• 封装多线程处理大量数据操作


    们需要解决WaitAny和取得异步执行的返回值的问题。地球人都知道Thread和ThreadPool接受的委托都是没有返回值的。要想取的返回值,我们就得自己动手了,我们需要构造一个AsyncContext类,由这个类来保存异步执行的状态以并存储返回值。

    using System; 
    using System.Collections.Generic; 
    using System.Text; 
    using System.Collections; 
    using System.Threading; 
    using System.Diagnostics; 
      
    namespace AppUtility 
    { 
     public delegate object DoGetObjTask(object state); 
        public static class AsyncHelper 
        { 
            /// <summary> 
            /// 执行多线程操作任务 
            /// </summary> 
            /// <param name="dataCollection">多线程操作的数据集合</param> 
            /// <param name="threadCn">分多少个线程来做</param> 
            /// <param name="processItemMethod">处理数据集合中单个数据使用的处理方法</param> 
            public static void DoAsync(IList dataCollection, int threadCn, WaitCallback processItemMethod) 
            { 
                DoAsync(dataCollection, threadCn, processItemMethod, true); 
            } 
      
      
            /// <summary> 
            /// 执行多线程操作任务 
            /// </summary> 
            /// <param name="dataCollection">多线程操作的数据集合</param> 
            /// <param name="threadCn">分多少个线程来做</param> 
            /// <param name="processItemMethod">处理数据集合中单个数据使用的处理方法</param> 
            /// <param name="needWaitAll">是否需要等待所有线程执行完毕才返回,为true时会等待所有线程执行完毕,否则则是在有一个线程执行完毕就返回</param> 
            public static void DoAsync(IList dataCollection, int threadCn, DoGetObjTask processItemMethod, bool needWaitAll, out Hashtable processResult) 
            { 
                DoAsyncPrivate(dataCollection, threadCn, null, processItemMethod, needWaitAll, true, out processResult); 
            } 
      
            /// <summary> 
            /// 执行多线程操作任务 
            /// </summary> 
            /// <param name="dataCollection">多线程操作的数据集合</param> 
            /// <param name="threadCn">分多少个线程来做</param> 
            /// <param name="processItemMethod">处理数据集合中单个数据使用的处理方法</param> 
            /// <param name="needWaitAll">是否需要等待所有线程执行完毕才返回,为true时会等待所有线程执行完毕,否则则是在有一个线程执行完毕就返回</param> 
            public static void DoAsync(IList dataCollection, int threadCn, DoGetObjTask processItemMethod, out Hashtable processResult) 
            { 
                DoAsyncPrivate(dataCollection, threadCn, null, processItemMethod, true, true, out processResult); 
            } 
      
            /// <summary> 
            /// 执行多线程操作任务 
            /// </summary> 
            /// <param name="dataCollection">多线程操作的数据集合</param> 
            /// <param name="threadCn">分多少个线程来做</param> 
            /// <param name="processItemMethod">处理数据集合中单个数据使用的处理方法</param> 
            /// <param name="needWaitAll">是否需要等待所有线程执行完毕才返回,为true时会等待所有线程执行完毕,否则则是在有一个线程执行完毕就返回</param> 
            public static void DoAsync(IList dataCollection, int threadCn, WaitCallback processItemMethod, bool needWaitAll) 
            { 
                Hashtable hash; 
                DoAsyncPrivate(dataCollection, threadCn, processItemMethod, null, needWaitAll, false, out hash); 
            } 
      
            private static void DoAsyncPrivate(IList dataCollection, int threadCn, WaitCallback processItemMethod, DoGetObjTask getObjMethod, bool needWaitAll, bool hasReturnValue, out Hashtable processResult) 
            { 
                if (dataCollection == null) throw new ArgumentNullException("dataCollection"); 
      
                if (threadCn >= 64 || threadCn < 2) 
                { 
                    throw new ArgumentOutOfRangeException("threadCn", "threadCn 参数必须在2和64之间"); 
                } 
      
                if (threadCn > dataCollection.Count) threadCn = dataCollection.Count; 
      
                IList[] colls = new ArrayList[threadCn]; 
      
                DataWithStateList dataWithStates = new DataWithStateList(); 
                AutoResetEvent[] evts = new AutoResetEvent[threadCn]; 
      
                for (int i = 0; i < threadCn; i++) 
                { 
                    colls[i] = new ArrayList(); 
                    evts[i] = new AutoResetEvent(false); 
                } 
      
                for (int i = 0; i < dataCollection.Count; i++) 
                { 
                    object obj = dataCollection[i]; 
                    int threadIndex = i % threadCn; 
                    colls[threadIndex].Add(obj); 
                    dataWithStates.Add(new DataWithState(obj, ProcessState.WaitForProcess)); 
                } 
      
                AsyncContext context = AsyncContext.GetContext(threadCn, dataWithStates, needWaitAll, hasReturnValue, processItemMethod, getObjMethod); 
      
                for (int i = 0; i < threadCn; i++) 
                { 
                    ThreadPool.QueueUserWorkItem(DoPrivate, new object[] {  
                        colls[i],context,evts[i] 
                    }); 
                } 
      
                if (needWaitAll) 
                { 
                    WaitHandle.WaitAll(evts); 
                } 
                else
                { 
                    WaitHandle.WaitAny(evts); 
                    context.SetBreakSignal(); 
                } 
                processResult = context.ProcessResult; 
            } 
      
            private class AsyncContext 
            { 
                static public AsyncContext GetContext( 
                    int threadCn, 
                    DataWithStateList dataWithStates, 
                    bool needWaitAll, 
                    bool hasReturnValue, 
                    WaitCallback processItemMethod, 
                    DoGetObjTask hasReturnValueMethod 
                    ) 
                { 
                    AsyncContext context = new AsyncContext(); 
                    context.ThreadCount = threadCn; 
                    context.DataWithStates = dataWithStates; 
                    context.NeedWaitAll = needWaitAll; 
                    if (hasReturnValue) 
                    { 
                        Hashtable processResult = Hashtable.Synchronized(new Hashtable()); 
                        context.ProcessResult = processResult; 
                        context.HasReturnValueMethod = hasReturnValueMethod; 
                    } 
                    else
                    { 
                        context.VoidMethod = processItemMethod; 
                    } 
                    context.HasReturnValue = hasReturnValue; 
                    return context; 
                } 
      
                internal int ThreadCount; 
      
                internal DataWithStateList DataWithStates; 
      
                internal bool NeedWaitAll; 
      
                internal bool HasReturnValue; 
      
                internal WaitCallback VoidMethod; 
      
                internal DoGetObjTask HasReturnValueMethod; 
      
                private bool _breakSignal; 
      
                private Hashtable _processResult; 
      
                internal Hashtable ProcessResult 
                { 
                    get { return _processResult; } 
                    set { _processResult = value; } 
                } 
      
                internal void SetReturnValue(object obj, object result) 
                { 
                    lock (_processResult.SyncRoot) 
                    { 
                        _processResult[obj] = result; 
                    } 
                } 
      
                internal void SetBreakSignal() 
                { 
                    if (NeedWaitAll) throw new NotSupportedException("设定为NeedWaitAll时不可设置BreakSignal"); 
      
                    _breakSignal = true; 
                } 
      
                internal bool NeedBreak 
                { 
                    get
                    { 
                        return !NeedWaitAll && _breakSignal; 
                    } 
                } 
      
                internal void Exec(object obj) 
                { 
                    if (HasReturnValue) 
                    { 
                        SetReturnValue(obj, HasReturnValueMethod(obj)); 
                    } 
                    else
                    { 
                        VoidMethod(obj); 
                    } 
                    DataWithStates.SetState(obj, ProcessState.Processed); 
                } 
            } 
      
            private enum ProcessState : byte
            { 
                WaitForProcess = 0, 
                Processing = 1, 
                Processed = 2 
            } 
      
            private class DataWithStateList : List<DataWithState> 
            { 
                public void SetState(object obj, ProcessState state) 
                { 
                    lock (((ICollection)this).SyncRoot) 
                    { 
                        DataWithState dws = this.Find(delegate(DataWithState i) { return Object.Equals(i.Data, obj); }); 
      
                        if (dws != null) 
                        { 
                            dws.State = state; 
                        } 
                    } 
                } 
      
                public ProcessState GetState(object obj) 
                { 
                    lock (((ICollection)this).SyncRoot) 
                    { 
                        DataWithState dws = this.Find(delegate(DataWithState i) { return Object.Equals(i.Data, obj); }); 
                        return dws.State; 
                    } 
                } 
      
                private int GetCount(ProcessState state) 
                { 
                    List<DataWithState> datas = this.FindAll(delegate(DataWithState i) { return i.State == state; }); 
                    if (datas == null) return 0; 
                    return datas.Count; 
                } 
      
                public int WaitForDataCount 
                { 
                    get
                    { 
                        return GetCount(ProcessState.WaitForProcess); 
                    } 
                } 
      
                internal object GetWaitForObject() 
                { 
                    lock (((ICollection)this).SyncRoot) 
                    { 
                        DataWithState dws = this.Find(delegate(DataWithState i) { return i.State == ProcessState.WaitForProcess; }); 
                        if (dws == null) return null; 
                        dws.State = ProcessState.Processing; 
                        return dws.Data; 
                    } 
                } 
      
                internal bool IsWaitForData(object obj, bool setState) 
                { 
                    lock (((ICollection)this).SyncRoot) 
                    { 
                        DataWithState dws = this.Find(delegate(DataWithState i) { return i.State == ProcessState.WaitForProcess; }); 
      
                        if (setState && dws != null) dws.State = ProcessState.Processing; 
      
                        return dws != null; 
                    } 
                } 
            } 
      
            private class DataWithState 
            { 
                public readonly object Data; 
                public ProcessState State; 
      
                public DataWithState(object data, ProcessState state) 
                { 
                    Data = data; 
                    State = state; 
                } 
            } 
      
            private static int _threadNo = 0; 
      
            private static void DoPrivate(object state) 
            { 
                object[] objs = state as object[]; 
      
                IList datas = objs[0] as IList; 
                AsyncContext context = objs[1] as AsyncContext; 
                AutoResetEvent evt = objs[2] as AutoResetEvent; 
      
                DataWithStateList objStates = context.DataWithStates; 
      
    #if DEBUG 
                Thread.CurrentThread.Name = "Thread " + _threadNo; 
      
                Interlocked.Increment(ref _threadNo); 
                string threadName = Thread.CurrentThread.Name + "[" + Thread.CurrentThread.ManagedThreadId + "]"; 
                Trace.WriteLine("线程ID:" + threadName); 
    #endif 
                if (datas != null) 
                { 
                    for (int i = 0; i < datas.Count; i++) 
                    { 
                        if (context.NeedBreak) 
                        { 
    #if DEBUG 
                            Trace.WriteLine("线程" + threadName + "未执行完跳出"); 
    #endif 
                            break; 
                        } 
                        object obj = datas[i]; 
                        if (objStates.IsWaitForData(obj, true)) 
                        { 
                            if (context.NeedBreak) 
                            { 
    #if DEBUG 
                                Trace.WriteLine("线程" + threadName + "未执行完跳出"); 
    #endif 
                                break; 
                            } 
      
                            context.Exec(obj); 
      
    #if DEBUG 
                            Trace.WriteLine(string.Format("线程{0}处理{1}", threadName, obj)); 
    #endif 
                        } 
                    } 
                } 
      
                if (context.NeedWaitAll) 
                { 
                    //如果执行完当前进程的数据,还要查看剩下多少没有做,如果还剩下超过ThreadCount个没有做 
                    while (objStates.WaitForDataCount > context.ThreadCount) 
                    { 
                        if (context.NeedBreak) break; 
      
                        object obj = objStates.GetWaitForObject(); 
                        if (obj != null && objStates.IsWaitForData(obj, false)) 
                        { 
                            if (context.NeedBreak) 
                            { 
    #if DEBUG 
                                Trace.WriteLine("线程" + threadName + "未执行完跳出"); 
    #endif 
                                break; 
                            } 
      
                            context.Exec(obj); 
      
    #if DEBUG 
                            Trace.WriteLine(string.Format("线程{0}执行另一个进程的数据{1}", threadName, obj)); 
    #endif 
                        } 
                    } 
                } 
      
                evt.Set(); 
            } 
      
      
        } 
    } 

    使用AsyncHelper类,请看下面的测试代码:

    using System; 
    using System.Collections.Generic; 
    using System.Text; 
    using System.Diagnostics; 
    using AppUtility; 
    using System.IO; 
    using System.Collections; 
    using System.Threading; 
      
    namespace ConsoleApplication2 
    { 
        class Program 
        { 
            static void Main(string[] args) 
            { 
                Stopwatch sw = new Stopwatch(); 
                sw.Start(); 
                /* 
                List<string> testFiles = new List<string>(); 
                for (int i = 0; i < 100; i++) 
                { 
                    testFiles.Add("D:\test\async\file_" + i.ToString() + ".log"); 
                } 
                AsyncHelper.DoAsync(testFiles, 10, WriteFile); 
      
                Console.WriteLine("异步写耗时"+sw.ElapsedMilliseconds + "ms"); 
                */
      
                List<string> testFiles = new List<string>(); 
                for (int i = 0; i < 200; i++) 
                { 
                    testFiles.Add("D:\test\async\file_" + i.ToString() + ".log"); 
                } 
                  
                Hashtable result; 
      
                AsyncHelper.DoAsync(testFiles, 20, WriteFileAndReturnRowCount,false,out result); 
      
                Console.WriteLine("异步写耗时" + sw.ElapsedMilliseconds + "ms"); 
      
                Thread.Sleep(10); 
      
                if (result != null) 
                { 
                    foreach (object key in result.Keys) 
                    { 
                        Console.WriteLine("{0}={1}",  key,result[key]); 
                    } 
      
                } 
      
                sw.Reset(); 
                sw.Start(); 
                for (int i = 0; i < 200; i++) 
                { 
                    WriteFile("D:\test\sync\file_" + i.ToString() + ".log"); 
                } 
      
                Console.WriteLine("同步写耗时" + sw.ElapsedMilliseconds + "ms"); 
      
                Console.Read(); 
            } 
      
            static void WriteFile(object objFilePath) 
            { 
                string filePath = (string)objFilePath; 
                string dir = Path.GetDirectoryName(filePath); 
                if (!Directory.Exists(dir)) 
                { 
                    Directory.CreateDirectory(dir); 
                } 
      
                //Random r = new Random(DateTime.Now.Minute); 
                int rowCn = 10000; 
                using (StreamWriter writer = new StreamWriter(filePath, false, Encoding.Default)) 
                { 
                    for (int i = 0; i < rowCn; i++) writer.WriteLine(Guid.NewGuid()); 
                } 
            } 
      
            static object WriteFileAndReturnRowCount(object objFilePath) 
            { 
                string filePath = (string)objFilePath; 
                string dir = Path.GetDirectoryName(filePath); 
                if (!Directory.Exists(dir)) 
                { 
                    Directory.CreateDirectory(dir); 
                } 
      
                //Random r = new Random(DateTime.Now.Minute); 
                int rowCn = 10000; 
                using (StreamWriter writer = new StreamWriter(filePath, false, Encoding.Default)) 
                { 
                    for (int i = 0; i < rowCn ; i++) writer.WriteLine(Guid.NewGuid()); 
                } 
                return DateTime.Now.ToLongTimeString(); 
            } 
        } 
    }

    转自:http://www.cnblogs.com/yukaizhao/archive/2009/12/18/system_threading_threadpool_AsyncHelper_2.html

    需求是从一个文件夹中读取N个文件,将N个文件平均分配到10个线程,然后分别读取每个文件里面的内容。经测试,发现在DoPrivate方法里面,每个datas.Count分配是对的,但到objStates.IsWaitForData(obj, true)方法总有不定数量返回false,因此最后没有完全执行,最后数量总少于N,而且不定,降至2个线程也一样,使用这种方法会造成漏数据,因此只能改进DoPrivate方法,去掉objStates.IsWaitForData方法判断。

  • 相关阅读:
    vue cli选择lint模式
    收集Java8 Lambda mapreduce代码
    k8s笔记
    Deep Reinforcement Learning with Double Qlearning
    Deep Spiking Neural Networks With Binary Weights for Object Recognition
    Implementing Spiking Neural Networks on Neuromorphic Architectures: A Review
    BPSTDP: Approximating Backpropagation using Spike Timing Dependent Plasticity
    Deep Spiking Neural Network model for timevariant signals classification: a realtime speech recognition approach
    SpikingYOLO: Spiking Neural Network for EnergyEfficient Object Detection
    政策、效果、请求、匹配 PERM (Policy, Effect, Request, Matcher)
  • 原文地址:https://www.cnblogs.com/zhangs1986/p/3625680.html
Copyright © 2020-2023  润新知