On Wednesday, 21 October 2015 at 20:50:29 UTC, Andrei
Alexandrescu wrote:
Better late than later.
http://wiki.dlang.org/Vision/2015H2_(draft)
Destroy. After we make this good I'll rename it and make it
official.
Andrei
Memory management
"No GC" remains a primary concern going forward for the standard
library and client code.
We need good language+library support and robust idioms to
follow for good no-GC code.
language-level smart pointers would be a good start. Library
level ones feel like an ugly hack honestly, its been repeatedly
beaten to death all the issues there is in implementing them at
library-level and the only solutions are more ugly DIPs to duct
tape it together - look, DIP25 required another DIP just to fix
the DIP itself. http://wiki.dlang.org/DIP77
Another major issue is the lack of -> like C++, which means that
methods on the smart pointer can't be disambiguated from methods
on the object it's wrapping.
D's entire memory management system almost feels C arcane
compared to where C++ is moving.
Scope as an attribute is still mostly unimplemented AFAIK.
Honestly, I think the new C++ guidelines have a lot of goodies in
it...
I'd like to just be able to use D without the GC. I don't mean a
gimped version of D where I might as well just be using another
language. D's GC is just slow, and honestly, not viable for most
tasks you're going to use a system's level language for. Yeah,
I'm going to get flak for that and people are going to cherrypick
data that shows that a GC'd language has been used in some OS
that was used by 7 people total or something, but lets not delude
ourselves - there's a reason why everytime D gets optimized for a
benchmark the first thing anyone does is essentially nix all
allocations and replace classes with structs.
