source: TFP-WebServer/SpaceWizards.HttpListener/src/System/HexConverter.cs@ 444

// Licensed to the .NET Foundation under one or more agreements.
// The .NET Foundation licenses this file to you under the MIT license.

// ReSharper disable RedundantCast

#nullable disable
using System.Diagnostics;
using System.Runtime.CompilerServices;

namespace System
{
    internal static class HexConverter
    {
        public enum Casing : uint
        {
            // Output [ '0' .. '9' ] and [ 'A' .. 'F' ].
            Upper = 0,

            // Output [ '0' .. '9' ] and [ 'a' .. 'f' ].
            // This works because values in the range [ 0x30 .. 0x39 ] ([ '0' .. '9' ])
            // already have the 0x20 bit set, so ORing them with 0x20 is a no-op,
            // while outputs in the range [ 0x41 .. 0x46 ] ([ 'A' .. 'F' ])
            // don't have the 0x20 bit set, so ORing them maps to
            // [ 0x61 .. 0x66 ] ([ 'a' .. 'f' ]), which is what we want.
            Lower = 0x2020U,
        }

        // We want to pack the incoming byte into a single integer [ 0000 HHHH 0000 LLLL ],
        // where HHHH and LLLL are the high and low nibbles of the incoming byte. Then
        // subtract this integer from a constant minuend as shown below.
        //
        //   [ 1000 1001 1000 1001 ]
        // - [ 0000 HHHH 0000 LLLL ]
        // =========================
        //   [ *YYY **** *ZZZ **** ]
        //
        // The end result of this is that YYY is 0b000 if HHHH <= 9, and YYY is 0b111 if HHHH >= 10.
        // Similarly, ZZZ is 0b000 if LLLL <= 9, and ZZZ is 0b111 if LLLL >= 10.
        // (We don't care about the value of asterisked bits.)
        //
        // To turn a nibble in the range [ 0 .. 9 ] into hex, we calculate hex := nibble + 48 (ascii '0').
        // To turn a nibble in the range [ 10 .. 15 ] into hex, we calculate hex := nibble - 10 + 65 (ascii 'A').
        //                                                                => hex := nibble + 55.
        // The difference in the starting ASCII offset is (55 - 48) = 7, depending on whether the nibble is <= 9 or >= 10.
        // Since 7 is 0b111, this conveniently matches the YYY or ZZZ value computed during the earlier subtraction.

        // The commented out code below is code that directly implements the logic described above.

        // uint packedOriginalValues = (((uint)value & 0xF0U) << 4) + ((uint)value & 0x0FU);
        // uint difference = 0x8989U - packedOriginalValues;
        // uint add7Mask = (difference & 0x7070U) >> 4; // line YYY and ZZZ back up with the packed values
        // uint packedResult = packedOriginalValues + add7Mask + 0x3030U /* ascii '0' */;

        // The code below is equivalent to the commented out code above but has been tweaked
        // to allow codegen to make some extra optimizations.

        // The low byte of the packed result contains the hex representation of the incoming byte's low nibble.
        // The adjacent byte of the packed result contains the hex representation of the incoming byte's high nibble.

        // Finally, write to the output buffer starting with the *highest* index so that codegen can
        // elide all but the first bounds check. (This only works if 'startingIndex' is a compile-time constant.)

        // The JIT can elide bounds checks if 'startingIndex' is constant and if the caller is
        // writing to a span of known length (or the caller has already checked the bounds of the
        // furthest access).
        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static void ToBytesBuffer(byte value, Span<byte> buffer, int startingIndex = 0, Casing casing = Casing.Upper)
        {
            uint difference = (((uint)value & 0xF0U) << 4) + ((uint)value & 0x0FU) - 0x8989U;
            uint packedResult = ((((uint)(-(int)difference) & 0x7070U) >> 4) + difference + 0xB9B9U) | (uint)casing;

            buffer[startingIndex + 1] = (byte)packedResult;
            buffer[startingIndex] = (byte)(packedResult >> 8);
        }

#if ALLOW_PARTIALLY_TRUSTED_CALLERS
        [System.Security.SecuritySafeCriticalAttribute]
#endif
        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static void ToCharsBuffer(byte value, Span<char> buffer, int startingIndex = 0, Casing casing = Casing.Upper)
        {
            uint difference = (((uint)value & 0xF0U) << 4) + ((uint)value & 0x0FU) - 0x8989U;
            uint packedResult = ((((uint)(-(int)difference) & 0x7070U) >> 4) + difference + 0xB9B9U) | (uint)casing;

            buffer[startingIndex + 1] = (char)(packedResult & 0xFF);
            buffer[startingIndex] = (char)(packedResult >> 8);
        }

        public static void EncodeToUtf16(ReadOnlySpan<byte> bytes, Span<char> chars, Casing casing = Casing.Upper)
        {
            Debug.Assert(chars.Length >= bytes.Length * 2);

            for (int pos = 0; pos < bytes.Length; ++pos)
            {
                ToCharsBuffer(bytes[pos], chars, pos * 2, casing);
            }
        }

#if !UNITY_NETFRAMEWORK
#if ALLOW_PARTIALLY_TRUSTED_CALLERS
        [System.Security.SecuritySafeCriticalAttribute]
#endif
        public static unsafe string ToString(ReadOnlySpan<byte> bytes, Casing casing = Casing.Upper)
        {
#if NETFRAMEWORK || NETSTANDARD1_0 || NETSTANDARD1_3 || NETSTANDARD2_0
            Span<char> result = stackalloc char[0];
            if (bytes.Length > 16)
            {
                var array = new char[bytes.Length * 2];
                result = array.AsSpan();
            }
            else
            {
                result = stackalloc char[bytes.Length * 2];
            }

            int pos = 0;
            foreach (byte b in bytes)
            {
                ToCharsBuffer(b, result, pos, casing);
                pos += 2;
            }
            return result.ToString();
#else
            fixed (byte* bytesPtr = bytes)
            {
                return string.Create(bytes.Length * 2, (Ptr: (IntPtr)bytesPtr, bytes.Length, casing), static (chars, args) =>
                {
                    var ros = new ReadOnlySpan<byte>((byte*)args.Ptr, args.Length);
                    EncodeToUtf16(ros, chars, args.casing);
                });
            }
#endif
        }
#endif

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static char ToCharUpper(int value)
        {
            value &= 0xF;
            value += '0';

            if (value > '9')
            {
                value += ('A' - ('9' + 1));
            }

            return (char)value;
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static char ToCharLower(int value)
        {
            value &= 0xF;
            value += '0';

            if (value > '9')
            {
                value += ('a' - ('9' + 1));
            }

            return (char)value;
        }

        public static bool TryDecodeFromUtf16(ReadOnlySpan<char> chars, Span<byte> bytes)
        {
            return TryDecodeFromUtf16(chars, bytes, out _);
        }

        public static bool TryDecodeFromUtf16(ReadOnlySpan<char> chars, Span<byte> bytes, out int charsProcessed)
        {
            Debug.Assert(chars.Length % 2 == 0, "Un-even number of characters provided");
            Debug.Assert(chars.Length / 2 == bytes.Length, "Target buffer not right-sized for provided characters");

            int i = 0;
            int j = 0;
            int byteLo = 0;
            int byteHi = 0;
            while (j < bytes.Length)
            {
                byteLo = FromChar(chars[i + 1]);
                byteHi = FromChar(chars[i]);

                // byteHi hasn't been shifted to the high half yet, so the only way the bitwise or produces this pattern
                // is if either byteHi or byteLo was not a hex character.
                if ((byteLo | byteHi) == 0xFF)
                    break;

                bytes[j++] = (byte)((byteHi << 4) | byteLo);
                i += 2;
            }

            if (byteLo == 0xFF)
                i++;

            charsProcessed = i;
            return (byteLo | byteHi) != 0xFF;
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static int FromChar(int c)
        {
            return c >= CharToHexLookup.Length ? 0xFF : CharToHexLookup[c];
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static int FromUpperChar(int c)
        {
            return c > 71 ? 0xFF : CharToHexLookup[c];
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static int FromLowerChar(int c)
        {
            if ((uint)(c - '0') <= '9' - '0')
                return c - '0';

            if ((uint)(c - 'a') <= 'f' - 'a')
                return c - 'a' + 10;

            return 0xFF;
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static bool IsHexChar(int c)
        {
            return FromChar(c) != 0xFF;
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static bool IsHexUpperChar(int c)
        {
            return (uint)(c - '0') <= 9 || (uint)(c - 'A') <= ('F' - 'A');
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static bool IsHexLowerChar(int c)
        {
            return (uint)(c - '0') <= 9 || (uint)(c - 'a') <= ('f' - 'a');
        }

        /// <summary>Map from an ASCII char to its hex value, e.g. arr['b'] == 11. 0xFF means it's not a hex digit.</summary>
        public static ReadOnlySpan<byte> CharToHexLookup => new byte[]
        {
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 15
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 31
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 47
            0x0,  0x1,  0x2,  0x3,  0x4,  0x5,  0x6,  0x7,  0x8,  0x9,  0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 63
            0xFF, 0xA,  0xB,  0xC,  0xD,  0xE,  0xF,  0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 79
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 95
            0xFF, 0xa,  0xb,  0xc,  0xd,  0xe,  0xf,  0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 111
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 127
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 143
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 159
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 175
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 191
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 207
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 223
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 239
            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF  // 255
        };
    }
}