Document multidimensional array op overloading

[quickfur]

Cf. dmd pull: dlang/dmd#443

[9rnsr]

Excellent!

An additional point: In arr[1, 2..3, $], arr is always evaluated only once, even when it is a function call. For example, foo()[1, 2..3, $; is rewritten to auto __tmp = foo(); __tmp.opIndex(1, __tmp.opSlice!1(2, 3), __tmp.opDollar!2());

Can you documenti the behavior?

[quickfur]

What about opDollar? Is it also only evaluated once if it occurs multiple times within the same parameter? E.g., is arr[$/2 - sqrt($)] equivalent to:

auto __tmp = arr.opDollar!1;
arr.opIndex(__tmp/2 - sqrt(__tmp));

?

[9rnsr]

Yes, opDollar is also evaluated at most once in each dimensions.

[quickfur]

Updated. Anything else?

[9rnsr]

Nothing. Thanks!

[MartinNowak]

There is no code example on how to integrate opIndex and opSlice, e.g. this does not work.

struct Array
{
    auto opIndex(size_t[2] bounds)
    {
        return ary[bounds[0] .. bounds[1]];
    }
    size_t[2] opSlice(size_t lwr, size_t upr)
    {
        return [lwr, upr];
    }
    ubyte[] ary;
}

But this does.

struct Array
{
    auto opIndex(size_t[2] bounds)
    {
        return ary[bounds[0] .. bounds[1]];
    }
    size_t[2] opSlice(size_t dim)(size_t lwr, size_t upr)
    {
        return [lwr, upr];
    }
    ubyte[] ary;
}

[quickfur]

#1095