mirror of
https://github.com/Barubary/dsdecmp.git
synced 2024-08-20 06:49:14 +02:00
460 lines
21 KiB
C#
460 lines
21 KiB
C#
using System;
|
|
using System.Collections.Generic;
|
|
using System.Text;
|
|
using System.IO;
|
|
|
|
namespace DSDecmp.Formats.Nitro
|
|
{
|
|
/// <summary>
|
|
/// Compressor and decompressor for the LZ-0x10 format used in many of the games for the
|
|
/// newer Nintendo consoles and handhelds.
|
|
/// </summary>
|
|
public class LZ10 : NitroCFormat
|
|
{
|
|
private static bool lookAhead = false;
|
|
/// <summary>
|
|
/// Sets the flag that determines if 'look-ahead'/DP should be used when compressing
|
|
/// with the LZ-10 format. The default is false, which is what is used in the original
|
|
/// implementation.
|
|
/// </summary>
|
|
public static bool LookAhead
|
|
{
|
|
set { lookAhead = value; }
|
|
}
|
|
|
|
public LZ10() : base(0x10) { }
|
|
|
|
#region 'Original' Decompression method
|
|
/// <summary>
|
|
/// Decompress a stream that is compressed in the LZ-10 format.
|
|
/// </summary>
|
|
/// <param name="instream">The compressed stream.</param>
|
|
/// <param name="inLength">The length of the input stream.</param>
|
|
/// <param name="outstream">The output stream, where the decompressed data is written to.</param>
|
|
public override long Decompress(Stream instream, long inLength,
|
|
Stream outstream)
|
|
{
|
|
#region format definition form GBATEK/NDSTEK
|
|
/* Data header (32bit)
|
|
Bit 0-3 Reserved
|
|
Bit 4-7 Compressed type (must be 1 for LZ77)
|
|
Bit 8-31 Size of decompressed data
|
|
Repeat below. Each Flag Byte followed by eight Blocks.
|
|
Flag data (8bit)
|
|
Bit 0-7 Type Flags for next 8 Blocks, MSB first
|
|
Block Type 0 - Uncompressed - Copy 1 Byte from Source to Dest
|
|
Bit 0-7 One data byte to be copied to dest
|
|
Block Type 1 - Compressed - Copy N+3 Bytes from Dest-Disp-1 to Dest
|
|
Bit 0-3 Disp MSBs
|
|
Bit 4-7 Number of bytes to copy (minus 3)
|
|
Bit 8-15 Disp LSBs
|
|
*/
|
|
#endregion
|
|
|
|
long readBytes = 0;
|
|
|
|
byte type = (byte)instream.ReadByte();
|
|
if (type != base.magicByte)
|
|
throw new InvalidDataException("The provided stream is not a valid LZ-0x10 "
|
|
+ "compressed stream (invalid type 0x" + type.ToString("X") + ")");
|
|
byte[] sizeBytes = new byte[3];
|
|
instream.Read(sizeBytes, 0, 3);
|
|
int decompressedSize = base.Bytes2Size(sizeBytes);
|
|
readBytes += 4;
|
|
if (decompressedSize == 0)
|
|
{
|
|
sizeBytes = new byte[4];
|
|
instream.Read(sizeBytes, 0, 4);
|
|
decompressedSize = base.Bytes2Size(sizeBytes);
|
|
readBytes += 4;
|
|
}
|
|
|
|
// the maximum 'DISP-1' is 0xFFF.
|
|
int bufferLength = 0x1000;
|
|
byte[] buffer = new byte[bufferLength];
|
|
int bufferOffset = 0;
|
|
|
|
|
|
int currentOutSize = 0;
|
|
int flags = 0, mask = 1;
|
|
while (currentOutSize < decompressedSize)
|
|
{
|
|
// (throws when requested new flags byte is not available)
|
|
#region Update the mask. If all flag bits have been read, get a new set.
|
|
// the current mask is the mask used in the previous run. So if it masks the
|
|
// last flag bit, get a new flags byte.
|
|
if (mask == 1)
|
|
{
|
|
if (readBytes >= inLength)
|
|
throw new NotEnoughDataException(currentOutSize, decompressedSize);
|
|
flags = instream.ReadByte(); readBytes++;
|
|
if (flags < 0)
|
|
throw new StreamTooShortException();
|
|
mask = 0x80;
|
|
}
|
|
else
|
|
{
|
|
mask >>= 1;
|
|
}
|
|
#endregion
|
|
|
|
// bit = 1 <=> compressed.
|
|
if ((flags & mask) > 0)
|
|
{
|
|
// (throws when < 2 bytes are available)
|
|
#region Get length and displacement('disp') values from next 2 bytes
|
|
// there are < 2 bytes available when the end is at most 1 byte away
|
|
if (readBytes + 1 >= inLength)
|
|
{
|
|
// make sure the stream is at the end
|
|
if (readBytes < inLength)
|
|
{
|
|
instream.ReadByte(); readBytes++;
|
|
}
|
|
throw new NotEnoughDataException(currentOutSize, decompressedSize);
|
|
}
|
|
int byte1 = instream.ReadByte(); readBytes++;
|
|
int byte2 = instream.ReadByte(); readBytes++;
|
|
if (byte2 < 0)
|
|
throw new StreamTooShortException();
|
|
|
|
// the number of bytes to copy
|
|
int length = byte1 >> 4;
|
|
length += 3;
|
|
|
|
// from where the bytes should be copied (relatively)
|
|
int disp = ((byte1 & 0x0F) << 8) | byte2;
|
|
disp += 1;
|
|
|
|
if (disp > currentOutSize)
|
|
throw new InvalidDataException("Cannot go back more than already written. "
|
|
+ "DISP = 0x" + disp.ToString("X") + ", #written bytes = 0x" + currentOutSize.ToString("X")
|
|
+ " at 0x" + instream.Position.ToString("X"));
|
|
#endregion
|
|
|
|
int bufIdx = bufferOffset + bufferLength - disp;
|
|
for (int i = 0; i < length; i++)
|
|
{
|
|
byte next = buffer[bufIdx % bufferLength];
|
|
bufIdx++;
|
|
outstream.WriteByte(next);
|
|
buffer[bufferOffset] = next;
|
|
bufferOffset = (bufferOffset + 1) % bufferLength;
|
|
}
|
|
currentOutSize += length;
|
|
}
|
|
else
|
|
{
|
|
if (readBytes >= inLength)
|
|
throw new NotEnoughDataException(currentOutSize, decompressedSize);
|
|
int next = instream.ReadByte(); readBytes++;
|
|
if (next < 0)
|
|
throw new StreamTooShortException();
|
|
|
|
currentOutSize++;
|
|
outstream.WriteByte((byte)next);
|
|
buffer[bufferOffset] = (byte)next;
|
|
bufferOffset = (bufferOffset + 1) % bufferLength;
|
|
}
|
|
outstream.Flush();
|
|
}
|
|
|
|
if (readBytes < inLength)
|
|
{
|
|
// the input may be 4-byte aligned.
|
|
if ((readBytes ^ (readBytes & 3)) + 4 < inLength)
|
|
throw new TooMuchInputException(readBytes, inLength);
|
|
}
|
|
|
|
return decompressedSize;
|
|
}
|
|
#endregion
|
|
|
|
#region Original Compress method
|
|
public unsafe override int Compress(Stream instream, long inLength, Stream outstream)
|
|
{
|
|
// make sure the decompressed size fits in 3 bytes.
|
|
// There should be room for four bytes, however I'm not 100% sure if that can be used
|
|
// in every game, as it may not be a built-in function.
|
|
if (inLength > 0xFFFFFF)
|
|
throw new InputTooLargeException();
|
|
|
|
// use the other method if lookahead is enabled
|
|
if (lookAhead)
|
|
{
|
|
return CompressWithLA(instream, inLength, outstream);
|
|
}
|
|
|
|
// save the input data in an array to prevent having to go back and forth in a file
|
|
byte[] indata = new byte[inLength];
|
|
int numReadBytes = instream.Read(indata, 0, (int)inLength);
|
|
if (numReadBytes != inLength)
|
|
throw new StreamTooShortException();
|
|
|
|
// write the compression header first
|
|
outstream.WriteByte(0x10);
|
|
outstream.WriteByte((byte)(inLength & 0xFF));
|
|
outstream.WriteByte((byte)((inLength >> 8) & 0xFF));
|
|
outstream.WriteByte((byte)((inLength >> 16) & 0xFF));
|
|
|
|
int compressedLength = 4;
|
|
|
|
fixed (byte* instart = &indata[0])
|
|
{
|
|
// we do need to buffer the output, as the first byte indicates which blocks are compressed.
|
|
// this version does not use a look-ahead, so we do not need to buffer more than 8 blocks at a time.
|
|
byte[] outbuffer = new byte[8 * 2 + 1];
|
|
outbuffer[0] = 0;
|
|
int bufferlength = 1, bufferedBlocks = 0;
|
|
int readBytes = 0;
|
|
while (readBytes < inLength)
|
|
{
|
|
#region If 8 blocks are bufferd, write them and reset the buffer
|
|
// we can only buffer 8 blocks at a time.
|
|
if (bufferedBlocks == 8)
|
|
{
|
|
outstream.Write(outbuffer, 0, bufferlength);
|
|
compressedLength += bufferlength;
|
|
// reset the buffer
|
|
outbuffer[0] = 0;
|
|
bufferlength = 1;
|
|
bufferedBlocks = 0;
|
|
}
|
|
#endregion
|
|
|
|
// determine if we're dealing with a compressed or raw block.
|
|
// it is a compressed block when the next 3 or more bytes can be copied from
|
|
// somewhere in the set of already compressed bytes.
|
|
int disp;
|
|
int oldLength = Math.Min(readBytes, 0x1000);
|
|
int length = this.GetOccurrenceLength(instart + readBytes, (int)Math.Min(inLength - readBytes, 0x12),
|
|
instart + readBytes - oldLength, oldLength, out disp);
|
|
|
|
// length not 3 or more? next byte is raw data
|
|
if (length < 3)
|
|
{
|
|
outbuffer[bufferlength++] = *(instart + (readBytes++));
|
|
}
|
|
else
|
|
{
|
|
// 3 or more bytes can be copied? next (length) bytes will be compressed into 2 bytes
|
|
readBytes += length;
|
|
|
|
// mark the next block as compressed
|
|
outbuffer[0] |= (byte)(1 << (7 - bufferedBlocks));
|
|
|
|
outbuffer[bufferlength] = (byte)(((length - 3) << 4) & 0xF0);
|
|
outbuffer[bufferlength] |= (byte)(((disp - 1) >> 8) & 0x0F);
|
|
bufferlength++;
|
|
outbuffer[bufferlength] = (byte)((disp - 1) & 0xFF);
|
|
bufferlength++;
|
|
}
|
|
bufferedBlocks++;
|
|
}
|
|
|
|
// copy the remaining blocks to the output
|
|
if (bufferedBlocks > 0)
|
|
{
|
|
outstream.Write(outbuffer, 0, bufferlength);
|
|
compressedLength += bufferlength;
|
|
/*/ make the compressed file 4-byte aligned.
|
|
while ((compressedLength % 4) != 0)
|
|
{
|
|
outstream.WriteByte(0);
|
|
compressedLength++;
|
|
}/**/
|
|
}
|
|
}
|
|
|
|
return compressedLength;
|
|
}
|
|
#endregion
|
|
|
|
#region Dynamic Programming compression method
|
|
/// <summary>
|
|
/// Variation of the original compression method, making use of Dynamic Programming to 'look ahead'
|
|
/// and determine the optimal 'length' values for the compressed blocks. Is not 100% optimal,
|
|
/// as the flag-bytes are not taken into account.
|
|
/// </summary>
|
|
private unsafe int CompressWithLA(Stream instream, long inLength, Stream outstream)
|
|
{
|
|
// save the input data in an array to prevent having to go back and forth in a file
|
|
byte[] indata = new byte[inLength];
|
|
int numReadBytes = instream.Read(indata, 0, (int)inLength);
|
|
if (numReadBytes != inLength)
|
|
throw new StreamTooShortException();
|
|
|
|
// write the compression header first
|
|
outstream.WriteByte(0x10);
|
|
outstream.WriteByte((byte)(inLength & 0xFF));
|
|
outstream.WriteByte((byte)((inLength >> 8) & 0xFF));
|
|
outstream.WriteByte((byte)((inLength >> 16) & 0xFF));
|
|
|
|
int compressedLength = 4;
|
|
|
|
fixed (byte* instart = &indata[0])
|
|
{
|
|
// we do need to buffer the output, as the first byte indicates which blocks are compressed.
|
|
// this version does not use a look-ahead, so we do not need to buffer more than 8 blocks at a time.
|
|
byte[] outbuffer = new byte[8 * 2 + 1];
|
|
outbuffer[0] = 0;
|
|
int bufferlength = 1, bufferedBlocks = 0;
|
|
int readBytes = 0;
|
|
|
|
// get the optimal choices for len and disp
|
|
int[] lengths, disps;
|
|
this.GetOptimalCompressionLengths(instart, indata.Length, out lengths, out disps);
|
|
while (readBytes < inLength)
|
|
{
|
|
// we can only buffer 8 blocks at a time.
|
|
if (bufferedBlocks == 8)
|
|
{
|
|
outstream.Write(outbuffer, 0, bufferlength);
|
|
compressedLength += bufferlength;
|
|
// reset the buffer
|
|
outbuffer[0] = 0;
|
|
bufferlength = 1;
|
|
bufferedBlocks = 0;
|
|
}
|
|
|
|
|
|
if (lengths[readBytes] == 1)
|
|
{
|
|
outbuffer[bufferlength++] = *(instart + (readBytes++));
|
|
}
|
|
else
|
|
{
|
|
// mark the next block as compressed
|
|
outbuffer[0] |= (byte)(1 << (7 - bufferedBlocks));
|
|
|
|
outbuffer[bufferlength] = (byte)(((lengths[readBytes] - 3) << 4) & 0xF0);
|
|
outbuffer[bufferlength] |= (byte)(((disps[readBytes] - 1) >> 8) & 0x0F);
|
|
bufferlength++;
|
|
outbuffer[bufferlength] = (byte)((disps[readBytes] - 1) & 0xFF);
|
|
bufferlength++;
|
|
|
|
readBytes += lengths[readBytes];
|
|
}
|
|
|
|
|
|
bufferedBlocks++;
|
|
}
|
|
|
|
// copy the remaining blocks to the output
|
|
if (bufferedBlocks > 0)
|
|
{
|
|
outstream.Write(outbuffer, 0, bufferlength);
|
|
compressedLength += bufferlength;
|
|
/*/ make the compressed file 4-byte aligned.
|
|
while ((compressedLength % 4) != 0)
|
|
{
|
|
outstream.WriteByte(0);
|
|
compressedLength++;
|
|
}/**/
|
|
}
|
|
}
|
|
|
|
return compressedLength;
|
|
}
|
|
#endregion
|
|
|
|
#region DP compression helper method; GetOptimalCompressionLengths
|
|
/// <summary>
|
|
/// Gets the optimal compression lengths for each start of a compressed block using Dynamic Programming.
|
|
/// This takes O(n^2) time.
|
|
/// </summary>
|
|
/// <param name="indata">The data to compress.</param>
|
|
/// <param name="inLength">The length of the data to compress.</param>
|
|
/// <param name="lengths">The optimal 'length' of the compressed blocks. For each byte in the input data,
|
|
/// this value is the optimal 'length' value. If it is 1, the block should not be compressed.</param>
|
|
/// <param name="disps">The 'disp' values of the compressed blocks. May be 0, in which case the
|
|
/// corresponding length will never be anything other than 1.</param>
|
|
private unsafe void GetOptimalCompressionLengths(byte* indata, int inLength, out int[] lengths, out int[] disps)
|
|
{
|
|
lengths = new int[inLength];
|
|
disps = new int[inLength];
|
|
int[] minLengths = new int[inLength];
|
|
|
|
for (int i = inLength - 1; i >= 0; i--)
|
|
{
|
|
// first get the compression length when the next byte is not compressed
|
|
minLengths[i] = int.MaxValue;
|
|
lengths[i] = 1;
|
|
if (i + 1 >= inLength)
|
|
minLengths[i] = 1;
|
|
else
|
|
minLengths[i] = 1 + minLengths[i + 1];
|
|
// then the optimal compressed length
|
|
int oldLength = Math.Min(0x1000, i);
|
|
// get the appropriate disp while at it. Takes at most O(n) time if oldLength is considered O(n)
|
|
// be sure to bound the input length with 0x12, as that's the maximum length for LZ-10 compressed blocks.
|
|
int maxLen = GetOccurrenceLength(indata + i, Math.Min(inLength - i, 0x12),
|
|
indata + i - oldLength, oldLength, out disps[i]);
|
|
if (disps[i] > i)
|
|
throw new Exception("disp is too large");
|
|
for (int j = 3; j <= maxLen; j++)
|
|
{
|
|
int newCompLen;
|
|
if (i + j >= inLength)
|
|
newCompLen = 2;
|
|
else
|
|
newCompLen = 2 + minLengths[i + j];
|
|
if (newCompLen < minLengths[i])
|
|
{
|
|
lengths[i] = j;
|
|
minLengths[i] = newCompLen;
|
|
}
|
|
}
|
|
}
|
|
|
|
// we could optimize this further to also optimize it with regard to the flag-bytes, but that would require 8 times
|
|
// more space and time (one for each position in the block) for only a potentially tiny increase in compression ratio.
|
|
}
|
|
#endregion
|
|
|
|
/// <summary>
|
|
/// Determine the maximum size of a LZ-compressed block starting at newPtr, using the already compressed data
|
|
/// starting at oldPtr. Takes O(inLength * oldLength) = O(n^2) time.
|
|
/// </summary>
|
|
/// <param name="newPtr">The start of the data that needs to be compressed.</param>
|
|
/// <param name="newLength">The number of bytes that still need to be compressed.</param>
|
|
/// <param name="oldPtr">The start of the raw file.</param>
|
|
/// <param name="oldLength">The number of bytes already compressed.</param>
|
|
/// <param name="disp">The offset of the start of the longest block to refer to.</param>
|
|
/// <returns>The length of the longest sequence of bytes that can be copied from the already decompressed data.</returns>
|
|
private unsafe int GetOccurrenceLength(byte* newPtr, int newLength, byte* oldPtr, int oldLength, out int disp)
|
|
{
|
|
disp = 0;
|
|
if (newLength == 0)
|
|
return 0;
|
|
int maxLength = 0;
|
|
// try every possible 'disp' value (disp = oldLength - i)
|
|
for (int i = 0; i < oldLength - 1; i++)
|
|
{
|
|
// work from the start of the old data to the end, to mimic the original implementation's behaviour
|
|
// (and going from start to end or from end to start does not influence the compression ratio anyway)
|
|
byte* currentOldStart = oldPtr + i;
|
|
int currentLength = 0;
|
|
// determine the length we can copy if we go back (oldLength - i) bytes
|
|
// always check the next 'newLength' bytes, and not just the available 'old' bytes,
|
|
// as the copied data can also originate from what we're currently trying to compress.
|
|
for (int j = 0; j < newLength; j++)
|
|
{
|
|
// stop when the bytes are no longer the same
|
|
if (*(currentOldStart + j) != *(newPtr + j))
|
|
break;
|
|
currentLength++;
|
|
}
|
|
|
|
// update the optimal value
|
|
if (currentLength > maxLength)
|
|
{
|
|
maxLength = currentLength;
|
|
disp = oldLength - i;
|
|
}
|
|
}
|
|
return maxLength;
|
|
}
|
|
}
|
|
}
|