我编写了一个下载某些文件的方法,现在我试图让它最多并行下载5个文件,其余的则等待之前的文件完成。 我正在使用ManualResetEvent,但是当我包含同步部分时,它不再下载任何东西(没有它可以工作)。
以下是方法的代码:
static readonly int maxFiles = 5;
static int files = 0;
static object filesLocker = new object();
static System.Threading.ManualResetEvent sync = new System.Threading.ManualResetEvent(true);
/// <summary>
/// Download a file from wikipedia asynchronously
/// </summary>
/// <param name="filename"></param>
public void DoanloadFileAsync(string filename)
{
...
System.Threading.ThreadPool.QueueUserWorkItem(
(o) =>
{
bool loop = true;
while (loop)
if (sync.WaitOne())
lock (filesLocker)
{
if (files < maxFiles)
{
++files;
if (files == maxFiles)
sync.Reset();
loop = false;
}
}
try
{
WebClient downloadClient = new WebClient();
downloadClient.OpenReadCompleted += new OpenReadCompletedEventHandler(downloadClient_OpenReadCompleted);
downloadClient.OpenReadAsync(new Uri(url, UriKind.Absolute));
//5 of them do get here
}
catch
{
lock (filesLocker)
{
--files;
sync.Set();
}
throw;
}
});
}
void downloadClient_OpenReadCompleted(object sender, OpenReadCompletedEventArgs e)
{
try
{
//but none of the 5 get here
...Download logic... //works without the ManualResetEvent
}
finally
{
lock (filesLocker)
{
--files;
sync.Set();
}
}
}
我做错了吗?
使用适用于Windows Phone 7的Silverlight 4编写。
编辑: Silverlight 4中没有信号量或SemaphoreSlim。
答案 0 :(得分:4)
我在评论中的意思是在使用lock
时使用慢Interlocked
。而且这种方式也会更加高效。
最多5个并行活动下载:
public class Downloader
{
private int fileCount = 0;
private AutoResetEvent sync = new AutoResetEvent(false);
private void StartNewDownload(object o)
{
if (Interlocked.Increment(ref this.fileCount) > 5) this.sync.WaitOne();
WebClient downloadClient = new WebClient();
downloadClient.OpenReadCompleted += downloadClient_OpenReadCompleted;
downloadClient.OpenReadAsync(new Uri(o.ToString(), UriKind.Absolute));
}
private void downloadClient_OpenReadCompleted(object sender, OpenReadCompletedEventArgs e)
{
try
{
// Download logic goes here.
}
catch {}
finally
{
this.sync.Set();
Interlocked.Decrement(ref this.fileCount);
}
}
public void Run()
{
string o = "url1";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
Thread.Sleep(100);
o = "url2";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
o = "url3";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
Thread.Sleep(200);
o = "url4";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
o = "url5";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
o = "url6";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
o = "url7";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
Thread.Sleep(200);
o = "url8";
System.Threading.ThreadPool.QueueUserWorkItem(this.StartNewDownload, o);
Thread.Sleep(400);
}
}
答案 1 :(得分:1)
您似乎正在尝试一次性限制可以进入关键部分的线程数量,即文件下载。不要试图手工制作,而是使用System.Threading.Semaphore
- 这就是它的作用!
答案 2 :(得分:0)
在您创建WaitOne()
之前,WebClient
看起来很像。由于调用Set()
的所有代码都在事件处理程序或异常处理程序中,因此它永远不会命中。
也许你错误地将线程池线程方法中的WebClient
代码包含在它之外
System.Threading.ThreadPool.QueueUserWorkItem(
(o) =>
{
bool loop = true;
while (loop)
if (sync.WaitOne())
lock (filesLocker)
{
if (files < maxFiles)
{
++files;
if (files == maxFiles)
sync.Reset();
loop = false;
}
}
});
//Have the try catch OUTSIDE the background thread.
try
{
WebClient downloadClient = new WebClient();
downloadClient.OpenReadCompleted += new OpenReadCompletedEventHandler(downloadClient_OpenReadCompleted);
downloadClient.OpenReadAsync(new Uri(url, UriKind.Absolute));
//5 of them do get here
}
catch
{
lock (filesLocker)
{
--files;
sync.Set();
}
throw;
}