kurl.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

    } else if (protocolIs("file")) {
        result.append('/');
        result.append('/');
    }

    append(result, path());

    if (m_pathEnd != m_queryEnd) {
        result.append('?');
        append(result, query());
    }

    if (m_fragmentEnd != m_queryEnd) {
        result.append('#');
        append(result, ref());
    }

    return String::adopt(result);
}

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

String decodeURLEscapeSequences(const String& str, const TextEncoding& encoding)
{
    Vector<UChar> result;

    CharBuffer buffer;

    int length = str.length();
    int decodedPosition = 0;
    int searchPosition = 0;
    int encodedRunPosition;
    while ((encodedRunPosition = str.find('%', searchPosition)) >= 0) {
        // Find the sequence of %-escape codes.
        int encodedRunEnd = encodedRunPosition;
        while (length - encodedRunEnd >= 3
                && str[encodedRunEnd] == '%'
                && isASCIIHexDigit(str[encodedRunEnd + 1])
                && isASCIIHexDigit(str[encodedRunEnd + 2]))
            encodedRunEnd += 3;
        if (encodedRunEnd == encodedRunPosition) {
            ++searchPosition;
            continue;
        }
        searchPosition = encodedRunEnd;

        // Decode the %-escapes into bytes.
        unsigned runLength = (encodedRunEnd - encodedRunPosition) / 3;
        buffer.resize(runLength);
        char* p = buffer.data();
        const UChar* q = str.characters() + encodedRunPosition;
        for (unsigned i = 0; i < runLength; ++i) {
            *p++ = (hexDigitValue(q[1]) << 4) | hexDigitValue(q[2]);
            q += 3;
        }

        // Decode the bytes into Unicode characters.
        String decoded = (encoding.isValid() ? encoding : UTF8Encoding()).decode(buffer.data(), p - buffer.data());
        if (decoded.isEmpty())
            continue;

        // Build up the string with what we just skipped and what we just decoded.
        result.append(str.characters() + decodedPosition, encodedRunPosition - decodedPosition);
        result.append(decoded.characters(), decoded.length());
        decodedPosition = encodedRunEnd;
    }

    result.append(str.characters() + decodedPosition, length - decodedPosition);

    return String::adopt(result);
}

bool KURL::isLocalFile() const
{
    // Including feed here might be a bad idea since drag and drop uses this check
    // and including feed would allow feeds to potentially let someone's blog
    // read the contents of the clipboard on a drag, even without a drop.
    // Likewise with using the FrameLoader::shouldTreatURLAsLocal() function.
    return protocolIs("file");
}

static void appendEscapingBadChars(char*& buffer, const char* strStart, size_t length)
{
    char* p = buffer;

    const char* str = strStart;
    const char* strEnd = strStart + length;
    while (str < strEnd) {
        unsigned char c = *str++;
        if (isBadChar(c)) {
            if (c == '%' || c == '?') {
                *p++ = c;
            } else if (c != 0x09 && c != 0x0a && c != 0x0d) {
                *p++ = '%';
                *p++ = hexDigits[c >> 4];
                *p++ = hexDigits[c & 0xF];
            }
        } else {
            *p++ = c;
        }
    }

    buffer = p;
}

// copy a path, accounting for "." and ".." segments
static int copyPathRemovingDots(char* dst, const char* src, int srcStart, int srcEnd)
{
    char* bufferPathStart = dst;

    // empty path is a special case, and need not have a leading slash
    if (srcStart != srcEnd) {
        const char* baseStringStart = src + srcStart;
        const char* baseStringEnd = src + srcEnd;
        const char* baseStringPos = baseStringStart;

        // this code is unprepared for paths that do not begin with a
        // slash and we should always have one in the source string
        ASSERT(baseStringPos[0] == '/');

        // copy the leading slash into the destination
        *dst = *baseStringPos;
        baseStringPos++;
        dst++;

        while (baseStringPos < baseStringEnd) {
            if (baseStringPos[0] == '.' && dst[-1] == '/') {
                if (baseStringPos[1] == '/' || baseStringPos + 1 == baseStringEnd) {
                    // skip over "." segment
                    baseStringPos += 2;
                    continue;
                } else if (baseStringPos[1] == '.' && (baseStringPos[2] == '/' ||
                                       baseStringPos + 2 == baseStringEnd)) {
                    // skip over ".." segment and rewind the last segment
                    // the RFC leaves it up to the app to decide what to do with excess
                    // ".." segments - we choose to drop them since some web content
                    // relies on this.
                    baseStringPos += 3;
                    if (dst > bufferPathStart + 1)
                        dst--;
                    // Note that these two while blocks differ subtly.
                    // The first helps to remove multiple adjoining slashes as we rewind.
                    // The +1 to bufferPathStart in the first while block prevents eating a leading slash
                    while (dst > bufferPathStart + 1 && dst[-1] == '/')
                        dst--;
                    while (dst > bufferPathStart && dst[-1] != '/')
                        dst--;
                    continue;
                }
            }

            *dst = *baseStringPos;
            baseStringPos++;
            dst++;
        }
    }
    *dst = '\0';
    return dst - bufferPathStart;
}

static inline bool hasSlashDotOrDotDot(const char* str)
{
    const unsigned char* p = reinterpret_cast<const unsigned char*>(str);
    if (!*p)
        return false;
    unsigned char pc = *p;
    while (unsigned char c = *++p) {
        if (c == '.' && (pc == '/' || pc == '.'))
            return true;
        pc = c;
    }
    return false;
}

static inline bool matchLetter(char c, char lowercaseLetter)
{
    return (c | 0x20) == lowercaseLetter;
}

void KURL::parse(const String& string)
{
    checkEncodedString(string);

    CharBuffer buffer(string.length() + 1);
    copyASCII(string.characters(), string.length(), buffer.data());
    buffer[string.length()] = '\0';
    parse(buffer.data(), &string);
}

void KURL::parse(const char* url, const String* originalString)
{
    if (!url || url[0] == '\0') {
        // valid URL must be non-empty
        m_string = originalString ? *originalString : url;
        invalidate();
        return;
    }

    if (!isSchemeFirstChar(url[0])) {
        // scheme must start with an alphabetic character
        m_string = originalString ? *originalString : url;
        invalidate();
        return;
    }

    int schemeEnd = 0;
    while (isSchemeChar(url[schemeEnd]))
        schemeEnd++;

    if (url[schemeEnd] != ':') {
        m_string = originalString ? *originalString : url;
        invalidate();
        return;
    }

    int userStart = schemeEnd + 1;
    int userEnd;
    int passwordStart;
    int passwordEnd;
    int hostStart;
    int hostEnd;
    int portStart;
    int portEnd;

    bool hierarchical = url[schemeEnd + 1] == '/';

    bool isFile = schemeEnd == 4
        && matchLetter(url[0], 'f')
        && matchLetter(url[1], 'i')
        && matchLetter(url[2], 'l')
        && matchLetter(url[3], 'e');

    m_protocolInHTTPFamily = matchLetter(url[0], 'h')
        && matchLetter(url[1], 't')
        && matchLetter(url[2], 't')
        && matchLetter(url[3], 'p')
        && (url[4] == ':' || (matchLetter(url[4], 's') && url[5] == ':'));

    if (hierarchical && url[schemeEnd + 2] == '/') {
        // The part after the scheme is either a net_path or an abs_path whose first path segment is empty.
        // Attempt to find an authority.

        // FIXME: Authority characters may be scanned twice, and it would be nice to be faster.
        userStart += 2;
        userEnd = userStart;

        int colonPos = 0;
        while (isUserInfoChar(url[userEnd])) {
            if (url[userEnd] == ':' && colonPos == 0)
                colonPos = userEnd;
            userEnd++;
        }

        if (url[userEnd] == '@') {
            // actual end of the userinfo, start on the host
            if (colonPos != 0) {
                passwordEnd = userEnd;
                userEnd = colonPos;
                passwordStart = colonPos + 1;
            } else
                passwordStart = passwordEnd = userEnd;

            hostStart = passwordEnd + 1;
        } else if (url[userEnd] == '[' || isPathSegmentEndChar(url[userEnd])) {
            // hit the end of the authority, must have been no user
            // or looks like an IPv6 hostname
            // either way, try to parse it as a hostname
            userEnd = userStart;
            passwordStart = passwordEnd = userEnd;
            hostStart = userStart;
        } else {
            // invalid character
            m_string = originalString ? *originalString : url;
            invalidate();
            return;
        }

        hostEnd = hostStart;

        // IPV6 IP address
        if (url[hostEnd] == '[') {
            hostEnd++;
            while (isIPv6Char(url[hostEnd]))
                hostEnd++;
            if (url[hostEnd] == ']')
                hostEnd++;
            else {
                // invalid character
                m_string = originalString ? *originalString : url;
                invalidate();
                return;
            }
        } else {
            while (isHostnameChar(url[hostEnd]))
                hostEnd++;
        }
        
        if (url[hostEnd] == ':') {
            portStart = portEnd = hostEnd + 1;
 
            // possible start of port
            portEnd = portStart;
            while (isASCIIDigit(url[portEnd]))
                portEnd++;
        } else
            portStart = portEnd = hostEnd;

        if (!isPathSegmentEndChar(url[portEnd])) {
            // invalid character
            m_string = originalString ? *originalString : url;
            invalidate();
            return;
        }

        if (userStart == portEnd && !m_protocolInHTTPFamily && !isFile) {
            // No authority found, which means that this is not a net_path, but rather an abs_path whose first two
            // path segments are empty. For file, http and https only, an empty authority is allowed.
            userStart -= 2;
            userEnd = userStart;
            passwordStart = userEnd;
            passwordEnd = passwordStart;
            hostStart = passwordEnd;
            hostEnd = hostStart;
            portStart = hostEnd;
            portEnd = hostEnd;
        }
    } else {
        // the part after the scheme must be an opaque_part or an abs_path
        userEnd = userStart;
        passwordStart = passwordEnd = userEnd;
        hostStart = hostEnd = passwordEnd;
        portStart = portEnd = hostEnd;
    }

    int pathStart = portEnd;
    int pathEnd = pathStart;
    while (url[pathEnd] && url[pathEnd] != '?' && url[pathEnd] != '#')
        pathEnd++;

    int queryStart = pathEnd;
    int queryEnd = queryStart;
    if (url[queryStart] == '?') {
        while (url[queryEnd] && url[queryEnd] != '#')
            queryEnd++;
    }

    int fragmentStart = queryEnd;
    int fragmentEnd = fragmentStart;
    if (url[fragmentStart] == '#') {
        fragmentStart++;
        fragmentEnd = fragmentStart;
        while (url[fragmentEnd])
            fragmentEnd++;
    }

    // assemble it all, remembering the real ranges

    Vector<char, 4096> buffer(fragmentEnd * 3 + 1);

    char *p = buffer.data();
    const char *strPtr = url;

    // copy in the scheme
    const char *schemeEndPtr = url + schemeEnd;
    while (strPtr < schemeEndPtr)
        *p++ = *strPtr++;
    m_schemeEnd = p - buffer.data();

    bool hostIsLocalHost = portEnd - userStart == 9
        && matchLetter(url[userStart], 'l')
        && matchLetter(url[userStart+1], 'o')
        && matchLetter(url[userStart+2], 'c')
        && matchLetter(url[userStart+3], 'a')
        && matchLetter(url[userStart+4], 'l')
        && matchLetter(url[userStart+5], 'h')
        && matchLetter(url[userStart+6], 'o')
        && matchLetter(url[userStart+7], 's')
        && matchLetter(url[userStart+8], 't');

    // File URLs need a host part unless it is just file:// or file://localhost
    bool degenFilePath = pathStart == pathEnd && (hostStart == hostEnd || hostIsLocalHost);

    bool haveNonHostAuthorityPart = userStart != userEnd || passwordStart != passwordEnd || portStart != portEnd;

    // add ":" after scheme
    *p++ = ':';

    // if we have at least one authority part or a file URL - add "//" and authority
    if (isFile ? !degenFilePath : (haveNonHostAuthorityPart || hostStart != hostEnd)) {
        *p++ = '/';
        *p++ = '/';

        m_userStart = p - buffer.data();

        // copy in the user
        strPtr = url + userStart;
        const char* userEndPtr = url + userEnd;
        while (strPtr < userEndPtr)
            *p++ = *strPtr++;
        m_userEnd = p - buffer.data();

        // copy in the password
        if (passwordEnd != passwordStart) {
            *p++ = ':';
            strPtr = url + passwordStart;
            const char* passwordEndPtr = url + passwordEnd;
            while (strPtr < passwordEndPtr)
                *p++ = *strPtr++;