public static void GetUtf16SurrogatesFromSupplementaryPlaneScalar(uint value, out char highSurrogateCodePoint, out char lowSurrogateCodePoint)
{
UnicodeDebug.AssertIsValidSupplementaryPlaneScalar(value);
// This calculation comes from the Unicode specification, Table 3-5.
highSurrogateCodePoint = (char)((value + ((0xD800u - 0x40u) << 10)) >> 10);
lowSurrogateCodePoint = (char)((value & 0x3FFu) + 0xDC00u);
}
/// <summary>
/// Given a Unicode scalar value, gets the number of UTF-16 code units required to represent this value.
/// </summary>
public static int GetUtf16SequenceLength(uint value)
{
UnicodeDebug.AssertIsValidScalar(value);
value -= 0x10000; // if value < 0x10000, high byte = 0xFF; else high byte = 0x00
value += (2 << 24); // if value < 0x10000, high byte = 0x01; else high byte = 0x02
value >>= 24; // shift high byte down
return((int)value); // and return it
}
/// <summary>
/// Returns a Unicode scalar value from two code points representing a UTF-16 surrogate pair.
/// </summary>
public static uint GetScalarFromUtf16SurrogatePair(uint highSurrogateCodePoint, uint lowSurrogateCodePoint)
{
UnicodeDebug.AssertIsHighSurrogateCodePoint(highSurrogateCodePoint);
UnicodeDebug.AssertIsLowSurrogateCodePoint(lowSurrogateCodePoint);
// This calculation comes from the Unicode specification, Table 3-5.
// Need to remove the D800 marker from the high surrogate and the DC00 marker from the low surrogate,
// then fix up the "wwww = uuuuu - 1" section of the bit distribution. The code is written as below
// to become just two instructions: shl, lea.
return((highSurrogateCodePoint << 10) + lowSurrogateCodePoint - ((0xD800U << 10) + 0xDC00U - (1 << 16)));
}
/// <summary>
/// Given a Unicode scalar value, gets the number of UTF-8 code units required to represent this value.
/// </summary>
public static int GetUtf8SequenceLength(uint value)
{
UnicodeDebug.AssertIsValidScalar(value);
// The logic below can handle all valid scalar values branchlessly.
// It gives generally good performance across all inputs, and on x86
// it's only six instructions: lea, sar, xor, add, shr, lea.
// 'a' will be -1 if input is < 0x800; else 'a' will be 0
// => 'a' will be -1 if input is 1 or 2 UTF-8 code units; else 'a' will be 0
int a = ((int)value - 0x0800) >> 31;
// The number of UTF-8 code units for a given scalar is as follows:
// - U+0000..U+007F => 1 code unit
// - U+0080..U+07FF => 2 code units
// - U+0800..U+FFFF => 3 code units
// - U+10000+ => 4 code units
//
// If we XOR the incoming scalar with 0xF800, the chart mutates:
// - U+0000..U+F7FF => 3 code units
// - U+F800..U+F87F => 1 code unit
// - U+F880..U+FFFF => 2 code units
// - U+10000+ => 4 code units
//
// Since the 1- and 3-code unit cases are now clustered, they can
// both be checked together very cheaply.
value ^= 0xF800u;
value -= 0xF880u; // if scalar is 1 or 3 code units, high byte = 0xFF; else high byte = 0x00
value += (4 << 24); // if scalar is 1 or 3 code units, high byte = 0x03; else high byte = 0x04
value >>= 24; // shift high byte down
// Final return value:
// - U+0000..U+007F => 3 + (-1) * 2 = 1
// - U+0080..U+07FF => 4 + (-1) * 2 = 2
// - U+0800..U+FFFF => 3 + ( 0) * 2 = 3
// - U+10000+ => 4 + ( 0) * 2 = 4
return((int)value + (a * 2));
}
// non-validating ctor
private Rune(uint scalarValue, bool unused)
{
UnicodeDebug.AssertIsValidScalar(scalarValue);
_value = scalarValue;
}
/// <summary>
/// Returns the Unicode plane (0 through 16, inclusive) which contains this code point.
/// </summary>
public static int GetPlane(uint codePoint)
{
UnicodeDebug.AssertIsValidCodePoint(codePoint);
return((int)(codePoint >> 16));
}
/// <summary>
/// Decodes the <see cref="Rune"/> at the beginning of the provided UTF-8 source buffer.
/// </summary>
/// <returns>
/// <para>
/// If the source buffer begins with a valid UTF-8 encoded scalar value, returns <see cref="OperationStatus.Done"/>,
/// and outs via <paramref name="result"/> the decoded <see cref="Rune"/> and via <paramref name="bytesConsumed"/> the
/// number of <see langword="byte"/>s used in the input buffer to encode the <see cref="Rune"/>.
/// </para>
/// <para>
/// If the source buffer is empty or contains only a partial UTF-8 subsequence, returns <see cref="OperationStatus.NeedMoreData"/>,
/// and outs via <paramref name="result"/> <see cref="ReplacementChar"/> and via <paramref name="bytesConsumed"/> the length of the input buffer.
/// </para>
/// <para>
/// If the source buffer begins with an ill-formed UTF-8 encoded scalar value, returns <see cref="OperationStatus.InvalidData"/>,
/// and outs via <paramref name="result"/> <see cref="ReplacementChar"/> and via <paramref name="bytesConsumed"/> the number of
/// <see langword="char"/>s used in the input buffer to encode the ill-formed sequence.
/// </para>
/// </returns>
/// <remarks>
/// The general calling convention is to call this method in a loop, slicing the <paramref name="source"/> buffer by
/// <paramref name="bytesConsumed"/> elements on each iteration of the loop. On each iteration of the loop <paramref name="result"/>
/// will contain the real scalar value if successfully decoded, or it will contain <see cref="ReplacementChar"/> if
/// the data could not be successfully decoded. This pattern provides convenient automatic U+FFFD substitution of
/// invalid sequences while iterating through the loop.
/// </remarks>
public static OperationStatus DecodeFromUtf8(ReadOnlySpan <byte> source, out Rune result, out int bytesConsumed)
{
// This method follows the Unicode Standard's recommendation for detecting
// the maximal subpart of an ill-formed subsequence. See The Unicode Standard,
// Ch. 3.9 for more details. In summary, when reporting an invalid subsequence,
// it tries to consume as many code units as possible as long as those code
// units constitute the beginning of a longer well-formed subsequence per Table 3-7.
int index = 0;
// Try reading input[0].
if ((uint)index >= (uint)source.Length)
{
goto NeedsMoreData;
}
uint tempValue = source[index];
if (!UnicodeUtility.IsAsciiCodePoint(tempValue))
{
goto NotAscii;
}
Finish:
bytesConsumed = index + 1;
Debug.Assert(1 <= bytesConsumed && bytesConsumed <= 4); // Valid subsequences are always length [1..4]
result = UnsafeCreate(tempValue);
return(OperationStatus.Done);
NotAscii:
// Per Table 3-7, the beginning of a multibyte sequence must be a code unit in
// the range [C2..F4]. If it's outside of that range, it's either a standalone
// continuation byte, or it's an overlong two-byte sequence, or it's an out-of-range
// four-byte sequence.
if (!UnicodeUtility.IsInRangeInclusive(tempValue, 0xC2, 0xF4))
{
goto FirstByteInvalid;
}
tempValue = (tempValue - 0xC2) << 6;
// Try reading input[1].
index++;
if ((uint)index >= (uint)source.Length)
{
goto NeedsMoreData;
}
// Continuation bytes are of the form [10xxxxxx], which means that their two's
// complement representation is in the range [-65..-128]. This allows us to
// perform a single comparison to see if a byte is a continuation byte.
int thisByteSignExtended = (sbyte)source[index];
if (thisByteSignExtended >= -64)
{
goto Invalid;
}
tempValue += (uint)thisByteSignExtended;
tempValue += 0x80; // remove the continuation byte marker
tempValue += (0xC2 - 0xC0) << 6; // remove the leading byte marker
if (tempValue < 0x0800)
{
Debug.Assert(UnicodeUtility.IsInRangeInclusive(tempValue, 0x0080, 0x07FF));
goto Finish; // this is a valid 2-byte sequence
}
// This appears to be a 3- or 4-byte sequence. Since per Table 3-7 we now have
// enough information (from just two code units) to detect overlong or surrogate
// sequences, we need to perform these checks now.
if (!UnicodeUtility.IsInRangeInclusive(tempValue, ((0xE0 - 0xC0) << 6) + (0xA0 - 0x80), ((0xF4 - 0xC0) << 6) + (0x8F - 0x80)))
{
// The first two bytes were not in the range [[E0 A0]..[F4 8F]].
// This is an overlong 3-byte sequence or an out-of-range 4-byte sequence.
goto Invalid;
}
if (UnicodeUtility.IsInRangeInclusive(tempValue, ((0xED - 0xC0) << 6) + (0xA0 - 0x80), ((0xED - 0xC0) << 6) + (0xBF - 0x80)))
{
// This is a UTF-16 surrogate code point, which is invalid in UTF-8.
goto Invalid;
}
if (UnicodeUtility.IsInRangeInclusive(tempValue, ((0xF0 - 0xC0) << 6) + (0x80 - 0x80), ((0xF0 - 0xC0) << 6) + (0x8F - 0x80)))
{
// This is an overlong 4-byte sequence.
goto Invalid;
}
// The first two bytes were just fine. We don't need to perform any other checks
// on the remaining bytes other than to see that they're valid continuation bytes.
// Try reading input[2].
index++;
if ((uint)index >= (uint)source.Length)
{
goto NeedsMoreData;
}
thisByteSignExtended = (sbyte)source[index];
if (thisByteSignExtended >= -64)
{
goto Invalid; // this byte is not a UTF-8 continuation byte
}
tempValue <<= 6;
tempValue += (uint)thisByteSignExtended;
tempValue += 0x80; // remove the continuation byte marker
tempValue -= (0xE0 - 0xC0) << 12; // remove the leading byte marker
if (tempValue <= 0xFFFF)
{
Debug.Assert(UnicodeUtility.IsInRangeInclusive(tempValue, 0x0800, 0xFFFF));
goto Finish; // this is a valid 3-byte sequence
}
// Try reading input[3].
index++;
if ((uint)index >= (uint)source.Length)
{
goto NeedsMoreData;
}
thisByteSignExtended = (sbyte)source[index];
if (thisByteSignExtended >= -64)
{
goto Invalid; // this byte is not a UTF-8 continuation byte
}
tempValue <<= 6;
tempValue += (uint)thisByteSignExtended;
tempValue += 0x80; // remove the continuation byte marker
tempValue -= (0xF0 - 0xE0) << 18; // remove the leading byte marker
UnicodeDebug.AssertIsValidSupplementaryPlaneScalar(tempValue);
goto Finish; // this is a valid 4-byte sequence
FirstByteInvalid:
index = 1; // Invalid subsequences are always at least length 1.
Invalid:
Debug.Assert(1 <= index && index <= 3); // Invalid subsequences are always length 1..3
bytesConsumed = index;
result = ReplacementChar;
return(OperationStatus.InvalidData);
NeedsMoreData:
Debug.Assert(0 <= index && index <= 3); // Incomplete subsequences are always length 0..3
bytesConsumed = index;
result = ReplacementChar;
return(OperationStatus.NeedMoreData);
}
