url/url_util.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 #include "url/url_util.h"
   6
   7 #include <string.h>
   8 #include <vector>
   9
  10 #include "base/debug/leak_annotations.h"
  11 #include "base/logging.h"
  12 #include "base/strings/string_util.h"
  13 #include "url/url_canon_internal.h"
  14 #include "url/url_file.h"
  15 #include "url/url_util_internal.h"
  16
  17 namespace url {
  18
  19 namespace {
  20
  21 const int kNumStandardURLSchemes = 8;
  22 const char* kStandardURLSchemes[kNumStandardURLSchemes] = {
  23   kHttpScheme,
  24   kHttpsScheme,
  25   kFileScheme,  // Yes, file urls can have a hostname!
  26   kFtpScheme,
  27   kGopherScheme,
  28   kWsScheme,    // WebSocket.
  29   kWssScheme,   // WebSocket secure.
  30   kFileSystemScheme,
  31 };
  32
  33 // List of the currently installed standard schemes. This list is lazily
  34 // initialized by InitStandardSchemes and is leaked on shutdown to prevent
  35 // any destructors from being called that will slow us down or cause problems.
  36 std::vector<const char*>* standard_schemes = NULL;
  37
  38 // See the LockStandardSchemes declaration in the header.
  39 bool standard_schemes_locked = false;
  40
  41 // This template converts a given character type to the corresponding
  42 // StringPiece type.
  43 template<typename CHAR> struct CharToStringPiece {
  44 };
  45 template<> struct CharToStringPiece<char> {
  46   typedef base::StringPiece Piece;
  47 };
  48 template<> struct CharToStringPiece<base::char16> {
  49   typedef base::StringPiece16 Piece;
  50 };
  51
  52 // Ensures that the standard_schemes list is initialized, does nothing if it
  53 // already has values.
  54 void InitStandardSchemes() {
  55   if (standard_schemes)
  56     return;
  57   standard_schemes = new std::vector<const char*>;
  58   for (int i = 0; i < kNumStandardURLSchemes; i++)
  59     standard_schemes->push_back(kStandardURLSchemes[i]);
  60 }
  61
  62 // Given a string and a range inside the string, compares it to the given
  63 // lower-case |compare_to| buffer.
  64 template<typename CHAR>
  65 inline bool DoCompareSchemeComponent(const CHAR* spec,
  66                                      const Component& component,
  67                                      const char* compare_to) {
  68   if (!component.is_nonempty())
  69     return compare_to[0] == 0;  // When component is empty, match empty scheme.
  70   return base::LowerCaseEqualsASCII(
  71       typename CharToStringPiece<CHAR>::Piece(
  72           &spec[component.begin], component.len),
  73       compare_to);
  74 }
  75
  76 // Returns true if the given scheme identified by |scheme| within |spec| is one
  77 // of the registered "standard" schemes.
  78 template<typename CHAR>
  79 bool DoIsStandard(const CHAR* spec, const Component& scheme) {
  80   if (!scheme.is_nonempty())
  81     return false;  // Empty or invalid schemes are non-standard.
  82
  83   InitStandardSchemes();
  84   for (size_t i = 0; i < standard_schemes->size(); i++) {
  85     if (base::LowerCaseEqualsASCII(
  86             typename CharToStringPiece<CHAR>::Piece(
  87                 &spec[scheme.begin], scheme.len),
  88             standard_schemes->at(i)))
  89       return true;
  90   }
  91   return false;
  92 }
  93
  94 template<typename CHAR>
  95 bool DoFindAndCompareScheme(const CHAR* str,
  96                             int str_len,
  97                             const char* compare,
  98                             Component* found_scheme) {
  99   // Before extracting scheme, canonicalize the URL to remove any whitespace.
 100   // This matches the canonicalization done in DoCanonicalize function.
 101   RawCanonOutputT<CHAR> whitespace_buffer;
 102   int spec_len;
 103   const CHAR* spec = RemoveURLWhitespace(str, str_len,
 104                                          &whitespace_buffer, &spec_len);
 105
 106   Component our_scheme;
 107   if (!ExtractScheme(spec, spec_len, &our_scheme)) {
 108     // No scheme.
 109     if (found_scheme)
 110       *found_scheme = Component();
 111     return false;
 112   }
 113   if (found_scheme)
 114     *found_scheme = our_scheme;
 115   return DoCompareSchemeComponent(spec, our_scheme, compare);
 116 }
 117
 118 template<typename CHAR>
 119 bool DoCanonicalize(const CHAR* in_spec,
 120                     int in_spec_len,
 121                     bool trim_path_end,
 122                     CharsetConverter* charset_converter,
 123                     CanonOutput* output,
 124                     Parsed* output_parsed) {
 125   // Remove any whitespace from the middle of the relative URL, possibly
 126   // copying to the new buffer.
 127   RawCanonOutputT<CHAR> whitespace_buffer;
 128   int spec_len;
 129   const CHAR* spec = RemoveURLWhitespace(in_spec, in_spec_len,
 130                                          &whitespace_buffer, &spec_len);
 131
 132   Parsed parsed_input;
 133 #ifdef WIN32
 134   // For Windows, we allow things that look like absolute Windows paths to be
 135   // fixed up magically to file URLs. This is done for IE compatability. For
 136   // example, this will change "c:/foo" into a file URL rather than treating
 137   // it as a URL with the protocol "c". It also works for UNC ("\\foo\bar.txt").
 138   // There is similar logic in url_canon_relative.cc for
 139   //
 140   // For Max & Unix, we don't do this (the equivalent would be "/foo/bar" which
 141   // has no meaning as an absolute path name. This is because browsers on Mac
 142   // & Unix don't generally do this, so there is no compatibility reason for
 143   // doing so.
 144   if (DoesBeginUNCPath(spec, 0, spec_len, false) ||
 145       DoesBeginWindowsDriveSpec(spec, 0, spec_len)) {
 146     ParseFileURL(spec, spec_len, &parsed_input);
 147     return CanonicalizeFileURL(spec, spec_len, parsed_input, charset_converter,
 148                                output, output_parsed);
 149   }
 150 #endif
 151
 152   Component scheme;
 153   if (!ExtractScheme(spec, spec_len, &scheme))
 154     return false;
 155
 156   // This is the parsed version of the input URL, we have to canonicalize it
 157   // before storing it in our object.
 158   bool success;
 159   if (DoCompareSchemeComponent(spec, scheme, url::kFileScheme)) {
 160     // File URLs are special.
 161     ParseFileURL(spec, spec_len, &parsed_input);
 162     success = CanonicalizeFileURL(spec, spec_len, parsed_input,
 163                                   charset_converter, output, output_parsed);
 164   } else if (DoCompareSchemeComponent(spec, scheme, url::kFileSystemScheme)) {
 165     // Filesystem URLs are special.
 166     ParseFileSystemURL(spec, spec_len, &parsed_input);
 167     success = CanonicalizeFileSystemURL(spec, spec_len, parsed_input,
 168                                         charset_converter, output,
 169                                         output_parsed);
 170
 171   } else if (DoIsStandard(spec, scheme)) {
 172     // All "normal" URLs.
 173     ParseStandardURL(spec, spec_len, &parsed_input);
 174     success = CanonicalizeStandardURL(spec, spec_len, parsed_input,
 175                                       charset_converter, output, output_parsed);
 176
 177   } else if (DoCompareSchemeComponent(spec, scheme, url::kMailToScheme)) {
 178     // Mailto are treated like a standard url with only a scheme, path, query
 179     ParseMailtoURL(spec, spec_len, &parsed_input);
 180     success = CanonicalizeMailtoURL(spec, spec_len, parsed_input, output,
 181                                     output_parsed);
 182
 183   } else {
 184     // "Weird" URLs like data: and javascript:
 185     ParsePathURL(spec, spec_len, trim_path_end, &parsed_input);
 186     success = CanonicalizePathURL(spec, spec_len, parsed_input, output,
 187                                   output_parsed);
 188   }
 189   return success;
 190 }
 191
 192 template<typename CHAR>
 193 bool DoResolveRelative(const char* base_spec,
 194                        int base_spec_len,
 195                        const Parsed& base_parsed,
 196                        const CHAR* in_relative,
 197                        int in_relative_length,
 198                        CharsetConverter* charset_converter,
 199                        CanonOutput* output,
 200                        Parsed* output_parsed) {
 201   // Remove any whitespace from the middle of the relative URL, possibly
 202   // copying to the new buffer.
 203   RawCanonOutputT<CHAR> whitespace_buffer;
 204   int relative_length;
 205   const CHAR* relative = RemoveURLWhitespace(in_relative, in_relative_length,
 206                                              &whitespace_buffer,
 207                                              &relative_length);
 208   bool base_is_authority_based = false;
 209   bool base_is_hierarchical = false;
 210   if (base_spec &&
 211       base_parsed.scheme.is_nonempty()) {
 212     int after_scheme = base_parsed.scheme.end() + 1;  // Skip past the colon.
 213     int num_slashes = CountConsecutiveSlashes(base_spec, after_scheme,
 214                                               base_spec_len);
 215     base_is_authority_based = num_slashes > 1;
 216     base_is_hierarchical = num_slashes > 0;
 217   }
 218
 219   bool standard_base_scheme =
 220       base_parsed.scheme.is_nonempty() &&
 221       DoIsStandard(base_spec, base_parsed.scheme);
 222
 223   bool is_relative;
 224   Component relative_component;
 225   if (!IsRelativeURL(base_spec, base_parsed, relative, relative_length,
 226                      (base_is_hierarchical || standard_base_scheme),
 227                      &is_relative, &relative_component)) {
 228     // Error resolving.
 229     return false;
 230   }
 231
 232   // Pretend for a moment that |base_spec| is a standard URL. Normally
 233   // non-standard URLs are treated as PathURLs, but if the base has an
 234   // authority we would like to preserve it.
 235   if (is_relative && base_is_authority_based && !standard_base_scheme) {
 236     Parsed base_parsed_authority;
 237     ParseStandardURL(base_spec, base_spec_len, &base_parsed_authority);
 238     if (base_parsed_authority.host.is_nonempty()) {
 239       RawCanonOutputT<char> temporary_output;
 240       bool did_resolve_succeed =
 241           ResolveRelativeURL(base_spec, base_parsed_authority, false, relative,
 242                              relative_component, charset_converter,
 243                              &temporary_output, output_parsed);
 244       // The output_parsed is incorrect at this point (because it was built
 245       // based on base_parsed_authority instead of base_parsed) and needs to be
 246       // re-created.
 247       DoCanonicalize(temporary_output.data(), temporary_output.length(), true,
 248                      charset_converter, output, output_parsed);
 249       return did_resolve_succeed;
 250     }
 251   } else if (is_relative) {
 252     // Relative, resolve and canonicalize.
 253     bool file_base_scheme = base_parsed.scheme.is_nonempty() &&
 254         DoCompareSchemeComponent(base_spec, base_parsed.scheme, kFileScheme);
 255     return ResolveRelativeURL(base_spec, base_parsed, file_base_scheme, relative,
 256                               relative_component, charset_converter, output,
 257                               output_parsed);
 258   }
 259
 260   // Not relative, canonicalize the input.
 261   return DoCanonicalize(relative, relative_length, true, charset_converter,
 262                         output, output_parsed);
 263 }
 264
 265 template<typename CHAR>
 266 bool DoReplaceComponents(const char* spec,
 267                          int spec_len,
 268                          const Parsed& parsed,
 269                          const Replacements<CHAR>& replacements,
 270                          CharsetConverter* charset_converter,
 271                          CanonOutput* output,
 272                          Parsed* out_parsed) {
 273   // If the scheme is overridden, just do a simple string substitution and
 274   // reparse the whole thing. There are lots of edge cases that we really don't
 275   // want to deal with. Like what happens if I replace "http://e:8080/foo"
 276   // with a file. Does it become "file:///E:/8080/foo" where the port number
 277   // becomes part of the path? Parsing that string as a file URL says "yes"
 278   // but almost no sane rule for dealing with the components individually would
 279   // come up with that.
 280   //
 281   // Why allow these crazy cases at all? Programatically, there is almost no
 282   // case for replacing the scheme. The most common case for hitting this is
 283   // in JS when building up a URL using the location object. In this case, the
 284   // JS code expects the string substitution behavior:
 285   //   http://www.w3.org/TR/2008/WD-html5-20080610/structured.html#common3
 286   if (replacements.IsSchemeOverridden()) {
 287     // Canonicalize the new scheme so it is 8-bit and can be concatenated with
 288     // the existing spec.
 289     RawCanonOutput<128> scheme_replaced;
 290     Component scheme_replaced_parsed;
 291     CanonicalizeScheme(replacements.sources().scheme,
 292                        replacements.components().scheme,
 293                        &scheme_replaced, &scheme_replaced_parsed);
 294
 295     // We can assume that the input is canonicalized, which means it always has
 296     // a colon after the scheme (or where the scheme would be).
 297     int spec_after_colon = parsed.scheme.is_valid() ? parsed.scheme.end() + 1
 298                                                     : 1;
 299     if (spec_len - spec_after_colon > 0) {
 300       scheme_replaced.Append(&spec[spec_after_colon],
 301                              spec_len - spec_after_colon);
 302     }
 303
 304     // We now need to completely re-parse the resulting string since its meaning
 305     // may have changed with the different scheme.
 306     RawCanonOutput<128> recanonicalized;
 307     Parsed recanonicalized_parsed;
 308     DoCanonicalize(scheme_replaced.data(), scheme_replaced.length(), true,
 309                    charset_converter,
 310                    &recanonicalized, &recanonicalized_parsed);
 311
 312     // Recurse using the version with the scheme already replaced. This will now
 313     // use the replacement rules for the new scheme.
 314     //
 315     // Warning: this code assumes that ReplaceComponents will re-check all
 316     // components for validity. This is because we can't fail if DoCanonicalize
 317     // failed above since theoretically the thing making it fail could be
 318     // getting replaced here. If ReplaceComponents didn't re-check everything,
 319     // we wouldn't know if something *not* getting replaced is a problem.
 320     // If the scheme-specific replacers are made more intelligent so they don't
 321     // re-check everything, we should instead recanonicalize the whole thing
 322     // after this call to check validity (this assumes replacing the scheme is
 323     // much much less common than other types of replacements, like clearing the
 324     // ref).
 325     Replacements<CHAR> replacements_no_scheme = replacements;
 326     replacements_no_scheme.SetScheme(NULL, Component());
 327     return DoReplaceComponents(recanonicalized.data(), recanonicalized.length(),
 328                                recanonicalized_parsed, replacements_no_scheme,
 329                                charset_converter, output, out_parsed);
 330   }
 331
 332   // If we get here, then we know the scheme doesn't need to be replaced, so can
 333   // just key off the scheme in the spec to know how to do the replacements.
 334   if (DoCompareSchemeComponent(spec, parsed.scheme, url::kFileScheme)) {
 335     return ReplaceFileURL(spec, parsed, replacements, charset_converter, output,
 336                           out_parsed);
 337   }
 338   if (DoCompareSchemeComponent(spec, parsed.scheme, url::kFileSystemScheme)) {
 339     return ReplaceFileSystemURL(spec, parsed, replacements, charset_converter,
 340                                 output, out_parsed);
 341   }
 342   if (DoIsStandard(spec, parsed.scheme)) {
 343     return ReplaceStandardURL(spec, parsed, replacements, charset_converter,
 344                               output, out_parsed);
 345   }
 346   if (DoCompareSchemeComponent(spec, parsed.scheme, url::kMailToScheme)) {
 347     return ReplaceMailtoURL(spec, parsed, replacements, output, out_parsed);
 348   }
 349
 350   // Default is a path URL.
 351   return ReplacePathURL(spec, parsed, replacements, output, out_parsed);
 352 }
 353
 354 }  // namespace
 355
 356 void Initialize() {
 357   InitStandardSchemes();
 358 }
 359
 360 void Shutdown() {
 361   if (standard_schemes) {
 362     delete standard_schemes;
 363     standard_schemes = NULL;
 364   }
 365 }
 366
 367 void AddStandardScheme(const char* new_scheme) {
 368   // If this assert triggers, it means you've called AddStandardScheme after
 369   // LockStandardSchemes have been called (see the header file for
 370   // LockStandardSchemes for more).
 371   //
 372   // This normally means you're trying to set up a new standard scheme too late
 373   // in your application's init process. Locate where your app does this
 374   // initialization and calls LockStandardScheme, and add your new standard
 375   // scheme there.
 376   DCHECK(!standard_schemes_locked) <<
 377       "Trying to add a standard scheme after the list has been locked.";
 378
 379   size_t scheme_len = strlen(new_scheme);
 380   if (scheme_len == 0)
 381     return;
 382
 383   // Dulicate the scheme into a new buffer and add it to the list of standard
 384   // schemes. This pointer will be leaked on shutdown.
 385   char* dup_scheme = new char[scheme_len + 1];
 386   ANNOTATE_LEAKING_OBJECT_PTR(dup_scheme);
 387   memcpy(dup_scheme, new_scheme, scheme_len + 1);
 388
 389   InitStandardSchemes();
 390   standard_schemes->push_back(dup_scheme);
 391 }
 392
 393 void LockStandardSchemes() {
 394   standard_schemes_locked = true;
 395 }
 396
 397 bool IsStandard(const char* spec, const Component& scheme) {
 398   return DoIsStandard(spec, scheme);
 399 }
 400
 401 bool IsStandard(const base::char16* spec, const Component& scheme) {
 402   return DoIsStandard(spec, scheme);
 403 }
 404
 405 bool FindAndCompareScheme(const char* str,
 406                           int str_len,
 407                           const char* compare,
 408                           Component* found_scheme) {
 409   return DoFindAndCompareScheme(str, str_len, compare, found_scheme);
 410 }
 411
 412 bool FindAndCompareScheme(const base::char16* str,
 413                           int str_len,
 414                           const char* compare,
 415                           Component* found_scheme) {
 416   return DoFindAndCompareScheme(str, str_len, compare, found_scheme);
 417 }
 418
 419 bool Canonicalize(const char* spec,
 420                   int spec_len,
 421                   bool trim_path_end,
 422                   CharsetConverter* charset_converter,
 423                   CanonOutput* output,
 424                   Parsed* output_parsed) {
 425   return DoCanonicalize(spec, spec_len, trim_path_end, charset_converter,
 426                         output, output_parsed);
 427 }
 428
 429 bool Canonicalize(const base::char16* spec,
 430                   int spec_len,
 431                   bool trim_path_end,
 432                   CharsetConverter* charset_converter,
 433                   CanonOutput* output,
 434                   Parsed* output_parsed) {
 435   return DoCanonicalize(spec, spec_len, trim_path_end, charset_converter,
 436                         output, output_parsed);
 437 }
 438
 439 bool ResolveRelative(const char* base_spec,
 440                      int base_spec_len,
 441                      const Parsed& base_parsed,
 442                      const char* relative,
 443                      int relative_length,
 444                      CharsetConverter* charset_converter,
 445                      CanonOutput* output,
 446                      Parsed* output_parsed) {
 447   return DoResolveRelative(base_spec, base_spec_len, base_parsed,
 448                            relative, relative_length,
 449                            charset_converter, output, output_parsed);
 450 }
 451
 452 bool ResolveRelative(const char* base_spec,
 453                      int base_spec_len,
 454                      const Parsed& base_parsed,
 455                      const base::char16* relative,
 456                      int relative_length,
 457                      CharsetConverter* charset_converter,
 458                      CanonOutput* output,
 459                      Parsed* output_parsed) {
 460   return DoResolveRelative(base_spec, base_spec_len, base_parsed,
 461                            relative, relative_length,
 462                            charset_converter, output, output_parsed);
 463 }
 464
 465 bool ReplaceComponents(const char* spec,
 466                        int spec_len,
 467                        const Parsed& parsed,
 468                        const Replacements<char>& replacements,
 469                        CharsetConverter* charset_converter,
 470                        CanonOutput* output,
 471                        Parsed* out_parsed) {
 472   return DoReplaceComponents(spec, spec_len, parsed, replacements,
 473                              charset_converter, output, out_parsed);
 474 }
 475
 476 bool ReplaceComponents(const char* spec,
 477                        int spec_len,
 478                        const Parsed& parsed,
 479                        const Replacements<base::char16>& replacements,
 480                        CharsetConverter* charset_converter,
 481                        CanonOutput* output,
 482                        Parsed* out_parsed) {
 483   return DoReplaceComponents(spec, spec_len, parsed, replacements,
 484                              charset_converter, output, out_parsed);
 485 }
 486
 487 void DecodeURLEscapeSequences(const char* input,
 488                               int length,
 489                               CanonOutputW* output) {
 490   RawCanonOutputT<char> unescaped_chars;
 491   for (int i = 0; i < length; i++) {
 492     if (input[i] == '%') {
 493       unsigned char ch;
 494       if (DecodeEscaped(input, &i, length, &ch)) {
 495         unescaped_chars.push_back(ch);
 496       } else {
 497         // Invalid escape sequence, copy the percent literal.
 498         unescaped_chars.push_back('%');
 499       }
 500     } else {
 501       // Regular non-escaped 8-bit character.
 502       unescaped_chars.push_back(input[i]);
 503     }
 504   }
 505
 506   // Convert that 8-bit to UTF-16. It's not clear IE does this at all to
 507   // JavaScript URLs, but Firefox and Safari do.
 508   for (int i = 0; i < unescaped_chars.length(); i++) {
 509     unsigned char uch = static_cast<unsigned char>(unescaped_chars.at(i));
 510     if (uch < 0x80) {
 511       // Non-UTF-8, just append directly
 512       output->push_back(uch);
 513     } else {
 514       // next_ch will point to the last character of the decoded
 515       // character.
 516       int next_character = i;
 517       unsigned code_point;
 518       if (ReadUTFChar(unescaped_chars.data(), &next_character,
 519                       unescaped_chars.length(), &code_point)) {
 520         // Valid UTF-8 character, convert to UTF-16.
 521         AppendUTF16Value(code_point, output);
 522         i = next_character;
 523       } else {
 524         // If there are any sequences that are not valid UTF-8, we keep
 525         // invalid code points and promote to UTF-16. We copy all characters
 526         // from the current position to the end of the identified sequence.
 527         while (i < next_character) {
 528           output->push_back(static_cast<unsigned char>(unescaped_chars.at(i)));
 529           i++;
 530         }
 531         output->push_back(static_cast<unsigned char>(unescaped_chars.at(i)));
 532       }
 533     }
 534   }
 535 }
 536
 537 void EncodeURIComponent(const char* input, int length, CanonOutput* output) {
 538   for (int i = 0; i < length; ++i) {
 539     unsigned char c = static_cast<unsigned char>(input[i]);
 540     if (IsComponentChar(c))
 541       output->push_back(c);
 542     else
 543       AppendEscapedChar(c, output);
 544   }
 545 }
 546
 547 bool CompareSchemeComponent(const char* spec,
 548                             const Component& component,
 549                             const char* compare_to) {
 550   return DoCompareSchemeComponent(spec, component, compare_to);
 551 }
 552
 553 bool CompareSchemeComponent(const base::char16* spec,
 554                             const Component& component,
 555                             const char* compare_to) {
 556   return DoCompareSchemeComponent(spec, component, compare_to);
 557 }
 558
 559 }  // namespace url