/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ /* vim: set ts=8 sts=2 et sw=2 tw=80: */ /* This Source Code Form is subject to the terms of the Mozilla Public * License, v. 2.0. If a copy of the MPL was not distributed with this * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ #include "mozilla/SnappyCompressOutputStream.h" #include #include "nsStreamUtils.h" #include "snappy/snappy.h" namespace mozilla { NS_IMPL_ISUPPORTS(SnappyCompressOutputStream, nsIOutputStream); // static const size_t SnappyCompressOutputStream::kMaxBlockSize = snappy::kBlockSize; SnappyCompressOutputStream::SnappyCompressOutputStream( nsIOutputStream* aBaseStream, size_t aBlockSize) : mBaseStream(aBaseStream), mBlockSize(std::min(aBlockSize, kMaxBlockSize)), mNextByte(0), mCompressedBufferLength(0), mStreamIdentifierWritten(false) { MOZ_ASSERT(mBlockSize > 0); // This implementation only supports sync base streams. Verify this in debug // builds. Note, this can be simpler than the check in // SnappyUncompressInputStream because we don't have to deal with the // nsStringInputStream oddness of being non-blocking and sync. #ifdef DEBUG bool baseNonBlocking; nsresult rv = mBaseStream->IsNonBlocking(&baseNonBlocking); MOZ_ASSERT(NS_SUCCEEDED(rv)); MOZ_ASSERT(!baseNonBlocking); #endif } size_t SnappyCompressOutputStream::BlockSize() const { return mBlockSize; } NS_IMETHODIMP SnappyCompressOutputStream::Close() { if (!mBaseStream) { return NS_OK; } nsresult rv = Flush(); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } mBaseStream->Close(); mBaseStream = nullptr; mBuffer = nullptr; mCompressedBuffer = nullptr; return NS_OK; } NS_IMETHODIMP SnappyCompressOutputStream::Flush() { if (!mBaseStream) { return NS_BASE_STREAM_CLOSED; } nsresult rv = FlushToBaseStream(); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } mBaseStream->Flush(); return NS_OK; } NS_IMETHODIMP SnappyCompressOutputStream::Write(const char* aBuf, uint32_t aCount, uint32_t* aResultOut) { return WriteSegments(NS_CopyBufferToSegment, const_cast(aBuf), aCount, aResultOut); } NS_IMETHODIMP SnappyCompressOutputStream::WriteFrom(nsIInputStream*, uint32_t, uint32_t*) { return NS_ERROR_NOT_IMPLEMENTED; } NS_IMETHODIMP SnappyCompressOutputStream::WriteSegments(nsReadSegmentFun aReader, void* aClosure, uint32_t aCount, uint32_t* aBytesWrittenOut) { *aBytesWrittenOut = 0; if (!mBaseStream) { return NS_BASE_STREAM_CLOSED; } if (!mBuffer) { mBuffer.reset(new (fallible) char[mBlockSize]); if (NS_WARN_IF(!mBuffer)) { return NS_ERROR_OUT_OF_MEMORY; } } while (aCount > 0) { // Determine how much space is left in our flat, uncompressed buffer. MOZ_ASSERT(mNextByte <= mBlockSize); uint32_t remaining = mBlockSize - mNextByte; // If it is full, then compress and flush the data to the base stream. if (remaining == 0) { nsresult rv = FlushToBaseStream(); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } // Now the entire buffer should be available for copying. MOZ_ASSERT(!mNextByte); remaining = mBlockSize; } uint32_t numToRead = std::min(remaining, aCount); uint32_t numRead = 0; nsresult rv = aReader(this, aClosure, &mBuffer[mNextByte], *aBytesWrittenOut, numToRead, &numRead); // As defined in nsIOutputStream.idl, do not pass reader func errors. if (NS_FAILED(rv)) { return NS_OK; } // End-of-file if (numRead == 0) { return NS_OK; } mNextByte += numRead; *aBytesWrittenOut += numRead; aCount -= numRead; } return NS_OK; } NS_IMETHODIMP SnappyCompressOutputStream::IsNonBlocking(bool* aNonBlockingOut) { *aNonBlockingOut = false; return NS_OK; } SnappyCompressOutputStream::~SnappyCompressOutputStream() { Close(); } nsresult SnappyCompressOutputStream::FlushToBaseStream() { MOZ_ASSERT(mBaseStream); // Lazily create the compressed buffer on our first flush. This // allows us to report OOM during stream operation. This buffer // will then get re-used until the stream is closed. if (!mCompressedBuffer) { mCompressedBufferLength = MaxCompressedBufferLength(mBlockSize); mCompressedBuffer.reset(new (fallible) char[mCompressedBufferLength]); if (NS_WARN_IF(!mCompressedBuffer)) { return NS_ERROR_OUT_OF_MEMORY; } } // The first chunk must be a StreamIdentifier chunk. Write it out // if we have not done so already. nsresult rv = MaybeFlushStreamIdentifier(); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } // Compress the data to our internal compressed buffer. size_t compressedLength; rv = WriteCompressedData(mCompressedBuffer.get(), mCompressedBufferLength, mBuffer.get(), mNextByte, &compressedLength); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } MOZ_ASSERT(compressedLength > 0); mNextByte = 0; // Write the compressed buffer out to the base stream. uint32_t numWritten = 0; rv = WriteAll(mCompressedBuffer.get(), compressedLength, &numWritten); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } MOZ_ASSERT(compressedLength == numWritten); return NS_OK; } nsresult SnappyCompressOutputStream::MaybeFlushStreamIdentifier() { MOZ_ASSERT(mCompressedBuffer); if (mStreamIdentifierWritten) { return NS_OK; } // Build the StreamIdentifier in our compressed buffer. size_t compressedLength; nsresult rv = WriteStreamIdentifier( mCompressedBuffer.get(), mCompressedBufferLength, &compressedLength); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } // Write the compressed buffer out to the base stream. uint32_t numWritten = 0; rv = WriteAll(mCompressedBuffer.get(), compressedLength, &numWritten); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } MOZ_ASSERT(compressedLength == numWritten); mStreamIdentifierWritten = true; return NS_OK; } nsresult SnappyCompressOutputStream::WriteAll(const char* aBuf, uint32_t aCount, uint32_t* aBytesWrittenOut) { *aBytesWrittenOut = 0; if (!mBaseStream) { return NS_BASE_STREAM_CLOSED; } uint32_t offset = 0; while (aCount > 0) { uint32_t numWritten = 0; nsresult rv = mBaseStream->Write(aBuf + offset, aCount, &numWritten); if (NS_WARN_IF(NS_FAILED(rv))) { return rv; } offset += numWritten; aCount -= numWritten; *aBytesWrittenOut += numWritten; } return NS_OK; } } // namespace mozilla