How to best translate this C++ algorithm into D?

[logicchains via Digitalmars-d]

On Saturday, 7 June 2014 at 08:27:52 UTC, deadalnix wrote:
> This may ends up doing many allocations. you could use an array 
> where you preallocate forests.length * 3 items via the reserve 
> function and use it as an output range.
>

I've changed filter to partition, which speeds it up slightly, 
but I'm not exactly sure how to use a pre-allocated array as an 
output range. I've tried:

forest_t[] newForests;
newForests.reserve(forests.length*3);
map!(a => [forest_t(-1, -1, +1)+a, forest_t(-1, +1, -1)+a, 
forest_t(+1, -1, -1)+a])(forests)
   .join
   .partition!(forest_invalid)
   .sort.uniq.copy(newForests);

But this leaves newForests with length zero, even though indexing 
it via for instance newForests[0] gives a correct forest.

> Try
> forests.map!(a => only(possible_meals[0]+a, possible_meals[1]+a,
> possible_meals[2]+a))

'only' doesn't seem to be working on my version of ldc (latest 
Arch Linux community build); even a simple `auto tst = only("1", 
"2", "3");` gives a compile error, even though the same thing 
works fine on dmd. The best I can get is:

  return map!(a => [forest_t(-1, -1, +1)+a, forest_t(-1, +1, 
-1)+a, forest_t(+1, -1, -1)+a])(forests)
   .join
   .partition!(forest_invalid)
   .sort.uniq.array;