ref for (const) variables

[via Digitalmars-d-learn]

On Monday, 16 March 2015 at 19:20:09 UTC, anonymous wrote:
> On Monday, 16 March 2015 at 18:47:00 UTC, Namespace wrote:
>>    const(Matrix)* m = &t.getCurrentModelViewMatrix(); // 
>> currently
>> }
>> ----
>>
>> But IMO it would be a lot nicer if I could store the reference 
>> like this:
>> ----
>> ref const(Matrix) m = t.getCurrentModelViewMatrix(); // nicer
>> ----
>>
>> [Of course the name is exaggerated for the purpose of 
>> demonstration.]
>>
>> May this be worth of an enhancement request?
>
> Maybe, but I think you'd have to present a better argument. 
> It's not obvious to me how `ref T x = y;` is supposed to be a 
> lot nicer than `T* x = &y;`.
>
>> Or was this  already rejected?
>
> I don't know. But since it's a C++ thing, it's probably been 
> discussed.

The last time we discussed about scope, Walter suggested allowing 
`ref` for local variables. He might not be opposed to extend it 
further to members. On the other hand, `ref` has gained different 
semantics with DIP25 (something like scope, more or less), so 
this might collide with such a change.

Anyway, the situation in D is not the same as in C++, because 
`ref` is not a type modifier, but a storage class.