Thread overview
getOverloads order
Jul 13, 2023
IchorDev
Jul 13, 2023
Dennis
Jul 13, 2023
IchorDev
Jul 13, 2023
Dennis
July 13, 2023

I've noticed that __traits(getOverloads) always returns the overloads in lexical order across DMD, LDC, and GDC. Is this reliable at all?

July 13, 2023

On Thursday, 13 July 2023 at 08:03:02 UTC, IchorDev wrote:

>

I've noticed that __traits(getOverloads) always returns the overloads in lexical order across DMD, LDC, and GDC. Is this reliable at all?

No. It depends on the order the compiler analyzes the symbols, which is often lexical order, but it can vary based on static if, mixin, forward references etc. Here's a counter example:

       void f(int  x);
mixin("void f(float y);");
       void f(char z);

Here you get overloads of f in the order (x, z, y) instead of (x, y, z).

July 13, 2023

On Thursday, 13 July 2023 at 10:53:49 UTC, Dennis wrote:

>

On Thursday, 13 July 2023 at 08:03:02 UTC, IchorDev wrote:

>

I've noticed that __traits(getOverloads) always returns the overloads in lexical order across DMD, LDC, and GDC. Is this reliable at all?

No. It depends on the order the compiler analyzes the symbols, which is often lexical order, but it can vary based on static if, mixin, forward references etc. Here's a counter example:

       void f(int  x);
mixin("void f(float y);");
       void f(char z);

Here you get overloads of f in the order (x, z, y) instead of (x, y, z).

Well that makes sense, but also wouldn't apply to the use-case that I was considering, since all the code would be in one mixin. However, the spec doesn't specify that this is how getOverloads must work; is this guaranteed behaviour but the spec simply omits it?

July 13, 2023

On Thursday, 13 July 2023 at 11:04:40 UTC, IchorDev wrote:

>

However, the spec doesn't specify that this is how getOverloads must work; is this guaranteed behaviour but the spec simply omits it?

The order is not guaranteed. I don't know why you need a specific order, but perhaps you can sort based on __traits(getLocation).