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 WebCore {
44
45 static const int maximumValidPortNumber = 0xFFFE;
46 static const int invalidPortNumber = 0xFFFF;
47
assertProtocolIsGood(const char * protocol)48 static void assertProtocolIsGood(const char* protocol)
49 {
50 #ifndef NDEBUG
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.
asURLChar8Subtle(const String & spec)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.
charactersOrEmpty(const StringUTF8Adaptor & string)73 static const char* charactersOrEmpty(const StringUTF8Adaptor& string)
74 {
75 static const char zero = 0;
76 return string.data() ? string.data() : &zero;
77 }
78
isSchemeFirstChar(char c)79 static bool isSchemeFirstChar(char c)
80 {
81 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z');
82 }
83
isSchemeChar(char c)84 static bool isSchemeChar(char c)
85 {
86 return isSchemeFirstChar(c) || (c >= '0' && c <= '9') || c == '.' || c == '-' || c == '+';
87 }
88
isUnicodeEncoding(const WTF::TextEncoding * encoding)89 static bool isUnicodeEncoding(const WTF::TextEncoding* encoding)
90 {
91 return encoding->encodingForFormSubmission() == UTF8Encoding();
92 }
93
94 namespace {
95
96 class KURLCharsetConverter : public url_canon::CharsetConverter {
97 public:
98 // The encoding parameter may be 0, but in this case the object must not be called.
KURLCharsetConverter(const WTF::TextEncoding * encoding)99 explicit KURLCharsetConverter(const WTF::TextEncoding* encoding)
100 : m_encoding(encoding)
101 {
102 }
103
ConvertFromUTF16(const url_parse::UTF16Char * input,int inputLength,url_canon::CanonOutput * output)104 virtual void ConvertFromUTF16(const url_parse::UTF16Char* input, int inputLength, url_canon::CanonOutput* output)
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
isValidProtocol(const String & protocol)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
strippedForUseAsReferrer() const131 String KURL::strippedForUseAsReferrer() const
132 {
133 if (isBlankURL() || protocolIs("data") || protocolIs("javascript"))
134 return String();
135
136 KURL referrer(*this);
137 referrer.setUser(String());
138 referrer.setPass(String());
139 referrer.removeFragmentIdentifier();
140 return referrer.string();
141 }
142
isLocalFile() const143 bool KURL::isLocalFile() const
144 {
145 // Including feed here might be a bad idea since drag and drop uses this check
146 // and including feed would allow feeds to potentially let someone's blog
147 // read the contents of the clipboard on a drag, even without a drop.
148 // Likewise with using the FrameLoader::shouldTreatURLAsLocal() function.
149 return protocolIs("file");
150 }
151
protocolIsJavaScript(const String & url)152 bool protocolIsJavaScript(const String& url)
153 {
154 return protocolIs(url, "javascript");
155 }
156
blankURL()157 const KURL& blankURL()
158 {
159 DEFINE_STATIC_LOCAL(KURL, staticBlankURL, (ParsedURLString, "about:blank"));
160 return staticBlankURL;
161 }
162
isBlankURL() const163 bool KURL::isBlankURL() const
164 {
165 return protocolIs("about");
166 }
167
elidedString() const168 String KURL::elidedString() const
169 {
170 if (string().length() <= 1024)
171 return string();
172
173 return string().left(511) + "..." + string().right(510);
174 }
175
176 // Initializes with a string representing an absolute URL. No encoding
177 // information is specified. This generally happens when a KURL is converted
178 // to a string and then converted back. In this case, the URL is already
179 // canonical and in proper escaped form so needs no encoding. We treat it as
180 // UTF-8 just in case.
KURL(ParsedURLStringTag,const String & url)181 KURL::KURL(ParsedURLStringTag, const String& url)
182 {
183 if (!url.isNull())
184 init(KURL(), url, 0);
185 else {
186 // WebCore expects us to preserve the nullness of strings when this
187 // constructor is used. In all other cases, it expects a non-null
188 // empty string, which is what init() will create.
189 m_isValid = false;
190 m_protocolIsInHTTPFamily = false;
191 }
192 }
193
createIsolated(ParsedURLStringTag,const String & url)194 KURL KURL::createIsolated(ParsedURLStringTag, const String& url)
195 {
196 // FIXME: We should be able to skip this extra copy and created an
197 // isolated KURL more efficiently.
198 return KURL(ParsedURLString, url).copy();
199 }
200
201 // Constructs a new URL given a base URL and a possibly relative input URL.
202 // This assumes UTF-8 encoding.
KURL(const KURL & base,const String & relative)203 KURL::KURL(const KURL& base, const String& relative)
204 {
205 init(base, relative, 0);
206 }
207
208 // Constructs a new URL given a base URL and a possibly relative input URL.
209 // Any query portion of the relative URL will be encoded in the given encoding.
KURL(const KURL & base,const String & relative,const WTF::TextEncoding & encoding)210 KURL::KURL(const KURL& base, const String& relative, const WTF::TextEncoding& encoding)
211 {
212 init(base, relative, &encoding.encodingForFormSubmission());
213 }
214
KURL(const AtomicString & canonicalString,const url_parse::Parsed & parsed,bool isValid)215 KURL::KURL(const AtomicString& canonicalString, const url_parse::Parsed& parsed, bool isValid)
216 : m_isValid(isValid)
217 , m_protocolIsInHTTPFamily(false)
218 , m_parsed(parsed)
219 , m_string(canonicalString)
220 {
221 initProtocolIsInHTTPFamily();
222 initInnerURL();
223 }
224
KURL(WTF::HashTableDeletedValueType)225 KURL::KURL(WTF::HashTableDeletedValueType)
226 : m_isValid(false)
227 , m_protocolIsInHTTPFamily(false)
228 , m_string(WTF::HashTableDeletedValue)
229 {
230 }
231
KURL(const KURL & other)232 KURL::KURL(const KURL& other)
233 : m_isValid(other.m_isValid)
234 , m_protocolIsInHTTPFamily(other.m_protocolIsInHTTPFamily)
235 , m_parsed(other.m_parsed)
236 , m_string(other.m_string)
237 {
238 if (other.m_innerURL.get())
239 m_innerURL = adoptPtr(new KURL(other.m_innerURL->copy()));
240 }
241
operator =(const KURL & other)242 KURL& KURL::operator=(const KURL& other)
243 {
244 m_isValid = other.m_isValid;
245 m_protocolIsInHTTPFamily = other.m_protocolIsInHTTPFamily;
246 m_parsed = other.m_parsed;
247 m_string = other.m_string;
248 if (other.m_innerURL)
249 m_innerURL = adoptPtr(new KURL(other.m_innerURL->copy()));
250 else
251 m_innerURL.clear();
252 return *this;
253 }
254
copy() const255 KURL KURL::copy() const
256 {
257 KURL result;
258 result.m_isValid = m_isValid;
259 result.m_protocolIsInHTTPFamily = m_protocolIsInHTTPFamily;
260 result.m_parsed = m_parsed;
261 result.m_string = m_string.isolatedCopy();
262 if (result.m_innerURL)
263 result.m_innerURL = adoptPtr(new KURL(m_innerURL->copy()));
264 return result;
265 }
266
isNull() const267 bool KURL::isNull() const
268 {
269 return m_string.isNull();
270 }
271
isEmpty() const272 bool KURL::isEmpty() const
273 {
274 return m_string.isEmpty();
275 }
276
isValid() const277 bool KURL::isValid() const
278 {
279 return m_isValid;
280 }
281
hasPort() const282 bool KURL::hasPort() const
283 {
284 return hostEnd() < pathStart();
285 }
286
protocolIsInHTTPFamily() const287 bool KURL::protocolIsInHTTPFamily() const
288 {
289 return m_protocolIsInHTTPFamily;
290 }
291
hasPath() const292 bool KURL::hasPath() const
293 {
294 // Note that http://www.google.com/" has a path, the path is "/". This can
295 // return false only for invalid or nonstandard URLs.
296 return m_parsed.path.len >= 0;
297 }
298
299 // We handle "parameters" separated by a semicolon, while KURL.cpp does not,
300 // which can lead to different results in some cases.
lastPathComponent() const301 String KURL::lastPathComponent() const
302 {
303 if (!m_isValid)
304 return stringForInvalidComponent();
305 ASSERT(!m_string.isNull());
306
307 // When the output ends in a slash, WebCore has different expectations than
308 // the GoogleURL library. For "/foo/bar/" the library will return the empty
309 // string, but WebCore wants "bar".
310 url_parse::Component path = m_parsed.path;
311 if (path.len > 0 && m_string[path.end() - 1] == '/')
312 path.len--;
313
314 url_parse::Component file;
315 if (m_string.is8Bit())
316 url_parse::ExtractFileName(asURLChar8Subtle(m_string), path, &file);
317 else
318 url_parse::ExtractFileName(m_string.characters16(), path, &file);
319
320 // Bug: https://bugs.webkit.org/show_bug.cgi?id=21015 this function returns
321 // a null string when the path is empty, which we duplicate here.
322 if (!file.is_nonempty())
323 return String();
324 return componentString(file);
325 }
326
protocol() const327 String KURL::protocol() const
328 {
329 return componentString(m_parsed.scheme);
330 }
331
host() const332 String KURL::host() const
333 {
334 return componentString(m_parsed.host);
335 }
336
337 // Returns 0 when there is no port.
338 //
339 // We treat URL's with out-of-range port numbers as invalid URLs, and they will
340 // be rejected by the canonicalizer. KURL.cpp will allow them in parsing, but
341 // return invalidPortNumber from this port() function, so we mirror that behavior here.
port() const342 unsigned short KURL::port() const
343 {
344 if (!m_isValid || m_parsed.port.len <= 0)
345 return 0;
346 ASSERT(!m_string.isNull());
347 int port = m_string.is8Bit() ?
348 url_parse::ParsePort(asURLChar8Subtle(m_string), m_parsed.port) :
349 url_parse::ParsePort(m_string.characters16(), m_parsed.port);
350 ASSERT(port != url_parse::PORT_UNSPECIFIED); // Checked port.len <= 0 before.
351
352 if (port == url_parse::PORT_INVALID || port > maximumValidPortNumber) // Mimic KURL::port()
353 port = invalidPortNumber;
354
355 return static_cast<unsigned short>(port);
356 }
357
pass() const358 String KURL::pass() const
359 {
360 // Bug: https://bugs.webkit.org/show_bug.cgi?id=21015 this function returns
361 // a null string when the password is empty, which we duplicate here.
362 if (!m_parsed.password.is_nonempty())
363 return String();
364 return componentString(m_parsed.password);
365 }
366
user() const367 String KURL::user() const
368 {
369 return componentString(m_parsed.username);
370 }
371
fragmentIdentifier() const372 String KURL::fragmentIdentifier() const
373 {
374 // Empty but present refs ("foo.com/bar#") should result in the empty
375 // string, which componentString will produce. Nonexistent refs
376 // should be the null string.
377 if (!m_parsed.ref.is_valid())
378 return String();
379 return componentString(m_parsed.ref);
380 }
381
hasFragmentIdentifier() const382 bool KURL::hasFragmentIdentifier() const
383 {
384 return m_parsed.ref.len >= 0;
385 }
386
baseAsString() const387 String KURL::baseAsString() const
388 {
389 // FIXME: There is probably a more efficient way to do this?
390 return m_string.left(pathAfterLastSlash());
391 }
392
query() const393 String KURL::query() const
394 {
395 if (m_parsed.query.len >= 0)
396 return componentString(m_parsed.query);
397
398 // Bug: https://bugs.webkit.org/show_bug.cgi?id=21015 this function returns
399 // an empty string when the query is empty rather than a null (not sure
400 // which is right).
401 // Returns a null if the query is not specified, instead of empty.
402 if (m_parsed.query.is_valid())
403 return emptyString();
404 return String();
405 }
406
path() const407 String KURL::path() const
408 {
409 return componentString(m_parsed.path);
410 }
411
setProtocol(const String & protocol)412 bool KURL::setProtocol(const String& protocol)
413 {
414 // Firefox and IE remove everything after the first ':'.
415 int separatorPosition = protocol.find(':');
416 String newProtocol = protocol.substring(0, separatorPosition);
417 StringUTF8Adaptor newProtocolUTF8(newProtocol);
418
419 // If KURL is given an invalid scheme, it returns failure without modifying
420 // the URL at all. This is in contrast to most other setters which modify
421 // the URL and set "m_isValid."
422 url_canon::RawCanonOutputT<char> canonProtocol;
423 url_parse::Component protocolComponent;
424 if (!url_canon::CanonicalizeScheme(newProtocolUTF8.data(), url_parse::Component(0, newProtocolUTF8.length()), &canonProtocol, &protocolComponent)
425 || !protocolComponent.is_nonempty())
426 return false;
427
428 url_canon::Replacements<char> replacements;
429 replacements.SetScheme(charactersOrEmpty(newProtocolUTF8), url_parse::Component(0, newProtocolUTF8.length()));
430 replaceComponents(replacements);
431
432 // isValid could be false but we still return true here. This is because
433 // WebCore or JS scripts can build up a URL by setting individual
434 // components, and a JS exception is based on the return value of this
435 // function. We want to throw the exception and stop the script only when
436 // its trying to set a bad protocol, and not when it maybe just hasn't
437 // finished building up its final scheme.
438 return true;
439 }
440
setHost(const String & host)441 void KURL::setHost(const String& host)
442 {
443 StringUTF8Adaptor hostUTF8(host);
444 url_canon::Replacements<char> replacements;
445 replacements.SetHost(charactersOrEmpty(hostUTF8), url_parse::Component(0, hostUTF8.length()));
446 replaceComponents(replacements);
447 }
448
parsePortFromStringPosition(const String & value,unsigned portStart)449 static String parsePortFromStringPosition(const String& value, unsigned portStart)
450 {
451 // "008080junk" needs to be treated as port "8080" and "000" as "0".
452 size_t length = value.length();
453 unsigned portEnd = portStart;
454 while (isASCIIDigit(value[portEnd]) && portEnd < length)
455 ++portEnd;
456 while (value[portStart] == '0' && portStart < portEnd - 1)
457 ++portStart;
458
459 // Required for backwards compat.
460 // https://www.w3.org/Bugs/Public/show_bug.cgi?id=23463
461 if (portStart == portEnd)
462 return "0";
463
464 return value.substring(portStart, portEnd - portStart);
465 }
466
setHostAndPort(const String & hostAndPort)467 void KURL::setHostAndPort(const String& hostAndPort)
468 {
469 size_t separator = hostAndPort.find(':');
470 if (!separator)
471 return;
472
473 if (separator == kNotFound) {
474 url_canon::Replacements<char> replacements;
475 StringUTF8Adaptor hostUTF8(hostAndPort);
476 replacements.SetHost(charactersOrEmpty(hostUTF8), url_parse::Component(0, hostUTF8.length()));
477 replaceComponents(replacements);
478 return;
479 }
480
481 String host = hostAndPort.substring(0, separator);
482 String port = parsePortFromStringPosition(hostAndPort, separator + 1);
483
484 StringUTF8Adaptor hostUTF8(host);
485 StringUTF8Adaptor portUTF8(port);
486
487 url_canon::Replacements<char> replacements;
488 replacements.SetHost(charactersOrEmpty(hostUTF8), url_parse::Component(0, hostUTF8.length()));
489 replacements.SetPort(charactersOrEmpty(portUTF8), url_parse::Component(0, portUTF8.length()));
490 replaceComponents(replacements);
491 }
492
removePort()493 void KURL::removePort()
494 {
495 if (!hasPort())
496 return;
497 url_canon::Replacements<char> replacements;
498 replacements.ClearPort();
499 replaceComponents(replacements);
500 }
501
setPort(const String & port)502 void KURL::setPort(const String& port)
503 {
504 String parsedPort = parsePortFromStringPosition(port, 0);
505 setPort(parsedPort.toUInt());
506 }
507
setPort(unsigned short port)508 void KURL::setPort(unsigned short port)
509 {
510 if (isDefaultPortForProtocol(port, protocol())) {
511 removePort();
512 return;
513 }
514
515 String portString = String::number(port);
516 ASSERT(portString.is8Bit());
517
518 url_canon::Replacements<char> replacements;
519 replacements.SetPort(reinterpret_cast<const char*>(portString.characters8()), url_parse::Component(0, portString.length()));
520 replaceComponents(replacements);
521 }
522
setUser(const String & user)523 void KURL::setUser(const String& user)
524 {
525 // This function is commonly called to clear the username, which we
526 // normally don't have, so we optimize this case.
527 if (user.isEmpty() && !m_parsed.username.is_valid())
528 return;
529
530 // The canonicalizer will clear any usernames that are empty, so we
531 // don't have to explicitly call ClearUsername() here.
532 StringUTF8Adaptor userUTF8(user);
533 url_canon::Replacements<char> replacements;
534 replacements.SetUsername(charactersOrEmpty(userUTF8), url_parse::Component(0, userUTF8.length()));
535 replaceComponents(replacements);
536 }
537
setPass(const String & pass)538 void KURL::setPass(const String& pass)
539 {
540 // This function is commonly called to clear the password, which we
541 // normally don't have, so we optimize this case.
542 if (pass.isEmpty() && !m_parsed.password.is_valid())
543 return;
544
545 // The canonicalizer will clear any passwords that are empty, so we
546 // don't have to explicitly call ClearUsername() here.
547 StringUTF8Adaptor passUTF8(pass);
548 url_canon::Replacements<char> replacements;
549 replacements.SetPassword(charactersOrEmpty(passUTF8), url_parse::Component(0, passUTF8.length()));
550 replaceComponents(replacements);
551 }
552
setFragmentIdentifier(const String & fragment)553 void KURL::setFragmentIdentifier(const String& fragment)
554 {
555 // This function is commonly called to clear the ref, which we
556 // normally don't have, so we optimize this case.
557 if (fragment.isNull() && !m_parsed.ref.is_valid())
558 return;
559
560 StringUTF8Adaptor fragmentUTF8(fragment);
561
562 url_canon::Replacements<char> replacements;
563 if (fragment.isNull())
564 replacements.ClearRef();
565 else
566 replacements.SetRef(charactersOrEmpty(fragmentUTF8), url_parse::Component(0, fragmentUTF8.length()));
567 replaceComponents(replacements);
568 }
569
removeFragmentIdentifier()570 void KURL::removeFragmentIdentifier()
571 {
572 url_canon::Replacements<char> replacements;
573 replacements.ClearRef();
574 replaceComponents(replacements);
575 }
576
setQuery(const String & query)577 void KURL::setQuery(const String& query)
578 {
579 StringUTF8Adaptor queryUTF8(query);
580 url_canon::Replacements<char> replacements;
581 if (query.isNull()) {
582 // KURL.cpp sets to null to clear any query.
583 replacements.ClearQuery();
584 } else if (query.length() > 0 && query[0] == '?') {
585 // WebCore expects the query string to begin with a question mark, but
586 // GoogleURL doesn't. So we trim off the question mark when setting.
587 replacements.SetQuery(charactersOrEmpty(queryUTF8), url_parse::Component(1, queryUTF8.length() - 1));
588 } else {
589 // When set with the empty string or something that doesn't begin with
590 // a question mark, KURL.cpp will add a question mark for you. The only
591 // way this isn't compatible is if you call this function with an empty
592 // string. KURL.cpp will leave a '?' with nothing following it in the
593 // URL, whereas we'll clear it.
594 // FIXME We should eliminate this difference.
595 replacements.SetQuery(charactersOrEmpty(queryUTF8), url_parse::Component(0, queryUTF8.length()));
596 }
597 replaceComponents(replacements);
598 }
599
setPath(const String & path)600 void KURL::setPath(const String& path)
601 {
602 // Empty paths will be canonicalized to "/", so we don't have to worry
603 // about calling ClearPath().
604 StringUTF8Adaptor pathUTF8(path);
605 url_canon::Replacements<char> replacements;
606 replacements.SetPath(charactersOrEmpty(pathUTF8), url_parse::Component(0, pathUTF8.length()));
607 replaceComponents(replacements);
608 }
609
decodeURLEscapeSequences(const String & string)610 String decodeURLEscapeSequences(const String& string)
611 {
612 return decodeURLEscapeSequences(string, UTF8Encoding());
613 }
614
615 // In KURL.cpp's implementation, this is called by every component getter.
616 // It will unescape every character, including '\0'. This is scary, and may
617 // cause security holes. We never call this function for components, and
618 // just return the ASCII versions instead.
619 //
620 // This function is also used to decode javascript: URLs and as a general
621 // purpose unescaping function.
622 //
623 // FIXME These should be merged to the KURL.cpp implementation.
decodeURLEscapeSequences(const String & string,const WTF::TextEncoding & encoding)624 String decodeURLEscapeSequences(const String& string, const WTF::TextEncoding& encoding)
625 {
626 // FIXME We can probably use KURL.cpp's version of this function
627 // without modification. However, I'm concerned about
628 // https://bugs.webkit.org/show_bug.cgi?id=20559 so am keeping this old
629 // custom code for now. Using their version will also fix the bug that
630 // we ignore the encoding.
631 //
632 // FIXME b/1350291: This does not get called very often. We just convert
633 // first to 8-bit UTF-8, then unescape, then back to 16-bit. This kind of
634 // sucks, and we don't use the encoding properly, which will make some
635 // obscure anchor navigations fail.
636 StringUTF8Adaptor stringUTF8(string);
637 url_canon::RawCanonOutputT<url_parse::UTF16Char> unescaped;
638 url_util::DecodeURLEscapeSequences(stringUTF8.data(), stringUTF8.length(), &unescaped);
639 return StringImpl::create8BitIfPossible(reinterpret_cast<UChar*>(unescaped.data()), unescaped.length());
640 }
641
encodeWithURLEscapeSequences(const String & notEncodedString)642 String encodeWithURLEscapeSequences(const String& notEncodedString)
643 {
644 CString utf8 = UTF8Encoding().normalizeAndEncode(notEncodedString, WTF::URLEncodedEntitiesForUnencodables);
645
646 url_canon::RawCanonOutputT<char> buffer;
647 int inputLength = utf8.length();
648 if (buffer.length() < inputLength * 3)
649 buffer.Resize(inputLength * 3);
650
651 url_util::EncodeURIComponent(utf8.data(), inputLength, &buffer);
652 String escaped(buffer.data(), buffer.length());
653 // Unescape '/'; it's safe and much prettier.
654 escaped.replace("%2F", "/");
655 return escaped;
656 }
657
isHierarchical() const658 bool KURL::isHierarchical() const
659 {
660 if (m_string.isNull() || !m_parsed.scheme.is_nonempty())
661 return false;
662 return m_string.is8Bit() ?
663 url_util::IsStandard(asURLChar8Subtle(m_string), m_parsed.scheme) :
664 url_util::IsStandard(m_string.characters16(), m_parsed.scheme);
665 }
666
667 #ifndef NDEBUG
print() const668 void KURL::print() const
669 {
670 printf("%s\n", m_string.utf8().data());
671 }
672 #endif
673
equalIgnoringFragmentIdentifier(const KURL & a,const KURL & b)674 bool equalIgnoringFragmentIdentifier(const KURL& a, const KURL& b)
675 {
676 // Compute the length of each URL without its ref. Note that the reference
677 // begin (if it exists) points to the character *after* the '#', so we need
678 // to subtract one.
679 int aLength = a.m_string.length();
680 if (a.m_parsed.ref.len >= 0)
681 aLength = a.m_parsed.ref.begin - 1;
682
683 int bLength = b.m_string.length();
684 if (b.m_parsed.ref.len >= 0)
685 bLength = b.m_parsed.ref.begin - 1;
686
687 if (aLength != bLength)
688 return false;
689
690 const String& aString = a.m_string;
691 const String& bString = b.m_string;
692 // FIXME: Abstraction this into a function in WTFString.h.
693 for (int i = 0; i < aLength; ++i) {
694 if (aString[i] != bString[i])
695 return false;
696 }
697 return true;
698 }
699
hostStart() const700 unsigned KURL::hostStart() const
701 {
702 return m_parsed.CountCharactersBefore(url_parse::Parsed::HOST, false);
703 }
704
hostEnd() const705 unsigned KURL::hostEnd() const
706 {
707 return m_parsed.CountCharactersBefore(url_parse::Parsed::PORT, true);
708 }
709
pathStart() const710 unsigned KURL::pathStart() const
711 {
712 return m_parsed.CountCharactersBefore(url_parse::Parsed::PATH, false);
713 }
714
pathEnd() const715 unsigned KURL::pathEnd() const
716 {
717 return m_parsed.CountCharactersBefore(url_parse::Parsed::QUERY, true);
718 }
719
pathAfterLastSlash() const720 unsigned KURL::pathAfterLastSlash() const
721 {
722 if (m_string.isNull())
723 return 0;
724 if (!m_isValid || !m_parsed.path.is_valid())
725 return m_parsed.CountCharactersBefore(url_parse::Parsed::PATH, false);
726 url_parse::Component filename;
727 if (m_string.is8Bit())
728 url_parse::ExtractFileName(asURLChar8Subtle(m_string), m_parsed.path, &filename);
729 else
730 url_parse::ExtractFileName(m_string.characters16(), m_parsed.path, &filename);
731 return filename.begin;
732 }
733
protocolIs(const String & url,const char * protocol)734 bool protocolIs(const String& url, const char* protocol)
735 {
736 assertProtocolIsGood(protocol);
737 if (url.isNull())
738 return false;
739 if (url.is8Bit())
740 return url_util::FindAndCompareScheme(asURLChar8Subtle(url), url.length(), protocol, 0);
741 return url_util::FindAndCompareScheme(url.characters16(), url.length(), protocol, 0);
742 }
743
init(const KURL & base,const String & relative,const WTF::TextEncoding * queryEncoding)744 void KURL::init(const KURL& base, const String& relative, const WTF::TextEncoding* queryEncoding)
745 {
746 if (!relative.isNull() && relative.is8Bit()) {
747 StringUTF8Adaptor relativeUTF8(relative);
748 init(base, relativeUTF8.data(), relativeUTF8.length(), queryEncoding);
749 } else
750 init(base, relative.characters16(), relative.length(), queryEncoding);
751 initProtocolIsInHTTPFamily();
752 initInnerURL();
753 }
754
755 template <typename CHAR>
init(const KURL & base,const CHAR * relative,int relativeLength,const WTF::TextEncoding * queryEncoding)756 void KURL::init(const KURL& base, const CHAR* relative, int relativeLength, const WTF::TextEncoding* queryEncoding)
757 {
758 // As a performance optimization, we do not use the charset converter
759 // if encoding is UTF-8 or other Unicode encodings. Note that this is
760 // per HTML5 2.5.3 (resolving URL). The URL canonicalizer will be more
761 // efficient with no charset converter object because it can do UTF-8
762 // internally with no extra copies.
763
764 // We feel free to make the charset converter object every time since it's
765 // just a wrapper around a reference.
766 KURLCharsetConverter charsetConverterObject(queryEncoding);
767 KURLCharsetConverter* charsetConverter = (!queryEncoding || isUnicodeEncoding(queryEncoding)) ? 0 : &charsetConverterObject;
768
769 StringUTF8Adaptor baseUTF8(base.string());
770
771 url_canon::RawCanonOutputT<char> output;
772 m_isValid = url_util::ResolveRelative(baseUTF8.data(), baseUTF8.length(), base.m_parsed, relative, relativeLength, charsetConverter, &output, &m_parsed);
773
774 // See FIXME in KURLPrivate in the header. If canonicalization has not
775 // changed the string, we can avoid an extra allocation by using assignment.
776 m_string = AtomicString::fromUTF8(output.data(), output.length());
777 }
778
initInnerURL()779 void KURL::initInnerURL()
780 {
781 if (!m_isValid) {
782 m_innerURL.clear();
783 return;
784 }
785 if (url_parse::Parsed* innerParsed = m_parsed.inner_parsed())
786 m_innerURL = adoptPtr(new KURL(ParsedURLString, m_string.substring(innerParsed->scheme.begin, innerParsed->Length() - innerParsed->scheme.begin)));
787 else
788 m_innerURL.clear();
789 }
790
791 template<typename CHAR>
internalProtocolIs(const url_parse::Component & scheme,const CHAR * spec,const char * protocol)792 bool internalProtocolIs(const url_parse::Component& scheme, const CHAR* spec, const char* protocol)
793 {
794 const CHAR* begin = spec + scheme.begin;
795 const CHAR* end = begin + scheme.len;
796
797 while (begin != end && *protocol) {
798 ASSERT(toASCIILower(*protocol) == *protocol);
799 if (toASCIILower(*begin++) != *protocol++)
800 return false;
801 }
802
803 // Both strings are equal (ignoring case) if and only if all of the characters were equal,
804 // and the end of both has been reached.
805 return begin == end && !*protocol;
806 }
807
808 template<typename CHAR>
checkIfProtocolIsInHTTPFamily(const url_parse::Component & scheme,const CHAR * spec)809 bool checkIfProtocolIsInHTTPFamily(const url_parse::Component& scheme, const CHAR* spec)
810 {
811 if (scheme.len == 4)
812 return internalProtocolIs(scheme, spec, "http");
813 if (scheme.len == 5)
814 return internalProtocolIs(scheme, spec, "https");
815 return false;
816 }
817
initProtocolIsInHTTPFamily()818 void KURL::initProtocolIsInHTTPFamily()
819 {
820 if (!m_isValid) {
821 m_protocolIsInHTTPFamily = false;
822 return;
823 }
824
825 ASSERT(!m_string.isNull());
826 m_protocolIsInHTTPFamily = m_string.is8Bit() ?
827 checkIfProtocolIsInHTTPFamily(m_parsed.scheme, m_string.characters8()) :
828 checkIfProtocolIsInHTTPFamily(m_parsed.scheme, m_string.characters16());
829 }
830
protocolIs(const char * protocol) const831 bool KURL::protocolIs(const char* protocol) const
832 {
833 assertProtocolIsGood(protocol);
834
835 // JavaScript URLs are "valid" and should be executed even if KURL decides they are invalid.
836 // The free function protocolIsJavaScript() should be used instead.
837 // FIXME: Chromium code needs to be fixed for this assert to be enabled. ASSERT(strcmp(protocol, "javascript"));
838
839 if (m_string.isNull() || m_parsed.scheme.len <= 0)
840 return *protocol == '\0';
841
842 return m_string.is8Bit() ?
843 internalProtocolIs(m_parsed.scheme, m_string.characters8(), protocol) :
844 internalProtocolIs(m_parsed.scheme, m_string.characters16(), protocol);
845 }
846
stringForInvalidComponent() const847 String KURL::stringForInvalidComponent() const
848 {
849 if (m_string.isNull())
850 return String();
851 return emptyString();
852 }
853
componentString(const url_parse::Component & component) const854 String KURL::componentString(const url_parse::Component& component) const
855 {
856 if (!m_isValid || component.len <= 0)
857 return stringForInvalidComponent();
858 // begin and len are in terms of bytes which do not match
859 // if string() is UTF-16 and input contains non-ASCII characters.
860 // However, the only part in urlString that can contain non-ASCII
861 // characters is 'ref' at the end of the string. In that case,
862 // begin will always match the actual value and len (in terms of
863 // byte) will be longer than what's needed by 'mid'. However, mid
864 // truncates len to avoid go past the end of a string so that we can
865 // get away without doing anything here.
866 return string().substring(component.begin, component.len);
867 }
868
869 template<typename CHAR>
replaceComponents(const url_canon::Replacements<CHAR> & replacements)870 void KURL::replaceComponents(const url_canon::Replacements<CHAR>& replacements)
871 {
872 url_canon::RawCanonOutputT<char> output;
873 url_parse::Parsed newParsed;
874
875 StringUTF8Adaptor utf8(m_string);
876 m_isValid = url_util::ReplaceComponents(utf8.data(), utf8.length(), m_parsed, replacements, 0, &output, &newParsed);
877
878 m_parsed = newParsed;
879 m_string = AtomicString::fromUTF8(output.data(), output.length());
880 }
881
isSafeToSendToAnotherThread() const882 bool KURL::isSafeToSendToAnotherThread() const
883 {
884 return m_string.isSafeToSendToAnotherThread()
885 && (!m_innerURL || m_innerURL->isSafeToSendToAnotherThread());
886 }
887
888 } // namespace WebCore
889