I do a lot of matrix calculations with a D library that I have written and it would be really nice to be able to overload a few more array operators, like assigned multiplication (i.e. mat1[i,j] += mat2[i,j];).  Has this feature been approved, discussed, or planned?
Thanks,
JC

Jonathan Crapuchettes wrote:

> I do a lot of matrix calculations with a D library that I have written and it would be really nice to be able to overload a few more array operators, like assigned multiplication (i.e. mat1[i,j] += mat2[i,j];).  Has this feature been approved, discussed, or planned? Thanks, JC

What is preventing you from doing so now? if opIndex returns a type that overloads opAddAssign your example should be possible right now.

Johan Granberg wrote:
> Jonathan Crapuchettes wrote:
> 
>> I do a lot of matrix calculations with a D library that I have written and
>> it would be really nice to be able to overload a few more array operators,
>> like assigned multiplication (i.e. mat1[i,j] += mat2[i,j];).  Has this
>> feature been approved, discussed, or planned? Thanks, JC
> 
> What is preventing you from doing so now? if opIndex returns a type that
> overloads opAddAssign your example should be possible right now.

Typically, if you have a matrix of T you'd want to have mat[i,j] (or mat[i][j]) return either a T or something that implicitly converts to one. Since you can only overload operators on aggregates (class, struct, union) but can't define implicit conversions for them...

Frits van Bommel wrote:

> Johan Granberg wrote:
>> Jonathan Crapuchettes wrote:
>> 
>>> I do a lot of matrix calculations with a D library that I have written
>>> and it would be really nice to be able to overload a few more array
>>> operators,
>>> like assigned multiplication (i.e. mat1[i,j] += mat2[i,j];).  Has this
>>> feature been approved, discussed, or planned? Thanks, JC
>> 
>> What is preventing you from doing so now? if opIndex returns a type that overloads opAddAssign your example should be possible right now.
> 
> Typically, if you have a matrix of T you'd want to have mat[i,j] (or mat[i][j]) return either a T or something that implicitly converts to one. Since you can only overload operators on aggregates (class, struct, union) but can't define implicit conversions for them...

But if T supports both opIndex and opAssign that wouldn't be a problem, right?

Johan Granberg wrote:
> Frits van Bommel wrote:
> 
>> Johan Granberg wrote:
>>> Jonathan Crapuchettes wrote:
>>>
>>>> I do a lot of matrix calculations with a D library that I have written
>>>> and it would be really nice to be able to overload a few more array
>>>> operators,
>>>> like assigned multiplication (i.e. mat1[i,j] += mat2[i,j];).  Has this
>>>> feature been approved, discussed, or planned? Thanks, JC
>>> What is preventing you from doing so now? if opIndex returns a type that
>>> overloads opAddAssign your example should be possible right now.
>> Typically, if you have a matrix of T you'd want to have mat[i,j] (or
>> mat[i][j]) return either a T or something that implicitly converts to
>> one. Since you can only overload operators on aggregates (class, struct,
>> union) but can't define implicit conversions for them...
> 
> But if T supports both opIndex and opAssign that wouldn't be a problem,
> right?

(Presumably you mean opAddAssign instead of opIndex?)

Only if T is a struct or union[1], and its author was aware of the need to put it in that particular Matrix implementation (or you're willing to modify it to fit[2]).
If it's a primitive type like (int, float, cfloat) and variants, or a class (hint: opAssign can't modify the reference) you're screwed.


[1]: Requirements: value type with possibility of operator overloading.
[2]: Hopefully by way of a mixin, since there's a lot of op<X>Assign operators.