FIX Target encoder const y (#28233)

s-banach · s-banach · adrinjalali · glemaitre · commit b279e4e14a3b · 2024-02-13T20:24:48.000+01:00
Co-authored-by: s-banach &lt;john@hopfensperger.family&gt;
Co-authored-by: Adrin Jalali &lt;adrin.jalali@gmail.com&gt;
Co-authored-by: Thomas J. Fan &lt;thomasjpfan@gmail.com&gt;
diff --git a/doc/whats_new/v1.4.rst b/doc/whats_new/v1.4.rst
@@ -131,6 +131,15 @@ Changelog
   and `axis=1`, as documented in the docstring.
   :pr:`28222` by :user:`Guillaume Lemaitre <glemaitre>`.
 
+
+:mod:`sklearn.preprocessing`
+............................
+
+- |Fix| :class:`preprocessing.TargetEncoder` no longer fails when
+  `target_type="continuous"` and the input is read-only. In particular, it now
+  works with pandas copy-on-write mode enabled.
+  :pr:`28233` by :user:`John Hopfensperger <s-banach>`.
+
 .. _changes_1_4:
 
 Version 1.4.0
diff --git a/sklearn/preprocessing/_target_encoder_fast.pyx b/sklearn/preprocessing/_target_encoder_fast.pyx
@@ -19,7 +19,7 @@ ctypedef fused Y_DTYPE:
 
 def _fit_encoding_fast(
     INT_DTYPE[:, ::1] X_int,
-    Y_DTYPE[:] y,
+    const Y_DTYPE[:] y,
     cnp.int64_t[::1] n_categories,
     double smooth,
     double y_mean,
@@ -79,7 +79,7 @@ def _fit_encoding_fast(
 
 def _fit_encoding_fast_auto_smooth(
     INT_DTYPE[:, ::1] X_int,
-    Y_DTYPE[:] y,
+    const Y_DTYPE[:] y,
     cnp.int64_t[::1] n_categories,
     double y_mean,
     double y_variance,
diff --git a/sklearn/preprocessing/tests/test_target_encoder.py b/sklearn/preprocessing/tests/test_target_encoder.py
@@ -701,3 +701,16 @@ def test_target_encoding_for_linear_regression(smooth, global_random_seed):
     # cardinality yet non-informative feature instead of the lower
     # cardinality yet informative feature:
     assert abs(coef[0]) < abs(coef[2])
+
+
+def test_pandas_copy_on_write():
+    """
+    Test target-encoder cython code when y is read-only.
+
+    The numpy array underlying df["y"] is read-only when copy-on-write is enabled.
+    Non-regression test for gh-27879.
+    """
+    pd = pytest.importorskip("pandas", minversion="2.0")
+    with pd.option_context("mode.copy_on_write", True):
+        df = pd.DataFrame({"x": ["a", "b", "b"], "y": [4.0, 5.0, 6.0]})
+        TargetEncoder(target_type="continuous").fit(df[["x"]], df["y"])
