void KURL::init(const KURL& base,
const CHAR* relative,
int relativeLength,
const WTF::TextEncoding* queryEncoding) {
// As a performance optimization, we do not use the charset converter
// if encoding is UTF-8 or other Unicode encodings. Note that this is
// per HTML5 2.5.3 (resolving URL). The URL canonicalizer will be more
// efficient with no charset converter object because it can do UTF-8
// internally with no extra copies.
// We feel free to make the charset converter object every time since it's
// just a wrapper around a reference.
KURLCharsetConverter charsetConverterObject(queryEncoding);
KURLCharsetConverter* charsetConverter =
(!queryEncoding || isUnicodeEncoding(queryEncoding))
? 0
: &charsetConverterObject;
StringUTF8Adaptor baseUTF8(base.getString());
url::RawCanonOutputT<char> output;
m_isValid = url::ResolveRelative(baseUTF8.data(), baseUTF8.length(),
base.m_parsed, relative, relativeLength,
charsetConverter, &output, &m_parsed);
// See FIXME in KURLPrivate in the header. If canonicalization has not
// changed the string, we can avoid an extra allocation by using assignment.
m_string = AtomicString::fromUTF8(output.data(), output.length());
}
void KURL::init(const KURL& base,
const String& relative,
const WTF::TextEncoding* queryEncoding) {
// As a performance optimization, we do not use the charset converter
// if encoding is UTF-8 or other Unicode encodings. Note that this is
// per HTML5 2.5.3 (resolving URL). The URL canonicalizer will be more
// efficient with no charset converter object because it can do UTF-8
// internally with no extra copies.
StringUTF8Adaptor baseUTF8(base.getString());
// We feel free to make the charset converter object every time since it's
// just a wrapper around a reference.
KURLCharsetConverter charsetConverterObject(queryEncoding);
KURLCharsetConverter* charsetConverter =
(!queryEncoding || isUnicodeEncoding(queryEncoding))
? 0
: &charsetConverterObject;
// Clamp to int max to avoid overflow.
url::RawCanonOutputT<char> output;
if (!relative.isNull() && relative.is8Bit()) {
StringUTF8Adaptor relativeUTF8(relative);
m_isValid = url::ResolveRelative(baseUTF8.data(), baseUTF8.length(),
base.m_parsed, relativeUTF8.data(),
clampTo<int>(relativeUTF8.length()),
charsetConverter, &output, &m_parsed);
} else {
m_isValid = url::ResolveRelative(baseUTF8.data(), baseUTF8.length(),
base.m_parsed, relative.characters16(),
clampTo<int>(relative.length()),
charsetConverter, &output, &m_parsed);
}
// AtomicString::fromUTF8 will re-hash the raw output and check the
// AtomicStringTable (addWithTranslator) for the string. This can be very
// expensive for large URLs. However, since many URLs are generated from
// existing AtomicStrings (which already have their hashes computed), this
// fast path is used if the input string is already canonicalized.
//
// Because this optimization does not apply to non-AtomicStrings, explicitly
// check that the input is Atomic before moving forward with it. If we mark
// non-Atomic input as Atomic here, we will render the (const) input string
// thread unsafe.
if (!relative.isNull() && relative.impl()->isAtomic() &&
StringView(output.data(), static_cast<unsigned>(output.length())) ==
relative) {
m_string = relative;
} else {
m_string = AtomicString::fromUTF8(output.data(), output.length());
}
initProtocolIsInHTTPFamily();
initInnerURL();
DCHECK_EQ(protocol(), protocol().lower());
}
// Note: code mostly duplicated below.
void KURLGooglePrivate::init(const KURL& base, const char* rel, int relLength,
const TextEncoding* queryEncoding)
{
// As a performance optimization, we do not use the charset converter if
// encoding is UTF-8 or other Unicode encodings. Note that this is
// per HTML5 2.5.3 (resolving URL). The URL canonicalizer will be
// more efficient with no charset converter object because it
// can do UTF-8 internally with no extra copies.
// We feel free to make the charset converter object every time since it's
// just a wrapper around a reference.
KURLCharsetConverter charsetConverterObject(queryEncoding);
KURLCharsetConverter* charsetConverter =
(!queryEncoding || isUnicodeEncoding(queryEncoding)) ? 0 :
&charsetConverterObject;
url_canon::RawCanonOutputT<char> output;
const CString& baseStr = base.m_url.utf8String();
m_isValid = url_util::ResolveRelative(baseStr.data(), baseStr.length(),
base.m_url.m_parsed, rel, relLength,
charsetConverter,
&output, &m_parsed);
// See FIXME in KURLGooglePrivate in the header. If canonicalization has not
// changed the string, we can avoid an extra allocation by using assignment.
//
// When KURL encounters an error such that the URL is invalid and empty
// (for example, resolving a relative URL on a non-hierarchical base), it
// will produce an isNull URL, and calling setUtf8 will produce an empty
// non-null URL. This is unlikely to affect anything, but we preserve this
// just in case.
if (m_isValid || output.length()) {
// Without ref, the whole url is guaranteed to be ASCII-only.
if (m_parsed.ref.is_nonempty())
setUtf8(CString(output.data(), output.length()));
else
setAscii(CString(output.data(), output.length()));
} else {
// WebCore expects resolved URLs to be empty rather than NULL.
setUtf8(CString("", 0));
}
}
// Note: code mostly duplicated above. See FIXMEs and comments there.
void KURLGooglePrivate::init(const KURL& base, const UChar* rel, int relLength,
const TextEncoding* queryEncoding)
{
KURLCharsetConverter charsetConverterObject(queryEncoding);
KURLCharsetConverter* charsetConverter =
(!queryEncoding || isUnicodeEncoding(queryEncoding)) ? 0 :
&charsetConverterObject;
url_canon::RawCanonOutputT<char> output;
const CString& baseStr = base.m_url.utf8String();
m_isValid = url_util::ResolveRelative(baseStr.data(), baseStr.length(),
base.m_url.m_parsed, rel, relLength,
charsetConverter,
&output, &m_parsed);
if (m_isValid || output.length()) {
if (m_parsed.ref.is_nonempty())
setUtf8(CString(output.data(), output.length()));
else
setAscii(CString(output.data(), output.length()));
} else
setUtf8(CString("", 0));
}