def test_flat_corr_nan_column():
nan_col_df = ex_df.copy(deep=True)
nan_col_df.iloc[:,0] = np.nan
assert flat_corr(nan_col_df).shape[0] == comb(nan_col_df.shape[1], 2, exact=True)
def test_flat_corr_int_columns():
assert flat_corr(ex_df, columnns=slice(1)).shape[0] == comb(ex_df.shape[1] - 1, 2, exact=True)
def test_flat_corr_arrangement():
res = flat_corr(ex_df)
res_asc = flat_corr(ex_df, ascending=True)
np.testing.assert_allclose(res_asc['pearson'], np.sort(squareform(ex_df.corr(), checks=False)))
assert res.iloc[0,0] > res.iloc[-1,0]
assert res_asc.iloc[0,0] < res_asc.iloc[-1,0]
def test_flat_corr_method():
assert flat_corr(ex_df, method='spearman').notna().all(None)
def test_flat_corr_shape_nans():
assert flat_corr(ex_missing_df).shape[0] == comb(ex_missing_df.shape[1], 2, exact=True)
assert flat_corr(ex_missing_df).shape[0] == squareform(ex_missing_df.corr(), checks=False).shape[0]
assert flat_corr(ex_missing_df, columnns=[*'ABC']).shape[0] == comb(3, 2, exact=True)