Re: How to flatten N-dimensional array?
On 5/24/20 2:37 AM, Pavel Shkadzko wrote:
On Saturday, 23 May 2020 at 19:59:30 UTC, Ali Çehreli wrote:
On 5/23/20 11:15 AM, Pavel Shkadzko wrote:> I have tried to implement
a simple flatten function for multidimensional
[...]
Thank you, I was lacking practical examples for templates with "if"
constructs, ehh.
Template constraints are great but 'static if' can be more useful as it
allows custom error messages:
auto makeNdim(size_t N)(size_t length) {
static if (N == 1) {
auto result = iota(value, value + length).array;
value += length;
return result;
} else static if (N > 1) {
return iota(N).map!(n => makeNdim!(N - 1)(length)).array;
} else {
static assert(false, "N cannot be 0.");
}
}
Ali
Re: How to flatten N-dimensional array?
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.
Re: How to flatten N-dimensional array?
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.
Re: How to flatten N-dimensional array?
On Saturday, 23 May 2020 at 19:59:30 UTC, Ali Çehreli wrote: On 5/23/20 11:15 AM, Pavel Shkadzko wrote:> I have tried to implement a simple flatten function for multidimensional [...] Thank you, I was lacking practical examples for templates with "if" constructs, ehh.
Re: How to flatten N-dimensional array?
On 5/23/20 11:15 AM, Pavel Shkadzko wrote:> I have tried to implement a
simple flatten function for multidimensional
> I'd like to clarify a couple of questions first.
>
> How come Phobos doesn't have "flatten" function for arrays?
We call in 'joiner'.
I wrote something like this:
import std.stdio;
import std.algorithm;
import std.range;
int value = 0;
auto makeNdim(size_t N)(size_t length)
if (N == 1) {
auto result = iota(value, value + length).array;
value += length;
return result;
}
auto makeNdim(size_t N)(size_t length)
if (N > 1) {
return iota(N).map!(n => makeNdim!(N - 1)(length)).array;
}
auto flatten(R)(R range)
if (!isInputRange!(ElementType!R)) {
return range.joiner;
}
auto flatten(R)(R range)
if (isInputRange!(ElementType!R)) {
return range.map!(r => r.joiner).joiner;
}
void main() {
auto a = makeNdim!3(5);
writefln!"Original : %s"(a);
writefln!"Flattened: %s"(a.flatten);
}
Output:
Original : [[[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]]]
Flattened: [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]
Ali
How to flatten N-dimensional array?
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.
