A opIndexUnary quiz

[Ali Çehreli]

On 01/02/2013 01:39 AM, Dmitry Olshansky wrote:
 > 1/2/2013 7:52 AM, bearophile пишет:

 >> struct Foo {
 >> int x;
 >> alias this = x;
 >
 > Implicit conversion to int...

 > }

 >> class Bar {
 >>      Foo[] data;
 >
 >>      this() {
 >>          data.length = 10;
 >>      }
 >
 >>      Foo opIndex(uint i) {
 >>          return data[i];
 >>      }
 >
 >> void opIndexUnary(string op)(uint i) if (op == "++") {
 >> data[i]++;
 >
 > ...and ++ somehow works with rvalue.

I am not sure that it is an rvalue in this case. I think x is used 
directly as an lvalue. Since the post-increment operator cannot be 
overloaded in D, the compiler writes the previous expression as the 
equivalent of the following:

{
     int oldValue = data[i];
     ++data[i];
     oldValue;
}

Since data[i].x is an lvalue int, there shouldn't be a problem.

The problem that I have seen here is that although the programmer thinks 
that the post-increment operation is being defined by 
opIndexUnaryRight(), opIndexUnaryRight() is not a special name. We have 
seen a similar issue recently with opGet(). These names happen to look 
like operator overloads but they are not.

Ali