kurlgoogle.cpp

linux下开源浏览器WebKit的源码,市面上的很多商用浏览器都是移植自WebKit

    if (ref.isNull())
        replacements.ClearRef();
    else
        replacements.SetRef(CharactersOrEmpty(ref), url_parse::Component(0, ref.length()));
    m_url.replaceComponents(replacements);
}

void KURL::removeRef()
{
    KURLGooglePrivate::Replacements replacements;
    replacements.ClearRef();
    m_url.replaceComponents(replacements);
}

void KURL::setQuery(const String& query)
{
    KURLGooglePrivate::Replacements replacements;
    if (query.isNull()) {
        // KURL.cpp sets to NULL to clear any query.
        replacements.ClearQuery();
    } else if (query.length() > 0 && query[0] == '?') {
        // WebCore expects the query string to begin with a question mark, but
        // GoogleURL doesn't. So we trim off the question mark when setting.
        replacements.SetQuery(CharactersOrEmpty(query),
                              url_parse::Component(1, query.length() - 1));
    } else {
        // When set with the empty string or something that doesn't begin with
        // a question mark, KURL.cpp will add a question mark for you. The only
        // way this isn't compatible is if you call this function with an empty
        // string. KURL.cpp will leave a '?' with nothing following it in the
        // URL, whereas we'll clear it.
        // FIXME We should eliminate this difference.
        replacements.SetQuery(CharactersOrEmpty(query),
                              url_parse::Component(0, query.length()));
    }
    m_url.replaceComponents(replacements);
}

void KURL::setPath(const String& path)
{
    // Empty paths will be canonicalized to "/", so we don't have to worry
    // about calling ClearPath().
    KURLGooglePrivate::Replacements replacements;
    replacements.SetPath(CharactersOrEmpty(path),
                         url_parse::Component(0, path.length()));
    m_url.replaceComponents(replacements);
}

// On Mac, this just seems to return the same URL, but with "/foo/bar" for
// file: URLs instead of file:///foo/bar. We don't bother with any of this,
// at least for now.
String KURL::prettyURL() const
{
    if (!m_url.m_isValid)
        return String();
    return m_url.string();
}

// We copied the KURL version here on Sept 12, 2008 while doing a WebKit
// merge.
// 
// FIXME Somehow share this with KURL? Like we'd theoretically merge with
// decodeURLEscapeSequences below?
String mimeTypeFromDataURL(const String& url)
{
    ASSERT(protocolIs(url, "data"));
    int index = url.find(';');
    if (index == -1)
        index = url.find(',');
    if (index != -1) {
        int len = index - 5;
        if (len > 0)
            return url.substring(5, len);
        return "text/plain"; // Data URLs with no MIME type are considered text/plain.
    }
    return "";
}

String decodeURLEscapeSequences(const String& str)
{
    return decodeURLEscapeSequences(str, UTF8Encoding());
}

// In KURL.cpp's implementation, this is called by every component getter.
// It will unescape every character, including NULL. This is scary, and may
// cause security holes. We never call this function for components, and
// just return the ASCII versions instead.
//
// However, this static function is called directly in some cases. It appears
// that this only happens for javascript: URLs, so this is essentially the
// JavaScript URL decoder. It assumes UTF-8 encoding.
//
// IE doesn't unescape %00, forcing you to use \x00 in JS strings, so we do
// the same. This also eliminates NULL-related problems should a consumer
// incorrectly call this function for non-JavaScript.
//
// FIXME These should be merged to the KURL.cpp implementation.
String decodeURLEscapeSequences(const String& str, const TextEncoding& encoding)
{
    // FIXME We can probably use KURL.cpp's version of this function
    // without modification. However, I'm concerned about
    // https://bugs.webkit.org/show_bug.cgi?id=20559 so am keeping this old
    // custom code for now. Using their version will also fix the bug that
    // we ignore the encoding.
    //
    // FIXME b/1350291: This does not get called very often. We just convert
    // first to 8-bit UTF-8, then unescape, then back to 16-bit. This kind of
    // sucks, and we don't use the encoding properly, which will make some
    // obscure anchor navigations fail.
    CString cstr = str.utf8();

    const char* input = cstr.data();
    int inputLength = cstr.length();
    url_canon::RawCanonOutputT<char> unescaped;
    for (int i = 0; i < inputLength; i++) {
        if (input[i] == '%') {
            unsigned char ch;
            if (url_canon::DecodeEscaped(input, &i, inputLength, &ch)) {
                if (!ch) {
                    // Never unescape NULLs.
                    unescaped.push_back('%');
                    unescaped.push_back('0');
                    unescaped.push_back('0');
                } else
                    unescaped.push_back(ch);
            } else {
                // Invalid escape sequence, copy the percent literal.
                unescaped.push_back('%');
            }
        } else {
            // Regular non-escaped 8-bit character.
            unescaped.push_back(input[i]);
        }
    }

    // Convert that 8-bit to UTF-16. It's not clear IE does this at all to
    // JavaScript URLs, but Firefox and Safari do.
    url_canon::RawCanonOutputT<url_parse::UTF16Char> utf16;
    for (int i = 0; i < unescaped.length(); i++) {
        unsigned char uch = static_cast<unsigned char>(unescaped.at(i));
        if (uch < 0x80) {
            // Non-UTF-8, just append directly
            utf16.push_back(uch);
        } else {
            // next_ch will point to the last character of the decoded
            // character.
            int nextCharacter = i;
            unsigned codePoint;
            if (url_canon::ReadUTFChar(unescaped.data(), &nextCharacter,
                                       unescaped.length(), &codePoint)) {
                // Valid UTF-8 character, convert to UTF-16.
                url_canon::AppendUTF16Value(codePoint, &utf16);
                i = nextCharacter;
            } else {
                // KURL.cpp strips any sequences that are not valid UTF-8. This
                // sounds scary. Instead, we just keep those invalid code
                // points and promote to UTF-16. We copy all characters from
                // the current position to the end of the identified sqeuqnce.
                while (i < nextCharacter) {
                    utf16.push_back(static_cast<unsigned char>(unescaped.at(i)));
                    i++;
                }
                utf16.push_back(static_cast<unsigned char>(unescaped.at(i)));
            }
        }
    }

    return String(reinterpret_cast<UChar*>(utf16.data()), utf16.length());
}

bool KURL::protocolIs(const char* protocol) const
{
    assertProtocolIsGood(protocol);
    if (m_url.m_parsed.scheme.len <= 0)
        return !protocol;
    return lowerCaseEqualsASCII(
        m_url.utf8String().data() + m_url.m_parsed.scheme.begin,
        m_url.utf8String().data() + m_url.m_parsed.scheme.end(),
        protocol);
}

bool KURL::isLocalFile() const
{
    return protocolIs("file");
}

// This is called to escape a URL string. It is only used externally when
// constructing mailto: links to set the query section. Since our query setter
// will automatically do the correct escaping, this function does not have to
// do any work.
//
// There is a possibility that a future called may use this function in other
// ways, and may expect to get a valid URL string. The dangerous thing we want
// to protect against here is accidentally getting NULLs in a string that is
// not supposed to have NULLs. Therefore, we escape NULLs here to prevent this.
String encodeWithURLEscapeSequences(const String& notEncodedString)
{
    CString utf8 = UTF8Encoding().encode(
        reinterpret_cast<const UChar*>(notEncodedString.characters()),
        notEncodedString.length(),
        URLEncodedEntitiesForUnencodables);
    const char* input = utf8.data();
    int inputLength = utf8.length();

    Vector<char, 2048> buffer;
    for (int i = 0; i < inputLength; i++) {
        if (!input[i])
            buffer.append("%00", 3);
        else
            buffer.append(input[i]);
    }
    return String(buffer.data(), buffer.size());
}

bool KURL::isHierarchical() const
{
    if (!m_url.m_parsed.scheme.is_nonempty())
        return false;
    return url_util::IsStandard(
        &m_url.utf8String().data()[m_url.m_parsed.scheme.begin],
        m_url.utf8String().length(),
        m_url.m_parsed.scheme);
}

#ifndef NDEBUG
void KURL::print() const
{
    printf("%s\n", m_url.utf8String().data());
}
#endif

void KURL::invalidate()
{
    // This is only called from the constructor so resetting the (automatically
    // initialized) string and parsed structure would be a waste of time.
    m_url.m_isValid = false;
    m_url.m_protocolInHTTPFamily = false;
}

// Equal up to reference fragments, if any.
bool equalIgnoringRef(const KURL& a, const KURL& b)
{
    // Compute the length of each URL without its ref. Note that the reference
    // begin (if it exists) points to the character *after* the '#', so we need
    // to subtract one.
    int aLength = a.m_url.utf8String().length();
    if (a.m_url.m_parsed.ref.len >= 0)
        aLength = a.m_url.m_parsed.ref.begin - 1;

    int bLength = b.m_url.utf8String().length();
    if (b.m_url.m_parsed.ref.len >= 0)
        bLength = b.m_url.m_parsed.ref.begin - 1;

    return aLength == bLength
        && !strncmp(a.m_url.utf8String().data(), b.m_url.utf8String().data(), aLength);
}

unsigned KURL::hostStart() const
{
    return m_url.m_parsed.CountCharactersBefore(url_parse::Parsed::HOST, false);
}

unsigned KURL::hostEnd() const
{
    return m_url.m_parsed.CountCharactersBefore(url_parse::Parsed::PORT, true);
}

unsigned KURL::pathStart() const
{
    return m_url.m_parsed.CountCharactersBefore(url_parse::Parsed::PATH, false);
}

unsigned KURL::pathEnd() const
{
    return m_url.m_parsed.CountCharactersBefore(url_parse::Parsed::QUERY, true);
}

unsigned KURL::pathAfterLastSlash() const
{
    // When there's no path, ask for what would be the beginning of it.
    if (!m_url.m_parsed.path.is_valid())
        return m_url.m_parsed.CountCharactersBefore(url_parse::Parsed::PATH, false);

    url_parse::Component filename;
    url_parse::ExtractFileName(m_url.utf8String().data(), m_url.m_parsed.path,
                               &filename);
    return filename.begin;
}

const KURL& blankURL()
{
    static KURL staticBlankURL("about:blank");
    return staticBlankURL;
}

bool protocolIs(const String& url, const char* protocol)
{
    // Do the comparison without making a new string object.
    assertProtocolIsGood(protocol);
    for (int i = 0; ; ++i) {
        if (!protocol[i])
            return url[i] == ':';
        if (toASCIILower(url[i]) != protocol[i])
            return false;
    }
}

inline bool KURL::protocolIs(const String& string, const char* protocol)
{
    return WebCore::protocolIs(string, protocol);
}

} // namespace WebCore

#endif // USE(GOOGLEURL)