Static inner functions

[Jonathan M Davis]

On Wednesday, December 22, 2010 10:17:11 bearophile wrote:
> A little D2 program:
> 
> void main() {
>            pure nothrow int foo1(immutable int x) { return x; }
>     static pure nothrow int foo2(immutable int x) { return x; }
> }
> 
> 
> This is the asm of the two inner functions:
> 
> _D6test4mainFZv4foo1MFNaNbyiZi    comdat
>         enter   4,0
>         mov EAX,8[EBP]
>         leave
>         ret 4
> 
> _D6test4mainFZv4foo2FNaNbyiZi comdat
>         enter   4,0
>         leave
>         ret
> 
> Is this a compiler mistake? Aren't strongly pure inner function static too?

Even if they are conceptually static (and I'm not sure if they are), the 
compiler would have to recognize them as being static, which it obviously 
doesn't. Presumably only considers a function static if it's marked that way. 
But arguably, it's _not_ the same because when dealing with a static inner 
function, you're essentially dealing with a function pointer whereas when 
dealing with a non-static inner function, you're essentially dealing with a 
delegate. Now, assuming that it really doesn't make sense for a delegate to be 
pure (presumably with the idea that accessing its outer scope would be impure), 
then I would think that it would make more sense to disallow pure delegates and 
pure non-static inner functions than to just make them static. But I'm not sure 
what all the implications of combining purity and delegates are, so without 
studying it a fair bit more, I'm not quite sure whether it really makes sense to 
have pure delegates.

- Jonathan M Davis