scipy.sparse.

diags_array#

scipy.sparse.diags_array(diagonals, /, *, offsets=0, shape=None, format=None, dtype=<object object>)[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):

  • k = 0 the main diagonal (default)

  • k > 0 the kth upper diagonal

  • k < 0 the kth lower diagonal

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. If dtype is None, the output data type is determined by the data type of the input diagonals.

Up until SciPy 1.19, the default behavior will be to return an array with an inexact (floating point) data type. In particular, integer input will be converted to double precision floating point. This behavior is deprecated, and in SciPy 1.19, the default behavior will be changed to return an array with the same data type as the input diagonals. To adopt this behavior before version 1.19, use dtype=None.

Returns:
new_arraydia_array

dia_array holding the values in diagonals offset from the main diagonal as indicated in offsets.

See also

dia_array

constructor for the sparse DIAgonal format.

Notes

Repeated diagonal offsets are disallowed.

The result from diags_array is the sparse equivalent of:

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

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 = [[1.0, 2.0, 3.0, 4.0], [1.0, 2.0, 3.0], [1.0, 2.0]]
>>> 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([1.0, -2.0, 1.0], 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([1.0, 2.0, 3.0], offsets=1).toarray()
array([[ 0.,  1.,  0.,  0.],
       [ 0.,  0.,  2.,  0.],
       [ 0.,  0.,  0.,  3.],
       [ 0.,  0.,  0.,  0.]])
On this page