我试图拼凑出一个接收管道的工作原理如下：

I'm attempting to piece together a Rx pipeline that works as follows:

在我写了一个函数，它在的IObservable为我提供一个包含关于公司的信息档案 我查询各种数据源的发现可能涉及公司简介，所有的并行。我合并入公司简介单一的IObservable。 在我回来后这些潜在相关的配置文件，我把它们比作是我已经观察到的轮廓，如果他们有针对性> 80％，是不一样的，因为我已经观察到的任何资料，我认为他们匹配。 我要养活匹配的公司回到步骤1，这样我就可以搜索相关的数据，这些新的配套文件。

我引导了一些已知的良好配置文件的过程。

I bootstrap the process with some known good profiles.

最终，没有那些尚未被看见更多匹配的轮廓，所以处理结束。

Eventually, there's no more matching profiles that haven't already been seen, and so the process ends.

我有麻烦的编程这一点。如果我用一个主题，让尾部的管道发送其配置文件的工作流程的开始，那么没有人会打电话OnCompleted，我从来没有发现，这个过程已经结束。如果我开发这个递归相反，我似乎永远结束了一个堆栈溢出，因为我想调用一个函数与它自己的返回值。

I'm having trouble programming this. If I use a Subject to allow the tail end of the pipeline to send its profiles to the beginning of the workflow, then no one is going to call OnCompleted and I never find out that the process has ended. If I develop this with recursion instead, I seem to always end up with a stack overflow since I'm trying to call a function with its own return value.

谁能帮我我怎么能在某种程度上，我可以确定的过程已经结束完成这项任务？

Can anyone help me with how I can accomplish this task in a way where I can determine that the process has ended?

推荐答案

这听起来像你想这样的数据流：

It sounds like you want a dataflow like this:

seed profiles --> source --> get related --> output
                     ^                    |
                     |                    v
                     -<--- transform <-----

这似乎是一个情况下，解决普遍的问题是不是具体的一个容易或更容易，所以我提出一个通用的反馈功能，应该给你你需要的构建块：

This seems like a case where solving the general problem is as easy or easier than the specific one, so I'll propose a generic "feedback" function that should give you the building blocks you need:

编辑：固定功能来完成的

IObservable<TResult> Feedback<T, TResult>(this IObservable<T> seed,
                                          Func<T, IObservable<TResult>> produce,
                                          Func<TResult, IObservable<T>> feed)
    {
        return Observable.Create<TResult>(
                obs =>
                {
                    var ret = new CompositeDisposable();
                    Action<IDisposable> partComplete = 
                        d =>
                        {
                            ret.Remove(d);
                            if (ret.Count == 0) obs.OnCompleted();
                        };
                    Action<IObservable<T>, Action<T>> ssub =
                        (o, n) =>
                        {
                            var disp = new SingleAssignmentDisposable();
                            ret.Add(disp);
                            disp.Disposable = o.Subscribe(n, obs.OnError, () => partComplete(disp));
                        };
                    Action<IObservable<TResult>, Action<TResult>> rsub =
                        (o, n) =>
                        {
                            var disp = new SingleAssignmentDisposable();
                            ret.Add(disp);
                            disp.Disposable = o.Subscribe(n, obs.OnError, () => partComplete(disp));
                        };

                    Action<T> recurse = null;
                    recurse = s =>
                              {
                                  rsub(produce(s),
                                       r => 
                                       {
                                           obs.OnNext(r);
                                           ssub(feed(r), recurse);
                                       });
                              };

                    ssub(seed, recurse);
                    return ret;
                });
    }

在你的情况， T 和 TResult 看起来是一样的，所以饲料将是身份的功能。 制作将用于实施步骤2和3的功能。

In your case, T and TResult appear to be the same, so feed will be the identity function. produce will be the functions used to implement step 2 and 3.

有些样品code我测试了此功能：

Some sample code I tested this function with:

void Main()
{
    var seed = new int[] { 1, 2, 3, 4, 5, 6 };
    var found = new HashSet<int>();
    var mults = seed.ToObservable()
                    .Feedback(i =>
                              {
                                  return Observable.Range(0, 5)
                                         .Select(r => r * i)
                                         .TakeWhile(v => v < 100)
                                         .Where(v => found.Add(v));
                              },
                              i => Observable.Return(i));

    using (var disp = mults.Dump())
    {
        Console.WriteLine("Press any key to stop");
        Console.ReadKey();
    }
    Console.WriteLine("Press any key to exit");
    Console.ReadKey();
}

static IDisposable Dump<T>(this IObservable<T> source)
{
    return source.Subscribe(item => Console.WriteLine(item),
                            ex => Console.WriteLine("Error occurred in dump observable: " + ex.ToString()),
                            () => Console.WriteLine("Dump completed"));
}