Is Scala mapValues lazy?

Question

When I call

System.err.println("Before")
System.err.flush()
val foo: Map[Int, T] = t mapValues (fn(_))
System.err.println(foo.head) //prevent optimiser from delaying the construction of 'foo' 
System.err.println("After")
System.err.flush()

with fn having a debug print statement inside, I get this output:

Before
...head item...
After
...debug print statement from fn...
...debug print statement from fn...

I don't understand why the debug print statements are being called after "After" is printed, and I don't understand why I'm getting it twice --- unless mapValues creates a lazy map?

Answer 1

Yes it is. It maps to an intermediate class that holds fn and doesn't evaluate until access (again and again).

def mapValues[W](f: V => W): Map[K, W] = new MappedValues(f)

Use a strict map if you don't want lazy evaluation. That is:

collection map { case (k, v) => (k, fn(v)) }

Answer 2

Keep in mind that the MappedValues implementation evaluates the function on every access -- different from a Scala lazy val that evaluates only once. You might be seeing the output twice when stepping through the code. Expanding the val foo in the debugger window will iterate over the values, calling function fn and generating debug output.

If you provide code for map t and function fn, then we might be able to help.