url/gurl.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 #ifdef WIN32
   6 #include <windows.h>
   7 #else
   8 #include <pthread.h>
   9 #endif
  10
  11 #include <algorithm>
  12 #include <ostream>
  13
  14 #include "url/gurl.h"
  15
  16 #include "base/logging.h"
  17 #include "url/url_canon_stdstring.h"
  18 #include "url/url_util.h"
  19
  20 namespace {
  21
  22 static std::string* empty_string = NULL;
  23 static GURL* empty_gurl = NULL;
  24
  25 #ifdef WIN32
  26
  27 // Returns a static reference to an empty string for returning a reference
  28 // when there is no underlying string.
  29 const std::string& EmptyStringForGURL() {
  30   // Avoid static object construction/destruction on startup/shutdown.
  31   if (!empty_string) {
  32     // Create the string. Be careful that we don't break in the case that this
  33     // is being called from multiple threads. Statics are not threadsafe.
  34     std::string* new_empty_string = new std::string;
  35     if (InterlockedCompareExchangePointer(
  36         reinterpret_cast<PVOID*>(&empty_string), new_empty_string, NULL)) {
  37       // The old value was non-NULL, so no replacement was done. Another
  38       // thread did the initialization out from under us.
  39       delete new_empty_string;
  40     }
  41   }
  42   return *empty_string;
  43 }
  44
  45 #else
  46
  47 static pthread_once_t empty_string_once = PTHREAD_ONCE_INIT;
  48 static pthread_once_t empty_gurl_once = PTHREAD_ONCE_INIT;
  49
  50 void EmptyStringForGURLOnce(void) {
  51   empty_string = new std::string;
  52 }
  53
  54 const std::string& EmptyStringForGURL() {
  55   // Avoid static object construction/destruction on startup/shutdown.
  56   pthread_once(&empty_string_once, EmptyStringForGURLOnce);
  57   return *empty_string;
  58 }
  59
  60 #endif  // WIN32
  61
  62 } // namespace
  63
  64 GURL::GURL() : is_valid_(false) {
  65 }
  66
  67 GURL::GURL(const GURL& other)
  68     : spec_(other.spec_),
  69       is_valid_(other.is_valid_),
  70       parsed_(other.parsed_) {
  71   if (other.inner_url_)
  72     inner_url_.reset(new GURL(*other.inner_url_));
  73   // Valid filesystem urls should always have an inner_url_.
  74   DCHECK(!is_valid_ || !SchemeIsFileSystem() || inner_url_);
  75 }
  76
  77 GURL::GURL(const std::string& url_string) {
  78   InitCanonical(url_string, true);
  79 }
  80
  81 GURL::GURL(const base::string16& url_string) {
  82   InitCanonical(url_string, true);
  83 }
  84
  85 GURL::GURL(const std::string& url_string, RetainWhiteSpaceSelector) {
  86   InitCanonical(url_string, false);
  87 }
  88
  89 GURL::GURL(const char* canonical_spec,
  90            size_t canonical_spec_len,
  91            const url::Parsed& parsed,
  92            bool is_valid)
  93     : spec_(canonical_spec, canonical_spec_len),
  94       is_valid_(is_valid),
  95       parsed_(parsed) {
  96   InitializeFromCanonicalSpec();
  97 }
  98
  99 GURL::GURL(std::string canonical_spec, const url::Parsed& parsed, bool is_valid)
 100     : is_valid_(is_valid),
 101       parsed_(parsed) {
 102   spec_.swap(canonical_spec);
 103   InitializeFromCanonicalSpec();
 104 }
 105
 106 template<typename STR>
 107 void GURL::InitCanonical(const STR& input_spec, bool trim_path_end) {
 108   // Reserve enough room in the output for the input, plus some extra so that
 109   // we have room if we have to escape a few things without reallocating.
 110   spec_.reserve(input_spec.size() + 32);
 111   url::StdStringCanonOutput output(&spec_);
 112   is_valid_ = url::Canonicalize(
 113       input_spec.data(), static_cast<int>(input_spec.length()), trim_path_end,
 114       NULL, &output, &parsed_);
 115
 116   output.Complete();  // Must be done before using string.
 117   if (is_valid_ && SchemeIsFileSystem()) {
 118     inner_url_.reset(new GURL(spec_.data(), parsed_.Length(),
 119                               *parsed_.inner_parsed(), true));
 120   }
 121 }
 122
 123 void GURL::InitializeFromCanonicalSpec() {
 124   if (is_valid_ && SchemeIsFileSystem()) {
 125     inner_url_.reset(
 126         new GURL(spec_.data(), parsed_.Length(),
 127                  *parsed_.inner_parsed(), true));
 128   }
 129
 130 #ifndef NDEBUG
 131   // For testing purposes, check that the parsed canonical URL is identical to
 132   // what we would have produced. Skip checking for invalid URLs have no meaning
 133   // and we can't always canonicalize then reproducabely.
 134   if (is_valid_) {
 135     url::Component scheme;
 136     // We can't do this check on the inner_url of a filesystem URL, as
 137     // canonical_spec actually points to the start of the outer URL, so we'd
 138     // end up with infinite recursion in this constructor.
 139     if (!url::FindAndCompareScheme(spec_.data(), spec_.length(),
 140                                    url::kFileSystemScheme, &scheme) ||
 141         scheme.begin == parsed_.scheme.begin) {
 142       // We need to retain trailing whitespace on path URLs, as the |parsed_|
 143       // spec we originally received may legitimately contain trailing white-
 144       // space on the path or  components e.g. if the #ref has been
 145       // removed from a "foo:hello #ref" URL (see http://crbug.com/291747).
 146       GURL test_url(spec_, RETAIN_TRAILING_PATH_WHITEPACE);
 147
 148       DCHECK(test_url.is_valid_ == is_valid_);
 149       DCHECK(test_url.spec_ == spec_);
 150
 151       DCHECK(test_url.parsed_.scheme == parsed_.scheme);
 152       DCHECK(test_url.parsed_.username == parsed_.username);
 153       DCHECK(test_url.parsed_.password == parsed_.password);
 154       DCHECK(test_url.parsed_.host == parsed_.host);
 155       DCHECK(test_url.parsed_.port == parsed_.port);
 156       DCHECK(test_url.parsed_.path == parsed_.path);
 157       DCHECK(test_url.parsed_.query == parsed_.query);
 158       DCHECK(test_url.parsed_.ref == parsed_.ref);
 159     }
 160   }
 161 #endif
 162 }
 163
 164 GURL::~GURL() {
 165 }
 166
 167 GURL& GURL::operator=(GURL other) {
 168   Swap(&other);
 169   return *this;
 170 }
 171
 172 const std::string& GURL::spec() const {
 173   if (is_valid_ || spec_.empty())
 174     return spec_;
 175
 176   DCHECK(false) << "Trying to get the spec of an invalid URL!";
 177   return EmptyStringForGURL();
 178 }
 179
 180 GURL GURL::Resolve(const std::string& relative) const {
 181   return ResolveWithCharsetConverter(relative, NULL);
 182 }
 183 GURL GURL::Resolve(const base::string16& relative) const {
 184   return ResolveWithCharsetConverter(relative, NULL);
 185 }
 186
 187 // Note: code duplicated below (it's inconvenient to use a template here).
 188 GURL GURL::ResolveWithCharsetConverter(
 189     const std::string& relative,
 190     url::CharsetConverter* charset_converter) const {
 191   // Not allowed for invalid URLs.
 192   if (!is_valid_)
 193     return GURL();
 194
 195   GURL result;
 196
 197   // Reserve enough room in the output for the input, plus some extra so that
 198   // we have room if we have to escape a few things without reallocating.
 199   result.spec_.reserve(spec_.size() + 32);
 200   url::StdStringCanonOutput output(&result.spec_);
 201
 202   if (!url::ResolveRelative(spec_.data(), static_cast<int>(spec_.length()),
 203                             parsed_, relative.data(),
 204                             static_cast<int>(relative.length()),
 205                             charset_converter, &output, &result.parsed_)) {
 206     // Error resolving, return an empty URL.
 207     return GURL();
 208   }
 209
 210   output.Complete();
 211   result.is_valid_ = true;
 212   if (result.SchemeIsFileSystem()) {
 213     result.inner_url_.reset(
 214         new GURL(result.spec_.data(), result.parsed_.Length(),
 215                  *result.parsed_.inner_parsed(), true));
 216   }
 217   return result;
 218 }
 219
 220 // Note: code duplicated above (it's inconvenient to use a template here).
 221 GURL GURL::ResolveWithCharsetConverter(
 222     const base::string16& relative,
 223     url::CharsetConverter* charset_converter) const {
 224   // Not allowed for invalid URLs.
 225   if (!is_valid_)
 226     return GURL();
 227
 228   GURL result;
 229
 230   // Reserve enough room in the output for the input, plus some extra so that
 231   // we have room if we have to escape a few things without reallocating.
 232   result.spec_.reserve(spec_.size() + 32);
 233   url::StdStringCanonOutput output(&result.spec_);
 234
 235   if (!url::ResolveRelative(spec_.data(), static_cast<int>(spec_.length()),
 236                             parsed_, relative.data(),
 237                             static_cast<int>(relative.length()),
 238                             charset_converter, &output, &result.parsed_)) {
 239     // Error resolving, return an empty URL.
 240     return GURL();
 241   }
 242
 243   output.Complete();
 244   result.is_valid_ = true;
 245   if (result.SchemeIsFileSystem()) {
 246     result.inner_url_.reset(
 247         new GURL(result.spec_.data(), result.parsed_.Length(),
 248                  *result.parsed_.inner_parsed(), true));
 249   }
 250   return result;
 251 }
 252
 253 // Note: code duplicated below (it's inconvenient to use a template here).
 254 GURL GURL::ReplaceComponents(
 255     const url::Replacements<char>& replacements) const {
 256   GURL result;
 257
 258   // Not allowed for invalid URLs.
 259   if (!is_valid_)
 260     return GURL();
 261
 262   // Reserve enough room in the output for the input, plus some extra so that
 263   // we have room if we have to escape a few things without reallocating.
 264   result.spec_.reserve(spec_.size() + 32);
 265   url::StdStringCanonOutput output(&result.spec_);
 266
 267   result.is_valid_ = url::ReplaceComponents(
 268       spec_.data(), static_cast<int>(spec_.length()), parsed_, replacements,
 269       NULL, &output, &result.parsed_);
 270
 271   output.Complete();
 272   if (result.is_valid_ && result.SchemeIsFileSystem()) {
 273     result.inner_url_.reset(new GURL(spec_.data(), result.parsed_.Length(),
 274                                      *result.parsed_.inner_parsed(), true));
 275   }
 276   return result;
 277 }
 278
 279 // Note: code duplicated above (it's inconvenient to use a template here).
 280 GURL GURL::ReplaceComponents(
 281     const url::Replacements<base::char16>& replacements) const {
 282   GURL result;
 283
 284   // Not allowed for invalid URLs.
 285   if (!is_valid_)
 286     return GURL();
 287
 288   // Reserve enough room in the output for the input, plus some extra so that
 289   // we have room if we have to escape a few things without reallocating.
 290   result.spec_.reserve(spec_.size() + 32);
 291   url::StdStringCanonOutput output(&result.spec_);
 292
 293   result.is_valid_ = url::ReplaceComponents(
 294       spec_.data(), static_cast<int>(spec_.length()), parsed_, replacements,
 295       NULL, &output, &result.parsed_);
 296
 297   output.Complete();
 298   if (result.is_valid_ && result.SchemeIsFileSystem()) {
 299     result.inner_url_.reset(new GURL(spec_.data(), result.parsed_.Length(),
 300                                      *result.parsed_.inner_parsed(), true));
 301   }
 302   return result;
 303 }
 304
 305 GURL GURL::GetOrigin() const {
 306   // This doesn't make sense for invalid or nonstandard URLs, so return
 307   // the empty URL
 308   if (!is_valid_ || !IsStandard())
 309     return GURL();
 310
 311   if (SchemeIsFileSystem())
 312     return inner_url_->GetOrigin();
 313
 314   url::Replacements<char> replacements;
 315   replacements.ClearUsername();
 316   replacements.ClearPassword();
 317   replacements.ClearPath();
 318   replacements.ClearQuery();
 319   replacements.ClearRef();
 320
 321   return ReplaceComponents(replacements);
 322 }
 323
 324 GURL GURL::GetAsReferrer() const {
 325   if (!is_valid_ || !SchemeIsHTTPOrHTTPS())
 326     return GURL();
 327
 328   if (!has_ref() && !has_username() && !has_password())
 329     return GURL(*this);
 330
 331   url::Replacements<char> replacements;
 332   replacements.ClearRef();
 333   replacements.ClearUsername();
 334   replacements.ClearPassword();
 335   return ReplaceComponents(replacements);
 336 }
 337
 338 GURL GURL::GetWithEmptyPath() const {
 339   // This doesn't make sense for invalid or nonstandard URLs, so return
 340   // the empty URL.
 341   if (!is_valid_ || !IsStandard())
 342     return GURL();
 343
 344   // We could optimize this since we know that the URL is canonical, and we are
 345   // appending a canonical path, so avoiding re-parsing.
 346   GURL other(*this);
 347   if (parsed_.path.len == 0)
 348     return other;
 349
 350   // Clear everything after the path.
 351   other.parsed_.query.reset();
 352   other.parsed_.ref.reset();
 353
 354   // Set the path, since the path is longer than one, we can just set the
 355   // first character and resize.
 356   other.spec_[other.parsed_.path.begin] = '/';
 357   other.parsed_.path.len = 1;
 358   other.spec_.resize(other.parsed_.path.begin + 1);
 359   return other;
 360 }
 361
 362 bool GURL::IsStandard() const {
 363   return url::IsStandard(spec_.data(), parsed_.scheme);
 364 }
 365
 366 bool GURL::SchemeIs(const char* lower_ascii_scheme) const {
 367   if (parsed_.scheme.len <= 0)
 368     return lower_ascii_scheme == NULL;
 369   return url::LowerCaseEqualsASCII(spec_.data() + parsed_.scheme.begin,
 370                                    spec_.data() + parsed_.scheme.end(),
 371                                    lower_ascii_scheme);
 372 }
 373
 374 bool GURL::SchemeIsHTTPOrHTTPS() const {
 375   return SchemeIs(url::kHttpScheme) || SchemeIs(url::kHttpsScheme);
 376 }
 377
 378 bool GURL::SchemeIsWSOrWSS() const {
 379   return SchemeIs(url::kWsScheme) || SchemeIs(url::kWssScheme);
 380 }
 381
 382 int GURL::IntPort() const {
 383   if (parsed_.port.is_nonempty())
 384     return url::ParsePort(spec_.data(), parsed_.port);
 385   return url::PORT_UNSPECIFIED;
 386 }
 387
 388 int GURL::EffectiveIntPort() const {
 389   int int_port = IntPort();
 390   if (int_port == url::PORT_UNSPECIFIED && IsStandard())
 391     return url::DefaultPortForScheme(spec_.data() + parsed_.scheme.begin,
 392                                      parsed_.scheme.len);
 393   return int_port;
 394 }
 395
 396 std::string GURL::ExtractFileName() const {
 397   url::Component file_component;
 398   url::ExtractFileName(spec_.data(), parsed_.path, &file_component);
 399   return ComponentString(file_component);
 400 }
 401
 402 std::string GURL::PathForRequest() const {
 403   DCHECK(parsed_.path.len > 0) << "Canonical path for requests should be non-empty";
 404   if (parsed_.ref.len >= 0) {
 405     // Clip off the reference when it exists. The reference starts after the #
 406     // sign, so we have to subtract one to also remove it.
 407     return std::string(spec_, parsed_.path.begin,
 408                        parsed_.ref.begin - parsed_.path.begin - 1);
 409   }
 410   // Compute the actual path length, rather than depending on the spec's
 411   // terminator.  If we're an inner_url, our spec continues on into our outer
 412   // url's path/query/ref.
 413   int path_len = parsed_.path.len;
 414   if (parsed_.query.is_valid())
 415     path_len = parsed_.query.end() - parsed_.path.begin;
 416
 417   return std::string(spec_, parsed_.path.begin, path_len);
 418 }
 419
 420 std::string GURL::HostNoBrackets() const {
 421   // If host looks like an IPv6 literal, strip the square brackets.
 422   url::Component h(parsed_.host);
 423   if (h.len >= 2 && spec_[h.begin] == '[' && spec_[h.end() - 1] == ']') {
 424     h.begin++;
 425     h.len -= 2;
 426   }
 427   return ComponentString(h);
 428 }
 429
 430 std::string GURL::GetContent() const {
 431   return is_valid_ ? ComponentString(parsed_.GetContent()) : std::string();
 432 }
 433
 434 bool GURL::HostIsIPAddress() const {
 435   if (!is_valid_ || spec_.empty())
 436      return false;
 437
 438   url::RawCanonOutputT<char, 128> ignored_output;
 439   url::CanonHostInfo host_info;
 440   url::CanonicalizeIPAddress(spec_.c_str(), parsed_.host, &ignored_output,
 441                              &host_info);
 442   return host_info.IsIPAddress();
 443 }
 444
 445 #ifdef WIN32
 446
 447 const GURL& GURL::EmptyGURL() {
 448   // Avoid static object construction/destruction on startup/shutdown.
 449   if (!empty_gurl) {
 450     // Create the string. Be careful that we don't break in the case that this
 451     // is being called from multiple threads.
 452     GURL* new_empty_gurl = new GURL;
 453     if (InterlockedCompareExchangePointer(
 454         reinterpret_cast<PVOID*>(&empty_gurl), new_empty_gurl, NULL)) {
 455       // The old value was non-NULL, so no replacement was done. Another
 456       // thread did the initialization out from under us.
 457       delete new_empty_gurl;
 458     }
 459   }
 460   return *empty_gurl;
 461 }
 462
 463 #else
 464
 465 void EmptyGURLOnce(void) {
 466   empty_gurl = new GURL;
 467 }
 468
 469 const GURL& GURL::EmptyGURL() {
 470   // Avoid static object construction/destruction on startup/shutdown.
 471   pthread_once(&empty_gurl_once, EmptyGURLOnce);
 472   return *empty_gurl;
 473 }
 474
 475 #endif  // WIN32
 476
 477 bool GURL::DomainIs(const char* lower_ascii_domain,
 478                     int domain_len) const {
 479   // Return false if this URL is not valid or domain is empty.
 480   if (!is_valid_ || !domain_len)
 481     return false;
 482
 483   // FileSystem URLs have empty parsed_.host, so check this first.
 484   if (SchemeIsFileSystem() && inner_url_)
 485     return inner_url_->DomainIs(lower_ascii_domain, domain_len);
 486
 487   if (!parsed_.host.is_nonempty())
 488     return false;
 489
 490   // Check whether the host name is end with a dot. If yes, treat it
 491   // the same as no-dot unless the input comparison domain is end
 492   // with dot.
 493   const char* last_pos = spec_.data() + parsed_.host.end() - 1;
 494   int host_len = parsed_.host.len;
 495   if ('.' == *last_pos && '.' != lower_ascii_domain[domain_len - 1]) {
 496     last_pos--;
 497     host_len--;
 498   }
 499
 500   // Return false if host's length is less than domain's length.
 501   if (host_len < domain_len)
 502     return false;
 503
 504   // Compare this url whether belong specific domain.
 505   const char* start_pos = spec_.data() + parsed_.host.begin +
 506                           host_len - domain_len;
 507
 508   if (!url::LowerCaseEqualsASCII(start_pos,
 509                                  last_pos + 1,
 510                                  lower_ascii_domain,
 511                                  lower_ascii_domain + domain_len))
 512     return false;
 513
 514   // Check whether host has right domain start with dot, make sure we got
 515   // right domain range. For example www.google.com has domain
 516   // "google.com" but www.iamnotgoogle.com does not.
 517   if ('.' != lower_ascii_domain[0] && host_len > domain_len &&
 518       '.' != *(start_pos - 1))
 519     return false;
 520
 521   return true;
 522 }
 523
 524 void GURL::Swap(GURL* other) {
 525   spec_.swap(other->spec_);
 526   std::swap(is_valid_, other->is_valid_);
 527   std::swap(parsed_, other->parsed_);
 528   inner_url_.swap(other->inner_url_);
 529 }
 530
 531 std::ostream& operator<<(std::ostream& out, const GURL& url) {
 532   return out << url.possibly_invalid_spec();
 533 }