• Scalaz(20)-Monad: Validation-Applicative版本的Either


      scalaz还提供了个type class叫Validation。乍看起来跟/没什么分别。实际上这个Validation是在/的基础上增加了Applicative功能,就是实现了ap函数。通过Applicative实例就可以同时运算多个Validation并返回多条异常信息。所以,/与Validation核心分别就在于Validation可以返回多条异常信息。Validation也是由两种状态组成:Success和Failure,分别与/的left和right相对应。Failure可以返回多个值。我们先来看看Validation在scalaz里的定义:scalaz/Validation.scala

    sealed abstract class Validation[+E, +A] extends Product with Serializable {
    ...
      def isSuccess: Boolean = this match {
        case Success(_) => true
        case Failure(_) => false
      }
    
      /** Return `true` if this validation is failure. */
      def isFailure: Boolean = !isSuccess
    ...
      /** Return the success value of this validation or the given default if failure. Alias for `|` */
      def getOrElse[AA >: A](x: => AA): AA =
        this match {
          case Failure(_) => x
          case Success(a) => a
        }
    
      /** Return the success value of this validation or the given default if failure. Alias for `getOrElse` */
      def |[AA >: A](x: => AA): AA =
        getOrElse(x)
    
      /** Return the success value of this validation or run the given function on the failure. */
      def valueOr[AA >: A](x: E => AA): AA =
        this match {
          case Failure(a) => x(a)
          case Success(b) => b
        }
    
      /** Return this if it is a success, otherwise, return the given value. Alias for `|||` */
      def orElse[EE >: E, AA >: A](x: => Validation[EE, AA]): Validation[EE, AA] =
        this match {
          case Failure(_) => x
          case Success(_) => this
        }
    
      /** Return this if it is a success, otherwise, return the given value. Alias for `orElse` */
      def |||[EE >: E, AA >: A](x: => Validation[EE, AA]): Validation[EE, AA] =
        orElse(x)
    ...

    与/非常相似,也是提供了getOrElse来获取Success[A]的A值。如果需要获取Failure[B]值则与/一样先用swap再用getOrElse:

    /** Flip the failure/success values in this validation. Alias for `unary_~` */
      def swap: Validation[A, E] =
        this match {
          case Failure(a) => Success(a)
          case Success(b) => Failure(b)
        }
    
    Success(3).getOrElse(0)                           //> res5: Int = 3
    Success("Three").getOrElse("Everything OK!")      //> res6: String = Three
    Failure("Something wrong!").swap.getOrElse("Everything OK!")
                                                      //> res7: String = Something wrong!
    (~Failure("Something wrong!")).getOrElse("Everything OK!")
                                                      //> res8: String = Something wrong!

     Validation的两个状态是这样定义的:

    final case class Success[A](a: A) extends Validation[Nothing, A]
    final case class Failure[E](e: E) extends Validation[E, Nothing]

    Validation也是一个Monad,可以在for-comprehension中实现Failure立即退出功能:

     1 for {
     2   a <- Success(3)
     3   b <- Success(2)
     4 } yield a + b                                     //> res5: scalaz.Validation[Nothing,Int] = Success(5)
     5 
     6 val valid= for {
     7   a <- Success(3)
     8   c <- Failure("oh, error!"): Validation[String,Int]
     9   d <- Failure("oh, error again!"): Validation[String,Int]
    10   b <- Success(2)
    11 } yield a + b                                     //> valid  : scalaz.Validation[String,Int] = Failure(oh, error!)
    12 if (valid.isFailure) valid.swap.getOrElse("no error")
    13                                                   //> res6: Any = oh, error!

    scalaz同样为所有类型值提供了注入方法:scalaz.syntax/ValidationOps.scala

    final class ValidationOps[A](self: A) {
      def success[X]: Validation[X, A] = Validation.success[X, A](self)
    
      def successNel[X]: ValidationNel[X, A] = success
    
      def failure[X]: Validation[A, X] = Validation.failure[A, X](self)
    
      @deprecated("use `failure` instead", "7.1")
      def fail[X]: Validation[A, X] = failure[X]
    
      def failureNel[X]: ValidationNel[A, X] = Validation.failureNel[A, X](self)
    
      @deprecated("use `failureNel` instead", "7.1")
      def failNel[X]: ValidationNel[A, X] = failureNel[X]
    }
    
    trait ToValidationOps {
      implicit def ToValidationOps[A](a: A) = new ValidationOps(a)
    }

    上面的例子也可以这样写:

     1 for {
     2   a <- 3.success
     3   b <- 2.success
     4 } yield a + b                                     //> res7: scalaz.Validation[Nothing,Int] = Success(5)
     5 
     6 val pv = for {
     7   a <-  3.success
     8   c <- "oh, error!".failure[String]
     9   d <- "oh, error again!".failure[String]
    10   b <-  2.success
    11 } yield a + b                                     //> pv  : scalaz.Validation[String,Int] = Failure(oh, error!)
    12 if (pv.isFailure) (~pv).getOrElse("no error")     //> res8: Any = oh, error!

    不过上面两条异常信息只返回了头一条,这与/并没有什么两样,因为它们的flatMap都是一样的:

    final class ValidationFlatMap[E, A] private[scalaz](val self: Validation[E, A]) {
      /** Bind through the success of this validation. */
      def flatMap[EE >: E, B](f: A => Validation[EE, B]): Validation[EE, B] =
        self match {
          case Success(a) => f(a)
          case e @ Failure(_) => e
        }
    }

    当前版本的scalaz已经放弃了flatMap用法:

      @deprecated("""flatMap does not accumulate errors, use `scalaz./` or `import scalaz.Validation.FlatMap._` instead""", "7.1")
      @inline implicit def ValidationFlatMapDeprecated[E, A](d: Validation[E, A]): ValidationFlatMap[E, A] = 
        new ValidationFlatMap(d)
    
      /** Import this if you wish to use `flatMap` without a deprecation
        * warning.
        */
      object FlatMap {
        @inline implicit def ValidationFlatMapRequested[E, A](d: Validation[E, A]): ValidationFlatMap[E, A] =
          new ValidationFlatMap(d)
      }

    因为Validation又是个Applicative。它实现了ap函数:

      /** Apply a function in the environment of the success of this validation, accumulating errors. */
      def ap[EE >: E, B](x: => Validation[EE, A => B])(implicit E: Semigroup[EE]): Validation[EE, B] = (this, x) match {
        case (Success(a), Success(f))   => Success(f(a))
        case (e @ Failure(_), Success(_)) => e
        case (Success(_), e @ Failure(_)) => e
        case (Failure(e1), Failure(e2)) => Failure(E.append(e2, e1))
      }

    我们可以同时运算几个Validation算法并返回所有异常信息:

    ((3.success : Validation[String,Int]) |@|
     ("oh, error1! ".failure : Validation[String,Int]) |@|
     (2.success : Validation[String,Int]) |@|
     ("oh, error2 again!".failure : Validation[String,Int])){_ + _ + _ + _}
     //> res13: scalaz.Validation[String,Int] = Failure(oh, error1! oh, error2 again!)

    我们看到即使其中两项运算出现异常但还是完成了所有运算并且返回了两条异常信息。不过这两条信息合并在了String里,可能不方便后续处理。Validation注入方法提供了failureNel函数。我们试着用用:

    ((3.successNel : ValidationNel[String,Int]) |@|
     ("oh, error1! ".failureNel : ValidationNel[String,Int]) |@|
     (2.successNel : ValidationNel[String,Int]) |@|
     ("oh, error2 again!".failureNel : ValidationNel[String,Int])){_ + _ + _ + _}
     //> res14: scalaz.Validation[scalaz.NonEmptyList[String],Int] = Failure(NonEmptyList(oh, error1! , oh, error2 again!))

    现在这两条信息被放进了NonEmptyList里。NonEmptyList就是一种List,不过没有Nil状态。看看它的定义:scalaz/NonEmptyList.scala

    /** A singly-linked list that is guaranteed to be non-empty. */
    final class NonEmptyList[+A] private[scalaz](val head: A, val tail: List[A]) {
    ...

    至少这个List含有head元素。NonEmptyList的构建器在注入方法中:scalaz/NonEmptyListOps.scala

    final class NelOps[A](self: A) {
      final def wrapNel: NonEmptyList[A] =
        NonEmptyList(self)
    }
    
    trait ToNelOps {
      implicit def ToNelOps[A](a: A) = new NelOps(a)
    }

    我们简单地试用这个NonEmptyList:

    1  val nel = 2 <:: 4 <:: 3.wrapNel                  //> nel  : scalaz.NonEmptyList[Int] = NonEmptyList(2, 4, 3)
    2  val snel = "one" <:: "two" <:: "three".wrapNel   //> snel  : scalaz.NonEmptyList[String] = NonEmptyList(one, two, three)
    3  nel.list                                         //> res17: List[Int] = List(2, 4, 3)
    4  snel.list                                        //> res18: List[String] = List(one, two, three)

    我们可以直接把它转成List再进行处理操作。

     

     

     

     

     

     

     

  • 相关阅读:
    Flink开发中的问题
    怎么确定一个Flink job的资源
    Flink统计日活
    Flink读取resource资源文件
    Spark-6-如何缓解消除数据倾斜
    Spark-5-如何定位导致数据倾斜的代码
    Spark-4-为何要处理数据倾斜
    集合遍历删除中遇到问题
    1206
    1205
  • 原文地址:https://www.cnblogs.com/tiger-xc/p/5054872.html
Copyright © 2020-2023  润新知