dnarray

Creates an N-dimensional array where each element is computed by a flat-index init function. Use the explicit-dimensions overload for 5+ dimensions, or the shape-based overload to create arrays of any dimension from a dynamically computed shape.

Signatures

From explicit dimensions (5+)

inline fun <reified T : Any> Multik.dnarray(
    sizeD1: Int,
    sizeD2: Int,
    sizeD3: Int,
    sizeD4: Int,
    vararg dims: Int,
    noinline init: (Int) -> T
): NDArray<T, DN>

From shape array

inline fun <reified T : Any, reified D : Dimension> Multik.dnarray(
    shape: IntArray,
    noinline init: (Int) -> T
): NDArray<T, D>

Parameters

Parameter	Type	Description
`sizeD1..sizeD4`	`Int`	First four axis sizes. Must be positive.
`dims`	`vararg Int`	Additional axis sizes beyond the fourth.
`shape`	`IntArray`	Complete shape array. All elements must be positive.
`init`	`(Int) -> T`	Lambda that receives a flat index. Elements are filled in row-major order.

Returns: NDArray<T, DN> (varargs version) or NDArray<T, D> (shape version)

Example

// 5D array
val a = mk.dnarray<Int>(2, 3, 4, 5, 6) { it }
// shape (2, 3, 4, 5, 6), 720 elements

// Using shape array with reified dimension
val b = mk.dnarray<Double, D3>(intArrayOf(2, 3, 4)) { it.toDouble() }
// shape (2, 3, 4)

28 February 2026