Question

我正在尝试实现以下要求（C＃4.0）;

提交上传请求的一个“生产者”（在UI线程上 - 由用户操作驱动）
一个ListView控件，它被数据绑定到待处理的上传请求集合
一个处理请求（FIFO）的消费者，并在项目状态更新时保持UI最新，或者请求得到满足。

到目前为止，我还没有能够在不使用两个集合的情况下解决上述问题，如下所示;

public void AddUploadRequest(UploadRequest uploadRequest)
{
    this.UploadRequests.Add(uploadRequest);
    this.uploadRequestsBlocking.Add(uploadRequest);
}

...其中UploadRequests是ObservableCollection<UploadRequest>而uploadRequestsBlocking是BlockingCollection<UploadRequest>

完整的代码可以在这里找到; http://pastebin.com/620EqaY5（忽略Dispatcher的大量注入 - 到目前为止这只是原型代码）

我通知我的更新UI如下;

this.dispatcher.Invoke(() => this.UploadRequests.Remove(uploadRequest));

有没有更好的方法来实现此功能？或者，更重要的是，这种方法有任何严重的缺点吗？

我考虑的一个可能的扩展是使用Parallel.ForEach而不是GetConsumingPartitioner（基于this示例），有多个消费者。关于当前方法的任何事情都会使这个不合适吗？它确实可以正常工作，但我并不是100％有信心我在这个过程中没有提供过一些主要的线程故障。

Answer 1

您只能将数据绑定到BlockingCollection的主要原因是它没有实现INotifyCollectionChanged，因此在添加/删除请求时UI不会更新。也就是说，BlockingCollection wrappers around on the net实现了INotifyCollectionChanged。或者您可以为BlockingCollection公开CollectionView，而不是从ObservableCollection调度remove，只需调度CollectionView的Refresh。

    public ICollectionView UploadRequestsView {get;set;}

    public UploadRequester(Dispatcher dispatcher)
    {
        this.dispatcher = dispatcher;           
        this.uploadRequestsBlocking = new BlockingCollection<UploadRequest>();

        UploadRequestsView = CollectionViewSource.GetDefaultView(uploadRequestsBlocking);

        this.consumerTask = Task.Factory.StartNew(this.ConsumeUploadRequests);
    }

    public void AddUploadRequest(UploadRequest uploadRequest)
    {
        uploadRequestsBlocking.Add(uploadRequest);
        UploadRequestsView.Refresh()
    }

    private void ConsumeUploadRequests()
    {
        foreach (var uploadRequest in this.uploadRequestsBlocking.GetConsumingEnumerable())
        {
            uploadRequest.Status = "Uploading...";

            Thread.Sleep(2000);
            uploadRequest.Status = "Successfully uploaded";

            Thread.Sleep(500);
            dispatcher.Invoke(() => UploadRequestsView.Refresh());
        }
    }

Answer 2

我会使用TPL而不需要注入Dispatcher，因为TaskScheduler.FromCurrentSynchronizationContext()会给我Dispatcher线程的当前MainUI。

public class UploadRequestProcessor
{
        public ObservableCollection<UploadRequest> UploadRequests { get; private set; }
        private readonly BlockingCollection<UploadRequest> uploadRequestsBlocking;
        private Task consumerTask;

        public UploadRequester()
        {
                this.UploadRequests = new ObservableCollection<UploadRequest>();
                this.uploadRequestsBlocking = new BlockingCollection<UploadRequest>();
                this.consumerTask = Task.Factory.StartNew(this.ConsumeUploadRequests).ContinueWith(nextTask => RemoveUploadRequests(nextTask.Result as BlockingCollection<UploadRequest>), TaskScheduler.FromCurrentSynchronizationContext());
        }

        public void AddUploadRequest(UploadRequest uploadRequest)
        {
                this.UploadRequests.Add(uploadRequest);
                this.uploadRequestsBlocking.Add(uploadRequest);
        }

        private void ConsumeUploadRequests()
        {
            var requestsToRemove = new BlockingCollection<UploadRequest>();
                foreach (var uploadRequest in this.uploadRequestsBlocking.GetConsumingEnumerable())
                {
                        uploadRequest.Status = "Uploading...";

                        Thread.Sleep(2000);
                        uploadRequest.Status = "Successfully uploaded";

                        Thread.Sleep(500);
                        requestsToRemove.Add(uploadRequest);

                }
            return requestsToRemove;
        }

        private void RemoveUploadRequests(BlockingCollection<UploadRequest> requestsToRemove)
        {
            foreach(var request in requestsToRemove)
            {
                UploadRequests.Remove(request);
            }
        }
}

数据绑定到BlockingCollection

2 个答案: