FIX Handles all numerical DataFrames with check_inverse=True in FunctionTransformer (scikit-learn#25274)

thomasjpfan · jeremiedbb · jjerphan · commit ae5a22f96214 · 2023-01-23T16:42:32.000+01:00
Co-authored-by: Jérémie du Boisberranger &lt;34657725+jeremiedbb@users.noreply.github.com&gt;
diff --git a/doc/whats_new/v1.2.rst b/doc/whats_new/v1.2.rst
@@ -27,6 +27,13 @@ Changelog
   `verbose` parameter set to a value greater than 0.
   :pr:`25250` by :user:`Jérémie Du Boisberranger <jeremiedbb>`.
 
+:mod:`sklearn.preprocessing`
+............................
+
+- |Fix| :meth:`preprocessing.FunctionTransformer.inverse_transform` correctly
+  supports DataFrames that are all numerical when `check_inverse=True`.
+  :pr:`25274` by `Thomas Fan`_.
+
 :mod:`sklearn.utils`
 ....................
 
diff --git a/sklearn/preprocessing/_function_transformer.py b/sklearn/preprocessing/_function_transformer.py
@@ -174,7 +174,13 @@ def _check_inverse_transform(self, X):
         idx_selected = slice(None, None, max(1, X.shape[0] // 100))
         X_round_trip = self.inverse_transform(self.transform(X[idx_selected]))
 
-        if not np.issubdtype(X.dtype, np.number):
+        if hasattr(X, "dtype"):
+            dtypes = [X.dtype]
+        elif hasattr(X, "dtypes"):
+            # Dataframes can have multiple dtypes
+            dtypes = X.dtypes
+
+        if not all(np.issubdtype(d, np.number) for d in dtypes):
             raise ValueError(
                 "'check_inverse' is only supported when all the elements in `X` is"
                 " numerical."
diff --git a/sklearn/preprocessing/tests/test_function_transformer.py b/sklearn/preprocessing/tests/test_function_transformer.py
@@ -217,6 +217,36 @@ def inverse_func(X):
         transformer.fit(data)
 
 
+def test_function_transformer_support_all_nummerical_dataframes_check_inverse_True():
+    """Check support for dataframes with only numerical values."""
+    pd = pytest.importorskip("pandas")
+
+    df = pd.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]})
+    transformer = FunctionTransformer(
+        func=lambda x: x + 2, inverse_func=lambda x: x - 2, check_inverse=True
+    )
+
+    # Does not raise an error
+    df_out = transformer.fit_transform(df)
+    assert_allclose_dense_sparse(df_out, df + 2)
+
+
+def test_function_transformer_with_dataframe_and_check_inverse_True():
+    """Check error is raised when check_inverse=True.
+
+    Non-regresion test for gh-25261.
+    """
+    pd = pytest.importorskip("pandas")
+    transformer = FunctionTransformer(
+        func=lambda x: x, inverse_func=lambda x: x, check_inverse=True
+    )
+
+    df_mixed = pd.DataFrame({"a": [1, 2, 3], "b": ["a", "b", "c"]})
+    msg = "'check_inverse' is only supported when all the elements in `X` is numerical."
+    with pytest.raises(ValueError, match=msg):
+        transformer.fit(df_mixed)
+
+
 @pytest.mark.parametrize(
     "X, feature_names_out, input_features, expected",
     [
