p;
3384                   continue;
3385                 }
3386 
3387                 // The RFCs don't permit underscores in hostnames, but URI has to because a certain
3388                 // large website doesn't seem to care about standards and specs.
3389                 // http://code.google.com/p/android/issues/detail?id=37577
3390                 // http://b/17579865
3391                 // http://b/18016625
3392                 // http://b/18023709
3393                 q = scan(p, n, L_ALPHANUM, H_ALPHANUM);
3394                 if (q <= p)
3395                     break;
3396                 l = p;
3397                 if (q > p) {
3398                     p = q;
3399                     q = scan(p, n, L_ALPHANUM | L_DASH | L_UNDERSCORE, H_ALPHANUM | H_DASH | H_UNDERSCORE);
3400                     if (q > p) {
3401                         if (charAt(q - 1) == '-')
3402                             fail("Illegal character in hostname", q - 1);
3403                         p = q;
3404                     }
3405                 }
3406                 q = scan(p, n, '.');
3407                 if (q <= p)
3408                     break;
3409                 p = q;
3410             } while (p < n);
3411 
3412             if ((p < n) && !at(p, n, ':'))
3413                 fail("Illegal character in hostname", p);
3414 
3415             if (l < 0)
3416                 failExpecting("hostname", start);
3417 
3418             // for a fully qualified hostname check that the rightmost
3419             // label starts with an alpha character.
3420             if (l > start && !match(charAt(l), L_ALPHA, H_ALPHA)) {
3421                 fail("Illegal character in hostname", l);
3422             }
3423 
3424             host = substring(start, p);
3425             return p;
3426         }
3427 
3428 
3429         // IPv6 address parsing, from RFC2373: IPv6 Addressing Architecture
3430         //
3431         // Bug: The grammar in RFC2373 Appendix B does not allow addresses of
3432         // the form ::12.34.56.78, which are clearly shown in the examples
3433         // earlier in the document.  Here is the original grammar:
3434         //
3435         //   IPv6address = hexpart [ ":" IPv4address ]
3436         //   hexpart     = hexseq | hexseq "::" [ hexseq ] | "::" [ hexseq ]
3437         //   hexseq      = hex4 *( ":" hex4)
3438         //   hex4        = 1*4HEXDIG
3439         //
3440         // We therefore use the following revised grammar:
3441         //
3442         //   IPv6address = hexseq [ ":" IPv4address ]
3443         //                 | hexseq [ "::" [ hexpost ] ]
3444         //                 | "::" [ hexpost ]
3445         //   hexpost     = hexseq | hexseq ":" IPv4address | IPv4address
3446         //   hexseq      = hex4 *( ":" hex4)
3447         //   hex4        = 1*4HEXDIG
3448         //
3449         // This covers all and only the following cases:
3450         //
3451         //   hexseq
3452         //   hexseq : IPv4address
3453         //   hexseq ::
3454         //   hexseq :: hexseq
3455         //   hexseq :: hexseq : IPv4address
3456         //   hexseq :: IPv4address
3457         //   :: hexseq
3458         //   :: hexseq : IPv4address
3459         //   :: IPv4address
3460         //   ::
3461         //
3462         // Additionally we constrain the IPv6 address as follows :-
3463         //
3464         //  i.  IPv6 addresses without compressed zeros should contain
3465         //      exactly 16 bytes.
3466         //
3467         //  ii. IPv6 addresses with compressed zeros should contain
3468         //      less than 16 bytes.
3469 
3470         private int ipv6byteCount = 0;
3471 
parseIPv6Reference(int start, int n)3472         private int parseIPv6Reference(int start, int n)
3473             throws URISyntaxException
3474         {
3475             int p = start;
3476             int q;
3477             boolean compressedZeros = false;
3478 
3479             q = scanHexSeq(p, n);
3480 
3481             if (q > p) {
3482                 p = q;
3483                 if (at(p, n, "::")) {
3484                     compressedZeros = true;
3485                     p = scanHexPost(p + 2, n);
3486                 } else if (at(p, n, ':')) {
3487                     p = takeIPv4Address(p + 1,  n, "IPv4 address");
3488                     ipv6byteCount += 4;
3489                 }
3490             } else if (at(p, n, "::")) {
3491                 compressedZeros = true;
3492                 p = scanHexPost(p + 2, n);
3493             }
3494             if (p < n)
3495                 fail("Malformed IPv6 address", start);
3496             if (ipv6byteCount > 16)
3497                 fail("IPv6 address too long", start);
3498             if (!compressedZeros && ipv6byteCount < 16)
3499                 fail("IPv6 address too short", start);
3500             if (compressedZeros && ipv6byteCount == 16)
3501                 fail("Malformed IPv6 address", start);
3502 
3503             return p;
3504         }
3505 
scanHexPost(int start, int n)3506         private int scanHexPost(int start, int n)
3507             throws URISyntaxException
3508         {
3509             int p = start;
3510             int q;
3511 
3512             if (p == n)
3513                 return p;
3514 
3515             q = scanHexSeq(p, n);
3516             if (q > p) {
3517                 p = q;
3518                 if (at(p, n, ':')) {
3519                     p++;
3520                     p = takeIPv4Address(p, n, "hex digits or IPv4 address");
3521                     ipv6byteCount += 4;
3522                 }
3523             } else {
3524                 p = takeIPv4Address(p, n, "hex digits or IPv4 address");
3525                 ipv6byteCount += 4;
3526             }
3527             return p;
3528         }
3529 
3530         // Scan a hex sequence; return -1 if one could not be scanned
3531         //
scanHexSeq(int start, int n)3532         private int scanHexSeq(int start, int n)
3533             throws URISyntaxException
3534         {
3535             int p = start;
3536             int q;
3537 
3538             q = scan(p, n, L_HEX, H_HEX);
3539             if (q <= p)
3540                 return -1;
3541             if (at(q, n, '.'))          // Beginning of IPv4 address
3542                 return -1;
3543             if (q > p + 4)
3544                 fail("IPv6 hexadecimal digit sequence too long", p);
3545             ipv6byteCount += 2;
3546             p = q;
3547             while (p < n) {
3548                 if (!at(p, n, ':'))
3549                     break;
3550                 if (at(p + 1, n, ':'))
3551                     break;              // "::"
3552                 p++;
3553                 q = scan(p, n, L_HEX, H_HEX);
3554                 if (q <= p)
3555                     failExpecting("digits for an IPv6 address", p);
3556                 if (at(q, n, '.')) {    // Beginning of IPv4 address
3557                     p--;
3558                     break;
3559                 }
3560                 if (q > p + 4)
3561                     fail("IPv6 hexadecimal digit sequence too long", p);
3562                 ipv6byteCount += 2;
3563                 p = q;
3564             }
3565 
3566             return p;
3567         }
3568 
3569     }
3570 
3571 }
3572