My goal is to be able to overload the "++" operator transparently to the code, but I can't.

import std.stdio;

struct Arc {
   int I = 0;
   // This is void, but the error appears under all return types
   void opUnary(string op)() if( op == "++" ) {
      ++I;
   }
}

struct HasArc {
    Arc myArc;
}

void main() {
   HasArc has;
   writefln(" Arc.I = %s", has.myArc.I); // Arc.I = 0

   has.myArc++; // Error: var has no effect in expression (__pitmp1481)

   // Okay, try this instead:
   auto UselessVar = has.myArc++; // Works fine

   writefln(" Arc.I = %s", has.myArc.I); // Arc.I = 1
}

Obviously the expression has an effect. Bug? If not, what is the right way to have the "++" operator pass that error test?

Thanks, Zach

On 02/08/2012 10:48 PM, Zach the Mystic wrote:
> My goal is to be able to overload the "++" operator transparently to the
> code, but I can't.
>
> import std.stdio;
>
> struct Arc {
> int I = 0;
> // This is void, but the error appears under all return types
> void opUnary(string op)() if( op == "++" ) {
> ++I;
> }
> }
>
> struct HasArc {
> Arc myArc;
> }
>
> void main() {
> HasArc has;
> writefln(" Arc.I = %s", has.myArc.I); // Arc.I = 0
>
> has.myArc++; // Error: var has no effect in expression (__pitmp1481)
>
> // Okay, try this instead:
> auto UselessVar = has.myArc++; // Works fine
>
> writefln(" Arc.I = %s", has.myArc.I); // Arc.I = 1
> }
>
> Obviously the expression has an effect. Bug? If not, what is the right
> way to have the "++" operator pass that error test?
>
> Thanks, Zach

This is indeed a bug.

The expression is rewritten internally into (pseudo code)

(auto __pitmp1481 = has.myArc, has.myArc.opUnary!"++"(), __pitmp1481);

This introduces a sub-expression with no effect. You may want to file a bug report. For now, just using ++has.myArc; should work around the issue.

On 2/8/12 4:58 PM, Timon Gehr wrote:
> This is indeed a bug.
>
> The expression is rewritten internally into (pseudo code)
>
> (auto __pitmp1481 = has.myArc, has.myArc.opUnary!"++"(), __pitmp1481);
>
> This introduces a sub-expression with no effect.

Is that because has.myArc.opUnary!"++"() is going to be ignored at that level of compiler activity? The compiler demands only that __pitmp1481 has a value at that point?

> You may want to file a
> bug report. For now, just using ++has.myArc; should work around the issue.

http://d.puremagic.com/issues/show_bug.cgi?id=7467

Thanks, man. You're awesome!

On 02/09/2012 12:47 AM, Zach the Mystic wrote:
> On 2/8/12 4:58 PM, Timon Gehr wrote:
>> This is indeed a bug.
>>
>> The expression is rewritten internally into (pseudo code)
>>
>> (auto __pitmp1481 = has.myArc, has.myArc.opUnary!"++"(), __pitmp1481);
>>
>> This introduces a sub-expression with no effect.
>
> Is that because has.myArc.opUnary!"++"() is going to be ignored at that
> level of compiler activity? The compiler demands only that __pitmp1481
> has a value at that point?

This is the issue:

void main(){
    int x;
    ++x, x; // error
}

While the whole expression has an effect, the single 'x' does not have an effect.

>
>> You may want to file a
>> bug report. For now, just using ++has.myArc; should work around the
>> issue.
>
> http://d.puremagic.com/issues/show_bug.cgi?id=7467
>
> Thanks, man. You're awesome!

You're welcome. =)