In documentation and forums I found some example for overloading opCmp for int values. But I couldn't see any examples for double values.

That is what I come up with my own:

struct AdjustableVal ( T = double )
{
	this ( T initVal )
	{
		curVal = initVal;
		initialVal = initVal;
	}
	
	
    int opCmp( T rhs ) const {
        auto diff = curVal - rhs;
        if (  fabs(diff) < 0.00000001 )
            return 0;
        else if ( diff < 0 )
            return -1;
        else
            return 1;
     }

	T curVal;
	T initialVal;
}

Do you guys see any problem with it or any suggestions?
Secondly if the value type(T) does not have "-" operator, is it possible to still get my code compiled somehow with static_if?


Erdem

On 12/24/2017 02:10 AM, kerdemdemir wrote:

>          if (  fabs(diff) < 0.00000001 )

I can't answer your question but I know that the comparison above is wrong because it can be meaningful only if the values are sufficiently larger than that hard-coded value.

I think feqrel() will be useful there:

  https://dlang.org/phobos/std_math.html#.feqrel

Ali

25.12.2017 09:50, Ali Çehreli пишет:
> On 12/24/2017 02:10 AM, kerdemdemir wrote:
> 
>  >          if (  fabs(diff) < 0.00000001 )
> 
> I can't answer your question but I know that the comparison above is wrong because it can be meaningful only if the values are sufficiently larger than that hard-coded value.
> 
> I think feqrel() will be useful there:
> 
>    https://dlang.org/phobos/std_math.html#.feqrel
> 
> Ali
> 
I'd do something like that (disclaimer - from memory):
```
    int opCmp( T rhs ) const {
        auto diff = curVal - rhs;
	auto epsilon = max(curVal.epsilon, rhs.epsilon);
        if (  fabs(diff) < epsilon )
            return 0;
        else if ( diff < 0 )
            return -1;
        else
            return 1;
    }
```
so in this case epsilon is also floating.

25.12.2017 10:40, drug пишет:
> 25.12.2017 09:50, Ali Çehreli пишет:
>> On 12/24/2017 02:10 AM, kerdemdemir wrote:
>>
>>  >          if (  fabs(diff) < 0.00000001 )
>>
>> I can't answer your question but I know that the comparison above is wrong because it can be meaningful only if the values are sufficiently larger than that hard-coded value.
>>
>> I think feqrel() will be useful there:
>>
>>    https://dlang.org/phobos/std_math.html#.feqrel
>>
>> Ali
>>
> I'd do something like that (disclaimer - from memory):
> ```
>      int opCmp( T rhs ) const {
>          auto diff = curVal - rhs;
>      auto epsilon = max(curVal.epsilon, rhs.epsilon);
>          if (  fabs(diff) < epsilon )
>              return 0;
>          else if ( diff < 0 )
>              return -1;
>          else
>              return 1;
>      }
> ```
> so in this case epsilon is also floating.

should be
 ```
      int opCmp( T rhs ) const {
          auto diff = curVal - rhs;
          // get error for curVal and rhs
          auto epsilon = max(curVal.epsilon * curVal, rhs.epsilon * rhs);
          if (  fabs(diff) < epsilon )
              return 0;
          else if ( diff < 0 )
              return -1;
          else
              return 1;
      }
 ```