BigInt -- a challenge for std.allocator

[Kagamin]

On Tuesday, 29 October 2013 at 21:59:38 UTC, Joseph Rushton 
Wakeling wrote:
> On 29/10/13 15:58, Dmitry Olshansky wrote:
>> Can't it use ref-counted COW? Then since there is only 1 
>> reference to the
>> original block it may mutate it in-place else it creates new 
>> chunk with refs=1
>> and decrements the original count. Maybe I'm missing something 
>> but it shouldn't
>> be that hard.
>
> Yes, that's pretty much what I was thinking of.  The question 
> is how to implement it in terms of std.allocator -- or is that 
> the wrong base to build ref=counted memory upon?

Implement a pool of bigints as an allocator: when refcount 
reaches zero, return the bigint buffer to the pool - a simple 
bump the pointer algorithm, when you need a new buffer, request 
it from the pool. A pool is a stack with fixed capacity, and 
buffers go in and out.