url/url_canon_stdurl.cc

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