Add options to brotli and implement clean brotli .br support

- allow to specify brotli window size
  - parameter -m0=brotli:long=n, BROTLI_MAX_WINDOW_BITS (24) used by default in brotli-mt, smaller == faster
  - note that :long can be set up to BROTLI_LARGE_MAX_WINDOW_BITS (30), whereas :wlog can be set up to BROTLI_MAX_WINDOW_BITS (24) only...
  - todo: check whether set of BROTLI_PARAM_LARGE_WINDOW to BROTLI_TRUE is needed if (lgwin > BROTLI_MAX_WINDOW_BITS)

- implementation of single-threaded brotli compression / decompression for .br data

Signed-off-by: Sergey G. Brester <info@sebres.de>
Reviewed-by: Tino Reichardt <milky-7zip@mcmilk.de>

CPP/7zip/Archive/BrotliHandler.cpp

@@ -0,0 +1,351 @@
+// BrotliHandler.cpp
+
+#include "StdAfx.h"
+
+#include "../../../C/CpuArch.h"
+#include "../../Common/ComTry.h"
+#include "../../Common/Defs.h"
+
+#include "../Common/ProgressUtils.h"
+#include "../Common/RegisterArc.h"
+#include "../Common/StreamUtils.h"
+
+#include "../Compress/BrotliDecoder.h"
+#include "../Compress/BrotliEncoder.h"
+#include "../Compress/CopyCoder.h"
+
+#include "Common/DummyOutStream.h"
+#include "Common/HandlerOut.h"
+
+using namespace NWindows;
+
+namespace NArchive {
+namespace NBROTLI {
+
+class CHandler:
+  public IInArchive,
+  public IArchiveOpenSeq,
+  public IOutArchive,
+  public ISetProperties,
+  public CMyUnknownImp
+{
+  CMyComPtr<IInStream> _stream;
+  CMyComPtr<ISequentialInStream> _seqStream;
+
+  bool _isArc;
+  bool _dataAfterEnd;
+  bool _needMoreInput;
+
+  bool _packSize_Defined;
+  bool _unpackSize_Defined;
+
+  UInt64 _packSize;
+  UInt64 _unpackSize;
+  UInt64 _numStreams;
+  UInt64 _numBlocks;
+
+  CSingleMethodProps _props;
+
+public:
+  MY_UNKNOWN_IMP4(
+      IInArchive,
+      IArchiveOpenSeq,
+      IOutArchive,
+      ISetProperties)
+  INTERFACE_IInArchive(;)
+  INTERFACE_IOutArchive(;)
+  STDMETHOD(OpenSeq)(ISequentialInStream *stream);
+  STDMETHOD(SetProperties)(const wchar_t * const *names, const PROPVARIANT *values, UInt32 numProps);
+
+  CHandler() { }
+};
+
+static const Byte kProps[] =
+{
+  kpidSize,
+  kpidPackSize
+};
+
+static const Byte kArcProps[] =
+{
+  kpidNumStreams,
+  kpidNumBlocks
+};
+
+IMP_IInArchive_Props
+IMP_IInArchive_ArcProps
+
+STDMETHODIMP CHandler::GetArchiveProperty(PROPID /*propID*/, PROPVARIANT * /*value*/)
+{
+  return S_OK;
+}
+
+STDMETHODIMP CHandler::GetNumberOfItems(UInt32 *numItems)
+{
+  *numItems = 1;
+  return S_OK;
+}
+
+STDMETHODIMP CHandler::GetProperty(UInt32 /* index */, PROPID propID, PROPVARIANT *value)
+{
+  NCOM::CPropVariant prop;
+  switch (propID)
+  {
+    case kpidPackSize: if (_packSize_Defined) prop = _packSize; break;
+    case kpidSize: if (_unpackSize_Defined) prop = _unpackSize; break;
+  }
+  prop.Detach(value);
+  return S_OK;
+}
+
+// brotli format has no signature
+// brotli stream can't be veriefied unless errors by unpack
+#if 0
+API_FUNC_static_IsArc IsArc_Brotli(const Byte *p, size_t size)
+{
+  return k_IsArc_Res_YES;
+}
+}
+#endif
+
+STDMETHODIMP CHandler::Open(IInStream *stream, const UInt64 *, IArchiveOpenCallback *)
+{
+  COM_TRY_BEGIN
+  Close();
+  {
+    _isArc = true;
+    _stream = stream;
+    _seqStream = stream;
+  }
+  return S_OK;
+  COM_TRY_END
+}
+
+STDMETHODIMP CHandler::OpenSeq(ISequentialInStream *stream)
+{
+  Close();
+  _isArc = true;
+  _seqStream = stream;
+  return S_OK;
+}
+
+STDMETHODIMP CHandler::Close()
+{
+  _isArc = false;
+  _dataAfterEnd = false;
+  _needMoreInput = false;
+
+  _packSize_Defined = false;
+  _unpackSize_Defined = false;
+
+  _packSize = 0;
+
+  _seqStream.Release();
+  _stream.Release();
+  return S_OK;
+}
+
+STDMETHODIMP CHandler::Extract(const UInt32 *indices, UInt32 numItems,
+    Int32 testMode, IArchiveExtractCallback *extractCallback)
+{
+  COM_TRY_BEGIN
+  if (numItems == 0)
+    return S_OK;
+  if (numItems != (UInt32)(Int32)-1 && (numItems != 1 || indices[0] != 0))
+    return E_INVALIDARG;
+
+  if (_packSize_Defined)
+    extractCallback->SetTotal(_packSize);
+
+  CMyComPtr<ISequentialOutStream> realOutStream;
+  Int32 askMode = testMode ?
+      NExtract::NAskMode::kTest :
+      NExtract::NAskMode::kExtract;
+  RINOK(extractCallback->GetStream(0, &realOutStream, askMode));
+  if (!testMode && !realOutStream)
+    return S_OK;
+
+  extractCallback->PrepareOperation(askMode);
+
+  Int32 opRes;
+
+  {
+
+  NCompress::NBROTLI::CDecoder *decoderSpec = new NCompress::NBROTLI::CDecoder;
+  decoderSpec->SetNumberOfThreads(0); /* .br - single threaded processing (without header/mt-frames) */
+  CMyComPtr<ICompressCoder> decoder = decoderSpec;
+  decoderSpec->SetInStream(_seqStream);
+
+  CDummyOutStream *outStreamSpec = new CDummyOutStream;
+  CMyComPtr<ISequentialOutStream> outStream(outStreamSpec);
+  outStreamSpec->SetStream(realOutStream);
+  outStreamSpec->Init();
+
+  realOutStream.Release();
+
+  CLocalProgress *lps = new CLocalProgress;
+  CMyComPtr<ICompressProgressInfo> progress = lps;
+  lps->Init(extractCallback, true);
+
+  UInt64 packSize = 0;
+  UInt64 unpackedSize = 0;
+
+  HRESULT result = S_OK;
+
+  for (;;)
+  {
+    lps->InSize = packSize;
+    lps->OutSize = unpackedSize;
+
+    RINOK(lps->SetCur());
+    result = decoderSpec->CodeResume(outStream, &unpackedSize, progress);
+    UInt64 streamSize = decoderSpec->GetInputProcessedSize();
+
+    if (result != S_FALSE && result != S_OK)
+      return result;
+
+    if (unpackedSize == 0)
+      break;
+
+    if (streamSize == packSize)
+    {
+      // no new bytes in input stream, So it's good end of archive.
+      result = S_OK;
+      break;
+    }
+
+    if (packSize > streamSize)
+      return E_FAIL;
+
+    if (result != S_OK)
+      break;
+  }
+
+  decoderSpec->ReleaseInStream();
+  outStream.Release();
+
+  if (!_isArc)
+    opRes = NExtract::NOperationResult::kIsNotArc;
+  else if (_needMoreInput)
+    opRes = NExtract::NOperationResult::kUnexpectedEnd;
+  else if (_dataAfterEnd)
+    opRes = NExtract::NOperationResult::kDataAfterEnd;
+  else if (result == S_FALSE)
+    opRes = NExtract::NOperationResult::kDataError;
+  else if (result == S_OK)
+    opRes = NExtract::NOperationResult::kOK;
+  else
+    return result;
+
+  }
+
+  return extractCallback->SetOperationResult(opRes);
+
+  COM_TRY_END
+}
+
+static HRESULT UpdateArchive(
+    UInt64 unpackSize,
+    ISequentialOutStream *outStream,
+    const CProps &props,
+    IArchiveUpdateCallback *updateCallback)
+{
+  RINOK(updateCallback->SetTotal(unpackSize));
+  CMyComPtr<ISequentialInStream> fileInStream;
+  RINOK(updateCallback->GetStream(0, &fileInStream));
+  CLocalProgress *localProgressSpec = new CLocalProgress;
+  CMyComPtr<ICompressProgressInfo> localProgress = localProgressSpec;
+  localProgressSpec->Init(updateCallback, true);
+  NCompress::NBROTLI::CEncoder *encoderSpec = new NCompress::NBROTLI::CEncoder;
+  encoderSpec->unpackSize = unpackSize;
+  encoderSpec->SetNumberOfThreads(0); /* .br - single threaded processing (without header/mt-frames) */
+  CMyComPtr<ICompressCoder> encoder = encoderSpec;
+  RINOK(props.SetCoderProps(encoderSpec, NULL));
+  RINOK(encoder->Code(fileInStream, outStream, NULL, NULL, localProgress));
+  return updateCallback->SetOperationResult(NArchive::NUpdate::NOperationResult::kOK);
+}
+
+STDMETHODIMP CHandler::GetFileTimeType(UInt32 *type)
+{
+  *type = NFileTimeType::kUnix;
+  return S_OK;
+}
+
+STDMETHODIMP CHandler::UpdateItems(ISequentialOutStream *outStream, UInt32 numItems,
+    IArchiveUpdateCallback *updateCallback)
+{
+  COM_TRY_BEGIN
+
+  if (numItems != 1)
+    return E_INVALIDARG;
+
+  Int32 newData, newProps;
+  UInt32 indexInArchive;
+  if (!updateCallback)
+    return E_FAIL;
+  RINOK(updateCallback->GetUpdateItemInfo(0, &newData, &newProps, &indexInArchive));
+ 
+  if ((newProps))
+  {
+    {
+      NCOM::CPropVariant prop;
+      RINOK(updateCallback->GetProperty(0, kpidIsDir, &prop));
+      if (prop.vt != VT_EMPTY)
+        if (prop.vt != VT_BOOL || prop.boolVal != VARIANT_FALSE)
+          return E_INVALIDARG;
+    }
+  }
+  
+  if ((newData))
+  {
+    UInt64 size;
+    {
+      NCOM::CPropVariant prop;
+      RINOK(updateCallback->GetProperty(0, kpidSize, &prop));
+      if (prop.vt != VT_UI8)
+        return E_INVALIDARG;
+      size = prop.uhVal.QuadPart;
+    }
+    return UpdateArchive(size, outStream, _props, updateCallback);
+  }
+
+  if (indexInArchive != 0)
+    return E_INVALIDARG;
+
+  CLocalProgress *lps = new CLocalProgress;
+  CMyComPtr<ICompressProgressInfo> progress = lps;
+  lps->Init(updateCallback, true);
+
+  CMyComPtr<IArchiveUpdateCallbackFile> opCallback;
+  updateCallback->QueryInterface(IID_IArchiveUpdateCallbackFile, (void **)&opCallback);
+  if (opCallback)
+  {
+    RINOK(opCallback->ReportOperation(
+        NEventIndexType::kInArcIndex, 0,
+        NUpdateNotifyOp::kReplicate))
+  }
+
+  if (_stream)
+    RINOK(_stream->Seek(0, STREAM_SEEK_SET, NULL));
+
+  return NCompress::CopyStream(_stream, outStream, progress);
+
+  COM_TRY_END
+}
+
+STDMETHODIMP CHandler::SetProperties(const wchar_t * const *names, const PROPVARIANT *values, UInt32 numProps)
+{
+  return _props.SetProperties(names, values, numProps);
+}
+
+IMP_CreateArcIn
+IMP_CreateArcOut
+REGISTER_ARC_R(
+  "brotli", "br brotli tbr", 0, 0x1F,
+  0, NULL,
+  0,
+  NArcInfoFlags::kKeepName | NArcInfoFlags::kPureStartOpen | NArcInfoFlags::kByExtOnlyOpen,
+  0,
+  CreateArc, CreateArcOut, 
+  /* IsArc_Brotli */)
+}}

CPP/7zip/Bundles/Format7zF/Arc.mak

@@ -70,6 +70,7 @@ AR_OBJS = \
   $O\ArHandler.obj \
   $O\ArjHandler.obj \
   $O\Base64Handler.obj \
+  $O\BrotliHandler.obj \
   $O\Bz2Handler.obj \
   $O\ComHandler.obj \
   $O\CpioHandler.obj \

CPP/7zip/Compress/BrotliDecoder.cpp

@@ -98,7 +98,7 @@ STDMETHODIMP CDecoder::SetDecoderProperties2(const Byte * prop, UInt32 size)
 STDMETHODIMP CDecoder::SetNumberOfThreads(UInt32 numThreads)
 {
   const UInt32 kNumThreadsMax = BROTLIMT_THREAD_MAX;
-  if (numThreads < 1) numThreads = 1;
+  if (numThreads < 0) numThreads = 0;
   if (numThreads > kNumThreadsMax) numThreads = kNumThreadsMax;
   _numThreads = numThreads;
   return S_OK;

CPP/7zip/Compress/BrotliEncoder.cpp

@@ -12,8 +12,10 @@ CEncoder::CEncoder():
   _processedIn(0),
   _processedOut(0),
   _inputSize(0),
-  _ctx(NULL),
-  _numThreads(NWindows::NSystem::GetNumberOfProcessors())
+  _numThreads(NWindows::NSystem::GetNumberOfProcessors()),
+  _Long(-1),
+  _WindowLog(-1),
+  _ctx(NULL)
 {
   _props.clear();
 }
@@ -52,6 +54,37 @@ STDMETHODIMP CEncoder::SetCoderProperties(const PROPID * propIDs, const PROPVARI
         SetNumberOfThreads(v);
         break;
       }
+    case NCoderPropID::kLong:
+      {
+        /* like --long in zstd cli program */
+        _Long = 1;
+        if (v == 0) {
+          if (prop.vt == VT_EMPTY) {
+            // m0=brotli:long -> long=default
+            _WindowLog = BROTLI_MAX_WINDOW_BITS; //BROTLI_DEFAULT_WINDOW;
+          } else {
+            // m0=brotli:long=0 -> long=auto
+            _WindowLog = 0;
+          }
+        } else if (v < BROTLI_MIN_WINDOW_BITS) {
+          // m0=brotli:long=9 -> long=10
+          _WindowLog = BROTLI_MIN_WINDOW_BITS;
+        } else if (v > BROTLI_LARGE_MAX_WINDOW_BITS) {
+          // m0=brotli:long=33 -> long=max
+          _WindowLog = BROTLI_LARGE_MAX_WINDOW_BITS;
+        } else {
+          // m0=brotli:long=15 -> long=value
+          _WindowLog = v;
+        }
+        break;
+      }
+    case NCoderPropID::kWindowLog:
+      {
+        if (v < BROTLI_MIN_WINDOW_BITS) v = BROTLI_MIN_WINDOW_BITS;
+        if (v > BROTLI_MAX_WINDOW_BITS) v = BROTLI_MAX_WINDOW_BITS;
+        _WindowLog = v;
+        break;
+      }
     default:
       {
         break;
@@ -102,11 +135,12 @@ STDMETHODIMP CEncoder::Code(ISequentialInStream *inStream,
 
   /* 2) create compression context, if needed */
   if (!_ctx)
-    _ctx = BROTLIMT_createCCtx(_numThreads, _props._level, _inputSize);
+    _ctx = BROTLIMT_createCCtx(_numThreads, unpackSize, _props._level, _inputSize, 
+      _WindowLog >= 0 ? _WindowLog : BROTLI_MAX_WINDOW_BITS);
   if (!_ctx)
     return S_FALSE;
 
-  /* 3) compress */
+  /* 4) compress */
   result = BROTLIMT_compressCCtx(_ctx, &rdwr);
   if (BROTLIMT_isError(result)) {
     if (result == (size_t)-BROTLIMT_error_canceled)
@@ -120,7 +154,7 @@ STDMETHODIMP CEncoder::Code(ISequentialInStream *inStream,
 STDMETHODIMP CEncoder::SetNumberOfThreads(UInt32 numThreads)
 {
   const UInt32 kNumThreadsMax = BROTLIMT_THREAD_MAX;
-  if (numThreads < 1) numThreads = 1;
+  if (numThreads < 0) numThreads = 0;
   if (numThreads > kNumThreadsMax) numThreads = kNumThreadsMax;
   _numThreads = numThreads;
   return S_OK;

CPP/7zip/Compress/BrotliEncoder.h

@@ -45,9 +45,15 @@ class CEncoder:
   UInt32 _inputSize;
   UInt32 _numThreads;
 
+  Int32 _Long;
+  Int32 _WindowLog;
+
   BROTLIMT_CCtx *_ctx;
 
 public:
+
+  UInt64 unpackSize;
+
   MY_QUERYINTERFACE_BEGIN2(ICompressCoder)
   MY_QUERYINTERFACE_ENTRY(ICompressSetCoderMt)
   MY_QUERYINTERFACE_ENTRY(ICompressSetCoderProperties)