Re: Mir Slice.shape is not consistent with the actual array shape

[Pavel Shkadzko via Digitalmars-d-learn]

On Sunday, 24 May 2020 at 14:35:33 UTC, jmh530 wrote:
On Sunday, 24 May 2020 at 14:21:26 UTC, Pavel Shkadzko wrote:
[snip]
Sorry for the typo. It should be "auto arrSlice = a.sliced;"

Try using fuse

/+dub.sdl:
dependency "mir-algorithm" version="*"
+/
import std.stdio;
import std.conv;
import std.array: array;
import std.range: chunks;
import mir.ndslice;

int[] getShape(T : int)(T obj, int[] dims = null)
{
    return dims;
}

// return arr shape
int[] getShape(T)(T obj, int[] dims = null)
{
    dims ~= obj.length.to!int;
    return getShape!(typeof(obj[0]))(obj[0], dims);
}

void main() {
    int[] arr = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 
15, 16];
    int[][][] a = arr.chunks(4).array.chunks(2).array;

    int err;
    writeln(arr);
    writeln(a.shape(err));

    auto aSlice = a.fuse;
    writeln(aSlice);
    writeln(aSlice.shape);

}

Yes, this works. It doesn't explain why sliced behaves like this 
though. Also, isn't sliced the default way of converting D arrays 
to Mir arrays? Above all arr.fuse.field flattens the array. So, 
fuse works as D join if combined with field but as slice 
allocator is used on a D array...