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.
5 #include "url/url_util.h"
10 #include "base/debug/leak_annotations.h"
11 #include "base/logging.h"
12 #include "base/strings/string_util.h"
13 #include "url/url_canon_internal.h"
14 #include "url/url_file.h"
15 #include "url/url_util_internal.h"
21 const int kNumStandardURLSchemes
= 8;
22 const SchemeWithType kStandardURLSchemes
[kNumStandardURLSchemes
] = {
23 {kHttpScheme
, SCHEME_WITH_PORT
},
24 {kHttpsScheme
, SCHEME_WITH_PORT
},
25 // Yes, file URLs can have a hostname, so file URLs should be handled as
26 // "standard". File URLs never have a port as specified by the SchemeType
28 {kFileScheme
, SCHEME_WITHOUT_PORT
},
29 {kFtpScheme
, SCHEME_WITH_PORT
},
30 {kGopherScheme
, SCHEME_WITH_PORT
},
31 {kWsScheme
, SCHEME_WITH_PORT
}, // WebSocket.
32 {kWssScheme
, SCHEME_WITH_PORT
}, // WebSocket secure.
33 {kFileSystemScheme
, SCHEME_WITHOUT_AUTHORITY
},
36 // List of the currently installed standard schemes. This list is lazily
37 // initialized by InitStandardSchemes and is leaked on shutdown to prevent
38 // any destructors from being called that will slow us down or cause problems.
39 std::vector
<SchemeWithType
>* standard_schemes
= NULL
;
41 // See the LockStandardSchemes declaration in the header.
42 bool standard_schemes_locked
= false;
44 // This template converts a given character type to the corresponding
46 template<typename CHAR
> struct CharToStringPiece
{
48 template<> struct CharToStringPiece
<char> {
49 typedef base::StringPiece Piece
;
51 template<> struct CharToStringPiece
<base::char16
> {
52 typedef base::StringPiece16 Piece
;
55 // Ensures that the standard_schemes list is initialized, does nothing if it
56 // already has values.
57 void InitStandardSchemes() {
60 standard_schemes
= new std::vector
<SchemeWithType
>;
61 for (int i
= 0; i
< kNumStandardURLSchemes
; i
++)
62 standard_schemes
->push_back(kStandardURLSchemes
[i
]);
65 // Given a string and a range inside the string, compares it to the given
66 // lower-case |compare_to| buffer.
67 template<typename CHAR
>
68 inline bool DoCompareSchemeComponent(const CHAR
* spec
,
69 const Component
& component
,
70 const char* compare_to
) {
71 if (!component
.is_nonempty())
72 return compare_to
[0] == 0; // When component is empty, match empty scheme.
73 return base::LowerCaseEqualsASCII(
74 typename CharToStringPiece
<CHAR
>::Piece(
75 &spec
[component
.begin
], component
.len
),
79 // Returns true and sets |type| to the SchemeType of the given scheme
80 // identified by |scheme| within |spec| if the scheme is one of the registered
81 // "standard" schemes.
82 template<typename CHAR
>
83 bool DoIsStandard(const CHAR
* spec
,
84 const Component
& scheme
,
86 if (!scheme
.is_nonempty())
87 return false; // Empty or invalid schemes are non-standard.
89 InitStandardSchemes();
90 for (size_t i
= 0; i
< standard_schemes
->size(); i
++) {
91 if (base::LowerCaseEqualsASCII(
92 typename CharToStringPiece
<CHAR
>::Piece(
93 &spec
[scheme
.begin
], scheme
.len
),
94 standard_schemes
->at(i
).scheme
)) {
95 *type
= standard_schemes
->at(i
).type
;
102 template<typename CHAR
>
103 bool DoFindAndCompareScheme(const CHAR
* str
,
106 Component
* found_scheme
) {
107 // Before extracting scheme, canonicalize the URL to remove any whitespace.
108 // This matches the canonicalization done in DoCanonicalize function.
109 RawCanonOutputT
<CHAR
> whitespace_buffer
;
111 const CHAR
* spec
= RemoveURLWhitespace(str
, str_len
,
112 &whitespace_buffer
, &spec_len
);
114 Component our_scheme
;
115 if (!ExtractScheme(spec
, spec_len
, &our_scheme
)) {
118 *found_scheme
= Component();
122 *found_scheme
= our_scheme
;
123 return DoCompareSchemeComponent(spec
, our_scheme
, compare
);
126 template<typename CHAR
>
127 bool DoCanonicalize(const CHAR
* in_spec
,
130 CharsetConverter
* charset_converter
,
132 Parsed
* output_parsed
) {
133 // Remove any whitespace from the middle of the relative URL, possibly
134 // copying to the new buffer.
135 RawCanonOutputT
<CHAR
> whitespace_buffer
;
137 const CHAR
* spec
= RemoveURLWhitespace(in_spec
, in_spec_len
,
138 &whitespace_buffer
, &spec_len
);
142 // For Windows, we allow things that look like absolute Windows paths to be
143 // fixed up magically to file URLs. This is done for IE compatibility. For
144 // example, this will change "c:/foo" into a file URL rather than treating
145 // it as a URL with the protocol "c". It also works for UNC ("\\foo\bar.txt").
146 // There is similar logic in url_canon_relative.cc for
148 // For Max & Unix, we don't do this (the equivalent would be "/foo/bar" which
149 // has no meaning as an absolute path name. This is because browsers on Mac
150 // & Unix don't generally do this, so there is no compatibility reason for
152 if (DoesBeginUNCPath(spec
, 0, spec_len
, false) ||
153 DoesBeginWindowsDriveSpec(spec
, 0, spec_len
)) {
154 ParseFileURL(spec
, spec_len
, &parsed_input
);
155 return CanonicalizeFileURL(spec
, spec_len
, parsed_input
, charset_converter
,
156 output
, output_parsed
);
161 if (!ExtractScheme(spec
, spec_len
, &scheme
))
164 // This is the parsed version of the input URL, we have to canonicalize it
165 // before storing it in our object.
167 SchemeType unused_scheme_type
= SCHEME_WITH_PORT
;
168 if (DoCompareSchemeComponent(spec
, scheme
, url::kFileScheme
)) {
169 // File URLs are special.
170 ParseFileURL(spec
, spec_len
, &parsed_input
);
171 success
= CanonicalizeFileURL(spec
, spec_len
, parsed_input
,
172 charset_converter
, output
, output_parsed
);
173 } else if (DoCompareSchemeComponent(spec
, scheme
, url::kFileSystemScheme
)) {
174 // Filesystem URLs are special.
175 ParseFileSystemURL(spec
, spec_len
, &parsed_input
);
176 success
= CanonicalizeFileSystemURL(spec
, spec_len
, parsed_input
,
177 charset_converter
, output
,
180 } else if (DoIsStandard(spec
, scheme
, &unused_scheme_type
)) {
181 // All "normal" URLs.
182 ParseStandardURL(spec
, spec_len
, &parsed_input
);
183 success
= CanonicalizeStandardURL(spec
, spec_len
, parsed_input
,
184 charset_converter
, output
, output_parsed
);
186 } else if (DoCompareSchemeComponent(spec
, scheme
, url::kMailToScheme
)) {
187 // Mailto URLs are treated like standard URLs, with only a scheme, path,
189 ParseMailtoURL(spec
, spec_len
, &parsed_input
);
190 success
= CanonicalizeMailtoURL(spec
, spec_len
, parsed_input
, output
,
194 // "Weird" URLs like data: and javascript:.
195 ParsePathURL(spec
, spec_len
, trim_path_end
, &parsed_input
);
196 success
= CanonicalizePathURL(spec
, spec_len
, parsed_input
, output
,
202 template<typename CHAR
>
203 bool DoResolveRelative(const char* base_spec
,
205 const Parsed
& base_parsed
,
206 const CHAR
* in_relative
,
207 int in_relative_length
,
208 CharsetConverter
* charset_converter
,
210 Parsed
* output_parsed
) {
211 // Remove any whitespace from the middle of the relative URL, possibly
212 // copying to the new buffer.
213 RawCanonOutputT
<CHAR
> whitespace_buffer
;
215 const CHAR
* relative
= RemoveURLWhitespace(in_relative
, in_relative_length
,
218 bool base_is_authority_based
= false;
219 bool base_is_hierarchical
= false;
221 base_parsed
.scheme
.is_nonempty()) {
222 int after_scheme
= base_parsed
.scheme
.end() + 1; // Skip past the colon.
223 int num_slashes
= CountConsecutiveSlashes(base_spec
, after_scheme
,
225 base_is_authority_based
= num_slashes
> 1;
226 base_is_hierarchical
= num_slashes
> 0;
229 SchemeType unused_scheme_type
= SCHEME_WITH_PORT
;
230 bool standard_base_scheme
=
231 base_parsed
.scheme
.is_nonempty() &&
232 DoIsStandard(base_spec
, base_parsed
.scheme
, &unused_scheme_type
);
235 Component relative_component
;
236 if (!IsRelativeURL(base_spec
, base_parsed
, relative
, relative_length
,
237 (base_is_hierarchical
|| standard_base_scheme
),
238 &is_relative
, &relative_component
)) {
243 // Pretend for a moment that |base_spec| is a standard URL. Normally
244 // non-standard URLs are treated as PathURLs, but if the base has an
245 // authority we would like to preserve it.
246 if (is_relative
&& base_is_authority_based
&& !standard_base_scheme
) {
247 Parsed base_parsed_authority
;
248 ParseStandardURL(base_spec
, base_spec_len
, &base_parsed_authority
);
249 if (base_parsed_authority
.host
.is_nonempty()) {
250 RawCanonOutputT
<char> temporary_output
;
251 bool did_resolve_succeed
=
252 ResolveRelativeURL(base_spec
, base_parsed_authority
, false, relative
,
253 relative_component
, charset_converter
,
254 &temporary_output
, output_parsed
);
255 // The output_parsed is incorrect at this point (because it was built
256 // based on base_parsed_authority instead of base_parsed) and needs to be
258 DoCanonicalize(temporary_output
.data(), temporary_output
.length(), true,
259 charset_converter
, output
, output_parsed
);
260 return did_resolve_succeed
;
262 } else if (is_relative
) {
263 // Relative, resolve and canonicalize.
264 bool file_base_scheme
= base_parsed
.scheme
.is_nonempty() &&
265 DoCompareSchemeComponent(base_spec
, base_parsed
.scheme
, kFileScheme
);
266 return ResolveRelativeURL(base_spec
, base_parsed
, file_base_scheme
, relative
,
267 relative_component
, charset_converter
, output
,
271 // Not relative, canonicalize the input.
272 return DoCanonicalize(relative
, relative_length
, true, charset_converter
,
273 output
, output_parsed
);
276 template<typename CHAR
>
277 bool DoReplaceComponents(const char* spec
,
279 const Parsed
& parsed
,
280 const Replacements
<CHAR
>& replacements
,
281 CharsetConverter
* charset_converter
,
283 Parsed
* out_parsed
) {
284 // If the scheme is overridden, just do a simple string substitution and
285 // re-parse the whole thing. There are lots of edge cases that we really don't
286 // want to deal with. Like what happens if I replace "http://e:8080/foo"
287 // with a file. Does it become "file:///E:/8080/foo" where the port number
288 // becomes part of the path? Parsing that string as a file URL says "yes"
289 // but almost no sane rule for dealing with the components individually would
290 // come up with that.
292 // Why allow these crazy cases at all? Programatically, there is almost no
293 // case for replacing the scheme. The most common case for hitting this is
294 // in JS when building up a URL using the location object. In this case, the
295 // JS code expects the string substitution behavior:
296 // http://www.w3.org/TR/2008/WD-html5-20080610/structured.html#common3
297 if (replacements
.IsSchemeOverridden()) {
298 // Canonicalize the new scheme so it is 8-bit and can be concatenated with
299 // the existing spec.
300 RawCanonOutput
<128> scheme_replaced
;
301 Component scheme_replaced_parsed
;
302 CanonicalizeScheme(replacements
.sources().scheme
,
303 replacements
.components().scheme
,
304 &scheme_replaced
, &scheme_replaced_parsed
);
306 // We can assume that the input is canonicalized, which means it always has
307 // a colon after the scheme (or where the scheme would be).
308 int spec_after_colon
= parsed
.scheme
.is_valid() ? parsed
.scheme
.end() + 1
310 if (spec_len
- spec_after_colon
> 0) {
311 scheme_replaced
.Append(&spec
[spec_after_colon
],
312 spec_len
- spec_after_colon
);
315 // We now need to completely re-parse the resulting string since its meaning
316 // may have changed with the different scheme.
317 RawCanonOutput
<128> recanonicalized
;
318 Parsed recanonicalized_parsed
;
319 DoCanonicalize(scheme_replaced
.data(), scheme_replaced
.length(), true,
321 &recanonicalized
, &recanonicalized_parsed
);
323 // Recurse using the version with the scheme already replaced. This will now
324 // use the replacement rules for the new scheme.
326 // Warning: this code assumes that ReplaceComponents will re-check all
327 // components for validity. This is because we can't fail if DoCanonicalize
328 // failed above since theoretically the thing making it fail could be
329 // getting replaced here. If ReplaceComponents didn't re-check everything,
330 // we wouldn't know if something *not* getting replaced is a problem.
331 // If the scheme-specific replacers are made more intelligent so they don't
332 // re-check everything, we should instead re-canonicalize the whole thing
333 // after this call to check validity (this assumes replacing the scheme is
334 // much much less common than other types of replacements, like clearing the
336 Replacements
<CHAR
> replacements_no_scheme
= replacements
;
337 replacements_no_scheme
.SetScheme(NULL
, Component());
338 return DoReplaceComponents(recanonicalized
.data(), recanonicalized
.length(),
339 recanonicalized_parsed
, replacements_no_scheme
,
340 charset_converter
, output
, out_parsed
);
343 // If we get here, then we know the scheme doesn't need to be replaced, so can
344 // just key off the scheme in the spec to know how to do the replacements.
345 if (DoCompareSchemeComponent(spec
, parsed
.scheme
, url::kFileScheme
)) {
346 return ReplaceFileURL(spec
, parsed
, replacements
, charset_converter
, output
,
349 if (DoCompareSchemeComponent(spec
, parsed
.scheme
, url::kFileSystemScheme
)) {
350 return ReplaceFileSystemURL(spec
, parsed
, replacements
, charset_converter
,
353 SchemeType unused_scheme_type
= SCHEME_WITH_PORT
;
354 if (DoIsStandard(spec
, parsed
.scheme
, &unused_scheme_type
)) {
355 return ReplaceStandardURL(spec
, parsed
, replacements
, charset_converter
,
358 if (DoCompareSchemeComponent(spec
, parsed
.scheme
, url::kMailToScheme
)) {
359 return ReplaceMailtoURL(spec
, parsed
, replacements
, output
, out_parsed
);
362 // Default is a path URL.
363 return ReplacePathURL(spec
, parsed
, replacements
, output
, out_parsed
);
369 InitStandardSchemes();
373 if (standard_schemes
) {
374 delete standard_schemes
;
375 standard_schemes
= NULL
;
379 void AddStandardScheme(const char* new_scheme
,
381 // If this assert triggers, it means you've called AddStandardScheme after
382 // LockStandardSchemes have been called (see the header file for
383 // LockStandardSchemes for more).
385 // This normally means you're trying to set up a new standard scheme too late
386 // in your application's init process. Locate where your app does this
387 // initialization and calls LockStandardSchemes, and add your new standard
389 DCHECK(!standard_schemes_locked
) <<
390 "Trying to add a standard scheme after the list has been locked.";
392 size_t scheme_len
= strlen(new_scheme
);
396 // Duplicate the scheme into a new buffer and add it to the list of standard
397 // schemes. This pointer will be leaked on shutdown.
398 char* dup_scheme
= new char[scheme_len
+ 1];
399 ANNOTATE_LEAKING_OBJECT_PTR(dup_scheme
);
400 memcpy(dup_scheme
, new_scheme
, scheme_len
+ 1);
402 InitStandardSchemes();
403 SchemeWithType scheme_with_type
;
404 scheme_with_type
.scheme
= dup_scheme
;
405 scheme_with_type
.type
= type
;
406 standard_schemes
->push_back(scheme_with_type
);
409 void LockStandardSchemes() {
410 standard_schemes_locked
= true;
413 bool IsStandard(const char* spec
, const Component
& scheme
) {
414 SchemeType unused_scheme_type
;
415 return DoIsStandard(spec
, scheme
, &unused_scheme_type
);
418 bool GetStandardSchemeType(const char* spec
,
419 const Component
& scheme
,
421 return DoIsStandard(spec
, scheme
, type
);
424 bool IsStandard(const base::char16
* spec
, const Component
& scheme
) {
425 SchemeType unused_scheme_type
;
426 return DoIsStandard(spec
, scheme
, &unused_scheme_type
);
429 bool FindAndCompareScheme(const char* str
,
432 Component
* found_scheme
) {
433 return DoFindAndCompareScheme(str
, str_len
, compare
, found_scheme
);
436 bool FindAndCompareScheme(const base::char16
* str
,
439 Component
* found_scheme
) {
440 return DoFindAndCompareScheme(str
, str_len
, compare
, found_scheme
);
443 bool Canonicalize(const char* spec
,
446 CharsetConverter
* charset_converter
,
448 Parsed
* output_parsed
) {
449 return DoCanonicalize(spec
, spec_len
, trim_path_end
, charset_converter
,
450 output
, output_parsed
);
453 bool Canonicalize(const base::char16
* spec
,
456 CharsetConverter
* charset_converter
,
458 Parsed
* output_parsed
) {
459 return DoCanonicalize(spec
, spec_len
, trim_path_end
, charset_converter
,
460 output
, output_parsed
);
463 bool ResolveRelative(const char* base_spec
,
465 const Parsed
& base_parsed
,
466 const char* relative
,
468 CharsetConverter
* charset_converter
,
470 Parsed
* output_parsed
) {
471 return DoResolveRelative(base_spec
, base_spec_len
, base_parsed
,
472 relative
, relative_length
,
473 charset_converter
, output
, output_parsed
);
476 bool ResolveRelative(const char* base_spec
,
478 const Parsed
& base_parsed
,
479 const base::char16
* relative
,
481 CharsetConverter
* charset_converter
,
483 Parsed
* output_parsed
) {
484 return DoResolveRelative(base_spec
, base_spec_len
, base_parsed
,
485 relative
, relative_length
,
486 charset_converter
, output
, output_parsed
);
489 bool ReplaceComponents(const char* spec
,
491 const Parsed
& parsed
,
492 const Replacements
<char>& replacements
,
493 CharsetConverter
* charset_converter
,
495 Parsed
* out_parsed
) {
496 return DoReplaceComponents(spec
, spec_len
, parsed
, replacements
,
497 charset_converter
, output
, out_parsed
);
500 bool ReplaceComponents(const char* spec
,
502 const Parsed
& parsed
,
503 const Replacements
<base::char16
>& replacements
,
504 CharsetConverter
* charset_converter
,
506 Parsed
* out_parsed
) {
507 return DoReplaceComponents(spec
, spec_len
, parsed
, replacements
,
508 charset_converter
, output
, out_parsed
);
511 void DecodeURLEscapeSequences(const char* input
,
513 CanonOutputW
* output
) {
514 RawCanonOutputT
<char> unescaped_chars
;
515 for (int i
= 0; i
< length
; i
++) {
516 if (input
[i
] == '%') {
518 if (DecodeEscaped(input
, &i
, length
, &ch
)) {
519 unescaped_chars
.push_back(ch
);
521 // Invalid escape sequence, copy the percent literal.
522 unescaped_chars
.push_back('%');
525 // Regular non-escaped 8-bit character.
526 unescaped_chars
.push_back(input
[i
]);
530 // Convert that 8-bit to UTF-16. It's not clear IE does this at all to
531 // JavaScript URLs, but Firefox and Safari do.
532 for (int i
= 0; i
< unescaped_chars
.length(); i
++) {
533 unsigned char uch
= static_cast<unsigned char>(unescaped_chars
.at(i
));
535 // Non-UTF-8, just append directly
536 output
->push_back(uch
);
538 // next_ch will point to the last character of the decoded
540 int next_character
= i
;
542 if (ReadUTFChar(unescaped_chars
.data(), &next_character
,
543 unescaped_chars
.length(), &code_point
)) {
544 // Valid UTF-8 character, convert to UTF-16.
545 AppendUTF16Value(code_point
, output
);
548 // If there are any sequences that are not valid UTF-8, we keep
549 // invalid code points and promote to UTF-16. We copy all characters
550 // from the current position to the end of the identified sequence.
551 while (i
< next_character
) {
552 output
->push_back(static_cast<unsigned char>(unescaped_chars
.at(i
)));
555 output
->push_back(static_cast<unsigned char>(unescaped_chars
.at(i
)));
561 void EncodeURIComponent(const char* input
, int length
, CanonOutput
* output
) {
562 for (int i
= 0; i
< length
; ++i
) {
563 unsigned char c
= static_cast<unsigned char>(input
[i
]);
564 if (IsComponentChar(c
))
565 output
->push_back(c
);
567 AppendEscapedChar(c
, output
);
571 bool CompareSchemeComponent(const char* spec
,
572 const Component
& component
,
573 const char* compare_to
) {
574 return DoCompareSchemeComponent(spec
, component
, compare_to
);
577 bool CompareSchemeComponent(const base::char16
* spec
,
578 const Component
& component
,
579 const char* compare_to
) {
580 return DoCompareSchemeComponent(spec
, component
, compare_to
);