def __init__(
self,
initial_value=None,
shape=None,
name=None,
base=None,
readonly=False,
offset=0,
):
if self.assert_named_signals:
assert name is not None
self._name = name
if initial_value is None:
assert shape is not None
initial_value = np.zeros(shape, dtype=rc.float_dtype)
elif shape is not None:
assert initial_value.shape == shape
self._initial_value = initial_value
if np.any(
np.isnan(self._initial_value.data if self.sparse else self._initial_value)
):
raise SignalError(f"{self!r} contains NaNs.")
if self.sparse:
assert initial_value.ndim == 2
assert offset == 0
assert base is None
if npext.is_spmatrix(initial_value):
self._initial_value.data.setflags(write=False)
else:
# To ensure we do not modify data passed into the signal,
# we make a view of the data and mark it as not writeable.
# Consumers (like SignalDict) are responsible for making copies
# that can be modified, or using the readonly view appropriately.
readonly_view = np.asarray(self._initial_value)
if readonly_view.ndim > 0 and base is None:
readonly_view = np.ascontiguousarray(readonly_view)
# Ensure we have a view and aren't modifying the original's flags
readonly_view = readonly_view.view()
readonly_view.setflags(write=False)
self._initial_value = readonly_view
if base is not None:
assert isinstance(base, Signal) and not base.is_view
# make sure initial_value uses the same data as base.initial_value
assert initial_value.base is base.initial_value.base
self._base = base
self._offset = offset
self._readonly = bool(readonly)
def is_sparse(obj):
"""Check if ``obj`` is a sparse matrix."""
return isinstance(obj, SparseMatrix) or npext.is_spmatrix(obj)
def is_sparse(obj):
return isinstance(obj, SparseMatrix) or npext.is_spmatrix(obj)