1 /* 2 * Copyright (C) 2004, 2007, 2008, 2011, 2012 Apple Inc. All rights reserved. 3 * Copyright (C) 2012 Research In Motion Limited. All rights reserved. 4 * Copyright (C) 2008, 2009, 2011 Google Inc. All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY 16 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 18 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE COMPUTER, INC. OR 19 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 20 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 21 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 22 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY 23 * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 25 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #include "config.h" 29 #include "platform/weborigin/KURL.h" 30 31 #include "platform/weborigin/KnownPorts.h" 32 #include "wtf/StdLibExtras.h" 33 #include "wtf/text/CString.h" 34 #include "wtf/text/StringHash.h" 35 #include "wtf/text/StringUTF8Adaptor.h" 36 #include "wtf/text/TextEncoding.h" 37 #include <algorithm> 38 #include <url/url_util.h> 39 #ifndef NDEBUG 40 #include <stdio.h> 41 #endif 42 43 namespace blink { 44 45 static const int maximumValidPortNumber = 0xFFFE; 46 static const int invalidPortNumber = 0xFFFF; 47 48 static void assertProtocolIsGood(const char* protocol) 49 { 50 #if ENABLE(ASSERT) 51 const char* p = protocol; 52 while (*p) { 53 ASSERT(*p > ' ' && *p < 0x7F && !(*p >= 'A' && *p <= 'Z')); 54 ++p; 55 } 56 #endif 57 } 58 59 // Note: You must ensure that |spec| is a valid canonicalized URL before calling this function. 60 static const char* asURLChar8Subtle(const String& spec) 61 { 62 ASSERT(spec.is8Bit()); 63 // characters8 really return characters in Latin-1, but because we canonicalize 64 // URL strings, we know that everything before the fragment identifier will 65 // actually be ASCII, which means this cast is safe as long as you don't look 66 // at the fragment component. 67 return reinterpret_cast<const char*>(spec.characters8()); 68 } 69 70 // Returns the characters for the given string, or a pointer to a static empty 71 // string if the input string is null. This will always ensure we have a non- 72 // null character pointer since ReplaceComponents has special meaning for null. 73 static const char* charactersOrEmpty(const StringUTF8Adaptor& string) 74 { 75 static const char zero = 0; 76 return string.data() ? string.data() : &zero; 77 } 78 79 static bool isSchemeFirstChar(char c) 80 { 81 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'); 82 } 83 84 static bool isSchemeChar(char c) 85 { 86 return isSchemeFirstChar(c) || (c >= '0' && c <= '9') || c == '.' || c == '-' || c == '+'; 87 } 88 89 static bool isUnicodeEncoding(const WTF::TextEncoding* encoding) 90 { 91 return encoding->encodingForFormSubmission() == UTF8Encoding(); 92 } 93 94 namespace { 95 96 class KURLCharsetConverter FINAL : public url::CharsetConverter { 97 public: 98 // The encoding parameter may be 0, but in this case the object must not be called. 99 explicit KURLCharsetConverter(const WTF::TextEncoding* encoding) 100 : m_encoding(encoding) 101 { 102 } 103 104 virtual void ConvertFromUTF16(const url::UTF16Char* input, int inputLength, url::CanonOutput* output) OVERRIDE 105 { 106 CString encoded = m_encoding->normalizeAndEncode(String(input, inputLength), WTF::URLEncodedEntitiesForUnencodables); 107 output->Append(encoded.data(), static_cast<int>(encoded.length())); 108 } 109 110 private: 111 const WTF::TextEncoding* m_encoding; 112 }; 113 114 } // namespace 115 116 bool isValidProtocol(const String& protocol) 117 { 118 // RFC3986: ALPHA *( ALPHA / DIGIT / "+" / "-" / "." ) 119 if (protocol.isEmpty()) 120 return false; 121 if (!isSchemeFirstChar(protocol[0])) 122 return false; 123 unsigned protocolLength = protocol.length(); 124 for (unsigned i = 1; i < protocolLength; i++) { 125 if (!isSchemeChar(protocol[i])) 126 return false; 127 } 128 return true; 129 } 130 131 String KURL::strippedForUseAsReferrer() const 132 { 133 if (protocolIsAbout() || protocolIs("data") || protocolIs("javascript")) 134 return String(); 135 136 if (m_parsed.username.is_nonempty() || m_parsed.password.is_nonempty() || m_parsed.ref.is_nonempty()) { 137 KURL referrer(*this); 138 referrer.setUser(String()); 139 referrer.setPass(String()); 140 referrer.removeFragmentIdentifier(); 141 return referrer.string(); 142 } 143 return string(); 144 } 145 146 bool KURL::isLocalFile() const 147 { 148 // Including feed here might be a bad idea since drag and drop uses this check 149 // and including feed would allow feeds to potentially let someone's blog 150 // read the contents of the clipboard on a drag, even without a drop. 151 // Likewise with using the FrameLoader::shouldTreatURLAsLocal() function. 152 return protocolIs("file"); 153 } 154 155 bool protocolIsJavaScript(const String& url) 156 { 157 return protocolIs(url, "javascript"); 158 } 159 160 const KURL& blankURL() 161 { 162 DEFINE_STATIC_LOCAL(KURL, staticBlankURL, (ParsedURLString, "about:blank")); 163 return staticBlankURL; 164 } 165 166 bool KURL::isAboutBlankURL() const 167 { 168 return *this == blankURL(); 169 } 170 171 String KURL::elidedString() const 172 { 173 if (string().length() <= 1024) 174 return string(); 175 176 return string().left(511) + "..." + string().right(510); 177 } 178 179 KURL::KURL() 180 : m_isValid(false) 181 , m_protocolIsInHTTPFamily(false) 182 { 183 } 184 185 // Initializes with a string representing an absolute URL. No encoding 186 // information is specified. This generally happens when a KURL is converted 187 // to a string and then converted back. In this case, the URL is already 188 // canonical and in proper escaped form so needs no encoding. We treat it as 189 // UTF-8 just in case. 190 KURL::KURL(ParsedURLStringTag, const String& url) 191 { 192 if (!url.isNull()) 193 init(KURL(), url, 0); 194 else { 195 // WebCore expects us to preserve the nullness of strings when this 196 // constructor is used. In all other cases, it expects a non-null 197 // empty string, which is what init() will create. 198 m_isValid = false; 199 m_protocolIsInHTTPFamily = false; 200 } 201 } 202 203 KURL KURL::createIsolated(ParsedURLStringTag, const String& url) 204 { 205 // FIXME: We should be able to skip this extra copy and created an 206 // isolated KURL more efficiently. 207 return KURL(ParsedURLString, url).copy(); 208 } 209 210 // Constructs a new URL given a base URL and a possibly relative input URL. 211 // This assumes UTF-8 encoding. 212 KURL::KURL(const KURL& base, const String& relative) 213 { 214 init(base, relative, 0); 215 } 216 217 // Constructs a new URL given a base URL and a possibly relative input URL. 218 // Any query portion of the relative URL will be encoded in the given encoding. 219 KURL::KURL(const KURL& base, const String& relative, const WTF::TextEncoding& encoding) 220 { 221 init(base, relative, &encoding.encodingForFormSubmission()); 222 } 223 224 KURL::KURL(const AtomicString& canonicalString, const url::Parsed& parsed, bool isValid) 225 : m_isValid(isValid) 226 , m_protocolIsInHTTPFamily(false) 227 , m_parsed(parsed) 228 , m_string(canonicalString) 229 { 230 initProtocolIsInHTTPFamily(); 231 initInnerURL(); 232 } 233 234 KURL::KURL(WTF::HashTableDeletedValueType) 235 : m_isValid(false) 236 , m_protocolIsInHTTPFamily(false) 237 , m_string(WTF::HashTableDeletedValue) 238 { 239 } 240 241 KURL::KURL(const KURL& other) 242 : m_isValid(other.m_isValid) 243 , m_protocolIsInHTTPFamily(other.m_protocolIsInHTTPFamily) 244 , m_parsed(other.m_parsed) 245 , m_string(other.m_string) 246 { 247 if (other.m_innerURL.get()) 248 m_innerURL = adoptPtr(new KURL(other.m_innerURL->copy())); 249 } 250 251 KURL& KURL::operator=(const KURL& other) 252 { 253 m_isValid = other.m_isValid; 254 m_protocolIsInHTTPFamily = other.m_protocolIsInHTTPFamily; 255 m_parsed = other.m_parsed; 256 m_string = other.m_string; 257 if (other.m_innerURL) 258 m_innerURL = adoptPtr(new KURL(other.m_innerURL->copy())); 259 else 260 m_innerURL.clear(); 261 return *this; 262 } 263 264 #if COMPILER_SUPPORTS(CXX_RVALUE_REFERENCES) 265 KURL::KURL(KURL&& other) 266 : m_isValid(other.m_isValid) 267 , m_protocolIsInHTTPFamily(other.m_protocolIsInHTTPFamily) 268 , m_parsed(other.m_parsed) 269 // FIXME: Instead of explicitly casting to String&& here, we should use std::move, but that requires us to 270 // have a standard library that supports move semantics. 271 , m_string(static_cast<String&&>(other.m_string)) 272 , m_innerURL(other.m_innerURL.release()) 273 { 274 } 275 276 KURL& KURL::operator=(KURL&& other) 277 { 278 m_isValid = other.m_isValid; 279 m_protocolIsInHTTPFamily = other.m_protocolIsInHTTPFamily; 280 m_parsed = other.m_parsed; 281 // FIXME: Instead of explicitly casting to String&& here, we should use std::move, but that requires us to 282 // have a standard library that supports move semantics. 283 m_string = static_cast<String&&>(other.m_string); 284 m_innerURL = other.m_innerURL.release(); 285 return *this; 286 } 287 #endif 288 289 KURL KURL::copy() const 290 { 291 KURL result; 292 result.m_isValid = m_isValid; 293 result.m_protocolIsInHTTPFamily = m_protocolIsInHTTPFamily; 294 result.m_parsed = m_parsed; 295 result.m_string = m_string.isolatedCopy(); 296 if (m_innerURL) 297 result.m_innerURL = adoptPtr(new KURL(m_innerURL->copy())); 298 return result; 299 } 300 301 bool KURL::isNull() const 302 { 303 return m_string.isNull(); 304 } 305 306 bool KURL::isEmpty() const 307 { 308 return m_string.isEmpty(); 309 } 310 311 bool KURL::isValid() const 312 { 313 return m_isValid; 314 } 315 316 bool KURL::hasPort() const 317 { 318 return hostEnd() < pathStart(); 319 } 320 321 bool KURL::protocolIsInHTTPFamily() const 322 { 323 return m_protocolIsInHTTPFamily; 324 } 325 326 bool KURL::hasPath() const 327 { 328 // Note that http://www.google.com/" has a path, the path is "/". This can 329 // return false only for invalid or nonstandard URLs. 330 return m_parsed.path.len >= 0; 331 } 332 333 // We handle "parameters" separated by a semicolon, while KURL.cpp does not, 334 // which can lead to different results in some cases. 335 String KURL::lastPathComponent() const 336 { 337 if (!m_isValid) 338 return stringForInvalidComponent(); 339 ASSERT(!m_string.isNull()); 340 341 // When the output ends in a slash, WebCore has different expectations than 342 // the GoogleURL library. For "/foo/bar/" the library will return the empty 343 // string, but WebCore wants "bar". 344 url::Component path = m_parsed.path; 345 if (path.len > 0 && m_string[path.end() - 1] == '/') 346 path.len--; 347 348 url::Component file; 349 if (m_string.is8Bit()) 350 url::ExtractFileName(asURLChar8Subtle(m_string), path, &file); 351 else 352 url::ExtractFileName(m_string.characters16(), path, &file); 353 354 // Bug: https://bugs.webkit.org/show_bug.cgi?id=21015 this function returns 355 // a null string when the path is empty, which we duplicate here. 356 if (!file.is_nonempty()) 357 return String(); 358 return componentString(file); 359 } 360 361 String KURL::protocol() const 362 { 363 return componentString(m_parsed.scheme); 364 } 365 366 String KURL::host() const 367 { 368 return componentString(m_parsed.host); 369 } 370 371 // Returns 0 when there is no port. 372 // 373 // We treat URL's with out-of-range port numbers as invalid URLs, and they will 374 // be rejected by the canonicalizer. KURL.cpp will allow them in parsing, but 375 // return invalidPortNumber from this port() function, so we mirror that behavior here. 376 unsigned short KURL::port() const 377 { 378 if (!m_isValid || m_parsed.port.len <= 0) 379 return 0; 380 ASSERT(!m_string.isNull()); 381 int port = m_string.is8Bit() ? 382 url::ParsePort(asURLChar8Subtle(m_string), m_parsed.port) : 383 url::ParsePort(m_string.characters16(), m_parsed.port); 384 ASSERT(port != url::PORT_UNSPECIFIED); // Checked port.len <= 0 before. 385 386 if (port == url::PORT_INVALID || port > maximumValidPortNumber) // Mimic KURL::port() 387 port = invalidPortNumber; 388 389 return static_cast<unsigned short>(port); 390 } 391 392 String KURL::pass() const 393 { 394 // Bug: https://bugs.webkit.org/show_bug.cgi?id=21015 this function returns 395 // a null string when the password is empty, which we duplicate here. 396 if (!m_parsed.password.is_nonempty()) 397 return String(); 398 return componentString(m_parsed.password); 399 } 400 401 String KURL::user() const 402 { 403 return componentString(m_parsed.username); 404 } 405 406 String KURL::fragmentIdentifier() const 407 { 408 // Empty but present refs ("foo.com/bar#") should result in the empty 409 // string, which componentString will produce. Nonexistent refs 410 // should be the null string. 411 if (!m_parsed.ref.is_valid()) 412 return String(); 413 return componentString(m_parsed.ref); 414 } 415 416 bool KURL::hasFragmentIdentifier() const 417 { 418 return m_parsed.ref.len >= 0; 419 } 420 421 String KURL::baseAsString() const 422 { 423 // FIXME: There is probably a more efficient way to do this? 424 return m_string.left(pathAfterLastSlash()); 425 } 426 427 String KURL::query() const 428 { 429 if (m_parsed.query.len >= 0) 430 return componentString(m_parsed.query); 431 432 // Bug: https://bugs.webkit.org/show_bug.cgi?id=21015 this function returns 433 // an empty string when the query is empty rather than a null (not sure 434 // which is right). 435 // Returns a null if the query is not specified, instead of empty. 436 if (m_parsed.query.is_valid()) 437 return emptyString(); 438 return String(); 439 } 440 441 String KURL::path() const 442 { 443 return componentString(m_parsed.path); 444 } 445 446 bool KURL::setProtocol(const String& protocol) 447 { 448 // Firefox and IE remove everything after the first ':'. 449 int separatorPosition = protocol.find(':'); 450 String newProtocol = protocol.substring(0, separatorPosition); 451 StringUTF8Adaptor newProtocolUTF8(newProtocol); 452 453 // If KURL is given an invalid scheme, it returns failure without modifying 454 // the URL at all. This is in contrast to most other setters which modify 455 // the URL and set "m_isValid." 456 url::RawCanonOutputT<char> canonProtocol; 457 url::Component protocolComponent; 458 if (!url::CanonicalizeScheme(newProtocolUTF8.data(), url::Component(0, newProtocolUTF8.length()), &canonProtocol, &protocolComponent) 459 || !protocolComponent.is_nonempty()) 460 return false; 461 462 url::Replacements<char> replacements; 463 replacements.SetScheme(charactersOrEmpty(newProtocolUTF8), url::Component(0, newProtocolUTF8.length())); 464 replaceComponents(replacements); 465 466 // isValid could be false but we still return true here. This is because 467 // WebCore or JS scripts can build up a URL by setting individual 468 // components, and a JS exception is based on the return value of this 469 // function. We want to throw the exception and stop the script only when 470 // its trying to set a bad protocol, and not when it maybe just hasn't 471 // finished building up its final scheme. 472 return true; 473 } 474 475 void KURL::setHost(const String& host) 476 { 477 StringUTF8Adaptor hostUTF8(host); 478 url::Replacements<char> replacements; 479 replacements.SetHost(charactersOrEmpty(hostUTF8), url::Component(0, hostUTF8.length())); 480 replaceComponents(replacements); 481 } 482 483 static String parsePortFromStringPosition(const String& value, unsigned portStart) 484 { 485 // "008080junk" needs to be treated as port "8080" and "000" as "0". 486 size_t length = value.length(); 487 unsigned portEnd = portStart; 488 while (isASCIIDigit(value[portEnd]) && portEnd < length) 489 ++portEnd; 490 while (value[portStart] == '0' && portStart < portEnd - 1) 491 ++portStart; 492 493 // Required for backwards compat. 494 // https://www.w3.org/Bugs/Public/show_bug.cgi?id=23463 495 if (portStart == portEnd) 496 return "0"; 497 498 return value.substring(portStart, portEnd - portStart); 499 } 500 501 void KURL::setHostAndPort(const String& hostAndPort) 502 { 503 size_t separator = hostAndPort.find(':'); 504 if (!separator) 505 return; 506 507 if (separator == kNotFound) { 508 url::Replacements<char> replacements; 509 StringUTF8Adaptor hostUTF8(hostAndPort); 510 replacements.SetHost(charactersOrEmpty(hostUTF8), url::Component(0, hostUTF8.length())); 511 replaceComponents(replacements); 512 return; 513 } 514 515 String host = hostAndPort.substring(0, separator); 516 String port = parsePortFromStringPosition(hostAndPort, separator + 1); 517 518 StringUTF8Adaptor hostUTF8(host); 519 StringUTF8Adaptor portUTF8(port); 520 521 url::Replacements<char> replacements; 522 replacements.SetHost(charactersOrEmpty(hostUTF8), url::Component(0, hostUTF8.length())); 523 replacements.SetPort(charactersOrEmpty(portUTF8), url::Component(0, portUTF8.length())); 524 replaceComponents(replacements); 525 } 526 527 void KURL::removePort() 528 { 529 if (!hasPort()) 530 return; 531 url::Replacements<char> replacements; 532 replacements.ClearPort(); 533 replaceComponents(replacements); 534 } 535 536 void KURL::setPort(const String& port) 537 { 538 String parsedPort = parsePortFromStringPosition(port, 0); 539 setPort(parsedPort.toUInt()); 540 } 541 542 void KURL::setPort(unsigned short port) 543 { 544 if (isDefaultPortForProtocol(port, protocol())) { 545 removePort(); 546 return; 547 } 548 549 String portString = String::number(port); 550 ASSERT(portString.is8Bit()); 551 552 url::Replacements<char> replacements; 553 replacements.SetPort(reinterpret_cast<const char*>(portString.characters8()), url::Component(0, portString.length())); 554 replaceComponents(replacements); 555 } 556 557 void KURL::setUser(const String& user) 558 { 559 // This function is commonly called to clear the username, which we 560 // normally don't have, so we optimize this case. 561 if (user.isEmpty() && !m_parsed.username.is_valid()) 562 return; 563 564 // The canonicalizer will clear any usernames that are empty, so we 565 // don't have to explicitly call ClearUsername() here. 566 StringUTF8Adaptor userUTF8(user); 567 url::Replacements<char> replacements; 568 replacements.SetUsername(charactersOrEmpty(userUTF8), url::Component(0, userUTF8.length())); 569 replaceComponents(replacements); 570 } 571 572 void KURL::setPass(const String& pass) 573 { 574 // This function is commonly called to clear the password, which we 575 // normally don't have, so we optimize this case. 576 if (pass.isEmpty() && !m_parsed.password.is_valid()) 577 return; 578 579 // The canonicalizer will clear any passwords that are empty, so we 580 // don't have to explicitly call ClearUsername() here. 581 StringUTF8Adaptor passUTF8(pass); 582 url::Replacements<char> replacements; 583 replacements.SetPassword(charactersOrEmpty(passUTF8), url::Component(0, passUTF8.length())); 584 replaceComponents(replacements); 585 } 586 587 void KURL::setFragmentIdentifier(const String& fragment) 588 { 589 // This function is commonly called to clear the ref, which we 590 // normally don't have, so we optimize this case. 591 if (fragment.isNull() && !m_parsed.ref.is_valid()) 592 return; 593 594 StringUTF8Adaptor fragmentUTF8(fragment); 595 596 url::Replacements<char> replacements; 597 if (fragment.isNull()) 598 replacements.ClearRef(); 599 else 600 replacements.SetRef(charactersOrEmpty(fragmentUTF8), url::Component(0, fragmentUTF8.length())); 601 replaceComponents(replacements); 602 } 603 604 void KURL::removeFragmentIdentifier() 605 { 606 url::Replacements<char> replacements; 607 replacements.ClearRef(); 608 replaceComponents(replacements); 609 } 610 611 void KURL::setQuery(const String& query) 612 { 613 StringUTF8Adaptor queryUTF8(query); 614 url::Replacements<char> replacements; 615 if (query.isNull()) { 616 // KURL.cpp sets to null to clear any query. 617 replacements.ClearQuery(); 618 } else if (query.length() > 0 && query[0] == '?') { 619 // WebCore expects the query string to begin with a question mark, but 620 // GoogleURL doesn't. So we trim off the question mark when setting. 621 replacements.SetQuery(charactersOrEmpty(queryUTF8), url::Component(1, queryUTF8.length() - 1)); 622 } else { 623 // When set with the empty string or something that doesn't begin with 624 // a question mark, KURL.cpp will add a question mark for you. The only 625 // way this isn't compatible is if you call this function with an empty 626 // string. KURL.cpp will leave a '?' with nothing following it in the 627 // URL, whereas we'll clear it. 628 // FIXME We should eliminate this difference. 629 replacements.SetQuery(charactersOrEmpty(queryUTF8), url::Component(0, queryUTF8.length())); 630 } 631 replaceComponents(replacements); 632 } 633 634 void KURL::setPath(const String& path) 635 { 636 // Empty paths will be canonicalized to "/", so we don't have to worry 637 // about calling ClearPath(). 638 StringUTF8Adaptor pathUTF8(path); 639 url::Replacements<char> replacements; 640 replacements.SetPath(charactersOrEmpty(pathUTF8), url::Component(0, pathUTF8.length())); 641 replaceComponents(replacements); 642 } 643 644 String decodeURLEscapeSequences(const String& string) 645 { 646 return decodeURLEscapeSequences(string, UTF8Encoding()); 647 } 648 649 // In KURL.cpp's implementation, this is called by every component getter. 650 // It will unescape every character, including '\0'. This is scary, and may 651 // cause security holes. We never call this function for components, and 652 // just return the ASCII versions instead. 653 // 654 // This function is also used to decode javascript: URLs and as a general 655 // purpose unescaping function. 656 // 657 // FIXME These should be merged to the KURL.cpp implementation. 658 String decodeURLEscapeSequences(const String& string, const WTF::TextEncoding& encoding) 659 { 660 // FIXME We can probably use KURL.cpp's version of this function 661 // without modification. However, I'm concerned about 662 // https://bugs.webkit.org/show_bug.cgi?id=20559 so am keeping this old 663 // custom code for now. Using their version will also fix the bug that 664 // we ignore the encoding. 665 // 666 // FIXME b/1350291: This does not get called very often. We just convert 667 // first to 8-bit UTF-8, then unescape, then back to 16-bit. This kind of 668 // sucks, and we don't use the encoding properly, which will make some 669 // obscure anchor navigations fail. 670 StringUTF8Adaptor stringUTF8(string); 671 url::RawCanonOutputT<url::UTF16Char> unescaped; 672 url::DecodeURLEscapeSequences(stringUTF8.data(), stringUTF8.length(), &unescaped); 673 return StringImpl::create8BitIfPossible(reinterpret_cast<UChar*>(unescaped.data()), unescaped.length()); 674 } 675 676 String encodeWithURLEscapeSequences(const String& notEncodedString) 677 { 678 CString utf8 = UTF8Encoding().normalizeAndEncode(notEncodedString, WTF::URLEncodedEntitiesForUnencodables); 679 680 url::RawCanonOutputT<char> buffer; 681 int inputLength = utf8.length(); 682 if (buffer.length() < inputLength * 3) 683 buffer.Resize(inputLength * 3); 684 685 url::EncodeURIComponent(utf8.data(), inputLength, &buffer); 686 String escaped(buffer.data(), buffer.length()); 687 // Unescape '/'; it's safe and much prettier. 688 escaped.replace("%2F", "/"); 689 return escaped; 690 } 691 692 bool KURL::isHierarchical() const 693 { 694 if (m_string.isNull() || !m_parsed.scheme.is_nonempty()) 695 return false; 696 return m_string.is8Bit() ? 697 url::IsStandard(asURLChar8Subtle(m_string), m_parsed.scheme) : 698 url::IsStandard(m_string.characters16(), m_parsed.scheme); 699 } 700 701 #ifndef NDEBUG 702 void KURL::print() const 703 { 704 printf("%s\n", m_string.utf8().data()); 705 } 706 #endif 707 708 bool equalIgnoringFragmentIdentifier(const KURL& a, const KURL& b) 709 { 710 // Compute the length of each URL without its ref. Note that the reference 711 // begin (if it exists) points to the character *after* the '#', so we need 712 // to subtract one. 713 int aLength = a.m_string.length(); 714 if (a.m_parsed.ref.len >= 0) 715 aLength = a.m_parsed.ref.begin - 1; 716 717 int bLength = b.m_string.length(); 718 if (b.m_parsed.ref.len >= 0) 719 bLength = b.m_parsed.ref.begin - 1; 720 721 if (aLength != bLength) 722 return false; 723 724 const String& aString = a.m_string; 725 const String& bString = b.m_string; 726 // FIXME: Abstraction this into a function in WTFString.h. 727 for (int i = 0; i < aLength; ++i) { 728 if (aString[i] != bString[i]) 729 return false; 730 } 731 return true; 732 } 733 734 unsigned KURL::hostStart() const 735 { 736 return m_parsed.CountCharactersBefore(url::Parsed::HOST, false); 737 } 738 739 unsigned KURL::hostEnd() const 740 { 741 return m_parsed.CountCharactersBefore(url::Parsed::PORT, true); 742 } 743 744 unsigned KURL::pathStart() const 745 { 746 return m_parsed.CountCharactersBefore(url::Parsed::PATH, false); 747 } 748 749 unsigned KURL::pathEnd() const 750 { 751 return m_parsed.CountCharactersBefore(url::Parsed::QUERY, true); 752 } 753 754 unsigned KURL::pathAfterLastSlash() const 755 { 756 if (m_string.isNull()) 757 return 0; 758 if (!m_isValid || !m_parsed.path.is_valid()) 759 return m_parsed.CountCharactersBefore(url::Parsed::PATH, false); 760 url::Component filename; 761 if (m_string.is8Bit()) 762 url::ExtractFileName(asURLChar8Subtle(m_string), m_parsed.path, &filename); 763 else 764 url::ExtractFileName(m_string.characters16(), m_parsed.path, &filename); 765 return filename.begin; 766 } 767 768 bool protocolIs(const String& url, const char* protocol) 769 { 770 assertProtocolIsGood(protocol); 771 if (url.isNull()) 772 return false; 773 if (url.is8Bit()) 774 return url::FindAndCompareScheme(asURLChar8Subtle(url), url.length(), protocol, 0); 775 return url::FindAndCompareScheme(url.characters16(), url.length(), protocol, 0); 776 } 777 778 void KURL::init(const KURL& base, const String& relative, const WTF::TextEncoding* queryEncoding) 779 { 780 if (!relative.isNull() && relative.is8Bit()) { 781 StringUTF8Adaptor relativeUTF8(relative); 782 init(base, relativeUTF8.data(), relativeUTF8.length(), queryEncoding); 783 } else 784 init(base, relative.characters16(), relative.length(), queryEncoding); 785 initProtocolIsInHTTPFamily(); 786 initInnerURL(); 787 } 788 789 template <typename CHAR> 790 void KURL::init(const KURL& base, const CHAR* relative, int relativeLength, const WTF::TextEncoding* queryEncoding) 791 { 792 // As a performance optimization, we do not use the charset converter 793 // if encoding is UTF-8 or other Unicode encodings. Note that this is 794 // per HTML5 2.5.3 (resolving URL). The URL canonicalizer will be more 795 // efficient with no charset converter object because it can do UTF-8 796 // internally with no extra copies. 797 798 // We feel free to make the charset converter object every time since it's 799 // just a wrapper around a reference. 800 KURLCharsetConverter charsetConverterObject(queryEncoding); 801 KURLCharsetConverter* charsetConverter = (!queryEncoding || isUnicodeEncoding(queryEncoding)) ? 0 : &charsetConverterObject; 802 803 StringUTF8Adaptor baseUTF8(base.string()); 804 805 url::RawCanonOutputT<char> output; 806 m_isValid = url::ResolveRelative(baseUTF8.data(), baseUTF8.length(), base.m_parsed, relative, relativeLength, charsetConverter, &output, &m_parsed); 807 808 // See FIXME in KURLPrivate in the header. If canonicalization has not 809 // changed the string, we can avoid an extra allocation by using assignment. 810 m_string = AtomicString::fromUTF8(output.data(), output.length()); 811 } 812 813 void KURL::initInnerURL() 814 { 815 if (!m_isValid) { 816 m_innerURL.clear(); 817 return; 818 } 819 if (url::Parsed* innerParsed = m_parsed.inner_parsed()) 820 m_innerURL = adoptPtr(new KURL(ParsedURLString, m_string.substring(innerParsed->scheme.begin, innerParsed->Length() - innerParsed->scheme.begin))); 821 else 822 m_innerURL.clear(); 823 } 824 825 template<typename CHAR> 826 bool internalProtocolIs(const url::Component& scheme, const CHAR* spec, const char* protocol) 827 { 828 const CHAR* begin = spec + scheme.begin; 829 const CHAR* end = begin + scheme.len; 830 831 while (begin != end && *protocol) { 832 ASSERT(toASCIILower(*protocol) == *protocol); 833 if (toASCIILower(*begin++) != *protocol++) 834 return false; 835 } 836 837 // Both strings are equal (ignoring case) if and only if all of the characters were equal, 838 // and the end of both has been reached. 839 return begin == end && !*protocol; 840 } 841 842 template<typename CHAR> 843 bool checkIfProtocolIsInHTTPFamily(const url::Component& scheme, const CHAR* spec) 844 { 845 if (scheme.len == 4) 846 return internalProtocolIs(scheme, spec, "http"); 847 if (scheme.len == 5) 848 return internalProtocolIs(scheme, spec, "https"); 849 return false; 850 } 851 852 void KURL::initProtocolIsInHTTPFamily() 853 { 854 if (!m_isValid) { 855 m_protocolIsInHTTPFamily = false; 856 return; 857 } 858 859 ASSERT(!m_string.isNull()); 860 m_protocolIsInHTTPFamily = m_string.is8Bit() ? 861 checkIfProtocolIsInHTTPFamily(m_parsed.scheme, m_string.characters8()) : 862 checkIfProtocolIsInHTTPFamily(m_parsed.scheme, m_string.characters16()); 863 } 864 865 bool KURL::protocolIs(const char* protocol) const 866 { 867 assertProtocolIsGood(protocol); 868 869 // JavaScript URLs are "valid" and should be executed even if KURL decides they are invalid. 870 // The free function protocolIsJavaScript() should be used instead. 871 // FIXME: Chromium code needs to be fixed for this assert to be enabled. ASSERT(strcmp(protocol, "javascript")); 872 873 if (m_string.isNull() || m_parsed.scheme.len <= 0) 874 return *protocol == '\0'; 875 876 return m_string.is8Bit() ? 877 internalProtocolIs(m_parsed.scheme, m_string.characters8(), protocol) : 878 internalProtocolIs(m_parsed.scheme, m_string.characters16(), protocol); 879 } 880 881 String KURL::stringForInvalidComponent() const 882 { 883 if (m_string.isNull()) 884 return String(); 885 return emptyString(); 886 } 887 888 String KURL::componentString(const url::Component& component) const 889 { 890 if (!m_isValid || component.len <= 0) 891 return stringForInvalidComponent(); 892 // begin and len are in terms of bytes which do not match 893 // if string() is UTF-16 and input contains non-ASCII characters. 894 // However, the only part in urlString that can contain non-ASCII 895 // characters is 'ref' at the end of the string. In that case, 896 // begin will always match the actual value and len (in terms of 897 // byte) will be longer than what's needed by 'mid'. However, mid 898 // truncates len to avoid go past the end of a string so that we can 899 // get away without doing anything here. 900 return string().substring(component.begin, component.len); 901 } 902 903 template<typename CHAR> 904 void KURL::replaceComponents(const url::Replacements<CHAR>& replacements) 905 { 906 url::RawCanonOutputT<char> output; 907 url::Parsed newParsed; 908 909 StringUTF8Adaptor utf8(m_string); 910 m_isValid = url::ReplaceComponents(utf8.data(), utf8.length(), m_parsed, replacements, 0, &output, &newParsed); 911 912 m_parsed = newParsed; 913 m_string = AtomicString::fromUTF8(output.data(), output.length()); 914 } 915 916 bool KURL::isSafeToSendToAnotherThread() const 917 { 918 return m_string.isSafeToSendToAnotherThread() 919 && (!m_innerURL || m_innerURL->isSafeToSendToAnotherThread()); 920 } 921 922 } // namespace blink 923