I may be just tired, but could somebody explain this behaviour to me? It seems odd to me:

----
interface I
{
}

class S : I
{
}

void test1(I[])
{
}

void test2(I)
{
}

void main()
{
	test1([new S()]); // Works
	test2(new S()); // Works
	
	I i = new S();
	test2(i); // Works
	
	S s = new S();
	test2(s); // Works
	
	I[] si = [new S()];
	test1(si); // Works
	
	S[] ss = [new S()];
	test1(ss); // Fails
}
----

Compiler output: test.d(32): Error: function test1 (I[] _param_0) is not callable using argument types (S[])

Why isn't the compiler able to deduce S[] => I[]? Or is it just me?
I've tried dmd 2.075

On Monday, 24 July 2017 at 17:29:55 UTC, Dgame wrote:
> 	S[] ss = [new S()];
> 	test1(ss); // Fails
>
> Why isn't the compiler able to deduce S[] => I[]? Or is it just me?

This is exactly because of polymorphism. Consider the following:

```
S[] ss = [new S()];
I[] i = ss; // pass it to the function or whatever for implicit conversion

class OtherDerived : I {}

i[0] = new OtherDerived(); // looks OK, otherDerived is also interface I
```

But now, ss[0], the same array as i, no longer points to an S! You broke the type system.

On Monday, 24 July 2017 at 17:33:48 UTC, Adam D. Ruppe wrote:
> On Monday, 24 July 2017 at 17:29:55 UTC, Dgame wrote:
>> 	S[] ss = [new S()];
>> 	test1(ss); // Fails
>>
>> Why isn't the compiler able to deduce S[] => I[]? Or is it just me?
>
> This is exactly because of polymorphism. Consider the following:
>
> ```
> S[] ss = [new S()];
> I[] i = ss; // pass it to the function or whatever for implicit conversion
>
> class OtherDerived : I {}
>
> i[0] = new OtherDerived(); // looks OK, otherDerived is also interface I
> ```
>
> But now, ss[0], the same array as i, no longer points to an S! You broke the type system.

So, tired it is. Thanks a lot.

On 7/24/17 1:29 PM, Dgame wrote:

> Why isn't the compiler able to deduce S[] => I[]? Or is it just me?
> I've tried dmd 2.075

I know you got the explanation already, but just in case you actually need to call something like test1 but only have an S[]:

test1(ss.map!((I i) => i).array)

-Steve

On Monday, 24 July 2017 at 18:15:20 UTC, Steven Schveighoffer wrote:
> On 7/24/17 1:29 PM, Dgame wrote:
>
>> Why isn't the compiler able to deduce S[] => I[]? Or is it just me?
>> I've tried dmd 2.075
>
> I know you got the explanation already, but just in case you actually need to call something like test1 but only have an S[]:
>
> test1(ss.map!((I i) => i).array)
>
> -Steve

Thanks.