url_canon_stdurl.cc revision c2e0dbddbe15c98d52c4786dac06cb8952a8ae6d 
1// Copyright 2013 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.
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
11namespace url_canon {
12
13namespace {
14
15template<typename CHAR, typename UCHAR>
16bool DoCanonicalizeStandardURL(const URLComponentSource<CHAR>& source,
17                               const url_parse::Parsed& parsed,
18                               CharsetConverter* query_converter,
19                               CanonOutput* output,
20                               url_parse::Parsed* new_parsed) {
21  // Scheme: this will append the colon.
22  bool success = CanonicalizeScheme(source.scheme, parsed.scheme,
23                                    output, &new_parsed->scheme);
24
25  // Authority (username, password, host, port)
26  bool have_authority;
27  if (parsed.username.is_valid() || parsed.password.is_valid() ||
28      parsed.host.is_nonempty() || parsed.port.is_valid()) {
29    have_authority = true;
30
31    // Only write the authority separators when we have a scheme.
32    if (parsed.scheme.is_valid()) {
33      output->push_back('/');
34      output->push_back('/');
35    }
36
37    // User info: the canonicalizer will handle the : and @.
38    success &= CanonicalizeUserInfo(source.username, parsed.username,
39                                    source.password, parsed.password,
40                                    output,
41                                    &new_parsed->username,
42                                    &new_parsed->password);
43
44    success &= CanonicalizeHost(source.host, parsed.host,
45                                output, &new_parsed->host);
46
47    // Host must not be empty for standard URLs.
48    if (!parsed.host.is_nonempty())
49      success = false;
50
51    // Port: the port canonicalizer will handle the colon.
52    int default_port = DefaultPortForScheme(
53        &output->data()[new_parsed->scheme.begin], new_parsed->scheme.len);
54    success &= CanonicalizePort(source.port, parsed.port, default_port,
55                                output, &new_parsed->port);
56  } else {
57    // No authority, clear the components.
58    have_authority = false;
59    new_parsed->host.reset();
60    new_parsed->username.reset();
61    new_parsed->password.reset();
62    new_parsed->port.reset();
63    success = false;  // Standard URLs must have an authority.
64  }
65
66  // Path
67  if (parsed.path.is_valid()) {
68    success &= CanonicalizePath(source.path, parsed.path,
69                                output, &new_parsed->path);
70  } else if (have_authority ||
71             parsed.query.is_valid() || parsed.ref.is_valid()) {
72    // When we have an empty path, make up a path when we have an authority
73    // or something following the path. The only time we allow an empty
74    // output path is when there is nothing else.
75    new_parsed->path = url_parse::Component(output->length(), 1);
76    output->push_back('/');
77  } else {
78    // No path at all
79    new_parsed->path.reset();
80  }
81
82  // Query
83  CanonicalizeQuery(source.query, parsed.query, query_converter,
84                    output, &new_parsed->query);
85
86  // Ref: ignore failure for this, since the page can probably still be loaded.
87  CanonicalizeRef(source.ref, parsed.ref, output, &new_parsed->ref);
88
89  return success;
90}
91
92}  // namespace
93
94
95// Returns the default port for the given canonical scheme, or PORT_UNSPECIFIED
96// if the scheme is unknown.
97int DefaultPortForScheme(const char* scheme, int scheme_len) {
98  int default_port = url_parse::PORT_UNSPECIFIED;
99  switch (scheme_len) {
100    case 4:
101      if (!strncmp(scheme, "http", scheme_len))
102        default_port = 80;
103      break;
104    case 5:
105      if (!strncmp(scheme, "https", scheme_len))
106        default_port = 443;
107      break;
108    case 3:
109      if (!strncmp(scheme, "ftp", scheme_len))
110        default_port = 21;
111      else if (!strncmp(scheme, "wss", scheme_len))
112        default_port = 443;
113      break;
114    case 6:
115      if (!strncmp(scheme, "gopher", scheme_len))
116        default_port = 70;
117      break;
118    case 2:
119      if (!strncmp(scheme, "ws", scheme_len))
120        default_port = 80;
121      break;
122  }
123  return default_port;
124}
125
126bool CanonicalizeStandardURL(const char* spec,
127                             int spec_len,
128                             const url_parse::Parsed& parsed,
129                             CharsetConverter* query_converter,
130                             CanonOutput* output,
131                             url_parse::Parsed* new_parsed) {
132  return DoCanonicalizeStandardURL<char, unsigned char>(
133      URLComponentSource<char>(spec), parsed, query_converter,
134      output, new_parsed);
135}
136
137bool CanonicalizeStandardURL(const char16* spec,
138                             int spec_len,
139                             const url_parse::Parsed& parsed,
140                             CharsetConverter* query_converter,
141                             CanonOutput* output,
142                             url_parse::Parsed* new_parsed) {
143  return DoCanonicalizeStandardURL<char16, char16>(
144      URLComponentSource<char16>(spec), parsed, query_converter,
145      output, new_parsed);
146}
147
148// It might be nice in the future to optimize this so unchanged components don't
149// need to be recanonicalized. This is especially true since the common case for
150// ReplaceComponents is removing things we don't want, like reference fragments
151// and usernames. These cases can become more efficient if we can assume the
152// rest of the URL is OK with these removed (or only the modified parts
153// recanonicalized). This would be much more complex to implement, however.
154//
155// You would also need to update DoReplaceComponents in url_util.cc which
156// relies on this re-checking everything (see the comment there for why).
157bool ReplaceStandardURL(const char* base,
158                        const url_parse::Parsed& base_parsed,
159                        const Replacements<char>& replacements,
160                        CharsetConverter* query_converter,
161                        CanonOutput* output,
162                        url_parse::Parsed* new_parsed) {
163  URLComponentSource<char> source(base);
164  url_parse::Parsed parsed(base_parsed);
165  SetupOverrideComponents(base, replacements, &source, &parsed);
166  return DoCanonicalizeStandardURL<char, unsigned char>(
167      source, parsed, query_converter, output, new_parsed);
168}
169
170// For 16-bit replacements, we turn all the replacements into UTF-8 so the
171// regular codepath can be used.
172bool ReplaceStandardURL(const char* base,
173                        const url_parse::Parsed& base_parsed,
174                        const Replacements<char16>& replacements,
175                        CharsetConverter* query_converter,
176                        CanonOutput* output,
177                        url_parse::Parsed* new_parsed) {
178  RawCanonOutput<1024> utf8;
179  URLComponentSource<char> source(base);
180  url_parse::Parsed parsed(base_parsed);
181  SetupUTF16OverrideComponents(base, replacements, &utf8, &source, &parsed);
182  return DoCanonicalizeStandardURL<char, unsigned char>(
183      source, parsed, query_converter, output, new_parsed);
184}
185
186}  // namespace url_canon
187