diags_array — SciPy v1.15.2 Manual (original) (raw)

scipy.sparse.

scipy.sparse.diags_array(diagonals, /, *, offsets=0, shape=None, format=None, dtype=None)[source]#

Construct a sparse array from diagonals.

Parameters:

diagonalssequence of array_like

Sequence of arrays containing the array diagonals, corresponding to offsets.

offsetssequence of int or an int, optional

Diagonals to set (repeated offsets are not allowed):

shapetuple of int, optional

Shape of the result. If omitted, a square array large enough to contain the diagonals is returned.

format{“dia”, “csr”, “csc”, “lil”, …}, optional

Matrix format of the result. By default (format=None) an appropriate sparse array format is returned. This choice is subject to change.

dtypedtype, optional

Data type of the array.

Notes

Repeated diagonal offsets are disallowed.

The result from diags_array is the sparse equivalent of:

np.diag(diagonals[0], offsets[0])

diags_array differs from dia_array in the way it handles off-diagonals. Specifically, dia_array assumes the data input includes padding (ignored values) at the start/end of the rows for positive/negative offset, while diags_array` assumes the input data has no padding. Each value in the input ``diagonals is used.

Added in version 1.11.

Examples

from scipy.sparse import diags_array diagonals = [[1, 2, 3, 4], [1, 2, 3], [1, 2]] diags_array(diagonals, offsets=[0, -1, 2]).toarray() array([[1., 0., 1., 0.], [1., 2., 0., 2.], [0., 2., 3., 0.], [0., 0., 3., 4.]])

Broadcasting of scalars is supported (but shape needs to be specified):

diags_array([1, -2, 1], offsets=[-1, 0, 1], shape=(4, 4)).toarray() array([[-2., 1., 0., 0.], [ 1., -2., 1., 0.], [ 0., 1., -2., 1.], [ 0., 0., 1., -2.]])

If only one diagonal is wanted (as in numpy.diag), the following works as well:

diags_array([1, 2, 3], offsets=1).toarray() array([[ 0., 1., 0., 0.], [ 0., 0., 2., 0.], [ 0., 0., 0., 3.], [ 0., 0., 0., 0.]])