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/logging.h" 11 #include "url/url_canon_internal.h" 12 #include "url/url_file.h" 13 #include "url/url_util_internal.h" 14 15 namespace url_util { 16 17 const char kFileScheme[] = "file"; 18 const char kFileSystemScheme[] = "filesystem"; 19 const char kMailtoScheme[] = "mailto"; 20 21 namespace { 22 23 // ASCII-specific tolower. The standard library's tolower is locale sensitive, 24 // so we don't want to use it here. 25 template <class Char> inline Char ToLowerASCII(Char c) { 26 return (c >= 'A' && c <= 'Z') ? (c + ('a' - 'A')) : c; 27 } 28 29 // Backend for LowerCaseEqualsASCII. 30 template<typename Iter> 31 inline bool DoLowerCaseEqualsASCII(Iter a_begin, Iter a_end, const char* b) { 32 for (Iter it = a_begin; it != a_end; ++it, ++b) { 33 if (!*b || ToLowerASCII(*it) != *b) 34 return false; 35 } 36 return *b == 0; 37 } 38 39 const int kNumStandardURLSchemes = 8; 40 const char* kStandardURLSchemes[kNumStandardURLSchemes] = { 41 "http", 42 "https", 43 kFileScheme, // Yes, file urls can have a hostname! 44 "ftp", 45 "gopher", 46 "ws", // WebSocket. 47 "wss", // WebSocket secure. 48 kFileSystemScheme, 49 }; 50 51 // List of the currently installed standard schemes. This list is lazily 52 // initialized by InitStandardSchemes and is leaked on shutdown to prevent 53 // any destructors from being called that will slow us down or cause problems. 54 std::vector<const char*>* standard_schemes = NULL; 55 56 // See the LockStandardSchemes declaration in the header. 57 bool standard_schemes_locked = false; 58 59 // Ensures that the standard_schemes list is initialized, does nothing if it 60 // already has values. 61 void InitStandardSchemes() { 62 if (standard_schemes) 63 return; 64 standard_schemes = new std::vector<const char*>; 65 for (int i = 0; i < kNumStandardURLSchemes; i++) 66 standard_schemes->push_back(kStandardURLSchemes[i]); 67 } 68 69 // Given a string and a range inside the string, compares it to the given 70 // lower-case |compare_to| buffer. 71 template<typename CHAR> 72 inline bool DoCompareSchemeComponent(const CHAR* spec, 73 const url_parse::Component& component, 74 const char* compare_to) { 75 if (!component.is_nonempty()) 76 return compare_to[0] == 0; // When component is empty, match empty scheme. 77 return LowerCaseEqualsASCII(&spec[component.begin], 78 &spec[component.end()], 79 compare_to); 80 } 81 82 // Returns true if the given scheme identified by |scheme| within |spec| is one 83 // of the registered "standard" schemes. 84 template<typename CHAR> 85 bool DoIsStandard(const CHAR* spec, const url_parse::Component& scheme) { 86 if (!scheme.is_nonempty()) 87 return false; // Empty or invalid schemes are non-standard. 88 89 InitStandardSchemes(); 90 for (size_t i = 0; i < standard_schemes->size(); i++) { 91 if (LowerCaseEqualsASCII(&spec[scheme.begin], &spec[scheme.end()], 92 standard_schemes->at(i))) 93 return true; 94 } 95 return false; 96 } 97 98 template<typename CHAR> 99 bool DoFindAndCompareScheme(const CHAR* str, 100 int str_len, 101 const char* compare, 102 url_parse::Component* found_scheme) { 103 // Before extracting scheme, canonicalize the URL to remove any whitespace. 104 // This matches the canonicalization done in DoCanonicalize function. 105 url_canon::RawCanonOutputT<CHAR> whitespace_buffer; 106 int spec_len; 107 const CHAR* spec = RemoveURLWhitespace(str, str_len, 108 &whitespace_buffer, &spec_len); 109 110 url_parse::Component our_scheme; 111 if (!url_parse::ExtractScheme(spec, spec_len, &our_scheme)) { 112 // No scheme. 113 if (found_scheme) 114 *found_scheme = url_parse::Component(); 115 return false; 116 } 117 if (found_scheme) 118 *found_scheme = our_scheme; 119 return DoCompareSchemeComponent(spec, our_scheme, compare); 120 } 121 122 template<typename CHAR> 123 bool DoCanonicalize(const CHAR* in_spec, int in_spec_len, 124 url_canon::CharsetConverter* charset_converter, 125 url_canon::CanonOutput* output, 126 url_parse::Parsed* output_parsed) { 127 // Remove any whitespace from the middle of the relative URL, possibly 128 // copying to the new buffer. 129 url_canon::RawCanonOutputT<CHAR> whitespace_buffer; 130 int spec_len; 131 const CHAR* spec = RemoveURLWhitespace(in_spec, in_spec_len, 132 &whitespace_buffer, &spec_len); 133 134 url_parse::Parsed parsed_input; 135 #ifdef WIN32 136 // For Windows, we allow things that look like absolute Windows paths to be 137 // fixed up magically to file URLs. This is done for IE compatability. For 138 // example, this will change "c:/foo" into a file URL rather than treating 139 // it as a URL with the protocol "c". It also works for UNC ("\\foo\bar.txt"). 140 // There is similar logic in url_canon_relative.cc for 141 // 142 // For Max & Unix, we don't do this (the equivalent would be "/foo/bar" which 143 // has no meaning as an absolute path name. This is because browsers on Mac 144 // & Unix don't generally do this, so there is no compatibility reason for 145 // doing so. 146 if (url_parse::DoesBeginUNCPath(spec, 0, spec_len, false) || 147 url_parse::DoesBeginWindowsDriveSpec(spec, 0, spec_len)) { 148 url_parse::ParseFileURL(spec, spec_len, &parsed_input); 149 return url_canon::CanonicalizeFileURL(spec, spec_len, parsed_input, 150 charset_converter, 151 output, output_parsed); 152 } 153 #endif 154 155 url_parse::Component scheme; 156 if (!url_parse::ExtractScheme(spec, spec_len, &scheme)) 157 return false; 158 159 // This is the parsed version of the input URL, we have to canonicalize it 160 // before storing it in our object. 161 bool success; 162 if (DoCompareSchemeComponent(spec, scheme, kFileScheme)) { 163 // File URLs are special. 164 url_parse::ParseFileURL(spec, spec_len, &parsed_input); 165 success = url_canon::CanonicalizeFileURL(spec, spec_len, parsed_input, 166 charset_converter, output, 167 output_parsed); 168 } else if (DoCompareSchemeComponent(spec, scheme, kFileSystemScheme)) { 169 // Filesystem URLs are special. 170 url_parse::ParseFileSystemURL(spec, spec_len, &parsed_input); 171 success = url_canon::CanonicalizeFileSystemURL(spec, spec_len, 172 parsed_input, 173 charset_converter, 174 output, output_parsed); 175 176 } else if (DoIsStandard(spec, scheme)) { 177 // All "normal" URLs. 178 url_parse::ParseStandardURL(spec, spec_len, &parsed_input); 179 success = url_canon::CanonicalizeStandardURL(spec, spec_len, parsed_input, 180 charset_converter, 181 output, output_parsed); 182 183 } else if (DoCompareSchemeComponent(spec, scheme, kMailtoScheme)) { 184 // Mailto are treated like a standard url with only a scheme, path, query 185 url_parse::ParseMailtoURL(spec, spec_len, &parsed_input); 186 success = url_canon::CanonicalizeMailtoURL(spec, spec_len, parsed_input, 187 output, output_parsed); 188 189 } else { 190 // "Weird" URLs like data: and javascript: 191 url_parse::ParsePathURL(spec, spec_len, &parsed_input); 192 success = url_canon::CanonicalizePathURL(spec, spec_len, parsed_input, 193 output, output_parsed); 194 } 195 return success; 196 } 197 198 template<typename CHAR> 199 bool DoResolveRelative(const char* base_spec, 200 int base_spec_len, 201 const url_parse::Parsed& base_parsed, 202 const CHAR* in_relative, 203 int in_relative_length, 204 url_canon::CharsetConverter* charset_converter, 205 url_canon::CanonOutput* output, 206 url_parse::Parsed* output_parsed) { 207 // Remove any whitespace from the middle of the relative URL, possibly 208 // copying to the new buffer. 209 url_canon::RawCanonOutputT<CHAR> whitespace_buffer; 210 int relative_length; 211 const CHAR* relative = RemoveURLWhitespace(in_relative, in_relative_length, 212 &whitespace_buffer, 213 &relative_length); 214 bool base_is_authority_based = false; 215 bool base_is_hierarchical = false; 216 if (base_spec && 217 base_parsed.scheme.is_nonempty()) { 218 int after_scheme = base_parsed.scheme.end() + 1; // Skip past the colon. 219 int num_slashes = url_parse::CountConsecutiveSlashes( 220 base_spec, after_scheme, base_spec_len); 221 base_is_authority_based = num_slashes > 1; 222 base_is_hierarchical = num_slashes > 0; 223 } 224 225 bool standard_base_scheme = 226 base_parsed.scheme.is_nonempty() && 227 DoIsStandard(base_spec, base_parsed.scheme); 228 229 bool is_relative; 230 url_parse::Component relative_component; 231 if (!url_canon::IsRelativeURL(base_spec, base_parsed, 232 relative, relative_length, 233 (base_is_hierarchical || standard_base_scheme), 234 &is_relative, 235 &relative_component)) { 236 // Error resolving. 237 return false; 238 } 239 240 // Pretend for a moment that |base_spec| is a standard URL. Normally 241 // non-standard URLs are treated as PathURLs, but if the base has an 242 // authority we would like to preserve it. 243 if (is_relative && base_is_authority_based && !standard_base_scheme) { 244 url_parse::Parsed base_parsed_authority; 245 ParseStandardURL(base_spec, base_spec_len, &base_parsed_authority); 246 if (base_parsed_authority.host.is_nonempty()) { 247 bool did_resolve_succeed = 248 url_canon::ResolveRelativeURL(base_spec, base_parsed_authority, 249 false, relative, 250 relative_component, charset_converter, 251 output, output_parsed); 252 // The output_parsed is incorrect at this point (because it was built 253 // based on base_parsed_authority instead of base_parsed) and needs to be 254 // re-created. 255 ParsePathURL(output->data(), output->length(), output_parsed); 256 return did_resolve_succeed; 257 } 258 } else if (is_relative) { 259 // Relative, resolve and canonicalize. 260 bool file_base_scheme = base_parsed.scheme.is_nonempty() && 261 DoCompareSchemeComponent(base_spec, base_parsed.scheme, kFileScheme); 262 return url_canon::ResolveRelativeURL(base_spec, base_parsed, 263 file_base_scheme, relative, 264 relative_component, charset_converter, 265 output, output_parsed); 266 } 267 268 // Not relative, canonicalize the input. 269 return DoCanonicalize(relative, relative_length, charset_converter, 270 output, output_parsed); 271 } 272 273 template<typename CHAR> 274 bool DoReplaceComponents(const char* spec, 275 int spec_len, 276 const url_parse::Parsed& parsed, 277 const url_canon::Replacements<CHAR>& replacements, 278 url_canon::CharsetConverter* charset_converter, 279 url_canon::CanonOutput* output, 280 url_parse::Parsed* out_parsed) { 281 // If the scheme is overridden, just do a simple string substitution and 282 // reparse the whole thing. There are lots of edge cases that we really don't 283 // want to deal with. Like what happens if I replace "http://e:8080/foo" 284 // with a file. Does it become "file:///E:/8080/foo" where the port number 285 // becomes part of the path? Parsing that string as a file URL says "yes" 286 // but almost no sane rule for dealing with the components individually would 287 // come up with that. 288 // 289 // Why allow these crazy cases at all? Programatically, there is almost no 290 // case for replacing the scheme. The most common case for hitting this is 291 // in JS when building up a URL using the location object. In this case, the 292 // JS code expects the string substitution behavior: 293 // http://www.w3.org/TR/2008/WD-html5-20080610/structured.html#common3 294 if (replacements.IsSchemeOverridden()) { 295 // Canonicalize the new scheme so it is 8-bit and can be concatenated with 296 // the existing spec. 297 url_canon::RawCanonOutput<128> scheme_replaced; 298 url_parse::Component scheme_replaced_parsed; 299 url_canon::CanonicalizeScheme( 300 replacements.sources().scheme, 301 replacements.components().scheme, 302 &scheme_replaced, &scheme_replaced_parsed); 303 304 // We can assume that the input is canonicalized, which means it always has 305 // a colon after the scheme (or where the scheme would be). 306 int spec_after_colon = parsed.scheme.is_valid() ? parsed.scheme.end() + 1 307 : 1; 308 if (spec_len - spec_after_colon > 0) { 309 scheme_replaced.Append(&spec[spec_after_colon], 310 spec_len - spec_after_colon); 311 } 312 313 // We now need to completely re-parse the resulting string since its meaning 314 // may have changed with the different scheme. 315 url_canon::RawCanonOutput<128> recanonicalized; 316 url_parse::Parsed recanonicalized_parsed; 317 DoCanonicalize(scheme_replaced.data(), scheme_replaced.length(), 318 charset_converter, 319 &recanonicalized, &recanonicalized_parsed); 320 321 // Recurse using the version with the scheme already replaced. This will now 322 // use the replacement rules for the new scheme. 323 // 324 // Warning: this code assumes that ReplaceComponents will re-check all 325 // components for validity. This is because we can't fail if DoCanonicalize 326 // failed above since theoretically the thing making it fail could be 327 // getting replaced here. If ReplaceComponents didn't re-check everything, 328 // we wouldn't know if something *not* getting replaced is a problem. 329 // If the scheme-specific replacers are made more intelligent so they don't 330 // re-check everything, we should instead recanonicalize the whole thing 331 // after this call to check validity (this assumes replacing the scheme is 332 // much much less common than other types of replacements, like clearing the 333 // ref). 334 url_canon::Replacements<CHAR> replacements_no_scheme = replacements; 335 replacements_no_scheme.SetScheme(NULL, url_parse::Component()); 336 return DoReplaceComponents(recanonicalized.data(), recanonicalized.length(), 337 recanonicalized_parsed, replacements_no_scheme, 338 charset_converter, output, out_parsed); 339 } 340 341 // If we get here, then we know the scheme doesn't need to be replaced, so can 342 // just key off the scheme in the spec to know how to do the replacements. 343 if (DoCompareSchemeComponent(spec, parsed.scheme, kFileScheme)) { 344 return url_canon::ReplaceFileURL(spec, parsed, replacements, 345 charset_converter, output, out_parsed); 346 } 347 if (DoCompareSchemeComponent(spec, parsed.scheme, kFileSystemScheme)) { 348 return url_canon::ReplaceFileSystemURL(spec, parsed, replacements, 349 charset_converter, output, 350 out_parsed); 351 } 352 if (DoIsStandard(spec, parsed.scheme)) { 353 return url_canon::ReplaceStandardURL(spec, parsed, replacements, 354 charset_converter, output, out_parsed); 355 } 356 if (DoCompareSchemeComponent(spec, parsed.scheme, kMailtoScheme)) { 357 return url_canon::ReplaceMailtoURL(spec, parsed, replacements, 358 output, out_parsed); 359 } 360 361 // Default is a path URL. 362 return url_canon::ReplacePathURL(spec, parsed, replacements, 363 output, out_parsed); 364 } 365 366 } // namespace 367 368 void Initialize() { 369 InitStandardSchemes(); 370 } 371 372 void Shutdown() { 373 if (standard_schemes) { 374 delete standard_schemes; 375 standard_schemes = NULL; 376 } 377 } 378 379 void AddStandardScheme(const char* new_scheme) { 380 // If this assert triggers, it means you've called AddStandardScheme after 381 // LockStandardSchemes have been called (see the header file for 382 // LockStandardSchemes for more). 383 // 384 // This normally means you're trying to set up a new standard scheme too late 385 // in your application's init process. Locate where your app does this 386 // initialization and calls LockStandardScheme, and add your new standard 387 // scheme there. 388 DCHECK(!standard_schemes_locked) << 389 "Trying to add a standard scheme after the list has been locked."; 390 391 size_t scheme_len = strlen(new_scheme); 392 if (scheme_len == 0) 393 return; 394 395 // Dulicate the scheme into a new buffer and add it to the list of standard 396 // schemes. This pointer will be leaked on shutdown. 397 char* dup_scheme = new char[scheme_len + 1]; 398 memcpy(dup_scheme, new_scheme, scheme_len + 1); 399 400 InitStandardSchemes(); 401 standard_schemes->push_back(dup_scheme); 402 } 403 404 void LockStandardSchemes() { 405 standard_schemes_locked = true; 406 } 407 408 bool IsStandard(const char* spec, const url_parse::Component& scheme) { 409 return DoIsStandard(spec, scheme); 410 } 411 412 bool IsStandard(const base::char16* spec, const url_parse::Component& scheme) { 413 return DoIsStandard(spec, scheme); 414 } 415 416 bool FindAndCompareScheme(const char* str, 417 int str_len, 418 const char* compare, 419 url_parse::Component* found_scheme) { 420 return DoFindAndCompareScheme(str, str_len, compare, found_scheme); 421 } 422 423 bool FindAndCompareScheme(const base::char16* str, 424 int str_len, 425 const char* compare, 426 url_parse::Component* found_scheme) { 427 return DoFindAndCompareScheme(str, str_len, compare, found_scheme); 428 } 429 430 bool Canonicalize(const char* spec, 431 int spec_len, 432 url_canon::CharsetConverter* charset_converter, 433 url_canon::CanonOutput* output, 434 url_parse::Parsed* output_parsed) { 435 return DoCanonicalize(spec, spec_len, charset_converter, 436 output, output_parsed); 437 } 438 439 bool Canonicalize(const base::char16* spec, 440 int spec_len, 441 url_canon::CharsetConverter* charset_converter, 442 url_canon::CanonOutput* output, 443 url_parse::Parsed* output_parsed) { 444 return DoCanonicalize(spec, spec_len, charset_converter, 445 output, output_parsed); 446 } 447 448 bool ResolveRelative(const char* base_spec, 449 int base_spec_len, 450 const url_parse::Parsed& base_parsed, 451 const char* relative, 452 int relative_length, 453 url_canon::CharsetConverter* charset_converter, 454 url_canon::CanonOutput* output, 455 url_parse::Parsed* output_parsed) { 456 return DoResolveRelative(base_spec, base_spec_len, base_parsed, 457 relative, relative_length, 458 charset_converter, output, output_parsed); 459 } 460 461 bool ResolveRelative(const char* base_spec, 462 int base_spec_len, 463 const url_parse::Parsed& base_parsed, 464 const base::char16* relative, 465 int relative_length, 466 url_canon::CharsetConverter* charset_converter, 467 url_canon::CanonOutput* output, 468 url_parse::Parsed* output_parsed) { 469 return DoResolveRelative(base_spec, base_spec_len, base_parsed, 470 relative, relative_length, 471 charset_converter, output, output_parsed); 472 } 473 474 bool ReplaceComponents(const char* spec, 475 int spec_len, 476 const url_parse::Parsed& parsed, 477 const url_canon::Replacements<char>& replacements, 478 url_canon::CharsetConverter* charset_converter, 479 url_canon::CanonOutput* output, 480 url_parse::Parsed* out_parsed) { 481 return DoReplaceComponents(spec, spec_len, parsed, replacements, 482 charset_converter, output, out_parsed); 483 } 484 485 bool ReplaceComponents( 486 const char* spec, 487 int spec_len, 488 const url_parse::Parsed& parsed, 489 const url_canon::Replacements<base::char16>& replacements, 490 url_canon::CharsetConverter* charset_converter, 491 url_canon::CanonOutput* output, 492 url_parse::Parsed* out_parsed) { 493 return DoReplaceComponents(spec, spec_len, parsed, replacements, 494 charset_converter, output, out_parsed); 495 } 496 497 // Front-ends for LowerCaseEqualsASCII. 498 bool LowerCaseEqualsASCII(const char* a_begin, 499 const char* a_end, 500 const char* b) { 501 return DoLowerCaseEqualsASCII(a_begin, a_end, b); 502 } 503 504 bool LowerCaseEqualsASCII(const char* a_begin, 505 const char* a_end, 506 const char* b_begin, 507 const char* b_end) { 508 while (a_begin != a_end && b_begin != b_end && 509 ToLowerASCII(*a_begin) == *b_begin) { 510 a_begin++; 511 b_begin++; 512 } 513 return a_begin == a_end && b_begin == b_end; 514 } 515 516 bool LowerCaseEqualsASCII(const base::char16* a_begin, 517 const base::char16* a_end, 518 const char* b) { 519 return DoLowerCaseEqualsASCII(a_begin, a_end, b); 520 } 521 522 void DecodeURLEscapeSequences(const char* input, int length, 523 url_canon::CanonOutputW* output) { 524 url_canon::RawCanonOutputT<char> unescaped_chars; 525 for (int i = 0; i < length; i++) { 526 if (input[i] == '%') { 527 unsigned char ch; 528 if (url_canon::DecodeEscaped(input, &i, length, &ch)) { 529 unescaped_chars.push_back(ch); 530 } else { 531 // Invalid escape sequence, copy the percent literal. 532 unescaped_chars.push_back('%'); 533 } 534 } else { 535 // Regular non-escaped 8-bit character. 536 unescaped_chars.push_back(input[i]); 537 } 538 } 539 540 // Convert that 8-bit to UTF-16. It's not clear IE does this at all to 541 // JavaScript URLs, but Firefox and Safari do. 542 for (int i = 0; i < unescaped_chars.length(); i++) { 543 unsigned char uch = static_cast<unsigned char>(unescaped_chars.at(i)); 544 if (uch < 0x80) { 545 // Non-UTF-8, just append directly 546 output->push_back(uch); 547 } else { 548 // next_ch will point to the last character of the decoded 549 // character. 550 int next_character = i; 551 unsigned code_point; 552 if (url_canon::ReadUTFChar(unescaped_chars.data(), &next_character, 553 unescaped_chars.length(), &code_point)) { 554 // Valid UTF-8 character, convert to UTF-16. 555 url_canon::AppendUTF16Value(code_point, output); 556 i = next_character; 557 } else { 558 // If there are any sequences that are not valid UTF-8, we keep 559 // invalid code points and promote to UTF-16. We copy all characters 560 // from the current position to the end of the identified sequence. 561 while (i < next_character) { 562 output->push_back(static_cast<unsigned char>(unescaped_chars.at(i))); 563 i++; 564 } 565 output->push_back(static_cast<unsigned char>(unescaped_chars.at(i))); 566 } 567 } 568 } 569 } 570 571 void EncodeURIComponent(const char* input, int length, 572 url_canon::CanonOutput* output) { 573 for (int i = 0; i < length; ++i) { 574 unsigned char c = static_cast<unsigned char>(input[i]); 575 if (url_canon::IsComponentChar(c)) 576 output->push_back(c); 577 else 578 AppendEscapedChar(c, output); 579 } 580 } 581 582 bool CompareSchemeComponent(const char* spec, 583 const url_parse::Component& component, 584 const char* compare_to) { 585 return DoCompareSchemeComponent(spec, component, compare_to); 586 } 587 588 bool CompareSchemeComponent(const base::char16* spec, 589 const url_parse::Component& component, 590 const char* compare_to) { 591 return DoCompareSchemeComponent(spec, component, compare_to); 592 } 593 594 } // namespace url_util 595