Fix missing void back(bool) in Array!bool.Range

[monarchdodra]

Forked from #1010

[ghost]

Is there a test-case already in place for this? I imagine not. :)

[monarchdodra]

Is there a test-case already in place for this? I imagine not. :)

Nope. I'll write the minimal test that covers all functions.

[monarchdodra]

Ended up adding some missing opDollar, a missing opSlice, and changing some size_t to ulong. Although arguably, I think I should have instead changed to size_t?

[ghost]

Why did it use ulong to begin with? I think we should use size_t.

[monarchdodra]

Why did it use ulong to begin with? I think we should use size_t.

The idea is that someone may want to allocate 8_000_000_000 bools on a 32 bit system, which is perfectly fine (takes 1G memory), but it requires a type larger than size_t to access all those bools.

That said, I don't think this special ability warrants the problems that come with not using size_t. C++'s vector<bool> also uses size_t: http://www.cplusplus.com/reference/vector/vector-bool/

I'll change everything back to size_t then. Seems smarter...

[ghost]

It seems like the job of Array!bool is to compact memory, but not necessarily allow a greater number of elements. So using size_t is appropriate IMO.

[jmdavis]

Any attempts to use anything other than size_t for length should be viewed with extreme prejudice. I don't know that we should never do it, but there had better be a very good reason for doing it, because dealing with it in generic code is going to be horrible. I think that it was a huge mistake to make it so that iota uses anything other than size_t for length, and I don't want to see the same mistake elsewhere (in fact, I'd love to see it removed in iota). If someone really needs that many elements on a 32-bit system, they should probably just roll their own solution. The cost to supporting it in the standard library is too high IMHO.

[monarchdodra]

Done. I also filed an issue for it:
http://d.puremagic.com/issues/show_bug.cgi?id=11349
Since I suppose it would warrant an entry in the change log.

Also, removed some gratuitous (IMO) Tuple usage, when a simple struct would have done.

I think this is good for review now.

[ghost]

LGTM.