Apple的LZF算法解析

    有關LZF算法的相關解析文檔比較少，但是Apple對LZF的開源，可以讓我們對該算法進行一個簡單的解析。LZFSE 基于 Lempel-Ziv ，并使用了有限狀態熵編碼。LZF采用類似lz77和lzss的混合編碼。使用3種“起始標記”來代表每段輸出的數據串。

    接下來看一下開源的LZF算法的實現源碼。

     1.定義的全局字段：

        private readonly long[] _hashTable = new long[Hsize];

        private const uint Hlog = 14;

        private const uint Hsize = (1 << 14);

        private const uint MaxLit = (1 << 5);

        private const uint MaxOff = (1 << 13);

        private const uint MaxRef = ((1 << 8) + (1 << 3));

  2.使用LibLZF算法壓縮數據：

        /// <summary>
        /// 使用LibLZF算法壓縮數據
        /// </summary>
        /// <param name="input">需要壓縮的數據</param>
        /// <param name="inputLength">要壓縮的數據的長度</param>
        /// <param name="output">引用將包含壓縮數據的緩沖區</param>
        /// <param name="outputLength">壓縮緩沖區的長度（應大于輸入緩沖區）</param>
        /// <returns>輸出緩沖區中壓縮歸檔的大小</returns>
        public int Compress(byte[] input, int inputLength, byte[] output, int outputLength)
        {
            Array.Clear(_hashTable, 0, (int)Hsize);
            uint iidx = 0;
            uint oidx = 0;
            var hval = (uint)(((input[iidx]) << 8) | input[iidx + 1]);
            var lit = 0;
            for (; ; )
            {
                if (iidx < inputLength - 2)
                {
                    hval = (hval << 8) | input[iidx + 2];
                    long hslot = ((hval ^ (hval << 5)) >> (int)(((3 * 8 - Hlog)) - hval * 5) & (Hsize - 1));
                    var reference = _hashTable[hslot];
                    _hashTable[hslot] = iidx;
                    long off;
                    if ((off = iidx - reference - 1) < MaxOff
                        && iidx + 4 < inputLength
                        && reference > 0
                        && input[reference + 0] == input[iidx + 0]
                        && input[reference + 1] == input[iidx + 1]
                        && input[reference + 2] == input[iidx + 2]
                        )
                    {
                        uint len = 2;
                        var maxlen = (uint)inputLength - iidx - len;
                        maxlen = maxlen > MaxRef ? MaxRef : maxlen;
                        if (oidx + lit + 1 + 3 >= outputLength)
                            return 0;
                        do
                            len++;
                        while (len < maxlen && input[reference + len] == input[iidx + len]);
                        if (lit != 0)
                        {
                            output[oidx++] = (byte)(lit - 1);
                            lit = -lit;
                            do
                                output[oidx++] = input[iidx + lit];
                            while ((++lit) != 0);
                        }
                        len -= 2;
                        iidx++;
                        if (len < 7)
                        {
                            output[oidx++] = (byte)((off >> 8) + (len << 5));
                        }
                        else
                        {
                            output[oidx++] = (byte)((off >> 8) + (7 << 5));
                            output[oidx++] = (byte)(len - 7);
                        }
                        output[oidx++] = (byte)off;
                        iidx += len - 1;
                        hval = (uint)(((input[iidx]) << 8) | input[iidx + 1]);
                        hval = (hval << 8) | input[iidx + 2];
                        _hashTable[((hval ^ (hval << 5)) >> (int)(((3 * 8 - Hlog)) - hval * 5) & (Hsize - 1))] = iidx;
                        iidx++;
                        hval = (hval << 8) | input[iidx + 2];
                        _hashTable[((hval ^ (hval << 5)) >> (int)(((3 * 8 - Hlog)) - hval * 5) & (Hsize - 1))] = iidx;
                        iidx++;
                        continue;
                    }
                }
                else if (iidx == inputLength)
                    break;
                lit++;
                iidx++;
                if (lit != MaxLit) continue;
                if (oidx + 1 + MaxLit >= outputLength)
                    return 0;

                output[oidx++] = (byte)(MaxLit - 1);
                lit = -lit;
                do
                    output[oidx++] = input[iidx + lit];
                while ((++lit) != 0);
            }
            if (lit == 0) return (int)oidx;
            if (oidx + lit + 1 >= outputLength)
                return 0;
            output[oidx++] = (byte)(lit - 1);
            lit = -lit;
            do
                output[oidx++] = input[iidx + lit];
            while ((++lit) != 0);

            return (int)oidx;
        }

 3.使用LibLZF算法重載壓縮數據：

        /// <summary>
        /// 使用LibLZF算法解壓縮數據
        /// </summary>
        /// <param name="input">參考數據進行解壓縮</param>
        /// <param name="inputLength">要解壓縮的數據的長度</param>
        /// <param name="output">引用包含解壓縮數據的緩沖區</param>
        /// <param name="outputLength">輸出緩沖區中壓縮歸檔的大小</param>
        /// <returns>返回解壓縮大小</returns>
        public int Decompress(byte[] input, int inputLength, byte[] output, int outputLength)
        {
            uint iidx = 0;
            uint oidx = 0;
            do
            {
                uint ctrl = input[iidx++];

                if (ctrl < (1 << 5))
                {
                    ctrl++;

                    if (oidx + ctrl > outputLength)
                    {
                        return 0;
                    }

                    do
                        output[oidx++] = input[iidx++];
                    while ((--ctrl) != 0);
                }
                else
                {
                    var len = ctrl >> 5;
                    var reference = (int)(oidx - ((ctrl & 0x1f) << 8) - 1);
                    if (len == 7)
                        len += input[iidx++];
                    reference -= input[iidx++];
                    if (oidx + len + 2 > outputLength)
                    {
                        return 0;
                    }
                    if (reference < 0)
                    {
                        return 0;
                    }
                    output[oidx++] = output[reference++];
                    output[oidx++] = output[reference++];
                    do
                        output[oidx++] = output[reference++];
                    while ((--len) != 0);
                }
            }
            while (iidx < inputLength);

            return (int)oidx;
        }

   以上是LZF算法的代碼。