return null from primitive method

[Sebastian Janisch]

Hey guys,

does anybody know why the compiler doesn't throw bricks at you when doing this:

public boolean weird(Object o)
{
    return o == null ? null : true;
}

At runtime, a NullPointerException will be thrown when o is null, but I wonder how come the compiler lets you get away with it.

[Greg Charles]

Yeah, that's a bit strange. My guess is the compiler sees the object type (null) and so decides it needs to autobox. Both null and the autoboxed true are valid instances of Boolean, so the compiler decides it's OK. On mine, I do get a warning that the unboxing may cause an NPE though.

[Wouter Oet]

The ternary operator is a weird thing. It can cause weird behaviour such as this and this:

Number n = true ? new Integer(6) : new Double(5.0);
System.out.println(n); // Prints 6.0

[Greg Charles]

Is that weird? That example seems perfectly reasonable to me, much more so than the OP. In fact, if you looked at the byte code, I bet you'd see that the compiler optimized it to:

  
Number n = new Integer(6);  
...

[Wouter Oet]

But it is printing "6.0" not "6". Which would be expected with an Integer.

public static void main(String[] args) {
   Number n1 = new Integer(6); 
   Number n2 = true ? new Integer(6) : new Double(5.0);
   System.out.println(n1);  // 6
   System.out.println(n2);  // 6.0
}    

In this case it is partially the fault of autoboxing but it only happens with the ternary operator so I blame the ternary operator.

[Greg Charles]

Oh, I see your point. Yes, it unboxes both, converts the int to double, then reboxes both as Double. That is a bit weird.