Cerner LZW Decompression

DecompressBlob.cs

using System.Text;

/* This code provided by Bruce Jackson (brucejackson.info) and is 
 * provided "AS IS" with no warranties, expressed or implied, as to
 * its effectivness or use.  
 * 
 * User agrees to edit the code to meet their partiular use and 
 * understands that this code has not been fully 
 * tested in a production environment.
 * 
 * The author expressly waives any rights of copyright to the code
 * and so any user is welcome to use this in whatever way they want.
 * */

namespace OcfLzw
{
    /// <summary>
    /// Alternate implementation of the OCF/LZW Decompression by Bruce Jackson (brucejackson.info).
    /// This code has not been tested in production and is provided for reference only.
    /// </summary>
    public class DecompressBlob
    {
        class lzwItem
        {
            public lzwItem(uint _prefix, uint _sufix)
            {
                Prefix = _prefix;
                Suffix = _sufix;
            }
            public uint Prefix;
            public uint Suffix;
        }

        private const int MAX_CODES = 8192;
        private uint[] tempDecompressBuffer = new uint[MAX_CODES];
        private lzwItem[] lzwLookupTable = new lzwItem[MAX_CODES];
        private int tempBufferIndex = 0;
        private int currentByteBufferIndex = 0;
        private int codeCount = 257;
        private byte[] finalByteBuffer = null;

        public byte[] Decompress(byte[] rawbytes, int stringSize)
        {
            finalByteBuffer = new byte[stringSize];

            int byteArrayIndex = 0,
                    shift = 1,
                    currentShift = 1;
            uint prevCode = 0,
                    middleCode = 0,
                    lookupIndex = 0,
                    firstCode = rawbytes[byteArrayIndex];

            while (true)
            {
                if (currentShift >= 9)
                {
                    currentShift -= 8;
                    if (firstCode != 0)
                    {
                        middleCode = rawbytes[++byteArrayIndex];
                        firstCode = (firstCode << currentShift + 8) | (middleCode << currentShift);
                        middleCode = rawbytes[++byteArrayIndex];
                        uint tempCode = middleCode >> (8 - currentShift);
                        lookupIndex = firstCode | tempCode;
                        goto skipit;
                    }
                    else
                    {
                        firstCode = rawbytes[++byteArrayIndex];
                        middleCode = rawbytes[++byteArrayIndex];
                    }
                }
                else
                {
                    middleCode = rawbytes[++byteArrayIndex];
                }

                lookupIndex = (firstCode << currentShift) | (middleCode >> 8 - currentShift);

                if (lookupIndex == 256)  // time to move to a new lookup table
                {
                    shift = 1;
                    currentShift++;
                    firstCode = rawbytes[byteArrayIndex];
                    tempDecompressBuffer = new uint[MAX_CODES];
                    tempBufferIndex = 0;
                    lzwLookupTable = new lzwItem[MAX_CODES];
                    codeCount = 257;
                    continue;
                }
                else if (lookupIndex == 257) // EOF marker, better than using the string size
                {
                    return finalByteBuffer;
                }

                skipit:
                if (prevCode == 0)
                {
                    tempDecompressBuffer[0] = lookupIndex;
                }
                if (lookupIndex < codeCount)
                {
                    SaveItemToLookupTable(lookupIndex);
                    if (codeCount < MAX_CODES)
                    {
                        lzwLookupTable[codeCount++] = new lzwItem(prevCode, tempDecompressBuffer[tempBufferIndex]);
                    }
                }
                else
                {
                    lzwLookupTable[codeCount++] = new lzwItem(prevCode, tempDecompressBuffer[tempBufferIndex]);
                    SaveItemToLookupTable(lookupIndex);
                }

                firstCode = (uint)(middleCode & (0xff >> currentShift));
                currentShift += shift;

                switch (codeCount)  // use the lookup table size and not the current byte count
                {
                    case 511:
                    case 1023:
                    case 2047:
                    case 4095:
                        shift++;
                        currentShift++;
                        break;
                }
                prevCode = lookupIndex;
            }
        }
        private void SaveItemToLookupTable(uint compressedCode)
        {
            tempBufferIndex = -1;
            while (compressedCode >= 258)
            {
                tempDecompressBuffer[++tempBufferIndex] = lzwLookupTable[compressedCode].Suffix;
                compressedCode = lzwLookupTable[compressedCode].Prefix;
            }
            tempDecompressBuffer[++tempBufferIndex] = compressedCode;
            for (int i = tempBufferIndex; i >= 0; i--)
            {
                finalByteBuffer[currentByteBufferIndex++] = (byte)tempDecompressBuffer[i];
            }
        }
    }
}

Just wanted to say thank you, and provide a python port https://github.com/plessbd/ocflzw-decompress

Thanks @plessbd, and thanks for sharing your port.

@pgodwin thank you for sharing this script. Have you any idea where I can get the stringSize from in Cerner db tables?

Thank you! Here's a clojure port, unprofiled just yet.

@thegoatherder: working on the CE_BLOB table, BLOB_LENGTH seems to return the uncompressed length when the blob is compressed (and the concatenated length when the blob is split into a sequence?). The docs aren't helpful unfortunately. blobgetlen() returns the actual length of BLOB_CONTENTS.

Thanks for sharing the script. It has been extremely helpful.
One issue we are having is joining split Blobs. ie where the Blob sequence number >1. Has anyone else tackled this?

@pgodwin Thank you for this! I also ported it to Java in case this is helpful to anyone else.

https://github.com/ARMoir/CernerBlobJava/blob/main/src/ocflzw/DecompressBlob.java