UnsupportedOperationException: empty.max

Question
1

Dear All
Excuse me if this question is not a good one. I am new to scala
I am having the error :UnsupportedOperationException: empty.max. I must have empty arrays as a result of a test. And according to my purpose, I will have many of these empty arrays.
Would you kindly help me solve this issue? My whole thesis depends on this.

Thanks

2

Well, don’t call max on an empty collection.

val max = if (list.nonEmpty) Some(list.max) else None

Or if you want to have some fun:

scala> :paste
// Entering paste mode (ctrl-D to finish)

import scala.collection.GenTraversableOnce

implicit class SafeMaxSyntax[T, A](t: T)(implicit ev: T => GenTraversableOnce[A], ord: Ordering[A]) {
  def safeMax: Option[A] = {
    val coll = ev(t)
    if (coll.nonEmpty) Some(coll.max) else None
  }
}

// Exiting paste mode, now interpreting.

import scala.collection.GenTraversableOnce
defined class SafeMaxSyntax

scala> Array(1,2,3).safeMax
res3: Option[Int] = Some(3)

scala> Array[Int]().safeMax
res4: Option[Int] = None

scala> List[Int]().safeMax
res5: Option[Int] = None

scala> "Hello World".safeMax
res6: Option[Char] = Some(r)
1 Like
3

Excuse me if this is not an acceptable question but, What is the result of empty.max? Is it a zero? a Nan?
This is the code, but if I put None as the result, it cannot make “<=” on the same val. No?

 val maxCategoriesPerFeature = numBins.values.max
          val maxCategory = numBins.find(_._2 == maxCategoriesPerFeature).get._1
          require(maxCategoriesPerFeature <= maxPossibleBins,
            s"FuzzyDecisionTree requires maxBins (= $maxPossibleBins) to be at least as large as the " +
            s"number of values in feature, when split is set to binary, but feature $maxCategory " +
            s"has $maxCategoriesPerFeature values. Considering remove this and other " +
            s"features with a large number of values, or add more training examples.\n" +
            s"For values, we refer to fuzzy sets and categories in case of " + 
             "continuous and categorical features respectively")
4

The result is an exception being thrown, because an empty list cannot have a max value (since it has no values). So you have to make sure your list isn’t empty before you call max. A Nan would make no sense, because max doesn’t have to be based on a numeric comparison. Similarly a zero.

So the point is, calling max on an empty collection gives you a computation that cannot sensibly be completed, hence you see the exception.

5

What @millross said.

If you return an Option instead of letting max throw an exception, you can use higher order functions to operate on that optional result. For instance:

val value = numBins.values
val maxCategoriesPerFeature = if (values.nonEmpty) Some(values.max) else None
val maxCategoryOpt = maxCategoriesPerFeature.flatMap { max => 
  numBins.find(_._2 == max)
}.map(_._1)
require(maxCategoriesPerFeature.map(_ <= maxPossibleBins).getOrElse(false))

If you think that in your case there is a sensible result possible for empty.max you can just do

val maxCategoriesPerFeature = if (values.nonEmpty) values.max else mySensibleResult

and continue as normal.

1 Like
6

The problem is that .max should return Option[A], as should .head, .tail, .reduce, and others. This is a design error that was discussed in the collections redesign but was rejected because reasons.

1 Like