I'm trying to write a function which will recursively find the largest element in a list of integers. I know how to do this in Java, but can't understand how to do this at Scala.

Here is what I have so far, but without recursion:

```
def max(xs: List[Int]): Int = {
if (xs.isEmpty) throw new java.util.NoSuchElementException();
else xs.max;
}
```

How can we find it recursively with Scala semantic.

This is the most minimal recursive implementation of max I've ever been able to think up:

```
def max(xs: List[Int]): Option[Int] = xs match {
case Nil => None
case List(x: Int) => Some(x)
case x :: y :: rest => max( (if (x > y) x else y) :: rest )
}
```

It works by comparing the first two elements on the list, discarding the smaller (or the first, if both are equal) and then calling itself on the remaining list. Eventually, this will reduce the list to one element which must be the largest.

I return an Option to deal with the case of being given an empty list without throwing an exception - which forces the calling code to recognise the possibility and deal with it (up to the caller if *they* want to throw an exception).

If you want it to be more generic, it should be written like this:

```
def max[A <% Ordered[A]](xs: List[A]): Option[A] = xs match {
case Nil => None
case x :: Nil => Some(x)
case x :: y :: rest => max( (if (x > y) x else y) :: rest )
}
```

Which will work with any type which either extends the `Ordered`

trait or for which there is an implicit conversion from `A`

to `Ordered[A]`

in scope. So by default it works for `Int`

, `BigInt`

, `Char`

, `String`

and so on, because scala.Predef defines conversions for them.

We can become yet more generic like this:

```
def max[A <% Ordered[A]](xs: Seq[A]): Option[A] = xs match {
case s if s.isEmpty || !s.hasDefiniteSize => None
case s if s.size == 1 => Some(s(0))
case s if s(0) <= s(1) => max(s drop 1)
case s => max((s drop 1).updated(0, s(0)))
}
```

Which will work not just with lists but vectors and any other collection which extends the `Seq`

trait. Note that I had to add a check to see if the sequence actually has a definite size - it might be an infinite stream, so we back away if that might be the case. If you are sure your stream will have a definite size, you can always force it before calling this function - it's going to work through the whole stream anyway. See notes at the end for why I *really* would not want to return `None`

for an indefinite stream, though. I'm doing it here purely for simplicity.

But this doesn't work for sets and maps. What to do? The next common supertype is `Iterable`

, but that doesn't support `updated`

or anything equivalent. Anything we construct might be very poorly performing for the actual type. So my clean no-helper-function recursion breaks down. We *could* change to using a helper function but there are plenty of examples in the other answers and I'm going to stick with a one-simple-function approach. So at this point, we can to switch to `reduceLeft`

(and while we are at it, let's go for `Traversable' and cater for *all* collections):

```
def max[A <% Ordered[A]](xs: Traversable[A]): Option[A] = {
if (xs.hasDefiniteSize)
xs reduceLeftOption({(b, a) => if (a >= b) a else b})
else None
}
```

but if you don't consider reduceLeft recursive, we can do this:

```
def max[A <% Ordered[A]](xs: Traversable[A]): Option[A] = xs match {
case i if i.isEmpty => None
case i if i.size == 1 => Some(i.head)
case i if (i collect { case x if x > i.head => x }).isEmpty => Some(i.head)
case _ => max(xs collect { case x if x > xs.head => x })
}
```

It uses the `collect`

combinator to avoid some clumsy method of bodging a new Iterator out of `xs.head`

and `xs drop 2`

.

Either of these will work safely with almost any collection of anything which has an order. Examples:

```
scala> max(Map(1 -> "two", 3 -> "Nine", 8 -> "carrot"))
res1: Option[(Int, String)] = Some((8,carrot))
scala> max("Supercalifragilisticexpialidocious")
res2: Option[Char] = Some(x)
```

I don't usually give these others as examples, because it requires more expert knowledge of Scala.

Also, do remember that the basic `Traversable`

trait provides a `max`

method, so this is all just for practice ;)

Note: I hope that all my examples show how careful choice of the sequence of your case expressions can make each individual case expression as simple as possible.

*More Important Note:* Oh, also, while I am intensely comfortable returning `None`

for an input of `Nil`

, in practice I'd be strongly inclined to throw an exception for `hasDefiniteSize == false`

. Firstly, a finite stream could have a definite or non-definite size dependent purely on the sequence of evaluation and this function would effectively randomly return `Option`

in those cases - which could take a long time to track down. Secondly, I would want people to be able to differentiate between having passed `Nil`

and having passed truly risk input (that is, an infinite stream). I only returned `Option`

in these demonstrations to keep the code as simple as possible.