proposed @noreturn attribute

[Meta via Digitalmars-d]

On Sunday, 9 July 2017 at 04:23:15 UTC, Meta wrote:
> On Sunday, 9 July 2017 at 02:25:50 UTC, Walter Bright wrote:
>> (D already has a `void` type, so can't use Haskell's word.)
>
> Just so we are all on the same page, from a type-theory 
> perspective void is a unit type (it has 1 value), not an 
> uninhabited type (it has no values, i.e. Haskell's _|_ (bottom) 
> type).
>
> A function with a return type of unit means "this function 
> returns no useful information", because the type only has one 
> possible value anyway. A function with a return type of bottom 
> means "this function can never return", because there is no 
> value of type bottom that could be returned. All that can be 
> done is for the function to diverge (throwing an exception, 
> ending the program, looping forever, etc.).
>
> We sort of have this already with `assert(0)`. The compiler 
> knows that no execution can take place after an `assert(0)` is 
> encountered (in other words, it knows that the function 
> diverges). We just don't have a corresponding type to represent 
> this (Rust uses ! but if I remember correctly it's not quite a 
> first class type).
>
> If we wanted to be cute we could use `typeof()` to represent 
> this type as there is no value you can give to typeof such that 
> it returns the bottom type. It also avoids having to come up 
> with some special symbol or name for it.

In C++/D terms, the unit type could be expressed as `enum Unit { 
unit }` while the bottom type would be `enum Bottom {}` (although 
only conceptually because in C++/D you can still do `Bottom b;`).