Skip to main content

​
ravel_multi_index

Convert a tuple of index arrays into a flat index array for a given shape. 
function ravel_multi_index(
  multi_index: ArrayLike[],
  dims: number[],
  mode?: 'raise' | 'wrap' | 'clip'
): NDArray
NameTypeDefaultDescription
multi_indexArrayLike[]—Tuple of integer arrays, one per dimension.
dimsnumber[]—Shape of the array into which the indices apply.
mode'raise' | 'wrap' | 'clip''raise'How to handle out-of-bounds indices. 'raise' throws an error, 'wrap' wraps around, 'clip' clips to the valid range.
Returns: NDArray — Array of flat indices corresponding to the multi-dimensional indices. 
import * as np from 'numpy-ts';

const flat = np.ravel_multi_index([[0, 1, 2], [1, 0, 2]], [3, 3]);
// array([1, 3, 8])

// With wrapping for out-of-bounds
const wrapped = np.ravel_multi_index([[3], [1]], [3, 3], 'wrap');
// array([1])  -- row 3 wraps to row 0

​
unravel_index

Convert flat indices into a tuple of coordinate arrays for a given shape. 
function unravel_index(indices: ArrayLike | number, shape: number[]): NDArray[]
NameTypeDefaultDescription
indicesArrayLike—Array of flat indices to convert.
shapenumber[]—Shape of the array used for unraveling.
Returns: NDArray[] — Tuple of index arrays, one per dimension. Each array has the same shape as indices. 
import * as np from 'numpy-ts';

const coords = np.unravel_index([1, 3, 8], [3, 3]);
// coords[0]: array([0, 1, 2])  -- row indices
// coords[1]: array([1, 0, 2])  -- column indices

// Find the position of the max element in a 2D array
const a = np.array([[1, 5, 3], [9, 2, 7]]);
const maxIdx = np.argmax(a);
const [row, col] = np.unravel_index([maxIdx], a.shape);
// row: array([1]), col: array([0])  -- max is at position [1, 0]

​
iindex

Perform integer (fancy) indexing on an array. Select elements from a using an array of integer indices. 
function iindex(a: ArrayLike, indices: ArrayLike | number[][], axis?: number): NDArray
NameTypeDefaultDescription
aArrayLike—Source array.
indicesArrayLike—Integer array of indices. The result has the same shape as indices.
axisnumberundefinedAxis along which to index.
Returns: NDArray — Array of elements selected from a at the given flat indices. 
import * as np from 'numpy-ts';

const a = np.array([10, 20, 30, 40, 50]);
np.iindex(a, [0, 2, 4]);
// array([10, 30, 50])

// Fancy indexing with a 2D index array
np.iindex(a, [[0, 1], [3, 4]]);
// array([[10, 20],
//        [40, 50]])

​
bindex

Perform boolean array indexing on an array. Select elements from a where the mask is true. 
function bindex(a: ArrayLike, mask: ArrayLike, axis?: number): NDArray
NameTypeDefaultDescription
aArrayLike—Source array.
maskArrayLike—Boolean array. Must be broadcastable to the shape of a.
axisnumberundefinedAxis along which to index.
Returns: NDArray — 1-D array of elements from a where mask is true. 
import * as np from 'numpy-ts';

const a = np.array([1, 2, 3, 4, 5]);
np.bindex(a, [true, false, true, false, true]);
// array([1, 3, 5])

// Common pattern: filter by condition
const mat = np.array([[1, 2], [3, 4]]);
np.bindex(mat, np.greater(mat, 2));
// array([3, 4])