Add tests for index_matrix and index_vector

metalearners/_utils.py

@@ -26,6 +26,7 @@
 
 
 def safe_len(X: Matrix) -> int:
+    """Determine the length of a Matrix."""
     if scipy.sparse.issparse(X):
         return X.shape[0]
     return len(X)

tests/test__utils.py

@@ -8,6 +8,7 @@
 import pytest
 from glum import GeneralizedLinearRegressor, GeneralizedLinearRegressorCV
 from lightgbm import LGBMClassifier, LGBMRegressor
+from scipy.sparse import csr_matrix
 from sklearn.ensemble import HistGradientBoostingClassifier
 from sklearn.linear_model import LinearRegression
 from xgboost import XGBClassifier, XGBRegressor
@@ -20,6 +21,8 @@
     convert_treatment,
     function_has_argument,
     get_linear_dimension,
+    index_matrix,
+    index_vector,
     supports_categoricals,
     validate_all_vectors_same_index,
     validate_model_and_predict_method,
@@ -345,3 +348,59 @@ def test_validate_valid_treatment_variant_not_control(
     else:
         with pytest.raises(ValueError, match="variant"):
             validate_valid_treatment_variant_not_control(treatment_variant, n_variants)
+
+
+@pytest.mark.parametrize("matrix_backend", [np.ndarray, pd.DataFrame, csr_matrix])
+@pytest.mark.parametrize("rows_backend", [np.array, pd.Series])
+def test_index_matrix(matrix_backend, rows_backend):
+    n_samples = 10
+    if matrix_backend == np.ndarray:
+        matrix = np.array(list(range(n_samples))).reshape((-1, 1))
+    elif matrix_backend == pd.DataFrame:
+        # We make sure that the index is not equal to the row number.
+        matrix = pd.DataFrame(
+            list(range(n_samples)), index=list(range(20, 20 + n_samples))
+        )
+    elif matrix_backend == csr_matrix:
+        matrix = csr_matrix(np.array(list(range(n_samples))).reshape((-1, 1)))
+    else:
+        raise ValueError()
+    rows = rows_backend([1, 4, 5])
+    result = index_matrix(matrix=matrix, rows=rows)
+
+    assert isinstance(result, matrix_backend)
+    assert result.shape[1] == matrix.shape[1]
+
+    if isinstance(result, pd.DataFrame):
+        processed_result = result.values[:, 0]
+    else:
+        processed_result = result[:, 0]
+
+    expected = np.array([1, 4, 5])
+    assert (processed_result == expected).sum() == len(expected)
+
+
+@pytest.mark.parametrize("vector_backend", [np.ndarray, pd.Series])
+@pytest.mark.parametrize("rows_backend", [np.array, pd.Series])
+def test_index_vector(vector_backend, rows_backend):
+    n_samples = 10
+    if vector_backend == np.ndarray:
+        vector = np.array(list(range(n_samples)))
+    elif vector_backend == pd.Series:
+        # We make sure that the index is not equal to the row number.
+        vector = pd.Series(
+            list(range(n_samples)), index=list(range(20, 20 + n_samples))
+        )
+    else:
+        raise ValueError()
+
+    rows = rows_backend([1, 4, 5])
+
+    result = index_vector(vector=vector, rows=rows)
+    assert isinstance(result, vector_backend)
+
+    if isinstance(result, pd.Series):
+        result = result.values
+
+    expected = np.array([1, 4, 5])
+    assert (result == expected).all()