I'm writing a simple Apache Spark utility that automatically creates an AccumulatorV2 based on provided initial value:
import java.lang
type Acc[T] = AccumulatorV2[T, T]
implicit val long1: Long => Acc[lang.Long] = _ => new LongAccumulator()LongAccumulator()
implicit def double1: Double => Acc[lang.Double] = _ => new DoubleAccumulator()
def accumulator[T1, T2](
initialValue: T1,
name: String
)(implicit
canBuildFrom: T1 => Acc[T2],
converter: T1 => T2,
sc: SparkContext = SparkContext.getOrCreate()
): Acc[T2] = {
val acc: Acc[T2] = canBuildFrom(initialValue)
acc.reset()
acc.add(initialValue)
sc.register(acc, name)
acc
}
The 2 implicit parameters: canBuildFrom and converter are supposed to be provided by implicit function and compiler predefined typecast respectively. However, when I tried to call it:
Metrics.accumulator(0L, "webDriverDispatched")
I got the following error information:
Error:(102, 84) No implicit view available from Long => com.tribbloids.spookystuff.Metrics.Acc[T2].
webDriverDispatched: Acc[lang.Long] = Metrics.accumulator(0L, "webDriverDispatched"),
Error:(102, 84) not enough arguments for method accumulator: (implicit canBuildFrom: Long => com.tribbloids.spookystuff.Metrics.Acc[T2], implicit converter: Long => T2, implicit sc: org.apache.spark.SparkContext)com.tribbloids.spookystuff.Metrics.Acc[T2].
Unspecified value parameters canBuildFrom, converter.
webDriverDispatched: Acc[lang.Long] = Metrics.accumulator(0L, "webDriverDispatched"),
Why scala compiler failed to find the right implicit value? And what should I do to fix it?