Question

我有一个程序，我从Internet下载文件并处理它们。以下是我编写的使用线程下载文件的函数。

Task<File> re = Task.Factory.StartNew(() => { /*Download the File*/ });
re.ContinueWith((x) => { /*Do another function*/ });

我现在希望它只使用10个线程进行下载。我查看了 ParallelOptions.MaxDegreeOfParallelism 属性，但在任务返回结果时我无法理解如何使用它。

Answer 1

您可以使用以下内容：

Func<File> work = () => {
    // Do something
    File file = ...
    return file
};

var maxNoOfWorkers = 10;    
IEnumerable<Task> tasks = Enumerable.Range(0, maxNoOfWorkers)
    .Select(s =>
    {
        var task = Task.Factory.StartNew<File>(work);
        return task.ContinueWith(ant => { /* do soemthing else */ });        
    });

这样TPL决定从threadpool获取多少线程，如果你真的想创建一个专用的（non-threadpool）线程，你可以使用：

IEnumerable<Task> tasks = Enumerable.Range(0, maxNoOfWorkers)
    .Select(s =>
    {
        var task = Task.Factory.StartNew<File>(
            work, 
            CancellationToken.None, 
            TaskCreationOptions.LongRunning, 
            TaskScheduler.Default);
        return task.ContinueWith(ant => { /* do soemthing else */ });
    });

您的其他选择是使用PLINQ或Paraller.For/ForEach，您可以使用MaxDegreeOfParallelism。

PLINQ示例可以是：

Func<File> work = () => {
    // Do something
    File file = ...
    return file
};

var maxNoOfWorkers = 10;
ParallelEnumerable.Range(0, maxNoOfWorkers)
    .WithDegreeOfParallelism(maxNoOfWorkers)
    .ForAll(x => { 
        var file = work();
        // Do something with file
    });

当然，我不了解您的示例的背景，因此您可能需要根据您的要求进行调整。

Answer 2

这样做的一个好方法是使用DataFlow API。要使用它，您必须安装Microsoft.Tpl.Dataflow Nuget package。

假设您有以下方法来下载和处理数据：

public async Task<DownloadResult> DownloadFile(string url)
{
    //Asynchronously download the file and return the result of the download.
    //You don't need a thread to download the file if you use asynchronous API.
}

public ProcessingResult ProcessDownloadResult(DownloadResult download_result)
{
    //Synchronously process the download result and produce a ProcessingResult.    
}

假设您有一个要下载的网址列表：

List<string> urls = new List<string>();

然后，您可以使用DataFlow API执行以下操作：

TransformBlock<string,DownloadResult> download_block =
    new TransformBlock<string, DownloadResult>(
        url => DownloadFile(url),
        new ExecutionDataflowBlockOptions
        {
            //Only 10 asynchronous download operations
            //can happen at any point in time.
            MaxDegreeOfParallelism = 10
        });

TransformBlock<DownloadResult, ProcessingResult> process_block =
    new TransformBlock<DownloadResult, ProcessingResult>(
        dr => ProcessDownloadResult(dr),
        new ExecutionDataflowBlockOptions
        {
            //We limit the number of CPU intensive operation
            //to the number of processors in the system.
            MaxDegreeOfParallelism = Environment.ProcessorCount
        });

download_block.LinkTo(process_block);

foreach(var url in urls)
{
    download_block.Post(url);
}

MaxDegreeOfParallelism with Task.Factory.StartNew（）

2 个答案: