On Sunday, 24 May 2020 at 11:21:00 UTC, 9il wrote:
On Saturday, 23 May 2020 at 18:15:32 UTC, Pavel Shkadzko wrote:
[...]
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.
Yep, I know about Mir fuse. I was more wondering about the
absence of flatten in Phobos. But on dlang forum some people
claimed that when it comes to anything multidimensional, use Mir.