Question

Consider the following code, which uses two slightly different methods to check _instance and assign it when not already set.

class InstantiationTest
{
    private Object _instance;

    public void Method1() {
        if(_instance == null) {
            _instance = new Object();
        }
    }

    public void Method2() {
        _instance = _instance ?? new Object();
    }
}

Either VS or Resharper keeps underlining my explicit null checks, and prompting me to refactor using the null-coalescing operator.

I wondered whether the compiler is smart enough to detect the case in Method2() where _instance is reassigned to itself (effectively a nop?) and rewrite Method2() into Method1().

I see this is not actually the case:

Test.Method1:
IL_0000:  ldarg.0     
IL_0001:  ldfld       UserQuery+Test._instance
IL_0006:  brtrue.s    IL_0013
IL_0008:  ldarg.0     
IL_0009:  newobj      System.Object..ctor
IL_000E:  stfld       UserQuery+Test._instance
IL_0013:  ret   

versus:

Test.Method2:
IL_0000:  ldarg.0     
IL_0001:  ldarg.0     
IL_0002:  ldfld       UserQuery+Test._instance
IL_0007:  dup         
IL_0008:  brtrue.s    IL_0010
IL_000A:  pop         
IL_000B:  newobj      System.Object..ctor
IL_0010:  stfld       UserQuery+Test._instance
IL_0015:  ret      

My question is why?

Is it tricky to implement at the compiler level, too trivial to be worth the effort of implementation, or something I’m missing?

Answer 1

Generally, the C# compiler does very little optimizing of the IL, leaving that up to the JIT, which optimizes things much better for a specific architecture. So it’s simply not been implemented within the compiler, as that would take time away from other things.