Java <-> Scala Collection conversions, Scala 2.10 [duplicate]

Answer 1

The two works in a different way:

• With JavaConverters your objects will be pimped into a class which support asScala and asJava, which let you programmatically convert your collection.

• With JavaConversions, a Java/Scala collection will be automatically converted when required

The risk with the latter is to obtain wrong or unnecessary conversions paying a performance fault. Additionally, at least in Scala 2.9 there is no caching of implicit conversion, i.e. if the same conversion is applied twice inside a method, the conversion code is called twice.

When you explicitely convert a collection, if you need it in the "Java" version, you will avoid to call twice .asScala in the same method.

Answer 2

I don't know of any such proclamation, but you should just always use JavaConverters, i.e. the ones that require you to indicate conversions with .asScala and .asJava.

As I understand it, JavaConverters were brought in in 2.8.1 because the JavaConversions in 2.8 were dangerous and made it easy to accidentally convert things where you weren't expecting it.

Answer 3

This is the poster child example for the dangers of import JavaConversions._:

scala> val m = Map(1 -> "one")
m: scala.collection.immutable.Map[Int,String] = Map(1 -> one)

scala> m.contains(1)
res0: Boolean = true

scala> m.contains("")
<console>:9: error: type mismatch;
 found   : String("")
 required: Int
              m.contains("")
                         ^

scala> import collection.JavaConversions._
import collection.JavaConversions._

scala> m.contains("")
res2: Boolean = false

Instead of issuing a type error, the compiler converts the Scala Map to to a java.util.Map, which has a looser signature that accepts Object.