“Chunked” MemoryStream
I\'m looking for the implementation of MemoryStream which does not allocate memory as one big block, but rather a collection of chunks. I want to store a few GB of data in m
-
I've found similar problem in my application. I've read large amount of compressed data and I suffered from OutOfMemoryException using MemoryStream. I've written my own implementation of "chunked" memory stream based on collection of byte arrays. If you have any idea how to make this memory stream more effective, please write me about it.
public sealed class ChunkedMemoryStream : Stream { #region Constants private const int BUFFER_LENGTH = 65536; private const byte ONE = 1; private const byte ZERO = 0; #endregion #region Readonly & Static Fields private readonly Collection<byte[]> _chunks; #endregion #region Fields private long _length; private long _position; private const byte TWO = 2; #endregion #region C'tors public ChunkedMemoryStream() { _chunks = new Collection<byte[]> { new byte[BUFFER_LENGTH], new byte[BUFFER_LENGTH] }; _position = ZERO; _length = ZERO; } #endregion #region Instance Properties public override bool CanRead { get { return true; } } public override bool CanSeek { get { return true; } } public override bool CanWrite { get { return true; } } public override long Length { get { return _length; } } public override long Position { get { return _position; } set { if (!CanSeek) throw new NotSupportedException(); _position = value; if (_position > _length) _position = _length - ONE; } } private byte[] CurrentChunk { get { long positionDividedByBufferLength = _position / BUFFER_LENGTH; var chunkIndex = Convert.ToInt32(positionDividedByBufferLength); byte[] chunk = _chunks[chunkIndex]; return chunk; } } private int PositionInChunk { get { int positionInChunk = Convert.ToInt32(_position % BUFFER_LENGTH); return positionInChunk; } } private int RemainingBytesInCurrentChunk { get { Contract.Ensures(Contract.Result<int>() > ZERO); int remainingBytesInCurrentChunk = CurrentChunk.Length - PositionInChunk; return remainingBytesInCurrentChunk; } } #endregion #region Instance Methods public override void Flush() { } public override int Read(byte[] buffer, int offset, int count) { if (offset + count > buffer.Length) throw new ArgumentException(); if (buffer == null) throw new ArgumentNullException(); if (offset < ZERO || count < ZERO) throw new ArgumentOutOfRangeException(); if (!CanRead) throw new NotSupportedException(); int bytesToRead = count; if (_length - _position < bytesToRead) bytesToRead = Convert.ToInt32(_length - _position); int bytesreaded = 0; while (bytesToRead > ZERO) { // get remaining bytes in current chunk // read bytes in current chunk // advance to next position int remainingBytesInCurrentChunk = RemainingBytesInCurrentChunk; if (remainingBytesInCurrentChunk > bytesToRead) remainingBytesInCurrentChunk = bytesToRead; Array.Copy(CurrentChunk, PositionInChunk, buffer, offset, remainingBytesInCurrentChunk); //move position in source _position += remainingBytesInCurrentChunk; //move position in target offset += remainingBytesInCurrentChunk; //bytesToRead is smaller bytesToRead -= remainingBytesInCurrentChunk; //count readed bytes; bytesreaded += remainingBytesInCurrentChunk; } return bytesreaded; } public override long Seek(long offset, SeekOrigin origin) { switch (origin) { case SeekOrigin.Begin: Position = offset; break; case SeekOrigin.Current: Position += offset; break; case SeekOrigin.End: Position = Length + offset; break; } return Position; } private long Capacity { get { int numberOfChunks = _chunks.Count; long capacity = numberOfChunks * BUFFER_LENGTH; return capacity; } } public override void SetLength(long value) { if (value > _length) { while (value > Capacity) { var item = new byte[BUFFER_LENGTH]; _chunks.Add(item); } } else if (value < _length) { var decimalValue = Convert.ToDecimal(value); var valueToBeCompared = decimalValue % BUFFER_LENGTH == ZERO ? Capacity : Capacity - BUFFER_LENGTH; //remove data chunks, but leave at least two chunks while (value < valueToBeCompared && _chunks.Count > TWO) { byte[] lastChunk = _chunks.Last(); _chunks.Remove(lastChunk); } } _length = value; if (_position > _length - ONE) _position = _length == 0 ? ZERO : _length - ONE; } public override void Write(byte[] buffer, int offset, int count) { if (!CanWrite) throw new NotSupportedException(); int bytesToWrite = count; while (bytesToWrite > ZERO) { //get remaining space in current chunk int remainingBytesInCurrentChunk = RemainingBytesInCurrentChunk; //if count of bytes to be written is fewer than remaining if (remainingBytesInCurrentChunk > bytesToWrite) remainingBytesInCurrentChunk = bytesToWrite; //if remaining bytes is still greater than zero if (remainingBytesInCurrentChunk > ZERO) { //write remaining bytes to current Chunk Array.Copy(buffer, offset, CurrentChunk, PositionInChunk, remainingBytesInCurrentChunk); //change offset of source array offset += remainingBytesInCurrentChunk; //change bytes to write bytesToWrite -= remainingBytesInCurrentChunk; //change length and position _length += remainingBytesInCurrentChunk; _position += remainingBytesInCurrentChunk; } if (Capacity == _position) _chunks.Add(new byte[BUFFER_LENGTH]); } } /// <summary> /// Gets entire content of stream regardless of Position value and return output as byte array /// </summary> /// <returns>byte array</returns> public byte[] ToArray() { var outputArray = new byte[Length]; if (outputArray.Length != ZERO) { long outputPosition = ZERO; foreach (byte[] chunk in _chunks) { var remainingLength = (Length - outputPosition) > chunk.Length ? chunk.Length : Length - outputPosition; Array.Copy(chunk, ZERO, outputArray, outputPosition, remainingLength); outputPosition = outputPosition + remainingLength; } } return outputArray; } /// <summary> /// Method set Position to first element and write entire stream to another /// </summary> /// <param name="stream">Target stream</param> public void WriteTo(Stream stream) { Contract.Requires(stream != null); Position = ZERO; var buffer = new byte[BUFFER_LENGTH]; int bytesReaded; do { bytesReaded = Read(buffer, ZERO, BUFFER_LENGTH); stream.Write(buffer, ZERO, bytesReaded); } while (bytesReaded > ZERO); } #endregion }讨论(0) -
You need to first determine if virtual address fragmentation is the problem.
If you are on a 64 bit machine (which you seem to indicate you are) I seriously doubt it is. Each 64 bit process has almost the the entire 64 bit virtual memory space available and your only worry is virtual address space fragmentation not physical memory fragmentation (which is what the operating system must worry about). The OS memory manager already pages memory under the covers. For the forseeable future you will not run out of virtual address space before you run out of physical memory. This is unlikely change before we both retire.
If you are have a 32 bit address space, then allocating contiguous large blocks of memory in the GB ramge you will encounter a fragmentation problem quite quickly. There is no stock chunk allocating memory stream in the CLR. There is one in the under the covers in ASP.NET (for other reasons) but it is not accessable. If you must travel this path you are probably better off writing one youself anyway because the usage pattern of your application is unlikely to be similar to many others and trying to fit your data into a 32bit address space will likely be your perf bottleneck.
I highly recommend requiring a 64 bit process if you are manipulating GBs of data. It will do a much better job than hand-rolled solutions to 32 bit address space fragmentation regardless of how cleaver you are.
讨论(0) -
you can use this
http://www.123aspx.com/Rotor/src/chunkedmemorystream.cs.txt
讨论(0) -
Here is a full implementation:
/// <summary> /// Defines a MemoryStream that does not sit on the Large Object Heap, thus avoiding memory fragmentation. /// </summary> public sealed class ChunkedMemoryStream : Stream { /// <summary> /// Defines the default chunk size. Currently defined as 0x10000. /// </summary> public const int DefaultChunkSize = 0x10000; // needs to be < 85000 private List<byte[]> _chunks = new List<byte[]>(); private long _position; private int _chunkSize; private int _lastChunkPos; private int _lastChunkPosIndex; /// <summary> /// Initializes a new instance of the <see cref="ChunkedMemoryStream"/> class. /// </summary> public ChunkedMemoryStream() : this(DefaultChunkSize) { } /// <summary> /// Initializes a new instance of the <see cref="ChunkedMemoryStream"/> class. /// </summary> /// <param name="chunkSize">Size of the underlying chunks.</param> public ChunkedMemoryStream(int chunkSize) : this(null) { } /// <summary> /// Initializes a new instance of the <see cref="ChunkedMemoryStream"/> class based on the specified byte array. /// </summary> /// <param name="buffer">The array of unsigned bytes from which to create the current stream.</param> public ChunkedMemoryStream(byte[] buffer) : this(DefaultChunkSize, buffer) { } /// <summary> /// Initializes a new instance of the <see cref="ChunkedMemoryStream"/> class based on the specified byte array. /// </summary> /// <param name="chunkSize">Size of the underlying chunks.</param> /// <param name="buffer">The array of unsigned bytes from which to create the current stream.</param> public ChunkedMemoryStream(int chunkSize, byte[] buffer) { FreeOnDispose = true; ChunkSize = chunkSize; _chunks.Add(new byte[chunkSize]); if (buffer != null) { Write(buffer, 0, buffer.Length); Position = 0; } } /// <summary> /// Gets or sets a value indicating whether to free the underlying chunks on dispose. /// </summary> /// <value><c>true</c> if [free on dispose]; otherwise, <c>false</c>.</value> public bool FreeOnDispose { get; set; } /// <summary> /// Releases the unmanaged resources used by the <see cref="T:System.IO.Stream"/> and optionally releases the managed resources. /// </summary> /// <param name="disposing">true to release both managed and unmanaged resources; false to release only unmanaged resources.</param> protected override void Dispose(bool disposing) { if (FreeOnDispose) { if (_chunks != null) { _chunks = null; _chunkSize = 0; _position = 0; } } base.Dispose(disposing); } /// <summary> /// When overridden in a derived class, clears all buffers for this stream and causes any buffered data to be written to the underlying device. /// This implementation does nothing. /// </summary> public override void Flush() { // do nothing } /// <summary> /// When overridden in a derived class, reads a sequence of bytes from the current stream and advances the position within the stream by the number of bytes read. /// </summary> /// <param name="buffer">An array of bytes. When this method returns, the buffer contains the specified byte array with the values between <paramref name="offset"/> and (<paramref name="offset"/> + <paramref name="count"/> - 1) replaced by the bytes read from the current source.</param> /// <param name="offset">The zero-based byte offset in <paramref name="buffer"/> at which to begin storing the data read from the current stream.</param> /// <param name="count">The maximum number of bytes to be read from the current stream.</param> /// <returns> /// The total number of bytes read into the buffer. This can be less than the number of bytes requested if that many bytes are not currently available, or zero (0) if the end of the stream has been reached. /// </returns> /// <exception cref="T:System.ArgumentException"> /// The sum of <paramref name="offset"/> and <paramref name="count"/> is larger than the buffer length. /// </exception> /// <exception cref="T:System.ArgumentNullException"> /// <paramref name="buffer"/> is null. /// </exception> /// <exception cref="T:System.ArgumentOutOfRangeException"> /// <paramref name="offset"/> or <paramref name="count"/> is negative. /// </exception> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override int Read(byte[] buffer, int offset, int count) { if (buffer == null) throw new ArgumentNullException("buffer"); if (offset < 0) throw new ArgumentOutOfRangeException("offset"); if (count < 0) throw new ArgumentOutOfRangeException("count"); if ((buffer.Length - offset) < count) throw new ArgumentException(null, "count"); CheckDisposed(); int chunkIndex = (int)(_position / ChunkSize); if (chunkIndex == _chunks.Count) return 0; int chunkPos = (int)(_position % ChunkSize); count = (int)Math.Min(count, Length - _position); if (count == 0) return 0; int left = count; int inOffset = offset; int total = 0; do { int toCopy = Math.Min(left, ChunkSize - chunkPos); Buffer.BlockCopy(_chunks[chunkIndex], chunkPos, buffer, inOffset, toCopy); inOffset += toCopy; left -= toCopy; total += toCopy; if ((chunkPos + toCopy) == ChunkSize) { if (chunkIndex == (_chunks.Count - 1)) { // last chunk break; } chunkPos = 0; chunkIndex++; } else { chunkPos += toCopy; } } while (left > 0); _position += total; return total; } /// <summary> /// Reads a byte from the stream and advances the position within the stream by one byte, or returns -1 if at the end of the stream. /// </summary> /// <returns> /// The unsigned byte cast to an Int32, or -1 if at the end of the stream. /// </returns> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override int ReadByte() { CheckDisposed(); if (_position >= Length) return -1; byte b = _chunks[(int)(_position / ChunkSize)][_position % ChunkSize]; _position++; return b; } /// <summary> /// When overridden in a derived class, sets the position within the current stream. /// </summary> /// <param name="offset">A byte offset relative to the <paramref name="origin"/> parameter.</param> /// <param name="origin">A value of type <see cref="T:System.IO.SeekOrigin"/> indicating the reference point used to obtain the new position.</param> /// <returns> /// The new position within the current stream. /// </returns> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override long Seek(long offset, SeekOrigin origin) { CheckDisposed(); switch (origin) { case SeekOrigin.Begin: Position = offset; break; case SeekOrigin.Current: Position += offset; break; case SeekOrigin.End: Position = Length + offset; break; } return Position; } private void CheckDisposed() { if (_chunks == null) throw new ObjectDisposedException(null, "Cannot access a disposed stream"); } /// <summary> /// When overridden in a derived class, sets the length of the current stream. /// </summary> /// <param name="value">The desired length of the current stream in bytes.</param> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override void SetLength(long value) { CheckDisposed(); if (value < 0) throw new ArgumentOutOfRangeException("value"); if (value > Length) throw new ArgumentOutOfRangeException("value"); long needed = value / ChunkSize; if ((value % ChunkSize) != 0) { needed++; } if (needed > int.MaxValue) throw new ArgumentOutOfRangeException("value"); if (needed < _chunks.Count) { int remove = (int)(_chunks.Count - needed); for (int i = 0; i < remove; i++) { _chunks.RemoveAt(_chunks.Count - 1); } } _lastChunkPos = (int)(value % ChunkSize); } /// <summary> /// Converts the current stream to a byte array. /// </summary> /// <returns>An array of bytes</returns> public byte[] ToArray() { CheckDisposed(); byte[] bytes = new byte[Length]; int offset = 0; for (int i = 0; i < _chunks.Count; i++) { int count = (i == (_chunks.Count - 1)) ? _lastChunkPos : _chunks[i].Length; if (count > 0) { Buffer.BlockCopy(_chunks[i], 0, bytes, offset, count); offset += count; } } return bytes; } /// <summary> /// When overridden in a derived class, writes a sequence of bytes to the current stream and advances the current position within this stream by the number of bytes written. /// </summary> /// <param name="buffer">An array of bytes. This method copies <paramref name="count"/> bytes from <paramref name="buffer"/> to the current stream.</param> /// <param name="offset">The zero-based byte offset in <paramref name="buffer"/> at which to begin copying bytes to the current stream.</param> /// <param name="count">The number of bytes to be written to the current stream.</param> /// <exception cref="T:System.ArgumentException"> /// The sum of <paramref name="offset"/> and <paramref name="count"/> is greater than the buffer length. /// </exception> /// <exception cref="T:System.ArgumentNullException"> /// <paramref name="buffer"/> is null. /// </exception> /// <exception cref="T:System.ArgumentOutOfRangeException"> /// <paramref name="offset"/> or <paramref name="count"/> is negative. /// </exception> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override void Write(byte[] buffer, int offset, int count) { if (buffer == null) throw new ArgumentNullException("buffer"); if (offset < 0) throw new ArgumentOutOfRangeException("offset"); if (count < 0) throw new ArgumentOutOfRangeException("count"); if ((buffer.Length - offset) < count) throw new ArgumentException(null, "count"); CheckDisposed(); int chunkPos = (int)(_position % ChunkSize); int chunkIndex = (int)(_position / ChunkSize); if (chunkIndex == _chunks.Count) { _chunks.Add(new byte[ChunkSize]); } int left = count; int inOffset = offset; do { int copied = Math.Min(left, ChunkSize - chunkPos); Buffer.BlockCopy(buffer, inOffset, _chunks[chunkIndex], chunkPos, copied); inOffset += copied; left -= copied; if ((chunkPos + copied) == ChunkSize) { chunkIndex++; chunkPos = 0; if (chunkIndex == _chunks.Count) { _chunks.Add(new byte[ChunkSize]); } } else { chunkPos += copied; } } while (left > 0); _position += count; if (chunkIndex == (_chunks.Count - 1)) { if ((chunkIndex > _lastChunkPosIndex) || ((chunkIndex == _lastChunkPosIndex) && (chunkPos > _lastChunkPos))) { _lastChunkPos = chunkPos; _lastChunkPosIndex = chunkIndex; } } } /// <summary> /// Writes a byte to the current position in the stream and advances the position within the stream by one byte. /// </summary> /// <param name="value">The byte to write to the stream.</param> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override void WriteByte(byte value) { CheckDisposed(); int chunkIndex = (int)(_position / ChunkSize); int chunkPos = (int)(_position % ChunkSize); if (chunkPos > (ChunkSize - 1)) //changed from (chunkPos >= (ChunkSize - 1)) { chunkIndex++; chunkPos = 0; if (chunkIndex == _chunks.Count) { _chunks.Add(new byte[ChunkSize]); } } _chunks[chunkIndex][chunkPos++] = value; _position++; if (chunkIndex == (_chunks.Count - 1)) { if ((chunkIndex > _lastChunkPosIndex) || ((chunkIndex == _lastChunkPosIndex) && (chunkPos > _lastChunkPos))) { _lastChunkPos = chunkPos; _lastChunkPosIndex = chunkIndex; } } } /// <summary> /// Writes to the specified stream. /// </summary> /// <param name="stream">The stream.</param> public void WriteTo(Stream stream) { if (stream == null) throw new ArgumentNullException("stream"); CheckDisposed(); for (int i = 0; i < _chunks.Count; i++) { int count = (i == (_chunks.Count - 1)) ? _lastChunkPos : _chunks[i].Length; stream.Write(_chunks[i], 0, count); } } /// <summary> /// When overridden in a derived class, gets a value indicating whether the current stream supports reading. /// </summary> /// <value></value> /// <returns>true if the stream supports reading; otherwise, false. /// </returns> public override bool CanRead { get { return true; } } /// <summary> /// When overridden in a derived class, gets a value indicating whether the current stream supports seeking. /// </summary> /// <value></value> /// <returns>true if the stream supports seeking; otherwise, false. /// </returns> public override bool CanSeek { get { return true; } } /// <summary> /// When overridden in a derived class, gets a value indicating whether the current stream supports writing. /// </summary> /// <value></value> /// <returns>true if the stream supports writing; otherwise, false. /// </returns> public override bool CanWrite { get { return true; } } /// <summary> /// When overridden in a derived class, gets the length in bytes of the stream. /// </summary> /// <value></value> /// <returns> /// A long value representing the length of the stream in bytes. /// </returns> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override long Length { get { CheckDisposed(); if (_chunks.Count == 0) return 0; return (_chunks.Count - 1) * ChunkSize + _lastChunkPos; } } /// <summary> /// Gets or sets the size of the underlying chunks. Cannot be greater than or equal to 85000. /// </summary> /// <value>The chunks size.</value> public int ChunkSize { get { return _chunkSize; } set { if ((value <= 0) || (value >= 85000)) throw new ArgumentOutOfRangeException("value"); _chunkSize = value; } } /// <summary> /// When overridden in a derived class, gets or sets the position within the current stream. /// </summary> /// <value></value> /// <returns> /// The current position within the stream. /// </returns> /// <exception cref="T:System.ObjectDisposedException"> /// Methods were called after the stream was closed. /// </exception> public override long Position { get { CheckDisposed(); return _position; } set { CheckDisposed(); if (value < 0) throw new ArgumentOutOfRangeException("value"); if (value > Length) throw new ArgumentOutOfRangeException("value"); _position = value; } } }讨论(0) -
The Bing team has released RecyclableMemoryStream and wrote about it here. The benefits they cite are:
- Eliminate Large Object Heap allocations by using pooled buffers
- Incur far fewer gen 2 GCs, and spend far less time paused due to GC
- Avoid memory leaks by having a bounded pool size
- Avoid memory fragmentation
- Provide excellent debuggability
- Provide metrics for performance tracking
讨论(0) -
SparseMemoryStream does this in .NET it's in buried deep down in an internal class library though -- the source code is available of course, since Microsoft put it all out there as open source.
You can grab the code for it here: http://www.dotnetframework.org/default.aspx/4@0/4@0/DEVDIV_TFS/Dev10/Releases/RTMRel/wpf/src/Base/MS/Internal/IO/Packaging/SparseMemoryStream@cs/1305600/SparseMemoryStream@cs
That being said, I highly recommend not using it as is -- At the very least remove all the calls to IsolatedStorage for starters, as this seems to be the cause of no end of bugs* in the framework's packaging API.
(*: In addition to spreading the data around in streams, if it gets too big, it basically reinvents swap files for some reason -- in the user's Isolated Storage no less -- and coincidentally, most MS products that allow for .NET based add-ins do not have their app domains setup in such a way that you can access Isolated Storage -- VSTO add-ins are notorious for suffering from this issue, for example.)
讨论(0)
加载中...
- 热议问题