考虑到反应式编程,我经常会遇到两个流相互依赖的情况.解决这些案件的惯用方法是什么？
一个最小的例子：有按钮A和B,都显示一个值.单击A必须将A的值增加B.单击B必须将B的值设置为A.
我能提出的第一个解决方案(F#中的示例,但欢迎使用任何语言的答案)：
let solution1 buttonA buttonB =
let mutable lastA = 0
let mutable lastB = 1
let a = new Subject<_> ()
let b = new Subject<_> ()
(OnClick buttonA).Subscribe(fun _ -> lastA <- lastA + lastB; a.OnNext lastA)
(OnClick buttonB).Subscribe(fun _ -> lastB <- lastA; b.OnNext lastB)
a.Subscribe(SetText buttonA)
b.Subscribe(SetText buttonA)
a.OnNext 0
b.OnNext 1
这个解决方案使用可变状态和主题,它不是非常易读并且看起来不惯用.
我尝试的第二个解决方案涉及创建一个将两个依赖流链接在一起的方法：
let dependency (aGivenB: IObservable<_> -> IObservable<_>) (bGivenA: IObservable<_> -> IObservable<_>) =
let bProxy = new ReplaySubject<_> ()
let a = aGivenB bProxy
let b = bGivenA a
b.Subscribe(bProxy.OnNext)
a,b
let solution2 buttonA buttonB =
let aGivenB b =
Observable.WithLatestFrom(OnClick buttonA,b,fun click bValue -> bValue)
.Scan(fun acc x -> acc + x)
.StartWith(0)
let bGivenA a =
Observable.Sample(a,OnClick buttonB)
.StartWith(1)
let a,b = dependency aGivenB bGivenA
a.Subscribe(SetText buttonA)
b.Subscribe(SetText buttonB)
这看起来好一点,但由于在反应库中不存在类似依赖的方法,我相信存在更惯用的解决方案.使用第二种方法也很容易引入无限递归.
建议的方法是解决涉及流之间循环依赖的问题,例如上面的例子,在反应式编程中？
解决方法
编辑：
这是一个F#解决方案：
type DU =
| A
| B
type State = { AValue : int; BValue : int }
let solution2 (aObservable:IObservable<_>,bObservable:IObservable<_>) =
let union = aObservable.Select(fun _ -> A).Merge(bObservable.Select(fun _ -> B))
let result = union.Scan({AValue = 0; BValue = 1},fun state du -> match du with
| A -> { state with AValue = state.AValue + state.BValue }
| B -> { state with BValue = state.AValue }
)
result
由于内置的歧视联盟和记录,F#实际上是一种很好的语言.这是一个用C#编写的答案,带有一个自定义的discriminated Union;我的F#相当生疏.
诀窍是使用区别联合将您的两个可观察对象转换为一个可观察对象.因此,基本上将a和b合并为一个受歧视联盟的可观察者：
a : *---*---*---** b : -*-*--*---*--- du: ab-ba-b-a-b-aa
完成后,您可以对项目是’A’推送还是’B’推送做出反应.
为了确认,我认为没有办法明确设置ButtonA / ButtonB中嵌入的值.如果存在,那些变化应该被建模为可观察的,并且也会受到歧视的联合.
var a = new Subject<Unit>();
var b = new Subject<Unit>();
var observable = a.discriminatedUnion(b)
.Scan(new State(0,1),(state,du) => du.unify(
/* A clicked case */_ => new State(state.A + state.B,state.B),/* B clicked case */_ => new State(state.A,state.A)
)
);
observable.Subscribe(state => Console.WriteLine($"a = {state.A},b = {state.B}"));
a.OnNext(Unit.Default);
a.OnNext(Unit.Default);
a.OnNext(Unit.Default);
a.OnNext(Unit.Default);
b.OnNext(Unit.Default);
a.OnNext(Unit.Default);
a.OnNext(Unit.Default);
a.OnNext(Unit.Default);
a.OnNext(Unit.Default);
b.OnNext(Unit.Default);
这是C#中依赖的类.其中大部分内容很容易转换为内置的F#类型.
public class State /*easily replaced with an F# record */
{
public State(int a,int b)
{
A = a;
B = b;
}
public int A { get; }
public int B { get; }
}
/* easily replaced with built-in discriminated unions and pattern matching */
public static class discriminatedUnionExtensions
{
public static IObservable<discriminatedUnionClass<T1,T2>> discriminatedUnion<T1,T2>(this IObservable<T1> a,IObservable<T2> b)
{
return Observable.Merge(
a.Select(t1 => discriminatedUnionClass<T1,T2>.Create(t1)),b.Select(t2 => discriminatedUnionClass<T1,T2>.Create(t2))
);
}
public static IObservable<TResult> unify<T1,T2,TResult>(this IObservable<discriminatedUnionClass<T1,T2>> source,Func<T1,TResult> f1,Func<T2,TResult> f2)
{
return source.Select(union => unify(union,f1,f2));
}
public static TResult unify<T1,TResult>(this discriminatedUnionClass<T1,T2> union,TResult> f2)
{
return union.Item == 1
? f1(union.Item1)
: f2(union.Item2)
;
}
}
public class discriminatedUnionClass<T1,T2>
{
private readonly T1 _t1;
private readonly T2 _t2;
private readonly int _item;
private discriminatedUnionClass(T1 t1,T2 t2,int item)
{
_t1 = t1;
_t2 = t2;
_item = item;
}
public int Item
{
get { return _item; }
}
public T1 Item1
{
get { return _t1; }
}
public T2 Item2
{
get { return _t2; }
}
public static discriminatedUnionClass<T1,T2> Create(T1 t1)
{
return new discriminatedUnionClass<T1,T2>(t1,default(T2),1);
}
public static discriminatedUnionClass<T1,T2> Create(T2 t2)
{
return new discriminatedUnionClass<T1,T2>(default(T1),t2,2);
}
}