Question

我想在Scala中实现自己的for-comprehension兼容monad和functor。

让我们以两个愚蠢的monad为例。一个monad是一个状态monad，包含＆＃34; Int＆＃34;你可以映射或平面图。

val maybe = IntMonad(5)
maybe flatMap( a => 3 * ( a map ( () => 2 * a ) ) )
// returns IntMonad(30)

另一个monad采取像这样的功能组合...

val func = FunctionMonad( () => println("foo") )
val fooBar = func map ( () => println("bar") )
fooBar()
// foo
// bar
// returns Unit

这个例子可能有一些错误，但你明白了。

我希望能够在Scala中使用这两种不同类型的Monad组成的for-understanding。像这样：

val myMonad = IntMonad(5)
for {
    a <- myMonad
    b <- a*2
    c <- IntMonad(b*2)
} yield c    
// returns IntMonad(20)

我不是Scala大师，但你明白了

Answer 1

对于要在for-comprehension中使用的类型，您实际上只需要为它返回返回相同类型实例的map和flatMap方法。从语法上讲，for-comprehension由编译器转换为一系列flatMap s，然后是map的最终yield。只要这些方法具有适当的签名，它就可以工作。

我不确定您的示例是什么，但这是一个相当于Option的简单示例：

sealed trait MaybeInt {
    def map(f: Int => Int): MaybeInt
    def flatMap(f: Int => MaybeInt): MaybeInt
}

case class SomeInt(i: Int) extends MaybeInt {
    def map(f: Int => Int): MaybeInt = SomeInt(f(i))
    def flatMap(f: Int => MaybeInt): MaybeInt = f(i)
}

case object NoInt extends MaybeInt {
    def map(f: Int => Int): MaybeInt = NoInt
    def flatMap(f: Int => MaybeInt): MaybeInt = NoInt
}

我有两个子类型的共同特征（尽管我可以拥有尽可能多的子类型）。公共特征MaybeInt强制每个子类型符合map / flatMap接口。

scala> val maybe = SomeInt(1)
maybe: SomeInt = SomeInt(1)

scala> val no = NoInt
no: NoInt.type = NoInt

for {
  a <- maybe
  b <- no
} yield a + b

res10: MaybeInt = NoInt

for {
  a <- maybe
  b <- maybe
} yield a + b

res12: MaybeInt = SomeInt(2)

此外，您还可以添加foreach和filter。如果你也想处理这个（没有收益）：

for(a <- maybe) println(a)

您可以添加foreach。如果你想使用if警卫：

for(a <- maybe if a > 2) yield a

您需要filter或withFilter。

一个完整的例子：

sealed trait MaybeInt { self =>
    def map(f: Int => Int): MaybeInt
    def flatMap(f: Int => MaybeInt): MaybeInt
    def filter(f: Int => Boolean): MaybeInt
    def foreach[U](f: Int => U): Unit
    def withFilter(p: Int => Boolean): WithFilter = new WithFilter(p)

    // Based on Option#withFilter
    class WithFilter(p: Int => Boolean) {
        def map(f: Int => Int): MaybeInt = self filter p map f
        def flatMap(f: Int => MaybeInt): MaybeInt = self filter p flatMap f
        def foreach[U](f: Int => U): Unit = self filter p foreach f
        def withFilter(q: Int => Boolean): WithFilter = new WithFilter(x => p(x) && q(x))
    }
}

case class SomeInt(i: Int) extends MaybeInt {
    def map(f: Int => Int): MaybeInt = SomeInt(f(i))
    def flatMap(f: Int => MaybeInt): MaybeInt = f(i)
    def filter(f: Int => Boolean): MaybeInt = if(f(i)) this else NoInt
    def foreach[U](f: Int => U): Unit = f(i)
}

case object NoInt extends MaybeInt {
    def map(f: Int => Int): MaybeInt = NoInt
    def flatMap(f: Int => MaybeInt): MaybeInt = NoInt
    def filter(f: Int => Boolean): MaybeInt = NoInt
    def foreach[U](f: Int => U): Unit = ()
}

如何制作符合你理解标准的scala monad？

1 个答案: