Hello. I just came upon a need in my program to make binary arithmetic operators valid between two real[] in my programs, and thought of writing a global opOpAssign, but then looked through the documentation, found nothing on operator overloading allowed at the global level (even within a single module), and searched through the forum and saw this thread: http://forum.dlang.org/post/jeyaozvegcnivcppucpo@forum.dlang.org

Why isn't global operator overloading allowed in D?

Having to construct a class just to write a quick operator overload for two built-in types isn't a clean solution, so please don't suggest that (like it was suggested in the other thread).

Thanks.

Shriramana Sharma.

Shriramana Sharma wrote:

> Hello. I just came upon a need in my program to make binary arithmetic operators valid between two real[] in my programs, and thought of writing a global opOpAssign, but then looked through the documentation, found nothing on operator overloading allowed at the global level (even within a single module), and searched through the forum and saw this thread: http://forum.dlang.org/post/jeyaozvegcnivcppucpo@forum.dlang.org
> 
> Why isn't global operator overloading allowed in D?

Hmm, just now saw this one too: http://forum.dlang.org/post/l7nnnb$1ma0$1@digitalmars.com -- posted that too hastily.

But really, it's too bad built-in types are different from user-made types unlike in Python. In Python, I can inherit from list and dict and whatnot (yes I have used them in real apps)...

On Wednesday, 14 October 2015 at 15:02:02 UTC, Shriramana Sharma wrote:
> Hello. I just came upon a need in my program to make binary arithmetic operators valid between two real[] in my programs

What binary arithmetic operators do you need that real[] doesn't already support?

On Wednesday, 14 October 2015 at 15:02:02 UTC, Shriramana Sharma wrote:
> Hello. I just came upon a need in my program to make binary arithmetic operators valid between two real[] in my programs, and thought of writing a global opOpAssign, but then looked through the documentation, found nothing on operator overloading allowed at the global level (even within a single module), and searched through the forum and saw this thread: http://forum.dlang.org/post/jeyaozvegcnivcppucpo@forum.dlang.org
>
> Why isn't global operator overloading allowed in D?
>
> Having to construct a class just to write a quick operator overload for two built-in types isn't a clean solution, so please don't suggest that (like it was suggested in the other thread).
>
> Thanks.
>
> Shriramana Sharma.

Is Writing a global function an option ? With UFCS it will still look correct...
And under the hood, an operator overload is a function anyway.
With string template parameter it could even display the symbol:

(not tested but you shoud get the idea)
---
void assignResultOf(string op, L, R)(ref L lhs, ref R rhs) {}

assignResultOf!"+"'(dest, src);
dest.assignResultOf!"+"'(src);
---

otherwise struct + alias this + member operator overload is the most obvious way to overcome the problem...but it seems you don't like this.

John Colvin wrote:

> On Wednesday, 14 October 2015 at 15:02:02 UTC, Shriramana Sharma
> wrote:
> What binary arithmetic operators do you need that real[] doesn't
> already support?

OMG silly me! I can already do a[] /= b[]... D is great! :-D Thanks a lot!

On Thursday, 15 October 2015 at 15:45:00 UTC, Shriramana Sharma wrote:
> John Colvin wrote:
>
>> On Wednesday, 14 October 2015 at 15:02:02 UTC, Shriramana Sharma
>> wrote:
>> What binary arithmetic operators do you need that real[] doesn't
>> already support?
>
> OMG silly me! I can already do a[] /= b[]... D is great! :-D Thanks a lot!

Also:

a[] = b[] + c[] * d[] - 42;

and so on... All that's required is that there is pre-allocated memory for the result to go in i.e. there has to be enough space in a[].

You should be aware that with DMD these array operations should be much faster than a straightforward loop, as they are done in handwritten asm using vector instructions. Be wary of using them on very small arrays, there is some overhead.

With LDC/GDC you probably wont see much difference either way, if I remember correctly they rely on the optimiser instead.