"The Case for D" on Ycombinator

[bearophile]

Andrei Alexandrescu:

> I read the exchange just now (nice job ponce) and am a bit surprised how 
> much others know about what I do, e.g how I don't use D. My entire 
> doctoral research is implemented in D (to say nothing about phobos). I 
> haven't used another language extensively for about three years.

Like everything else, templates too have some disadvantages and some advantages. For the purposes of D I think that using templates is often positive. Haskell, Java, C and C# show worse alternatives when you want to define typed collections.

And some things in future may be improved a little to make D template usage a bit simpler or less visually cluttered. Introducing private/public names in templates is a possible positive step. Other ideas are possible, we have to think about them. (Time ago I have suggested a "type" type, to replace some functional-style template usage with simpler imperative-style compile time functions, but it was not appreciated. Maybe also because it introduces yet another way to do the same thing, despite it looks like a simpler way).

Regarding Phobos2, some of its details may not be "battle tested", and some its APIs may not be the best yet, some names may not be the best chosen (for example some people think that "iota" is not the best name), but such bad details may be fixed in future if you and others are willing refine the code. I don't see basic things wrong in Phobos2.

(There's only an interesting change I'd like to propose for Phobos2, I'll talk about it in another post. And it may be a bad/useless idea).

Bye,
bearophile