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