Using DList with capacity constraint (aka circular buffer)?

[bitwise via Digitalmars-d]

On Wednesday, 24 February 2016 at 14:46:12 UTC, Seb wrote:
> Hey all,
>
> this is yet another post about phobos (missing) data structures 
> ;-)
> I know this has been discussed quite a bit - [1,2,3] to name a 
> few.
>
> While it would be nice to have those "trivially to implement" 
> wrappers for some common use cases (map, unordered map, set, 
> multiset, ...) [1], this question focuses solely on dequeues.
>
> It is great that we have DList, so having a circular buffer 
> (aka constrained queue) should be easy.
>
> I do understand that baking this into DList  (as e.g. Python 
> does) might (a) make things more complex or (b) add overhead 
> that isn't needed for every user.
>
> However my question is: why is there not a neat wrapper around 
> DList with a capacity constraint?
>
> Unfortunately we don't have inheritance for structs, but 
> proxying most methods should work too (see e.g [4]).
>
> 1) Has anyone else missed deque with capacity constraints?
> 2) Would such a wrapper fit into phobos?
> 3) Would you prefer (a) a wrapper around DList [4], (b) around 
> arrays [5] or (c) a "vanilla" circular queue?
> (b is slower, but allows indexing)
>
> Best wishes,
>
> [1] 
> https://stackoverflow.com/questions/7162274/why-is-d-missing-container-classes
> [2] 
> http://forum.dlang.org/thread/mailman.852.1359488662.22503.digitalmars-d@puremagic.com
> [3] 
> http://forum.dlang.org/post/mailman.394.1358112013.22503.digitalmars-d@puremagic.com
> [4] http://dpaste.dzfl.pl/d8de9325e9a3
> [5] http://rosettacode.org/wiki/Queue/Usage#Faster_Version

Andrei is working on containers, but struggling with trying to 
make them @safe without comprimising efficiency or utility, 
AFAIK. Getting the containers done to his liking may require the 
work on lifetimes(language supported ref counting) to be 
complete. IIRC, there was a circular buffer on code.dlang.org 
somewhere. I've been planning on adding one to my container 
set(not on code.dlang.org yet) which would be implemented on a 
contiguous array, with the contents wrapping around as items are 
added/removed. I believe it's the most efficient approach.

    Bit