BUG: sparse.dok_matrix.fromkeys method totally nonfunctional

Describe your issue.

The fromkeys method of sparse.dok_matrix returns an error. If it is called in the form scipy.sparse.dok_matrix.fromkeys(keys, values=1) it throws TypeError: dok_matrix.fromkeys() takes no keyword arguments. If called as scipy.sparse.dok_matrix.fromkeys(keys) or scipy.sparse.dok_matrix.fromkeys(keys, 1) it throws TypeError: __init__() missing 1 required positional argument: 'arg1'. The content of keys and values do not seem to matter.

The first error contradicts the documentation for the method, which seems to have been inherited from the docs for Python dicts. The two disagree on whether values is a positional or keyword argument.

Reproducing Code Example

import scipy
edges = [(0,1),(1,2),(2,0)]
scipy.sparse.dok_matrix.fromkeys(edges, values=1)
scipy.sparse.dok_matrix.fromkeys(edges, 1)
scipy.sparse.dok_matrix.fromkeys(edges)

Error message

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: dok_matrix.fromkeys() takes no keyword arguments

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: __init__() missing 1 required positional argument: 'arg1'

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: __init__() missing 1 required positional argument: 'arg1'

SciPy/NumPy/Python version information

1.7.2 1.21.4 sys.version_info(major=3, minor=9, micro=8, releaselevel=‘final’, serial=0)

[perimosocordiae]

One way to address this issue might be to override the fromkeys classmethod and have it return a NotImplementedError that points users toward the above workaround.

[perimosocordiae]

Here’s an implementation that should work:

@classmethod
def fromkeys(cls, iterable, value=1):
    d = dict.fromkeys(iterable, value)
    num_rows = 1 + max(k[0] for k in d.keys())
    num_cols = 1 + max(k[1] for k in d.keys())
    result = cls.__init__((num_rows, num_cols))
    result._update(d)
    return result