Add fast-mode Uri class.

sdk/lib/core/uri.dart

 // Copyright (c) 2012, the Dart project authors.  Please see the AUTHORS file
 // for details. All rights reserved. Use of this source code is governed by a
 // BSD-style license that can be found in the LICENSE file.
 
 part of dart.core;
 
+// Frequently used character codes.
+const int _SPACE = 0x20;
+const int _PERCENT = 0x25;
+const int _PLUS = 0x2B;
+const int _DOT = 0x2E;
+const int _SLASH = 0x2F;
+const int _COLON = 0x3A;
+const int _UPPER_CASE_A = 0x41;
+const int _UPPER_CASE_Z = 0x5A;
+const int _LEFT_BRACKET = 0x5B;
+const int _BACKSLASH = 0x5C;
+const int _RIGHT_BRACKET = 0x5D;
+const int _LOWER_CASE_A = 0x61;
+const int _LOWER_CASE_F = 0x66;
+const int _LOWER_CASE_Z = 0x7A;
+
+const String _hexDigits = "0123456789ABCDEF";
+
 /**
  * A parsed URI, such as a URL.
  *
  * **See also:**
  *
  * * [URIs][uris] in the [library tour][libtour]
  * * [RFC-3986](http://tools.ietf.org/html/rfc3986)
  *
  * [uris]: https://www.dartlang.org/docs/dart-up-and-running/ch03.html#uris
  * [libtour]: https://www.dartlang.org/docs/dart-up-and-running/contents/ch03.html
  */
-class Uri {
+abstract class Uri {
   /**
-   * The scheme component of the URI.
+   * Returns the natural base URI for the current platform.
    *
-   * Returns the empty string if there is no scheme component.
+   * When running in a browser this is the current URL of the current page
+   * (from `window.location.href`).
    *
-   * A URI scheme is case insensitive.
-   * The returned scheme is canonicalized to lowercase letters.
+   * When not running in a browser this is the file URI referencing
+   * the current working directory.
    */
-  // We represent the missing scheme as an empty string.
-  // A valid scheme cannot be empty.
-  final String scheme;
-
-  /**
-   * The user-info part of the authority.
-   *
-   * Does not distinguish between an empty user-info and an absent one.
-   * The value is always non-null.
-   * Is considered absent if [_host] is `null`.
-   */
-  final String _userInfo;
-
-  /**
-   * The host name of the URI.
-   *
-   * Set to `null` if there is no authority in the URI.
-   * The host name is the only mandatory part of an authority, so we use
-   * it to mark whether an authority part was present or not.
-   */
-  final String _host;
-
-  /**
-   * The port number part of the authority.
-   *
-   * The port. Set to null if there is no port. Normalized to null if
-   * the port is the default port for the scheme.
-   */
-  int _port;
-
-  /**
-   * The path of the URI.
-   *
-   * Always non-null.
-   */
-  String _path;
-
-  // The query content, or null if there is no query.
-  final String _query;
-
-  // The fragment content, or null if there is no fragment.
-  final String _fragment;
-
-  /**
-   * Cache the computed return value of [pathSegements].
-   */
-  List<String> _pathSegments;
-
-  /**
-   * Cache the computed return value of [queryParameters].
-   */
-  Map<String, String> _queryParameters;
-  Map<String, List<String>> _queryParameterLists;
-
-  /// Internal non-verifying constructor. Only call with validated arguments.
-  Uri._internal(this.scheme,
-                this._userInfo,
-                this._host,
-                this._port,
-                this._path,
-                this._query,
-                this._fragment);
+  external static Uri get base;
 
   /**
    * Creates a new URI from its components.
    *
    * Each component is set through a named argument. Any number of
    * components can be provided. The [path] and [query] components can be set
    * using either of two different named arguments.
    *
    * The scheme component is set through [scheme]. The scheme is
    * normalized to all lowercase letters. If the scheme is omitted or empty,
@@ skipping 52 matching lines @@
    * use an empty map for `queryParameters`.
    *
    * If both `query` and `queryParameters` are omitted or `null`,
    * the URI has no query part.
    *
    * The fragment component is set through [fragment].
    * It should be a valid URI fragment, but invalid characters other than
    * general delimiters, are escaped if necessary.
    * If `fragment` is omitted or `null`, the URI has no fragment part.
    */
-  factory Uri({String scheme : "",
-               String userInfo : "",
+  factory Uri({String scheme,
+               String userInfo,
                String host,
                int port,
                String path,
                Iterable<String> pathSegments,
                String query,
                Map<String, dynamic/*String|Iterable<String>*/> queryParameters,
-               String fragment}) {
-    scheme = _makeScheme(scheme, 0, _stringOrNullLength(scheme));
-    userInfo = _makeUserInfo(userInfo, 0, _stringOrNullLength(userInfo));
-    host = _makeHost(host, 0, _stringOrNullLength(host), false);
-    // Special case this constructor for backwards compatibility.
-    if (query == "") query = null;
-    query = _makeQuery(query, 0, _stringOrNullLength(query), queryParameters);
-    fragment = _makeFragment(fragment, 0, _stringOrNullLength(fragment));
-    port = _makePort(port, scheme);
-    bool isFile = (scheme == "file");
-    if (host == null &&
-        (userInfo.isNotEmpty || port != null || isFile)) {
-      host = "";
-    }
-    bool hasAuthority = (host != null);
-    path = _makePath(path, 0, _stringOrNullLength(path), pathSegments,
-                     scheme, hasAuthority);
-    if (scheme.isEmpty && host == null && !path.startsWith('/')) {
-      path = _normalizeRelativePath(path);
-    } else {
-      path = _removeDotSegments(path);
-    }
-    return new Uri._internal(scheme, userInfo, host, port,
-                             path, query, fragment);
-  }
+               String fragment}) = _Uri;
 
   /**
    * Creates a new `http` URI from authority, path and query.
    *
    * Examples:
    *
    * ```
    * // http://example.org/path?q=dart.
    * new Uri.http("google.com", "/search", { "q" : "dart" });
    *
@@ skipping 16 matching lines @@
    *
    * The `path` component is set from the [unencodedPath]
    * argument. The path passed must not be encoded as this constructor
    * encodes the path.
    *
    * The `query` component is set from the optional [queryParameters]
    * argument.
    */
   factory Uri.http(String authority,
                    String unencodedPath,
-                   [Map<String, String> queryParameters]) {
-    return _makeHttpUri("http", authority, unencodedPath, queryParameters);
-  }
+                   [Map<String, String> queryParameters]) = _Uri.http;
 
   /**
    * Creates a new `https` URI from authority, path and query.
    *
    * This constructor is the same as [Uri.http] except for the scheme
    * which is set to `https`.
    */
   factory Uri.https(String authority,
                     String unencodedPath,
-                    [Map<String, String> queryParameters]) {
-    return _makeHttpUri("https", authority, unencodedPath, queryParameters);
+                    [Map<String, String> queryParameters]) = _Uri.https;
+
+  /**
+   * Creates a new file URI from an absolute or relative file path.
+   *
+   * The file path is passed in [path].
+   *
+   * This path is interpreted using either Windows or non-Windows
+   * semantics.
+   *
+   * With non-Windows semantics the slash ("/") is used to separate
+   * path segments.
+   *
+   * With Windows semantics, backslash ("\") and forward-slash ("/")
+   * are used to separate path segments, except if the path starts
+   * with "\\?\" in which case, only backslash ("\") separates path
+   * segments.
+   *
+   * If the path starts with a path separator an absolute URI is
+   * created. Otherwise a relative URI is created. One exception from
+   * this rule is that when Windows semantics is used and the path
+   * starts with a drive letter followed by a colon (":") and a
+   * path separator then an absolute URI is created.
+   *
+   * The default for whether to use Windows or non-Windows semantics
+   * determined from the platform Dart is running on. When running in
+   * the standalone VM this is detected by the VM based on the
+   * operating system. When running in a browser non-Windows semantics
+   * is always used.
+   *
+   * To override the automatic detection of which semantics to use pass
+   * a value for [windows]. Passing `true` will use Windows
+   * semantics and passing `false` will use non-Windows semantics.
+   *
+   * Examples using non-Windows semantics:
+   *
+   * ```
+   * // xxx/yyy
+   * new Uri.file("xxx/yyy", windows: false);
+   *
+   * // xxx/yyy/
+   * new Uri.file("xxx/yyy/", windows: false);
+   *
+   * // file:///xxx/yyy
+   * new Uri.file("/xxx/yyy", windows: false);
+   *
+   * // file:///xxx/yyy/
+   * new Uri.file("/xxx/yyy/", windows: false);
+   *
+   * // C:
+   * new Uri.file("C:", windows: false);
+   * ```
+   *
+   * Examples using Windows semantics:
+   *
+   * ```
+   * // xxx/yyy
+   * new Uri.file(r"xxx\yyy", windows: true);
+   *
+   * // xxx/yyy/
+   * new Uri.file(r"xxx\yyy\", windows: true);
+   *
+   * file:///xxx/yyy
+   * new Uri.file(r"\xxx\yyy", windows: true);
+   *
+   * file:///xxx/yyy/
+   * new Uri.file(r"\xxx\yyy/", windows: true);
+   *
+   * // file:///C:/xxx/yyy
+   * new Uri.file(r"C:\xxx\yyy", windows: true);
+   *
+   * // This throws an error. A path with a drive letter is not absolute.
+   * new Uri.file(r"C:", windows: true);
+   *
+   * // This throws an error. A path with a drive letter is not absolute.
+   * new Uri.file(r"C:xxx\yyy", windows: true);
+   *
+   * // file://server/share/file
+   * new Uri.file(r"\\server\share\file", windows: true);
+   * ```
+   *
+   * If the path passed is not a legal file path [ArgumentError] is thrown.
+   */
+  factory Uri.file(String path, {bool windows}) = _Uri.file;
+
+  /**
+   * Like [Uri.file] except that a non-empty URI path ends in a slash.
+   *
+   * If [path] is not empty, and it doesn't end in a directory separator,
+   * then a slash is added to the returned URI's path.
+   * In all other cases, the result is the same as returned by `Uri.file`.
+   */
+  factory Uri.directory(String path, {bool windows}) = _Uri.directory;
+
+  /**
+   * Creates a `data:` URI containing the [content] string.
+   *
+   * Converts the content to a bytes using [encoding] or the charset specified
+   * in [parameters] (defaulting to US-ASCII if not specified or unrecognized),
+   * then encodes the bytes into the resulting data URI.
+   *
+   * Defaults to encoding using percent-encoding (any non-ASCII or non-URI-valid
+   * bytes is replaced by a percent encoding). If [base64] is true, the bytes
+   * are instead encoded using [BASE64].
+   *
+   * If [encoding] is not provided and [parameters] has a `charset` entry,
+   * that name is looked up using [Encoding.getByName],
+   * and if the lookup returns an encoding, that encoding is used to convert
+   * [content] to bytes.
+   * If providing both an [encoding] and a charset [parameter], they should
+   * agree, otherwise decoding won't be able to use the charset parameter
+   * to determine the encoding.
+   *
+   * If [mimeType] and/or [parameters] are supplied, they are added to the
+   * created URI. If any of these contain characters that are not allowed
+   * in the data URI, the character is percent-escaped. If the character is
+   * non-ASCII, it is first UTF-8 encoded and then the bytes are percent
+   * encoded. An omitted [mimeType] in a data URI means `text/plain`, just
+   * as an omitted `charset` parameter defaults to meaning `US-ASCII`.
+   *
+   * To read the content back, use [UriData.contentAsString].
+   */
+  factory Uri.dataFromString(String content,
+                             {String mimeType,
+                              Encoding encoding,
+                              Map<String, String> parameters,
+                              bool base64: false}) {
+    UriData data =  new UriData.fromString(content,
+                                           mimeType: mimeType,
+                                           encoding: encoding,
+                                           parameters: parameters,
+                                           base64: base64);
+    return data.uri;
   }
 
   /**
+   * Creates a `data:` URI containing an encoding of [bytes].
+   *
+   * Defaults to Base64 encoding the bytes, but if [percentEncoded]
+   * is `true`, the bytes will instead be percent encoded (any non-ASCII
+   * or non-valid-ASCII-character byte is replaced by a percent encoding).
+   *
+   * To read the bytes back, use [UriData.contentAsBytes].
+   *
+   * It defaults to having the mime-type `application/octet-stream`.
+   * The [mimeType] and [parameters] are added to the created URI.
+   * If any of these contain characters that are not allowed
+   * in the data URI, the character is percent-escaped. If the character is
+   * non-ASCII, it is first UTF-8 encoded and then the bytes are percent
+   * encoded.
+   */
+  factory Uri.dataFromBytes(List<int> bytes,
+                            {mimeType: "application/octet-stream",
+                             Map<String, String> parameters,
+                             percentEncoded: false}) {
+    UriData data = new UriData.fromBytes(bytes,
+                                         mimeType: mimeType,
+                                         parameters: parameters,
+                                         percentEncoded: percentEncoded);
+    return data.uri;
+  }
+
+  /**
+   * The scheme component of the URI.
+   *
+   * Returns the empty string if there is no scheme component.
+   *
+   * A URI scheme is case insensitive.
+   * The returned scheme is canonicalized to lowercase letters.
+   */
+  String get scheme;
+
+  /**
    * Returns the authority component.
    *
    * The authority is formatted from the [userInfo], [host] and [port]
    * parts.
    *
    * Returns the empty string if there is no authority component.
    */
-  String get authority {
-    if (!hasAuthority) return "";
-    var sb = new StringBuffer();
-    _writeAuthority(sb);
-    return sb.toString();
-  }
+  String get authority;
 
   /**
    * Returns the user info part of the authority component.
    *
    * Returns the empty string if there is no user info in the
    * authority component.
    */
-  String get userInfo => _userInfo;
+  String get userInfo;
 
   /**
    * Returns the host part of the authority component.
    *
    * Returns the empty string if there is no authority component and
    * hence no host.
    *
    * If the host is an IP version 6 address, the surrounding `[` and `]` is
    * removed.
    *
    * The host string is case-insensitive.
    * The returned host name is canonicalized to lower-case
    * with upper-case percent-escapes.
    */
-  String get host {
-    if (_host == null) return "";
-    if (_host.startsWith('[')) {
-      return _host.substring(1, _host.length - 1);
-    }
-    return _host;
-  }
+  String get host;
 
   /**
    * Returns the port part of the authority component.
    *
    * Returns the defualt port if there is no port number in the authority
    * component. That's 80 for http, 443 for https, and 0 for everything else.
    */
-  int get port {
-    if (_port == null) return _defaultPort(scheme);
-    return _port;
-  }
-
-  // The default port for the scheme of this Uri..
-  static int _defaultPort(String scheme) {
-    if (scheme == "http") return 80;
-    if (scheme == "https") return 443;
-    return 0;
-  }
+  int get port;
 
   /**
    * Returns the path component.
    *
    * The returned path is encoded. To get direct access to the decoded
    * path use [pathSegments].
    *
    * Returns the empty string if there is no path component.
    */
-  String get path => _path;
+  String get path;
 
   /**
    * Returns the query component. The returned query is encoded. To get
    * direct access to the decoded query use [queryParameters].
    *
    * Returns the empty string if there is no query component.
    */
-  String get query => (_query == null) ? "" : _query;
+  String get query;
 
   /**
    * Returns the fragment identifier component.
    *
    * Returns the empty string if there is no fragment identifier
    * component.
    */
-  String get fragment => (_fragment == null) ? "" : _fragment;
+  String get fragment;
+
+  /**
+   * Returns the URI path split into its segments. Each of the segments in the
+   * returned list have been decoded. If the path is empty the empty list will
+   * be returned. A leading slash `/` does not affect the segments returned.
+   *
+   * The returned list is unmodifiable and will throw [UnsupportedError] on any
+   * calls that would mutate it.
+   */
+  List<String> get pathSegments;
+
+  /**
+   * Returns the URI query split into a map according to the rules
+   * specified for FORM post in the [HTML 4.01 specification section
+   * 17.13.4](http://www.w3.org/TR/REC-html40/interact/forms.html#h-17.13.4 "HTML 4.01 section 17.13.4").
+   * Each key and value in the returned map has been decoded.
+   * If there is no query the empty map is returned.
+   *
+   * Keys in the query string that have no value are mapped to the
+   * empty string.
+   * If a key occurs more than once in the query string, it is mapped to
+   * an arbitrary choice of possible value.
+   * The [queryParametersAll] getter can provide a map
+   * that maps keys to all of their values.
+   *
+   * The returned map is unmodifiable.
+   */
+  Map<String, String> get queryParameters;
+
+  /**
+   * Returns the URI query split into a map according to the rules
+   * specified for FORM post in the [HTML 4.01 specification section
+   * 17.13.4](http://www.w3.org/TR/REC-html40/interact/forms.html#h-17.13.4 "HTML 4.01 section 17.13.4").
+   * Each key and value in the returned map has been decoded. If there is no
+   * query the empty map is returned.
+   *
+   * Keys are mapped to lists of their values. If a key occurs only once,
+   * its value is a singleton list. If a key occurs with no value, the
+   * empty string is used as the value for that occurrence.
+   *
+   * The returned map and the lists it contains are unmodifiable.
+   */
+  Map<String, List<String>> get queryParametersAll;
+
+  /**
+   * Returns whether the URI is absolute.
+   *
+   * A URI is an absolute URI in the sense of RFC 3986 if it has a scheme
+   * and no fragment.
+   */
+  bool get isAbsolute;
+
+  /**
+   * Returns whether the URI has a [scheme] component.
+   */
+  bool get hasScheme => scheme.isNotEmpty;
+
+  /**
+   * Returns whether the URI has an [authority] component.
+   */
+  bool get hasAuthority;
+
+  /**
+   * Returns whether the URI has an explicit port.
+   *
+   * If the port number is the default port number
+   * (zero for unrecognized schemes, with http (80) and https (443) being
+   * recognized),
+   * then the port is made implicit and omitted from the URI.
+   */
+  bool get hasPort;
+
+  /**
+   * Returns whether the URI has a query part.
+   */
+  bool get hasQuery;
+
+  /**
+   * Returns whether the URI has a fragment part.
+   */
+  bool get hasFragment;
+
+  /**
+   * Returns whether the URI has an empty path.
+   */
+  bool get hasEmptyPath;
+
+  /**
+   * Returns whether the URI has an absolute path (starting with '/').
+   */
+  bool get hasAbsolutePath;
+
+  /**
+   * Returns the origin of the URI in the form scheme://host:port for the
+   * schemes http and https.
+   *
+   * It is an error if the scheme is not "http" or "https".
+   *
+   * See: http://www.w3.org/TR/2011/WD-html5-20110405/origin-0.html#origin
+   */
+  String get origin;
+
+  /**
+   * Returns the file path from a file URI.
+   *
+   * The returned path has either Windows or non-Windows
+   * semantics.
+   *
+   * For non-Windows semantics the slash ("/") is used to separate
+   * path segments.
+   *
+   * For Windows semantics the backslash ("\") separator is used to
+   * separate path segments.
+   *
+   * If the URI is absolute the path starts with a path separator
+   * unless Windows semantics is used and the first path segment is a
+   * drive letter. When Windows semantics is used a host component in
+   * the uri in interpreted as a file server and a UNC path is
+   * returned.
+   *
+   * The default for whether to use Windows or non-Windows semantics
+   * determined from the platform Dart is running on. When running in
+   * the standalone VM this is detected by the VM based on the
+   * operating system. When running in a browser non-Windows semantics
+   * is always used.
+   *
+   * To override the automatic detection of which semantics to use pass
+   * a value for [windows]. Passing `true` will use Windows
+   * semantics and passing `false` will use non-Windows semantics.
+   *
+   * If the URI ends with a slash (i.e. the last path component is
+   * empty) the returned file path will also end with a slash.
+   *
+   * With Windows semantics URIs starting with a drive letter cannot
+   * be relative to the current drive on the designated drive. That is
+   * for the URI `file:///c:abc` calling `toFilePath` will throw as a
+   * path segment cannot contain colon on Windows.
+   *
+   * Examples using non-Windows semantics (resulting of calling
+   * toFilePath in comment):
+   *
+   *     Uri.parse("xxx/yyy");  // xxx/yyy
+   *     Uri.parse("xxx/yyy/");  // xxx/yyy/
+   *     Uri.parse("file:///xxx/yyy");  // /xxx/yyy
+   *     Uri.parse("file:///xxx/yyy/");  // /xxx/yyy/
+   *     Uri.parse("file:///C:");  // /C:
+   *     Uri.parse("file:///C:a");  // /C:a
+   *
+   * Examples using Windows semantics (resulting URI in comment):
+   *
+   *     Uri.parse("xxx/yyy");  // xxx\yyy
+   *     Uri.parse("xxx/yyy/");  // xxx\yyy\
+   *     Uri.parse("file:///xxx/yyy");  // \xxx\yyy
+   *     Uri.parse("file:///xxx/yyy/");  // \xxx\yyy/
+   *     Uri.parse("file:///C:/xxx/yyy");  // C:\xxx\yyy
+   *     Uri.parse("file:C:xxx/yyy");  // Throws as a path segment
+   *                                   // cannot contain colon on Windows.
+   *     Uri.parse("file://server/share/file");  // \\server\share\file
+   *
+   * If the URI is not a file URI calling this throws
+   * [UnsupportedError].
+   *
+   * If the URI cannot be converted to a file path calling this throws
+   * [UnsupportedError].
+   */
+  // TODO(lrn): Deprecate and move functionality to File class or similar.
+  // The core libraries should not worry about the platform.
+  String toFilePath({bool windows});
+
+  /**
+   * Access the structure of a `data:` URI.
+   *
+   * Returns a [UriData] object for `data:` URIs and `null` for all other
+   * URIs.
+   * The [UriData] object can be used to access the media type and data
+   * of a `data:` URI.
+   */
+  UriData get data;
+
+  /// Returns a hash code computed as `toString().hashCode`.
+  ///
+  /// This guarantees that URIs with the same normalized
+  int get hashCode;
+
+  /// A URI is equal to another URI with the same normalized representation.
+  bool operator==(Object other);
+
+  /// Returns the normalized string representation of the URI.
+  String toString();
+
+  /**
+   * Returns a new `Uri` based on this one, but with some parts replaced.
+   *
+   * This method takes the same parameters as the [new Uri] constructor,
+   * and they have the same meaning.
+   *
+   * At most one of [path] and [pathSegments] must be provided.
+   * Likewise, at most one of [query] and [queryParameters] must be provided.
+   *
+   * Each part that is not provided will default to the corresponding
+   * value from this `Uri` instead.
+   *
+   * This method is different from [Uri.resolve] which overrides in a
+   * hierarchial manner,
+   * and can instead replace each part of a `Uri` individually.
+   *
+   * Example:
+   *
+   *     Uri uri1 = Uri.parse("a://b@c:4/d/e?f#g");
+   *     Uri uri2 = uri1.replace(scheme: "A", path: "D/E/E", fragment: "G");
+   *     print(uri2);  // prints "A://b@c:4/D/E/E/?f#G"
+   *
+   * This method acts similarly to using the `new Uri` constructor with
+   * some of the arguments taken from this `Uri` . Example:
+   *
+   *     Uri uri3 = new Uri(
+   *         scheme: "A",
+   *         userInfo: uri1.userInfo,
+   *         host: uri1.host,
+   *         port: uri1.port,
+   *         path: "D/E/E",
+   *         query: uri1.query,
+   *         fragment: "G");
+   *     print(uri3);  // prints "A://b@c:4/D/E/E/?f#G"
+   *     print(uri2 == uri3);  // prints true.
+   *
+   * Using this method can be seen as a shorthand for the `Uri` constructor
+   * call above, but may also be slightly faster because the parts taken
+   * from this `Uri` need not be checked for validity again.
+   */
+  Uri replace({String scheme,
+               String userInfo,
+               String host,
+               int port,
+               String path,
+               Iterable<String> pathSegments,
+               String query,
+               Map<String, dynamic/*String|Iterable<String>*/> queryParameters,
+               String fragment});
+
+  /**
+   * Returns a `Uri` that differs from this only in not having a fragment.
+   *
+   * If this `Uri` does not have a fragment, it is itself returned.
+   */
+  Uri removeFragment();
+
+  /**
+   * Resolve [reference] as an URI relative to `this`.
+   *
+   * First turn [reference] into a URI using [Uri.parse]. Then resolve the
+   * resulting URI relative to `this`.
+   *
+   * Returns the resolved URI.
+   *
+   * See [resolveUri] for details.
+   */
+  Uri resolve(String reference);
+
+  /**
+   * Resolve [reference] as an URI relative to `this`.
+   *
+   * Returns the resolved URI.
+   *
+   * The algorithm "Transform Reference" for resolving a reference is described
+   * in [RFC-3986 Section 5](http://tools.ietf.org/html/rfc3986#section-5 "RFC-1123").
+   *
+   * Updated to handle the case where the base URI is just a relative path -
+   * that is: when it has no scheme or authority and the path does not start
+   * with a slash.
+   * In that case, the paths are combined without removing leading "..", and
+   * an empty path is not converted to "/".
+   */
+  Uri resolveUri(Uri reference);
+
+  /**
+   * Returns a URI where the path has been normalized.
+   *
+   * A normalized path does not contain `.` segments or non-leading `..`
+   * segments.
+   * Only a relative path with no scheme or authority may contain
+   * leading `..` segments,
+   * a path that starts with `/` will also drop any leading `..` segments.
+   *
+   * This uses the same normalization strategy as `new Uri().resolve(this)`.
+   *
+   * Does not change any part of the URI except the path.
+   *
+   * The default implementation of `Uri` always normalizes paths, so calling
+   * this function has no effect.
+   */
+  Uri normalizePath();
 
   /**
    * Creates a new `Uri` object by parsing a URI string.
    *
    * If [start] and [end] are provided, only the substring from `start`
    * to `end` is parsed as a URI.
    *
    * If the string is not valid as a URI or URI reference,
    * a [FormatException] is thrown.
    */
@@ skipping 44 matching lines @@
     // segment       = *pchar
     // segment-nz    = 1*pchar
     // segment-nz-nc = 1*( unreserved / pct-encoded / sub-delims / "@" )
     //               ; non-zero-length segment without any colon ":"
     //
     // pchar         = unreserved / pct-encoded / sub-delims / ":" / "@"
     //
     // query         = *( pchar / "/" / "?" )
     //
     // fragment      = *( pchar / "/" / "?" )
-    const int EOI = -1;
-
-    String scheme = "";
-    String userinfo = "";
-    String host = null;
-    int port = null;
-    String path = null;
-    String query = null;
-    String fragment = null;
-    if (end == null) end = uri.length;
-
-    int index = start;
-    int pathStart = start;
-    // End of input-marker.
-    int char = EOI;
-
-    void parseAuth() {
-      if (index == end) {
-        char = EOI;
-        return;
-      }
-      int authStart = index;
-      int lastColon = -1;
-      int lastAt = -1;
-      char = uri.codeUnitAt(index);
-      while (index < end) {
-        char = uri.codeUnitAt(index);
-        if (char == _SLASH || char == _QUESTION || char == _NUMBER_SIGN) {
-          break;
-        }
-        if (char == _AT_SIGN) {
-          lastAt = index;
-          lastColon = -1;
-        } else if (char == _COLON) {
-          lastColon = index;
-        } else if (char == _LEFT_BRACKET) {
-          lastColon = -1;
-          int endBracket = uri.indexOf(']', index + 1);
-          if (endBracket == -1) {
-            index = end;
-            char = EOI;
-            break;
-          } else {
-            index = endBracket;
+    end ??= uri.length;
+
+    // Special case data:URIs. Ignore case when testing.
+    if (end >= start + 5) {
+      int dataDelta = _startsWithData(uri, start);
+      if (dataDelta == 0) {
+        // The case is right.
+        if (start > 0 || end < uri.length) uri = uri.substring(start, end);
+        return UriData._parse(uri, 5, null).uri;
+      } else if (dataDelta == 0x20) {
+        return UriData._parse(uri.substring(start + 5, end), 0, null).uri;
+      }
+      // Otherwise the URI doesn't start with "data:" or any case variant of it.
+    }
+
+    // TODO(lrn): Consider inlining _scanUri here.
+    // The following index-normalization belongs with the scanning, but is
+    // easier to do here because we already have extracted variables from the

[floitsch, 2016/06/29 23:41:47]

I think you should do the normalizations in _scanUri. It can't cost that much to
read it out there too, and to write it back.

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

It probably shouldn't but in practice I see a ~10% regression on my benchmark
(which is basically parsing all URIs seen in import statements in the Dart SDK
repository, weighted by count).

I went in the other direction and inlined _scanUri. It might still cost 2-3% but
it might also just be noise.

+    // indices list.
+    var indices = _scanUri(uri, start, end);
+
+    int schemeEnd     = indices[_schemeEndIndex];
+    int hostStart     = indices[_hostStartIndex] + 1;
+    int portStart     = indices[_portStartIndex];
+    int pathStart     = indices[_pathStartIndex];
+    int queryStart    = indices[_queryStartIndex];
+    int fragmentStart = indices[_fragmentStartIndex];
+
+    // We may discover scheme while handling special cases.
+    String scheme;
+
+    // Derive some indices that weren't set to normalize the indices.
+    // If fragment but no query, set query to start at fragment.
+    if (fragmentStart < queryStart) queryStart = fragmentStart;
+    // If scheme but no authority, the pathStart isn't set.
+    if (schemeEnd >= start && hostStart <= start) pathStart = schemeEnd + 1;
+    // If scheme or authority but pathStart isn't set.
+    if (pathStart == start && (schemeEnd >= start || hostStart > start)) {
+      pathStart = queryStart;
+    }
+    // If authority and no port.
+    // (including when user-info contains : and portStart >= 0).
+    if (portStart < hostStart) portStart = pathStart;
+
+    assert(hostStart == start || schemeEnd <= hostStart);
+    assert(hostStart <= portStart);
+    assert(schemeEnd <= pathStart);
+    assert(portStart <= pathStart);
+    assert(pathStart <= queryStart);
+    assert(queryStart <= fragmentStart);
+
+    bool isSimple = indices[_notSimpleIndex] < start;
+
+    if (isSimple) {
+      // Check/do normalizations that weren't detected by the scanner.
+      // This includes removal of empty port or userInfo,
+      // or scheme specific port and path normalizations.
+      if (hostStart > schemeEnd + 3) {
+        // Always be non-simple if URI contains user-info.
+        // The scanner doesn't set the not-simple position in this case because
+        // it's setting the host-start position instead.
+        isSimple = false;
+      } else if (portStart > start && portStart + 1 == pathStart) {
+        // If the port is empty, it should be omitted.
+        // Pathological case, don't bother correcting it.
+        isSimple = false;
+      } else if (hostStart == schemeEnd + 4) {
+        // If the userInfo is empty, it should be omitted.
+        // (4 is length of "://@").
+        // Pathological case, don't bother correcting it.
+        isSimple = false;
+      } else if (queryStart < end &&
+                 (queryStart == pathStart + 2 &&
+                  uri.startsWith("..", pathStart)) ||
+                 (queryStart > pathStart + 2 &&
+                  uri.startsWith("/..", queryStart - 3))) {
+        // The path ends in a ".." segment. This should be normalized to "../".
+        // We didn't detect this while scanning because a query or fragment was
+        // detected at the same time (which is why we only need to check this
+        // if there is something after the path).
+        isSimple = false;
+      } else {
+        // There are a few scheme-based normalizations that
+        // the scanner couldn't check.
+        // That means that the input is very close to simple, so just do
+        // the normalizations.
+        if (schemeEnd == start + 4) {
+          // Do scheme based normalizations for file, http.
+          if (uri.startsWith("file", start)) {
+            scheme = "file";
+            if (hostStart <= start) {
+              // File URIs should have an authority.
+              // Paths after an authority should be absolute.
+              String schemeAuth = "file://";
+              int delta = 2;
+              if (!uri.startsWith("/", pathStart)) {
+                schemeAuth = "file:///";
+                delta = 3;
+              }
+              uri = schemeAuth + uri.substring(pathStart, end);
+              schemeEnd -= start;
+              hostStart = 7;
+              portStart = 7;
+              pathStart = 7;
+              queryStart += delta - start;
+              fragmentStart += delta - start;
+              start = 0;
+              end = uri.length;
+            } else if (pathStart == queryStart) {
+              // Uri has authority and empty path. Add "/" as path.
+              uri = "${uri.substring(start, pathStart)}/"
+                    "${uri.substring(queryStart, end)}";
+              schemeEnd -= start;
+              hostStart -= start;
+              portStart -= start;
+              pathStart -= start;
+              queryStart += 1 - start;
+              fragmentStart += 1 - start;
+              start = 0;
+              end = uri.length;
+            }
+          } else if (uri.startsWith("http", start)) {
+            scheme = "http";
+            // HTTP URIs should not have an explicit port of 80.
+            if (portStart > start && portStart + 3 == pathStart &&
+                uri.startsWith("80", portStart + 1)) {
+              uri = uri.substring(start, portStart) +
+                    uri.substring(pathStart, end);
+              schemeEnd -= start;
+              hostStart -= start;
+              portStart -= start;
+              pathStart -= 3 + start;
+              queryStart -= 3 + start;
+              fragmentStart -= 3 + start;
+              start = 0;
+              end = uri.length;
+            }
+          }
+        } else if (schemeEnd == start + 5 && uri.startsWith("https", start)) {
+          scheme = "https";
+          // HTTPS URIs should not have an explicit port of 443.
+          if (portStart > start && portStart + 4 == pathStart &&
+              uri.startsWith("443", portStart + 1)) {
+            uri = uri.substring(start, portStart) +
+                  uri.substring(pathStart, end);
+            schemeEnd -= start;
+            hostStart -= start;
+            portStart -= start;
+            pathStart -= 4 + start;
+            queryStart -= 4 + start;
+            fragmentStart -= 4 + start;
+            start = 0;
+            end = uri.length;
           }
         }
-        index++;
-        char = EOI;
-      }
-      int hostStart = authStart;
-      int hostEnd = index;
-      if (lastAt >= 0) {
-        userinfo = _makeUserInfo(uri, authStart, lastAt);
-        hostStart = lastAt + 1;
-      }
-      if (lastColon >= 0) {
-        int portNumber;
-        if (lastColon + 1 < index) {
-          portNumber = 0;
-          for (int i = lastColon + 1; i < index; i++) {
-            int digit = uri.codeUnitAt(i);
-            if (_ZERO > digit || _NINE < digit) {
-              _fail(uri, i, "Invalid port number");
-            }
-            portNumber = portNumber * 10 + (digit - _ZERO);
+      }
+    }
+
+    if (isSimple) {
+      if (start > 0 || end < uri.length) {
+        uri = uri.substring(start, end);
+        if (schemeEnd >= 0) schemeEnd -= start;
+        if (hostStart > 0) {
+          hostStart -= start;
+          portStart -= start;
+        }
+        pathStart -= start;
+        queryStart -= start;
+        fragmentStart -= start;
+      }
+      return new _SimpleUri(uri, schemeEnd, hostStart, portStart, pathStart,
+                            queryStart, fragmentStart, scheme);
+
+    }
+
+    return new _Uri.notSimple(uri, start, end, schemeEnd, hostStart, portStart,
+                              pathStart, queryStart, fragmentStart, scheme);
+  }
+
+  /**
+   * Encode the string [component] using percent-encoding to make it
+   * safe for literal use as a URI component.
+   *
+   * All characters except uppercase and lowercase letters, digits and
+   * the characters `-_.!~*'()` are percent-encoded. This is the
+   * set of characters specified in RFC 2396 and the which is
+   * specified for the encodeUriComponent in ECMA-262 version 5.1.
+   *
+   * When manually encoding path segments or query components remember
+   * to encode each part separately before building the path or query
+   * string.
+   *
+   * For encoding the query part consider using
+   * [encodeQueryComponent].
+   *
+   * To avoid the need for explicitly encoding use the [pathSegments]
+   * and [queryParameters] optional named arguments when constructing
+   * a [Uri].
+   */
+  static String encodeComponent(String component) {
+    return _Uri._uriEncode(_Uri._unreserved2396Table, component, UTF8, false);
+  }
+
+  /**
+   * Encode the string [component] according to the HTML 4.01 rules
+   * for encoding the posting of a HTML form as a query string
+   * component.
+   *
+   * Encode the string [component] according to the HTML 4.01 rules
+   * for encoding the posting of a HTML form as a query string
+   * component.
+
+   * The component is first encoded to bytes using [encoding].
+   * The default is to use [UTF8] encoding, which preserves all
+   * the characters that don't need encoding.
+
+   * Then the resulting bytes are "percent-encoded". This transforms
+   * spaces (U+0020) to a plus sign ('+') and all bytes that are not
+   * the ASCII decimal digits, letters or one of '-._~' are written as
+   * a percent sign '%' followed by the two-digit hexadecimal
+   * representation of the byte.
+
+   * Note that the set of characters which are percent-encoded is a
+   * superset of what HTML 4.01 requires, since it refers to RFC 1738
+   * for reserved characters.
+   *
+   * When manually encoding query components remember to encode each
+   * part separately before building the query string.
+   *
+   * To avoid the need for explicitly encoding the query use the
+   * [queryParameters] optional named arguments when constructing a
+   * [Uri].
+   *
+   * See http://www.w3.org/TR/html401/interact/forms.html#h-17.13.4.2 for more
+   * details.
+   */
+  static String encodeQueryComponent(String component,
+                                     {Encoding encoding: UTF8}) {
+    return _Uri._uriEncode(_Uri._unreservedTable, component, encoding, true);
+  }
+
+  /**
+   * Decodes the percent-encoding in [encodedComponent].
+   *
+   * Note that decoding a URI component might change its meaning as
+   * some of the decoded characters could be characters with are
+   * delimiters for a given URI componene type. Always split a URI
+   * component using the delimiters for the component before decoding
+   * the individual parts.
+   *
+   * For handling the [path] and [query] components consider using
+   * [pathSegments] and [queryParameters] to get the separated and
+   * decoded component.
+   */
+  static String decodeComponent(String encodedComponent) {
+    return _Uri._uriDecode(encodedComponent, 0, encodedComponent.length,
+                           UTF8, false);
+  }
+
+  /**
+   * Decodes the percent-encoding in [encodedComponent], converting
+   * pluses to spaces.
+   *
+   * It will create a byte-list of the decoded characters, and then use
+   * [encoding] to decode the byte-list to a String. The default encoding is
+   * UTF-8.
+   */
+  static String decodeQueryComponent(
+      String encodedComponent,
+      {Encoding encoding: UTF8}) {
+    return _Uri._uriDecode(encodedComponent, 0, encodedComponent.length,
+                           encoding, true);
+  }
+
+  /**
+   * Encode the string [uri] using percent-encoding to make it
+   * safe for literal use as a full URI.
+   *
+   * All characters except uppercase and lowercase letters, digits and
+   * the characters `!#$&'()*+,-./:;=?@_~` are percent-encoded. This
+   * is the set of characters specified in in ECMA-262 version 5.1 for
+   * the encodeURI function .
+   */
+  static String encodeFull(String uri) {
+    return _Uri._uriEncode(_Uri._encodeFullTable, uri, UTF8, false);
+  }
+
+  /**
+   * Decodes the percent-encoding in [uri].
+   *
+   * Note that decoding a full URI might change its meaning as some of
+   * the decoded characters could be reserved characters. In most
+   * cases an encoded URI should be parsed into components using
+   * [Uri.parse] before decoding the separate components.
+   */
+  static String decodeFull(String uri) {
+    return _Uri._uriDecode(uri, 0, uri.length, UTF8, false);
+  }
+
+  /**
+   * Returns the [query] split into a map according to the rules
+   * specified for FORM post in the [HTML 4.01 specification section
+   * 17.13.4](http://www.w3.org/TR/REC-html40/interact/forms.html#h-17.13.4 "HTML 4.01 section 17.13.4").
+   * Each key and value in the returned map has been decoded. If the [query]
+   * is the empty string an empty map is returned.
+   *
+   * Keys in the query string that have no value are mapped to the
+   * empty string.
+   *
+   * Each query component will be decoded using [encoding]. The default encoding
+   * is UTF-8.
+   */
+  static Map<String, String> splitQueryString(String query,
+                                              {Encoding encoding: UTF8}) {
+    return query.split("&").fold({}, (map, element) {
+      int index = element.indexOf("=");
+      if (index == -1) {
+        if (element != "") {
+          map[decodeQueryComponent(element, encoding: encoding)] = "";
+        }
+      } else if (index != 0) {
+        var key = element.substring(0, index);
+        var value = element.substring(index + 1);
+        map[decodeQueryComponent(key, encoding: encoding)] =
+            decodeQueryComponent(value, encoding: encoding);
+      }
+      return map;
+    });
+  }
+
+
+  /**
+   * Parse the [host] as an IP version 4 (IPv4) address, returning the address
+   * as a list of 4 bytes in network byte order (big endian).
+   *
+   * Throws a [FormatException] if [host] is not a valid IPv4 address
+   * representation.
+   */
+  static List<int> parseIPv4Address(String host) {
+    void error(String msg) {
+      throw new FormatException('Illegal IPv4 address, $msg', host);
+    }
+    var bytes = host.split('.');
+    if (bytes.length != 4) {
+      error('IPv4 address should contain exactly 4 parts');
+    }
+    var result = new Uint8List(4);
+    for (int i = 0; i < 4; i++) {
+      var byteString = bytes[i];
+      int byte = int.parse(byteString);
+      if (byte < 0 || byte > 255) {
+        error('each part must be in the range of `0..255`');
+      }
+      result[i] = byte;
+    }
+    return result;
+  }
+
+  /**
+   * Parse the [host] as an IP version 6 (IPv6) address, returning the address
+   * as a list of 16 bytes in network byte order (big endian).
+   *
+   * Throws a [FormatException] if [host] is not a valid IPv6 address
+   * representation.
+   *
+   * Acts on the substring from [start] to [end]. If [end] is omitted, it
+   * defaults ot the end of the string.
+   *
+   * Some examples of IPv6 addresses:
+   *  * ::1
+   *  * FEDC:BA98:7654:3210:FEDC:BA98:7654:3210
+   *  * 3ffe:2a00:100:7031::1
+   *  * ::FFFF:129.144.52.38
+   *  * 2010:836B:4179::836B:4179
+   */
+  static List<int> parseIPv6Address(String host, [int start = 0, int end]) {
+    if (end == null) end = host.length;
+    // An IPv6 address consists of exactly 8 parts of 1-4 hex digits, seperated
+    // by `:`'s, with the following exceptions:
+    //
+    //  - One (and only one) wildcard (`::`) may be present, representing a fill
+    //    of 0's. The IPv6 `::` is thus 16 bytes of `0`.
+    //  - The last two parts may be replaced by an IPv4 address.
+    void error(String msg, [position]) {
+      throw new FormatException('Illegal IPv6 address, $msg', host, position);
+    }
+    int parseHex(int start, int end) {

[floitsch, 2016/06/29 23:41:47]

newline before and after nested functions.

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Done.

+      if (end - start > 4) {
+        error('an IPv6 part can only contain a maximum of 4 hex digits', start);
+      }
+      int value = int.parse(host.substring(start, end), radix: 16);
+      if (value < 0 || value > (1 << 16) - 1) {
+        error('each part must be in the range of `0x0..0xFFFF`', start);
+      }
+      return value;
+    }
+    if (host.length < 2) error('address is too short');
+    List<int> parts = [];
+    bool wildcardSeen = false;
+    int partStart = start;
+    // Parse all parts, except a potential last one.
+    for (int i = start; i < end; i++) {
+      if (host.codeUnitAt(i) == _COLON) {
+        if (i == start) {
+          // If we see a `:` in the beginning, expect wildcard.
+          i++;
+          if (host.codeUnitAt(i) != _COLON) {
+            error('invalid start colon.', i);
           }
+          partStart = i;
         }
-        port = _makePort(portNumber, scheme);
-        hostEnd = lastColon;
-      }
-      host = _makeHost(uri, hostStart, hostEnd, true);
-      if (index < end) {
-        char = uri.codeUnitAt(index);
-      }
-    }
-
-    // When reaching path parsing, the current character is known to not
-    // be part of the path.
-    const int NOT_IN_PATH = 0;
-    // When reaching path parsing, the current character is part
-    // of the a non-empty path.
-    const int IN_PATH = 1;
-    // When reaching authority parsing, authority is possible.
-    // This is only true at start or right after scheme.
-    const int ALLOW_AUTH = 2;
-
-    // Current state.
-    // Initialized to the default value that is used when exiting the
-    // scheme loop by reaching the end of input.
-    // All other breaks set their own state.
-    int state = NOT_IN_PATH;
-    int i = index;  // Temporary alias for index to avoid bug 19550 in dart2js.
-    while (i < end) {
-      char = uri.codeUnitAt(i);
-      if (char == _QUESTION || char == _NUMBER_SIGN) {
-        state = NOT_IN_PATH;
-        break;
-      }
-      if (char == _SLASH) {
-        state = (i == start) ? ALLOW_AUTH : IN_PATH;
-        break;
-      }
-      if (char == _COLON) {
-        if (i == start) _fail(uri, start, "Invalid empty scheme");
-        scheme = _makeScheme(uri, start, i);
-        i++;
-        if (scheme == "data") {
-          // This generates a URI that is (potentially) not path normalized.
-          // Applying part normalization to a non-hierarchial URI isn't
-          // meaningful.
-          return UriData._parse(uri, i, null).uri;
+        if (i == partStart) {
+          // Wildcard. We only allow one.
+          if (wildcardSeen) {
+            error('only one wildcard `::` is allowed', i);
+          }
+          wildcardSeen = true;
+          parts.add(-1);
+        } else {
+          // Found a single colon. Parse [partStart..i] as a hex entry.
+          parts.add(parseHex(partStart, i));
         }
-        pathStart = i;
-        if (i == end) {
-          char = EOI;
-          state = NOT_IN_PATH;
-        } else {
-          char = uri.codeUnitAt(i);
-          if (char == _QUESTION || char == _NUMBER_SIGN) {
-            state = NOT_IN_PATH;
-          } else if (char == _SLASH) {
-            state = ALLOW_AUTH;
-          } else {
-            state = IN_PATH;
-          }
+        partStart = i + 1;
+      }
+    }
+    if (parts.length == 0) error('too few parts');
+    bool atEnd = (partStart == end);
+    bool isLastWildcard = (parts.last == -1);
+    if (atEnd && !isLastWildcard) {
+      error('expected a part after last `:`', end);
+    }
+    if (!atEnd) {
+      try {

[floitsch, 2016/06/29 23:41:47]

I would assume that the try/catch costs performance.
The first one shouldn't even be an exceptions since having an ip4 address is
valid.

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Rewritten to not use try/catch.
This means that the error message for an invalid IPv4 address inside an IPv6
address changes to say "invalid IPv4 address", but I guess that's acceptable.
It only happens if the IPv6 address part contains a ".", so it is likely to be
intended as an IPv4 address.

+        parts.add(parseHex(partStart, end));
+      } catch (e) {
+        // Failed to parse the last chunk as hex. Try IPv4.
+        try {
+          List<int> last = parseIPv4Address(host.substring(partStart, end));
+          parts.add(last[0] << 8 | last[1]);
+          parts.add(last[2] << 8 | last[3]);
+        } catch (e) {
+          error('invalid end of IPv6 address.', partStart);
         }
-        break;
-      }
-      i++;
-      char = EOI;
-    }
-    index = i;  // Remove alias when bug is fixed.
-
-    if (state == ALLOW_AUTH) {
-      assert(char == _SLASH);
-      // Have seen one slash either at start or right after scheme.
-      // If two slashes, it's an authority, otherwise it's just the path.
-      index++;
-      if (index == end) {
-        char = EOI;
-        state = NOT_IN_PATH;
+      }
+    }
+    if (wildcardSeen) {
+      if (parts.length > 7) {
+        error('an address with a wildcard must have less than 7 parts');
+      }
+    } else if (parts.length != 8) {
+      error('an address without a wildcard must contain exactly 8 parts');
+    }
+    List<int> bytes = new Uint8List(16);
+    for (int i = 0, index = 0; i < parts.length; i++) {
+      int value = parts[i];
+      if (value == -1) {
+        int wildCardLength = 9 - parts.length;
+        for (int j = 0; j < wildCardLength; j++) {
+          bytes[index] = 0;
+          bytes[index + 1] = 0;
+          index += 2;
+        }
       } else {
-        char = uri.codeUnitAt(index);
-        if (char == _SLASH) {
-          index++;
-          parseAuth();
-          pathStart = index;
-        }
-        if (char == _QUESTION || char == _NUMBER_SIGN || char == EOI) {
-          state = NOT_IN_PATH;
-        } else {
-          state = IN_PATH;
-        }
-      }
-    }
-
-    assert(state == IN_PATH || state == NOT_IN_PATH);
-    if (state == IN_PATH) {
-      // Characters from pathStart to index (inclusive) are known
-      // to be part of the path.
-      while (++index < end) {
-        char = uri.codeUnitAt(index);
-        if (char == _QUESTION || char == _NUMBER_SIGN) {
-          break;
-        }
-        char = EOI;
-      }
-      state = NOT_IN_PATH;
-    }
-
-    assert(state == NOT_IN_PATH);
+        bytes[index] = value >> 8;
+        bytes[index + 1] = value & 0xff;
+        index += 2;
+      }
+    }
+    return bytes;
+  }
+}
+
+class _Uri implements Uri {
+  // We represent the missing scheme as an empty string.
+  // A valid scheme cannot be empty.
+  final String scheme;
+
+  /**
+   * The user-info part of the authority.
+   *
+   * Does not distinguish between an empty user-info and an absent one.
+   * The value is always non-null.
+   * Is considered absent if [_host] is `null`.
+   */
+  final String _userInfo;
+
+  /**
+   * The host name of the URI.
+   *
+   * Set to `null` if there is no authority in the URI.
+   * The host name is the only mandatory part of an authority, so we use
+   * it to mark whether an authority part was present or not.
+   */
+  final String _host;
+
+  /**
+   * The port number part of the authority.
+   *
+   * The port. Set to null if there is no port. Normalized to null if
+   * the port is the default port for the scheme.
+   */
+  int _port;
+
+  /**
+   * The path of the URI.
+   *
+   * Always non-null.
+   */
+  String _path;
+
+  // The query content, or null if there is no query.
+  final String _query;
+
+  // The fragment content, or null if there is no fragment.
+  final String _fragment;
+
+  /**
+   * Cache the computed return value of [pathSegements].
+   */
+  List<String> _pathSegments;
+
+  /**
+   * Cache of the full normalized text representation of the URI.
+   */
+  String _text;
+
+  /**
+   * Cache the computed return value of [queryParameters].
+   */
+  Map<String, String> _queryParameters;
+  Map<String, List<String>> _queryParameterLists;
+
+  /// Internal non-verifying constructor. Only call with validated arguments.
+  _Uri._internal(this.scheme,
+                 this._userInfo,
+                 this._host,
+                 this._port,
+                 this._path,
+                 this._query,
+                 this._fragment);
+
+  /// Create a [_Uri] from parts of [uri].
+  ///
+  /// The parameters specify the start/end of particular components of the URI.
+  /// The [scheme] may contain a string representing a normalized scheme
+  /// component if one has already been discovered.
+  factory _Uri.notSimple(String uri, int start, int end, int schemeEnd,
+                        int hostStart, int portStart, int pathStart,
+                        int queryStart, int fragmentStart, String scheme) {
+    if (scheme == null) {
+      scheme = "";
+      if (schemeEnd > start) {
+        scheme = _makeScheme(uri, start, schemeEnd);
+      } else if (schemeEnd == start) {
+        _fail(uri, start, "Invalid empty scheme");
+      }
+    }
+    String userInfo = "";
+    String host;
+    int port;
+    if (hostStart > start) {
+      int userInfoStart = schemeEnd + 3;
+      if (userInfoStart < hostStart) {
+        userInfo = _makeUserInfo(uri, userInfoStart, hostStart - 1);
+      }
+      host = _makeHost(uri, hostStart, portStart, false);
+      if (portStart + 1 < pathStart) {
+        // Should throw because invalid.
+        port = int.parse(uri.substring(portStart + 1, pathStart), onError: (_) {
+          throw new FormatException("Invalid port", uri, portStart + 1);
+        });
+        port = _makePort(port, scheme);
+      }
+    }
+    String path = _makePath(uri, pathStart, queryStart, null,
+                            scheme, host != null);
+    String query;
+    if (queryStart < fragmentStart) {
+      query = _makeQuery(uri, queryStart + 1, fragmentStart, null);
+    }
+    String fragment;
+    if (fragmentStart < end) {
+      fragment = _makeFragment(uri, fragmentStart + 1, end);
+    }
+    return new _Uri._internal(scheme,
+                              userInfo,
+                              host,
+                              port,
+                              path,
+                              query,
+                              fragment);
+  }
+
+  /// Implementation of [Uri.Uri].
+  factory _Uri({String scheme,
+                String userInfo,
+                String host,
+                int port,
+                String path,
+                Iterable<String> pathSegments,
+                String query,
+                Map<String, dynamic/*String|Iterable<String>*/> queryParameters,
+                String fragment}) {
+    scheme = _makeScheme(scheme, 0, _stringOrNullLength(scheme));
+    userInfo = _makeUserInfo(userInfo, 0, _stringOrNullLength(userInfo));
+    host = _makeHost(host, 0, _stringOrNullLength(host), false);
+    // Special case this constructor for backwards compatibility.
+    if (query == "") query = null;
+    query = _makeQuery(query, 0, _stringOrNullLength(query), queryParameters);
+    fragment = _makeFragment(fragment, 0, _stringOrNullLength(fragment));
+    port = _makePort(port, scheme);
+    bool isFile = (scheme == "file");
+    if (host == null &&
+        (userInfo.isNotEmpty || port != null || isFile)) {
+      host = "";
+    }
     bool hasAuthority = (host != null);
-    path = _makePath(uri, pathStart, index, null, scheme, hasAuthority);
-
-    if (char == _QUESTION) {
-      int numberSignIndex = -1;
-      for (int i = index + 1; i < end; i++) {
-        if (uri.codeUnitAt(i) == _NUMBER_SIGN) {
-          numberSignIndex = i;
-          break;
-        }
-      }
-      if (numberSignIndex < 0) {
-        query = _makeQuery(uri, index + 1, end, null);
-      } else {
-        query = _makeQuery(uri, index + 1, numberSignIndex, null);
-        fragment = _makeFragment(uri, numberSignIndex + 1, end);
-      }
-    } else if (char == _NUMBER_SIGN) {
-      fragment = _makeFragment(uri, index + 1, end);
-    }
-    return new Uri._internal(scheme,
-                             userinfo,
-                             host,
-                             port,
-                             path,
-                             query,
-                             fragment);
-  }
+    path = _makePath(path, 0, _stringOrNullLength(path), pathSegments,
+                     scheme, hasAuthority);
+    if (scheme.isEmpty && host == null && !path.startsWith('/')) {
+      path = _normalizeRelativePath(path);
+    } else {
+      path = _removeDotSegments(path);
+    }
+    return new _Uri._internal(scheme, userInfo, host, port,
+                              path, query, fragment);
+  }
+
+  /// Implementation of [Uri.http].
+  factory _Uri.http(String authority,
+                    String unencodedPath,
+                    [Map<String, String> queryParameters]) {
+    return _makeHttpUri("http", authority, unencodedPath, queryParameters);
+  }
+
+  /// Implementation of [Uri.https].
+  factory _Uri.https(String authority,
+                     String unencodedPath,
+                     [Map<String, String> queryParameters]) {
+    return _makeHttpUri("https", authority, unencodedPath, queryParameters);
+  }
+
+  String get authority {
+    if (!hasAuthority) return "";
+    var sb = new StringBuffer();
+    _writeAuthority(sb);
+    return sb.toString();
+  }
+
+  String get userInfo => _userInfo;
+
+  String get host {
+    if (_host == null) return "";
+    if (_host.startsWith('[')) {
+      return _host.substring(1, _host.length - 1);
+    }
+    return _host;
+  }
+
+  int get port {
+    if (_port == null) return _defaultPort(scheme);
+    return _port;
+  }
+
+  // The default port for the scheme of this Uri..

[floitsch, 2016/06/29 23:41:47]

Remove trailing ".".

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Done.

+  static int _defaultPort(String scheme) {
+    if (scheme == "http") return 80;
+    if (scheme == "https") return 443;
+    return 0;
+  }
+
+  String get path => _path;
+
+  String get query => (_query == null) ? "" : _query;

[floitsch, 2016/06/29 23:41:47]

_query ?? ""

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Done.

+
+  String get fragment => (_fragment == null) ? "" : _fragment;

[floitsch, 2016/06/29 23:41:47]

_fragment ?? ""

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Done.

 
   // Report a parse failure.
   static void _fail(String uri, int index, String message) {
     throw new FormatException(message, uri, index);
   }
 
   static Uri _makeHttpUri(String scheme,
                           String authority,
                           String unencodedPath,
                           Map<String, String> queryParameters) {
     var userInfo = "";
     var host = null;
     var port = null;
 
     if (authority != null && authority.isNotEmpty) {
       var hostStart = 0;
       // Split off the user info.
       bool hasUserInfo = false;
       for (int i = 0; i < authority.length; i++) {
-        if (authority.codeUnitAt(i) == _AT_SIGN) {
+        const int atSign = 0x40;
+        if (authority.codeUnitAt(i) == atSign) {
           hasUserInfo = true;
           userInfo = authority.substring(0, i);
           hostStart = i + 1;
           break;
         }
       }
       var hostEnd = hostStart;
       if (hostStart < authority.length &&
           authority.codeUnitAt(hostStart) == _LEFT_BRACKET) {
         // IPv6 host.
         for (; hostEnd < authority.length; hostEnd++) {
           if (authority.codeUnitAt(hostEnd) == _RIGHT_BRACKET) break;
         }
         if (hostEnd == authority.length) {
           throw new FormatException("Invalid IPv6 host entry.",
                                     authority, hostStart);
         }
-        parseIPv6Address(authority, hostStart + 1, hostEnd);
+        Uri.parseIPv6Address(authority, hostStart + 1, hostEnd);
         hostEnd++;  // Skip the closing bracket.
         if (hostEnd != authority.length &&
             authority.codeUnitAt(hostEnd) != _COLON) {
           throw new FormatException("Invalid end of authority",
                                     authority, hostEnd);
         }
       }
       // Split host and port.
       bool hasPort = false;
       for (; hostEnd < authority.length; hostEnd++) {
         if (authority.codeUnitAt(hostEnd) == _COLON) {
           var portString = authority.substring(hostEnd + 1);
           // We allow the empty port - falling back to initial value.
           if (portString.isNotEmpty) port = int.parse(portString);
           break;
         }
       }
       host = authority.substring(hostStart, hostEnd);
     }
     return new Uri(scheme: scheme,
                    userInfo: userInfo,
                    host: host,
                    port: port,
                    pathSegments: unencodedPath.split("/"),
                    queryParameters: queryParameters);
   }
 
-  /**
-   * Creates a new file URI from an absolute or relative file path.
-   *
-   * The file path is passed in [path].
-   *
-   * This path is interpreted using either Windows or non-Windows
-   * semantics.
-   *
-   * With non-Windows semantics the slash ("/") is used to separate
-   * path segments.
-   *
-   * With Windows semantics, backslash ("\") and forward-slash ("/")
-   * are used to separate path segments, except if the path starts
-   * with "\\?\" in which case, only backslash ("\") separates path
-   * segments.
-   *
-   * If the path starts with a path separator an absolute URI is
-   * created. Otherwise a relative URI is created. One exception from
-   * this rule is that when Windows semantics is used and the path
-   * starts with a drive letter followed by a colon (":") and a
-   * path separator then an absolute URI is created.
-   *
-   * The default for whether to use Windows or non-Windows semantics
-   * determined from the platform Dart is running on. When running in
-   * the standalone VM this is detected by the VM based on the
-   * operating system. When running in a browser non-Windows semantics
-   * is always used.
-   *
-   * To override the automatic detection of which semantics to use pass
-   * a value for [windows]. Passing `true` will use Windows
-   * semantics and passing `false` will use non-Windows semantics.
-   *
-   * Examples using non-Windows semantics:
-   *
-   * ```
-   * // xxx/yyy
-   * new Uri.file("xxx/yyy", windows: false);
-   *
-   * // xxx/yyy/
-   * new Uri.file("xxx/yyy/", windows: false);
-   *
-   * // file:///xxx/yyy
-   * new Uri.file("/xxx/yyy", windows: false);
-   *
-   * // file:///xxx/yyy/
-   * new Uri.file("/xxx/yyy/", windows: false);
-   *
-   * // C:
-   * new Uri.file("C:", windows: false);
-   * ```
-   *
-   * Examples using Windows semantics:
-   *
-   * ```
-   * // xxx/yyy
-   * new Uri.file(r"xxx\yyy", windows: true);
-   *
-   * // xxx/yyy/
-   * new Uri.file(r"xxx\yyy\", windows: true);
-   *
-   * file:///xxx/yyy
-   * new Uri.file(r"\xxx\yyy", windows: true);
-   *
-   * file:///xxx/yyy/
-   * new Uri.file(r"\xxx\yyy/", windows: true);
-   *
-   * // file:///C:/xxx/yyy
-   * new Uri.file(r"C:\xxx\yyy", windows: true);
-   *
-   * // This throws an error. A path with a drive letter is not absolute.
-   * new Uri.file(r"C:", windows: true);
-   *
-   * // This throws an error. A path with a drive letter is not absolute.
-   * new Uri.file(r"C:xxx\yyy", windows: true);
-   *
-   * // file://server/share/file
-   * new Uri.file(r"\\server\share\file", windows: true);
-   * ```
-   *
-   * If the path passed is not a legal file path [ArgumentError] is thrown.
-   */
-  factory Uri.file(String path, {bool windows}) {
-    windows = (windows == null) ? Uri._isWindows : windows;
+  /// Implementation of [Uri.file].
+  factory _Uri.file(String path, {bool windows}) {
+    windows = (windows == null) ? _Uri._isWindows : windows;
     return windows ? _makeWindowsFileUrl(path, false)
                    : _makeFileUri(path, false);
   }
 
-  /**
-   * Like [Uri.file] except that a non-empty URI path ends in a slash.
-   *
-   * If [path] is not empty, and it doesn't end in a directory separator,
-   * then a slash is added to the returned URI's path.
-   * In all other cases, the result is the same as returned by `Uri.file`.
-   */
-  factory Uri.directory(String path, {bool windows}) {
-    windows = (windows == null) ? Uri._isWindows : windows;
+  /// Implementation of [Uri.directory].
+  factory _Uri.directory(String path, {bool windows}) {
+    windows = (windows == null) ? _Uri._isWindows : windows;
     return windows ? _makeWindowsFileUrl(path, true)
                    : _makeFileUri(path, true);
   }
 
-  /**
-   * Creates a `data:` URI containing the [content] string.
-   *
-   * Converts the content to a bytes using [encoding] or the charset specified
-   * in [parameters] (defaulting to US-ASCII if not specified or unrecognized),
-   * then encodes the bytes into the resulting data URI.
-   *
-   * Defaults to encoding using percent-encoding (any non-ASCII or non-URI-valid
-   * bytes is replaced by a percent encoding). If [base64] is true, the bytes
-   * are instead encoded using [BASE64].
-   *
-   * If [encoding] is not provided and [parameters] has a `charset` entry,
-   * that name is looked up using [Encoding.getByName],
-   * and if the lookup returns an encoding, that encoding is used to convert
-   * [content] to bytes.
-   * If providing both an [encoding] and a charset [parameter], they should
-   * agree, otherwise decoding won't be able to use the charset parameter
-   * to determine the encoding.
-   *
-   * If [mimeType] and/or [parameters] are supplied, they are added to the
-   * created URI. If any of these contain characters that are not allowed
-   * in the data URI, the character is percent-escaped. If the character is
-   * non-ASCII, it is first UTF-8 encoded and then the bytes are percent
-   * encoded. An omitted [mimeType] in a data URI means `text/plain`, just
-   * as an omitted `charset` parameter defaults to meaning `US-ASCII`.
-   *
-   * To read the content back, use [UriData.contentAsString].
-   */
-  factory Uri.dataFromString(String content,
-                             {String mimeType,
-                              Encoding encoding,
-                              Map<String, String> parameters,
-                              bool base64: false}) {
-    UriData data =  new UriData.fromString(content,
-                                           mimeType: mimeType,
-                                           encoding: encoding,
-                                           parameters: parameters,
-                                           base64: base64);
-    return data.uri;
-  }
 
-  /**
-   * Creates a `data:` URI containing an encoding of [bytes].
-   *
-   * Defaults to Base64 encoding the bytes, but if [percentEncoded]
-   * is `true`, the bytes will instead be percent encoded (any non-ASCII
-   * or non-valid-ASCII-character byte is replaced by a percent encoding).
-   *
-   * To read the bytes back, use [UriData.contentAsBytes].
-   *
-   * It defaults to having the mime-type `application/octet-stream`.
-   * The [mimeType] and [parameters] are added to the created URI.
-   * If any of these contain characters that are not allowed
-   * in the data URI, the character is percent-escaped. If the character is
-   * non-ASCII, it is first UTF-8 encoded and then the bytes are percent
-   * encoded.
-   */
-  factory Uri.dataFromBytes(List<int> bytes,
-                            {mimeType: "application/octet-stream",
-                             Map<String, String> parameters,
-                             percentEncoded: false}) {
-    UriData data = new UriData.fromBytes(bytes,
-                                         mimeType: mimeType,
-                                         parameters: parameters,
-                                         percentEncoded: percentEncoded);
-    return data.uri;
-  }
-
-  /**
-   * Returns the natural base URI for the current platform.
-   *
-   * When running in a browser this is the current URL (from
-   * `window.location.href`).
-   *
-   * When not running in a browser this is the file URI referencing
-   * the current working directory.
-   */
-  external static Uri get base;
-
+  /// Used internally in path-related constructors.
   external static bool get _isWindows;
 
   static _checkNonWindowsPathReservedCharacters(List<String> segments,
                                                 bool argumentError) {
     segments.forEach((segment) {
       if (segment.contains("/")) {
         if (argumentError) {
           throw new ArgumentError("Illegal path character $segment");
         } else {
           throw new UnsupportedError("Illegal path character $segment");
@@ skipping 112 matching lines @@
       _checkWindowsPathReservedCharacters(pathSegments, true);
       if (slashTerminated &&
           pathSegments.isNotEmpty &&
           pathSegments.last.isNotEmpty) {
         pathSegments.add("");  // Extra separator at end.
       }
       return new Uri(pathSegments: pathSegments);
     }
   }
 
-  /**
-   * Returns a new `Uri` based on this one, but with some parts replaced.
-   *
-   * This method takes the same parameters as the [new Uri] constructor,
-   * and they have the same meaning.
-   *
-   * At most one of [path] and [pathSegments] must be provided.
-   * Likewise, at most one of [query] and [queryParameters] must be provided.
-   *
-   * Each part that is not provided will default to the corresponding
-   * value from this `Uri` instead.
-   *
-   * This method is different from [Uri.resolve] which overrides in a
-   * hierarchial manner,
-   * and can instead replace each part of a `Uri` individually.
-   *
-   * Example:
-   *
-   *     Uri uri1 = Uri.parse("a://b@c:4/d/e?f#g");
-   *     Uri uri2 = uri1.replace(scheme: "A", path: "D/E/E", fragment: "G");
-   *     print(uri2);  // prints "A://b@c:4/D/E/E/?f#G"
-   *
-   * This method acts similarly to using the `new Uri` constructor with
-   * some of the arguments taken from this `Uri` . Example:
-   *
-   *     Uri uri3 = new Uri(
-   *         scheme: "A",
-   *         userInfo: uri1.userInfo,
-   *         host: uri1.host,
-   *         port: uri1.port,
-   *         path: "D/E/E",
-   *         query: uri1.query,
-   *         fragment: "G");
-   *     print(uri3);  // prints "A://b@c:4/D/E/E/?f#G"
-   *     print(uri2 == uri3);  // prints true.
-   *
-   * Using this method can be seen as a shorthand for the `Uri` constructor
-   * call above, but may also be slightly faster because the parts taken
-   * from this `Uri` need not be checked for validity again.
-   */
   Uri replace({String scheme,
                String userInfo,
                String host,
                int port,
                String path,
                Iterable<String> pathSegments,
                String query,
                Map<String, dynamic/*String|Iterable<String>*/> queryParameters,
                String fragment}) {
     // Set to true if the scheme has (potentially) changed.
@@ skipping 46 matching lines @@
     } else {
       query = this._query;
     }
 
     if (fragment != null) {
       fragment = _makeFragment(fragment, 0, fragment.length);
     } else {
       fragment = this._fragment;
     }
 
-    return new Uri._internal(
+    return new _Uri._internal(
         scheme, userInfo, host, port, path, query, fragment);
   }
 
-  /**
-   * Returns a `Uri` that differs from this only in not having a fragment.
-   *
-   * If this `Uri` does not have a fragment, it is itself returned.
-   */
   Uri removeFragment() {
     if (!this.hasFragment) return this;
-    return new Uri._internal(scheme, _userInfo, _host, _port,
+    return new _Uri._internal(scheme, _userInfo, _host, _port,
                              _path, _query, null);
   }
 
-  /**
-   * Returns the URI path split into its segments. Each of the segments in the
-   * returned list have been decoded. If the path is empty the empty list will
-   * be returned. A leading slash `/` does not affect the segments returned.
-   *
-   * The returned list is unmodifiable and will throw [UnsupportedError] on any
-   * calls that would mutate it.
-   */
   List<String> get pathSegments {
     var result = _pathSegments;
     if (result != null) return result;
 
     var pathToSplit = path;
     if (pathToSplit.isNotEmpty && pathToSplit.codeUnitAt(0) == _SLASH) {
       pathToSplit = pathToSplit.substring(1);
     }
     result = (pathToSplit == "")
         ? const<String>[]
         : new List<String>.unmodifiable(
               pathToSplit.split("/").map(Uri.decodeComponent));
     _pathSegments = result;
     return result;
   }
 
-  /**
-   * Returns the URI query split into a map according to the rules
-   * specified for FORM post in the [HTML 4.01 specification section
-   * 17.13.4](http://www.w3.org/TR/REC-html40/interact/forms.html#h-17.13.4 "HTML 4.01 section 17.13.4").
-   * Each key and value in the returned map has been decoded.
-   * If there is no query the empty map is returned.
-   *
-   * Keys in the query string that have no value are mapped to the
-   * empty string.
-   * If a key occurs more than once in the query string, it is mapped to
-   * an arbitrary choice of possible value.
-   * The [queryParametersAll] getter can provide a map
-   * that maps keys to all of their values.
-   *
-   * The returned map is unmodifiable.
-   */
   Map<String, String> get queryParameters {
     if (_queryParameters == null) {
       _queryParameters =
-          new UnmodifiableMapView<String, String>(splitQueryString(query));
+          new UnmodifiableMapView<String, String>(Uri.splitQueryString(query));
     }
     return _queryParameters;
   }
 
-  /**
-   * Returns the URI query split into a map according to the rules
-   * specified for FORM post in the [HTML 4.01 specification section
-   * 17.13.4](http://www.w3.org/TR/REC-html40/interact/forms.html#h-17.13.4 "HTML 4.01 section 17.13.4").
-   * Each key and value in the returned map has been decoded. If there is no
-   * query the empty map is returned.
-   *
-   * Keys are mapped to lists of their values. If a key occurs only once,
-   * its value is a singleton list. If a key occurs with no value, the
-   * empty string is used as the value for that occurrence.
-   *
-   * The returned map and the lists it contains are unmodifiable.
-   */
   Map<String, List<String>> get queryParametersAll {
     if (_queryParameterLists == null) {
       Map queryParameterLists = _splitQueryStringAll(query);
       for (var key in queryParameterLists.keys) {
         queryParameterLists[key] =
             new List<String>.unmodifiable(queryParameterLists[key]);
       }
       _queryParameterLists =
           new Map<String, List<String>>.unmodifiable(queryParameterLists);
     }
     return _queryParameterLists;
   }
 
-  /**
-   * Returns a URI where the path has been normalized.
-   *
-   * A normalized path does not contain `.` segments or non-leading `..`
-   * segments.
-   * Only a relative path with no scheme or authority may contain
-   * leading `..` segments,
-   * a path that starts with `/` will also drop any leading `..` segments.
-   *
-   * This uses the same normalization strategy as `new Uri().resolve(this)`.
-   *
-   * Does not change any part of the URI except the path.
-   *
-   * The default implementation of `Uri` always normalizes paths, so calling
-   * this function has no effect.
-   */
   Uri normalizePath() {
     String path = _normalizePath(_path, scheme, hasAuthority);
     if (identical(path, _path)) return this;
     return this.replace(path: path);
   }
 
   static int _makePort(int port, String scheme) {
     // Perform scheme specific normalization.
     if (port != null && port == _defaultPort(scheme)) return null;
     return port;
@@ skipping 12 matching lines @@
    */
   static String _makeHost(String host, int start, int end, bool strictIPv6) {
     // TODO(lrn): Should we normalize IPv6 addresses according to RFC 5952?
     if (host == null) return null;
     if (start == end) return "";
     // Host is an IPv6 address if it starts with '[' or contains a colon.
     if (host.codeUnitAt(start) == _LEFT_BRACKET) {
       if (host.codeUnitAt(end - 1) != _RIGHT_BRACKET) {
         _fail(host, start, 'Missing end `]` to match `[` in host');
       }
-      parseIPv6Address(host, start + 1, end - 1);
+      Uri.parseIPv6Address(host, start + 1, end - 1);
       // RFC 5952 requires hex digits to be lower case.
       return host.substring(start, end).toLowerCase();
     }
     if (!strictIPv6) {
       // TODO(lrn): skip if too short to be a valid IPv6 address?
       for (int i = start; i < end; i++) {
         if (host.codeUnitAt(i) == _COLON) {
-          parseIPv6Address(host, start, end);
+          Uri.parseIPv6Address(host, start, end);
           return '[$host]';
         }
       }
     }
     return _normalizeRegName(host, start, end);
   }
 
   static bool _isRegNameChar(int char) {
     return char < 127 && (_regNameTable[char >> 4] & (1 << (char & 0xf))) != 0;
   }
@@ skipping 92 matching lines @@
       final int codeUnit = scheme.codeUnitAt(i);
       if (!_isSchemeCharacter(codeUnit)) {
         _fail(scheme, i, "Illegal scheme character");
       }
       if (_UPPER_CASE_A <= codeUnit && codeUnit <= _UPPER_CASE_Z) {
         containsUpperCase = true;
       }
     }
     scheme = scheme.substring(start, end);
     if (containsUpperCase) scheme = scheme.toLowerCase();
+    return _canonicalizeScheme(scheme);
+  }
+
+  // Canonicalize a few often-used scheme strings.
+  //
+  // This improves memory usage and makes comparison faster.
+  static String _canonicalizeScheme(String scheme) {
+    if (scheme == "http") return "http";
+    if (scheme == "file") return "file";
+    if (scheme == "https") return "https";
+    if (scheme == "package") return "package";
     return scheme;
   }
 
   static String _makeUserInfo(String userInfo, int start, int end) {
     if (userInfo == null) return "";
     return _normalize(userInfo, start, end, _userinfoTable);
   }
 
   static String _makePath(String path, int start, int end,
                           Iterable<String> pathSegments,
@@ skipping 66 matching lines @@
       }
     });
     return result.toString();
   }
 
   static String _makeFragment(String fragment, int start, int end) {
     if (fragment == null) return null;
     return _normalize(fragment, start, end, _queryCharTable);
   }
 
-  static int _stringOrNullLength(String s) => (s == null) ? 0 : s.length;
-
   /**
    * Performs RFC 3986 Percent-Encoding Normalization.
    *
    * Returns a replacement string that should be replace the original escape.
    * Returns null if no replacement is necessary because the escape is
    * not for an unreserved character and is already non-lower-case.
    *
    * Returns "%" if the escape is invalid (not two valid hex digits following
    * the percent sign). The calling code should replace the percent
    * sign with "%25", but leave the following two characters unmodified.
@@ skipping 24 matching lines @@
       return source.substring(index, index + 3).toUpperCase();
     }
     // Escape is retained, and is already non-lower case, so return null to
     // represent "no replacement necessary".
     return null;
   }
 
   // Converts a UTF-16 code-unit to its value as a hex digit.
   // Returns -1 for non-hex digits.
   static int _parseHexDigit(int char) {
-    int digit = char ^ Uri._ZERO;
+    const int zeroDigit = 0x30;
+    int digit = char ^ zeroDigit;
     if (digit <= 9) return digit;
     int lowerCase = char | 0x20;
-    if (Uri._LOWER_CASE_A <= lowerCase && lowerCase <= _LOWER_CASE_F) {
+    if (_LOWER_CASE_A <= lowerCase && lowerCase <= _LOWER_CASE_F) {
       return lowerCase - (_LOWER_CASE_A - 10);
     }
     return -1;
   }
 
   static String _escapeChar(int char) {
     assert(char <= 0x10ffff);  // It's a valid unicode code point.
     List<int> codeUnits;
     if (char < 0x80) {
       // ASCII, a single percent encoded sequence.
@@ skipping 214 matching lines @@
         output.add(segment);
       }
     }
     if (output.isEmpty || (output.length == 1 && output[0].isEmpty)) {
       return "./";
     }
     if (appendSlash || output.last == '..') output.add("");
     return output.join("/");
   }
 
-  /**
-   * Resolve [reference] as an URI relative to `this`.
-   *
-   * First turn [reference] into a URI using [Uri.parse]. Then resolve the
-   * resulting URI relative to `this`.
-   *
-   * Returns the resolved URI.
-   *
-   * See [resolveUri] for details.
-   */
   Uri resolve(String reference) {
     return resolveUri(Uri.parse(reference));
   }
 
-  /**
-   * Resolve [reference] as an URI relative to `this`.
-   *
-   * Returns the resolved URI.
-   *
-   * The algorithm "Transform Reference" for resolving a reference is described
-   * in [RFC-3986 Section 5](http://tools.ietf.org/html/rfc3986#section-5 "RFC-1123").
-   *
-   * Updated to handle the case where the base URI is just a relative path -
-   * that is: when it has no scheme or authority and the path does not start
-   * with a slash.
-   * In that case, the paths are combined without removing leading "..", and
-   * an empty path is not converted to "/".
-   */
   Uri resolveUri(Uri reference) {
     // From RFC 3986.
     String targetScheme;
     String targetUserInfo = "";
     String targetHost;
     int targetPort;
     String targetPath;
     String targetQuery;
     if (reference.scheme.isNotEmpty) {
       targetScheme = reference.scheme;
@@ skipping 37 matching lines @@
                 targetPath = reference.path;
               } else {
                 // Add path normalization on top of RFC algorithm.
                 targetPath = _removeDotSegments("/" + reference.path);
               }
             } else {
               var mergedPath = _mergePaths(this._path, reference.path);
               if (this.hasScheme || this.hasAuthority || this.hasAbsolutePath) {
                 targetPath = _removeDotSegments(mergedPath);
               } else {
-                // Non-RFC 3986 beavior. If both base and reference are relative
-                // path, allow the merged path to start with "..".
+                // Non-RFC 3986 behavior.
+                // If both base and reference are relative paths,
+                // allow the merged path to start with "..".
                 // The RFC only specifies the case where the base has a scheme.
                 targetPath = _normalizeRelativePath(mergedPath);
               }
             }
           }
           if (reference.hasQuery) targetQuery = reference.query;
         }
       }
     }
     String fragment = reference.hasFragment ? reference.fragment : null;
-    return new Uri._internal(targetScheme,
-                             targetUserInfo,
-                             targetHost,
-                             targetPort,
-                             targetPath,
-                             targetQuery,
-                             fragment);
+    return new _Uri._internal(targetScheme,
+                              targetUserInfo,
+                              targetHost,
+                              targetPort,
+                              targetPath,
+                              targetQuery,
+                              fragment);
   }
 
-  /**
-   * Returns whether the URI has a [scheme] component.
-   */
   bool get hasScheme => scheme.isNotEmpty;
 
-  /**
-   * Returns whether the URI has an [authority] component.
-   */
   bool get hasAuthority => _host != null;
 
-  /**
-   * Returns whether the URI has an explicit port.
-   *
-   * If the port number is the default port number
-   * (zero for unrecognized schemes, with http (80) and https (443) being
-   * recognized),
-   * then the port is made implicit and omitted from the URI.
-   */
   bool get hasPort => _port != null;
 
-  /**
-   * Returns whether the URI has a query part.
-   */
   bool get hasQuery => _query != null;
 
-  /**
-   * Returns whether the URI has a fragment part.
-   */
   bool get hasFragment => _fragment != null;
 
-  /**
-   * Returns whether the URI has an empty path.
-   */
   bool get hasEmptyPath => _path.isEmpty;
 
-  /**
-   * Returns whether the URI has an absolute path (starting with '/').
-   */
   bool get hasAbsolutePath => _path.startsWith('/');
 
-  /**
-   * Returns the origin of the URI in the form scheme://host:port for the
-   * schemes http and https.
-   *
-   * It is an error if the scheme is not "http" or "https".
-   *
-   * See: http://www.w3.org/TR/2011/WD-html5-20110405/origin-0.html#origin
-   */
   String get origin {
     if (scheme == "" || _host == null || _host == "") {
       throw new StateError("Cannot use origin without a scheme: $this");
     }
     if (scheme != "http" && scheme != "https") {
       throw new StateError(
         "Origin is only applicable schemes http and https: $this");
     }
     if (_port == null) return "$scheme://$_host";
     return "$scheme://$_host:$_port";
   }
 
-  /**
-   * Returns the file path from a file URI.
-   *
-   * The returned path has either Windows or non-Windows
-   * semantics.
-   *
-   * For non-Windows semantics the slash ("/") is used to separate
-   * path segments.
-   *
-   * For Windows semantics the backslash ("\") separator is used to
-   * separate path segments.
-   *
-   * If the URI is absolute the path starts with a path separator
-   * unless Windows semantics is used and the first path segment is a
-   * drive letter. When Windows semantics is used a host component in
-   * the uri in interpreted as a file server and a UNC path is
-   * returned.
-   *
-   * The default for whether to use Windows or non-Windows semantics
-   * determined from the platform Dart is running on. When running in
-   * the standalone VM this is detected by the VM based on the
-   * operating system. When running in a browser non-Windows semantics
-   * is always used.
-   *
-   * To override the automatic detection of which semantics to use pass
-   * a value for [windows]. Passing `true` will use Windows
-   * semantics and passing `false` will use non-Windows semantics.
-   *
-   * If the URI ends with a slash (i.e. the last path component is
-   * empty) the returned file path will also end with a slash.
-   *
-   * With Windows semantics URIs starting with a drive letter cannot
-   * be relative to the current drive on the designated drive. That is
-   * for the URI `file:///c:abc` calling `toFilePath` will throw as a
-   * path segment cannot contain colon on Windows.
-   *
-   * Examples using non-Windows semantics (resulting of calling
-   * toFilePath in comment):
-   *
-   *     Uri.parse("xxx/yyy");  // xxx/yyy
-   *     Uri.parse("xxx/yyy/");  // xxx/yyy/
-   *     Uri.parse("file:///xxx/yyy");  // /xxx/yyy
-   *     Uri.parse("file:///xxx/yyy/");  // /xxx/yyy/
-   *     Uri.parse("file:///C:");  // /C:
-   *     Uri.parse("file:///C:a");  // /C:a
-   *
-   * Examples using Windows semantics (resulting URI in comment):
-   *
-   *     Uri.parse("xxx/yyy");  // xxx\yyy
-   *     Uri.parse("xxx/yyy/");  // xxx\yyy\
-   *     Uri.parse("file:///xxx/yyy");  // \xxx\yyy
-   *     Uri.parse("file:///xxx/yyy/");  // \xxx\yyy/
-   *     Uri.parse("file:///C:/xxx/yyy");  // C:\xxx\yyy
-   *     Uri.parse("file:C:xxx/yyy");  // Throws as a path segment
-   *                                   // cannot contain colon on Windows.
-   *     Uri.parse("file://server/share/file");  // \\server\share\file
-   *
-   * If the URI is not a file URI calling this throws
-   * [UnsupportedError].
-   *
-   * If the URI cannot be converted to a file path calling this throws
-   * [UnsupportedError].
-   */
   String toFilePath({bool windows}) {
     if (scheme != "" && scheme != "file") {
       throw new UnsupportedError(
           "Cannot extract a file path from a $scheme URI");
     }
     if (query != "") {
       throw new UnsupportedError(
           "Cannot extract a file path from a URI with a query component");
     }
     if (fragment != "") {
       throw new UnsupportedError(
           "Cannot extract a file path from a URI with a fragment component");
     }
     if (windows == null) windows = _isWindows;
-    return windows ? _toWindowsFilePath() : _toFilePath();
+    return windows ? _toWindowsFilePath(this) : _toFilePath();
   }
 
   String _toFilePath() {
-    if (host != "") {
+    if (hasAuthority && host != "") {
       throw new UnsupportedError(
           "Cannot extract a non-Windows file path from a file URI "
           "with an authority");
     }
+    // Use path segments to have any escapes unescaped.
+    var pathSegments = this.pathSegments;
     _checkNonWindowsPathReservedCharacters(pathSegments, false);
     var result = new StringBuffer();
-    if (_isPathAbsolute) result.write("/");
+    if (hasAbsolutePath) result.write("/");
     result.writeAll(pathSegments, "/");
     return result.toString();
   }
 
-  String _toWindowsFilePath() {
+  static String _toWindowsFilePath(Uri uri) {
     bool hasDriveLetter = false;
-    var segments = pathSegments;
+    var segments = uri.pathSegments;
     if (segments.length > 0 &&
         segments[0].length == 2 &&
         segments[0].codeUnitAt(1) == _COLON) {
       _checkWindowsDriveLetter(segments[0].codeUnitAt(0), false);
       _checkWindowsPathReservedCharacters(segments, false, 1);
       hasDriveLetter = true;
     } else {
-      _checkWindowsPathReservedCharacters(segments, false);
+      _checkWindowsPathReservedCharacters(segments, false, 0);
     }
     var result = new StringBuffer();
-    if (_isPathAbsolute && !hasDriveLetter) result.write("\\");
-    if (host != "") {
-      result.write("\\");
-      result.write(host);
-      result.write("\\");
+    if (uri.hasAbsolutePath && !hasDriveLetter) result.write(r"\");
+    if (uri.hasAuthority) {
+      var host = uri.host;
+      if (host.isNotEmpty) {
+        result.write(r"\");
+        result.write(host);
+        result.write(r"\");
+      }
     }
-    result.writeAll(segments, "\\");
-    if (hasDriveLetter && segments.length == 1) result.write("\\");
+    result.writeAll(segments, r"\");
+    if (hasDriveLetter && segments.length == 1) result.write(r"\");
     return result.toString();
   }
 
   bool get _isPathAbsolute {
-    if (path == null || path.isEmpty) return false;
-    return path.startsWith('/');
+    return _path != null && _path.startsWith('/');
   }
 
   void _writeAuthority(StringSink ss) {
     if (_userInfo.isNotEmpty) {
       ss.write(_userInfo);
       ss.write("@");
     }
     if (_host != null) ss.write(_host);
     if (_port != null) {
       ss.write(":");
       ss.write(_port);
     }
   }
 
   /**
    * Access the structure of a `data:` URI.
    *
    * Returns a [UriData] object for `data:` URIs and `null` for all other
    * URIs.
    * The [UriData] object can be used to access the media type and data
    * of a `data:` URI.
    */
   UriData get data => (scheme == "data") ? new UriData.fromUri(this) : null;
 
   String toString() {
+    if (_text != null) return _text;
     StringBuffer sb = new StringBuffer();
-    _addIfNonEmpty(sb, scheme, scheme, ':');
+    if (scheme.isNotEmpty) sb..write(scheme)..write(":");
     if (hasAuthority || path.startsWith("//") || (scheme == "file")) {
       // File URIS always have the authority, even if it is empty.
       // The empty URI means "localhost".
       sb.write("//");
       _writeAuthority(sb);
     }
     sb.write(path);
-    if (_query != null) { sb..write("?")..write(_query); }
-    if (_fragment != null) { sb..write("#")..write(_fragment); }
-    return sb.toString();
+    if (_query != null) sb..write("?")..write(_query);
+    if (_fragment != null) sb..write("#")..write(_fragment);
+    _text = sb.toString();
+    return _text;
   }
 
   bool operator==(other) {
-    if (other is! Uri) return false;
-    Uri uri = other;
-    return scheme       == uri.scheme       &&
-           hasAuthority == uri.hasAuthority &&
-           userInfo     == uri.userInfo     &&
-           host         == uri.host         &&
-           port         == uri.port         &&
-           path         == uri.path         &&
-           hasQuery     == uri.hasQuery     &&
-           query        == uri.query        &&
-           hasFragment  == uri.hasFragment  &&
-           fragment     == uri.fragment;
+    if (identical(this, other)) return true;
+    if (other is Uri) {
+      Uri uri = other;
+      return scheme       == uri.scheme       &&
+             hasAuthority == uri.hasAuthority &&
+             userInfo     == uri.userInfo     &&
+             host         == uri.host         &&
+             port         == uri.port         &&
+             path         == uri.path         &&
+             hasQuery     == uri.hasQuery     &&
+             query        == uri.query        &&
+             hasFragment  == uri.hasFragment  &&
+             fragment     == uri.fragment;
+    }
+    return false;
   }
 
   int get hashCode {
-    int combine(part, current) {
-      // The sum is truncated to 30 bits to make sure it fits into a Smi.
-      return (current * 31 + part.hashCode) & 0x3FFFFFFF;
-    }
-    return combine(scheme, combine(userInfo, combine(host, combine(port,
-        combine(path, combine(query, combine(fragment, 1)))))));
-  }
-
-  static void _addIfNonEmpty(StringBuffer sb, String test,
-                             String first, String second) {
-    if ("" != test) {
-      sb.write(first);
-      sb.write(second);
-    }
-  }
-
-  /**
-   * Encode the string [component] using percent-encoding to make it
-   * safe for literal use as a URI component.
-   *
-   * All characters except uppercase and lowercase letters, digits and
-   * the characters `-_.!~*'()` are percent-encoded. This is the
-   * set of characters specified in RFC 2396 and the which is
-   * specified for the encodeUriComponent in ECMA-262 version 5.1.
-   *
-   * When manually encoding path segments or query components remember
-   * to encode each part separately before building the path or query
-   * string.
-   *
-   * For encoding the query part consider using
-   * [encodeQueryComponent].
-   *
-   * To avoid the need for explicitly encoding use the [pathSegments]
-   * and [queryParameters] optional named arguments when constructing
-   * a [Uri].
-   */
-  static String encodeComponent(String component) {
-    return _uriEncode(_unreserved2396Table, component, UTF8, false);
-  }
-
-  /**
-   * Encode the string [component] according to the HTML 4.01 rules
-   * for encoding the posting of a HTML form as a query string
-   * component.
-   *
-   * Encode the string [component] according to the HTML 4.01 rules
-   * for encoding the posting of a HTML form as a query string
-   * component.
-
-   * The component is first encoded to bytes using [encoding].
-   * The default is to use [UTF8] encoding, which preserves all
-   * the characters that don't need encoding.
-
-   * Then the resulting bytes are "percent-encoded". This transforms
-   * spaces (U+0020) to a plus sign ('+') and all bytes that are not
-   * the ASCII decimal digits, letters or one of '-._~' are written as
-   * a percent sign '%' followed by the two-digit hexadecimal
-   * representation of the byte.
-
-   * Note that the set of characters which are percent-encoded is a
-   * superset of what HTML 4.01 requires, since it refers to RFC 1738
-   * for reserved characters.
-   *
-   * When manually encoding query components remember to encode each
-   * part separately before building the query string.
-   *
-   * To avoid the need for explicitly encoding the query use the
-   * [queryParameters] optional named arguments when constructing a
-   * [Uri].
-   *
-   * See http://www.w3.org/TR/html401/interact/forms.html#h-17.13.4.2 for more
-   * details.
-   */
-  static String encodeQueryComponent(String component,
-                                     {Encoding encoding: UTF8}) {
-    return _uriEncode(_unreservedTable, component, encoding, true);
-  }
-
-  /**
-   * Decodes the percent-encoding in [encodedComponent].
-   *
-   * Note that decoding a URI component might change its meaning as
-   * some of the decoded characters could be characters with are
-   * delimiters for a given URI componene type. Always split a URI
-   * component using the delimiters for the component before decoding
-   * the individual parts.
-   *
-   * For handling the [path] and [query] components consider using
-   * [pathSegments] and [queryParameters] to get the separated and
-   * decoded component.
-   */
-  static String decodeComponent(String encodedComponent) {
-    return _uriDecode(encodedComponent, 0, encodedComponent.length,
-                      UTF8, false);
-  }
-
-  /**
-   * Decodes the percent-encoding in [encodedComponent], converting
-   * pluses to spaces.
-   *
-   * It will create a byte-list of the decoded characters, and then use
-   * [encoding] to decode the byte-list to a String. The default encoding is
-   * UTF-8.
-   */
-  static String decodeQueryComponent(
-      String encodedComponent,
-      {Encoding encoding: UTF8}) {
-    return _uriDecode(encodedComponent, 0, encodedComponent.length,
-                      encoding, true);
-  }
-
-  /**
-   * Encode the string [uri] using percent-encoding to make it
-   * safe for literal use as a full URI.
-   *
-   * All characters except uppercase and lowercase letters, digits and
-   * the characters `!#$&'()*+,-./:;=?@_~` are percent-encoded. This
-   * is the set of characters specified in in ECMA-262 version 5.1 for
-   * the encodeURI function .
-   */
-  static String encodeFull(String uri) {
-    return _uriEncode(_encodeFullTable, uri, UTF8, false);
-  }
-
-  /**
-   * Decodes the percent-encoding in [uri].
-   *
-   * Note that decoding a full URI might change its meaning as some of
-   * the decoded characters could be reserved characters. In most
-   * cases an encoded URI should be parsed into components using
-   * [Uri.parse] before decoding the separate components.
-   */
-  static String decodeFull(String uri) {
-    return _uriDecode(uri, 0, uri.length, UTF8, false);
-  }
-
-  /**
-   * Returns the [query] split into a map according to the rules
-   * specified for FORM post in the [HTML 4.01 specification section
-   * 17.13.4](http://www.w3.org/TR/REC-html40/interact/forms.html#h-17.13.4 "HTML 4.01 section 17.13.4").
-   * Each key and value in the returned map has been decoded. If the [query]
-   * is the empty string an empty map is returned.
-   *
-   * Keys in the query string that have no value are mapped to the
-   * empty string.
-   *
-   * Each query component will be decoded using [encoding]. The default encoding
-   * is UTF-8.
-   */
-  static Map<String, String> splitQueryString(String query,
-                                              {Encoding encoding: UTF8}) {
-    return query.split("&").fold({}, (map, element) {
-      int index = element.indexOf("=");
-      if (index == -1) {
-        if (element != "") {
-          map[decodeQueryComponent(element, encoding: encoding)] = "";
-        }
-      } else if (index != 0) {
-        var key = element.substring(0, index);
-        var value = element.substring(index + 1);
-        map[Uri.decodeQueryComponent(key, encoding: encoding)] =
-            decodeQueryComponent(value, encoding: encoding);
-      }
-      return map;
-    });
+    return (_text ?? toString()).hashCode;
   }
 
   static List _createList() => [];
 
   static Map _splitQueryStringAll(
       String query, {Encoding encoding: UTF8}) {
     Map result = {};
     int i = 0;
     int start = 0;
     int equalsIndex = -1;
@@ skipping 22 matching lines @@
         parsePair(start, equalsIndex, i);
         start = i + 1;
         equalsIndex = -1;
       }
       i++;
     }
     parsePair(start, equalsIndex, i);
     return result;
   }
 
-  /**
-   * Parse the [host] as an IP version 4 (IPv4) address, returning the address
-   * as a list of 4 bytes in network byte order (big endian).
-   *
-   * Throws a [FormatException] if [host] is not a valid IPv4 address
-   * representation.
-   */
-  static List<int> parseIPv4Address(String host) {
-    void error(String msg) {
-      throw new FormatException('Illegal IPv4 address, $msg');
-    }
-    var bytes = host.split('.');
-    if (bytes.length != 4) {
-      error('IPv4 address should contain exactly 4 parts');
-    }
-    // TODO(ajohnsen): Consider using Uint8List.
-    return bytes
-        .map((byteString) {
-          int byte = int.parse(byteString);
-          if (byte < 0 || byte > 255) {
-            error('each part must be in the range of `0..255`');
-          }
-          return byte;
-        })
-        .toList();
-  }
-
-  /**
-   * Parse the [host] as an IP version 6 (IPv6) address, returning the address
-   * as a list of 16 bytes in network byte order (big endian).
-   *
-   * Throws a [FormatException] if [host] is not a valid IPv6 address
-   * representation.
-   *
-   * Acts on the substring from [start] to [end]. If [end] is omitted, it
-   * defaults ot the end of the string.
-   *
-   * Some examples of IPv6 addresses:
-   *  * ::1
-   *  * FEDC:BA98:7654:3210:FEDC:BA98:7654:3210
-   *  * 3ffe:2a00:100:7031::1
-   *  * ::FFFF:129.144.52.38
-   *  * 2010:836B:4179::836B:4179
-   */
-  static List<int> parseIPv6Address(String host, [int start = 0, int end]) {
-    if (end == null) end = host.length;
-    // An IPv6 address consists of exactly 8 parts of 1-4 hex digits, seperated
-    // by `:`'s, with the following exceptions:
-    //
-    //  - One (and only one) wildcard (`::`) may be present, representing a fill
-    //    of 0's. The IPv6 `::` is thus 16 bytes of `0`.
-    //  - The last two parts may be replaced by an IPv4 address.
-    void error(String msg, [position]) {
-      throw new FormatException('Illegal IPv6 address, $msg', host, position);
-    }
-    int parseHex(int start, int end) {
-      if (end - start > 4) {
-        error('an IPv6 part can only contain a maximum of 4 hex digits', start);
-      }
-      int value = int.parse(host.substring(start, end), radix: 16);
-      if (value < 0 || value > (1 << 16) - 1) {
-        error('each part must be in the range of `0x0..0xFFFF`', start);
-      }
-      return value;
-    }
-    if (host.length < 2) error('address is too short');
-    List<int> parts = [];
-    bool wildcardSeen = false;
-    int partStart = start;
-    // Parse all parts, except a potential last one.
-    for (int i = start; i < end; i++) {
-      if (host.codeUnitAt(i) == _COLON) {
-        if (i == start) {
-          // If we see a `:` in the beginning, expect wildcard.
-          i++;
-          if (host.codeUnitAt(i) != _COLON) {
-            error('invalid start colon.', i);
-          }
-          partStart = i;
-        }
-        if (i == partStart) {
-          // Wildcard. We only allow one.
-          if (wildcardSeen) {
-            error('only one wildcard `::` is allowed', i);
-          }
-          wildcardSeen = true;
-          parts.add(-1);
-        } else {
-          // Found a single colon. Parse [partStart..i] as a hex entry.
-          parts.add(parseHex(partStart, i));
-        }
-        partStart = i + 1;
-      }
-    }
-    if (parts.length == 0) error('too few parts');
-    bool atEnd = (partStart == end);
-    bool isLastWildcard = (parts.last == -1);
-    if (atEnd && !isLastWildcard) {
-      error('expected a part after last `:`', end);
-    }
-    if (!atEnd) {
-      try {
-        parts.add(parseHex(partStart, end));
-      } catch (e) {
-        // Failed to parse the last chunk as hex. Try IPv4.
-        try {
-          List<int> last = parseIPv4Address(host.substring(partStart, end));
-          parts.add(last[0] << 8 | last[1]);
-          parts.add(last[2] << 8 | last[3]);
-        } catch (e) {
-          error('invalid end of IPv6 address.', partStart);
-        }
-      }
-    }
-    if (wildcardSeen) {
-      if (parts.length > 7) {
-        error('an address with a wildcard must have less than 7 parts');
-      }
-    } else if (parts.length != 8) {
-      error('an address without a wildcard must contain exactly 8 parts');
-    }
-    List<int> bytes = new Uint8List(16);
-    for (int i = 0, index = 0; i < parts.length; i++) {
-      int value = parts[i];
-      if (value == -1) {
-        int wildCardLength = 9 - parts.length;
-        for (int j = 0; j < wildCardLength; j++) {
-          bytes[index] = 0;
-          bytes[index + 1] = 0;
-          index += 2;
-        }
-      } else {
-        bytes[index] = value >> 8;
-        bytes[index + 1] = value & 0xff;
-        index += 2;
-      }
-    }
-    return bytes;
-  }
-
-  // Frequently used character codes.
-  static const int _SPACE = 0x20;
-  static const int _DOUBLE_QUOTE = 0x22;
-  static const int _NUMBER_SIGN = 0x23;
-  static const int _PERCENT = 0x25;
-  static const int _ASTERISK = 0x2A;
-  static const int _PLUS = 0x2B;
-  static const int _DOT = 0x2E;
-  static const int _SLASH = 0x2F;
-  static const int _ZERO = 0x30;
-  static const int _NINE = 0x39;
-  static const int _COLON = 0x3A;
-  static const int _LESS = 0x3C;
-  static const int _GREATER = 0x3E;
-  static const int _QUESTION = 0x3F;
-  static const int _AT_SIGN = 0x40;
-  static const int _UPPER_CASE_A = 0x41;
-  static const int _UPPER_CASE_F = 0x46;
-  static const int _UPPER_CASE_Z = 0x5A;
-  static const int _LEFT_BRACKET = 0x5B;
-  static const int _BACKSLASH = 0x5C;
-  static const int _RIGHT_BRACKET = 0x5D;
-  static const int _LOWER_CASE_A = 0x61;
-  static const int _LOWER_CASE_F = 0x66;
-  static const int _LOWER_CASE_Z = 0x7A;
-  static const int _BAR = 0x7C;
-
-  static const String _hexDigits = "0123456789ABCDEF";
-
   external static String _uriEncode(List<int> canonicalTable,
                                     String text,
                                     Encoding encoding,
                                     bool spaceToPlus);
 
   /**
    * Convert a byte (2 character hex sequence) in string [s] starting
    * at position [pos] to its ordinal value
    */
   static int _hexCharPairToByte(String s, int pos) {
@@ skipping 501 matching lines @@
       mimeType = "";
     }
     if (mimeType.isEmpty || identical(mimeType, "application/octet-stream")) {
       buffer.write(mimeType);  // Common cases need no escaping.
     } else {
       int slashIndex = _validateMimeType(mimeType);
       if (slashIndex < 0) {
         throw new ArgumentError.value(mimeType, "mimeType",
                                       "Invalid MIME type");
       }
-      buffer.write(Uri._uriEncode(_tokenCharTable,
-                                  mimeType.substring(0, slashIndex),
-                                  UTF8, false));
+      buffer.write(_Uri._uriEncode(_tokenCharTable,
+                                   mimeType.substring(0, slashIndex),
+                                   UTF8, false));
       buffer.write("/");
-      buffer.write(Uri._uriEncode(_tokenCharTable,
-                                  mimeType.substring(slashIndex + 1),
-                                  UTF8, false));
+      buffer.write(_Uri._uriEncode(_tokenCharTable,
+                                   mimeType.substring(slashIndex + 1),
+                                   UTF8, false));
     }
     if (charsetName != null) {
       if (indices != null) {
         indices..add(buffer.length)
                ..add(buffer.length + 8);
       }
       buffer.write(";charset=");
-      buffer.write(Uri._uriEncode(_tokenCharTable, charsetName, UTF8, false));
+      buffer.write(_Uri._uriEncode(_tokenCharTable, charsetName, UTF8, false));
     }
     parameters?.forEach((var key, var value) {
       if (key.isEmpty) {
         throw new ArgumentError.value("", "Parameter names must not be empty");
       }
       if (value.isEmpty) {
         throw new ArgumentError.value("", "Parameter values must not be empty",
                                       'parameters["$key"]');
       }
       if (indices != null) indices.add(buffer.length);
       buffer.write(';');
       // Encode any non-RFC2045-token character and both '%' and '#'.
-      buffer.write(Uri._uriEncode(_tokenCharTable, key, UTF8, false));
+      buffer.write(_Uri._uriEncode(_tokenCharTable, key, UTF8, false));
       if (indices != null) indices.add(buffer.length);
       buffer.write('=');
-      buffer.write(Uri._uriEncode(_tokenCharTable, value, UTF8, false));
+      buffer.write(_Uri._uriEncode(_tokenCharTable, value, UTF8, false));
     });
   }
 
   /**
    * Checks mimeType is valid-ish (`token '/' token`).
    *
    * Returns the index of the slash, or -1 if the mime type is not
    * considered valid.
    *
    * Currently only looks for slashes, all other characters will be
    * percent-encoded as UTF-8 if necessary.
    */
   static int _validateMimeType(String mimeType) {
     int slashIndex = -1;
     for (int i = 0; i < mimeType.length; i++) {
       var char = mimeType.codeUnitAt(i);
-      if (char != Uri._SLASH) continue;
+      if (char != _SLASH) continue;
       if (slashIndex < 0) {
         slashIndex = i;
         continue;
       }
       return -1;
     }
     return slashIndex;
   }
 
   /**
    * Parses a string as a `data` URI.
    *
    * The string must have the format:
    *
    * ```
    * 'data:' (type '/' subtype)? (';' attribute '=' value)* (';base64')? ',' data
    * ````
    *
    * where `type`, `subtype`, `attribute` and `value` are specified in RFC-2045,
-   * and `data` is a sequnce of URI-characters (RFC-2396 `uric`).
+   * and `data` is a sequence of URI-characters (RFC-2396 `uric`).
    *
    * This means that all the characters must be ASCII, but the URI may contain
    * percent-escapes for non-ASCII byte values that need an interpretation
    * to be converted to the corresponding string.
    *
    * Parsing doesn't check the validity of any part, it just checks that the
    * input has the correct structure with the correct sequence of `/`, `;`, `=`
    * and `,` delimiters.
    *
    * Accessing the individual parts may fail later if they turn out to have
-   * content that can't be decoded sucessfully as a string.
+   * content that can't be decoded successfully as a string.
    */
   static UriData parse(String uri) {
-    if (!uri.startsWith("data:")) {
-      throw new FormatException("Does not start with 'data:'", uri, 0);
+    if (uri.length >= 5) {
+      int dataDelta = _startsWithData(uri, 0);

[floitsch, 2016/06/29 23:41:47]

startsWithData should return a bool, and you shouldn't take the substring. (see
my other comments).

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Wouldn't work. I need to distinguish "data:" from "Data:".

+      if (dataDelta == 0) {
+        // Exact match on "data:".
+        return _parse(uri, 5, null);
+      }
+      if (dataDelta == 0x20) {
+        // Starts with a non-normalized "data" scheme containing upper-case
+        // letters. Parse anyway, but throw away the scheme.
+        return _parse(uri.substring(5), 0, null);
+      }
     }
-    return _parse(uri, 5, null);
+    throw new FormatException("Does not start with 'data:'", uri, 0);
   }
 
   /**
    * The [Uri] that this `UriData` is giving access to.
    *
    * Returns a `Uri` with scheme `data` and the remainder of the data URI
    * as path.
    */
   Uri get uri {
     if (_uriCache != null) return _uriCache;
     String path = _text;
     String query = null;
     int colonIndex = _separatorIndices[0];
     int queryIndex = _text.indexOf('?', colonIndex + 1);
     int end = null;
     if (queryIndex >= 0) {
       query = _text.substring(queryIndex + 1);
       end = queryIndex;
     }
     path = _text.substring(colonIndex + 1, end);
     // TODO(lrn): This can generate a URI that isn't path normalized.
     // That's perfectly reasonable - data URIs are not hierarchical,
     // but it may make some consumers stumble.
     // Should we at least do escape normalization?
-    _uriCache = new Uri._internal("data", "", null, null, path, query, null);
+    _uriCache = new _Uri._internal("data", "", null, null, path, query, null);
     return _uriCache;
   }
 
   /**
    * The MIME type of the data URI.
    *
    * A data URI consists of a "media type" followed by data.
    * The media type starts with a MIME type and can be followed by
    * extra parameters.
    *
    * Example:
    *
    *     data:text/plain;charset=utf-8,Hello%20World!
    *
    * This data URI has the media type `text/plain;charset=utf-8`, which is the
    * MIME type `text/plain` with the parameter `charset` with value `utf-8`.
    * See [RFC 2045](https://tools.ietf.org/html/rfc2045) for more detail.
    *
    * If the first part of the data URI is empty, it defaults to `text/plain`.
    */
   String get mimeType {
     int start = _separatorIndices[0] + 1;
     int end = _separatorIndices[1];
     if (start == end) return "text/plain";
-    return Uri._uriDecode(_text, start, end, UTF8, false);
+    return _Uri._uriDecode(_text, start, end, UTF8, false);
   }
 
   /**
    * The charset parameter of the media type.
    *
    * If the parameters of the media type contains a `charset` parameter
    * then this returns its value, otherwise it returns `US-ASCII`,
    * which is the default charset for data URIs.
    */
   String get charset {
     int parameterStart = 1;
     int parameterEnd = _separatorIndices.length - 1;  // The ',' before data.
     if (isBase64) {
       // There is a ";base64" separator, so subtract one for that as well.
       parameterEnd -= 1;
     }
     for (int i = parameterStart; i < parameterEnd; i += 2) {
       var keyStart = _separatorIndices[i] + 1;
       var keyEnd = _separatorIndices[i + 1];
       if (keyEnd == keyStart + 7 && _text.startsWith("charset", keyStart)) {
-        return Uri._uriDecode(_text, keyEnd + 1, _separatorIndices[i + 2],
-                              UTF8, false);
+        return _Uri._uriDecode(_text, keyEnd + 1, _separatorIndices[i + 2],
+                               UTF8, false);
       }
     }
     return "US-ASCII";
   }
 
   /**
    * Whether the data is Base64 encoded or not.
    */
   bool get isBase64 => _separatorIndices.length.isOdd;
 
@@ skipping 37 matching lines @@
       result.setRange(0, length, text.codeUnits, start);
       return result;
     }
     int index = 0;
     for (int i = start; i < text.length; i++) {
       var codeUnit = text.codeUnitAt(i);
       if (codeUnit != percent) {
         result[index++] = codeUnit;
       } else {
         if (i + 2 < text.length) {
-          var digit1 = Uri._parseHexDigit(text.codeUnitAt(i + 1));
-          var digit2 = Uri._parseHexDigit(text.codeUnitAt(i + 2));
+          var digit1 = _Uri._parseHexDigit(text.codeUnitAt(i + 1));
+          var digit2 = _Uri._parseHexDigit(text.codeUnitAt(i + 2));
           if (digit1 >= 0 && digit2 >= 0) {
             int byte = digit1 * 16 + digit2;
             result[index++] = byte;
             i += 2;
             continue;
           }
         }
         throw new FormatException("Invalid percent escape", text, i);
       }
     }
     assert(index == result.length);
     return result;
   }
 
   /**
    * Returns a string created from the content of the data URI.
    *
    * If the content is Base64 encoded, it will be decoded to bytes and then
    * decoded to a string using [encoding].
    * If encoding is omitted, the value of a `charset` parameter is used
-   * if it is recongized by [Encoding.getByName], otherwise it defaults to
+   * if it is recognized by [Encoding.getByName], otherwise it defaults to
    * the [ASCII] encoding, which is the default encoding for data URIs
    * that do not specify an encoding.
    *
    * If the content is not Base64 encoded, it will first have percent-escapes
    * converted to bytes and then the character codes and byte values are
    * decoded using [encoding].
    */
   String contentAsString({Encoding encoding}) {
     if (encoding == null) {
       var charset = this.charset;  // Returns "US-ASCII" if not present.
       encoding = Encoding.getByName(charset);
       if (encoding == null) {
         throw new UnsupportedError("Unknown charset: $charset");
       }
     }
     String text = _text;
     int start = _separatorIndices.last + 1;
     if (isBase64) {
       var converter = BASE64.decoder.fuse(encoding.decoder);
       return converter.convert(text.substring(start));
     }
-    return Uri._uriDecode(text, start, text.length, encoding, false);
+    return _Uri._uriDecode(text, start, text.length, encoding, false);
   }
 
   /**
    * A map representing the parameters of the media type.
    *
    * A data URI may contain parameters between the MIME type and the
    * data. This converts these parameters to a map from parameter name
    * to parameter value.
    * The map only contains parameters that actually occur in the URI.
    * The `charset` parameter has a default value even if it doesn't occur
    * in the URI, which is reflected by the [charset] getter. This means that
    * [charset] may return a value even if `parameters["charset"]` is `null`.
    *
    * If the values contain non-ASCII values or percent escapes, they default
    * to being decoded as UTF-8.
    */
   Map<String, String> get parameters {
     var result = <String, String>{};
     for (int i = 3; i < _separatorIndices.length; i += 2) {
       var start = _separatorIndices[i - 2] + 1;
       var equals = _separatorIndices[i - 1];
       var end = _separatorIndices[i];
-      String key = Uri._uriDecode(_text, start, equals, UTF8, false);
-      String value = Uri._uriDecode(_text,equals + 1, end, UTF8, false);
+      String key = _Uri._uriDecode(_text, start, equals, UTF8, false);
+      String value = _Uri._uriDecode(_text,equals + 1, end, UTF8, false);
       result[key] = value;
     }
     return result;
   }
 
   static UriData _parse(String text, int start, Uri sourceUri) {
     assert(start == 0 || start == 5);
     assert((start == 5) == text.startsWith("data:"));

[floitsch, 2016/06/29 23:41:47]

(start == 5) == text.toLowerCase().startsWith("data:");

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

We *need* to have a lower-case scheme. We promise to do scheme-normalization, so
code will expect .toString() to return something starting with "data:", and the
code is allowed to it against lower-case strings. So, storing "Data:..." is a
complete waste of space because we have to ignore it everywhere and remove it
when doing toString.

 
     /// Character codes.
     const int comma     = 0x2c;
     const int slash     = 0x2f;
     const int semicolon = 0x3b;
     const int equals    = 0x3d;
     List<int> indices = [start - 1];
     int slashIndex = -1;
     var char;
     int i = start;
@@ skipping 54 matching lines @@
     // Encode the string into bytes then generate an ASCII only string
     // by percent encoding selected bytes.
     int byteOr = 0;
     for (int i = 0; i < bytes.length; i++) {
       int byte = bytes[i];
       byteOr |= byte;
       if (byte < 128 &&
           ((canonicalTable[byte >> 4] & (1 << (byte & 0x0f))) != 0)) {
         buffer.writeCharCode(byte);
       } else {
-        buffer.writeCharCode(Uri._PERCENT);
-        buffer.writeCharCode(Uri._hexDigits.codeUnitAt(byte >> 4));
-        buffer.writeCharCode(Uri._hexDigits.codeUnitAt(byte & 0x0f));
+        buffer.writeCharCode(_PERCENT);
+        buffer.writeCharCode(_hexDigits.codeUnitAt(byte >> 4));
+        buffer.writeCharCode(_hexDigits.codeUnitAt(byte & 0x0f));
       }
     }
     if ((byteOr & ~0xFF) != 0) {
       for (int i = 0; i < bytes.length; i++) {
         var byte = bytes[i];
         if (byte < 0 || byte > 255) {
           throw new ArgumentError.value(byte, "non-byte value");
         }
       }
     }
@@ skipping 28 matching lines @@
       0x7fff];  // 0x70 - 0x7f  11111111 11111110
 
   // All non-escape RFC-2396 uric characters.
   //
   //  uric        =  reserved | unreserved | escaped
   //  reserved    =  ";" | "/" | "?" | ":" | "@" | "&" | "=" | "+" | "$" | ","
   //  unreserved  =  alphanum | mark
   //  mark        =  "-" | "_" | "." | "!" | "~" | "*" | "'" | "(" | ")"
   //
   // This is the same characters as in a URI query (which is URI pchar plus '?')
-  static const _uricTable = Uri._queryCharTable;
+  static const _uricTable = _Uri._queryCharTable;
 }
+
+// --------------------------------------------------------------------
+// Constants used to read the scanner result.
+// The indices points into the table returned by [_scanUri] which contains
+// recognized positions in the scanned URI.
+// The `0` index is only used internally.
+
+/// Index of the position of that `:` after a scheme.
+const int _schemeEndIndex     = 1;
+/// Index of the position of the character just before the host name.
+const int _hostStartIndex     = 2;
+/// Index of the position of the `:` before a port value.
+const int _portStartIndex     = 3;
+/// Index of the position of the first character of a path.
+const int _pathStartIndex     = 4;
+/// Index of the position of the `?` before a query.
+const int _queryStartIndex    = 5;
+/// Index of the position of the `#` before a fragment.
+const int _fragmentStartIndex = 6;
+/// Index of a position where the URI was determined to be "non-simple".
+const int _notSimpleIndex     = 7;
+
+// Initial state for scanner.
+const int _uriStart           = 00;
+
+// If scanning of a URI terminates in this state or above,
+// consider the URI non-simple
+const int _nonSimpleEndStates = 14;
+
+// Initial state for scheme validation.
+const int _schemeStart        = 20;
+
+/// Transition tables used to scan a URI to determine its structure.
+///
+/// The tables represent a state machine with output.
+///
+/// To scan the URI, start in the [_uriStart] state, then read each character
+/// of the URI in order, from start to end, and for each character perform a
+/// transition to a new state while writing the current position into the output
+/// buffer at a designated index.
+///
+/// Each state, represented by an integer which is an index into
+/// [_scannerTables], has a set of transitions, one for each character.
+/// The transitions are encoded as a 5-bit integer representing the next state
+/// and a 3-bit index into the output table.
+///
+/// For URI scanning, only characters in the range U+0020 through U+007E are
+/// interesting, all characters outside that range are treated the same.
+/// The tables only contain 96 entries, representing that characters in the
+/// interesting range, plus one more to represent all values outside the range.
+/// The character entries are stored in one `Uint8List` per state, with the
+/// transition for a character at position `character ^ 0x60`,
+/// which maps the range U+0020 .. U+007F into positions 0 .. 95.
+/// All remaining characters are mapped to position 31 (`0x7f ^ 0x60`) which
+/// represents the transition for all remaining characters.
+final List<Uint8List> _scannerTables = _createTables();
+
+// ----------------------------------------------------------------------
+// Code to create the URI scanner table.
+
+/// Creates the tables for [_scannerTables] used by [_scanUri].
+///
+/// See [_scannerTables] for the generated format.
+///
+/// The concrete tables are chosen as a trade-off between the number of states
+/// needed and the precision of the result.
+/// This allows definitely recognizing the general structure of the URI
+/// (presence and location of scheme, user-info, host, port, path, query and
+/// fragment) while at the same time detecting that some components are not
+/// in canonical form (anything containing a `%`, a host-name containing a
+/// capital letter). Since the scanner doesn't know whether something is a
+/// scheme or a path until it sees `:`, or user-info or host until it sees
+/// a `@`, a second pass is needed to validate the scheme and any user-info
+/// is considered non-canonical by default.
+List<Uint8List> _createTables() {
+  // TODO(lrn): Use a precomputed table.
+
+  // Total number of states for the scanner.
+  const int stateCount         = 22;
+
+  // States used to scan a URI from scratch.
+  const int schemeOrPath       = 01;
+  const int authOrPath         = 02;
+  const int authOrPathSlash    = 03;
+  const int uinfoOrHost0       = 04;
+  const int uinfoOrHost        = 05;
+  const int uinfoOrPort0       = 06;
+  const int uinfoOrPort        = 07;
+  const int ipv6Host           = 08;
+  const int relPathSeg         = 09;
+  const int pathSeg            = 10;
+  const int path               = 11;
+  const int query              = 12;
+  const int fragment           = 13;
+  const int schemeOrPathDot    = 14;
+  const int schemeOrPathDot2   = 15;
+  const int relPathSegDot      = 16;
+  const int relPathSegDot2     = 17;
+  const int pathSegDot         = 18;
+  const int pathSegDot2        = 19;
+
+  // States used to validate a scheme after its end position has been found.
+  const int scheme0            = _schemeStart;
+  const int scheme             = 21;
+
+  // Constants encoding the write-index for the state transition into the top 5
+  // bits of a byte.
+  const int schemeEnd          = _schemeEndIndex     << 5;
+  const int hostStart          = _hostStartIndex     << 5;
+  const int portStart          = _portStartIndex     << 5;
+  const int pathStart          = _pathStartIndex     << 5;
+  const int queryStart         = _queryStartIndex    << 5;
+  const int fragmentStart      = _fragmentStartIndex << 5;
+  const int notSimple          = _notSimpleIndex     << 5;
+
+  /// The `unreserved` characters of RFC 3986.
+  const unreserved =
+      "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz-._~"  ;
+  /// The `sub-delim` characters of RFC 3986.
+  const subDelims = r"!$&'()*+,;=";
+  // The `pchar` characters of RFC 3986: characters that may occur in a path,
+  // excluding escapes.
+  const pchar = "$unreserved$subDelims";
+
+  var tables = new List<Uint8List>.generate(stateCount,
+      (_) => new Uint8List(96));
+
+  // Helper function which initialize the table for [state] with a default
+  // transition and returns the table.
+  Uint8List build(state, defaultTransition) =>
+      tables[state]..fillRange(0, 96, defaultTransition);
+
+  // Helper function which sets the transition for each character in [chars]
+  // to [transition] in the [target] table.
+  // The [chars] string must contain only characters in the U+0020 .. U+007E
+  // range.
+  void setChars(Uint8List target, String chars, int transition) {
+    for (int i = 0; i < chars.length; i++) {
+      var char = chars.codeUnitAt(i);
+      target[char ^ 0x60] = transition;
+    }
+  }
+
+  /// Helper function which sets the transition for all characters in the
+  /// range from `range[0]` to `range[1]` to [transition] in the [target] table.
+  ///
+  /// The [range] must be a two-character string where both characters are in
+  /// the U+0020 .. U+007E range and the former character must have a lower
+  /// code point than the latter.
+  void setRange(Uint8List target, String range, int transition) {
+    for (int i = range.codeUnitAt(0), n = range.codeUnitAt(1); i <= n; i++) {
+      target[i ^ 0x60] = transition;
+    }
+  }
+
+  // Create the transitions for each state.
+  var b;
+
+  // Validate as path, if it is a scheme, we handle it later.
+  b = build(_uriStart, schemeOrPath | notSimple);
+  setChars(b, pchar, schemeOrPath);
+  setChars(b, ".", schemeOrPathDot);
+  setChars(b, ":", authOrPath | schemeEnd);  // Handle later.
+  setChars(b, "/", authOrPathSlash);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(schemeOrPathDot, schemeOrPath | notSimple);
+  setChars(b, pchar, schemeOrPath);
+  setChars(b, ".", schemeOrPathDot2);
+  setChars(b, ':', authOrPath | schemeEnd);
+  setChars(b, "/", pathSeg | notSimple);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(schemeOrPathDot2, schemeOrPath | notSimple);
+  setChars(b, pchar, schemeOrPath);
+  setChars(b, "%", schemeOrPath | notSimple);
+  setChars(b, ':', authOrPath | schemeEnd);
+  setChars(b, "/", relPathSeg);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(schemeOrPath, schemeOrPath | notSimple);
+  setChars(b, pchar, schemeOrPath);
+  setChars(b, ':', authOrPath | schemeEnd);
+  setChars(b, "/", pathSeg);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(authOrPath, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, "/", authOrPathSlash);
+  setChars(b, ".", pathSegDot);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(authOrPathSlash, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, "/", uinfoOrHost0 | hostStart);
+  setChars(b, ".", pathSegDot);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(uinfoOrHost0, uinfoOrHost | notSimple);
+  setChars(b, pchar, uinfoOrHost);
+  setRange(b, "AZ", uinfoOrHost | notSimple);
+  setChars(b, ":", uinfoOrPort0 | portStart);
+  setChars(b, "@", uinfoOrHost0 | hostStart);
+  setChars(b, "[", ipv6Host | notSimple);
+  setChars(b, "/", pathSeg | pathStart);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(uinfoOrHost, uinfoOrHost | notSimple);
+  setChars(b, pchar, uinfoOrHost);
+  setRange(b, "AZ", uinfoOrHost | notSimple);
+  setChars(b, ":", uinfoOrPort0 | portStart);
+  setChars(b, "@", uinfoOrHost0 | hostStart);
+  setChars(b, "/", pathSeg | pathStart);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(uinfoOrPort0, uinfoOrPort | notSimple);
+  setRange(b, "19", uinfoOrPort);
+  setChars(b, "@", uinfoOrHost0 | hostStart);
+  setChars(b, "/", pathSeg | pathStart);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(uinfoOrPort, uinfoOrPort | notSimple);
+  setRange(b, "09", uinfoOrPort);
+  setChars(b, "@", uinfoOrHost0 | hostStart);
+  setChars(b, "/", pathSeg | pathStart);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(ipv6Host, ipv6Host);
+  setChars(b, "]", uinfoOrHost);
+
+  b = build(relPathSeg, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, ".", relPathSegDot);
+  setChars(b, "/", pathSeg | notSimple);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(relPathSegDot, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, ".", relPathSegDot2);
+  setChars(b, "/", pathSeg | notSimple);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(relPathSegDot2, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, "/", relPathSeg);
+  setChars(b, "?", query | queryStart);  // This should be non-simple.
+  setChars(b, "#", fragment | fragmentStart);  // This should be non-simple.
+
+  b = build(pathSeg, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, ".", pathSegDot);
+  setChars(b, "/", pathSeg | notSimple);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(pathSegDot, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, ".", pathSegDot2);
+  setChars(b, "/", pathSeg | notSimple);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(pathSegDot2, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, "/", pathSeg | notSimple);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(path, path | notSimple);
+  setChars(b, pchar, path);
+  setChars(b, "/", pathSeg);
+  setChars(b, "?", query | queryStart);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(query, query | notSimple);
+  setChars(b, pchar, query);
+  setChars(b, "?", query);
+  setChars(b, "#", fragment | fragmentStart);
+
+  b = build(fragment, fragment | notSimple);
+  setChars(b, pchar, fragment);
+  setChars(b, "?", fragment);
+
+  // A separate two-state validator for lower-case scheme names.
+  // Any non-scheme character or upper-case letter is marked as non-simple.
+  b = build(scheme0, scheme | notSimple);
+  setRange(b, "az", scheme);
+
+  b = build(scheme, scheme | notSimple);
+  setRange(b, "az", scheme);
+  setRange(b, "09", scheme);
+  setChars(b, "+-.", scheme);
+
+  return tables;
+}
+
+// --------------------------------------------------------------------
+// Code that uses the URI scanner table.
+
+/// Scan a string using the [_scannerTables] state machine.
+///
+/// Scans [uri] from [start] to [end], startig in state [state] and
+/// writing output into [indices].
+///
+/// Returns the final state.
+int _scan(String uri, int start, int end, int state, List<int> indices) {
+  var tables = _scannerTables;
+  assert(end <= uri.length);
+  for (int i = start; i < end; i++) {
+    var table = tables[state];
+    // Xor with 0x60 to move range 0x20-0x7f into 0x00-0x5f
+    int char = uri.codeUnitAt(i) ^ 0x60;
+    // Use 0x1f (nee 0x7f) to represent all unhandled characters.
+    if (char > 0x5f) char = 0x1f;
+    int transition = table[char];
+    state = transition & 0x1f;
+    indices[transition >> 5] = i;
+  }
+  return state;
+}
+
+/// Scan a URI to determine its structure.
+///
+/// Scans the [uri] from [start] to [end] and returns a list of indices
+/// representing significant positions in the URI. The positions are
+/// sufficient to split the URI into its structural parts.
+///
+/// The returned list of positions have the following values at the given
+/// indices:
+///
+/// - [_schemeEndIndex]: Is `start-1` if the URI has no scheme,
+///   otherwise it is the position of the `:` after the scheme.
+/// - [_hostStartIndex]: Either [start] if the URI has no authority component,
+///   otherwise either the position of the `@` after a user-info part or the
+///   position of the final `/` in the `//` leading the authority.
+/// - [_portStartIndex]: Either [start] if the URI has no authority component, or

[floitsch, 2016/06/29 23:41:47]

long line.

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Reworded and moved after inlining _scanUri.

+///   the position of the `:` leading a port, or the same position as the
+///   start of the path if the URI has no port.
+/// - [_pathStartIndex]: The position of the initial character of the path.
+///   If the path is empty, then the position may be [start] or the position
+///   of the initial `/` in the `//` leading an authority.
+///   If there is a scheme or authority, and the path start is before the
+///   query end or host start, the actual path is empty.
+/// - [_queryStartIndex]: The position of the `?` leading a query if the URI
+///   contains a query, otherwise [end].
+/// - [_fragmentStartIndex]: The position of the `#` leading a fragment if the
+///   URI contains a fragment, otherwise [end].
+/// - [_notSimpleIndex]: Equal to `start - 1` unless the URI is considered
+///   "not simple". This is elaborated below.
+///
+/// The returned positions are limited by the scanners ability to write only
+/// one position per character, and only the current position.
+/// Scanning from left to right, we only know whether something is a scheme
+/// or a path when we see a `:` or `/`, and likewise we only know if the first
+/// `/` is part of the path or is leading an authority component when we see
+/// the next character.
+///
+/// # Simple URIs
+/// A URI is considered "simple" if it is in a normalized form containing no
+/// escapes. This allows us to skip normalization and checking whether escapes
+/// are valid, and to extract components without worrying about unescaping.
+///
+/// The scanner computes a conservative approximation of being "simple".
+/// It rejects any URI with an escape, with a user-info component (mainly
+/// because they are rare and would increase the number of states in the
+/// scanner significantly), with an IPV6 host or with a capital letter in
+/// the scheme or host name.
+/// Further, paths containing `..` or `.` path segments are considered
+/// non-simple except for relative paths (no scheme or authority) starting
+/// with a sequence of "../" segments.
+///
+/// There are some scheme-based normalizations for `file`, `http` and `https`
+/// URIs that have to be handled after scanning, to either consider the URI
+/// non-simple or normalize the string.
+List<int> _scanUri(String uri, int start, int end) {
+  var indices = new List<int>.filled(8, start - 1);
+  // For some positions, pick a better default which will be the correct value
+  // in some cases where the value isn't updated by the scanner.
+  indices[_portStartIndex] = start;
+  indices[_pathStartIndex] = start;
+  indices[_queryStartIndex] = end;
+  indices[_fragmentStartIndex] = end;
+  var state = _scan(uri, start, end, _uriStart, indices);
+  // Some states that should be non-simple, but the URI ended early.
+  // In particular paths that end at a ".." must be normalized to end in "../".
+  if (state >= _nonSimpleEndStates) {
+    indices[_notSimpleIndex] = end;
+  }
+  int schemeEnd = indices[_schemeEndIndex];
+  if (schemeEnd >= start) {
+    // Rescan the scheme part now that we know it's not a path.
+    state = _scan(uri, start, schemeEnd, _schemeStart, indices);
+    if (state == _schemeStart) {
+      // Empty scheme.
+      indices[_notSimpleIndex] = schemeEnd;
+    }
+  }
+  return indices;
+}
+
+class _SimpleUri implements Uri {
+  final String _uri;
+  final int _schemeEnd;
+  final int _hostStart;
+  final int _portStart;
+  final int _pathStart;
+  final int _queryStart;
+  final int _fragmentStart;
+  /// The scheme is often used to distinguish URIs.
+  /// To make comparisons more efficient, we cache the value, and
+  /// canonicalize a few known types.
+  String _schemeCache;
+
+  _SimpleUri(
+      this._uri,
+      this._schemeEnd,
+      this._hostStart,
+      this._portStart,
+      this._pathStart,
+      this._queryStart,
+      this._fragmentStart,
+      this._schemeCache);
+
+  bool get hasScheme => _schemeEnd > 0;
+  bool get hasAuthority => _hostStart > 0;
+  bool get hasUserInfo => _hostStart > _schemeEnd + 4;
+  bool get hasPort => _hostStart > 0 && _portStart + 1 < _pathStart;
+  bool get hasQuery => _queryStart < _fragmentStart;
+  bool get hasFragment => _fragmentStart < _uri.length;
+
+  bool get _isFile => _schemeEnd == 4 && _uri.startsWith("file");
+  bool get _isHttp => _schemeEnd == 4 && _uri.startsWith("http");
+  bool get _isHttps => _schemeEnd == 5 && _uri.startsWith("https");
+  bool get _isPackage => _schemeEnd == 7 && _uri.startsWith("package");
+  bool _isScheme(String scheme) =>
+    _schemeEnd == scheme.length && _uri.startsWith(scheme);
+
+  bool get hasAbsolutePath => _uri.startsWith("/", _pathStart);
+  bool get hasEmptyPath => _pathStart == _queryStart;
+
+  bool get isAbsolute => hasScheme && !hasFragment;
+
+  String get scheme {
+    if (_schemeEnd <= 0) return "";
+    if (_schemeCache != null) return _schemeCache;
+    if (_isHttp) {
+      _schemeCache = "http";
+    } else if (_isHttps) {
+      _schemeCache = "https";
+    } else if (_isFile) {
+      _schemeCache = "file";
+    } else if (_isPackage) {
+      _schemeCache = "package";
+    } else {
+      _schemeCache = _uri.substring(0, _schemeEnd);
+    }
+    return _schemeCache;
+  }
+  String get authority => _hostStart > 0 ?
+      _uri.substring(_schemeEnd + 3, _pathStart) : "";
+  String get userInfo => (_hostStart > _schemeEnd + 3) ?
+      _uri.substring(_schemeEnd + 3, _hostStart - 1) : "";
+  String get host =>
+      _hostStart > 0 ? _uri.substring(_hostStart, _portStart) : "";
+  int get port {
+    if (hasPort) return int.parse(_uri.substring(_portStart + 1, _pathStart));
+    if (_isHttp) return 80;
+    if (_isHttps) return 443;
+    return 0;
+  }
+  String get path =>_uri.substring(_pathStart, _queryStart);
+  String get query => (_queryStart < _fragmentStart) ?
+     _uri.substring(_queryStart + 1, _fragmentStart) : "";
+  String get fragment => (_fragmentStart < _uri.length) ?
+     _uri.substring(_fragmentStart + 1) : "";
+
+  String get origin {
+    // Check original behavior - W3C spec is wonky!
+    bool isHttp = _isHttp;
+    if (_schemeEnd < 0 || _hostStart == _portStart) {
+      throw new StateError("Cannot use origin without a scheme: $this");
+    }
+    if (!isHttp && !_isHttps) {
+      throw new StateError(
+        "Origin is only applicable schemes http and https: $this");
+    }
+    if (_hostStart == _schemeEnd + 3) {
+      return _uri.substring(0, _pathStart);
+    }
+    // Need to drop anon-empty userInfo.
+    return _uri.substring(0, _schemeEnd + 3) +
+           _uri.substring(_hostStart, _pathStart);
+  }
+
+  List<String> get pathSegments {
+    int start = _pathStart;
+    int end = _queryStart;
+    if (_uri.startsWith("/", start)) start++;
+    if (start == end) return const <String>[];
+    List<String> parts = [];
+    for (int i = start; i < end; i++) {
+      var char = _uri.codeUnitAt(i);
+      if (char == _SLASH) {
+        parts.add(_uri.substring(start, i));
+        start = i + 1;
+      }
+    }
+    parts.add(_uri.substring(start, end));
+    return new List<String>.unmodifiable(parts);
+  }
+
+  Map<String, String> get queryParameters {
+    if (!hasQuery) return const <String, String>{};
+    return new UnmodifiableMapView<String, String>(
+        Uri.splitQueryString(query));
+  }
+
+  Map<String, List<String>> get queryParametersAll {
+    if (!hasQuery) return const <String, List<String>>{};
+    Map queryParameterLists = _Uri._splitQueryStringAll(query);
+    for (var key in queryParameterLists.keys) {
+      queryParameterLists[key] =
+          new List<String>.unmodifiable(queryParameterLists[key]);
+    }
+    return new Map<String, List<String>>.unmodifiable(queryParameterLists);
+  }
+
+  bool _isPort(String port) {
+    int portDigitStart = _portStart + 1;
+    return portDigitStart + port.length == _pathStart &&
+           _uri.startsWith(port, portDigitStart);
+  }
+
+  Uri normalizePath() => this;
+
+  Uri removeFragment() {
+    if (!hasFragment) return this;
+    return new _SimpleUri(
+      _uri.substring(0, _fragmentStart),
+      _schemeEnd, _hostStart, _portStart,
+      _pathStart, _queryStart, _fragmentStart, _schemeCache);
+  }
+
+  Uri replace({String scheme,
+               String userInfo,
+               String host,
+               int port,
+               String path,
+               Iterable<String> pathSegments,
+               String query,
+               Map<String, dynamic/*String|Iterable<String>*/> queryParameters,
+               String fragment}) {
+    bool schemeChanged = false;
+    if (scheme != null) {
+      scheme = _Uri._makeScheme(scheme, 0, scheme.length);
+      schemeChanged = !_isScheme(scheme);
+    } else {
+      scheme = this.scheme;
+    }
+    bool isFile = (scheme == "file");
+    if (userInfo != null) {
+      userInfo = _Uri._makeUserInfo(userInfo, 0, userInfo.length);
+    } else if (_hostStart > 0) {
+      userInfo = _uri.substring(_schemeEnd + 3, _hostStart);
+    } else {
+      userInfo = "";
+    }
+    if (port != null) {
+      port = _Uri._makePort(port, scheme);
+    } else {
+      port = this.hasPort ? this.port : null;
+      if (schemeChanged) {
+        // The default port might have changed.
+        port = _Uri._makePort(port, scheme);
+      }
+    }
+    if (host != null) {
+      host = _Uri._makeHost(host, 0, host.length, false);
+    } else if (_hostStart > 0) {
+      host = _uri.substring(_hostStart, _portStart);
+    } else if (userInfo.isNotEmpty || port != null || isFile) {
+      host = "";
+    }
+
+    bool hasAuthority = host != null;
+    if (path != null || pathSegments != null) {
+      path = _Uri._makePath(path, 0, _stringOrNullLength(path), pathSegments,
+                           scheme, hasAuthority);
+    } else {
+      path = _uri.substring(_pathStart, _queryStart);
+      if ((isFile || (hasAuthority && !path.isEmpty)) &&
+          !path.startsWith('/')) {
+        path = "/" + path;
+      }
+    }
+
+    if (query != null || queryParameters != null) {
+      query =
+          _Uri._makeQuery(query, 0, _stringOrNullLength(query), queryParameters);

[floitsch, 2016/06/29 23:41:47]

long line.

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Done.

+    } else if (_queryStart < _fragmentStart) {
+      query = _uri.substring(_queryStart, _fragmentStart);
+    }
+
+    if (fragment != null) {
+      fragment = _Uri._makeFragment(fragment, 0, fragment.length);
+    } else if (_fragmentStart < _uri.length) {
+      fragment = _uri.substring(_fragmentStart);
+    }
+
+    return new _Uri._internal(
+        scheme, userInfo, host, port, path, query, fragment);
+  }
+
+  Uri resolve(String reference) {
+    return resolveUri(Uri.parse(reference));
+  }
+
+  Uri resolveUri(Uri reference) {
+    if (reference is _SimpleUri) {
+      return _simpleMerge(this, reference);
+    }
+    return _toNonSimple().resolveUri(reference);
+  }
+
+  // Merge two simple URIs. This should always result in a prefix of
+  // one concatentated with a suffix of the other, which is again simple.
+  // In a few cases, there might be a need for extra normalization, when
+  // resolving on top of a known scheme.
+  Uri _simpleMerge(_SimpleUri base, _SimpleUri ref) {
+    if (ref.hasScheme) return ref;
+    if (ref.hasAuthority) {
+      if (!base.hasScheme) return ref;
+      bool isSimple = true;
+      if (base._isFile) {
+        isSimple = !ref.hasEmptyPath;
+      } else if (base._isHttp) {
+        isSimple = !ref._isPort("80");
+      } else if (base._isHttps) {
+        isSimple = !ref._isPort("443");
+      }
+      if (isSimple) {
+        var delta = base._schemeEnd + 1;
+        var newUri = base._uri.substring(0, base._schemeEnd + 1) +
+                     ref._uri.substring(ref._schemeEnd + 1);
+        return new _SimpleUri(newUri,
+           base._schemeEnd,
+           ref._hostStart + delta,
+           ref._portStart + delta,
+           ref._pathStart + delta,
+           ref._queryStart + delta,
+           ref._fragmentStart + delta,
+           base._schemeCache);
+      } else {
+        // This will require normalization, so use the _Uri implementation.
+        return _toNonSimple().resolveUri(ref);
+      }
+    }
+    if (ref.hasEmptyPath) {
+      if (ref.hasQuery) {
+        int delta = base._queryStart - ref._queryStart;
+        var newUri = base._uri.substring(0, base._queryStart) +
+                     ref._uri.substring(ref._queryStart);
+        return new _SimpleUri(newUri,
+           base._schemeEnd,
+           base._hostStart,
+           base._portStart,
+           base._pathStart,
+           ref._queryStart + delta,
+           ref._fragmentStart + delta,
+           base._schemeCache);
+      }
+      if (ref.hasFragment) {
+        int delta = base._fragmentStart - ref._fragmentStart;
+        var newUri = base._uri.substring(0, base._fragmentStart) +
+                     ref._uri.substring(ref._fragmentStart);
+        return new _SimpleUri(newUri,
+           base._schemeEnd,
+           base._hostStart,
+           base._portStart,
+           base._pathStart,
+           base._queryStart,
+           ref._fragmentStart + delta,
+           base._schemeCache);
+      }
+      return base.removeFragment();
+    }
+    if (ref.hasAbsolutePath) {
+      var delta = base._pathStart - ref._pathStart;
+      var newUri = base._uri.substring(0, base._pathStart) +
+                   ref._uri.substring(ref._pathStart);
+      return new _SimpleUri(newUri,
+        base._schemeEnd,
+        base._hostStart,
+        base._portStart,
+        base._pathStart,
+        ref._queryStart + delta,
+        ref._fragmentStart + delta,
+        base._schemeCache);
+    }
+    if (base.hasEmptyPath && base.hasAuthority) {
+      // ref has relative non-empty path.
+      var delta = base._pathStart - ref._pathStart + 1;
+      var newUri = "${base._uri.substring(0, base._pathStart)}/"
+                   "${ref._uri.substring(ref._pathStart)}";
+      return new _SimpleUri(newUri,
+        base._schemeEnd,
+        base._hostStart,
+        base._portStart,
+        base._pathStart,
+        ref._queryStart + delta,
+        ref._fragmentStart + delta,
+        base._schemeCache);
+    }
+    // Merge paths.
+    if (base._uri.startsWith("../", base._pathStart)) {
+      // Complex rare case, go slow.
+      return _toNonSimple().resolveUri(ref);
+    }
+
+    // The RFC 3986 algorithm merges the base path without its final segment
+    // (anything after the final "/", or everything if the base path doesn't
+    // contain any "/"), and the reference path.
+    // Then it removes "." and ".." segments using the remove-dot-segment
+    // algorithm.
+    // This code combines the two steps. It is simplified by knowing that
+    // the base path contains no "." or ".." segments, and the reference
+    // path can only contain leading ".." segments.
+
+    String baseUri = base._uri;
+    String refUri = ref._uri;
+    int baseStart = base._pathStart;
+    int baseEnd = base._queryStart;
+    int refStart = ref._pathStart;
+    int refEnd = ref._queryStart;
+    int backCount = 1;
+
+    int slashCount = 0;
+
+    // Count leading ".." segments in reference path.
+    while (refStart + 3 <= refEnd && refUri.startsWith("../", refStart)) {
+      refStart += 3;
+      backCount += 1;
+    }
+
+    // Extra slash inserted between base and reference path parts if
+    // the base path contains any slashes.
+    // (We could use a slash from the base path in most cases, but not if
+    // we remove the entire base path).
+    String insert = "";
+    while (baseEnd > baseStart) {
+      baseEnd--;
+      int char = baseUri.codeUnitAt(baseEnd);
+      if (char == _SLASH) {
+        insert = "/";
+        backCount--;
+        if (backCount == 0) break;
+      }
+    }
+
+    var delta = baseEnd - refStart + insert.length;
+    var newUri = "${base._uri.substring(0, baseEnd)}$insert"
+               "${ref._uri.substring(refStart)}";
+
+    return new _SimpleUri(newUri,
+      base._schemeEnd,
+      base._hostStart,
+      base._portStart,
+      base._pathStart,
+      ref._queryStart + delta,
+      ref._fragmentStart + delta,
+      base._schemeCache);
+  }
+
+  String toFilePath({bool windows}) {
+    if (_schemeEnd >= 0 && !_isFile) {
+      throw new UnsupportedError(
+          "Cannot extract a file path from a $scheme URI");
+    }
+    if (_queryStart < _uri.length) {
+      if (_queryStart < _fragmentStart) {
+        throw new UnsupportedError(
+            "Cannot extract a file path from a URI with a query component");
+      }
+      throw new UnsupportedError(
+          "Cannot extract a file path from a URI with a fragment component");
+    }
+    if (windows == null) windows = _Uri._isWindows;
+    return windows ? _Uri._toWindowsFilePath(this) : _toFilePath();
+  }
+
+  String _toFilePath() {
+    if (_hostStart < _portStart) {
+      // Has authority and non-empty host.
+      throw new UnsupportedError(
+        "Cannot extract a non-Windows file path from a file URI "
+        "with an authority");
+    }
+    return this.path;
+  }
+
+  UriData get data {
+    assert(scheme != "data");
+    return null;
+  }
+
+  int get hashCode => _uri.hashCode;
+
+  bool operator==(Object other) {
+    if (identical(this, other)) return true;
+    if (other is Uri) return _uri == other.toString();
+    return false;
+  }
+
+  Uri _toNonSimple() {
+    return new _Uri._internal(
+      this.scheme,
+      this.userInfo,
+      this.hasAuthority ? this.host: null,
+      this.hasPort ? this.port : null,
+      this.path,
+      this.hasQuery ? this.query : null,
+      this.hasFragment ? this.fragment : null
+    );
+  }
+
+  String toString() => _uri;
+}
+
+/// Checks whether [text] starts with "data:" at position [start].
+///
+/// The text must be long enough to allow reading five characters
+/// from the [start] position.
+///
+/// Returns an integer value which is zero if text starts with all-lowercase
+/// "data:" and 0x20 if the text starts with "data:" that isn't all lower-case.
+/// All other values means the text starts with some other character.
+int _startsWithData(String text, int start) {

[floitsch, 2016/06/29 23:41:47]

This should return a bool. (return delta == 0 || delta == 0x20)

[Lasse Reichstein Nielsen, 2016/06/30 10:27:31]

Again, I actually need to distinguish the two cases.
This may be an overly tricky way to do it, but it gets the job done.
Would it be better if I had named constants for "starts with lower case data"
(0) and "starts with non-lower case data" (0x20)?

[floitsch, 2016/06/30 18:15:22]

I think that would be better. Even just changing the name of the function would
already be an improvement.

+  // Multiply by 3 to avoid a non-colon character making delta be 0x20.
+  int delta = (text.codeUnitAt(start + 4) ^ _COLON) * 3;
+  delta |= text.codeUnitAt(start)     ^ 0x64 /*d*/;
+  delta |= text.codeUnitAt(start + 1) ^ 0x61 /*a*/;
+  delta |= text.codeUnitAt(start + 2) ^ 0x74 /*t*/;
+  delta |= text.codeUnitAt(start + 3) ^ 0x61 /*a*/;
+  return delta;
+}
+
+/// Helper function returning the length of a string, or `0` for `null`.
+int _stringOrNullLength(String s) => (s == null) ? 0 : s.length;