Why does Scala compiler for .NET ignore the meaning of val?

Question

I am playing around with Scala. And I found 3 interesting things(the title is the third one ).

1 a local variable declared as val is not interpreted as final.

class HowAreVarAndValImplementedInScala {
  var v1 = 123
  val v2 = 456

  def method1() = {
    var v3 = 123
    val v4 = 456
    println(v3 + v4)
  }
}

if i compile the above scala code into bytecode and then decompile that into java, it looks like this:

public class HowAreVarAndValImplementedInScala
{
  private int v1 = 123;
  private final int v2 = 456;

  public int v1()
  {
    return this.v1;
  }

  public void v1_$eq(int x$1) { this.v1 = x$1; }

  public int v2() { return this.v2; }

  public void method1() {
    int v3 = 123;
    int v4 = 456;
    Predef..MODULE$.println(BoxesRunTime.boxToInteger(v3 + v4));
  }
}

we can see v2 is final, but v4 is not, why?

2 scala compiler for .net adds override keyword to a lot of(if not all) public instance methods

if we compile the scala code shown above into CIL and then decompile it into C#, it's like this:

public class HowAreVarAndValImplementedInScala : ScalaObject
{
  private int v1;
  private int v2;

  public override int v1()
  {
    return this.v1;
  }

  public override void v1_$eq(int x$1)
  {
    this.v1 = x$1;
  }

  public override int v2()
  {
    return this.v2;
  }

  public override void method1()
  {
    int v3 = 123;
    int v4 = 456;
    Predef$.MODULE$.println(v3 + v4);
  }

  public HowAreVarAndValImplementedInScala()
  {
    this.v1 = 123;
    this.v2 = 456;
  }
}

all public instance methods(exclude the constructor) are marked as override, why? is that necessary?

3 Scala compiler for .net looses the meaning of val

in the above c# code, we can see that v2 is just a normal field, while in the java couter part, v2 is marked as final, shouldn't v2 be marked as readonly by the Scala compiler for .net?(a bug?)

As far as I know, final doesn't have runtime implications in local variables but just emphasized that you won't change a value. final fields in contrast imply for instance that they're initialized after the constructor has been executed. I therefore suppose it's omitted for v4 since it is irrelevant to the machine, whereas v2 must be final in Java. I'm not sure enough to make this an answer, however. — Matthias Meid
@AustinSalonen but i can reuse the code by posting only once :) — Cui Pengfei 崔鹏飞
public override void method1() will not compile in a C# class with no parent method to override. The compiler message is "no suitable method found to override". If the above decompiled code compiles with the C# compiler there must have been a method1 defined in ScalaObject or other parent. — Christian Maslen

Antimony Antimony · Accepted Answer · 2013-05-08T17:24:40

At the bytecode level, final doesn't exist for local variables. In fact the concept of local variables itself doesn't really exist either. Marking a local variable as final is purely for compile time checking. Since the information isn't present in the classfile, the decompiler has no way of guessing it.

As for the second two questions, I'm not as familiar with CIL bytecode, but if I had to guess I'd say that there's no reason not to add override, and readonly might have different semantics.

Edit: After looking through the CIL specification, here's what I've found.

The CIL equivalent of Java's final flag for fields is initonly, which appears to have the same semantics. It's not clear why the Scala compiler doesn't emit this. Perhaps they just didn't get around to it? Or perhaps the .net decompiler you used didn't reflect this. If you want to see what the compiler actually generates, you're best off looking at the bytecode directly.

Why does Scala compiler for .NET ignore the meaning of val?

1 Answers