lzmaencoder.cpp

压缩软件源码

// LZMA/Encoder.cpp

#include "StdAfx.h"

#include "../../../Common/Defs.h"

#include "LZMAEncoder.h"

// for minimal compressing code size define these:
// #define COMPRESS_MF_BT
// #define COMPRESS_MF_BT4

#if !defined(COMPRESS_MF_BT) && !defined(COMPRESS_MF_PAT) && !defined(COMPRESS_MF_HC)
#define COMPRESS_MF_BT
#define COMPRESS_MF_PAT
#define COMPRESS_MF_HC
#endif

#ifdef COMPRESS_MF_BT
#if !defined(COMPRESS_MF_BT2) && !defined(COMPRESS_MF_BT3) && !defined(COMPRESS_MF_BT4) && !defined(COMPRESS_MF_BT4B)
#define COMPRESS_MF_BT2
#define COMPRESS_MF_BT3
#define COMPRESS_MF_BT4
#define COMPRESS_MF_BT4B
#endif
#ifdef COMPRESS_MF_BT2
#include "../LZ/BinTree/BinTree2.h"
#endif
#ifdef COMPRESS_MF_BT3
#include "../LZ/BinTree/BinTree3.h"
#endif
#ifdef COMPRESS_MF_BT4
#include "../LZ/BinTree/BinTree4.h"
#endif
#ifdef COMPRESS_MF_BT4B
#include "../LZ/BinTree/BinTree4b.h"
#endif
#endif

#ifdef COMPRESS_MF_PAT
#include "../LZ/Patricia/Pat2.h"
#include "../LZ/Patricia/Pat2H.h"
#include "../LZ/Patricia/Pat3H.h"
#include "../LZ/Patricia/Pat4H.h"
#include "../LZ/Patricia/Pat2R.h"
#endif

#ifdef COMPRESS_MF_HC
#include "../LZ/HashChain/HC3.h"
#include "../LZ/HashChain/HC4.h"
#endif

#ifdef COMPRESS_MF_MT
#include "../LZ/MT/MT.h"
#endif

namespace NCompress {
namespace NLZMA {

const int kDefaultDictionaryLogSize = 20;
const UInt32 kNumFastBytesDefault = 0x20;

enum 
{
  kBT2,
  kBT3,
  kBT4,
  kBT4B,
  kPat2,
  kPat2H,
  kPat3H,
  kPat4H,
  kPat2R,
  kHC3,
  kHC4
};

static const wchar_t *kMatchFinderIDs[] = 
{
  L"BT2",
  L"BT3",
  L"BT4",
  L"BT4B",
  L"PAT2",
  L"PAT2H",
  L"PAT3H",
  L"PAT4H",
  L"PAT2R",
  L"HC3",
  L"HC4"
};

Byte g_FastPos[1024];

class CFastPosInit
{
public:
  CFastPosInit() { Init(); }
  void Init()
  {
    const Byte kFastSlots = 20;
    int c = 2;
    g_FastPos[0] = 0;
    g_FastPos[1] = 1;

    for (Byte slotFast = 2; slotFast < kFastSlots; slotFast++)
    {
      UInt32 k = (1 << ((slotFast >> 1) - 1));
      for (UInt32 j = 0; j < k; j++, c++)
        g_FastPos[c] = slotFast;
    }
  }
} g_FastPosInit;


void CLiteralEncoder2::Encode(NRangeCoder::CEncoder *rangeEncoder, Byte symbol)
{
  UInt32 context = 1;
  int i = 8;
  do 
  {
    i--;
    UInt32 bit = (symbol >> i) & 1;
    _encoders[context].Encode(rangeEncoder, bit);
    context = (context << 1) | bit;
  }
  while(i != 0);
}

void CLiteralEncoder2::EncodeMatched(NRangeCoder::CEncoder *rangeEncoder, 
    Byte matchByte, Byte symbol)
{
  UInt32 context = 1;
  int i = 8;
  do 
  {
    i--;
    UInt32 bit = (symbol >> i) & 1;
    UInt32 matchBit = (matchByte >> i) & 1;
    _encoders[0x100 + (matchBit << 8) + context].Encode(rangeEncoder, bit);
    context = (context << 1) | bit;
    if (matchBit != bit)
    {
      while(i != 0)
      {
        i--;
        UInt32 bit = (symbol >> i) & 1;
        _encoders[context].Encode(rangeEncoder, bit);
        context = (context << 1) | bit;
      }
      break;
    }
  }
  while(i != 0);
}

UInt32 CLiteralEncoder2::GetPrice(bool matchMode, Byte matchByte, Byte symbol) const
{
  UInt32 price = 0;
  UInt32 context = 1;
  int i = 8;
  if (matchMode)
  {
    do 
    {
      i--;
      UInt32 matchBit = (matchByte >> i) & 1;
      UInt32 bit = (symbol >> i) & 1;
      price += _encoders[0x100 + (matchBit << 8) + context].GetPrice(bit);
      context = (context << 1) | bit;
      if (matchBit != bit)
        break;
    }
    while (i != 0);
  }
  while(i != 0)
  {
    i--;
    UInt32 bit = (symbol >> i) & 1;
    price += _encoders[context].GetPrice(bit);
    context = (context << 1) | bit;
  }
  return price;
};


namespace NLength {

void CEncoder::Init(UInt32 numPosStates)
{
  _choice.Init();
  _choice2.Init();
  for (UInt32 posState = 0; posState < numPosStates; posState++)
  {
    _lowCoder[posState].Init();
    _midCoder[posState].Init();
  }
  _highCoder.Init();
}

void CEncoder::Encode(NRangeCoder::CEncoder *rangeEncoder, UInt32 symbol, UInt32 posState)
{
  if(symbol < kNumLowSymbols)
  {
    _choice.Encode(rangeEncoder, 0);
    _lowCoder[posState].Encode(rangeEncoder, symbol);
  }
  else
  {
    _choice.Encode(rangeEncoder, 1);
    if(symbol < kNumLowSymbols + kNumMidSymbols)
    {
      _choice2.Encode(rangeEncoder, 0);
      _midCoder[posState].Encode(rangeEncoder, symbol - kNumLowSymbols);
    }
    else
    {
      _choice2.Encode(rangeEncoder, 1);
      _highCoder.Encode(rangeEncoder, symbol - kNumLowSymbols - kNumMidSymbols);
    }
  }
}

UInt32 CEncoder::GetPrice(UInt32 symbol, UInt32 posState) const
{
  if(symbol < kNumLowSymbols)
    return _choice.GetPrice0() + _lowCoder[posState].GetPrice(symbol);
  UInt32 price = _choice.GetPrice1();
  if(symbol < kNumLowSymbols + kNumMidSymbols)
  {
    price += _choice2.GetPrice0();
    price += _midCoder[posState].GetPrice(symbol - kNumLowSymbols);
  }
  else
  {
    price += _choice2.GetPrice1();
    price += _highCoder.GetPrice(symbol - kNumLowSymbols - kNumMidSymbols);
  }
  return price;
}

}
CEncoder::CEncoder():
  _numFastBytes(kNumFastBytesDefault),
  _distTableSize(kDefaultDictionaryLogSize * 2),
  _posStateBits(2),
  _posStateMask(4 - 1),
  _numLiteralPosStateBits(0),
  _numLiteralContextBits(3),
  _dictionarySize(1 << kDefaultDictionaryLogSize),
  _dictionarySizePrev(UInt32(-1)),
  _numFastBytesPrev(UInt32(-1)),
  _matchFinderIndex(kBT4),
   #ifdef COMPRESS_MF_MT
  _multiThread(false),
   #endif
  _writeEndMark(false)
{
  _maxMode = false;
  _fastMode = false;
}

HRESULT CEncoder::Create()
{
  if (!_rangeEncoder.Create(1 << 20))
    return E_OUTOFMEMORY;
  if (!_matchFinder)
  {
    switch(_matchFinderIndex)
    {
      #ifdef COMPRESS_MF_BT
      #ifdef COMPRESS_MF_BT2
      case kBT2:
        _matchFinder = new NBT2::CMatchFinderBinTree;
        break;
      #endif
      #ifdef COMPRESS_MF_BT3
      case kBT3:
        _matchFinder = new NBT3::CMatchFinderBinTree;
        break;
      #endif
      #ifdef COMPRESS_MF_BT4
      case kBT4:
        _matchFinder = new NBT4::CMatchFinderBinTree;
        break;
      #endif
      #ifdef COMPRESS_MF_BT4B
      case kBT4B:
        _matchFinder = new NBT4B::CMatchFinderBinTree;
        break;
      #endif
      #endif
      
      #ifdef COMPRESS_MF_PAT
      case kPat2:
        _matchFinder = new NPat2::CPatricia;
        break;
      case kPat2H:
        _matchFinder = new NPat2H::CPatricia;
        break;
      case kPat3H:
        _matchFinder = new NPat3H::CPatricia;
        break;
      case kPat4H:
        _matchFinder = new NPat4H::CPatricia;
        break;
      case kPat2R:
        _matchFinder = new NPat2R::CPatricia;
        break;
      #endif

      #ifdef COMPRESS_MF_HC
      case kHC3:
        _matchFinder = new NHC3::CMatchFinderHC;
        break;
      case kHC4:
        _matchFinder = new NHC4::CMatchFinderHC;
        break;
      #endif
    }
    if (_matchFinder == 0)
      return E_OUTOFMEMORY;

    #ifdef COMPRESS_MF_MT
    if (_multiThread && !(_fastMode && (_matchFinderIndex == kHC3 || _matchFinderIndex == kHC4)))
    {
      CMatchFinderMT *mfSpec = new CMatchFinderMT;
      if (mfSpec == 0)
        return E_OUTOFMEMORY;
      CMyComPtr<IMatchFinder> mf = mfSpec;
      RINOK(mfSpec->SetMatchFinder(_matchFinder));
      _matchFinder.Release();
      _matchFinder = mf;
    }
    #endif
  }
  
  if (!_literalEncoder.Create(_numLiteralPosStateBits, _numLiteralContextBits))
    return E_OUTOFMEMORY;

  if (_dictionarySize == _dictionarySizePrev && _numFastBytesPrev == _numFastBytes)
    return S_OK;
  RINOK(_matchFinder->Create(_dictionarySize, kNumOpts, _numFastBytes, 
      kMatchMaxLen - _numFastBytes));
  _dictionarySizePrev = _dictionarySize;
  _numFastBytesPrev = _numFastBytes;
  return S_OK;
}

static bool AreStringsEqual(const wchar_t *base, const wchar_t *testString)
{
  while (true)
  {
    wchar_t c = *testString;
    if (c >= 'a' && c <= 'z')
      c -= 0x20;
    if (*base != c)
      return false;
    if (c == 0)
      return true;
    base++;
    testString++;
  }
}

static int FindMatchFinder(const wchar_t *s)
{
  for (int m = 0; m < (int)(sizeof(kMatchFinderIDs) / sizeof(kMatchFinderIDs[0])); m++)
    if (AreStringsEqual(kMatchFinderIDs[m], s))
      return m;
  return -1;
}

STDMETHODIMP CEncoder::SetCoderProperties(const PROPID *propIDs, 
    const PROPVARIANT *properties, UInt32 numProperties)
{
  for (UInt32 i = 0; i < numProperties; i++)
  {
    const PROPVARIANT &prop = properties[i];
    switch(propIDs[i])
    {
      case NCoderPropID::kNumFastBytes:
      {
        if (prop.vt != VT_UI4)
          return E_INVALIDARG;
        UInt32 numFastBytes = prop.ulVal;
        if(numFastBytes < 5 || numFastBytes > kMatchMaxLen)
          return E_INVALIDARG;
        _numFastBytes = numFastBytes;
        break;
      }
      case NCoderPropID::kAlgorithm:
      {
        if (prop.vt != VT_UI4)
          return E_INVALIDARG;
        UInt32 maximize = prop.ulVal;
        _fastMode = (maximize == 0); 
        _maxMode = (maximize >= 2);
        break;
      }
      case NCoderPropID::kMatchFinder:
      {
        if (prop.vt != VT_BSTR)
          return E_INVALIDARG;
        int matchFinderIndexPrev = _matchFinderIndex;
        int m = FindMatchFinder(prop.bstrVal);
        if (m < 0)
          return E_INVALIDARG;
        _matchFinderIndex = m;
        if (_matchFinder && matchFinderIndexPrev != _matchFinderIndex)
        {
          _dictionarySizePrev = UInt32(-1);
          _matchFinder.Release();
        }
        break;
      }
      #ifdef COMPRESS_MF_MT
      case NCoderPropID::kMultiThread:
      {
        if (prop.vt != VT_BOOL)
          return E_INVALIDARG;
        bool newMultiThread = (prop.boolVal == VARIANT_TRUE);
        if (newMultiThread != _multiThread)
        {
          _dictionarySizePrev = UInt32(-1);
          _matchFinder.Release();
        }
        _multiThread = newMultiThread;
        break;
      }
      #endif
      case NCoderPropID::kDictionarySize:
      {
        const int kDicLogSizeMaxCompress = 28;
        if (prop.vt != VT_UI4)
          return E_INVALIDARG;
        UInt32 dictionarySize = prop.ulVal;
        if (dictionarySize < UInt32(1 << kDicLogSizeMin) ||
            dictionarySize > UInt32(1 << kDicLogSizeMaxCompress))
          return E_INVALIDARG;
        _dictionarySize = dictionarySize;
        UInt32 dicLogSize;
        for(dicLogSize = 0; dicLogSize < (UInt32)kDicLogSizeMaxCompress; dicLogSize++)
          if (dictionarySize <= (UInt32(1) << dicLogSize))
            break;
        _distTableSize = dicLogSize * 2;
        break;
      }
      case NCoderPropID::kPosStateBits:
      {
        if (prop.vt != VT_UI4)
          return E_INVALIDARG;
        UInt32 value = prop.ulVal;
        if (value > (UInt32)NLength::kNumPosStatesBitsEncodingMax)
          return E_INVALIDARG;
        _posStateBits = value;
        _posStateMask = (1 << _posStateBits) - 1;
        break;
      }
      case NCoderPropID::kLitPosBits:
      {
        if (prop.vt != VT_UI4)
          return E_INVALIDARG;
        UInt32 value = prop.ulVal;
        if (value > (UInt32)kNumLitPosStatesBitsEncodingMax)
          return E_INVALIDARG;
        _numLiteralPosStateBits = value;
        break;
      }
      case NCoderPropID::kLitContextBits:
      {
        if (prop.vt != VT_UI4)
          return E_INVALIDARG;
        UInt32 value = prop.ulVal;
        if (value > (UInt32)kNumLitContextBitsMax)
          return E_INVALIDARG;
        _numLiteralContextBits = value;
        break;
      }
      case NCoderPropID::kEndMarker:
      {
        if (prop.vt != VT_BOOL)
          return E_INVALIDARG;
        SetWriteEndMarkerMode(prop.boolVal == VARIANT_TRUE);
        break;
      }
      default:
        return E_INVALIDARG;
    }
  }
  return S_OK;
}

STDMETHODIMP CEncoder::WriteCoderProperties(ISequentialOutStream *outStream)
{