传统异常处理的局限性

try {
  1024 / 0
} catch {
  case e: Throwable => e.printStackTrace()
}

用try-catch的模式,异常必须在抛出的时候马上处理.
然而在分布式计算中,我们很可能希望将异常集中到一起处理,来避免需要到每台机器上单独看错误日志的窘态.

 val seq = Seq(0, 1, 2, 3, 4)
 //seq: Seq[Int] = List(0, 1, 2, 3, 4)

val seqTry = seq.map(x => Try {
  20 / x
})
//seqTry: Seq[scala.util.Try[Int]] = List(Failure(java.lang.ArithmeticException: devide by zero),Success(20), Success(10), Success(6), Success(5))

val succSeq = seqTry.flatMap(_.toOption)
//succSeq: Seq[Int] = List(20, 10, 6, 5) Try可以转换成Option
val succSeq2 = seqTry.collect {
  case Success(x) => x
}
//succSeq2: Seq[Int] = List(20, 10, 6, 5) 和上一个是一样的
val failSeq: Seq[Throwable] = seqTry.collect {
  case Failure(e) => e
}
//failSeq: Seq[Throwable] = List(java.lang.ArithmeticException: devide by zero)

Try实例可以序列化,并且在机器间传送.