Improve zip file iteration performance

benchmark/ICSharpCode.SharpZipLib.Benchmark/ICSharpCode.SharpZipLib.Benchmark.csproj

@@ -1,18 +1,16 @@
 <Project Sdk="Microsoft.NET.Sdk">
 
-  <PropertyGroup>
-    <OutputType>Exe</OutputType>
-    <TargetFrameworks>net6.0;netcoreapp3.1;net462</TargetFrameworks>
-  </PropertyGroup>
+	<PropertyGroup>
+		<OutputType>Exe</OutputType>
+		<TargetFrameworks>net462;net6.0</TargetFrameworks>
+	</PropertyGroup>
 
 	<ItemGroup>
-		<PackageReference Include="BenchmarkDotNet">
-			<Version>0.12.1</Version>
-		</PackageReference>
+		<PackageReference Include="BenchmarkDotNet" Version="0.13.7" />
 	</ItemGroup>
 
 	<ItemGroup>
-	  <ProjectReference Include="..\..\src\ICSharpCode.SharpZipLib\ICSharpCode.SharpZipLib.csproj" />
+		<ProjectReference Include="..\..\src\ICSharpCode.SharpZipLib\ICSharpCode.SharpZipLib.csproj" />
 	</ItemGroup>
 
 </Project>

benchmark/ICSharpCode.SharpZipLib.Benchmark/Program.cs

@@ -9,9 +9,8 @@ public class MultipleRuntimes : ManualConfig
 	{
 		public MultipleRuntimes()
 		{
-			AddJob(Job.Default.WithToolchain(CsProjClassicNetToolchain.Net461).AsBaseline()); // NET 4.6.1
-			AddJob(Job.Default.WithToolchain(CsProjCoreToolchain.NetCoreApp21)); // .NET Core 2.1
-			AddJob(Job.Default.WithToolchain(CsProjCoreToolchain.NetCoreApp31)); // .NET Core 3.1
+			AddJob(Job.Default.WithToolchain(CsProjClassicNetToolchain.Net462).AsBaseline()); // NET 4.6.2
+			AddJob(Job.Default.WithToolchain(CsProjCoreToolchain.NetCoreApp60)); // .NET 6.0
 		}
 	}
 

benchmark/ICSharpCode.SharpZipLib.Benchmark/Zip/ZipFile.cs

@@ -0,0 +1,63 @@
+using System;
+using System.IO;
+using System.Net.Http;
+using System.Threading.Tasks;
+using BenchmarkDotNet.Attributes;
+using ICSharpCode.SharpZipLib.Zip;
+
+namespace ICSharpCode.SharpZipLib.Benchmark.Zip
+{
+	[MemoryDiagnoser]
+	[Config(typeof(MultipleRuntimes))]
+	public class ZipFile
+	{
+		private readonly byte[] readBuffer = new byte[4096];
+		private string zipFileWithLargeAmountOfEntriesPath;
+
+		[GlobalSetup]
+		public async Task GlobalSetup()
+		{
+			SharpZipLibOptions.InflaterPoolSize = 4;
+
+			// large real-world test file from test262 repository
+			string commitSha = "2e4e0e6b8ebe3348a207144204cb6d7a5571c863";
+			zipFileWithLargeAmountOfEntriesPath = Path.Combine(Path.GetTempPath(), $"{commitSha}.zip");
+			if (!File.Exists(zipFileWithLargeAmountOfEntriesPath))
+			{
+				var uri = $"https://github.com/tc39/test262/archive/{commitSha}.zip";
+
+				Console.WriteLine("Loading test262 repository archive from {0}", uri);
+
+				using (var client = new HttpClient())
+				{
+					using (var downloadStream = await client.GetStreamAsync(uri))
+					{
+						using (var writeStream = File.OpenWrite(zipFileWithLargeAmountOfEntriesPath))
+						{
+							await downloadStream.CopyToAsync(writeStream);
+							Console.WriteLine("File downloaded and saved to {0}", zipFileWithLargeAmountOfEntriesPath);
+						}
+					}
+				}
+			}
+
+		}
+
+		[Benchmark]
+		public void ReadLargeZipFile()
+		{
+			using (var file = new SharpZipLib.Zip.ZipFile(zipFileWithLargeAmountOfEntriesPath))
+			{
+				foreach (ZipEntry entry in file)
+				{
+					using (var stream = file.GetInputStream(entry))
+					{
+						while (stream.Read(readBuffer, 0, readBuffer.Length) > 0)
+						{
+						}
+					}
+				}
+			}
+		}
+	}
+}

benchmark/ICSharpCode.SharpZipLib.Benchmark/Zip/ZipInputStream.cs

@@ -1,5 +1,4 @@
-using System;
-using System.IO;
+using System.IO;
 using BenchmarkDotNet.Attributes;
 
 namespace ICSharpCode.SharpZipLib.Benchmark.Zip
@@ -15,7 +14,8 @@ public class ZipInputStream
 		byte[] zippedData;
 		byte[] readBuffer = new byte[4096];
 
-		public ZipInputStream()
+		[GlobalSetup]
+		public void GlobalSetup()
 		{
 			using (var memoryStream = new MemoryStream())
 			{

benchmark/ICSharpCode.SharpZipLib.Benchmark/Zip/ZipOutputStream.cs

@@ -1,5 +1,4 @@
-using System;
-using System.IO;
+using System.IO;
 using System.Threading.Tasks;
 using BenchmarkDotNet.Attributes;
 
@@ -16,7 +15,8 @@ public class ZipOutputStream
 		byte[] outputBuffer;
 		byte[] inputBuffer;
 
-		public ZipOutputStream()
+		[GlobalSetup]
+		public void GlobalSetup()
 		{
 			inputBuffer = new byte[ChunkSize];
 			outputBuffer = new byte[N];

src/ICSharpCode.SharpZipLib/Core/InflaterPool.cs

@@ -0,0 +1,66 @@
+using System;
+using System.Collections.Concurrent;
+using ICSharpCode.SharpZipLib.Zip.Compression;
+
+namespace ICSharpCode.SharpZipLib.Core
+{
+	/// <summary>
+	/// Pool for <see cref="Inflater"/> instances as they can be costly due to byte array allocations.
+	/// </summary>
+	internal sealed class InflaterPool
+	{
+		private readonly ConcurrentQueue<PooledInflater> noHeaderPool = new ConcurrentQueue<PooledInflater>();
+		private readonly ConcurrentQueue<PooledInflater> headerPool = new ConcurrentQueue<PooledInflater>();
+
+		internal static InflaterPool Instance { get; } = new InflaterPool();
+
+		private InflaterPool()
+		{
+		}
+
+		internal Inflater Rent(bool noHeader = false)
+		{
+			if (SharpZipLibOptions.InflaterPoolSize <= 0)
+			{
+				return new Inflater(noHeader);
+			}
+
+			var pool = GetPool(noHeader);
+
+			PooledInflater inf;
+			if (pool.TryDequeue(out var inflater))
+			{
+				inf = inflater;
+				inf.Reset();
+			}
+			else
+			{
+				inf = new PooledInflater(noHeader);
+			}
+
+			return inf;
+		}
+
+		internal void Return(Inflater inflater)
+		{
+			if (SharpZipLibOptions.InflaterPoolSize <= 0)
+			{
+				return;
+			}
+
+			if (!(inflater is PooledInflater pooledInflater))
+			{
+				throw new ArgumentException("Returned inflater was not a pooled one");
+			}
+
+			var pool = GetPool(inflater.noHeader);
+			if (pool.Count < SharpZipLibOptions.InflaterPoolSize)
+			{
+				pooledInflater.Reset();
+				pool.Enqueue(pooledInflater);
+			}
+		}
+
+		private ConcurrentQueue<PooledInflater> GetPool(bool noHeader) => noHeader ? noHeaderPool : headerPool;
+	}
+}

src/ICSharpCode.SharpZipLib/GZip/GzipInputStream.cs

@@ -4,6 +4,7 @@
 using System;
 using System.IO;
 using System.Text;
+using ICSharpCode.SharpZipLib.Core;
 
 namespace ICSharpCode.SharpZipLib.GZip
 {
@@ -82,7 +83,7 @@ public GZipInputStream(Stream baseInputStream)
 		/// Size of the buffer to use
 		/// </param>
 		public GZipInputStream(Stream baseInputStream, int size)
-			: base(baseInputStream, new Inflater(true), size)
+			: base(baseInputStream, InflaterPool.Instance.Rent(true), size)
 		{
 		}
 

src/ICSharpCode.SharpZipLib/SharpZipLibOptions.cs

@@ -0,0 +1,15 @@
+using ICSharpCode.SharpZipLib.Zip.Compression;
+
+namespace ICSharpCode.SharpZipLib
+{
+	/// <summary>
+	/// Global options to alter behavior.
+	/// </summary>
+	public static class SharpZipLibOptions
+	{
+		/// <summary>
+		/// The max pool size allowed for reusing <see cref="Inflater"/> instances, defaults to 0 (disabled).
+		/// </summary>
+		public static int InflaterPoolSize { get; set; } = 0;
+	}
+}

src/ICSharpCode.SharpZipLib/Zip/Compression/Inflater.cs

@@ -137,7 +137,7 @@ public class Inflater
 		/// True means, that the inflated stream doesn't contain a Zlib header or
 		/// footer.
 		/// </summary>
-		private bool noHeader;
+		internal bool noHeader;
 
 		private readonly StreamManipulator input;
 		private OutputWindow outputWindow;

src/ICSharpCode.SharpZipLib/Zip/Compression/PooledInflater.cs

@@ -0,0 +1,14 @@
+using ICSharpCode.SharpZipLib.Core;
+
+namespace ICSharpCode.SharpZipLib.Zip.Compression
+{
+	/// <summary>
+	/// A marker type for pooled version of an inflator that we can return back to <see cref="InflaterPool"/>.
+	/// </summary>
+	internal sealed class PooledInflater : Inflater
+	{
+		public PooledInflater(bool noHeader) : base(noHeader)
+		{
+		}
+	}
+}

src/ICSharpCode.SharpZipLib/Zip/Compression/Streams/InflaterInputStream.cs

@@ -1,6 +1,7 @@
 using System;
 using System.IO;
 using System.Security.Cryptography;
+using ICSharpCode.SharpZipLib.Core;
 
 namespace ICSharpCode.SharpZipLib.Zip.Compression.Streams
 {
@@ -339,7 +340,7 @@ public class InflaterInputStream : Stream
 		/// The InputStream to read bytes from
 		/// </param>
 		public InflaterInputStream(Stream baseInputStream)
-			: this(baseInputStream, new Inflater(), 4096)
+			: this(baseInputStream, InflaterPool.Instance.Rent(), 4096)
 		{
 		}
 
@@ -630,6 +631,12 @@ protected override void Dispose(bool disposing)
 					baseInputStream.Dispose();
 				}
 			}
+
+			if (inf is PooledInflater inflater)
+			{
+				InflaterPool.Instance.Return(inflater);
+			}
+			inf = null;
 		}
 
 		/// <summary>