How to flatten N-dimensional array?

[9il]

On Saturday, 23 May 2020 at 18:15:32 UTC, Pavel Shkadzko wrote:
> I have tried to implement a simple flatten function for 
> multidimensional arrays with recursive templates but got stuck. 
> Then I googled a little and stumped into complex 
> https://rosettacode.org/wiki/Flatten_a_list#D implementation 
> which requires your arrays to be either TreeList or Algebraic. 
> That is, it does not work out-of-the-box on something like 
> int[][][].
>
> I'd like to clarify a couple of questions first.
>
> How come Phobos doesn't have "flatten" function for arrays?
>
> Is there an implementation of flatten that works out-of-the-box 
> on N-dim arrays outside of Phobos? Excluding "flattened" from 
> mir.ndslice since it works on Slices.

If the common nd-array isn't jugged (a parallelotop), you can use 
fuse function.

----------
/+dub.sdl:
dependency "mir-algorithm" version="~>3.8.12"
+/
import std.stdio: writeln;
import mir.ndslice;

void main() {
     auto arr =
     [[[0, 1, 2, 3, 4],
       [5, 6, 7, 8, 9]],
      [[10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19]],
      [[20, 21, 22, 23, 24],
       [25, 26, 27, 28, 29]]];
     auto flatten = arr.fuse.field;

     static assert(is(typeof(flatten) == int[]));
    	assert(flatten == 30.iota);
}
----------

It performs exactly one allocation.