1 // Copyright 2013 The Chromium Authors
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4
5 // Functions to canonicalize "standard" URLs, which are ones that have an
6 // authority section including a host name.
7
8 #include "url/url_canon.h"
9 #include "url/url_canon_internal.h"
10 #include "url/url_constants.h"
11
12 namespace url {
13
14 namespace {
15
16 template <typename CHAR>
DoCanonicalizeStandardURL(const URLComponentSource<CHAR> & source,const Parsed & parsed,SchemeType scheme_type,CharsetConverter * query_converter,CanonOutput * output,Parsed * new_parsed)17 bool DoCanonicalizeStandardURL(const URLComponentSource<CHAR>& source,
18 const Parsed& parsed,
19 SchemeType scheme_type,
20 CharsetConverter* query_converter,
21 CanonOutput* output,
22 Parsed* new_parsed) {
23 DCHECK(!parsed.has_opaque_path);
24
25 // Scheme: this will append the colon.
26 bool success = CanonicalizeScheme(source.scheme, parsed.scheme,
27 output, &new_parsed->scheme);
28
29 bool scheme_supports_user_info =
30 (scheme_type == SCHEME_WITH_HOST_PORT_AND_USER_INFORMATION);
31 bool scheme_supports_ports =
32 (scheme_type == SCHEME_WITH_HOST_PORT_AND_USER_INFORMATION ||
33 scheme_type == SCHEME_WITH_HOST_AND_PORT);
34
35 // Authority (username, password, host, port)
36 bool have_authority;
37 if ((scheme_supports_user_info &&
38 (parsed.username.is_valid() || parsed.password.is_valid())) ||
39 parsed.host.is_nonempty() ||
40 (scheme_supports_ports && parsed.port.is_valid())) {
41 have_authority = true;
42
43 // Only write the authority separators when we have a scheme.
44 if (parsed.scheme.is_valid()) {
45 output->push_back('/');
46 output->push_back('/');
47 }
48
49 // User info: the canonicalizer will handle the : and @.
50 if (scheme_supports_user_info) {
51 success &= CanonicalizeUserInfo(
52 source.username, parsed.username, source.password, parsed.password,
53 output, &new_parsed->username, &new_parsed->password);
54 } else {
55 new_parsed->username.reset();
56 new_parsed->password.reset();
57 }
58
59 success &= CanonicalizeHost(source.host, parsed.host,
60 output, &new_parsed->host);
61
62 // Host must not be empty for standard URLs.
63 if (parsed.host.is_empty())
64 success = false;
65
66 // Port: the port canonicalizer will handle the colon.
67 if (scheme_supports_ports) {
68 int default_port = DefaultPortForScheme(
69 &output->data()[new_parsed->scheme.begin], new_parsed->scheme.len);
70 success &= CanonicalizePort(source.port, parsed.port, default_port,
71 output, &new_parsed->port);
72 } else {
73 new_parsed->port.reset();
74 }
75 } else {
76 // No authority, clear the components.
77 have_authority = false;
78 new_parsed->host.reset();
79 new_parsed->username.reset();
80 new_parsed->password.reset();
81 new_parsed->port.reset();
82 success = false; // Standard URLs must have an authority.
83 }
84
85 // Path
86 if (parsed.path.is_valid()) {
87 success &= CanonicalizePath(source.path, parsed.path,
88 output, &new_parsed->path);
89 } else if (have_authority ||
90 parsed.query.is_valid() || parsed.ref.is_valid()) {
91 // When we have an empty path, make up a path when we have an authority
92 // or something following the path. The only time we allow an empty
93 // output path is when there is nothing else.
94 new_parsed->path = Component(output->length(), 1);
95 output->push_back('/');
96 } else {
97 // No path at all
98 new_parsed->path.reset();
99 }
100
101 // Query
102 CanonicalizeQuery(source.query, parsed.query, query_converter,
103 output, &new_parsed->query);
104
105 // Ref: ignore failure for this, since the page can probably still be loaded.
106 CanonicalizeRef(source.ref, parsed.ref, output, &new_parsed->ref);
107
108 // Carry over the flag for potentially dangling markup:
109 if (parsed.potentially_dangling_markup)
110 new_parsed->potentially_dangling_markup = true;
111
112 return success;
113 }
114
115 } // namespace
116
117 // Returns the default port for the given canonical scheme, or PORT_UNSPECIFIED
118 // if the scheme is unknown.
119 //
120 // Please keep blink::DefaultPortForProtocol and url::DefaultPortForProtocol in
121 // sync.
DefaultPortForScheme(const char * scheme,int scheme_len)122 int DefaultPortForScheme(const char* scheme, int scheme_len) {
123 int default_port = PORT_UNSPECIFIED;
124 switch (scheme_len) {
125 case 4:
126 if (!strncmp(scheme, kHttpScheme, scheme_len))
127 default_port = 80;
128 break;
129 case 5:
130 if (!strncmp(scheme, kHttpsScheme, scheme_len))
131 default_port = 443;
132 break;
133 case 3:
134 if (!strncmp(scheme, kFtpScheme, scheme_len))
135 default_port = 21;
136 else if (!strncmp(scheme, kWssScheme, scheme_len))
137 default_port = 443;
138 break;
139 case 2:
140 if (!strncmp(scheme, kWsScheme, scheme_len))
141 default_port = 80;
142 break;
143 }
144 return default_port;
145 }
146
CanonicalizeStandardURL(const char * spec,const Parsed & parsed,SchemeType scheme_type,CharsetConverter * query_converter,CanonOutput * output,Parsed * new_parsed)147 bool CanonicalizeStandardURL(const char* spec,
148 const Parsed& parsed,
149 SchemeType scheme_type,
150 CharsetConverter* query_converter,
151 CanonOutput* output,
152 Parsed* new_parsed) {
153 return DoCanonicalizeStandardURL(URLComponentSource(spec), parsed,
154 scheme_type, query_converter, output,
155 new_parsed);
156 }
157
CanonicalizeStandardURL(const char16_t * spec,const Parsed & parsed,SchemeType scheme_type,CharsetConverter * query_converter,CanonOutput * output,Parsed * new_parsed)158 bool CanonicalizeStandardURL(const char16_t* spec,
159 const Parsed& parsed,
160 SchemeType scheme_type,
161 CharsetConverter* query_converter,
162 CanonOutput* output,
163 Parsed* new_parsed) {
164 return DoCanonicalizeStandardURL(URLComponentSource(spec), parsed,
165 scheme_type, query_converter, output,
166 new_parsed);
167 }
168
169 // It might be nice in the future to optimize this so unchanged components don't
170 // need to be recanonicalized. This is especially true since the common case for
171 // ReplaceComponents is removing things we don't want, like reference fragments
172 // and usernames. These cases can become more efficient if we can assume the
173 // rest of the URL is OK with these removed (or only the modified parts
174 // recanonicalized). This would be much more complex to implement, however.
175 //
176 // You would also need to update DoReplaceComponents in url_util.cc which
177 // relies on this re-checking everything (see the comment there for why).
ReplaceStandardURL(const char * base,const Parsed & base_parsed,const Replacements<char> & replacements,SchemeType scheme_type,CharsetConverter * query_converter,CanonOutput * output,Parsed * new_parsed)178 bool ReplaceStandardURL(const char* base,
179 const Parsed& base_parsed,
180 const Replacements<char>& replacements,
181 SchemeType scheme_type,
182 CharsetConverter* query_converter,
183 CanonOutput* output,
184 Parsed* new_parsed) {
185 URLComponentSource<char> source(base);
186 Parsed parsed(base_parsed);
187 SetupOverrideComponents(base, replacements, &source, &parsed);
188 return DoCanonicalizeStandardURL(source, parsed, scheme_type, query_converter,
189 output, new_parsed);
190 }
191
192 // For 16-bit replacements, we turn all the replacements into UTF-8 so the
193 // regular code path can be used.
ReplaceStandardURL(const char * base,const Parsed & base_parsed,const Replacements<char16_t> & replacements,SchemeType scheme_type,CharsetConverter * query_converter,CanonOutput * output,Parsed * new_parsed)194 bool ReplaceStandardURL(const char* base,
195 const Parsed& base_parsed,
196 const Replacements<char16_t>& replacements,
197 SchemeType scheme_type,
198 CharsetConverter* query_converter,
199 CanonOutput* output,
200 Parsed* new_parsed) {
201 RawCanonOutput<1024> utf8;
202 URLComponentSource<char> source(base);
203 Parsed parsed(base_parsed);
204 SetupUTF16OverrideComponents(base, replacements, &utf8, &source, &parsed);
205 return DoCanonicalizeStandardURL(source, parsed, scheme_type, query_converter,
206 output, new_parsed);
207 }
208
209 } // namespace url
210