Source Home >> Java Source 1.6.0 >> java.net.URI V 0.09

0001/*
0002 * @(#)URI.java	1.48 06/06/12
0003 *
0004 * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
0005 * SUN PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
0006 */
0007
0008package java.net;
0009
0010import java.io.IOException;
0011import java.io.InvalidObjectException;
0012import java.io.ObjectInputStream;
0013import java.io.ObjectOutputStream;
0014import java.io.Serializable;
0015import java.nio.ByteBuffer;
0016import java.nio.CharBuffer;
0017import java.nio.charset.CharsetDecoder;
0018import java.nio.charset.CharsetEncoder;
0019import java.nio.charset.CoderResult;
0020import java.nio.charset.CodingErrorAction;
0021import java.nio.charset.CharacterCodingException;
0022import java.text.Normalizer;
0023import sun.nio.cs.ThreadLocalCoders;
0024
0025import java.lang.Character;		// for javadoc
0026import java.lang.NullPointerException;	// for javadoc
0027
0028
0029/**
0030 * Represents a Uniform Resource Identifier (URI) reference.
0031 *
0032 * <p> Aside from some minor deviations noted below, an instance of this 
0033 * class represents a URI reference as defined by
0034 * <a href="http://www.ietf.org/rfc/rfc2396.txt""><i>RFC 2396: Uniform
0035 * Resource Identifiers (URI): Generic Syntax</i></a>, amended by <a
0036 * href="http://www.ietf.org/rfc/rfc2732.txt"><i>RFC 2732: Format for
0037 * Literal IPv6 Addresses in URLs</i></a>. The Literal IPv6 address format
0038 * also supports scope_ids. The syntax and usage of scope_ids is described
0039 * <a href="Inet6Address.html#scoped">here</a>.
0040 * This class provides constructors for creating URI instances from
0041 * their components or by parsing their string forms, methods for accessing the
0042 * various components of an instance, and methods for normalizing, resolving,
0043 * and relativizing URI instances.  Instances of this class are immutable.
0044 *
0045 *
0046 * <h4> URI syntax and components </h4>
0047 *
0048 * At the highest level a URI reference (hereinafter simply "URI") in string
0049 * form has the syntax
0050 *
0051 * <blockquote>
0052 * [<i>scheme</i><tt><b>:</b></tt><i></i>]<i>scheme-specific-part</i>[<tt><b>#</b></tt><i>fragment</i>]
0053 * </blockquote>
0054 *
0055 * where square brackets [...] delineate optional components and the characters
0056 * <tt><b>:</b></tt> and <tt><b>#</b></tt> stand for themselves.
0057 *
0058 * <p> An <i>absolute</i> URI specifies a scheme; a URI that is not absolute is
0059 * said to be <i>relative</i>.  URIs are also classified according to whether
0060 * they are <i>opaque</i> or <i>hierarchical</i>.
0061 *
0062 * <p> An <i>opaque</i> URI is an absolute URI whose scheme-specific part does
0063 * not begin with a slash character (<tt>'/'</tt>).  Opaque URIs are not
0064 * subject to further parsing.  Some examples of opaque URIs are:
0065 *
0066 * <blockquote><table cellpadding=0 cellspacing=0 summary="layout">
0067 * <tr><td><tt>mailto:java-net@java.sun.com</tt><td></tr>
0068 * <tr><td><tt>news:comp.lang.java</tt><td></tr>
0069 * <tr><td><tt>urn:isbn:096139210x</tt></td></tr>
0070 * </table></blockquote>
0071 *
0072 * <p> A <i>hierarchical</i> URI is either an absolute URI whose
0073 * scheme-specific part begins with a slash character, or a relative URI, that
0074 * is, a URI that does not specify a scheme.  Some examples of hierarchical
0075 * URIs are:
0076 *
0077 * <blockquote>
0078 * <tt>http://java.sun.com/j2se/1.3/</tt><br>
0079 * <tt>docs/guide/collections/designfaq.html#28</tt><br>
0080 * <tt>../../../demo/jfc/SwingSet2/src/SwingSet2.java</tt><br>
0081 * <tt>file:///~/calendar</tt>
0082 * </blockquote>
0083 *
0084 * <p> A hierarchical URI is subject to further parsing according to the syntax
0085 *
0086 * <blockquote>
0087 * [<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>]
0088 * </blockquote>
0089 *
0090 * where the characters <tt><b>:</b></tt>, <tt><b>/</b></tt>,
0091 * <tt><b>?</b></tt>, and <tt><b>#</b></tt> stand for themselves.  The
0092 * scheme-specific part of a hierarchical URI consists of the characters
0093 * between the scheme and fragment components.
0094 *
0095 * <p> The authority component of a hierarchical URI is, if specified, either
0096 * <i>server-based</i> or <i>registry-based</i>.  A server-based authority
0097 * parses according to the familiar syntax
0098 *
0099 * <blockquote>
0100 * [<i>user-info</i><tt><b>@</b></tt>]<i>host</i>[<tt><b>:</b></tt><i>port</i>]
0101 * </blockquote>
0102 *
0103 * where the characters <tt><b>@</b></tt> and <tt><b>:</b></tt> stand for
0104 * themselves.  Nearly all URI schemes currently in use are server-based.  An
0105 * authority component that does not parse in this way is considered to be
0106 * registry-based.
0107 *
0108 * <p> The path component of a hierarchical URI is itself said to be absolute
0109 * if it begins with a slash character (<tt>'/'</tt>); otherwise it is
0110 * relative.  The path of a hierarchical URI that is either absolute or
0111 * specifies an authority is always absolute.
0112 *
0113 * <p> All told, then, a URI instance has the following nine components:
0114 *
0115 * <blockquote><table summary="Describes the components of a URI:scheme,scheme-specific-part,authority,user-info,host,port,path,query,fragment">
0116 * <tr><th><i>Component</i></th><th><i>Type</i></th></tr>
0117 * <tr><td>scheme</td><td><tt>String</tt></td></tr>
0118 * <tr><td>scheme-specific-part    </td><td><tt>String</tt></td></tr>
0119 * <tr><td>authority</td><td><tt>String</tt></td></tr>
0120 * <tr><td>user-info</td><td><tt>String</tt></td></tr>
0121 * <tr><td>host</td><td><tt>String</tt></td></tr>
0122 * <tr><td>port</td><td><tt>int</tt></td></tr>
0123 * <tr><td>path</td><td><tt>String</tt></td></tr>
0124 * <tr><td>query</td><td><tt>String</tt></td></tr>
0125 * <tr><td>fragment</td><td><tt>String</tt></td></tr>
0126 * </table></blockquote>
0127 *
0128 * In a given instance any particular component is either <i>undefined</i> or
0129 * <i>defined</i> with a distinct value.  Undefined string components are
0130 * represented by <tt>null</tt>, while undefined integer components are
0131 * represented by <tt>-1</tt>.  A string component may be defined to have the
0132 * empty string as its value; this is not equivalent to that component being
0133 * undefined.
0134 *
0135 * <p> Whether a particular component is or is not defined in an instance
0136 * depends upon the type of the URI being represented.  An absolute URI has a
0137 * scheme component.  An opaque URI has a scheme, a scheme-specific part, and
0138 * possibly a fragment, but has no other components.  A hierarchical URI always
0139 * has a path (though it may be empty) and a scheme-specific-part (which at
0140 * least contains the path), and may have any of the other components.  If the
0141 * authority component is present and is server-based then the host component
0142 * will be defined and the user-information and port components may be defined.
0143 *
0144 *
0145 * <h4> Operations on URI instances </h4>
0146 *
0147 * The key operations supported by this class are those of
0148 * <i>normalization</i>, <i>resolution</i>, and <i>relativization</i>.
0149 *
0150 * <p> <i>Normalization</i> is the process of removing unnecessary <tt>"."</tt>
0151 * and <tt>".."</tt> segments from the path component of a hierarchical URI.
0152 * Each <tt>"."</tt> segment is simply removed.  A <tt>".."</tt> segment is
0153 * removed only if it is preceded by a non-<tt>".."</tt> segment.
0154 * Normalization has no effect upon opaque URIs.
0155 *
0156 * <p> <i>Resolution</i> is the process of resolving one URI against another,
0157 * <i>base</i> URI.  The resulting URI is constructed from components of both
0158 * URIs in the manner specified by RFC 2396, taking components from the
0159 * base URI for those not specified in the original.  For hierarchical URIs,
0160 * the path of the original is resolved against the path of the base and then
0161 * normalized.  The result, for example, of resolving
0162 *
0163 * <blockquote>
0164 * <tt>docs/guide/collections/designfaq.html#28          </tt>(1)
0165 * </blockquote>
0166 *
0167 * against the base URI <tt>http://java.sun.com/j2se/1.3/</tt> is the result
0168 * URI
0169 *
0170 * <blockquote>
0171 * <tt>http://java.sun.com/j2se/1.3/docs/guide/collections/designfaq.html#28</tt>
0172 * </blockquote>
0173 *
0174 * Resolving the relative URI
0175 *
0176 * <blockquote>
0177 * <tt>../../../demo/jfc/SwingSet2/src/SwingSet2.java    </tt>(2)
0178 * </blockquote>
0179 *
0180 * against this result yields, in turn,
0181 *
0182 * <blockquote>
0183 * <tt>http://java.sun.com/j2se/1.3/demo/jfc/SwingSet2/src/SwingSet2.java</tt>
0184 * </blockquote>
0185 *
0186 * Resolution of both absolute and relative URIs, and of both absolute and
0187 * relative paths in the case of hierarchical URIs, is supported.  Resolving
0188 * the URI <tt>file:///~calendar</tt> against any other URI simply yields the
0189 * original URI, since it is absolute.  Resolving the relative URI (2) above
0190 * against the relative base URI (1) yields the normalized, but still relative,
0191 * URI
0192 *
0193 * <blockquote>
0194 * <tt>demo/jfc/SwingSet2/src/SwingSet2.java</tt>
0195 * </blockquote>
0196 *
0197 * <p> <i>Relativization</i>, finally, is the inverse of resolution: For any
0198 * two normalized URIs <i>u</i> and <i>v</i>,
0199 *
0200 * <blockquote>
0201 *   <i>u</i><tt>.relativize(</tt><i>u</i><tt>.resolve(</tt><i>v</i><tt>)).equals(</tt><i>v</i><tt>)</tt>  and<br>
0202 *   <i>u</i><tt>.resolve(</tt><i>u</i><tt>.relativize(</tt><i>v</i><tt>)).equals(</tt><i>v</i><tt>)</tt>  .<br>
0203 * </blockquote>
0204 *
0205 * This operation is often useful when constructing a document containing URIs
0206 * that must be made relative to the base URI of the document wherever
0207 * possible.  For example, relativizing the URI
0208 *
0209 * <blockquote>
0210 * <tt>http://java.sun.com/j2se/1.3/docs/guide/index.html</tt>
0211 * </blockquote>
0212 *
0213 * against the base URI
0214 *
0215 * <blockquote>
0216 * <tt>http://java.sun.com/j2se/1.3</tt>
0217 * </blockquote>
0218 *
0219 * yields the relative URI <tt>docs/guide/index.html</tt>.
0220 *
0221 *
0222 * <h4> Character categories </h4>
0223 *
0224 * RFC 2396 specifies precisely which characters are permitted in the
0225 * various components of a URI reference.  The following categories, most of
0226 * which are taken from that specification, are used below to describe these
0227 * constraints:
0228 *
0229 * <blockquote><table cellspacing=2 summary="Describes categories alpha,digit,alphanum,unreserved,punct,reserved,escaped,and other">
0230 *   <tr><th valign=top><i>alpha</i></th>
0231 *       <td>The US-ASCII alphabetic characters,
0232 * 	  <tt>'A'</tt> through <tt>'Z'</tt>
0233 * 	  and <tt>'a'</tt> through <tt>'z'</tt></td></tr>
0234 *   <tr><th valign=top><i>digit</i></th>
0235 *       <td>The US-ASCII decimal digit characters,
0236 *       <tt>'0'</tt> through <tt>'9'</tt></td></tr>
0237 *   <tr><th valign=top><i>alphanum</i></th>
0238 *       <td>All <i>alpha</i> and <i>digit</i> characters</td></tr>
0239 *   <tr><th valign=top><i>unreserved</i>    </th>
0240 *       <td>All <i>alphanum</i> characters together with those in the string
0241 * 	  <tt>"_-!.~'()*"</tt></td></tr>
0242 *   <tr><th valign=top><i>punct</i></th>
0243 *       <td>The characters in the string <tt>",;:$&+="</tt></td></tr>
0244 *   <tr><th valign=top><i>reserved</i></th>
0245 *       <td>All <i>punct</i> characters together with those in the string
0246 * 	  <tt>"?/[]@"</tt></td></tr>
0247 *   <tr><th valign=top><i>escaped</i></th>
0248 *       <td>Escaped octets, that is, triplets consisting of the percent
0249 *           character (<tt>'%'</tt>) followed by two hexadecimal digits
0250 *           (<tt>'0'</tt>-<tt>'9'</tt>, <tt>'A'</tt>-<tt>'F'</tt>, and
0251 *           <tt>'a'</tt>-<tt>'f'</tt>)</td></tr>
0252 *   <tr><th valign=top><i>other</i></th>
0253 *       <td>The Unicode characters that are not in the US-ASCII character set,
0254 *           are not control characters (according to the {@link
0255 *           java.lang.Character#isISOControl(char) Character.isISOControl}
0256 * 	     method), and are not space characters (according to the {@link
0257 * 	     java.lang.Character#isSpaceChar(char) Character.isSpaceChar}
0258 * 	     method)  <i>(<b>Deviation from RFC 2396</b>, which is
0259 * 	     limited to US-ASCII)</i></td></tr>
0260 * </table></blockquote>
0261 *
0262 * <p><a name="legal-chars"></a> The set of all legal URI characters consists of
0263 * the <i>unreserved</i>, <i>reserved</i>, <i>escaped</i>, and <i>other</i>
0264 * characters.
0265 *
0266 *
0267 * <h4> Escaped octets, quotation, encoding, and decoding </h4>
0268 *
0269 * RFC 2396 allows escaped octets to appear in the user-info, path, query, and
0270 * fragment components.  Escaping serves two purposes in URIs:
0271 *
0272 * <ul>
0273 *
0274 *   <li><p> To <i>encode</i> non-US-ASCII characters when a URI is required to
0275 *   conform strictly to RFC 2396 by not containing any <i>other</i>
0276 *   characters.  </p></li>
0277 *
0278 *   <li><p> To <i>quote</i> characters that are otherwise illegal in a
0279 *   component.  The user-info, path, query, and fragment components differ
0280 *   slightly in terms of which characters are considered legal and illegal.
0281 *   </p></li>
0282 *
0283 * </ul>
0284 *
0285 * These purposes are served in this class by three related operations:
0286 *
0287 * <ul>
0288 *
0289 *   <li><p><a name="encode"></a> A character is <i>encoded</i> by replacing it
0290 *   with the sequence of escaped octets that represent that character in the
0291 *   UTF-8 character set.  The Euro currency symbol (<tt>'\u20AC'</tt>),
0292 *   for example, is encoded as <tt>"%E2%82%AC"</tt>.  <i>(<b>Deviation from
0293 *   RFC 2396</b>, which does not specify any particular character
0294 *   set.)</i> </p></li>
0295 *
0296 *   <li><p><a name="quote"></a> An illegal character is <i>quoted</i> simply by
0297 *   encoding it.  The space character, for example, is quoted by replacing it
0298 *   with <tt>"%20"</tt>.  UTF-8 contains US-ASCII, hence for US-ASCII
0299 *   characters this transformation has exactly the effect required by
0300 *   RFC 2396. </p></li>
0301 *
0302 *   <li><p><a name="decode"></a>
0303 *   A sequence of escaped octets is <i>decoded</i> by
0304 *   replacing it with the sequence of characters that it represents in the
0305 *   UTF-8 character set.  UTF-8 contains US-ASCII, hence decoding has the
0306 *   effect of de-quoting any quoted US-ASCII characters as well as that of
0307 *   decoding any encoded non-US-ASCII characters.  If a <a
0308 *   href="../nio/charset/CharsetDecoder.html#ce">decoding error</a> occurs
0309 *   when decoding the escaped octets then the erroneous octets are replaced by
0310 *   <tt>'\uFFFD'</tt>, the Unicode replacement character.  </p></li>
0311 *
0312 * </ul>
0313 *
0314 * These operations are exposed in the constructors and methods of this class
0315 * as follows:
0316 *
0317 * <ul>
0318 *
0319 *   <li><p> The {@link #URI(java.lang.String) <code>single-argument
0320 *   constructor</code>} requires any illegal characters in its argument to be
0321 *   quoted and preserves any escaped octets and <i>other</i> characters that
0322 *   are present.  </p></li>
0323 *
0324 *   <li><p> The {@link
0325 *   #URI(java.lang.String,java.lang.String,java.lang.String,int,java.lang.String,java.lang.String,java.lang.String)
0326 *   <code>multi-argument constructors</code>} quote illegal characters as
0327 *   required by the components in which they appear.  The percent character
0328 *   (<tt>'%'</tt>) is always quoted by these constructors.  Any <i>other</i>
0329 *   characters are preserved.  </p></li>
0330 *
0331 *   <li><p> The {@link #getRawUserInfo() getRawUserInfo}, {@link #getRawPath()
0332 *   getRawPath}, {@link #getRawQuery() getRawQuery}, {@link #getRawFragment()
0333 *   getRawFragment}, {@link #getRawAuthority() getRawAuthority}, and {@link
0334 *   #getRawSchemeSpecificPart() getRawSchemeSpecificPart} methods return the
0335 *   values of their corresponding components in raw form, without interpreting
0336 *   any escaped octets.  The strings returned by these methods may contain
0337 *   both escaped octets and <i>other</i> characters, and will not contain any
0338 *   illegal characters.  </p></li>
0339 *
0340 *   <li><p> The {@link #getUserInfo() getUserInfo}, {@link #getPath()
0341 *   getPath}, {@link #getQuery() getQuery}, {@link #getFragment()
0342 *   getFragment}, {@link #getAuthority() getAuthority}, and {@link
0343 *   #getSchemeSpecificPart() getSchemeSpecificPart} methods decode any escaped
0344 *   octets in their corresponding components.  The strings returned by these
0345 *   methods may contain both <i>other</i> characters and illegal characters,
0346 *   and will not contain any escaped octets.  </p></li>
0347 *
0348 *   <li><p> The {@link #toString() toString} method returns a URI string with
0349 *   all necessary quotation but which may contain <i>other</i> characters.
0350 *   </p></li>
0351 *
0352 *   <li><p> The {@link #toASCIIString() toASCIIString} method returns a fully
0353 *   quoted and encoded URI string that does not contain any <i>other</i>
0354 *   characters.  </p></li>
0355 *
0356 * </ul>
0357 *
0358 *
0359 * <h4> Identities </h4>
0360 *
0361 * For any URI <i>u</i>, it is always the case that
0362 *
0363 * <blockquote>
0364 * <tt>new URI(</tt><i>u</i><tt>.toString()).equals(</tt><i>u</i><tt>)</tt> .
0365 * </blockquote>
0366 *
0367 * For any URI <i>u</i> that does not contain redundant syntax such as two
0368 * slashes before an empty authority (as in <tt>file:///tmp/</tt> ) or a
0369 * colon following a host name but no port (as in
0370 * <tt>http://java.sun.com:</tt> ), and that does not encode characters
0371 * except those that must be quoted, the following identities also hold:
0372 *
0373 * <blockquote>
0374 * <tt>new URI(</tt><i>u</i><tt>.getScheme(),<br>
0375 *         </tt><i>u</i><tt>.getSchemeSpecificPart(),<br>
0376 *         </tt><i>u</i><tt>.getFragment())<br>
0377 * .equals(</tt><i>u</i><tt>)</tt>
0378 * </blockquote>
0379 *
0380 * in all cases,
0381 *
0382 * <blockquote>
0383 * <tt>new URI(</tt><i>u</i><tt>.getScheme(),<br>
0384 *         </tt><i>u</i><tt>.getUserInfo(), </tt><i>u</i><tt>.getAuthority(),<br>
0385 *         </tt><i>u</i><tt>.getPath(), </tt><i>u</i><tt>.getQuery(),<br>
0386 *         </tt><i>u</i><tt>.getFragment())<br>
0387 * .equals(</tt><i>u</i><tt>)</tt>
0388 * </blockquote>
0389 *
0390 * if <i>u</i> is hierarchical, and
0391 *
0392 * <blockquote>
0393 * <tt>new URI(</tt><i>u</i><tt>.getScheme(),<br>
0394 *         </tt><i>u</i><tt>.getUserInfo(), </tt><i>u</i><tt>.getHost(), </tt><i>u</i><tt>.getPort(),<br>
0395 *         </tt><i>u</i><tt>.getPath(), </tt><i>u</i><tt>.getQuery(),<br>
0396 *         </tt><i>u</i><tt>.getFragment())<br>
0397 * .equals(</tt><i>u</i><tt>)</tt>
0398 * </blockquote>
0399 *
0400 * if <i>u</i> is hierarchical and has either no authority or a server-based
0401 * authority.
0402 *
0403 *
0404 * <h4> URIs, URLs, and URNs </h4>
0405 *
0406 * A URI is a uniform resource <i>identifier</i> while a URL is a uniform
0407 * resource <i>locator</i>.  Hence every URL is a URI, abstractly speaking, but
0408 * not every URI is a URL.  This is because there is another subcategory of
0409 * URIs, uniform resource <i>names</i> (URNs), which name resources but do not
0410 * specify how to locate them.  The <tt>mailto</tt>, <tt>news</tt>, and
0411 * <tt>isbn</tt> URIs shown above are examples of URNs.
0412 *
0413 * <p> The conceptual distinction between URIs and URLs is reflected in the
0414 * differences between this class and the {@link URL} class.
0415 *
0416 * <p> An instance of this class represents a URI reference in the syntactic
0417 * sense defined by RFC 2396.  A URI may be either absolute or relative.
0418 * A URI string is parsed according to the generic syntax without regard to the
0419 * scheme, if any, that it specifies.  No lookup of the host, if any, is
0420 * performed, and no scheme-dependent stream handler is constructed.  Equality,
0421 * hashing, and comparison are defined strictly in terms of the character
0422 * content of the instance.  In other words, a URI instance is little more than
0423 * a structured string that supports the syntactic, scheme-independent
0424 * operations of comparison, normalization, resolution, and relativization.
0425 *
0426 * <p> An instance of the {@link URL} class, by contrast, represents the
0427 * syntactic components of a URL together with some of the information required
0428 * to access the resource that it describes.  A URL must be absolute, that is,
0429 * it must always specify a scheme.  A URL string is parsed according to its
0430 * scheme.  A stream handler is always established for a URL, and in fact it is
0431 * impossible to create a URL instance for a scheme for which no handler is
0432 * available.  Equality and hashing depend upon both the scheme and the
0433 * Internet address of the host, if any; comparison is not defined.  In other
0434 * words, a URL is a structured string that supports the syntactic operation of
0435 * resolution as well as the network I/O operations of looking up the host and
0436 * opening a connection to the specified resource.
0437 *
0438 *
0439 * @version 1.48, 06/06/12
0440 * @author Mark Reinhold
0441 * @since 1.4
0442 *
0443 * @see <a href="http://ietf.org/rfc/rfc2279.txt"><i>RFC 2279: UTF-8, a
0444 * transformation format of ISO 10646</i></a>, <br><a
0445 * href="http://www.ietf.org/rfc/rfc2373.txt"><i>RFC 2373: IPv6 Addressing
0446 * Architecture</i></a>, <br><a
0447 * href="http://www.ietf.org/rfc/rfc2396.txt""><i>RFC 2396: Uniform
0448 * Resource Identifiers (URI): Generic Syntax</i></a>, <br><a
0449 * href="http://www.ietf.org/rfc/rfc2732.txt"><i>RFC 2732: Format for
0450 * Literal IPv6 Addresses in URLs</i></a>, <br><a
0451 * href="URISyntaxException.html">URISyntaxException</a>
0452 */
0453
0454public final class URI
0455    implements Comparable<URI>, Serializable
0456{
0457
0458    // Note: Comments containing the word "ASSERT" indicate places where a
0459    // throw of an InternalError should be replaced by an appropriate assertion
0460    // statement once asserts are enabled in the build.
0461
0462    static final long serialVersionUID = -6052424284110960213L;
0463
0464
0465    // -- Properties and components of this instance --
0466
0467    // Components of all URIs: [<scheme>:]<scheme-specific-part>[#<fragment>]
0468    private transient String scheme;		// null ==> relative URI
0469    private transient String fragment;
0470
0471    // Hierarchical URI components: [//<authority>]<path>[?<query>]
0472    private transient String authority;		// Registry or server
0473
0474    // Server-based authority: [<userInfo>@]<host>[:<port>]
0475    private transient String userInfo;
0476    private transient String host;		// null ==> registry-based
0477    private transient int port = -1;		// -1 ==> undefined
0478
0479    // Remaining components of hierarchical URIs
0480    private transient String path;		// null ==> opaque
0481    private transient String query;
0482
0483    // The remaining fields may be computed on demand
0484
0485    private volatile transient String schemeSpecificPart;
0486    private volatile transient int hash;	// Zero ==> undefined
0487
0488    private volatile transient String decodedUserInfo = null;
0489    private volatile transient String decodedAuthority = null;
0490    private volatile transient String decodedPath = null;
0491    private volatile transient String decodedQuery = null;
0492    private volatile transient String decodedFragment = null;
0493    private volatile transient String decodedSchemeSpecificPart = null;
0494
0495    /**
0496     * The string form of this URI.
0497     *
0498     * @serial
0499     */
0500    private volatile String string;		// The only serializable field
0501
0502
0503
0504    // -- Constructors and factories --
0505
0506    private URI() { }				// Used internally
0507
0508    /**
0509     * Constructs a URI by parsing the given string.
0510     *
0511     * <p> This constructor parses the given string exactly as specified by the
0512     * grammar in <a
0513     * href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>,
0514     * Appendix A, <b><i>except for the following deviations:</i></b> </p>
0515     *
0516     * <ul type=disc>
0517     *
0518     *   <li><p> An empty authority component is permitted as long as it is
0519     *   followed by a non-empty path, a query component, or a fragment
0520     *   component.  This allows the parsing of URIs such as
0521     *   <tt>"file:///foo/bar"</tt>, which seems to be the intent of
0522     *   RFC 2396 although the grammar does not permit it.  If the
0523     *   authority component is empty then the user-information, host, and port
0524     *   components are undefined. </p></li>
0525     *
0526     *   <li><p> Empty relative paths are permitted; this seems to be the
0527     *   intent of RFC 2396 although the grammar does not permit it.  The
0528     *   primary consequence of this deviation is that a standalone fragment
0529     *   such as <tt>"#foo"</tt> parses as a relative URI with an empty path
0530     *   and the given fragment, and can be usefully <a
0531     *   href="#resolve-frag">resolved</a> against a base URI.
0532     *
0533     *   <li><p> IPv4 addresses in host components are parsed rigorously, as
0534     *   specified by <a
0535     *   href="http://www.ietf.org/rfc/rfc2732.txt">RFC 2732</a>: Each
0536     *   element of a dotted-quad address must contain no more than three
0537     *   decimal digits.  Each element is further constrained to have a value
0538     *   no greater than 255. </p></li>
0539     *
0540     *   <li> <p> Hostnames in host components that comprise only a single
0541     *   domain label are permitted to start with an <i>alphanum</i> 
0542     *   character. This seems to be the intent of <a
0543     *   href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>
0544     *   section 3.2.2 although the grammar does not permit it. The
0545     *   consequence of this deviation is that the authority component of a
0546     *   hierarchical URI such as <tt>s://123</tt>, will parse as a server-based 
0547     *   authority. </p></li>
0548     *
0549     *   <li><p> IPv6 addresses are permitted for the host component.  An IPv6
0550     *   address must be enclosed in square brackets (<tt>'['</tt> and
0551     *   <tt>']'</tt>) as specified by <a
0552     *   href="http://www.ietf.org/rfc/rfc2732.txt">RFC 2732</a>.  The
0553     *   IPv6 address itself must parse according to <a
0554     *   href="http://www.ietf.org/rfc/rfc2373.txt">RFC 2373</a>.  IPv6
0555     *   addresses are further constrained to describe no more than sixteen
0556     *   bytes of address information, a constraint implicit in RFC 2373
0557     *   but not expressible in the grammar. </p></li>
0558     *
0559     *   <li><p> Characters in the <i>other</i> category are permitted wherever
0560     *   RFC 2396 permits <i>escaped</i> octets, that is, in the
0561     *   user-information, path, query, and fragment components, as well as in
0562     *   the authority component if the authority is registry-based.  This
0563     *   allows URIs to contain Unicode characters beyond those in the US-ASCII
0564     *   character set. </p></li>
0565     *
0566     * </ul>
0567     *
0568     * @param  str   The string to be parsed into a URI
0569     *
0570     * @throws  NullPointerException
0571     *          If <tt>str</tt> is <tt>null</tt>
0572     *
0573     * @throws  URISyntaxException
0574     *          If the given string violates RFC 2396, as augmented
0575     *          by the above deviations
0576     */
0577    public URI(String str) throws URISyntaxException {
0578	new Parser(str).parse(false);
0579    }
0580
0581    /**
0582     * Constructs a hierarchical URI from the given components.
0583     *
0584     * <p> If a scheme is given then the path, if also given, must either be
0585     * empty or begin with a slash character (<tt>'/'</tt>).  Otherwise a
0586     * component of the new URI may be left undefined by passing <tt>null</tt>
0587     * for the corresponding parameter or, in the case of the <tt>port</tt>
0588     * parameter, by passing <tt>-1</tt>.
0589     *
0590     * <p> This constructor first builds a URI string from the given components
0591     * according to the rules specified in <a
0592     * href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>,
0593     * section 5.2, step 7: </p>
0594     *
0595     * <ol>
0596     *
0597     *   <li><p> Initially, the result string is empty. </p></li>
0598     *
0599     *   <li><p> If a scheme is given then it is appended to the result,
0600     *   followed by a colon character (<tt>':'</tt>).  </p></li>
0601     *
0602     *   <li><p> If user information, a host, or a port are given then the
0603     *   string <tt>"//"</tt> is appended.  </p></li>
0604     *
0605     *   <li><p> If user information is given then it is appended, followed by
0606     *   a commercial-at character (<tt>'@'</tt>).  Any character not in the
0607     *   <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
0608     *   categories is <a href="#quote">quoted</a>.  </p></li>
0609     *
0610     *   <li><p> If a host is given then it is appended.  If the host is a
0611     *   literal IPv6 address but is not enclosed in square brackets
0612     *   (<tt>'['</tt> and <tt>']'</tt>) then the square brackets are added.
0613     *   </p></li>
0614     *
0615     *   <li><p> If a port number is given then a colon character
0616     *   (<tt>':'</tt>) is appended, followed by the port number in decimal.
0617     *   </p></li>
0618     *
0619     *   <li><p> If a path is given then it is appended.  Any character not in
0620     *   the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
0621     *   categories, and not equal to the slash character (<tt>'/'</tt>) or the
0622     *   commercial-at character (<tt>'@'</tt>), is quoted.  </p></li>
0623     *
0624     *   <li><p> If a query is given then a question-mark character
0625     *   (<tt>'?'</tt>) is appended, followed by the query.  Any character that
0626     *   is not a <a href="#legal-chars">legal URI character</a> is quoted.
0627     *   </p></li>
0628     *
0629     *   <li><p> Finally, if a fragment is given then a hash character
0630     *   (<tt>'#'</tt>) is appended, followed by the fragment.  Any character
0631     *   that is not a legal URI character is quoted.  </p></li>
0632     *
0633     * </ol>
0634     *
0635     * <p> The resulting URI string is then parsed as if by invoking the {@link
0636     * #URI(String)} constructor and then invoking the {@link
0637     * #parseServerAuthority()} method upon the result; this may cause a {@link
0638     * URISyntaxException} to be thrown.  </p>
0639     *
0640     * @param   scheme    Scheme name
0641     * @param   userInfo  User name and authorization information
0642     * @param   host      Host name
0643     * @param   port      Port number
0644     * @param   path      Path
0645     * @param   query     Query
0646     * @param   fragment  Fragment
0647     *
0648     * @throws URISyntaxException
0649     *         If both a scheme and a path are given but the path is relative,
0650     *         if the URI string constructed from the given components violates
0651     *         RFC 2396, or if the authority component of the string is
0652     *         present but cannot be parsed as a server-based authority
0653     */
0654    public URI(String scheme,
0655               String userInfo, String host, int port,
0656               String path, String query, String fragment)
0657	throws URISyntaxException
0658    {
0659	String s = toString(scheme, null,
0660			    null, userInfo, host, port,
0661			    path, query, fragment);
0662	checkPath(s, scheme, path);
0663	new Parser(s).parse(true);
0664    }
0665
0666    /**
0667     * Constructs a hierarchical URI from the given components.
0668     *
0669     * <p> If a scheme is given then the path, if also given, must either be
0670     * empty or begin with a slash character (<tt>'/'</tt>).  Otherwise a
0671     * component of the new URI may be left undefined by passing <tt>null</tt>
0672     * for the corresponding parameter.
0673     *
0674     * <p> This constructor first builds a URI string from the given components
0675     * according to the rules specified in <a
0676     * href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>,
0677     * section 5.2, step 7: </p>
0678     *
0679     * <ol>
0680     *
0681     *   <li><p> Initially, the result string is empty.  </p></li>
0682     *
0683     *   <li><p> If a scheme is given then it is appended to the result,
0684     *   followed by a colon character (<tt>':'</tt>).  </p></li>
0685     *
0686     *   <li><p> If an authority is given then the string <tt>"//"</tt> is
0687     *   appended, followed by the authority.  If the authority contains a
0688     *   literal IPv6 address then the address must be enclosed in square
0689     *   brackets (<tt>'['</tt> and <tt>']'</tt>).  Any character not in the
0690     *   <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
0691     *   categories, and not equal to the commercial-at character
0692     *   (<tt>'@'</tt>), is <a href="#quote">quoted</a>.  </p></li>
0693     *
0694     *   <li><p> If a path is given then it is appended.  Any character not in
0695     *   the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, or <i>other</i>
0696     *   categories, and not equal to the slash character (<tt>'/'</tt>) or the
0697     *   commercial-at character (<tt>'@'</tt>), is quoted.  </p></li>
0698     *
0699     *   <li><p> If a query is given then a question-mark character
0700     *   (<tt>'?'</tt>) is appended, followed by the query.  Any character that
0701     *   is not a <a href="#legal-chars">legal URI character</a> is quoted.
0702     *   </p></li>
0703     *
0704     *   <li><p> Finally, if a fragment is given then a hash character
0705     *   (<tt>'#'</tt>) is appended, followed by the fragment.  Any character
0706     *   that is not a legal URI character is quoted.  </p></li>
0707     *
0708     * </ol>
0709     *
0710     * <p> The resulting URI string is then parsed as if by invoking the {@link
0711     * #URI(String)} constructor and then invoking the {@link
0712     * #parseServerAuthority()} method upon the result; this may cause a {@link
0713     * URISyntaxException} to be thrown.  </p>
0714     *
0715     * @param   scheme     Scheme name
0716     * @param   authority  Authority
0717     * @param   path       Path
0718     * @param   query      Query
0719     * @param   fragment   Fragment
0720     *
0721     * @throws URISyntaxException
0722     *         If both a scheme and a path are given but the path is relative,
0723     *         if the URI string constructed from the given components violates
0724     *         RFC 2396, or if the authority component of the string is
0725     *         present but cannot be parsed as a server-based authority
0726     */
0727    public URI(String scheme,
0728	       String authority,
0729	       String path, String query, String fragment)
0730	throws URISyntaxException
0731    {
0732	String s = toString(scheme, null,
0733			    authority, null, null, -1,
0734			    path, query, fragment);
0735	checkPath(s, scheme, path);
0736	new Parser(s).parse(false);
0737    }
0738
0739    /**
0740     * Constructs a hierarchical URI from the given components.
0741     *
0742     * <p> A component may be left undefined by passing <tt>null</tt>.
0743     *
0744     * <p> This convenience constructor works as if by invoking the
0745     * seven-argument constructor as follows:
0746     *
0747     * <blockquote><tt>
0748     * new {@link #URI(String, String, String, int, String, String, String)
0749     * URI}(scheme, null, host, -1, path, null, fragment);
0750     * </tt></blockquote>
0751     *
0752     * @param   scheme    Scheme name
0753     * @param   host      Host name
0754     * @param   path      Path
0755     * @param   fragment  Fragment
0756     *
0757     * @throws  URISyntaxException
0758     *          If the URI string constructed from the given components
0759     *          violates RFC 2396
0760     */
0761    public URI(String scheme, String host, String path, String fragment)
0762	throws URISyntaxException
0763    {
0764	this(scheme, null, host, -1, path, null, fragment);
0765    }
0766
0767    /**
0768     * Constructs a URI from the given components.
0769     *
0770     * <p> A component may be left undefined by passing <tt>null</tt>.
0771     *
0772     * <p> This constructor first builds a URI in string form using the given
0773     * components as follows:  </p>
0774     *
0775     * <ol>
0776     *
0777     *   <li><p> Initially, the result string is empty.  </p></li>
0778     *
0779     *   <li><p> If a scheme is given then it is appended to the result,
0780     *   followed by a colon character (<tt>':'</tt>).  </p></li>
0781     *
0782     *   <li><p> If a scheme-specific part is given then it is appended.  Any
0783     *   character that is not a <a href="#legal-chars">legal URI character</a>
0784     *   is <a href="#quote">quoted</a>.  </p></li>
0785     *
0786     *   <li><p> Finally, if a fragment is given then a hash character
0787     *   (<tt>'#'</tt>) is appended to the string, followed by the fragment.
0788     *   Any character that is not a legal URI character is quoted.  </p></li>
0789     *
0790     * </ol>
0791     *
0792     * <p> The resulting URI string is then parsed in order to create the new
0793     * URI instance as if by invoking the {@link #URI(String)} constructor;
0794     * this may cause a {@link URISyntaxException} to be thrown.  </p>
0795     *
0796     * @param   scheme    Scheme name
0797     * @param   ssp       Scheme-specific part
0798     * @param   fragment  Fragment
0799     *
0800     * @throws  URISyntaxException
0801     *          If the URI string constructed from the given components
0802     *          violates RFC 2396
0803     */
0804    public URI(String scheme, String ssp, String fragment)
0805	throws URISyntaxException
0806    {
0807	new Parser(toString(scheme, ssp,
0808			    null, null, null, -1,
0809			    null, null, fragment))
0810	    .parse(false);
0811    }
0812
0813    /**
0814     * Creates a URI by parsing the given string.
0815     *
0816     * <p> This convenience factory method works as if by invoking the {@link
0817     * #URI(String)} constructor; any {@link URISyntaxException} thrown by the
0818     * constructor is caught and wrapped in a new {@link
0819     * IllegalArgumentException} object, which is then thrown.
0820     *
0821     * <p> This method is provided for use in situations where it is known that
0822     * the given string is a legal URI, for example for URI constants declared
0823     * within in a program, and so it would be considered a programming error
0824     * for the string not to parse as such.  The constructors, which throw
0825     * {@link URISyntaxException} directly, should be used situations where a
0826     * URI is being constructed from user input or from some other source that
0827     * may be prone to errors.  </p>
0828     *
0829     * @param  str   The string to be parsed into a URI
0830     * @return The new URI
0831     *
0832     * @throws  NullPointerException
0833     *          If <tt>str</tt> is <tt>null</tt>
0834     *
0835     * @throws  IllegalArgumentException
0836     *          If the given string violates RFC 2396
0837     */
0838    public static URI create(String str) {
0839	try {
0840	    return new URI(str);
0841	} catch (URISyntaxException x) {
0842	    IllegalArgumentException y = new IllegalArgumentException();
0843	    y.initCause(x);
0844	    throw y;
0845	}
0846    }
0847
0848
0849    // -- Operations --
0850
0851    /**
0852     * Attempts to parse this URI's authority component, if defined, into
0853     * user-information, host, and port components.
0854     *
0855     * <p> If this URI's authority component has already been recognized as
0856     * being server-based then it will already have been parsed into
0857     * user-information, host, and port components.  In this case, or if this
0858     * URI has no authority component, this method simply returns this URI.
0859     *
0860     * <p> Otherwise this method attempts once more to parse the authority
0861     * component into user-information, host, and port components, and throws
0862     * an exception describing why the authority component could not be parsed
0863     * in that way.
0864     *
0865     * <p> This method is provided because the generic URI syntax specified in
0866     * <a href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>
0867     * cannot always distinguish a malformed server-based authority from a
0868     * legitimate registry-based authority.  It must therefore treat some
0869     * instances of the former as instances of the latter.  The authority
0870     * component in the URI string <tt>"//foo:bar"</tt>, for example, is not a
0871     * legal server-based authority but it is legal as a registry-based
0872     * authority.
0873     *
0874     * <p> In many common situations, for example when working URIs that are
0875     * known to be either URNs or URLs, the hierarchical URIs being used will
0876     * always be server-based.  They therefore must either be parsed as such or
0877     * treated as an error.  In these cases a statement such as
0878     *
0879     * <blockquote>
0880     * <tt>URI </tt><i>u</i><tt> = new URI(str).parseServerAuthority();</tt>
0881     * </blockquote>
0882     *
0883     * <p> can be used to ensure that <i>u</i> always refers to a URI that, if
0884     * it has an authority component, has a server-based authority with proper
0885     * user-information, host, and port components.  Invoking this method also
0886     * ensures that if the authority could not be parsed in that way then an
0887     * appropriate diagnostic message can be issued based upon the exception
0888     * that is thrown. </p>
0889     *
0890     * @return  A URI whose authority field has been parsed
0891     *          as a server-based authority
0892     *
0893     * @throws  URISyntaxException
0894     *          If the authority component of this URI is defined
0895     *          but cannot be parsed as a server-based authority
0896     *          according to RFC 2396
0897     */
0898    public URI parseServerAuthority()
0899	throws URISyntaxException
0900    {
0901	// We could be clever and cache the error message and index from the
0902	// exception thrown during the original parse, but that would require
0903	// either more fields or a more-obscure representation.
0904	if ((host != null) || (authority == null))
0905	    return this;
0906	defineString();
0907	new Parser(string).parse(true);
0908	return this;
0909    }
0910
0911    /**
0912     * Normalizes this URI's path.
0913     *
0914     * <p> If this URI is opaque, or if its path is already in normal form,
0915     * then this URI is returned.  Otherwise a new URI is constructed that is
0916     * identical to this URI except that its path is computed by normalizing
0917     * this URI's path in a manner consistent with <a
0918     * href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>,
0919     * section 5.2, step 6, sub-steps c through f; that is:
0920     * </p>
0921     *
0922     * <ol>
0923     *
0924     *   <li><p> All <tt>"."</tt> segments are removed. </p></li>
0925     *
0926     *   <li><p> If a <tt>".."</tt> segment is preceded by a non-<tt>".."</tt>
0927     *   segment then both of these segments are removed.  This step is
0928     *   repeated until it is no longer applicable. </p></li>
0929     *
0930     *   <li><p> If the path is relative, and if its first segment contains a
0931     *   colon character (<tt>':'</tt>), then a <tt>"."</tt> segment is
0932     *   prepended.  This prevents a relative URI with a path such as
0933     *   <tt>"a:b/c/d"</tt> from later being re-parsed as an opaque URI with a
0934     *   scheme of <tt>"a"</tt> and a scheme-specific part of <tt>"b/c/d"</tt>.
0935     *   <b><i>(Deviation from RFC 2396)</i></b> </p></li>
0936     *
0937     * </ol>
0938     *
0939     * <p> A normalized path will begin with one or more <tt>".."</tt> segments
0940     * if there were insufficient non-<tt>".."</tt> segments preceding them to
0941     * allow their removal.  A normalized path will begin with a <tt>"."</tt>
0942     * segment if one was inserted by step 3 above.  Otherwise, a normalized
0943     * path will not contain any <tt>"."</tt> or <tt>".."</tt> segments. </p>
0944     *
0945     * @return  A URI equivalent to this URI,
0946     *          but whose path is in normal form
0947     */
0948    public URI normalize() {
0949	return normalize(this);
0950    }
0951
0952    /**
0953     * Resolves the given URI against this URI.
0954     *
0955     * <p> If the given URI is already absolute, or if this URI is opaque, then
0956     * the given URI is returned.
0957     *
0958     * <p><a name="resolve-frag"></a> If the given URI's fragment component is
0959     * defined, its path component is empty, and its scheme, authority, and
0960     * query components are undefined, then a URI with the given fragment but
0961     * with all other components equal to those of this URI is returned.  This
0962     * allows a URI representing a standalone fragment reference, such as
0963     * <tt>"#foo"</tt>, to be usefully resolved against a base URI.
0964     *
0965     * <p> Otherwise this method constructs a new hierarchical URI in a manner
0966     * consistent with <a
0967     * href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>,
0968     * section 5.2; that is: </p>
0969     *
0970     * <ol>
0971     *
0972     *   <li><p> A new URI is constructed with this URI's scheme and the given
0973     *   URI's query and fragment components. </p></li>
0974     *
0975     *   <li><p> If the given URI has an authority component then the new URI's
0976     *   authority and path are taken from the given URI. </p></li>
0977     *
0978     *   <li><p> Otherwise the new URI's authority component is copied from
0979     *   this URI, and its path is computed as follows: </p></li>
0980     *
0981     *   <ol type=a>
0982     *
0983     *     <li><p> If the given URI's path is absolute then the new URI's path
0984     *     is taken from the given URI. </p></li>
0985     *
0986     *     <li><p> Otherwise the given URI's path is relative, and so the new
0987     *     URI's path is computed by resolving the path of the given URI
0988     *     against the path of this URI.  This is done by concatenating all but
0989     *     the last segment of this URI's path, if any, with the given URI's
0990     *     path and then normalizing the result as if by invoking the {@link
0991     *     #normalize() normalize} method. </p></li>
0992     *
0993     *   </ol>
0994     *
0995     * </ol>
0996     *
0997     * <p> The result of this method is absolute if, and only if, either this
0998     * URI is absolute or the given URI is absolute.  </p>
0999     *
1000     * @param  uri  The URI to be resolved against this URI
1001     * @return The resulting URI
1002     *
1003     * @throws  NullPointerException
1004     *          If <tt>uri</tt> is <tt>null</tt>
1005     */
1006    public URI resolve(URI uri) {
1007	return resolve(this, uri);
1008    }
1009
1010    /**
1011     * Constructs a new URI by parsing the given string and then resolving it
1012     * against this URI.
1013     *
1014     * <p> This convenience method works as if invoking it were equivalent to
1015     * evaluating the expression <tt>{@link #resolve(java.net.URI)
1016     * resolve}(URI.{@link #create(String) create}(str))</tt>. </p>
1017     *
1018     * @param  str   The string to be parsed into a URI
1019     * @return The resulting URI
1020     *
1021     * @throws  NullPointerException
1022     *          If <tt>str</tt> is <tt>null</tt>
1023     *
1024     * @throws  IllegalArgumentException
1025     *          If the given string violates RFC 2396
1026     */
1027    public URI resolve(String str) {
1028	return resolve(URI.create(str));
1029    }
1030
1031    /**
1032     * Relativizes the given URI against this URI.
1033     *
1034     * <p> The relativization of the given URI against this URI is computed as
1035     * follows: </p>
1036     *
1037     * <ol>
1038     *
1039     *   <li><p> If either this URI or the given URI are opaque, or if the
1040     *   scheme and authority components of the two URIs are not identical, or
1041     *   if the path of this URI is not a prefix of the path of the given URI,
1042     *   then the given URI is returned. </p></li>
1043     *
1044     *   <li><p> Otherwise a new relative hierarchical URI is constructed with
1045     *   query and fragment components taken from the given URI and with a path
1046     *   component computed by removing this URI's path from the beginning of
1047     *   the given URI's path. </p></li>
1048     *
1049     * </ol>
1050     *
1051     * @param  uri  The URI to be relativized against this URI
1052     * @return The resulting URI
1053     *
1054     * @throws  NullPointerException
1055     *          If <tt>uri</tt> is <tt>null</tt>
1056     */
1057    public URI relativize(URI uri) {
1058	return relativize(this, uri);
1059    }
1060
1061    /**
1062     * Constructs a URL from this URI.
1063     *
1064     * <p> This convenience method works as if invoking it were equivalent to
1065     * evaluating the expression <tt>new URL(this.toString())</tt> after
1066     * first checking that this URI is absolute. </p>
1067     *
1068     * @return  A URL constructed from this URI
1069     *
1070     * @throws  IllegalArgumentException
1071     *          If this URL is not absolute
1072     *
1073     * @throws  MalformedURLException
1074     *          If a protocol handler for the URL could not be found,
1075     *          or if some other error occurred while constructing the URL
1076     */
1077    public URL toURL()
1078	throws MalformedURLException {
1079	if (!isAbsolute())
1080	    throw new IllegalArgumentException("URI is not absolute");
1081	return new URL(toString());
1082    }
1083
1084    // -- Component access methods --
1085
1086    /**
1087     * Returns the scheme component of this URI.
1088     *
1089     * <p> The scheme component of a URI, if defined, only contains characters
1090     * in the <i>alphanum</i> category and in the string <tt>"-.+"</tt>.  A
1091     * scheme always starts with an <i>alpha</i> character. <p>
1092     *
1093     * The scheme component of a URI cannot contain escaped octets, hence this
1094     * method does not perform any decoding.
1095     *
1096     * @return  The scheme component of this URI,
1097     *          or <tt>null</tt> if the scheme is undefined
1098     */
1099    public String getScheme() {
1100	return scheme;
1101    }
1102
1103    /**
1104     * Tells whether or not this URI is absolute.
1105     *
1106     * <p> A URI is absolute if, and only if, it has a scheme component. </p>
1107     *
1108     * @return  <tt>true</tt> if, and only if, this URI is absolute
1109     */
1110    public boolean isAbsolute() {
1111	return scheme != null;
1112    }
1113
1114    /**
1115     * Tells whether or not this URI is opaque.
1116     *
1117     * <p> A URI is opaque if, and only if, it is absolute and its
1118     * scheme-specific part does not begin with a slash character ('/').
1119     * An opaque URI has a scheme, a scheme-specific part, and possibly
1120     * a fragment; all other components are undefined. </p>
1121     *
1122     * @return  <tt>true</tt> if, and only if, this URI is opaque
1123     */
1124    public boolean isOpaque() {
1125        return path == null;
1126    }
1127
1128    /**
1129     * Returns the raw scheme-specific part of this URI.  The scheme-specific
1130     * part is never undefined, though it may be empty.
1131     *
1132     * <p> The scheme-specific part of a URI only contains legal URI
1133     * characters. </p>
1134     *
1135     * @return  The raw scheme-specific part of this URI
1136     *          (never <tt>null</tt>)
1137     */
1138    public String getRawSchemeSpecificPart() {
1139	defineSchemeSpecificPart();
1140	return schemeSpecificPart;
1141    }
1142
1143    /**
1144     * Returns the decoded scheme-specific part of this URI.
1145     *
1146     * <p> The string returned by this method is equal to that returned by the
1147     * {@link #getRawSchemeSpecificPart() getRawSchemeSpecificPart} method
1148     * except that all sequences of escaped octets are <a
1149     * href="#decode">decoded</a>.  </p>
1150     *
1151     * @return  The decoded scheme-specific part of this URI
1152     *          (never <tt>null</tt>)
1153     */
1154    public String getSchemeSpecificPart() {
1155	if (decodedSchemeSpecificPart == null)
1156	    decodedSchemeSpecificPart = decode(getRawSchemeSpecificPart());
1157	return decodedSchemeSpecificPart;
1158    }
1159
1160    /**
1161     * Returns the raw authority component of this URI.
1162     *
1163     * <p> The authority component of a URI, if defined, only contains the
1164     * commercial-at character (<tt>'@'</tt>) and characters in the
1165     * <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, and <i>other</i>
1166     * categories.  If the authority is server-based then it is further
1167     * constrained to have valid user-information, host, and port
1168     * components. </p>
1169     *
1170     * @return  The raw authority component of this URI,
1171     *          or <tt>null</tt> if the authority is undefined
1172     */
1173    public String getRawAuthority() {
1174	return authority;
1175    }
1176
1177    /**
1178     * Returns the decoded authority component of this URI.
1179     *
1180     * <p> The string returned by this method is equal to that returned by the
1181     * {@link #getRawAuthority() getRawAuthority} method except that all
1182     * sequences of escaped octets are <a href="#decode">decoded</a>.  </p>
1183     *
1184     * @return  The decoded authority component of this URI,
1185     *          or <tt>null</tt> if the authority is undefined
1186     */
1187    public String getAuthority() {
1188	if (decodedAuthority == null)
1189	    decodedAuthority = decode(authority);
1190	return decodedAuthority;
1191    }
1192
1193    /**
1194     * Returns the raw user-information component of this URI.
1195     *
1196     * <p> The user-information component of a URI, if defined, only contains
1197     * characters in the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>, and
1198     * <i>other</i> categories. </p>
1199     *
1200     * @return  The raw user-information component of this URI,
1201     *          or <tt>null</tt> if the user information is undefined
1202     */
1203    public String getRawUserInfo() {
1204	return userInfo;
1205    }
1206
1207    /**
1208     * Returns the decoded user-information component of this URI.
1209     *
1210     * <p> The string returned by this method is equal to that returned by the
1211     * {@link #getRawUserInfo() getRawUserInfo} method except that all
1212     * sequences of escaped octets are <a href="#decode">decoded</a>.  </p>
1213     *
1214     * @return  The decoded user-information component of this URI,
1215     *          or <tt>null</tt> if the user information is undefined
1216     */
1217    public String getUserInfo() {
1218	if ((decodedUserInfo == null) && (userInfo != null))
1219	    decodedUserInfo = decode(userInfo);
1220	return decodedUserInfo;
1221    }
1222
1223    /**
1224     * Returns the host component of this URI.
1225     *
1226     * <p> The host component of a URI, if defined, will have one of the
1227     * following forms: </p>
1228     *
1229     * <ul type=disc>
1230     *
1231     *   <li><p> A domain name consisting of one or more <i>labels</i>
1232     *   separated by period characters (<tt>'.'</tt>), optionally followed by
1233     *   a period character.  Each label consists of <i>alphanum</i> characters
1234     *   as well as hyphen characters (<tt>'-'</tt>), though hyphens never
1235     *   occur as the first or last characters in a label. The rightmost
1236     *   label of a domain name consisting of two or more labels, begins
1237     *   with an <i>alpha</i> character. </li>
1238     *
1239     *   <li><p> A dotted-quad IPv4 address of the form
1240     *   <i>digit</i><tt>+.</tt><i>digit</i><tt>+.</tt><i>digit</i><tt>+.</tt><i>digit</i><tt>+</tt>,
1241     *   where no <i>digit</i> sequence is longer than three characters and no
1242     *   sequence has a value larger than 255. </p></li>
1243     *
1244     *   <li><p> An IPv6 address enclosed in square brackets (<tt>'['</tt> and
1245     *   <tt>']'</tt>) and consisting of hexadecimal digits, colon characters
1246     *   (<tt>':'</tt>), and possibly an embedded IPv4 address.  The full
1247     *   syntax of IPv6 addresses is specified in <a
1248     *   href="http://www.ietf.org/rfc/rfc2373.txt"><i>RFC 2373: IPv6
1249     *   Addressing Architecture</i></a>.  </p></li>
1250     *
1251     * </ul>
1252     *
1253     * The host component of a URI cannot contain escaped octets, hence this
1254     * method does not perform any decoding.
1255     *
1256     * @return  The host component of this URI,
1257     *          or <tt>null</tt> if the host is undefined
1258     */
1259    public String getHost() {
1260	return host;
1261    }
1262
1263    /**
1264     * Returns the port number of this URI.
1265     *
1266     * <p> The port component of a URI, if defined, is a non-negative
1267     * integer. </p>
1268     *
1269     * @return  The port component of this URI,
1270     *          or <tt>-1</tt> if the port is undefined
1271     */
1272    public int getPort() {
1273	return port;
1274    }
1275
1276    /**
1277     * Returns the raw path component of this URI.
1278     *
1279     * <p> The path component of a URI, if defined, only contains the slash
1280     * character (<tt>'/'</tt>), the commercial-at character (<tt>'@'</tt>),
1281     * and characters in the <i>unreserved</i>, <i>punct</i>, <i>escaped</i>,
1282     * and <i>other</i> categories. </p>
1283     *
1284     * @return  The path component of this URI,
1285     *          or <tt>null</tt> if the path is undefined
1286     */
1287    public String getRawPath() {
1288	return path;
1289    }
1290
1291    /**
1292     * Returns the decoded path component of this URI.
1293     *
1294     * <p> The string returned by this method is equal to that returned by the
1295     * {@link #getRawPath() getRawPath} method except that all sequences of
1296     * escaped octets are <a href="#decode">decoded</a>.  </p>
1297     *
1298     * @return  The decoded path component of this URI,
1299     *          or <tt>null</tt> if the path is undefined
1300     */
1301    public String getPath() {
1302	if ((decodedPath == null) && (path != null))
1303	    decodedPath = decode(path);
1304	return decodedPath;
1305    }
1306
1307    /**
1308     * Returns the raw query component of this URI.
1309     *
1310     * <p> The query component of a URI, if defined, only contains legal URI
1311     * characters. </p>
1312     *
1313     * @return  The raw query component of this URI,
1314     *          or <tt>null</tt> if the query is undefined
1315     */
1316    public String getRawQuery() {
1317	return query;
1318    }
1319
1320    /**
1321     * Returns the decoded query component of this URI.
1322     *
1323     * <p> The string returned by this method is equal to that returned by the
1324     * {@link #getRawQuery() getRawQuery} method except that all sequences of
1325     * escaped octets are <a href="#decode">decoded</a>.  </p>
1326     *
1327     * @return  The decoded query component of this URI,
1328     *          or <tt>null</tt> if the query is undefined
1329     */
1330    public String getQuery() {
1331	if ((decodedQuery == null) && (query != null))
1332	    decodedQuery = decode(query);
1333	return decodedQuery;
1334    }
1335
1336    /**
1337     * Returns the raw fragment component of this URI.
1338     *
1339     * <p> The fragment component of a URI, if defined, only contains legal URI
1340     * characters. </p>
1341     *
1342     * @return  The raw fragment component of this URI,
1343     *          or <tt>null</tt> if the fragment is undefined
1344     */
1345    public String getRawFragment() {
1346	return fragment;
1347    }
1348
1349    /**
1350     * Returns the decoded fragment component of this URI.
1351     *
1352     * <p> The string returned by this method is equal to that returned by the
1353     * {@link #getRawFragment() getRawFragment} method except that all
1354     * sequences of escaped octets are <a href="#decode">decoded</a>.  </p>
1355     *
1356     * @return  The decoded fragment component of this URI,
1357     *          or <tt>null</tt> if the fragment is undefined
1358     */
1359    public String getFragment() {
1360	if ((decodedFragment == null) && (fragment != null))
1361	    decodedFragment = decode(fragment);
1362	return decodedFragment;
1363    }
1364
1365
1366    // -- Equality, comparison, hash code, toString, and serialization --
1367
1368    /**
1369     * Tests this URI for equality with another object.
1370     *
1371     * <p> If the given object is not a URI then this method immediately
1372     * returns <tt>false</tt>.
1373     *
1374     * <p> For two URIs to be considered equal requires that either both are
1375     * opaque or both are hierarchical.  Their schemes must either both be
1376     * undefined or else be equal without regard to case. Their fragments
1377     * must either both be undefined or else be equal.
1378     *
1379     * <p> For two opaque URIs to be considered equal, their scheme-specific
1380     * parts must be equal.
1381     *
1382     * <p> For two hierarchical URIs to be considered equal, their paths must
1383     * be equal and their queries must either both be undefined or else be
1384     * equal.  Their authorities must either both be undefined, or both be
1385     * registry-based, or both be server-based.  If their authorities are
1386     * defined and are registry-based, then they must be equal.  If their
1387     * authorities are defined and are server-based, then their hosts must be
1388     * equal without regard to case, their port numbers must be equal, and
1389     * their user-information components must be equal.
1390     *
1391     * <p> When testing the user-information, path, query, fragment, authority,
1392     * or scheme-specific parts of two URIs for equality, the raw forms rather
1393     * than the encoded forms of these components are compared and the
1394     * hexadecimal digits of escaped octets are compared without regard to
1395     * case.
1396     *
1397     * <p> This method satisfies the general contract of the {@link
1398     * java.lang.Object#equals(Object) Object.equals} method. </p>
1399     *
1400     * @param   ob   The object to which this object is to be compared
1401     *
1402     * @return  <tt>true</tt> if, and only if, the given object is a URI that
1403     *          is identical to this URI
1404     */
1405    public boolean equals(Object ob) {
1406	if (ob == this)
1407	    return true;
1408	if (!(ob instanceof URI))
1409	    return false;
1410	URI that = (URI)ob;
1411	if (this.isOpaque() != that.isOpaque()) return false;
1412	if (!equalIgnoringCase(this.scheme, that.scheme)) return false;
1413	if (!equal(this.fragment, that.fragment)) return false;
1414
1415	// Opaque
1416	if (this.isOpaque())
1417	    return equal(this.schemeSpecificPart, that.schemeSpecificPart);
1418
1419	// Hierarchical
1420	if (!equal(this.path, that.path)) return false;
1421	if (!equal(this.query, that.query)) return false;
1422
1423	// Authorities
1424	if (this.authority == that.authority) return true;
1425	if (this.host != null) {
1426	    // Server-based
1427	    if (!equal(this.userInfo, that.userInfo)) return false;
1428	    if (!equalIgnoringCase(this.host, that.host)) return false;
1429	    if (this.port != that.port) return false;
1430	} else if (this.authority != null) {
1431	    // Registry-based
1432	    if (!equal(this.authority, that.authority)) return false;
1433	} else if (this.authority != that.authority) {
1434	    return false;
1435	}
1436
1437	return true;
1438    }
1439
1440    /**
1441     * Returns a hash-code value for this URI.  The hash code is based upon all
1442     * of the URI's components, and satisfies the general contract of the
1443     * {@link java.lang.Object#hashCode() Object.hashCode} method.
1444     *
1445     * @return  A hash-code value for this URI
1446     */
1447    public int hashCode() {
1448	if (hash != 0)
1449	    return hash;
1450	int h = hashIgnoringCase(0, scheme);
1451	h = hash(h, fragment);
1452	if (isOpaque()) {
1453	    h = hash(h, schemeSpecificPart);
1454	} else {
1455	    h = hash(h, path);
1456	    h = hash(h, query);
1457	    if (host != null) {
1458		h = hash(h, userInfo);
1459		h = hashIgnoringCase(h, host);
1460		h += 1949 * port;
1461	    } else {
1462		h = hash(h, authority);
1463	    }
1464	}
1465	hash = h;
1466	return h;
1467    }
1468
1469    /**
1470     * Compares this URI to another object, which must be a URI.
1471     *
1472     * <p> When comparing corresponding components of two URIs, if one
1473     * component is undefined but the other is defined then the first is
1474     * considered to be less than the second.  Unless otherwise noted, string
1475     * components are ordered according to their natural, case-sensitive
1476     * ordering as defined by the {@link java.lang.String#compareTo(Object)
1477     * String.compareTo} method.  String components that are subject to
1478     * encoding are compared by comparing their raw forms rather than their
1479     * encoded forms.
1480     *
1481     * <p> The ordering of URIs is defined as follows: </p>
1482     *
1483     * <ul type=disc>
1484     *
1485     *   <li><p> Two URIs with different schemes are ordered according the
1486     *   ordering of their schemes, without regard to case. </p></li>
1487     *
1488     *   <li><p> A hierarchical URI is considered to be less than an opaque URI
1489     *   with an identical scheme. </p></li>
1490     *
1491     *   <li><p> Two opaque URIs with identical schemes are ordered according
1492     *   to the ordering of their scheme-specific parts. </p></li>
1493     *
1494     *   <li><p> Two opaque URIs with identical schemes and scheme-specific
1495     *   parts are ordered according to the ordering of their
1496     *   fragments. </p></li>
1497     *
1498     *   <li><p> Two hierarchical URIs with identical schemes are ordered
1499     *   according to the ordering of their authority components: </p></li>
1500     *
1501     *   <ul type=disc>
1502     *
1503     *     <li><p> If both authority components are server-based then the URIs
1504     *     are ordered according to their user-information components; if these
1505     *     components are identical then the URIs are ordered according to the
1506     *     ordering of their hosts, without regard to case; if the hosts are
1507     *     identical then the URIs are ordered according to the ordering of
1508     *     their ports. </p></li>
1509     *
1510     *     <li><p> If one or both authority components are registry-based then
1511     *     the URIs are ordered according to the ordering of their authority
1512     *     components. </p></li>
1513     *
1514     *   </ul>
1515     *
1516     *   <li><p> Finally, two hierarchical URIs with identical schemes and
1517     *   authority components are ordered according to the ordering of their
1518     *   paths; if their paths are identical then they are ordered according to
1519     *   the ordering of their queries; if the queries are identical then they
1520     *   are ordered according to the order of their fragments. </p></li>
1521     *
1522     * </ul>
1523     *
1524     * <p> This method satisfies the general contract of the {@link
1525     * java.lang.Comparable#compareTo(Object) Comparable.compareTo}
1526     * method. </p>
1527     *
1528     * @param   that
1529     *          The object to which this URI is to be compared
1530     *
1531     * @return  A negative integer, zero, or a positive integer as this URI is
1532     *          less than, equal to, or greater than the given URI
1533     *
1534     * @throws  ClassCastException
1535     *          If the given object is not a URI
1536     */
1537    public int compareTo(URI that) {
1538	int c;
1539
1540	if ((c = compareIgnoringCase(this.scheme, that.scheme)) != 0)
1541	    return c;
1542
1543	if (this.isOpaque()) {
1544	    if (that.isOpaque()) {
1545		// Both opaque
1546		if ((c = compare(this.schemeSpecificPart,
1547				 that.schemeSpecificPart)) != 0)
1548		    return c;
1549		return compare(this.fragment, that.fragment);
1550	    }
1551	    return +1;			// Opaque > hierarchical
1552	} else if (that.isOpaque()) {
1553	    return -1;			// Hierarchical < opaque
1554	}
1555
1556	// Hierarchical
1557	if ((this.host != null) && (that.host != null)) {
1558	    // Both server-based
1559	    if ((c = compare(this.userInfo, that.userInfo)) != 0)
1560		return c;
1561	    if ((c = compareIgnoringCase(this.host, that.host)) != 0)
1562		return c;
1563	    if ((c = this.port - that.port) != 0)
1564		return c;
1565	} else {
1566	    // If one or both authorities are registry-based then we simply
1567	    // compare them in the usual, case-sensitive way.  If one is
1568	    // registry-based and one is server-based then the strings are
1569	    // guaranteed to be unequal, hence the comparison will never return
1570	    // zero and the compareTo and equals methods will remain
1571	    // consistent.
1572	    if ((c = compare(this.authority, that.authority)) != 0) return c;
1573	}
1574
1575	if ((c = compare(this.path, that.path)) != 0) return c;
1576	if ((c = compare(this.query, that.query)) != 0) return c;
1577	return compare(this.fragment, that.fragment);
1578    }
1579
1580    /**
1581     * Returns the content of this URI as a string.
1582     *
1583     * <p> If this URI was created by invoking one of the constructors in this
1584     * class then a string equivalent to the original input string, or to the
1585     * string computed from the originally-given components, as appropriate, is
1586     * returned.  Otherwise this URI was created by normalization, resolution,
1587     * or relativization, and so a string is constructed from this URI's
1588     * components according to the rules specified in <a
1589     * href="http://www.ietf.org/rfc/rfc2396.txt">RFC 2396</a>,
1590     * section 5.2, step 7. </p>
1591     *
1592     * @return  The string form of this URI
1593     */
1594    public String toString() {
1595	defineString();
1596	return string;
1597    }
1598
1599    /**
1600     * Returns the content of this URI as a US-ASCII string.
1601     *
1602     * <p> If this URI does not contain any characters in the <i>other</i>
1603     * category then an invocation of this method will return the same value as
1604     * an invocation of the {@link #toString() toString} method.  Otherwise
1605     * this method works as if by invoking that method and then <a
1606     * href="#encode">encoding</a> the result.  </p>
1607     *
1608     * @return  The string form of this URI, encoded as needed
1609     *          so that it only contains characters in the US-ASCII
1610     *          charset
1611     */
1612    public String toASCIIString() {
1613	defineString();
1614	return encode(string);
1615    }
1616
1617
1618    // -- Serialization support --
1619
1620    /**
1621     * Saves the content of this URI to the given serial stream.
1622     *
1623     * <p> The only serializable field of a URI instance is its <tt>string</tt>
1624     * field.  That field is given a value, if it does not have one already,
1625     * and then the {@link java.io.ObjectOutputStream#defaultWriteObject()}
1626     * method of the given object-output stream is invoked. </p>
1627     *
1628     * @param  os  The object-output stream to which this object
1629     *             is to be written
1630     */
1631    private void writeObject(ObjectOutputStream os)
1632	throws IOException
1633    {
1634	defineString();
1635	os.defaultWriteObject();	// Writes the string field only
1636    }
1637
1638    /**
1639     * Reconstitutes a URI from the given serial stream.
1640     *
1641     * <p> The {@link java.io.ObjectInputStream#defaultReadObject()} method is
1642     * invoked to read the value of the <tt>string</tt> field.  The result is
1643     * then parsed in the usual way.
1644     *
1645     * @param  is  The object-input stream from which this object
1646     *             is being read
1647     */
1648    private void readObject(ObjectInputStream is)
1649	throws ClassNotFoundException, IOException
1650    {
1651	port = -1;			// Argh
1652	is.defaultReadObject();
1653	try {
1654	    new Parser(string).parse(false);
1655	} catch (URISyntaxException x) {
1656	    IOException y = new InvalidObjectException("Invalid URI");
1657	    y.initCause(x);
1658	    throw y;
1659	}
1660    }
1661
1662
1663    // -- End of public methods --
1664
1665
1666    // -- Utility methods for string-field comparison and hashing --
1667
1668    // These methods return appropriate values for null string arguments,
1669    // thereby simplifying the equals, hashCode, and compareTo methods.
1670    //
1671    // The case-ignoring methods should only be applied to strings whose
1672    // characters are all known to be US-ASCII.  Because of this restriction,
1673    // these methods are faster than the similar methods in the String class.
1674
1675    // US-ASCII only
1676    private static int toLower(char c) {
1677	if ((c >= 'A') && (c <= 'Z'))
1678	    return c + ('a' - 'A');
1679	return c;
1680    }
1681
1682    private static boolean equal(String s, String t) {
1683	if (s == t) return true;
1684	if ((s != null) && (t != null)) {
1685	    if (s.length() != t.length())
1686		return false;
1687	    if (s.indexOf('%') < 0)
1688		return s.equals(t);
1689	    int n = s.length();
1690	    for (int i = 0; i < n;) {
1691		char c = s.charAt(i);
1692		char d = t.charAt(i);
1693		if (c != '%') {
1694		    if (c != d)
1695			return false;
1696		    i++;
1697		    continue;
1698		}
1699		i++;
1700		if (toLower(s.charAt(i)) != toLower(t.charAt(i)))
1701		    return false;
1702		i++;
1703		if (toLower(s.charAt(i)) != toLower(t.charAt(i)))
1704		    return false;
1705		i++;
1706	    }
1707	    return true;
1708	}
1709	return false;
1710    }
1711
1712    // US-ASCII only
1713    private static boolean equalIgnoringCase(String s, String t) {
1714	if (s == t) return true;
1715	if ((s != null) && (t != null)) {
1716	    int n = s.length();
1717	    if (t.length() != n)
1718		return false;
1719	    for (int i = 0; i < n; i++) {
1720		if (toLower(s.charAt(i)) != toLower(t.charAt(i)))
1721		    return false;
1722	    }
1723	    return true;
1724	}
1725	return false;
1726    }
1727
1728    private static int hash(int hash, String s) {
1729	if (s == null) return hash;
1730	return hash * 127 + s.hashCode();
1731    }
1732
1733    // US-ASCII only
1734    private static int hashIgnoringCase(int hash, String s) {
1735	if (s == null) return hash;
1736	int h = hash;
1737	int n = s.length();
1738	for (int i = 0; i < n; i++)
1739	    h = 31 * h + toLower(s.charAt(i));
1740	return h;
1741    }
1742
1743    private static int compare(String s, String t) {
1744	if (s == t) return 0;
1745	if (s != null) {
1746	    if (t != null)
1747		return s.compareTo(t);
1748	    else
1749		return +1;
1750	} else {
1751	    return -1;
1752	}
1753    }
1754
1755    // US-ASCII only
1756    private static int compareIgnoringCase(String s, String t) {
1757	if (s == t) return 0;
1758	if (s != null) {
1759	    if (t != null) {
1760		int sn = s.length();
1761		int tn = t.length();
1762		int n = sn < tn ? sn : tn;
1763		for (int i = 0; i < n; i++) {
1764		    int c = toLower(s.charAt(i)) - toLower(t.charAt(i));
1765		    if (c != 0)
1766			return c;
1767		}
1768		return sn - tn;
1769	    }
1770	    return +1;
1771	} else {
1772	    return -1;
1773	}
1774    }
1775
1776
1777    // -- String construction --
1778
1779    // If a scheme is given then the path, if given, must be absolute
1780    //
1781    private static void checkPath(String s, String scheme, String path)
1782	throws URISyntaxException
1783    {
1784	if (scheme != null) {
1785	    if ((path != null)
1786		&& ((path.length() > 0) && (path.charAt(0) != '/')))
1787		throw new URISyntaxException(s,
1788					     "Relative path in absolute URI");
1789	}
1790    }
1791
1792    private void appendAuthority(StringBuffer sb,
1793				 String authority,
1794				 String userInfo,
1795				 String host,
1796				 int port)
1797    {
1798	if (host != null) {
1799	    sb.append("//");
1800	    if (userInfo != null) {
1801		sb.append(quote(userInfo, L_USERINFO, H_USERINFO));
1802		sb.append('@');
1803	    }
1804	    boolean needBrackets = ((host.indexOf(':') >= 0)
1805				    && !host.startsWith("[")
1806				    && !host.endsWith("]"));
1807	    if (needBrackets) sb.append('[');
1808	    sb.append(host);
1809	    if (needBrackets) sb.append(']');
1810	    if (port != -1) {
1811		sb.append(':');
1812		sb.append(port);
1813	    }
1814	} else if (authority != null) {
1815	    sb.append("//");
1816	    if (authority.startsWith("[")) {
1817		int end = authority.indexOf("]");
1818		if (end != -1 && authority.indexOf(":")!=-1) {
1819		    String doquote, dontquote;
1820		    if (end == authority.length()) {
1821			dontquote = authority;
1822			doquote = "";
1823		    } else {
1824		    	dontquote = authority.substring(0,end+1);
1825			doquote = authority.substring(end+1);
1826		    }
1827		    sb.append (dontquote);
1828	    	    sb.append(quote(doquote, 
1829			    L_REG_NAME | L_SERVER,
1830			    H_REG_NAME | H_SERVER));
1831		}
1832	    } else {
1833	    	sb.append(quote(authority,
1834			    L_REG_NAME | L_SERVER,
1835			    H_REG_NAME | H_SERVER));
1836	    }
1837	}
1838    }
1839
1840    private void appendSchemeSpecificPart(StringBuffer sb,
1841					  String opaquePart,
1842					  String authority,
1843					  String userInfo,
1844					  String host,
1845					  int port,
1846					  String path,
1847					  String query)
1848    {
1849	if (opaquePart != null) {
1850	    /* check if SSP begins with an IPv6 address
1851	     * because we must not quote a literal IPv6 address
1852	     */
1853	    if (opaquePart.startsWith("//[")) {
1854		int end =  opaquePart.indexOf("]");
1855		if (end != -1 && opaquePart.indexOf(":")!=-1) {
1856		    String doquote, dontquote;
1857		    if (end == opaquePart.length()) {
1858			dontquote = opaquePart;
1859			doquote = "";
1860		    } else {
1861		    	dontquote = opaquePart.substring(0,end+1);
1862			doquote = opaquePart.substring(end+1);
1863		    }
1864		    sb.append (dontquote);
1865	    	    sb.append(quote(doquote, L_URIC, H_URIC));
1866		}
1867	    } else {
1868	    	sb.append(quote(opaquePart, L_URIC, H_URIC));
1869	    }
1870	} else {
1871	    appendAuthority(sb, authority, userInfo, host, port);
1872	    if (path != null) 
1873		sb.append(quote(path, L_PATH, H_PATH));
1874	    if (query != null) {
1875		sb.append('?');
1876		sb.append(quote(query, L_URIC, H_URIC));
1877	    }
1878	}
1879    }
1880
1881    private void appendFragment(StringBuffer sb, String fragment) {
1882	if (fragment != null) {
1883	    sb.append('#');
1884	    sb.append(quote(fragment, L_URIC, H_URIC));
1885	}
1886    }
1887
1888    private String toString(String scheme,
1889			    String opaquePart,
1890			    String authority,
1891			    String userInfo,
1892			    String host,
1893			    int port,
1894			    String path,
1895			    String query,
1896			    String fragment)
1897    {
1898	StringBuffer sb = new StringBuffer();
1899	if (scheme != null) {
1900	    sb.append(scheme);
1901	    sb.append(':');
1902	}
1903	appendSchemeSpecificPart(sb, opaquePart,
1904				 authority, userInfo, host, port,
1905				 path, query);
1906	appendFragment(sb, fragment);
1907	return sb.toString();
1908    }
1909
1910    private void defineSchemeSpecificPart() {
1911	if (schemeSpecificPart != null) return;
1912	StringBuffer sb = new StringBuffer();
1913	appendSchemeSpecificPart(sb, null, getAuthority(), getUserInfo(),
1914				 host, port, getPath(), getQuery());
1915	if (sb.length() == 0) return;
1916	schemeSpecificPart = sb.toString();
1917    }
1918
1919    private void defineString() {
1920	if (string != null) return;
1921
1922	StringBuffer sb = new StringBuffer();
1923        if (scheme != null) {
1924            sb.append(scheme);
1925            sb.append(':');
1926        }
1927	if (isOpaque()) {
1928            sb.append(schemeSpecificPart);
1929        } else {
1930	    if (host != null) {
1931                sb.append("//");
1932                if (userInfo != null) {
1933                    sb.append(userInfo);
1934                    sb.append('@');
1935                }
1936                boolean needBrackets = ((host.indexOf(':') >= 0)
1937                                    && !host.startsWith("[")
1938                                    && !host.endsWith("]"));
1939                if (needBrackets) sb.append('[');
1940                sb.append(host);
1941                if (needBrackets) sb.append(']');
1942                if (port != -1) {
1943                    sb.append(':');
1944                    sb.append(port);
1945                }
1946            } else if (authority != null) {
1947                sb.append("//");
1948                sb.append(authority);
1949	    }
1950            if (path != null)
1951                sb.append(path);
1952            if (query != null) {
1953                sb.append('?');
1954                sb.append(query);
1955            }
1956        }
1957	if (fragment != null) {
1958            sb.append('#');
1959            sb.append(fragment);
1960	}
1961	string = sb.toString();
1962    }
1963
1964
1965    // -- Normalization, resolution, and relativization --
1966
1967    // RFC2396 5.2 (6)
1968    private static String resolvePath(String base, String child,
1969				      boolean absolute)
1970    {
1971        int i = base.lastIndexOf('/');
1972	int cn = child.length();
1973	String path = "";
1974
1975	if (cn == 0) {
1976	    // 5.2 (6a)
1977	    if (i >= 0)
1978		path = base.substring(0, i + 1);
1979	} else {
1980	    StringBuffer sb = new StringBuffer(base.length() + cn);
1981	    // 5.2 (6a)
1982	    if (i >= 0)
1983		sb.append(base.substring(0, i + 1));
1984	    // 5.2 (6b)
1985	    sb.append(child);
1986	    path = sb.toString();
1987	}
1988
1989	// 5.2 (6c-f)
1990	String np = normalize(path);
1991
1992	// 5.2 (6g): If the result is absolute but the path begins with "../",
1993	// then we simply leave the path as-is
1994
1995	return np;
1996    }
1997
1998    // RFC2396 5.2
1999    private static URI resolve(URI base, URI child) {
2000	// check if child if opaque first so that NPE is thrown 
2001	// if child is null.
2002	if (child.isOpaque() || base.isOpaque())
2003	    return child;
2004
2005	// 5.2 (2): Reference to current document (lone fragment)
2006	if ((child.scheme == null) && (child.authority == null)
2007	    && child.path.equals("") && (child.fragment != null)
2008	    && (child.query == null)) {
2009	    if ((base.fragment != null)
2010		&& child.fragment.equals(base.fragment)) {
2011		return base;
2012	    }
2013	    URI ru = new URI();
2014	    ru.scheme = base.scheme;
2015	    ru.authority = base.authority;
2016	    ru.userInfo = base.userInfo;
2017	    ru.host = base.host;
2018	    ru.port = base.port;
2019	    ru.path = base.path;
2020	    ru.fragment = child.fragment;
2021	    ru.query = base.query;
2022	    return ru;
2023	}
2024
2025	// 5.2 (3): Child is absolute
2026	if (child.scheme != null)
2027	    return child;
2028
2029	URI ru = new URI();		// Resolved URI
2030	ru.scheme = base.scheme;
2031	ru.query = child.query;
2032	ru.fragment = child.fragment;
2033
2034	// 5.2 (4): Authority
2035	if (child.authority == null) {
2036	    ru.authority = base.authority;
2037	    ru.host = base.host;
2038	    ru.userInfo = base.userInfo;
2039	    ru.port = base.port;
2040
2041	    String cp = (child.path == null) ? "" : child.path;
2042	    if ((cp.length() > 0) && (cp.charAt(0) == '/')) {
2043		// 5.2 (5): Child path is absolute
2044		ru.path = child.path;
2045	    } else {
2046		// 5.2 (6): Resolve relative path
2047		ru.path = resolvePath(base.path, cp, base.isAbsolute());
2048	    }
2049	} else {
2050	    ru.authority = child.authority;
2051	    ru.host = child.host;
2052	    ru.userInfo = child.userInfo;
2053	    ru.host = child.host;
2054	    ru.port = child.port;
2055	    ru.path = child.path;
2056	}
2057
2058	// 5.2 (7): Recombine (nothing to do here)
2059	return ru;
2060    }
2061
2062    // If the given URI's path is normal then return the URI;
2063    // o.w., return a new URI containing the normalized path.
2064    //
2065    private static URI normalize(URI u) {
2066	if (u.isOpaque() || (u.path == null) || (u.path.length() == 0))
2067	    return u;
2068
2069	String np = normalize(u.path);
2070	if (np == u.path)
2071	    return u;
2072
2073	URI v = new URI();
2074	v.scheme = u.scheme;
2075	v.fragment = u.fragment;
2076	v.authority = u.authority;
2077	v.userInfo = u.userInfo;
2078	v.host = u.host;
2079	v.port = u.port;
2080	v.path = np;
2081	v.query = u.query;
2082	return v;
2083    }
2084
2085    // If both URIs are hierarchical, their scheme and authority components are
2086    // identical, and the base path is a prefix of the child's path, then
2087    // return a relative URI that, when resolved against the base, yields the
2088    // child; otherwise, return the child.
2089    //
2090    private static URI relativize(URI base, URI child) {
2091	// check if child if opaque first so that NPE is thrown 
2092        // if child is null.
2093	if (child.isOpaque() || base.isOpaque())
2094	    return child;
2095	if (!equalIgnoringCase(base.scheme, child.scheme)
2096	    || !equal(base.authority, child.authority))
2097	    return child;
2098
2099	String bp = normalize(base.path);
2100	String cp = normalize(child.path);
2101	if (!bp.equals(cp)) {
2102	    if (!bp.endsWith("/"))
2103		bp = bp + "/";
2104	    if (!cp.startsWith(bp))
2105		return child;
2106	}
2107
2108	URI v = new URI();
2109	v.path = cp.substring(bp.length());
2110	v.query = child.query;
2111	v.fragment = child.fragment;
2112	return v;
2113    }
2114
2115
2116
2117    // -- Path normalization --
2118
2119    // The following algorithm for path normalization avoids the creation of a
2120    // string object for each segment, as well as the use of a string buffer to
2121    // compute the final result, by using a single char array and editing it in
2122    // place.  The array is first split into segments, replacing each slash
2123    // with '\0' and creating a segment-index array, each element of which is
2124    // the index of the first char in the corresponding segment.  We then walk
2125    // through both arrays, removing ".", "..", and other segments as necessary
2126    // by setting their entries in the index array to -1.  Finally, the two
2127    // arrays are used to rejoin the segments and compute the final result.
2128    //
2129    // This code is based upon src/solaris/native/java/io/canonicalize_md.c
2130
2131
2132    // Check the given path to see if it might need normalization.  A path
2133    // might need normalization if it contains duplicate slashes, a "."
2134    // segment, or a ".." segment.  Return -1 if no further normalization is
2135    // possible, otherwise return the number of segments found.
2136    //
2137    // This method takes a string argument rather than a char array so that
2138    // this test can be performed without invoking path.toCharArray().
2139    //
2140    static private int needsNormalization(String path) {
2141	boolean normal = true;
2142	int ns = 0;			// Number of segments
2143	int end = path.length() - 1;	// Index of last char in path
2144	int p = 0;			// Index of next char in path
2145
2146	// Skip initial slashes
2147	while (p <= end) {
2148	    if (path.charAt(p) != '/') break;
2149	    p++;
2150	}
2151	if (p > 1) normal = false;
2152
2153	// Scan segments
2154	while (p <= end) {
2155
2156	    // Looking at "." or ".." ?
2157	    if ((path.charAt(p) == '.')
2158		&& ((p == end)
2159		    || ((path.charAt(p + 1) == '/')
2160			|| ((path.charAt(p + 1) == '.')
2161			    && ((p + 1 == end)
2162				|| (path.charAt(p + 2) == '/')))))) {
2163		normal = false;
2164	    }
2165	    ns++;
2166
2167	    // Find beginning of next segment
2168	    while (p <= end) {
2169		if (path.charAt(p++) != '/')
2170		    continue;
2171
2172		// Skip redundant slashes
2173		while (p <= end) {
2174		    if (path.charAt(p) != '/') break;
2175		    normal = false;
2176		    p++;
2177		}
2178
2179		break;
2180	    }
2181	}
2182
2183	return normal ? -1 : ns;
2184    }
2185
2186
2187    // Split the given path into segments, replacing slashes with nulls and
2188    // filling in the given segment-index array.
2189    //
2190    // Preconditions:
2191    //   segs.length == Number of segments in path
2192    //
2193    // Postconditions:
2194    //   All slashes in path replaced by '\0'
2195    //   segs[i] == Index of first char in segment i (0 <= i < segs.length)
2196    //
2197    static private void split(char[] path, int[] segs) {
2198	int end = path.length - 1;	// Index of last char in path
2199	int p = 0;			// Index of next char in path
2200	int i = 0;			// Index of current segment
2201
2202	// Skip initial slashes
2203	while (p <= end) {
2204	    if (path[p] != '/') break;
2205	    path[p] = '\0';
2206	    p++;
2207	}
2208
2209	while (p <= end) {
2210
2211	    // Note start of segment
2212	    segs[i++] = p++;
2213
2214	    // Find beginning of next segment
2215	    while (p <= end) {
2216		if (path[p++] != '/')
2217		    continue;
2218		path[p - 1] = '\0';
2219
2220		// Skip redundant slashes
2221		while (p <= end) {
2222		    if (path[p] != '/') break;
2223		    path[p++] = '\0';
2224		}
2225		break;
2226	    }
2227	}
2228
2229	if (i != segs.length)
2230	    throw new InternalError();	// ASSERT
2231    }
2232
2233
2234    // Join the segments in the given path according to the given segment-index
2235    // array, ignoring those segments whose index entries have been set to -1,
2236    // and inserting slashes as needed.  Return the length of the resulting
2237    // path.
2238    //
2239    // Preconditions:
2240    //   segs[i] == -1 implies segment i is to be ignored
2241    //   path computed by split, as above, with '\0' having replaced '/'
2242    //
2243    // Postconditions:
2244    //   path[0] .. path[return value] == Resulting path
2245    //
2246    static private int join(char[] path, int[] segs) {
2247	int ns = segs.length;		// Number of segments
2248	int end = path.length - 1;	// Index of last char in path
2249	int p = 0;			// Index of next path char to write
2250
2251	if (path[p] == '\0') {
2252	    // Restore initial slash for absolute paths
2253	    path[p++] = '/';
2254	}
2255
2256	for (int i = 0; i < ns; i++) {
2257	    int q = segs[i];		// Current segment
2258	    if (q == -1)
2259		// Ignore this segment
2260		continue;
2261
2262	    if (p == q) {
2263		// We're already at this segment, so just skip to its end
2264		while ((p <= end) && (path[p] != '\0'))
2265		    p++;
2266		if (p <= end) {
2267		    // Preserve trailing slash
2268		    path[p++] = '/';
2269		}
2270	    } else if (p < q) {
2271		// Copy q down to p
2272		while ((q <= end) && (path[q] != '\0'))
2273		    path[p++] = path[q++];
2274		if (q <= end) {
2275		    // Preserve trailing slash
2276		    path[p++] = '/';
2277		}
2278	    } else
2279		throw new InternalError(); // ASSERT false
2280	}
2281
2282	return p;
2283    }
2284
2285
2286    // Remove "." segments from the given path, and remove segment pairs
2287    // consisting of a non-".." segment followed by a ".." segment.
2288    //
2289    private static void removeDots(char[] path, int[] segs) {
2290	int ns = segs.length;
2291	int end = path.length - 1;
2292
2293	for (int i = 0; i < ns; i++) {
2294	    int dots = 0;		// Number of dots found (0, 1, or 2)
2295
2296	    // Find next occurrence of "." or ".."
2297	    do {
2298		int p = segs[i];
2299		if (path[p] == '.') {
2300		    if (p == end) {
2301			dots = 1;
2302			break;
2303		    } else if (path[p + 1] == '\0') {
2304			dots = 1;
2305			break;
2306		    } else if ((path[p + 1] == '.')
2307			       && ((p + 1 == end)
2308				   || (path[p + 2] == '\0'))) {
2309			dots = 2;
2310			break;
2311		    }
2312		}
2313		i++;
2314	    } while (i < ns);
2315	    if ((i > ns) || (dots == 0))
2316		break;
2317
2318	    if (dots == 1) {
2319		// Remove this occurrence of "."
2320		segs[i] = -1;
2321	    } else {
2322		// If there is a preceding non-".." segment, remove both that
2323		// segment and this occurrence of ".."; otherwise, leave this
2324		// ".." segment as-is.
2325		int j;
2326		for (j = i - 1; j >= 0; j--) {
2327		    if (segs[j] != -1) break;
2328		}
2329		if (j >= 0) {
2330		    int q = segs[j];
2331		    if (!((path[q] == '.')
2332			  && (path[q + 1] == '.')
2333			  && (path[q + 2] == '\0'))) {
2334			segs[i] = -1;
2335			segs[j] = -1;
2336		    }
2337		}
2338	    }
2339	}
2340    }
2341
2342
2343    // DEVIATION: If the normalized path is relative, and if the first
2344    // segment could be parsed as a scheme name, then prepend a "." segment
2345    //
2346    private static void maybeAddLeadingDot(char[] path, int[] segs) {
2347
2348	if (path[0] == '\0')
2349	    // The path is absolute
2350	    return;
2351
2352	int ns = segs.length;
2353	int f = 0;			// Index of first segment
2354	while (f < ns) {
2355	    if (segs[f] >= 0)
2356		break;
2357	    f++;
2358	}
2359	if ((f >= ns) || (f == 0))
2360	    // The path is empty, or else the original first segment survived,
2361	    // in which case we already know that no leading "." is needed
2362	    return;
2363
2364	int p = segs[f];
2365	while ((p < path.length) && (path[p] != ':') && (path[p] != '\0')) p++;
2366	if (p >= path.length || path[p] == '\0')
2367	    // No colon in first segment, so no "." needed
2368	    return;
2369
2370	// At this point we know that the first segment is unused,
2371	// hence we can insert a "." segment at that position
2372	path[0] = '.';
2373	path[1] = '\0';
2374	segs[0] = 0;
2375    }
2376
2377
2378    // Normalize the given path string.  A normal path string has no empty
2379    // segments (i.e., occurrences of "//"), no segments equal to ".", and no
2380    // segments equal to ".." that are preceded by a segment not equal to "..".
2381    // In contrast to Unix-style pathname normalization, for URI paths we
2382    // always retain trailing slashes.
2383    //
2384    private static String normalize(String ps) {
2385
2386	// Does this path need normalization?
2387	int ns = needsNormalization(ps);	// Number of segments
2388	if (ns < 0)
2389	    // Nope -- just return it
2390	    return ps;
2391
2392	char[] path = ps.toCharArray();		// Path in char-array form
2393
2394	// Split path into segments
2395	int[] segs = new int[ns];		// Segment-index array
2396	split(path, segs);
2397
2398	// Remove dots
2399	removeDots(path, segs);
2400
2401	// Prevent scheme-name confusion
2402	maybeAddLeadingDot(path, segs);
2403
2404	// Join the remaining segments and return the result
2405	String s = new String(path, 0, join(path, segs));
2406	if (s.equals(ps)) {
2407	    // string was already normalized
2408	    return ps;
2409	}
2410	return s;
2411    }
2412
2413
2414
2415    // -- Character classes for parsing --
2416
2417    // RFC2396 precisely specifies which characters in the US-ASCII charset are
2418    // permissible in the various components of a URI reference.  We here
2419    // define a set of mask pairs to aid in enforcing these restrictions.  Each
2420    // mask pair consists of two longs, a low mask and a high mask.  Taken
2421    // together they represent a 128-bit mask, where bit i is set iff the
2422    // character with value i is permitted.
2423    //
2424    // This approach is more efficient than sequentially searching arrays of
2425    // permitted characters.  It could be made still more efficient by
2426    // precompiling the mask information so that a character's presence in a
2427    // given mask could be determined by a single table lookup.
2428
2429    // Compute the low-order mask for the characters in the given string
2430    private static long lowMask(String chars) {
2431	int n = chars.length();
2432	long m = 0;
2433	for (int i = 0; i < n; i++) {
2434	    char c = chars.charAt(i);
2435	    if (c < 64)
2436		m |= (1L << c);
2437	}
2438	return m;
2439    }
2440
2441    // Compute the high-order mask for the characters in the given string
2442    private static long highMask(String chars) {
2443	int n = chars.length();
2444	long m = 0;
2445	for (int i = 0; i < n; i++) {
2446	    char c = chars.charAt(i);
2447	    if ((c >= 64) && (c < 128))
2448		m |= (1L << (c - 64));
2449	}
2450	return m;
2451    }
2452
2453    // Compute a low-order mask for the characters
2454    // between first and last, inclusive
2455    private static long lowMask(char first, char last) {
2456	long m = 0;
2457	int f = Math.max(Math.min(first, 63), 0);
2458	int l = Math.max(Math.min(last, 63), 0);
2459	for (int i = f; i <= l; i++)
2460	    m |= 1L << i;
2461	return m;
2462    }
2463
2464    // Compute a high-order mask for the characters
2465    // between first and last, inclusive
2466    private static long highMask(char first, char last) {
2467	long m = 0;
2468	int f = Math.max(Math.min(first, 127), 64) - 64;
2469	int l = Math.max(Math.min(last, 127), 64) - 64;
2470	for (int i = f; i <= l; i++)
2471	    m |= 1L << i;
2472	return m;
2473    }
2474
2475    // Tell whether the given character is permitted by the given mask pair
2476    private static boolean match(char c, long lowMask, long highMask) {
2477	if (c < 64)
2478	    return ((1L << c) & lowMask) != 0;
2479	if (c < 128)
2480	    return ((1L << (c - 64)) & highMask) != 0;
2481	return false;
2482    }
2483
2484    // Character-class masks, in reverse order from RFC2396 because
2485    // initializers for static fields cannot make forward references.
2486
2487    // digit    = "0" | "1" | "2" | "3" | "4" | "5" | "6" | "7" |
2488    //            "8" | "9"
2489    private static final long L_DIGIT = lowMask('0', '9');
2490    private static final long H_DIGIT = 0L;
2491
2492    // upalpha  = "A" | "B" | "C" | "D" | "E" | "F" | "G" | "H" | "I" |
2493    //            "J" | "K" | "L" | "M" | "N" | "O" | "P" | "Q" | "R" |
2494    //            "S" | "T" | "U" | "V" | "W" | "X" | "Y" | "Z"
2495    private static final long L_UPALPHA = 0L;
2496    private static final long H_UPALPHA = highMask('A', 'Z');
2497
2498    // lowalpha = "a" | "b" | "c" | "d" | "e" | "f" | "g" | "h" | "i" |
2499    //            "j" | "k" | "l" | "m" | "n" | "o" | "p" | "q" | "r" |
2500    //            "s" | "t" | "u" | "v" | "w" | "x" | "y" | "z"
2501    private static final long L_LOWALPHA = 0L;
2502    private static final long H_LOWALPHA = highMask('a', 'z');
2503
2504    // alpha         = lowalpha | upalpha
2505    private static final long L_ALPHA = L_LOWALPHA | L_UPALPHA;
2506    private static final long H_ALPHA = H_LOWALPHA | H_UPALPHA;
2507
2508    // alphanum      = alpha | digit
2509    private static final long L_ALPHANUM = L_DIGIT | L_ALPHA;
2510    private static final long H_ALPHANUM = H_DIGIT | H_ALPHA;
2511
2512    // hex           = digit | "A" | "B" | "C" | "D" | "E" | "F" |
2513    //                         "a" | "b" | "c" | "d" | "e" | "f"
2514    private static final long L_HEX = L_DIGIT;
2515    private static final long H_HEX = highMask('A', 'F') | highMask('a', 'f');
2516
2517    // mark          = "-" | "_" | "." | "!" | "~" | "*" | "'" |
2518    //                 "(" | ")"
2519    private static final long L_MARK = lowMask("-_.!~*'()");
2520    private static final long H_MARK = highMask("-_.!~*'()");
2521
2522    // unreserved    = alphanum | mark
2523    private static final long L_UNRESERVED = L_ALPHANUM | L_MARK;
2524    private static final long H_UNRESERVED = H_ALPHANUM | H_MARK;
2525
2526    // reserved      = ";" | "/" | "?" | ":" | "@" | "&" | "=" | "+" |
2527    //                 "$" | "," | "[" | "]"
2528    // Added per RFC2732: "[", "]"
2529    private static final long L_RESERVED = lowMask(";/?:@&=+$,[]");
2530    private static final long H_RESERVED = highMask(";/?:@&=+$,[]");
2531
2532    // The zero'th bit is used to indicate that escape pairs and non-US-ASCII
2533    // characters are allowed; this is handled by the scanEscape method below.
2534    private static final long L_ESCAPED = 1L;
2535    private static final long H_ESCAPED = 0L;
2536
2537    // uric          = reserved | unreserved | escaped
2538    private static final long L_URIC = L_RESERVED | L_UNRESERVED | L_ESCAPED;
2539    private static final long H_URIC = H_RESERVED | H_UNRESERVED | H_ESCAPED;
2540
2541    // pchar         = unreserved | escaped |
2542    //                 ":" | "@" | "&" | "=" | "+" | "$" | ","
2543    private static final long L_PCHAR
2544	= L_UNRESERVED | L_ESCAPED | lowMask(":@&=+$,");
2545    private static final long H_PCHAR
2546	= H_UNRESERVED | H_ESCAPED | highMask(":@&=+$,");
2547
2548    // All valid path characters
2549    private static final long L_PATH = L_PCHAR | lowMask(";/");
2550    private static final long H_PATH = H_PCHAR | highMask(";/");
2551
2552    // Dash, for use in domainlabel and toplabel
2553    private static final long L_DASH = lowMask("-");
2554    private static final long H_DASH = highMask("-");
2555
2556    // Dot, for use in hostnames
2557    private static final long L_DOT = lowMask(".");
2558    private static final long H_DOT = highMask(".");
2559
2560    // userinfo      = *( unreserved | escaped |
2561    //                    ";" | ":" | "&" | "=" | "+" | "$" | "," )
2562    private static final long L_USERINFO
2563	= L_UNRESERVED | L_ESCAPED | lowMask(";:&=+$,");
2564    private static final long H_USERINFO
2565	= H_UNRESERVED | H_ESCAPED | highMask(";:&=+$,");
2566
2567    // reg_name      = 1*( unreserved | escaped | "$" | "," |
2568    //                     ";" | ":" | "@" | "&" | "=" | "+" )
2569    private static final long L_REG_NAME
2570	= L_UNRESERVED | L_ESCAPED | lowMask("$,;:@&=+");
2571    private static final long H_REG_NAME
2572	= H_UNRESERVED | H_ESCAPED | highMask("$,;:@&=+");
2573
2574    // All valid characters for server-based authorities
2575    private static final long L_SERVER
2576	= L_USERINFO | L_ALPHANUM | L_DASH | lowMask(".:@[]");
2577    private static final long H_SERVER
2578	= H_USERINFO | H_ALPHANUM | H_DASH | highMask(".:@[]");
2579
2580    // Special case of server authority that represents an IPv6 address
2581    // In this case, a % does not signify an escape sequence
2582    private static final long L_SERVER_PERCENT
2583	= L_SERVER | lowMask("%");
2584    private static final long H_SERVER_PERCENT
2585	= H_SERVER | highMask("%");
2586    private static final long L_LEFT_BRACKET = lowMask("[");
2587    private static final long H_LEFT_BRACKET = highMask("[");
2588
2589    // scheme        = alpha *( alpha | digit | "+" | "-" | "." )
2590    private static final long L_SCHEME = L_ALPHA | L_DIGIT | lowMask("+-.");
2591    private static final long H_SCHEME = H_ALPHA | H_DIGIT | highMask("+-.");
2592
2593    // uric_no_slash = unreserved | escaped | ";" | "?" | ":" | "@" |
2594    //                 "&" | "=" | "+" | "$" | ","
2595    private static final long L_URIC_NO_SLASH
2596	= L_UNRESERVED | L_ESCAPED | lowMask(";?:@&=+$,");
2597    private static final long H_URIC_NO_SLASH
2598	= H_UNRESERVED | H_ESCAPED | highMask(";?:@&=+$,");
2599
2600
2601    // -- Escaping and encoding --
2602
2603    private final static char[] hexDigits = {
2604	'0', '1', '2', '3', '4', '5', '6', '7',
2605	'8', '9', 'A', 'B', 'C', 'D', 'E', 'F'
2606    };
2607
2608    private static void appendEscape(StringBuffer sb, byte b) {
2609	sb.append('%');
2610	sb.append(hexDigits[(b >> 4) & 0x0f]);
2611	sb.append(hexDigits[(b >> 0) & 0x0f]);
2612    }
2613
2614    private static void appendEncoded(StringBuffer sb, char c) {
2615	ByteBuffer bb = null;
2616	try {
2617	    bb = ThreadLocalCoders.encoderFor("UTF-8")
2618		.encode(CharBuffer.wrap("" + c));
2619	} catch (CharacterCodingException x) {
2620	    assert false;
2621	}
2622	while (bb.hasRemaining()) {
2623	    int b = bb.get() & 0xff;
2624	    if (b >= 0x80)
2625		appendEscape(sb, (byte)b);
2626	    else
2627		sb.append((char)b);
2628	}
2629    }
2630
2631    // Quote any characters in s that are not permitted
2632    // by the given mask pair
2633    //
2634    private static String quote(String s, long lowMask, long highMask) {
2635	int n = s.length();
2636	StringBuffer sb = null;
2637	boolean allowNonASCII = ((lowMask & L_ESCAPED) != 0);
2638	for (int i = 0; i < s.length(); i++) {
2639	    char c = s.charAt(i);
2640	    if (c < '\u0080') {
2641		if (!match(c, lowMask, highMask)) {
2642		    if (sb == null) {
2643			sb = new StringBuffer();
2644			sb.append(s.substring(0, i));
2645		    }
2646		    appendEscape(sb, (byte)c);
2647		} else {
2648		    if (sb != null)
2649			sb.append(c);
2650		}
2651	    } else if (allowNonASCII
2652		       && (Character.isSpaceChar(c)
2653			   || Character.isISOControl(c))) {
2654		if (sb == null) {
2655		    sb = new StringBuffer();
2656		    sb.append(s.substring(0, i));
2657		}
2658		appendEncoded(sb, c);
2659	    } else {
2660		if (sb != null)
2661		    sb.append(c);
2662	    }
2663	}
2664	return (sb == null) ? s : sb.toString();
2665    }
2666
2667    // Encodes all characters >= \u0080 into escaped, normalized UTF-8 octets,
2668    // assuming that s is otherwise legal
2669    //
2670    private static String encode(String s) {
2671	int n = s.length();
2672	if (n == 0)
2673	    return s;
2674
2675	// First check whether we actually need to encode
2676	for (int i = 0;;) {
2677	    if (s.charAt(i) >= '\u0080')
2678		break;
2679	    if (++i >= n)
2680		return s;
2681	}
2682
2683	String ns = Normalizer.normalize(s, Normalizer.Form.NFC);
2684	ByteBuffer bb = null;
2685	try {
2686	    bb = ThreadLocalCoders.encoderFor("UTF-8")
2687		.encode(CharBuffer.wrap(ns));
2688	} catch (CharacterCodingException x) {
2689	    assert false;
2690	}
2691
2692	StringBuffer sb = new StringBuffer();
2693	while (bb.hasRemaining()) {
2694	    int b = bb.get() & 0xff;
2695	    if (b >= 0x80)
2696		appendEscape(sb, (byte)b);
2697	    else
2698		sb.append((char)b);
2699	}
2700	return sb.toString();
2701    }
2702
2703    private static int decode(char c) {
2704	if ((c >= '0') && (c <= '9'))
2705	    return c - '0';
2706	if ((c >= 'a') && (c <= 'f'))
2707	    return c - 'a' + 10;
2708	if ((c >= 'A') && (c <= 'F'))
2709	    return c - 'A' + 10;
2710	assert false;
2711	return -1;
2712    }
2713
2714    private static byte decode(char c1, char c2) {
2715	return (byte)(  ((decode(c1) & 0xf) << 4)
2716		      | ((decode(c2) & 0xf) << 0));
2717    }
2718
2719    // Evaluates all escapes in s, applying UTF-8 decoding if needed.  Assumes
2720    // that escapes are well-formed syntactically, i.e., of the form %XX.  If a
2721    // sequence of escaped octets is not valid UTF-8 then the erroneous octets
2722    // are replaced with '\uFFFD'.
2723    // Exception: any "%" found between "[]" is left alone. It is an IPv6 literal
2724    //            with a scope_id
2725    //
2726    private static String decode(String s) {
2727	if (s == null)
2728	    return s;
2729	int n = s.length();
2730	if (n == 0)
2731	    return s;
2732	if (s.indexOf('%') < 0)
2733	    return s;
2734
2735	byte[] ba = new byte[n];
2736	StringBuffer sb = new StringBuffer(n);
2737	ByteBuffer bb = ByteBuffer.allocate(n);
2738	CharBuffer cb = CharBuffer.allocate(n);
2739	CharsetDecoder dec = ThreadLocalCoders.decoderFor("UTF-8")
2740	    .onMalformedInput(CodingErrorAction.REPLACE)
2741	    .onUnmappableCharacter(CodingErrorAction.REPLACE);
2742
2743	// This is not horribly efficient, but it will do for now
2744	char c = s.charAt(0);
2745    	boolean betweenBrackets = false;
2746
2747	for (int i = 0; i < n;) {
2748	    assert c == s.charAt(i);	// Loop invariant
2749	    if (c == '[') {
2750		betweenBrackets = true;
2751	    } else if (betweenBrackets && c == ']') {
2752		betweenBrackets = false;
2753	    }
2754	    if (c != '%' || betweenBrackets) {
2755		sb.append(c);
2756		if (++i >= n)
2757		    break;
2758		c = s.charAt(i);
2759		continue;
2760	    }
2761	    bb.clear();
2762	    int ui = i;
2763	    for (;;) {
2764		assert (n - i >= 2);
2765		bb.put(decode(s.charAt(++i), s.charAt(++i)));
2766		if (++i >= n)
2767		    break;
2768		c = s.charAt(i);
2769		if (c != '%')
2770		    break;
2771	    }
2772	    bb.flip();
2773	    cb.clear();
2774	    dec.reset();
2775	    CoderResult cr = dec.decode(bb, cb, true);
2776	    assert cr.isUnderflow();
2777	    cr = dec.flush(cb);
2778	    assert cr.isUnderflow();
2779	    sb.append(cb.flip().toString());
2780	}
2781
2782	return sb.toString();
2783    }
2784
2785
2786    // -- Parsing --
2787
2788    // For convenience we wrap the input URI string in a new instance of the
2789    // following internal class.  This saves always having to pass the input
2790    // string as an argument to each internal scan/parse method.
2791
2792    private class Parser {
2793
2794	private String input;		// URI input string
2795	private boolean requireServerAuthority = false;
2796
2797	Parser(String s) {
2798	    input = s;
2799	    string = s;
2800	}
2801
2802	// -- Methods for throwing URISyntaxException in various ways --
2803
2804	private void fail(String reason) throws URISyntaxException {
2805	    throw new URISyntaxException(input, reason);
2806	}
2807
2808	private void fail(String reason, int p) throws URISyntaxException {
2809	    throw new URISyntaxException(input, reason, p);
2810	}
2811
2812	private void failExpecting(String expected, int p)
2813	    throws URISyntaxException
2814	{
2815	    fail("Expected " + expected, p);
2816	}
2817
2818	private void failExpecting(String expected, String prior, int p)
2819	    throws URISyntaxException
2820	{
2821	    fail("Expected " + expected + " following " + prior, p);
2822	}
2823
2824
2825	// -- Simple access to the input string --
2826
2827	// Return a substring of the input string
2828	//
2829	private String substring(int start, int end) {
2830	    return input.substring(start, end);
2831	}
2832
2833	// Return the char at position p,
2834	// assuming that p < input.length()
2835	//
2836	private char charAt(int p) {
2837	    return input.charAt(p);
2838	}
2839
2840	// Tells whether start < end and, if so, whether charAt(start) == c
2841	//
2842	private boolean at(int start, int end, char c) {
2843	    return (start < end) && (charAt(start) == c);
2844	}
2845
2846	// Tells whether start + s.length() < end and, if so,
2847	// whether the chars at the start position match s exactly
2848	//
2849	private boolean at(int start, int end, String s) {
2850	    int p = start;
2851	    int sn = s.length();
2852	    if (sn > end - p)
2853		return false;
2854	    int i = 0;
2855	    while (i < sn) {
2856		if (charAt(p++) != s.charAt(i)) {
2857		    break;
2858		}
2859		i++;
2860	    }
2861	    return (i == sn);
2862	}
2863
2864
2865	// -- Scanning --
2866
2867	// The various scan and parse methods that follow use a uniform
2868	// convention of taking the current start position and end index as
2869	// their first two arguments.  The start is inclusive while the end is
2870	// exclusive, just as in the String class, i.e., a start/end pair
2871	// denotes the left-open interval [start, end) of the input string.
2872	//
2873	// These methods never proceed past the end position.  They may return
2874	// -1 to indicate outright failure, but more often they simply return
2875	// the position of the first char after the last char scanned.  Thus
2876	// a typical idiom is
2877	//
2878	//     int p = start;
2879	//     int q = scan(p, end, ...);
2880	//     if (q > p)
2881	//         // We scanned something
2882	//         ...;
2883	//     else if (q == p)
2884	//         // We scanned nothing
2885	//         ...;
2886	//     else if (q == -1)
2887	//         // Something went wrong
2888	//         ...;
2889
2890
2891	// Scan a specific char: If the char at the given start position is
2892	// equal to c, return the index of the next char; otherwise, return the
2893	// start position.
2894	//
2895	private int scan(int start, int end, char c) {
2896	    if ((start < end) && (charAt(start) == c))
2897		return start + 1;
2898	    return start;
2899	}
2900
2901	// Scan forward from the given start position.  Stop at the first char
2902	// in the err string (in which case -1 is returned), or the first char
2903	// in the stop string (in which case the index of the preceding char is
2904	// returned), or the end of the input string (in which case the length
2905	// of the input string is returned).  May return the start position if
2906	// nothing matches.
2907	//
2908	private int scan(int start, int end, String err, String stop) {
2909	    int p = start;
2910	    while (p < end) {
2911		char c = charAt(p);
2912		if (err.indexOf(c) >= 0)
2913		    return -1;
2914		if (stop.indexOf(c) >= 0)
2915		    break;
2916		p++;
2917	    }
2918	    return p;
2919	}
2920
2921	// Scan a potential escape sequence, starting at the given position,
2922	// with the given first char (i.e., charAt(start) == c).
2923	//
2924	// This method assumes that if escapes are allowed then visible
2925	// non-US-ASCII chars are also allowed.
2926	//
2927	private int scanEscape(int start, int n, char first)
2928	    throws URISyntaxException
2929	{
2930	    int p = start;
2931	    char c = first;
2932	    if (c == '%') {
2933		// Process escape pair
2934		if ((p + 3 <= n)
2935		    && match(charAt(p + 1), L_HEX, H_HEX)
2936		    && match(charAt(p + 2), L_HEX, H_HEX)) {
2937		    return p + 3;
2938		}
2939		fail("Malformed escape pair", p);
2940	    } else if ((c > 128)
2941		       && !Character.isSpaceChar(c)
2942		       && !Character.isISOControl(c)) {
2943		// Allow unescaped but visible non-US-ASCII chars
2944		return p + 1;
2945	    }
2946	    return p;
2947	}
2948
2949	// Scan chars that match the given mask pair
2950	//
2951	private int scan(int start, int n, long lowMask, long highMask)
2952	    throws URISyntaxException
2953	{
2954	    int p = start;
2955	    while (p < n) {
2956		char c = charAt(p);
2957		if (match(c, lowMask, highMask)) {
2958		    p++;
2959		    continue;
2960		}
2961		if ((lowMask & L_ESCAPED) != 0) {
2962		    int q = scanEscape(p, n, c);
2963		    if (q > p) {
2964			p = q;
2965			continue;
2966		    }
2967		}
2968		break;
2969	    }
2970	    return p;
2971	}
2972
2973	// Check that each of the chars in [start, end) matches the given mask
2974	//
2975	private void checkChars(int start, int end,
2976				long lowMask, long highMask,
2977				String what)
2978	    throws URISyntaxException
2979	{
2980	    int p = scan(start, end, lowMask, highMask);
2981	    if (p < end)
2982		fail("Illegal character in " + what, p);
2983	}
2984
2985	// Check that the char at position p matches the given mask
2986	//
2987	private void checkChar(int p,
2988			       long lowMask, long highMask,
2989			       String what)
2990	    throws URISyntaxException
2991	{
2992	    checkChars(p, p + 1, lowMask, highMask, what);
2993	}
2994
2995
2996	// -- Parsing --
2997
2998	// [<scheme>:]<scheme-specific-part>[#<fragment>]
2999	//
3000	void parse(boolean rsa) throws URISyntaxException {
3001	    requireServerAuthority = rsa;
3002	    int ssp;			// Start of scheme-specific part
3003	    int n = input.length();
3004	    int p = scan(0, n, "/?#", ":");
3005	    if ((p >= 0) && at(p, n, ':')) {
3006		if (p == 0)
3007		    failExpecting("scheme name", 0);
3008		checkChar(0, L_ALPHA, H_ALPHA, "scheme name");
3009		checkChars(1, p, L_SCHEME, H_SCHEME, "scheme name");
3010		scheme = substring(0, p);
3011		p++;			// Skip ':'
3012		ssp = p;
3013		if (at(p, n, '/')) {
3014		    p = parseHierarchical(p, n);
3015		} else {
3016		    int q = scan(p, n, "", "#");
3017		    if (q <= p)
3018			failExpecting("scheme-specific part", p);
3019		    checkChars(p, q, L_URIC, H_URIC, "opaque part");
3020		    p = q;
3021		}
3022	    } else {
3023		ssp = 0;
3024		p = parseHierarchical(0, n);
3025	    }
3026	    schemeSpecificPart = substring(ssp, p);
3027	    if (at(p, n, '#')) {
3028		checkChars(p + 1, n, L_URIC, H_URIC, "fragment");
3029		fragment = substring(p + 1, n);
3030		p = n;
3031	    }
3032	    if (p < n)
3033		fail("end of URI", p);
3034	}
3035
3036	// [//authority]<path>[?<query>]
3037	//
3038	// DEVIATION from RFC2396: We allow an empty authority component as
3039	// long as it's followed by a non-empty path, query component, or
3040	// fragment component.  This is so that URIs such as "file:///foo/bar"
3041	// will parse.  This seems to be the intent of RFC2396, though the
3042	// grammar does not permit it.  If the authority is empty then the
3043	// userInfo, host, and port components are undefined.
3044	//
3045	// DEVIATION from RFC2396: We allow empty relative paths.  This seems
3046	// to be the intent of RFC2396, but the grammar does not permit it.
3047	// The primary consequence of this deviation is that "#f" parses as a
3048	// relative URI with an empty path.
3049	//
3050	private int parseHierarchical(int start, int n)
3051	    throws URISyntaxException
3052	{
3053	    int p = start;
3054	    if (at(p, n, '/') && at(p + 1, n, '/')) {
3055		p += 2;
3056		int q = scan(p, n, "", "/?#");
3057		if (q > p) {
3058		    p = parseAuthority(p, q);
3059		} else if (q < n) {
3060		    // DEVIATION: Allow empty authority prior to non-empty 
3061		    // path, query component or fragment identifier
3062		} else
3063		    failExpecting("authority", p);
3064	    }
3065	    int q = scan(p, n, "", "?#"); // DEVIATION: May be empty
3066	    checkChars(p, q, L_PATH, H_PATH, "path");
3067	    path = substring(p, q);
3068	    p = q;
3069	    if (at(p, n, '?')) {
3070		p++;
3071		q = scan(p, n, "", "#");
3072		checkChars(p, q, L_URIC, H_URIC, "query");
3073		query = substring(p, q);
3074		p = q;
3075	    }
3076	    return p;
3077	}
3078
3079	// authority     = server | reg_name
3080	//
3081	// Ambiguity: An authority that is a registry name rather than a server
3082	// might have a prefix that parses as a server.  We use the fact that
3083	// the authority component is always followed by '/' or the end of the
3084	// input string to resolve this: If the complete authority did not
3085	// parse as a server then we try to parse it as a registry name.
3086	//
3087	private int parseAuthority(int start, int n)
3088	    throws URISyntaxException
3089	{
3090	    int p = start;
3091	    int q = p;
3092	    URISyntaxException ex = null;
3093
3094	    boolean serverChars;
3095	    boolean regChars;
3096
3097	    if (scan(p, n, "", "]") > p) {
3098		// contains a literal IPv6 address, therefore % is allowed
3099	    	serverChars = (scan(p, n, L_SERVER_PERCENT, H_SERVER_PERCENT) == n);
3100	    } else {
3101	    	serverChars = (scan(p, n, L_SERVER, H_SERVER) == n);
3102	    }
3103	    regChars = (scan(p, n, L_REG_NAME, H_REG_NAME) == n);
3104
3105	    if (regChars && !serverChars) {
3106		// Must be a registry-based authority
3107		authority = substring(p, n);
3108		return n;
3109	    }
3110
3111	    if (serverChars) {
3112		// Might be (probably is) a server-based authority, so attempt
3113		// to parse it as such.  If the attempt fails, try to treat it
3114		// as a registry-based authority.
3115		try {
3116		    q = parseServer(p, n);
3117		    if (q < n)
3118			failExpecting("end of authority", q);
3119		    authority = substring(p, n);
3120		} catch (URISyntaxException x) {
3121		    // Undo results of failed parse
3122		    userInfo = null;
3123		    host = null;
3124		    port = -1;
3125		    if (requireServerAuthority) {
3126			// If we're insisting upon a server-based authority,
3127			// then just re-throw the exception
3128			throw x;
3129		    } else {
3130			// Save the exception in case it doesn't parse as a
3131			// registry either
3132			ex = x;
3133			q = p;
3134		    }
3135		}
3136	    }
3137
3138	    if (q < n) {
3139		if (regChars) {
3140		    // Registry-based authority
3141		    authority = substring(p, n);
3142		} else if (ex != null) {
3143		    // Re-throw exception; it was probably due to
3144		    // a malformed IPv6 address
3145		    throw ex;
3146		} else {
3147		    fail("Illegal character in authority", q);
3148		}
3149	    }
3150
3151	    return n;
3152	}
3153
3154
3155	// [<userinfo>@]<host>[:<port>]
3156	//
3157	private int parseServer(int start, int n)
3158	    throws URISyntaxException
3159	{
3160	    int p = start;
3161	    int q;
3162
3163	    // userinfo
3164	    q = scan(p, n, "/?#", "@");
3165	    if ((q >= p) && at(q, n, '@')) {
3166		checkChars(p, q, L_USERINFO, H_USERINFO, "user info");
3167		userInfo = substring(p, q);
3168		p = q + 1;		// Skip '@'
3169	    }
3170
3171	    // hostname, IPv4 address, or IPv6 address
3172	    if (at(p, n, '[')) {
3173		// DEVIATION from RFC2396: Support IPv6 addresses, per RFC2732
3174		p++;
3175		q = scan(p, n, "/?#", "]");
3176		if ((q > p) && at(q, n, ']')) {
3177		    // look for a "%" scope id
3178		    int r = scan (p, q, "", "%");
3179		    if (r > p) {
3180		    	parseIPv6Reference(p, r);
3181			if (r+1 == q) {
3182			    fail ("scope id expected");
3183			}
3184			checkChars (r+1, q, L_ALPHANUM, H_ALPHANUM, 
3185						"scope id");
3186		    } else {
3187		    	parseIPv6Reference(p, q);
3188		    }
3189	    	    host = substring(p-1, q+1);
3190		    p = q + 1;
3191		} else {
3192		    failExpecting("closing bracket for IPv6 address", q);
3193		}
3194	    } else {
3195		q = parseIPv4Address(p, n);
3196		if (q <= p)
3197		    q = parseHostname(p, n);
3198		p = q;
3199	    }
3200
3201	    // port
3202	    if (at(p, n, ':')) {
3203		p++;
3204		q = scan(p, n, "", "/");
3205		if (q > p) {
3206		    checkChars(p, q, L_DIGIT, H_DIGIT, "port number");
3207		    try {
3208			port = Integer.parseInt(substring(p, q));
3209		    } catch (NumberFormatException x) {
3210			fail("Malformed port number", p);
3211		    }
3212		    p = q;
3213		}
3214	    }
3215	    if (p < n)
3216		failExpecting("port number", p);
3217
3218	    return p;
3219	}
3220
3221	// Scan a string of decimal digits whose value fits in a byte
3222	//
3223	private int scanByte(int start, int n)
3224	    throws URISyntaxException
3225	{
3226	    int p = start;
3227	    int q = scan(p, n, L_DIGIT, H_DIGIT);
3228	    if (q <= p) return q;
3229	    if (Integer.parseInt(substring(p, q)) > 255) return p;
3230	    return q;
3231	}
3232
3233	// Scan an IPv4 address.
3234	//
3235	// If the strict argument is true then we require that the given
3236	// interval contain nothing besides an IPv4 address; if it is false
3237	// then we only require that it start with an IPv4 address.
3238	//
3239	// If the interval does not contain or start with (depending upon the
3240	// strict argument) a legal IPv4 address characters then we return -1
3241	// immediately; otherwise we insist that these characters parse as a
3242	// legal IPv4 address and throw an exception on failure.
3243	//
3244	// We assume that any string of decimal digits and dots must be an IPv4
3245	// address.  It won't parse as a hostname anyway, so making that
3246	// assumption here allows more meaningful exceptions to be thrown.
3247	//
3248	private int scanIPv4Address(int start, int n, boolean strict)
3249	    throws URISyntaxException
3250	{
3251	    int p = start;
3252	    int q;
3253	    int m = scan(p, n, L_DIGIT | L_DOT, H_DIGIT | H_DOT);
3254	    if ((m <= p) || (strict && (m != n)))
3255		return -1;
3256	    for (;;) {
3257		// Per RFC2732: At most three digits per byte
3258		// Further constraint: Each element fits in a byte
3259		if ((q = scanByte(p, m)) <= p) break;   p = q;
3260		if ((q = scan(p, m, '.')) <= p) break;  p = q;
3261		if ((q = scanByte(p, m)) <= p) break;   p = q;
3262		if ((q = scan(p, m, '.')) <= p) break;  p = q;
3263		if ((q = scanByte(p, m)) <= p) break;   p = q;
3264		if ((q = scan(p, m, '.')) <= p) break;  p = q;
3265		if ((q = scanByte(p, m)) <= p) break;   p = q;
3266		if (q < m) break;
3267		return q;
3268	    }
3269	    fail("Malformed IPv4 address", q);
3270	    return -1;
3271	}
3272
3273	// Take an IPv4 address: Throw an exception if the given interval
3274	// contains anything except an IPv4 address
3275	//
3276	private int takeIPv4Address(int start, int n, String expected)
3277	    throws URISyntaxException
3278	{
3279	    int p = scanIPv4Address(start, n, true);
3280	    if (p <= start)
3281		failExpecting(expected, start);
3282	    return p;
3283	}
3284
3285	// Attempt to parse an IPv4 address, returning -1 on failure but
3286	// allowing the given interval to contain [:<characters>] after
3287	// the IPv4 address.
3288	//
3289	private int parseIPv4Address(int start, int n) {
3290	    int p;
3291
3292	    try {
3293	        p = scanIPv4Address(start, n, false);
3294	    } catch (URISyntaxException x) {
3295		return -1;
3296            } catch (NumberFormatException nfe) {
3297		return -1;
3298            }
3299
3300	    if (p > start && p < n) {
3301	        // IPv4 address is followed by something - check that
3302		// it's a ":" as this is the only valid character to
3303		// follow an address.
3304		if (charAt(p) != ':') {
3305		    p = -1;
3306		}
3307	    }
3308
3309	    if (p > start)
3310		host = substring(start, p);
3311
3312	    return p;
3313	}
3314
3315	// hostname      = domainlabel [ "." ] | 1*( domainlabel "." ) toplabel [ "." ] 
3316	// domainlabel   = alphanum | alphanum *( alphanum | "-" ) alphanum
3317        // toplabel      = alpha | alpha *( alphanum | "-" ) alphanum
3318	//
3319	private int parseHostname(int start, int n)
3320	    throws URISyntaxException
3321	{
3322	    int p = start;
3323	    int q;
3324	    int l = -1;			// Start of last parsed label
3325
3326	    do {
3327		// domainlabel = alphanum [ *( alphanum | "-" ) alphanum ]
3328		q = scan(p, n, L_ALPHANUM, H_ALPHANUM);
3329		if (q <= p)
3330		    break;
3331		l = p;
3332		if (q > p) {
3333		    p = q;
3334		    q = scan(p, n, L_ALPHANUM | L_DASH, H_ALPHANUM | H_DASH);
3335		    if (q > p) {
3336			if (charAt(q - 1) == '-')
3337			    fail("Illegal character in hostname", q - 1);
3338			p = q;
3339		    }
3340		}
3341		q = scan(p, n, '.');
3342		if (q <= p)
3343		    break;
3344		p = q;
3345	    } while (p < n);
3346
3347	    if ((p < n) && !at(p, n, ':'))
3348		fail("Illegal character in hostname", p);
3349
3350	    if (l < 0)
3351		failExpecting("hostname", start);
3352
3353	    // for a fully qualified hostname check that the rightmost
3354	    // label starts with an alpha character.
3355	    if (l > start && !match(charAt(l), L_ALPHA, H_ALPHA)) {
3356		fail("Illegal character in hostname", l);
3357	    }
3358
3359	    host = substring(start, p);
3360	    return p;
3361	}
3362
3363
3364	// IPv6 address parsing, from RFC2373: IPv6 Addressing Architecture
3365	//
3366	// Bug: The grammar in RFC2373 Appendix B does not allow addresses of
3367	// the form ::12.34.56.78, which are clearly shown in the examples
3368	// earlier in the document.  Here is the original grammar:
3369	//
3370	//   IPv6address = hexpart [ ":" IPv4address ]
3371	//   hexpart     = hexseq | hexseq "::" [ hexseq ] | "::" [ hexseq ]
3372	//   hexseq      = hex4 *( ":" hex4)
3373	//   hex4        = 1*4HEXDIG
3374	//
3375	// We therefore use the following revised grammar:
3376	//
3377	//   IPv6address = hexseq [ ":" IPv4address ]
3378	//                 | hexseq [ "::" [ hexpost ] ]
3379	//                 | "::" [ hexpost ]
3380	//   hexpost     = hexseq | hexseq ":" IPv4address | IPv4address
3381	//   hexseq      = hex4 *( ":" hex4)
3382	//   hex4        = 1*4HEXDIG
3383	//
3384	// This covers all and only the following cases:
3385	//
3386	//   hexseq
3387	//   hexseq : IPv4address
3388	//   hexseq ::
3389	//   hexseq :: hexseq
3390	//   hexseq :: hexseq : IPv4address
3391	//   hexseq :: IPv4address
3392	//   :: hexseq
3393	//   :: hexseq : IPv4address
3394	//   :: IPv4address
3395	//   ::
3396	//
3397	// Additionally we constrain the IPv6 address as follows :-
3398	//
3399	//  i.  IPv6 addresses without compressed zeros should contain
3400	//      exactly 16 bytes.
3401	//
3402	//  ii. IPv6 addresses with compressed zeros should contain
3403	//      less than 16 bytes.
3404
3405	private int ipv6byteCount = 0;
3406
3407	private int parseIPv6Reference(int start, int n)
3408	    throws URISyntaxException
3409	{
3410	    int p = start;
3411	    int q;
3412	    boolean compressedZeros = false;
3413
3414	    q = scanHexSeq(p, n);
3415
3416	    if (q > p) {
3417		p = q;
3418		if (at(p, n, "::")) {
3419		    compressedZeros = true;
3420		    p = scanHexPost(p + 2, n);
3421		} else if (at(p, n, ':')) {
3422		    p = takeIPv4Address(p + 1,  n, "IPv4 address");
3423		    ipv6byteCount += 4;
3424		}
3425	    } else if (at(p, n, "::")) {
3426		compressedZeros = true;
3427		p = scanHexPost(p + 2, n);
3428	    }
3429	    if (p < n)
3430		fail("Malformed IPv6 address", start);
3431	    if (ipv6byteCount > 16)
3432		fail("IPv6 address too long", start);
3433	    if (!compressedZeros && ipv6byteCount < 16) 
3434		fail("IPv6 address too short", start);
3435	    if (compressedZeros && ipv6byteCount == 16)
3436		fail("Malformed IPv6 address", start);
3437
3438	    return p;
3439	}
3440
3441	private int scanHexPost(int start, int n)
3442	    throws URISyntaxException
3443	{
3444	    int p = start;
3445	    int q;
3446
3447	    if (p == n)
3448		return p;
3449
3450	    q = scanHexSeq(p, n);
3451	    if (q > p) {
3452		p = q;
3453		if (at(p, n, ':')) {
3454		    p++;
3455		    p = takeIPv4Address(p, n, "hex digits or IPv4 address");
3456		    ipv6byteCount += 4;
3457		}
3458	    } else {
3459		p = takeIPv4Address(p, n, "hex digits or IPv4 address");
3460		ipv6byteCount += 4;
3461	    }
3462	    return p;
3463	}
3464
3465	// Scan a hex sequence; return -1 if one could not be scanned
3466	//
3467	private int scanHexSeq(int start, int n)
3468	    throws URISyntaxException
3469	{
3470	    int p = start;
3471	    int q;
3472
3473	    q = scan(p, n, L_HEX, H_HEX);
3474	    if (q <= p)
3475		return -1;
3476	    if (at(q, n, '.'))		// Beginning of IPv4 address
3477		return -1;
3478	    if (q > p + 4)
3479                fail("IPv6 hexadecimal digit sequence too long", p);
3480	    ipv6byteCount += 2;
3481	    p = q;
3482	    while (p < n) {
3483		if (!at(p, n, ':'))
3484		    break;
3485		if (at(p + 1, n, ':'))
3486		    break;		// "::"
3487		p++;
3488		q = scan(p, n, L_HEX, H_HEX);
3489		if (q <= p)
3490		    failExpecting("digits for an IPv6 address", p);
3491		if (at(q, n, '.')) {	// Beginning of IPv4 address
3492		    p--;
3493		    break;
3494		}
3495		if (q > p + 4)
3496		    fail("IPv6 hexadecimal digit sequence too long", p);
3497		ipv6byteCount += 2;
3498		p = q;
3499	    }
3500
3501	    return p;
3502	}
3503
3504    }
3505
3506}