具有反应性扩展的阈值检测

问题描述

我在响应式扩展上花了一些时间后重写了这个问题 - 我可能已经知道足够现在非常危险，所以请耐心等待并提前致谢。

我有以下检测水平的代码：

static class LevelExtensions
{
    public static IEnumerable<double> GetCrossovers(this double[] self,double x1,double x2)
    {
        return from level in self
               where level >= x1 && level <= x2 || level <= x1 && level >= x2
               select level;
    }

    public static IObservable<ThresholdCrossedEvent> ThresholdDetection(this IObservable<double> source,double[] thresholds)
    {
        return source
                .Buffer(2,1)
                .Where(x => x.Count == 2)
                .Select(x => new { LevelsCrossed = thresholds.GetCrossovers(x[0],x[1]),PrevIoUs = x[0],Current = x[1] })
                .Where(x => x.LevelsCrossed.ToList().Count > 0)
                .SelectMany(x => from level in x.LevelsCrossed select new ThresholdCrossedEvent(level,x.PrevIoUs,x.Current));
    }
}

public class ThresholdCrossedEvent
{
    public ThresholdCrossedEvent(double level,double prevIoUs,double current)
    {

        Threshold = level;
        PrevIoUs = prevIoUs;
        Current = current;
    }

    public double Threshold { get; set; }
    public double PrevIoUs { get; set; }
    public double Current { get; set; }
    public Direction SlopeDirection => Current >= PrevIoUs ? Direction.Up : Direction.Down;
    public bool IsTouching => Current == Threshold || PrevIoUs == Threshold;
}


var Feed = new double[] { 1,2,3,5,5.5,6,9,12,10,7.5,6.5,7 }.ToObservable();
var levels = new double[] { 5,7,8 };

Feed
  .ThresholdDetection(levels)
  .Subscribe(x => Console.WriteLine(JsonConvert.SerializeObject(x)));

它按预期产生事件，到目前为止还好：

{"Threshold":5.0,"PrevIoUs":3.0,"Current":5.0,"IsUpward":true,"IsTouching":true}
{"Threshold":5.0,"PrevIoUs":5.0,"Current":5.5,"IsTouching":true}
{"Threshold":7.0,"PrevIoUs":6.0,"Current":9.0,"IsTouching":false}
{"Threshold":8.0,"PrevIoUs":9.0,"Current":7.5,"IsUpward":false,"IsTouching":false}
{"Threshold":7.0,"PrevIoUs":7.5,"Current":6.5,"PrevIoUs":6.5,"Current":7.0,"IsTouching":true}

表演有什么味道吗？

没有过早优化 - 还有什么会引起性能问题（因为我每秒将处理 40,000 多条消息）？

我注意到我的 GetLevelsBetweenPair 平交道检测将依赖于 Linq.Where，每秒更新数千次，我的 levels 序列应该从低到高排序 - Linq为我优化或我应该考虑的任何事情？

我将如何处理要在运行时更新的 levels 列表？

我在我的 OnNext 列表上调用 Feed 从消息总线发布值，然后通过用户界面/命令我添加/删除 levels 列表上的元素（并且它修改后需要自动对自身重新排序）。

鉴于不断检查这将是一个竞争激烈的数组（但根本不会经常更新），此级别列表更新过程应考虑哪些事项？

哪种结构可以最有效的方式允许门禁访问？

拥有不同的 levels 列表并对其调用相同的订阅

我会有不同的“类别”级别，每个都创建相同的 ThresholdCrossedEvent，但意味着不同的东西，即历史阈值或周阈值。

是否可以（优选）将以下内容作为多个订阅进行操作，或者我是否应该将此阈值类型数据编码到警报中并将其从 double[] 更改为包含必需字段的类对象？

Feed
  .ThresholdDetection(historicalThresholds)
  .Subscribe(x => Console.WriteLine("Historical " + JsonConvert.SerializeObject(x)));

Feed
  .ThresholdDetection(weeklyThresholds)
  .Subscribe(x => Console.WriteLine("Weekly: " + JsonConvert.SerializeObject(x)));

解决方法

您的实现看起来不错。一些事情：

1. 您还想将不断变化的 levels 建模为可观察的。
2. 我不是最好的 Linq2Objects 专家，但我认为列表/数组转换是不必要的，并且会轻微影响性能（如果您担心的话）。显然是测试。
3. 我讨厌 Linq 语法，所以我在下面的示例中将其删除。我认为这不重要。

解决方案如下所示：

static class LevelExtensions
{
    public static IEnumerable<double> GetCrossovers(this IEnumerable<double> self,double x1,double x2)
    {
        return self
            .Where(level => level >= x1 && level <= x2 || level <= x1 && level >= x2);
    }

    public static IObservable<ThresholdCrossedEvent> ThresholdDetection(this IObservable<double> source,IObservable<double[]> thresholdSource)
    {
        return thresholdSource
            .Select(thresholds => source
                .Buffer(2,1)
                .Where(x => x.Count == 2)
                .Select(x => new { LevelsCrossed = thresholds.GetCrossovers(x[0],x[1]),Previous = x[0],Current = x[1] })
                .Where(x => x.LevelsCrossed.Count() > 0)
                .SelectMany(x => x.LevelsCrossed.Select(lc => new ThresholdCrossedEvent(lc,x.Previous,x.Current)))
            )
            .Switch();
    }
}

这是对您的代码的直接改编：唯一的区别是外部 Select、Switch 和类型。对于每个 levels 通知，新的 Select 都会创建一个新的 Observable。 Switch 表示总是切换到最新的，并丢弃旧的。因此，您始终使用最新的 levels 阈值。

我通过运行以下代码对其进行了测试：

var feed = new double[] { 1,2,3,5,5.5,6,9,12,10,7.5,6.5,7 };
var levels = new double[] { 5,7,8 };

var feedSubject = new Subject<double>();
var levelsSubject = new BehaviorSubject<double[]>(levels);

feedSubject
    .ThresholdDetection(levelsSubject)
    .Subscribe(x => Console.WriteLine(JsonConvert.SerializeObject(x)));
    
foreach (var d in feed)
    feedSubject.OnNext(d);

levelsSubject.OnNext(new double[] { 5,8 });
Console.WriteLine("---");

foreach (var d in feed)
    feedSubject.OnNext(d);

它使用您的原始阈值运行一次该提要数组，然后我在删除 7 个阈值的情况下运行它。我的输出是这样的：

{"Threshold":5.0,"Previous":3.0,"Current":5.0,"SlopeDirection":0,"IsTouching":true}
{"Threshold":5.0,"Previous":5.0,"Current":5.5,"IsTouching":true}
{"Threshold":7.0,"Previous":6.0,"Current":9.0,"IsTouching":false}
{"Threshold":8.0,"Previous":9.0,"Current":7.5,"SlopeDirection":1,"IsTouching":false}
{"Threshold":7.0,"Previous":7.5,"Current":6.5,"Previous":6.5,"Current":7.0,"IsTouching":true}
---
{"Threshold":5.0,"IsTouching":true}
{"Threshold":8.0,"IsTouching":false}