The implementation of std.container.Array!bool makes use of ulong instead of size_t as far as sizes and indices are concerned. This makes Array!bool inconsistent with other array-like types (including Array!T if !is(T==bool)) in 32-bit code. What would be the disadvantages of using size_t?

Thanks,
Nicolas

On 15.05.2012 13:44, Nicolas Sicard wrote:
> The implementation of std.container.Array!bool makes use of ulong
> instead of size_t as far as sizes and indices are concerned. This makes
> Array!bool inconsistent with other array-like types (including Array!T
> if !is(T==bool)) in 32-bit code. What would be the disadvantages of
> using size_t?
>
> Thanks,
> Nicolas
>
theoretically you can pack 8*2^30+ bits in 1Gb+ of memory. While it's doubtful if you can always fetch such big piece of contiguous memory from OS it's still possible. And 32bit bit-indexes just fail to address it.

-- 
Dmitry Olshansky

On Tuesday, 15 May 2012 at 10:04:52 UTC, Dmitry Olshansky wrote:
> On 15.05.2012 13:44, Nicolas Sicard wrote:
>> The implementation of std.container.Array!bool makes use of ulong
>> instead of size_t as far as sizes and indices are concerned. This makes
>> Array!bool inconsistent with other array-like types (including Array!T
>> if !is(T==bool)) in 32-bit code. What would be the disadvantages of
>> using size_t?
>>
>> Thanks,
>> Nicolas
>>
> theoretically you can pack 8*2^30+ bits in 1Gb+ of memory. While it's doubtful if you can always fetch such big piece of contiguous memory from OS it's still possible. And 32bit bit-indexes just fail to address it.

Yes, I understand this limitation. But from another theoretical point of view, length not returning a size_t is odd for a container (std.bitmanip.BitArray.length did).

But I am ok if it is more needful for an array of bool to be able to address the full memory...

Thanks,
Nicolas