def SparseDataFrame(X, default_fill_value=0.0):
if sparse.issparse(X):
X = pd.DataFrame.sparse.from_spmatrix(X)
X.sparse.fill_value = default_fill_value
elif is_SparseDataFrame(X) or not isinstance(X, pd.DataFrame):
X = pd.DataFrame(X)
return X.astype(pd.SparseDtype(float, fill_value=default_fill_value))
def _typename(X):
if (
isinstance(X, pd.DataFrame)
and not is_SparseDataFrame(X)
and hasattr(X, "sparse")
):
return "DataFrame[SparseArray]"
else:
return type(X).__name__
def assert_matrix_class_equivalent(X, Y):
"""Check the format of X and Y are the same.
We expect:
* shape hasn't changed
* sparsity hasn't changed
* class hasn't changed (except perhaps which kind of spmatrix)
* column and index names haven't changed
"""
assert X.shape == Y.shape
if sparse.issparse(X):
assert sparse.issparse(Y)
assert X.tocoo().nnz == Y.tocoo().nnz
elif is_SparseDataFrame(X):
assert _is_sparse_dataframe(Y)
else:
assert type(X) == type(Y), (type(X), type(Y))
if _is_sparse_dataframe(X):
assert _sparse_dataframe_density(X) == _sparse_dataframe_density(Y)
assert _sparse_dataframe_density(X) == _sparse_dataframe_density(Y)
if isinstance(X, pd.DataFrame):
assert np.all(X.columns == Y.columns)
assert np.all(X.index == Y.index)
return True
def _is_sparse_dataframe(X):
return is_SparseDataFrame(X) or (
isinstance(X, pd.DataFrame) and hasattr(X, "sparse")
)