DOK

class sparse.DOK(shape, data=None, dtype=None)[source]

A class for building sparse multidimensional arrays.

Parameters:
  • shape (tuple[int] (DOK.ndim,)) – The shape of the array.
  • data (dict, optional) – The key-value pairs for the data in this array.
  • dtype (np.dtype, optional) – The data type of this array. If left empty, it is inferred from the first element.
dtype

numpy.dtype – The datatype of this array. Can be None if no elements have been set yet.

shape

tuple[int] – The shape of this array.

data

dict – The keys of this dictionary contain all the indices and the values contain the nonzero entries.

See also

COO
A read-only sparse array.

Examples

You can create DOK objects from Numpy arrays.

>>> x = np.eye(5, dtype=np.uint8)
>>> x[2, 3] = 5
>>> s = DOK.from_numpy(x)
>>> s
<DOK: shape=(5, 5), dtype=uint8, nnz=6>

You can also create them from just shapes, and use slicing assignment.

>>> s2 = DOK((5, 5), dtype=np.int64)
>>> s2[1:3, 1:3] = [[4, 5], [6, 7]]
>>> s2
<DOK: shape=(5, 5), dtype=int64, nnz=4>

You can convert DOK arrays to COO arrays, or numpy.ndarray objects.

>>> from sparse import COO
>>> s3 = COO(s2)
>>> s3
<COO: shape=(5, 5), dtype=int64, nnz=4, sorted=False, duplicates=False>
>>> s2.todense()  
array([[0, 0, 0, 0, 0],
       [0, 4, 5, 0, 0],
       [0, 6, 7, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0]])
>>> s4 = COO.from_numpy(np.eye(4, dtype=np.uint8))
>>> s4
<COO: shape=(4, 4), dtype=uint8, nnz=4, sorted=True, duplicates=False>
>>> s5 = DOK.from_coo(s4)
>>> s5
<DOK: shape=(4, 4), dtype=uint8, nnz=4>

You can also create DOK arrays from a shape and a dict of values. Zeros are automatically ignored.

>>> values = {
...     (1, 2, 3): 4,
...     (3, 2, 1): 0,
... }
>>> s6 = DOK((5, 5, 5), values)
>>> s6
<DOK: shape=(5, 5, 5), dtype=int64, nnz=1>

Attributes

DOK.density The ratio of nonzero to all elements in this array.
DOK.ndim The number of dimensions of this array.
DOK.nnz The number of nonzero elements in this array.
DOK.size The number of all elements (including zeros) in this array.

Methods

DOK.from_coo(x) Get a DOK array from a COO array.
DOK.from_numpy(x) Get a DOK array from a Numpy array.
DOK.to_coo() Convert this DOK array to a COO array.
DOK.todense() Convert this DOK array into a Numpy array.