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.
16 #include "base/logging.h"
17 #include "url/url_canon_stdstring.h"
18 #include "url/url_util.h"
22 static std::string
* empty_string
= NULL
;
23 static GURL
* empty_gurl
= NULL
;
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.
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
;
47 static pthread_once_t empty_string_once
= PTHREAD_ONCE_INIT
;
48 static pthread_once_t empty_gurl_once
= PTHREAD_ONCE_INIT
;
50 void EmptyStringForGURLOnce(void) {
51 empty_string
= new std::string
;
54 const std::string
& EmptyStringForGURL() {
55 // Avoid static object construction/destruction on startup/shutdown.
56 pthread_once(&empty_string_once
, EmptyStringForGURLOnce
);
64 GURL::GURL() : is_valid_(false) {
67 GURL::GURL(const GURL
& other
)
69 is_valid_(other
.is_valid_
),
70 parsed_(other
.parsed_
) {
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_
);
77 GURL::GURL(const std::string
& url_string
) {
78 InitCanonical(url_string
, true);
81 GURL::GURL(const base::string16
& url_string
) {
82 InitCanonical(url_string
, true);
85 GURL::GURL(const std::string
& url_string
, RetainWhiteSpaceSelector
) {
86 InitCanonical(url_string
, false);
89 GURL::GURL(const char* canonical_spec
, size_t canonical_spec_len
,
90 const url_parse::Parsed
& parsed
, bool is_valid
)
91 : spec_(canonical_spec
, canonical_spec_len
),
94 InitializeFromCanonicalSpec();
97 GURL::GURL(std::string canonical_spec
,
98 const url_parse::Parsed
& parsed
, bool is_valid
)
99 : is_valid_(is_valid
),
101 spec_
.swap(canonical_spec
);
102 InitializeFromCanonicalSpec();
105 template<typename STR
>
106 void GURL::InitCanonical(const STR
& input_spec
, bool trim_path_end
) {
107 // Reserve enough room in the output for the input, plus some extra so that
108 // we have room if we have to escape a few things without reallocating.
109 spec_
.reserve(input_spec
.size() + 32);
110 url_canon::StdStringCanonOutput
output(&spec_
);
111 is_valid_
= url_util::Canonicalize(
112 input_spec
.data(), static_cast<int>(input_spec
.length()), trim_path_end
,
113 NULL
, &output
, &parsed_
);
115 output
.Complete(); // Must be done before using string.
116 if (is_valid_
&& SchemeIsFileSystem()) {
117 inner_url_
.reset(new GURL(spec_
.data(), parsed_
.Length(),
118 *parsed_
.inner_parsed(), true));
122 void GURL::InitializeFromCanonicalSpec() {
123 if (is_valid_
&& SchemeIsFileSystem()) {
125 new GURL(spec_
.data(), parsed_
.Length(),
126 *parsed_
.inner_parsed(), true));
130 // For testing purposes, check that the parsed canonical URL is identical to
131 // what we would have produced. Skip checking for invalid URLs have no meaning
132 // and we can't always canonicalize then reproducabely.
134 url_parse::Component scheme
;
135 // We can't do this check on the inner_url of a filesystem URL, as
136 // canonical_spec actually points to the start of the outer URL, so we'd
137 // end up with infinite recursion in this constructor.
138 if (!url_util::FindAndCompareScheme(spec_
.data(), spec_
.length(),
139 "filesystem", &scheme
) ||
140 scheme
.begin
== parsed_
.scheme
.begin
) {
141 // We need to retain trailing whitespace on path URLs, as the |parsed_|
142 // spec we originally received may legitimately contain trailing white-
143 // space on the path or components e.g. if the #ref has been
144 // removed from a "foo:hello #ref" URL (see http://crbug.com/291747).
145 GURL
test_url(spec_
, RETAIN_TRAILING_PATH_WHITEPACE
);
147 DCHECK(test_url
.is_valid_
== is_valid_
);
148 DCHECK(test_url
.spec_
== spec_
);
150 DCHECK(test_url
.parsed_
.scheme
== parsed_
.scheme
);
151 DCHECK(test_url
.parsed_
.username
== parsed_
.username
);
152 DCHECK(test_url
.parsed_
.password
== parsed_
.password
);
153 DCHECK(test_url
.parsed_
.host
== parsed_
.host
);
154 DCHECK(test_url
.parsed_
.port
== parsed_
.port
);
155 DCHECK(test_url
.parsed_
.path
== parsed_
.path
);
156 DCHECK(test_url
.parsed_
.query
== parsed_
.query
);
157 DCHECK(test_url
.parsed_
.ref
== parsed_
.ref
);
166 GURL
& GURL::operator=(GURL other
) {
171 const std::string
& GURL::spec() const {
172 if (is_valid_
|| spec_
.empty())
175 DCHECK(false) << "Trying to get the spec of an invalid URL!";
176 return EmptyStringForGURL();
179 GURL
GURL::Resolve(const std::string
& relative
) const {
180 return ResolveWithCharsetConverter(relative
, NULL
);
182 GURL
GURL::Resolve(const base::string16
& relative
) const {
183 return ResolveWithCharsetConverter(relative
, NULL
);
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.
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_
);
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.
210 result
.is_valid_
= true;
211 if (result
.SchemeIsFileSystem()) {
212 result
.inner_url_
.reset(
213 new GURL(result
.spec_
.data(), result
.parsed_
.Length(),
214 *result
.parsed_
.inner_parsed(), true));
219 // Note: code duplicated above (it's inconvenient to use a template here).
220 GURL
GURL::ResolveWithCharsetConverter(
221 const base::string16
& relative
,
222 url_canon::CharsetConverter
* charset_converter
) const {
223 // Not allowed for invalid URLs.
229 // Reserve enough room in the output for the input, plus some extra so that
230 // we have room if we have to escape a few things without reallocating.
231 result
.spec_
.reserve(spec_
.size() + 32);
232 url_canon::StdStringCanonOutput
output(&result
.spec_
);
234 if (!url_util::ResolveRelative(
235 spec_
.data(), static_cast<int>(spec_
.length()), parsed_
,
236 relative
.data(), static_cast<int>(relative
.length()),
237 charset_converter
, &output
, &result
.parsed_
)) {
238 // Error resolving, return an empty URL.
243 result
.is_valid_
= true;
244 if (result
.SchemeIsFileSystem()) {
245 result
.inner_url_
.reset(
246 new GURL(result
.spec_
.data(), result
.parsed_
.Length(),
247 *result
.parsed_
.inner_parsed(), true));
252 // Note: code duplicated below (it's inconvenient to use a template here).
253 GURL
GURL::ReplaceComponents(
254 const url_canon::Replacements
<char>& replacements
) const {
257 // Not allowed for invalid URLs.
261 // Reserve enough room in the output for the input, plus some extra so that
262 // we have room if we have to escape a few things without reallocating.
263 result
.spec_
.reserve(spec_
.size() + 32);
264 url_canon::StdStringCanonOutput
output(&result
.spec_
);
266 result
.is_valid_
= url_util::ReplaceComponents(
267 spec_
.data(), static_cast<int>(spec_
.length()), parsed_
, replacements
,
268 NULL
, &output
, &result
.parsed_
);
271 if (result
.is_valid_
&& result
.SchemeIsFileSystem()) {
272 result
.inner_url_
.reset(new GURL(spec_
.data(), result
.parsed_
.Length(),
273 *result
.parsed_
.inner_parsed(), true));
278 // Note: code duplicated above (it's inconvenient to use a template here).
279 GURL
GURL::ReplaceComponents(
280 const url_canon::Replacements
<base::char16
>& replacements
) const {
283 // Not allowed for invalid URLs.
287 // Reserve enough room in the output for the input, plus some extra so that
288 // we have room if we have to escape a few things without reallocating.
289 result
.spec_
.reserve(spec_
.size() + 32);
290 url_canon::StdStringCanonOutput
output(&result
.spec_
);
292 result
.is_valid_
= url_util::ReplaceComponents(
293 spec_
.data(), static_cast<int>(spec_
.length()), parsed_
, replacements
,
294 NULL
, &output
, &result
.parsed_
);
297 if (result
.is_valid_
&& result
.SchemeIsFileSystem()) {
298 result
.inner_url_
.reset(new GURL(spec_
.data(), result
.parsed_
.Length(),
299 *result
.parsed_
.inner_parsed(), true));
304 GURL
GURL::GetOrigin() const {
305 // This doesn't make sense for invalid or nonstandard URLs, so return
307 if (!is_valid_
|| !IsStandard())
310 if (SchemeIsFileSystem())
311 return inner_url_
->GetOrigin();
313 url_canon::Replacements
<char> replacements
;
314 replacements
.ClearUsername();
315 replacements
.ClearPassword();
316 replacements
.ClearPath();
317 replacements
.ClearQuery();
318 replacements
.ClearRef();
320 return ReplaceComponents(replacements
);
323 GURL
GURL::GetAsReferrer() const {
325 (!has_ref() && !has_username() && !has_password()))
328 url_canon::Replacements
<char> replacements
;
329 replacements
.ClearRef();
330 replacements
.ClearUsername();
331 replacements
.ClearPassword();
332 return ReplaceComponents(replacements
);
335 GURL
GURL::GetWithEmptyPath() const {
336 // This doesn't make sense for invalid or nonstandard URLs, so return
338 if (!is_valid_
|| !IsStandard())
341 // We could optimize this since we know that the URL is canonical, and we are
342 // appending a canonical path, so avoiding re-parsing.
344 if (parsed_
.path
.len
== 0)
347 // Clear everything after the path.
348 other
.parsed_
.query
.reset();
349 other
.parsed_
.ref
.reset();
351 // Set the path, since the path is longer than one, we can just set the
352 // first character and resize.
353 other
.spec_
[other
.parsed_
.path
.begin
] = '/';
354 other
.parsed_
.path
.len
= 1;
355 other
.spec_
.resize(other
.parsed_
.path
.begin
+ 1);
359 bool GURL::IsStandard() const {
360 return url_util::IsStandard(spec_
.data(), parsed_
.scheme
);
363 bool GURL::SchemeIs(const char* lower_ascii_scheme
) const {
364 if (parsed_
.scheme
.len
<= 0)
365 return lower_ascii_scheme
== NULL
;
366 return url_util::LowerCaseEqualsASCII(spec_
.data() + parsed_
.scheme
.begin
,
367 spec_
.data() + parsed_
.scheme
.end(),
371 bool GURL::SchemeIsHTTPOrHTTPS() const {
372 return SchemeIs("http") || SchemeIs("https");
375 bool GURL::SchemeIsWSOrWSS() const {
376 return SchemeIs("ws") || SchemeIs("wss");
379 int GURL::IntPort() const {
380 if (parsed_
.port
.is_nonempty())
381 return url_parse::ParsePort(spec_
.data(), parsed_
.port
);
382 return url_parse::PORT_UNSPECIFIED
;
385 int GURL::EffectiveIntPort() const {
386 int int_port
= IntPort();
387 if (int_port
== url_parse::PORT_UNSPECIFIED
&& IsStandard())
388 return url_canon::DefaultPortForScheme(spec_
.data() + parsed_
.scheme
.begin
,
393 std::string
GURL::ExtractFileName() const {
394 url_parse::Component file_component
;
395 url_parse::ExtractFileName(spec_
.data(), parsed_
.path
, &file_component
);
396 return ComponentString(file_component
);
399 std::string
GURL::PathForRequest() const {
400 DCHECK(parsed_
.path
.len
> 0) << "Canonical path for requests should be non-empty";
401 if (parsed_
.ref
.len
>= 0) {
402 // Clip off the reference when it exists. The reference starts after the #
403 // sign, so we have to subtract one to also remove it.
404 return std::string(spec_
, parsed_
.path
.begin
,
405 parsed_
.ref
.begin
- parsed_
.path
.begin
- 1);
407 // Compute the actual path length, rather than depending on the spec's
408 // terminator. If we're an inner_url, our spec continues on into our outer
409 // url's path/query/ref.
410 int path_len
= parsed_
.path
.len
;
411 if (parsed_
.query
.is_valid())
412 path_len
= parsed_
.query
.end() - parsed_
.path
.begin
;
414 return std::string(spec_
, parsed_
.path
.begin
, path_len
);
417 std::string
GURL::HostNoBrackets() const {
418 // If host looks like an IPv6 literal, strip the square brackets.
419 url_parse::Component
h(parsed_
.host
);
420 if (h
.len
>= 2 && spec_
[h
.begin
] == '[' && spec_
[h
.end() - 1] == ']') {
424 return ComponentString(h
);
427 std::string
GURL::GetContent() const {
428 return is_valid_
? ComponentString(parsed_
.GetContent()) : std::string();
431 bool GURL::HostIsIPAddress() const {
432 if (!is_valid_
|| spec_
.empty())
435 url_canon::RawCanonOutputT
<char, 128> ignored_output
;
436 url_canon::CanonHostInfo host_info
;
437 url_canon::CanonicalizeIPAddress(spec_
.c_str(), parsed_
.host
,
438 &ignored_output
, &host_info
);
439 return host_info
.IsIPAddress();
444 const GURL
& GURL::EmptyGURL() {
445 // Avoid static object construction/destruction on startup/shutdown.
447 // Create the string. Be careful that we don't break in the case that this
448 // is being called from multiple threads.
449 GURL
* new_empty_gurl
= new GURL
;
450 if (InterlockedCompareExchangePointer(
451 reinterpret_cast<PVOID
*>(&empty_gurl
), new_empty_gurl
, NULL
)) {
452 // The old value was non-NULL, so no replacement was done. Another
453 // thread did the initialization out from under us.
454 delete new_empty_gurl
;
462 void EmptyGURLOnce(void) {
463 empty_gurl
= new GURL
;
466 const GURL
& GURL::EmptyGURL() {
467 // Avoid static object construction/destruction on startup/shutdown.
468 pthread_once(&empty_gurl_once
, EmptyGURLOnce
);
474 bool GURL::DomainIs(const char* lower_ascii_domain
,
475 int domain_len
) const {
476 // Return false if this URL is not valid or domain is empty.
477 if (!is_valid_
|| !domain_len
)
480 // FileSystem URLs have empty parsed_.host, so check this first.
481 if (SchemeIsFileSystem() && inner_url_
)
482 return inner_url_
->DomainIs(lower_ascii_domain
, domain_len
);
484 if (!parsed_
.host
.is_nonempty())
487 // Check whether the host name is end with a dot. If yes, treat it
488 // the same as no-dot unless the input comparison domain is end
490 const char* last_pos
= spec_
.data() + parsed_
.host
.end() - 1;
491 int host_len
= parsed_
.host
.len
;
492 if ('.' == *last_pos
&& '.' != lower_ascii_domain
[domain_len
- 1]) {
497 // Return false if host's length is less than domain's length.
498 if (host_len
< domain_len
)
501 // Compare this url whether belong specific domain.
502 const char* start_pos
= spec_
.data() + parsed_
.host
.begin
+
503 host_len
- domain_len
;
505 if (!url_util::LowerCaseEqualsASCII(start_pos
,
508 lower_ascii_domain
+ domain_len
))
511 // Check whether host has right domain start with dot, make sure we got
512 // right domain range. For example www.google.com has domain
513 // "google.com" but www.iamnotgoogle.com does not.
514 if ('.' != lower_ascii_domain
[0] && host_len
> domain_len
&&
515 '.' != *(start_pos
- 1))
521 void GURL::Swap(GURL
* other
) {
522 spec_
.swap(other
->spec_
);
523 std::swap(is_valid_
, other
->is_valid_
);
524 std::swap(parsed_
, other
->parsed_
);
525 inner_url_
.swap(other
->inner_url_
);
528 std::ostream
& operator<<(std::ostream
& out
, const GURL
& url
) {
529 return out
<< url
.possibly_invalid_spec();