1 // Copyright 2015 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
5 #ifndef URL_SCHEME_HOST_PORT_H_
6 #define URL_SCHEME_HOST_PORT_H_
12 #include "base/strings/string_piece.h"
13 #include "url/url_export.h"
21 // This class represents a (scheme, host, port) tuple extracted from a URL.
23 // The primary purpose of this class is to represent relevant network-authority
24 // information for a URL. It is _not_ an Origin, as described in RFC 6454. In
25 // particular, it is generally NOT the right thing to use for security
28 // Instead, this class is a mechanism for simplifying URLs with standard schemes
29 // (that is, those which follow the generic syntax of RFC 3986) down to the
30 // uniquely identifying information necessary for network fetches. This makes it
31 // suitable as a cache key for a collection of active connections, for instance.
32 // It may, however, be inappropriate to use as a cache key for persistent
33 // storage associated with a host.
35 // In particular, note that:
37 // * SchemeHostPort can only represent schemes which follow the RFC 3986 syntax
38 // (e.g. those registered with GURL as "standard schemes"). Non-standard
39 // schemes such as "blob", "filesystem", "data", and "javascript" can only be
40 // represented as invalid SchemeHostPort objects.
42 // * For example, the "file" scheme follows the standard syntax, but it is
43 // important to note that the authority portion (host, port) is optional.
44 // URLs without an authority portion will be represented with an empty string
45 // for the host, and a port of 0 (e.g. "file:///etc/hosts" =>
46 // ("file", "", 0)), and URLs with a host-only authority portion will be
47 // represented with a port of 0 (e.g. "file://example.com/etc/hosts" =>
48 // ("file", "example.com", 0)). See Section 3 of RFC 3986 to better understand
51 // * SchemeHostPort has no notion of the Origin concept (RFC 6454), and in
52 // particular, it has no notion of a "unique" Origin. If you need to take
53 // uniqueness into account (and, if you're making security-relevant decisions
54 // then you absolutely do), please use 'url::Origin' instead.
58 // * SchemeHostPort objects are commonly created from GURL objects:
60 // GURL url("https://example.com/");
61 // url::SchemeHostPort tuple(url);
62 // tuple.scheme(); // "https"
63 // tuple.host(); // "example.com"
64 // tuple.port(); // 443
66 // * Objects may also be explicitly created and compared:
68 // url::SchemeHostPort tuple(url::kHttpsScheme, "example.com", 443);
69 // tuple.scheme(); // "https"
70 // tuple.host(); // "example.com"
71 // tuple.port(); // 443
73 // GURL url("https://example.com/");
74 // tuple.Equals(url::SchemeHostPort(url)); // true
75 class URL_EXPORT SchemeHostPort {
77 // Creates an invalid (scheme, host, port) tuple, which represents an invalid
78 // or non-standard URL.
81 // Creates a (scheme, host, port) tuple. |host| must be a canonicalized
82 // A-label (that is, '☃.net' must be provided as 'xn--n3h.net'). |scheme|
83 // must be a standard scheme. |port| must not be 0, unless |scheme| does not
84 // support ports (e.g. 'file'). In that case, |port| must be 0.
86 // Copies the data in |scheme| and |host|.
87 SchemeHostPort(base::StringPiece scheme,
88 base::StringPiece host,
91 // Metadata influencing whether or not the constructor should sanity check
92 // host canonicalization.
93 enum ConstructPolicy { CHECK_CANONICALIZATION, ALREADY_CANONICALIZED };
95 // Creates a (scheme, host, port) tuple without performing sanity checking
96 // that the host and port are canonicalized. This should only be used when
97 // converting between already normalized types, and should NOT be used for
99 SchemeHostPort(std::string scheme,
102 ConstructPolicy policy);
104 // Creates a (scheme, host, port) tuple from |url|, as described at
105 // https://tools.ietf.org/html/rfc6454#section-4
107 // If |url| is invalid or non-standard, the result will be an invalid
108 // SchemeHostPort object.
109 explicit SchemeHostPort(const GURL& url);
111 // Copyable and movable.
112 SchemeHostPort(const SchemeHostPort&) = default;
113 SchemeHostPort& operator=(const SchemeHostPort&) = default;
114 SchemeHostPort(SchemeHostPort&&) = default;
115 SchemeHostPort& operator=(SchemeHostPort&&) = default;
119 // Returns the host component, in URL form. That is all IDN domain names will
120 // be expressed as A-Labels ('☃.net' will be returned as 'xn--n3h.net'), and
121 // and all IPv6 addresses will be enclosed in brackets ("[2001:db8::1]").
122 const std::string& host() const { return host_; }
123 const std::string& scheme() const { return scheme_; }
124 uint16_t port() const { return port_; }
125 bool IsInvalid() const;
127 // Serializes the SchemeHostPort tuple to a canonical form.
129 // While this string form resembles the Origin serialization specified in
130 // Section 6.2 of RFC 6454, it is important to note that invalid
131 // SchemeHostPort tuples serialize to the empty string, rather than being
132 // serialized as a unique Origin.
133 std::string Serialize() const;
135 // Efficiently returns what GURL(Serialize()) would return, without needing to
139 // Two SchemeHostPort objects are "equal" iff their schemes, hosts, and ports
140 // are exact matches.
142 // Note that this comparison is _not_ the same as an origin-based comparison.
143 // In particular, invalid SchemeHostPort objects match each other (and
144 // themselves). Unique origins, on the other hand, would not.
145 bool Equals(const SchemeHostPort& other) const;
147 // Allows SchemeHostPort to be used as a key in STL (for example, a std::set
149 bool operator<(const SchemeHostPort& other) const;
152 std::string SerializeInternal(url::Parsed* parsed) const;
161 #endif // URL_SCHEME_HOST_PORT_H_