I'm porting to kotlin this code ```snap_descriptor...

# science

[elect]

I'm porting to kotlin this code

snap_descriptors_np = extract_compute_np(
    lmp,
    "bgrid",
    0,
    2,
    (nz, ny, nx, self.fingerprint_length),
    use_fp64=use_fp64,
)
lmp.close()

# switch from x-fastest to z-fastest order (swaps 0th and 2nd dimension)
snap_descriptors_np = snap_descriptors_np.transpose([2, 1, 0, 3])

where extract_compute_np returns a flat doubleArray representing an ndarray of [nz, ny, nx, self.fingerprint_length]
now, I have to swap the 0 and 2nd dimensions and I'm trying to do this:

Array(nx) { x ->
    Array(ny) { y ->
        Array(nz) { z ->
            val ofsX = x * (ny * nz * fingerprintLength)
            val ofsY = y * (nz * fingerprintLength)
            val ofsZ = z * fingerprintLength
            DoubleArray(fingerprintLength) { doubles[ofsX + ofsY + ofsZ + it] }
        }
    }
}

but it looks like it's wrong, anyone spots the problem?

[Gleb Minaev]

I tried solving the problem myself. All you have to do is to write a transformation to Array<Array<Array<DoubleArray>>> with initial ndarray [nz, ny, nx, fingerprintLength] and then replace Array(nx) { x -> with Array(nz) { z -> and vice versa to swap x and z dimensions.

Here is how I solved it: https://pl.kotl.in/y3FM_tGzS

[elect]

thanks Gleb, that looks working!

[elect]

I was going a little mad

[altavir]

I think that @Ролан did something similar in tensor extensions for KMath, but it makes sense only if you need a zero-copy solution.